Methods in

Molecular Biology 1615

Laure Journet
Eric Cascales Editors

Bacterial Protein
Secretion
Systems
Methods and Protocols
Methods in Molecular Biology

Series editor
John M. Walker
School of Life and Medical Sciences
University of Hertfordshire
Hatfield, Hertfordshire, AL10 9AB, UK

For further volumes:

http://www.springer.com/series/7651
 Bacterial Protein Secretion
Systems

Methods and Protocols

Editors

Laure Journet and Eric Cascales

Institut de Microbiologie de la Méditerranée, Aix-Marseille Univ - CNRS, Marseille, France
Editors
Laure Journet Eric Cascales
Institut de Microbiologie de la Méditerranée Institut de Microbiologie de la Méditerranée
Aix-Marseille Univ - CNRS Aix-Marseille Univ - CNRS
Marseille, France Marseille, France

ISSN 1064-3745     ISSN 1940-6029 (electronic)

Methods in Molecular Biology
ISBN 978-1-4939-7031-5    ISBN 978-1-4939-7033-9 (eBook)
DOI 10.1007/978-1-4939-7033-9

Library of Congress Control Number: 2017944103

© Springer Science+Business Media LLC 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction
on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation,
computer software, or by similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not
imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and
regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to
be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty,
express or implied, with respect to the material contained herein or for any errors or omissions that may have been made.
The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Printed on acid-free paper

This Humana Press imprint is published by Springer Nature

The registered company is Springer Science+Business Media LLC
The registered company address is: 233 Spring Street, New York, NY 10013, U.S.A.
Preface

In their ecological niches, bacteria are in contact with other prokaryotic and eukaryotic
cells. Bacteria therefore evolved mechanisms to communicate and collaborate with these
cells. They also developed belligerent behaviours to eliminate competitors and to infect
eukaryotic host cells. These aggressive actions are mediated by effector toxins with specific
activities which will ultimately cause target cell lysis or re-routing of metabolic or trafficking
pathways in the host.
The proper delivery of bacterial effectors into the milieu or directly into target cells is
assured by dedicated structures called secretion systems. Up to now, nine secretion systems
have been described in bacteria, and additional systems allow the exposition of toxins or
adhesins at the cell surface or at the extremity of a pilus structure. These multi-protein
trans-envelope complexes differ in composition, mechanism of assembly, and mode of
recruitment and transport of toxins. However, studying these macromolecular complexes
requires common techniques, ranging from the bioinformatic identification of machine
components and effectors to methods of defining interactions amongst the different sub-
units and the development of reporters to follow effector translocation in vivo. Finally,
state-of-the-art techniques have made significant progress in the analysis of these large
complexes directly in the bacterial cell or from purified samples.
The purpose of this book on bacterial secretion systems is to provide protocols that cover
the broad arsenal of techniques used to study a secretion system from A to Z: identifying and
localising the different subunits, defining interactions within subunits, monitoring conforma-
tional changes, purifying and imaging large complexes, defining the assembly pathway by
fluorescence microscopy and the role of energy during assembly or secretion, and identifying
secreted effectors as well as using reporters to follow effector transport. Most of these tech-
niques are not restricted to the study of secretion systems but are also of specific interest for
any researcher interested in multi-protein complexes of the bacterial cell envelope.
The book starts with a chapter describing a recently developed software program aimed
at identifying gene clusters encoding secretion systems within bacterial genomes. Then six
chapters describe methods for defining the subcellular localisation of the different subunits
of a multi-protein system: prediction programs, fractionation, cell surface exposition and
isopycnic density gradients to partition inner and outer membranes. Three techniques to
determine the topology of membrane proteins (substituted cysteine accessibility, protease
accessibility and reporter fusions) are then detailed. This is followed by eleven chapters
covering genetic, cytologic and biochemical methods used to study protein–protein and
protein–peptidoglycan interactions. Two chapters are dedicated to methods for unveiling
the role of energy and conformational changes, and then six chapters describe techniques
for purifying and imaging large complexes. Finally, methods to identify effectors as well as
the reporter fusions used to validate effector secretion and translocation are presented.

Marseille, France Laure Journet

Marseille, France Eric Cascales

v
Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

  1 Identification of Protein Secretion Systems in Bacterial

Genomes Using MacSyFinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Sophie S. Abby and Eduardo P.C. Rocha
  2 Protein Sorting Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Henrik Nielsen
  3 Cell Fractionation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Melissa Petiti, Laetitia Houot, and Denis Duché
  4 Defining Lipoprotein Localisation by Fluorescence Microscopy . . . . . . . . . . . . . . . 65
Maria Guillermina Casabona, Mylène Robert-Genthon,
Didier Grunwald, and Ina Attrée
  5 Identification of Lipoproteins Using Globomycin and 
Radioactive Palmitate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Nienke Buddelmeijer
  6 Defining Membrane Protein Localization by Isopycnic Density Gradients . . . . . . . 81
Rhys A. Dunstan, Iain D. Hay, and Trevor Lithgow
  7 Cell Surface Exposure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Anna Konovalova
  8 Probing Inner Membrane Protein Topology by Proteolysis . . . . . . . . . . . . . . . . . . 97
Maxence S. Vincent and Eric Cascales
  9 Mapping of Membrane Protein Topology by Substituted
Cysteine Accessibility Method (SCAM™) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Mikhail Bogdanov
10 Defining Membrane Protein Topology Using pho-lac Reporter Fusions . . . . . . . . . 129
Gouzel Karimova and Daniel Ladant
11 In Vivo and In Vitro Protein–Peptidoglycan Interactions . . . . . . . . . . . . . . . . . . . . 143
Gang Li and S. Peter Howard
12 Measure of Peptidoglycan Hydrolase Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Yoann G. Santin and Eric Cascales
13 Protein–Protein Interaction: Bacterial Two-Hybrid . . . . . . . . . . . . . . . . . . . . . . . . 159
Gouzel Karimova, Emilie Gauliard, Marilyne Davi,
Scot P. Ouellette, and Daniel Ladant
14 Protein–Protein Interactions: Yeast Two-Hybrid System . . . . . . . . . . . . . . . . . . . . 177
Jer-Sheng Lin and Erh-Min Lai
15 Protein–Protein Interactions: Cytology Two-Hybrid . . . . . . . . . . . . . . . . . . . . . . . 189
Krishnamohan Atmakuri
16 Fusion Reporter Approaches to Monitoring Transmembrane Helix
Interactions in Bacterial Membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Laureen Logger, Abdelrahim Zoued, and Eric Cascales

vii
viii Contents

17 Protein–Protein Interactions: Co-Immunoprecipitation . . . . . . . . . . . . . . . . . . . . . 211

Jer-Sheng Lin and Erh-Min Lai
18 Protein–Protein Interaction: Tandem Affinity Purification in Bacteria . . . . . . . . . . . 221
Julie P.M. Viala and Emmanuelle Bouveret
19 Site-Directed and Time-Resolved Photocrosslinking in Cells
Metabolically Labeled with Radioisotopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
Raffaele Ieva
20 Protein–Protein Interactions: Pull-Down Assays . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Arthur Louche, Suzana P. Salcedo, and Sarah Bigot
21 Protein–Protein Interactions: Surface Plasmon Resonance . . . . . . . . . . . . . . . . . . . 257
Badreddine Douzi
22 Assessing Energy-Dependent Protein Conformational
Changes in the TonB System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
Ray A. Larsen
23 Defining Assembly Pathways by Fluorescence Microscopy . . . . . . . . . . . . . . . . . . . 289
Abdelrahim Zoued and Andreas Diepold
24 Large Complexes: Cloning Strategy, Production, and Purification . . . . . . . . . . . . . 299
Eric Durand and Roland Lloubes
25 Shearing and Enrichment of Extracellular Type IV Pili . . . . . . . . . . . . . . . . . . . . . . 311
Alba Katiria Gonzalez Rivera and Katrina T. Forest
26 Blue Native PAGE Analysis of Bacterial Secretion Complexes . . . . . . . . . . . . . . . . . 321
Susann Zilkenat, Tobias Dietsche, Julia V. Monjarás Feria,
Claudia E. Torres-Vargas, Mehari Tesfazgi Mebrhatu, and Samuel Wagner
27 In Situ Imaging of Bacterial Secretion Systems by Electron Cryotomography . . . . . 353
Gregor L. Weiss, João M. Medeiros, and Martin Pilhofer
28 Structural Analysis of Protein Complexes by Cryo Electron Microscopy . . . . . . . . . 377
Tiago R.D. Costa, Athanasios Ignatiou, and Elena V. Orlova
29 Bacterial Filamentous Appendages Investigated by Solid-­State
NMR Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
Birgit Habenstein and Antoine Loquet
30 Energy Requirements for Protein Secretion via the 
Flagellar Type III Secretion System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
Marc Erhardt
31 Identification of Effectors: Precipitation of Supernatant Material . . . . . . . . . . . . . . 459
Nicolas Flaugnatti and Laure Journet
32 Screening for Secretion of the Type VI Secretion System Protein Hcp
by Enzyme-Linked Immunosorbent Assay and Colony Blot . . . . . . . . . . . . . . . . . . 465
Brent S. Weber, Pek Man Ly, and Mario F. Feldman
33 Effector Translocation: Cya Reporter Assay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473
Suma Chakravarthy, Bethany Huot, and Brian H. Kvitko
34 Monitoring Effector Translocation using the TEM-1 Beta-­Lactamase
Reporter System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 489
Julie Allombert, Anne Vianney, and Xavier Charpentier
 Contents ix

35 Effector Translocation Assay: Differential Solubilization . . . . . . . . . . . . . . . . . . . . . 501

Irina S. Franco, Sara V. Pais, Nuno Charro, and Luís Jaime Mota
36 Quantitative Determination of Anti-bacterial Activity During
Bacterial Co-culture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517
Juliana Alcoforado Diniz, Birte Hollmann, and Sarah J. Coulthurst

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
Contributors

Sophie S. Abby  •  Microbial Evolutionary Genomics, Institut Pasteur, Paris, France;

CNRS, UMR3525, Paris, France; Université Grenoble Alpes, Laboratoire Techniques
de l’Ingénierie Médicale et de la Complexité - Informatique, Mathématiques et
Applications, Grenoble (TIMC-IMAG), F-38000 Grenoble, France; Centre National
de la Recherche Scientifique (CNRS), TIMC-IMAG, Grenoble, France
Juliana Alcoforado Diniz  •  Division of Molecular Microbiology, School of Life Sciences,
University of Dundee, Dundee, UK
Julie Allombert  •  CIRI, Centre International de Recherche en Infectiologie, Inserm,
U1111, Université Claude Bernard Lyon 1, CNRS, UMR5308, École Normale
Supérieure de Lyon, Univ Lyon, Villeurbanne, France
Krishnamohan Atmakuri  •  Translational Health Science and Technology Institute, NCR
Biotech Science Cluster, Faridabad, India
Ina Attrée  •  Bacterial Pathogenesis and Cellular Responses, Centre National pour la
Recherche Scientifique (CNRS), University Grenoble Alpes, INSERM, Biosciences and
Biotechnology Institut, CEA Grenoble, Grenoble, France
Sarah Bigot  •  Molecular Microbiology and Structural Biochemistry, CNRS UMR 5086,
Université Lyon 1, Institut de Biologie et Chimie des Protéines, Lyon, France
Mikhail Bogdanov  •  Department of Biochemistry and Molecular Biology, University of Texas
Health Science Center at Houston, McGovern Medical School, Houston, TX, USA
Emmanuelle Bouveret  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires,
UMR7255, Institut de Microbiologie de la Méditerranée, Aix-Marseille
University—CNRS, Marseille, France
Nienke Buddelmeijer  •  Institut Pasteur, Biology and Genetics of the Bacterial Cell Wall
Unit, Inserm Group Avenir, Paris, France
Maria Guillermina Casabona  •  Bacterial Pathogenesis and Cellular Responses, Centre
National pour la Recherche Scientifique (CNRS), University Grenoble Alpes,
INSERM, Biosciences and Biotechnology Institut, CEA Grenoble, Grenoble, France;
Division of Molecular Microbiology, School of Life Sciences, University of Dundee,
Dundee, Scotland, UK
Eric Cascales  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires, (LISM,
UMR7255), Institut de Microbiologie de la Méditerranée (IMM), Aix-Marseille
Université—Centre National de la Recherche Scientifique (CNRS), Marseille, France
Suma Chakravarthy  •  Plant Pathology and Plant-Microbe Biology Section, School of
Integrative Plant Science, Cornell University, Ithaca, NY, USA
Xavier Charpentier  •  CIRI, Centre International de Recherche en Infectiologie, Inserm,
U1111, Université Claude Bernard Lyon 1, CNRS, UMR5308, École Normale
Supérieure de Lyon, Univ Lyon, Villeurbanne, France
Nuno Charro  •  UCIBIO—REQUIMTE, Faculdade de Ciências e Tecnologia,
Universidade NOVA de Lisboa (FCT NOVA), Caparica, Portugal
Tiago R.D. Costa  •  Institute for Structural and Molecular Biology, School of Biological
Sciences, Birkbeck College, London, UK

xi
xii Contributors

Sarah J. Coulthurst  •  Division of Molecular Microbiology, School of Life Sciences,

University of Dundee, Dundee, UK
Marilyne Davi  •  Unité de Biochimie des Interactions Macromoléculaires, Département de
Biologie Structurale et Chimie, Institut Pasteur, CNRS, UMR 3528, Paris, France
Andreas Diepold  •  Department of Biochemistry, University of Oxford, Oxford, UK;
Department of Ecophysiology, Max Planck Institute for Terrestrial Microbiology, Karl-
von-Frisch-Str., Marburg, Germany
Tobias Dietsche  •  Section of Cellular and Molecular Microbiology, Interfaculty Institute
of Microbiology and Infection Medicine (IMIT), University of Tübingen, Tübingen,
Germany
Badreddine Douzi  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires (LISM,
UMR 7255), Institut de Microbiologie de la Méditerranée (IMM), Aix-Marseille
Université—Centre National de la Recherche Scientifique (CNRS), Marseille, France
Denis Duché  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires (LISM, UMR
7255), Institut de Microbiologie de la Méditerranée (IMM), Aix-Marseille Université—
Centre National de la Recherche Scientifique (CNRS), Marseille, France
Rhys A. Dunstan  •  Department of Microbiology and Infection and Immunity Program,
Biomedicine Discovery Institute, Monash University, Melbourne, VIC, Australia
Eric Durand  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires (LISM,
UMR7255), Institut de Microbiologie de la Méditerranée (IMM), Aix-Marseille Univ
and CNRS, Marseille, France
Marc Erhardt  •  Helmholtz Centre for Infection Research, Braunschweig, Germany
Mario F. Feldman  •  Department of Molecular Microbiology, Washington University School
of Medicine, Saint Louis, MO, USA
Nicolas Flaugnatti  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires, (LISM,
UMR7255), Institut de Microbiologie de la Méditerranée, Aix-Marseille
Univ—CNRS, Marseille, France
Katrina T. Forest  •  Department of Bacteriology and Biophysics Program, University of
Wisconsin-Madison, Madison, WI, USA
Irina S. Franco  •  UCIBIO—REQUIMTE, Faculdade de Ciências e Tecnologia,
Universidade NOVA de Lisboa (FCT NOVA), Caparica, Portugal
Emilie Gauliard  •  Unité de Biochimie des Interactions Macromoléculaires, Département
de Biologie Structurale et Chimie, Institut Pasteur, CNRS, UMR 3528, Paris, France;
Université Paris Diderot, Sorbonne Paris Cité, Cellule Pasteur, Paris, France
Alba Katiria Gonzalez Rivera  •  Department of Bacteriology and Biophysics Program,
University of Wisconsin-Madison, Madison, WI, USA
Didier Grunwald  •  Bacterial Pathogenesis and Cellular Responses, Centre National pour
la Recherche Scientifique (CNRS), University Grenoble Alpes, INSERM, Biosciences and
Biotechnology Institut, CEA Grenoble, Grenoble, France
Birgit Habenstein  •  Institute of Chemistry & Biology of Membranes & Nanoobjects
(UMR5248 CBMN), CNRS, University of Bordeaux, Institut Européen de Chimie et
Biologie, Pessac, France
Iain D. Hay  •  Department of Microbiology and Infection and Immunity Program,
Biomedicine Discovery Institute, Monash University, Melbourne, VIC, Australia
Birte Hollmann  •  Division of Molecular Microbiology, School of Life Sciences, University of
Dundee, Dundee, UK
 Contributors xiii

Laetitia Houot  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires (LISM, UMR

7255), Institut de Microbiologie de la Méditerranée (IMM), Aix-Marseille Université—
Centre National de la Recherche Scientifique (CNRS), Marseille, France
S. Peter Howard  •  Department of Microbiology and Immunology, College of Medicine,
University of Saskatchewan, Saskatoon, SK, Canada
Bethany Huot  •  Department of Energy, Plant Research Laboratory, Michigan State
University, East Lansing, MI, USA
Raffaele Ieva  •  Laboratoire de Microbiologie et de Génétique Moléculaires, Centre de
Biologie Intégrative (CBI), Université de Toulouse, CNRS, UPS, Toulouse, France
Athanasios Ignatiou  •  Institute for Structural and Molecular Biology, School of Biological
Sciences, Birkbeck College, London, UK
Laure Journet  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires, (LISM,
UMR7255), Institut de Microbiologie de la Méditerranée, Aix-Marseille Univ—CNRS,
Marseille, France
Gouzel Karimova  •  Unité de Biochimie des Interactions Macromoléculaires, Département
de Biologie Structurale et Chimie, Institut Pasteur, CNRS, UMR 3528, Paris, France
Anna Konovalova  •  Lewis Thomas Laboratory, Department of Molecular Biology,
Princeton University, Princeton, NJ, USA
Brian H. Kvitko  •  Department of Plant Pathology, University of Georgia, Athens,
GA, USA
Daniel Ladant  •  Unité de Biochimie des Interactions Macromoléculaires, Département de
Biologie Structurale et Chimie, Institut Pasteur, CNRS, UMR 3528, Paris, France
Erh-Min Lai  •  Institute of Plant and Microbial Biology, Academia Sinica, Taipei, Taiwan
Ray A. Larsen  •  Department of Biological Sciences, Bowling Green State University,
Bowling Green, OH, USA
Gang Li  •  Department of Microbiology and Immunology, College of Medicine, University of
Saskatchewan, Saskatoon, SK, Canada
Jer-Sheng Lin  •  Institute of Plant and Microbial Biology, Academia Sinica, Taipei,
Taiwan
Trevor Lithgow  •  Department of Microbiology, and Infection and Immunity Program,
Biomedicine Discovery Institute, Monash University, Melbourne, VIC, Australia
Roland Lloubes  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires, (LISM,
UMR7255), Institut de Microbiologie de la Méditerranée, Aix-Marseille Univ—CNRS,
Marseille, France
Laureen Logger  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires, UMR7255,
Institut de Microbiologie de la Méditerranée (IMM), Aix-Marseille Univ and CNRS,
Marseille, France
Antoine Loquet  •  Institute of Chemistry & Biology of Membranes & Nanoobjects
(UMR5248 CBMN), CNRS, University of Bordeaux, Institut Européen de Chimie et
Biologie, Pessac, France
Arthur Louche  •  Molecular Microbiology and Structural Biochemistry, CNRS UMR
5086, Université Lyon 1, Institut de Biologie et Chimie des Protéines, Lyon, France
Pek Man Ly  •  Department of Molecular Microbiology, Washington University School of
Medicine, Saint Louis, MO, USA
João M. Medeiros  •  Department of Biology, ETH Zürich, Institute of Molecular Biology and
Biophysics, Zürich, Switzerland
xiv Contributors

Julia V. Monjarás Feria  •  Section of Cellular and Molecular Microbiology, Interfaculty

Institute of Microbiology and Infection Medicine (IMIT), University of Tübingen,
Tübingen, Germany
Luís Jaime Mota  •  UCIBIO—REQUIMTE, Faculdade de Ciências e Tecnologia,
Universidade Nova de Lisboa (FCT NOVA), Caparica, Portugal
Henrik Nielsen  •  Technical University of Denmark, Lyngby, Denmark
Elena V. Orlova  •  Institute for Structural and Molecular Biology, School of Biological
Sciences, Birkbeck College, London, UK
Scot P. Ouellette  •  Division of Basic Biomedical Sciences, Sanford School of Medicine,
University of South Dakota, Vermillion, SD, USA
Sara V. Pais  •  UCIBIO—REQUIMTE, Faculdade de Ciências e Tecnologia, Universidade
Nova de Lisboa (FCT NOVA), Caparica, Portugal
Melissa Petiti  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires, (LISM,
UMR7255), Institut de Microbiologie de la Méditerranée, Aix-Marseille Univ—CNRS,
Marseille, France
Martin Pilhofer  •  Department of Biology, ETH Zürich, Institute of Molecular Biology
and Biophysics, Zürich, Switzerland
Mylène Robert-Genthon  •  Bacterial Pathogenesis and Cellular Responses, Centre
National pour la Recherche Scientifique (CNRS), University Grenoble Alpes, INSERM,
Biosciences and Biotechnology Institut, CEA Grenoble, Grenoble, France
Eduardo P.C. Rocha  •  Microbial Evolutionary Genomics, Institut Pasteur, Paris, France;
CNRS, UMR3525, Paris, France
Suzana P. Salcedo  •  Molecular Microbiology and Structural Biochemistry, CNRS UMR
5086, Université Lyon 1, Institut de Biologie et Chimie des Protéines, Lyon, France
Yoann G. Santin  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires, (LISM,
UMR7255), Institut de Microbiologie de la Méditerranée, Aix-Marseille University—
CNRS, Marseille, France
Mehari Tesfazgi Mebrhatu  •  Section of Cellular and Molecular Microbiology,
Interfaculty Institute of Microbiology and Infection Medicine (IMIT), University of
Tübingen, Tübingen, Germany
Claudia E. Torres-Vargas  •  Section of Cellular and Molecular Microbiology, Interfaculty
Institute of Microbiology and Infection Medicine (IMIT), University of Tübingen,
Tübingen, Germany
Julie P.M. Viala  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires, UMR7255,
Institut de Microbiologie de la Méditerranée (IMM), Aix-Marseille Université—Centre
National de la Recherche Scientifique (CNRS), Marseille, France
Anne Vianney  •  CIRI, Centre International de Recherche en Infectiologie, Inserm,
U1111, Université Claude Bernard Lyon 1, CNRS, UMR5308, École Normale
Supérieure de Lyon, Univ Lyon, Villeurbanne, France
Maxence S. Vincent  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires, (LISM,
UMR7255), Institut de Microbiologie de la Méditerranée (IMM), Aix-Marseille
Université—Centre National de la Recherche Scientifique (CNRS), Marseille, France
Samuel Wagner  •  Section of Cellular and Molecular Microbiology, Interfaculty Institute of
Microbiology and Infection Medicine (IMIT), University of Tübingen, Tübingen,
Germany; German Center for Infection Research (DZIF), Tübingen, Germany
Brent S. Weber  •  Department of Molecular Microbiology, Washington University School of
Medicine, Saint Louis, MO, USA
 Contributors xv

Gregor L. Weiss  •  Department of Biology, ETH Zürich, Institute of Molecular Biology and

Biophysics, Zürich, Switzerland
Susann Zilkenat  •  Section of Cellular and Molecular Microbiology, Interfaculty Institute
of Microbiology and Infection Medicine (IMIT), University of Tübingen, Tübingen,
Germany
Abdelrahim Zoued  •  Laboratoire d’Ingénierie des Systèmes Macromoléculaires, UMR7255,
Institut de Microbiologie de la Méditerranée, Aix-Marseille Univ—CNRS, Marseille,
France; Division of Infectious Diseases and Harvard Medical School, Department of
Microbiology and Immunobiology, Howard Hughes Medical Institute, Brigham and
Women’s Hospital, Boston, MA, USA
 Chapter 1

Identification of Protein Secretion Systems in Bacterial

Genomes Using MacSyFinder
Sophie S. Abby and Eduardo P.C. Rocha

Abstract
Protein secretion systems are complex molecular machineries that translocate proteins through the outer
membrane, and sometimes through multiple other barriers. They have evolved by co-option of compo-
nents from other envelope-associated cellular machineries, making them sometimes difficult to identify and
discriminate. Here, we describe how to identify protein secretion systems in bacterial genomes using
MacSyFinder. This flexible computational tool uses the knowledge stemming from experimental studies to
identify homologous systems in genome data. It can be used with a set of predefined models—“TXSScan”—
to identify all major secretion systems of diderm bacteria (i.e., with inner and with LPS-containing outer
membranes). For this, it identifies and clusters colocalized components of secretion systems using sequence
similarity searches with hidden Markov model protein profiles. Finally, it checks whether the genetic con-
tent and organization of clusters satisfy the constraints of the model. TXSScan models can be customized
to search for variants of known systems. The models can also be built from scratch to identify novel sys-
tems. In this chapter, we describe a complete pipeline of analysis, including the identification of a reference
set of experimentally studied systems, the identification of components and the construction of their pro-
tein profiles, the definition of the models, their optimization, and, finally, their use as tools to search
genomic data.

Key words Comparative genomics, Genome annotation, Bioinformatics detection, Macromolecular

systems, Bioinformatic modeling

1  Introduction

Bacteria produce proteins to interact with other individuals, pro-

karyotes or eukaryotes, to effect changes in their local environ-
ment, or to take up resources. Many of the proteins involved in
these processes need to be secreted to the outside of the cell.
Bacteria with an LPS-containing outer membrane (henceforth
called diderms) face a formidable challenge in secreting these pro-
teins because they must transport them through the inner mem-
brane, the cell wall, the outer membrane, and eventually other
barriers such as the bacterial capsule and membranes of other cells.
The complexity of these molecular processes and the key roles of

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_1, © Springer Science+Business Media LLC 2017

1
2 Sophie S. Abby and Eduardo P.C. Rocha

protein secretion systems in bacterial ecology and virulence have

spurred much interest in their study (for recent reviews see refs.
[1–8]). There are nine well-known protein secretion systems in
diderms (numbered from T1SS to T9SS), but others probably
remain undiscovered [9].
Few computational tools exist to identify and characterize pro-
tein secretion systems in bacterial genomes (for a list see Table 1 of
[9]). Their development becomes urgent in light of the availability
of many thousands of genomes and the ease with which new ones
can be sequenced. These tools should be able to identify compo-
nents of the protein secretion systems and to assess whether they
are sufficient at defining an instance of a given system. When the
components are highly conserved proteins, they can be identified
with high sensitivity. The identification of fast-evolving compo-
nents might be more complicated because of poor sequence con-
servation. Additionally, some components may not be strictly
necessary for a functional system, and it may be difficult to know
which of the two factors explains their absence from an instance of
a system. Under these conditions, it is useful for bioinformatics
purposes to split the components of secretion systems into those
that should be present in the instance (“mandatory”) and those
that may be absent (“accessory”). The former correspond to highly
conserved, easily identifiable components, the latter to compo-
nents that may be lacking in systems either because they are miss-
ing or because they were not detected. This nomenclature presumes
nothing about the biological role of the accessory components:
they may be biologically essential but unidentifiable by sequence
similarity searches. The goal of this classification is to describe the
system in a way that facilitates its identification in genomes. We will
use it throughout this text.
The evolution of secretion systems has involved the co-option
of many components from other molecular machineries. These
components have sometimes been co-opted in their turn for other
cellular machineries. As a result, many components of protein
secretion systems have homologs in other systems [10, 11]. This
increases the risk of misidentifications. For example, the T3SS and
the flagellum are evolutionarily related, and several of their core
components belong to homologous families [10]. In this specific
case, discrimination between the two systems is facilitated by the
existence of flagellum-associated mandatory proteins that are
always absent from T3SS (e.g., FlgB), and vice versa (e.g., secre-
tin). These components can be qualified as “forbidden” in the
other system to prevent misidentifications.
The discrimination between protein secretion systems and
other molecular systems can also be improved by the analysis of the
genetic context. For example, T3SS components are usually
encoded in a single locus, and this facilitates their discrimination
from the components of the flagellum [12]. Another example is
 In Silico Identification of Secretion Systems 3

provided by the three mandatory components of T1SS, all of which

have homologs in other systems [13], even if none of the other
systems actually includes all three of them [14]. Importantly, the
abc and mfp components are systematically encoded in the same
locus in T1SS (and only in this system). Hence, three types of
information facilitate the unambiguous detection of secretion sys-
tems: the identification of pertinent and forbidden components,
the completeness of the set of components, and their genetic orga-
nization. This information can be put together in a model of the
system that can be used by a computer program called MacSyFinder.
This program searches genomes for instances satisfying the charac-
teristics described in the model [15].
Sometimes good computational models of protein secretion
systems are not available. The creation of novel (or better) models
requires the identification of the relevant components and their
genetic organization. Most secretion systems are studied on a small
number of bacteria, where they are sometimes remarkably well
characterized in terms of their components, their genetic regula-
tion, their structure, and sometimes their assembly pathways. In
contrast, the other instances of the systems can be very poorly
characterized. The challenge posed to the researcher interested in
identifying novel instances of a given type of system is thus to pro-
duce models with pertinent descriptions of the current knowledge
of the system. This is difficult because the number of components
and their organization may vary widely. For example, the T4SS
locus of Legionella pneumophila is encoded by more than twice the
number of genes of the vir T4SS of Agrobacterium tumefaciens
[16, 17]. Most T4SS are encoded in a locus, but there are intracel-
lular pathogens where they are encoded in several distant loci [18].
The key point in the production of novel models is thus the iden-
tification of traits that are conserved and can be most useful to
identify a certain type of system.
The production of models involves generalizing knowledge
obtained from specific examples. These models are quantitative
representations of the composition and organization of known sys-
tems. When they work, they dramatically facilitate the identifica-
tion of homologous systems. When they fail, they highlight gaps in
our understanding of systems, which often raises interesting bio-
logical questions.
This chapter shows how to use MacSyFinder to identify pro-
tein secretion systems with the predefined models of TXSScan (see
Fig. 1, Subsection 3.3). These models define the components of
the system, the minimum number of mandatory and accessory
components (quorum), and their genetic organization. They have
been validated and shown to perform well: they identify the vast
majority of known systems [9]. They have recently been used to
identify over 10,000 systems in bacteria (available in MacSyDB/
TXSSdb; see Table 1). Yet these models may be inadequate in
4 Sophie S. Abby and Eduardo P.C. Rocha

Fig. 1 Models of protein secretion systems available in TXSScan. The T1SS model is displayed in Fig. 2. Each
box represents a component and its status in the system’s model: “mandatory” (plain), “accessory” (dashed),
or “forbidden” (red cross). Within a system panel: the families of homologous proteins are represented in col-
umns and colored identically. When applicable, the system of origin is indicated next to the box(es). The colo-
calization parameter of the system is indicated (d). When this parameter is specific to a gene (e.g., the relaxase
in T4SS), it is indicated by a subscript (e.g., drel). Additional features specific to a component can be found in
Table 2. Curved double-headed arrows indicate exchangeable components. Figures and legends are freely
reproduced with modification from reference [9] [as specified by the Creative Commons Attribution (CC BY)
license version 4.0]

c­ertain situations. We show how they can be modified or built

from scratch to identify novel variants of a secretion system. The
models can then be easily shared.
MacSyFinder uses hidden Markov model (HMM) protein pro-
files specified in a model to search for components of a system in a
file of protein sequences. It collects all clusters of colocalized com-
ponents in the genome and checks whether they are valid relative
to the specifications of the model. MacSyFinder then outputs the
results of the protein profile searches and the information about
the identified secretion systems. In the next section, we indicate
 In Silico Identification of Secretion Systems 5

Table 1
Useful online resources to design models for the detection of secretion systems

Resource [reference] Type Commands

NCBI/Blast [32] Sequence database indexing makeblastdb
Sequence similarity search blastp
Silix [25] Sequence clustering silix
MAFFT [21] Multiple sequence alignment mafft
Muscle [22] Multiple sequence alignment muscle
Seaview [23] Sequence alignment, edition, and phylogeny –
Jalview [24] Sequence alignment, edition, and analysis –
HMMER [20] Build HMM profiles and use them for sequence hmmbuild
similarity searches hmmsearch
PFAM [34] Database of HMM protein profiles –
TIGRFAM [35] Database of HMM protein profiles –
InterProScan [36] Identify conserved domains using several resources –
MacSyFinder [15] Macromolecular system detection from models of macsyfinder
systems
MacSyView [15] Visualization of MacSyFinder’s results in a web browser –
TXSScan [9] MacSyFinder-based tools to detect T1SS-T9SS and related macsyfinder
appendages (T4P, Tad, bacterial flagellum)
TXSSdb [9] Database of secretion systems detected with TXSScan in –
1,528 genomes of diderm bacteria

the data and software required to use and design models for
MacSyFinder. In the last section, we describe how to define models
and protein profiles to identify protein secretion systems.

2  Materials

2.1  Sequence Data MacSyFinder analyzes protein sequences to identify protein secre-
tion systems. The protein sequences should all be stored in a single
file in Fasta format. This file represents one of several types of
information that must be specified by the “--db-type” option:
●● When the protein sequences are from diverse sources (e.g.,
peptides from a metagenome), the file type is “unordered.” In
this case, the program simply outputs the results of the identi-
fication of the systems’ components using the protein
profiles.
6 Sophie S. Abby and Eduardo P.C. Rocha

●● When the proteins are from a single genome but the relative
order of the corresponding genes in the genome is not known,
the file type is “unordered_replicon.” In this case, the program
can identify components and check whether the quorum of
components is respected. It cannot, however, check the genetic
organization of the model.
●● When proteins are from a single genome and are ordered fol-
lowing the position of genes in the genome, then the file type
is “ordered_replicon.” The “ordered_replicon” mode allows
the use of all available criteria (quorum of components and
genetic organization) to identify instances of the system and as
such is the most powerful mode. It should be used whenever
possible to create, test, and validate a new model and the cor-
responding protein profiles. It will constitute the focus of this
protocol.
●● The “gembase” type is similar to the “ordered_replicon” but
has special identifiers that allow the analysis of multiple
“ordered” genomes in a single step (see MacSyFinder
documentation).

2.2  Predefined TXSScan is a set of predefined MacSyFinder models to detect the

Models Available best-studied protein secretion systems in diderms (T1SS, T2SS,
in TXSScan T3SS, T4SS, T5SS, T6SS, T9SS, and related appendages). These
models are used as examples throughout the following sections.
The files of TXSScan (HMM profiles and model files) should be
placed at a recognizable location (see Subsection 3.3.1) to be used
locally with the standalone program MacSyFinder. One can add to
these directory profiles built by the user or retrieved from public
databases. MacSyFinder can also be used online with the TXSScan
models (see Table 1). Currently, only the standalone version allows
the modification of TXSScan models and the introduction of novel
protein profiles.

2.3  Software Table 1 displays resources of interest for this protocol. To run
MacSyFinder, one needs to install the NCBI/BLAST tools (in par-
ticular, makeblastdb version 2.8 or higher or formatdb), HMMER,
and MacSyFinder [15, 19, 20]. The latter requires a Python inter-
preter (version 2.7) that must be installed beforehand. See
MacSyFinder’s online documentation for more details [15].
To build novel HMM protein profiles, one also needs a pro-
gram to make multiple sequence alignments (e.g., MAFFT or
Muscle [21, 22]), an alignment editor (e.g., Seaview or Jalview
[23, 24]), and a program to cluster proteins by sequence similarity
(e.g., Silix or MCL [25, 26]). The program MacSyView can be
used to visualize the results of MacSyFinder.
 In Silico Identification of Secretion Systems 7

3  Methods

The procedure to design a novel model follows a number of steps

(see Fig. 2). Briefly, it starts with the identification of the relevant
components, and their genetic organization, from the reference
data set of experimentally validated instances of the system. The
HMM protein profiles for each component can be built from the
reference data set or retrieved from public databases. These three
types of information (component list, genetic architecture, and
profiles) can then be used to formulate the model. Finally, the
model is used to analyze an independent data set of experimentally
validated systems. At the end of this process, one should be able to
use the model to identify novel instances of the system and know
the sensitivity of the procedure.

components genetic organization

abc mfp ‘loner’
Biological OMF d>5
expertise abc mfp omf d≤5
MFP or omf ‘multi_system’
ABC d≤5
abc mfp

Reference systems in fasta files Genomes (protiens) in fasta files

Data extraction
from databanks
(NCBI,EBI,JGI)

1) protein family clustering

Protein profiles 2) multiple sequence alignment
design 3) alignment curation
or or 4) HMMER
Extraction from TIGRFAM
databanks or
PFAM abc.hmm mfp.hmm omf.hmm

T1SS
Decision rules ABC (essential)
MFP (essential) components list
Model design quorum rules
OMF (essential, multiple systems)
d ≤ 5 genes, OMF in trans co-localization rules

<system inter_gene_max_space= “5”> T1SS.xml

Model transcription <gene name = “abc” presence = “mandatory”/>
XML file writing <gene name = “mfp” presence = “mandatory”/>
<gene name = “omf” presence = “mandatory” multi_system= “1” loner = “1”/>
</system>

Fig. 2 Overview of the protocol

8 Sophie S. Abby and Eduardo P.C. Rocha

3.1  Compilation Gathering the current biological knowledge of the system, espe-
of Available cially of its experimentally studied instances, may not be simple.
Information The genes and proteins of experimentally studied systems are often
on the System named differently; sequence similarity searches may thus be neces-
sary to establish which components in a system are homologous to
those of other systems. Occasionally, gene fusions and fissions fur-
ther complicate the identification of families of homologs. Once
the relationships of homology between known instances of the sys-
tem are well established, its key components can be inventoried
and qualified (as mandatory, accessory, and forbidden). One can
then characterize the system in terms of its genetic organization
and quorum (i.e., how many components are required for a sys-
tem, see Subsection 3.1.4). The set of experimentally validated pro-
tein secretion systems should be split into two independent data
sets (see Note 1). The reference data set is used to characterize the
protein secretion systems, build the models, and construct the pro-
tein profiles (if required; see Notes 2 and 3). The validation data
set is used to validate the final model (see Subsection 3.5). There
are alternative procedures for systems with few experimentally vali-
dated systems (see Note 4).

3.1.1  Identification The most frequently encountered components of a secretion

and Classification system are typically classed as “mandatory.” They may be
of Secretion System inferred from the frequency of each protein family in the reference
Components data set or retrieved from the literature (see Subsections 3.1.3
and 3.1.4) (see Note 2). The other components are classed as
“accessory.” When it is necessary to distinguish one system from
other systems with many homologous components, it may be
relevant to introduce and class some genes specific to the other
systems as ­“forbidden,” as for the T2SS (e.g., PilM) or the T3SS
(FlgB) (see Fig. 1). Table 2 lists the features and terms available
to describe a model.

3.1.2  Extraction of HMM Each component (called a “gene” in the model) must be associ-
Protein Profiles ated with at least one HMM protein profile. Several public data-
from Databanks banks of protein profiles can be queried by keyword (e.g., the name
of the gene) or by sequence for a match to the component (see
Table 1). InterProScan is a good starting point for this task since it
integrates the information on protein profiles from many data-
banks, including PANTHER, PFAM, SUPERFAMILY, and
TIGRFAM [27].
At the end of this procedure, one often has many profiles for
some components and none for others. One can pick the best
matching profile for each component (see Note 1). However,
sometimes there is no profile that is clearly better than the others.
This occurs when profiles match only certain subfamilies of com-
ponents, for example, because they are very divergent in sequence.
In these cases, it is possible to specify multiple profiles (see Note 5).
Table 2
Keywords to specify models

Keyword
(Command-line override) Required? (Default) Level(s) Description
system Yes Upper (first) Keyword needed to start definition of
system model
inter_gene_max_space Yes system, gene “colocalization” parameter. Defines
(--inter-gene-max-space) maximum number of genes between
two consecutive components of
system for a cluster of components to
be created
min_genes_required No (number of mandatory genes or min_ system Minimum number of mandatory and
(--min-genes-required) mandatory_genes_required if declared) accessory components
min_mandatory_genes_required No (number of mandatory genes, or min_ system Minimum number of mandatory
(--min-mandatory-genes-required) mandatory_genes_required if declared) components
multi_loci No (false, i.e., “0”) Whether system can be encoded on
(--multi-loci) several main loci
gene Yes Second Component of system
multi_system No (false, i.e., “0”) gene Whether gene can participate in several
instances of a system
name Yes gene Name of gene to be used in report files,
and base name of corresponding
HMM protein profile

(continued)
In Silico Identification of Secretion Systems
9
Table 2
10

(continued)

Keyword
(Command-line override) Required? (Default) Level(s) Description
loner No (false, i.e., “0”) gene Whether gene can be found encoded
outside of system’s main loci, i.e., an
exception to colocalization rule
defined at system level
exchangeable No gene Whether gene can be replaced by
another gene in list of components
(quorum); the gene can then be
exchanged with one of its
“homologs” or “analogs”
analogs No gene List of “analogs” to gene
Sophie S. Abby and Eduardo P.C. Rocha

homologs No gene List of “homologs” to gene

system_ref No gene Specifies original system of gene when it
was described in another system’s
model
For more details, see http://macsyfinder.readthedocs.org/en/latest/system_definition.html#the-xml-hierarchy
 In Silico Identification of Secretion Systems 11

The user must build a novel protein profile in one of three

typical situations: (1) when none is available in the databases, (2)
when one profile can replace a large set of very specific profiles, or
(3) when one wishes to build profiles that are specific to a system
because existing profiles also match components of other types of
systems. The new profile should be built using the sequences of the
reference data set (see also Notes 1 and 3).

3.1.3  Establishing By default, MacSyFinder searches for clusters of colocalized genes

the Model of the Genetic encoding components of the system (“ordered” data sets) (see
Architecture Subsection 2.1). The maximum distance allowed between consec-
utive components (“inter_gene_max_space”) can be the same for
all components or be specific to a particular component. The dis-
tance is measured in genes, i.e., a distance of three means that
there can be up to three genes between two consecutive compo-
nents of the system. This distance can be inferred from the refer-
ence data set or from the secretion systems identified in bacterial
genomes (see Note 6). Additionally, one may define some genes as
“loners,” in which case the model allows them to be encoded out-
side the clusters of colocalized genes (Subsection 3.2.1).
The overall genetic architecture of a system can be specified
through the “multi_loci” attribute. The default value (False or
“0”) indicates that the system is encoded in a single locus (except
for the loner components), whereas the alternative (“1”) autho-
rizes the existence of multiple clusters for an instance of the system.
For example, this option is useful for describing the type IV pili or
T9SS, since they are often encoded in several loci [9].

3.1.4  Defining The quorum of the model is the minimum number of components
the Quorum required to validate an instance of the system. It is defined by two
of Components parameters: the minimum number of mandatory components
(“min_mandatory_genes_required”) and the minimum number of
mandatory and accessory components (“min_genes_required”).
The values of both parameters are set by default to the number of
mandatory components, and in this case the accessory components
do not actually count in the quorum. To make them count, one
must specify higher values for the second than for the first crite-
rion. It is particularly important to use prior knowledge to assess
the relevance of a locus lacking a mandatory gene.
The quorum can be optimized using the information from the
secretion systems identified in bacterial genomes (see Note 7).
Changes in the values of the quorum affect the sensitivity and spec-
ificity of the method. Low values authorize the validation of sys-
tems with fewer components that are less similar to the reference
data set, thereby increasing the number of detected systems. This
might come at the cost of misidentifications or the validation of
nonfunctional systems. Running the models with higher values of
the quorum results in the identification of instances that are more
12 Sophie S. Abby and Eduardo P.C. Rocha

similar to the reference data set and thus more likely to be true. On
the other hand, this results in the identification of fewer instances
and may exclude those that are too distinct from those in the refer-
ence data set.

3.2  Model After gathering the available information on the protein secretion
Formulation system, one must write down the model in the relatively simple
XML format predefined for MacSyFinder. A list of the keywords of
this hierarchical XML grammar is presented in Table 2 and in
MacSyFinder’s online documentation.

3.2.1  Defining the Model We illustrate the formulation of a simple model with the example
in an XML Text File: of the type I secretion system. This system has three mandatory
Example 1—T1SS components (Fig. 2): an ABC transporter (“abc”), a membrane
fusion protein (“mfp”), and an outer membrane porin (“omf”).
The latter can be colocalized with the two other components (less
than six genes apart, “inter_gene_max_space” set to 5), or encoded
apart in the genome (loner attribute set to True: “1”). In addition,
omf can be involved in several occurrences of the system (“multi-­
system” set to True (“1”)). The three mandatory components are
required to form a full system. The quorum is therefore not speci-
fied; it is left at its default value. The file with the model should be
named after the system (“T1SS.xml”). Its content is displayed in
Fig. 2.

3.2.2  Defining the Model Larger systems tend to require more complex models. This is well
in an XML Text File: exemplified by the model to identify T9SS (file “T9SS.xml” [9]).
Example 2—T9SS This model, in line with the existing literature [28–31], states
that T9SS consists of several mandatory and accessory components
encoded in multiple loci (multi_loci = “1”). The quorum allows
one mandatory and several accessory components to be missing.
Some components form clusters, whereas others are defined as lon-
ers (Fig. 1). The SprA component was matched by several profiles
of the PFAM and TIGRFAM databanks; we used the “exchange-
able” and “homologs” attributes to include them all (see Note 5).

3.3  Running HMM protein profiles should be in individual files named after the
MacSyFinder corresponding component and placed in the same directory. The
path to this directory must be provided with the option “-p”. All
3.3.1  Organizing
profiles must have the same file name extension (“.hmm” by
the Data
default, though it can be changed with the option “--profile-­
suffix”). For the T1SS model mentioned earlier, the program
requires three HMM files in a “profiles” directory: “abc.hmm,”
“mfp.hmm,” and “omf.hmm”.
The path to the directory within the current working directory
containing the XML model file must also be given in the input
(option “-d,” named “definitions,” for example).
 In Silico Identification of Secretion Systems 13

3.3.2  Identification In this section, we provide an example of the procedure to identify

of Secretion Systems instances of T1SS in a protein file of the type “ordered_replicon”
(“CP000521_proteins.fasta”). The command line to launch the
program is as follows:
macsyfinder -p profiles -d definitions --db-
type ordered_replicon --replicon-­ topology
circular --sequence-db CP000521_proteins.
fasta T1SS
In our experience, the identification of multiple unrelated
secretion systems is best done independently, i.e., in successive runs
for the different models. This is especially important when the sys-
tems to be searched for have many homologous components. Yet,
when systems are unrelated, one can identify them at the same time.
For example, to identify both T1SS and T9SS, one should type:
macsyfinder -p profiles -d definitions --db-
type ordered_replicon --replicon-­ topology
circular --sequence-db CP000521_proteins.
fasta T1SS T9SS

3.4  Output Files, The results of the detection are printed to files in simple text for-
and Visualization mat that can be read by any word-processing application. They are
of Results stored in the MacSyFinder’s output directory (specified using the
with MacSyView “-o” option, or named automatically). They include configuration
and log files, along with the results of MacSyFinder and
3.4.1  Finding Relevant HMMER. The description of the output files is detailed in
Information MacSyFinder’s documentation: http:// macsyfinder.readthedocs.
in MacSyFinder’s Output org/en/latest/outputs.html.
Files Some output files are particularly useful for improving the
design of a new model. The standard output (“macsyfinder.out”)
shows how the tool processed the instances of the different com-
ponents to validate the systems. Together with the files storing raw
or filtered HMMER hits (in the “hmmer_results” directory), it
gives an overview of the detection process, from the detection of
components to the validation of the clusters. For automated down-
stream analyses of MacSyFinder’s results, the files “macsyfinder.
report” and “macsyfinder.summary” are of particular interest. The
former contains the list of components identified in each instance
with information on the quality of the HMMER matches. The lat-
ter describes the content of each instance of a secretion system and
can be loaded as a dictionary object using a Python script. It can be
particularly useful in the optimization steps (see Notes 6 and 7).

3.4.2  Visualization MacSyView allows the user to visualize the results of MacSyFinder
of Results with MacSyView in a web browser. It reads the “results.macsyfinder.json” output
file. Clicking on a system opens a window with three panels show-
ing the number of components per category, the genomic context,
and the statistics of the detection of each component (Fig. 3).
14 Sophie S. Abby and Eduardo P.C. Rocha

Fig. 3 Screenshots of MacSyView application. (a) The application reads the “results.macsyfinder.json” file from
MacSyFinder’s results directory. (b and c) The selection from a list of detected systems leads to a page with a
description of the instance in relation to the model. (d) The graphical representation of the instance in its
genomic context can be downloaded. (e) A table details the identified components (e.g., name, length) along
with their HMMER scores and other statistics

3.5  Optimization The initial models should be very general to assess the diversity of
and Validation the instances of the system. They can then be optimized iteratively
depending on the results obtained in the analysis of the reference
3.5.1  Optimization
data set (and external data). The output files discussed in Subsection
3.4.2 can be used to identify the parts of the model that must be
improved. The specific case of model optimization for poorly stud-
ied systems is described in Notes 6 and 7.

3.5.2  Validation The final model must be tested on an independent validation data
set (see Note 1). This procedure allows the quantification of the
sensitivity of the model, i.e., to know how well it identifies a secre-
tion system and its different components. It is more difficult to
assess its specificity, i.e., its ability to separate true from false
instances, since usually there is no reliable information on invalid
protein secretion systems.
At the end of the validation process, it might seem tempting
(or necessary) to correct the initial model to identify a larger frac-
tion of the instances of the validation data set. It must be borne in
mind that this violates the hypothesis of independence between the
model and the validation data set. Hence, if the model is changed
 In Silico Identification of Secretion Systems 15

to fit the validation data set, then it can no longer be validated

with the same data set. One can build an additional independent
validation data set to circumvent this difficulty.

4  Notes

1. Constructing a reference or validation set of secretion systems

The diversity of the protein secretion systems in the reference
and validation data sets should encompass their diversity in the
genomes where they are going to be searched for. Otherwise,
the model will miss the instances that are very divergent, in
sequence or genetic organization. One can increase the diver-
sity of these sets by sampling instances from all previously
described subtypes of a system (e.g., SPI1, SPI2, Hrp1, Hrp2,
Ysc for T3SS [12]). When this information is unavailable, one
can use all the known systems to build phylogenetic trees of
previously identified key components and then sample repre-
sentatives of every major clade to build the reference data set.
2. Building protein families for the reference secretion systems
We propose the following procedure to build and analyze fam-
ilies of homologous proteins:
(a) Store the protein sequences of the reference set of secre-
tion systems in a multi-fasta file (e.g., “reference_systems.
fasta”). Index the file using the makeblastdb command:
makeblastdb -dbtype prot -in reference_
systems.fasta
(b) Run Blastp for each pair of proteins and generate a tabu-
lated output. Pick a statistical threshold of significance for
the hits (here E-value <10−6):
blastp -query reference_systems.fas-
ta -db reference_systems.fasta -evalue
0.000001 -outfmt 6 -out blastall_refer-
ence_systems.out
Alternatively, this can be done using other sequence similarity
search tools, such as psiblast or jackhmmer.
(c) Identify the protein families present in the data set by
clustering the blastp results with Silix.
silix reference_systems.fasta blastall_
reference_systems.out -f FAM > refer-
ence_systems.fnodes
(d) Save the different families obtained in separate fasta files:
silix-split reference_systems.fasta ref-
erence_systems.fnodes
16 Sophie S. Abby and Eduardo P.C. Rocha

(e) Annotate/label the families. If different components are

clustered together, they must be disambiguated. One can
split families by clustering the proteins with parameters
enforcing higher similarity or alignment coverage. A phy-
logenetic analysis may also show how to separate two pro-
tein families that cluster together (not covered here). In
some cases, clearly separate protein families (and, thus,
protein profiles) are unobtainable. The best solution is to
list them all as a single component in the model and spec-
ify a quorum of one (see Note 5).
3. Designing HMM protein profiles from families of homologous
proteins
Follow these steps to build HMM profiles for the protein fami-
lies (e.g., those obtained following Note 2):
(a) Store the sequences of each protein family in a different
text file in Fasta format. Name the file in accordance with
its content, e.g., “abc.fasta” for the ABC-transporter pro-
tein of T1SS.
(b) Make a multiple-sequence alignment for each family using
MAFFT [21] (or other analogous program):
mafft abc.fasta > abc.aln-­
mafft.fasta
(c) Check each multiple alignment (e.g., using Seaview [23]).
Trim its extremities if they are poorly conserved. Do not
remove columns inside the region of the alignment that is
going to be used to construct the profile. Save the edited
alignment in a new file, e.g., “abc.mafft-edit.fasta.”
(d) Run hmmbuild from the HMMER package [20] on the
edited alignment to create the HMM protein profile:
hmmbuild --informat afa --amino abc.hmm
abc.mafft-edit.fasta
4. Analyzing poorly characterized systems
The typical model-building procedure cannot be used to
model systems lacking enough experimentally validated
instances to make a reference data set. In this case, one must
use the few known instances to collect homologs from the
genome databases by sequence similarity searches (e.g., using
blast [32], see Note 2). The hits can be used to build protein
families and HMM profiles. A careful analysis of the patterns of
co-­occurrence of the components usually highlights those that
should be specified in the model because they are sufficiently
conserved.
It is usually better to build simple models with weak con-
straints in terms of quorum and colocalization when the refer-
ence data sets are small. For example, every component could
be set as “loner” and the quorum set to one or two. The model
can then be optimized iteratively (see Subsection 3.5.1).
 In Silico Identification of Secretion Systems 17

Naturally, one should be cautious when drawing conclusions

from studies using models that could not be validated with an
independent data set of experimentally validated systems.
5. Identifying a component using multiple protein profiles
A single protein profile may not be sufficient to identify all
instances of a given component. In such cases, one can associ-
ate several protein profiles to a single component in the quo-
rum of the system. These genes should be qualified as
“exchangeable,” and “homologs” or “analogs.” For example,
T3SS has one of three different subfamilies of secretins
(T3SS_sctC, T2SS_gspD, Tad_rcpA, depending on the T3SS
subtype, [12]). To detect them correctly, one can define it in
the T3SS model as follows:
<gene name="T3SS_sctC" presence="mandatory"
exchangeable ="1">
<homologs>
<gene name="T2SS_gspD" system_ref="T2SS"/>
<gene name="Tad_rcpA" system_ref="T4P"/>
</homologs>
</gene>
With this model, MacSyFinder will identify a secretin when it
finds a hit for any of these three protein profiles. The quorum
will increase by one when there is one or multiple hits to any of
these profiles. The keyword “system_ref” indicates that the two
genes T2SS_gspD and Tad_rcpA are defined in the models of
T2SS and Tad, respectively. When the other models are accessi-
ble to MacSyFinder, the components do not need to be rede-
fined in the “homologs” or “analogs” section of T3SS. When
the other models are unavailable, then the components must be
described in the system’s model. We show an example of the lat-
ter case concerning the SprA gene of T9SS in Subsection 3.2.2.
6. Optimizing the colocalization criterion
The procedure for optimizing the colocalization criterion starts
by setting it to a value higher than expected (but not too high;
otherwise, several occurrences of the system could be agglom-
erated). This produces large clusters that are expected to con-
tain all relevant colocalized genes. This parameter may be
subsequently refined by plotting the distribution of the maxi-
mum distance found between two consecutive components in
the clusters of the reference data set. To illustrate, we searched
for “single-locus” T6SS in a set of 1,528 bacterial genomes
using a minimum colocalization distance of 20 genes [9]:
macsyfinder -p profiles -d definitions --db-
type gembase --sequence-db bacterial_genom-
es_proteins.fasta –o macsyfinder_opt_coloc_
T6SS --inter-gene-­ max-space T6SS 20 T6SS
18 Sophie S. Abby and Eduardo P.C. Rocha

Fig. 4 Optimization of the quorum and colocalization parameters for T6SSi. (a) Distribution of number of differ-
ent components of T6SSi colocalized (d ≤ 20). (b) Distribution of maximum distance between consecutive
components in T6SSi detected with d ≤ 20. The minimum number of components required for a T6SSi was set
to 11 in the T6SSi model, which corresponds to the start of the second peak in the distribution represented in
panel (a). T6SSi detected as full systems are colored in blue. Figures and legends are freely reproduced with
modification from reference [9] (as specified by the Creative Commons Attribution (CC BY) license version 4.0)

The distribution of distances actually observed in genomes

suggests that a smaller value (e.g., 14) is enough to identify
most of the relevant clusters (Fig. 4b).
7. Optimizing the quorum criterion
The quorum can be optimized in multiple ways, depending on
the distribution of the secretion systems across genomes.
If the system is encoded at a single locus (but may be found
in several copies per replicon), the quorum can be optimized
by studying the distribution of the number of different compo-
nents detected in each cluster. For this purpose, one can use a
model with a very relaxed quorum criterion (e.g., set to “1”)
and draw the distribution of the number of components found
in each cluster (with at least one component) (Fig. 4a). This
can be done directly in the command line, for example, using
the T6SS model available in TXSScan:
macsyfinder -p profiles -d definitions --db-
type gembase --sequence-­
db bacterial_genomes_
proteins.fasta –o macsyfinder_opt_quorum_T6SS
--min-genes-required T6SS 1 --min-mandatory-
genes-required T6SS 1 T6SS
The number of (accessory and mandatory) components in each
cluster is indicated in the “macsyfinder.summary” output file in
the “macsyfinder_opt_quorum_T6SS” folder, in the eighth
 In Silico Identification of Secretion Systems 19

column, “Nb_Ref_Genes_detected_NR.” In this specific exam-

ple, it shows many clusters with more than 11 components
(Fig. 4a). This is in line with the presence of more than 13 core
components in most T6SS [33]. A final model with a quorum
set to 11 would thus be able to identify novel instances of the
T6SS accurately [9].
When the system is typically encoded in a single copy per
genome scattered in several loci (“multi_loci”), the analysis of
the clusters is less informative, especially if there are other sys-
tems in the genome with homologs to these components.
Nevertheless, it can be complemented with the information on
the number of components per replicon. This analysis should
use low stringency colocalization and quorum parameters, for
example, all genes can be set to “loner” and the quorum to a
small value. We illustrate this by searching to optimize the T9SS
model (see Fig. 1 and Subsection 3.2.2):
(a) Copy the “T9SS.xml” file as “T9SS_loner.xml.”
(b) Add “loner = ‘1’” to the definition line of each gene.
(c) Alter the values of the parameters “min_genes_required”
and “min_mandatory_genes_required,” either by using
the command line (as in the following command line) or
by modifying the XML file in the system definition line
(“min_mandatory_genes_required = ‘1′ min_genes_
required = ‘1′”).
(d) Run MacSyFinder on the modified version of the model:
macsyfinder -p profiles -d definitions
--db-type gembase --sequence-­ db bacte-
rial_genomes_proteins.fasta –o macsyfind-
er_opt_quorum_T9SS --min-genes-required
T9SS_loner 1 --min-mandatory-genes-re-
quired T9SS_loner 1 T9SS_loner
The distribution of the number of (accessory and mandatory)
components shows many replicons with more than seven com-
ponents. The model using a quorum of seven is able to identify
novel instances of T9SS accurately [9].

References
1. Kanonenberg K, Schwarz CK, Schmitt L molecular architecture and mechanism. Nat
(2013) Type I secretion systems—a story of Rev Microbiol 10(5):336–351
appendices. Res Microbiol 164(6):596–604 4. Galan JE, Lara-Tejero M, Marlovits TC,
2. Campos M, Cisneros DA, Nivaskumar M, Wagner S (2014) Bacterial type III secretion
Francetic O (2013) The type II secretion sys- systems: specialized nanomachines for protein
tem—a dynamic fiber assembly nanomachine. delivery into target cells. Annu Rev Microbiol
Res Microbiol 164(6):545–555 68:415–438
3. Korotkov KV, Sandkvist M, Hol WG (2012) 5. Alvarez-Martinez CE, Christie PJ (2009)
The type II secretion system: biogenesis, Biological diversity of prokaryotic type IV
20 Sophie S. Abby and Eduardo P.C. Rocha

secretion systems. Microbiol Mol Biol Rev type IV secretion system. Infect Immun
73:775–808 78(5):1809–1823
6. van Ulsen P, Rahman S, Jong WS, Daleke-­ 19. Camacho C, Coulouris G, Avagyan V, Ma N,
Schermerhorn MH, Luirink J (2014) Type V Papadopoulos J, Bealer K, Madden TL (2009)
secretion: from biogenesis to biotechnology. BLAST+: architecture and applications. BMC
Biochim Biophys Acta 1843(8):1592–1611 Bioinformatics 10:421
7. Zoued A, Brunet YR, Durand E, Aschtgen 20. Eddy SR (2011) Accelerated profile HMM
MS, Logger L, Douzi B, Journet L, Cambillau searches. PLoS Comput Biol 7(10):e1002195
C, Cascales E (2014) Architecture and assem- 21. Katoh K, Toh H (2010) Parallelization of the
bly of the Type VI secretion system. Biochim MAFFT multiple sequence alignment pro-
Biophys Acta 1843(8):1664–1673 gram. Bioinformatics 26(15):1899–1900
8. McBride MJ, Nakane D (2015) Flavobacterium 22. Edgar RC (2004) MUSCLE: multiple sequence
gliding motility and the type IX secretion sys- alignment with high accuracy and high
tem. Curr Opin Microbiol 28:72–77 throughput. Nucleic Acids Res 32:1792–1797
9. Abby SS, Cury J, Guglielmini J, Néron B, 23. Gouy M, Guindon S, Gascuel O (2010)
Touchon M, Rocha EPC (2016) Identification SeaView version 4: a multiplatform graphi-
of protein secretion systems in bacterial cal user interface for sequence alignment and
genomes. Sci Rep 6:23080 phylogenetic tree building. Mol Biol Evol
10. Ginocchio CC, Olmsted SB, Wells CL, Galan 27(2):221–224
JE (1994) Contact with epithelial cells induces 24. Waterhouse AM, Procter JB, Martin DM, Clamp
the formation of surface appendages on M, Barton GJ (2009) Jalview Version 2—a mul-
Salmonella typhimurium. Cell 76(4):717–724 tiple sequence alignment editor and analysis
11. Peabody CR, Chung YJ, Yen MR, Vidal-­ workbench. Bioinformatics 25(9):1189–1191
Ingigliardi D, Pugsley AP, Saier MH Jr (2003) 25. Miele V, Penel S, Duret L (2011) Ultra-fast
Type II protein secretion and its relationship sequence clustering from similarity networks
to bacterial type IV pili and archaeal flagella. with SiLiX. BMC Bioinformatics 12:116
Microbiology 149(Pt 11):3051–3072 26. Enright AJ, Van Dongen S, Ouzounis CA
12. Abby SS, Rocha EP (2012) The non-flagellar (2002) An efficient algorithm for large-scale
type III secretion system evolved from the bac- detection of protein families. Nucleic Acids Res
terial flagellum and diversified into host-cell 30(7):1575–1584
adapted systems. PLoS Genet 8(9):e1002983 27. Mitchell A, Chang HY, Daugherty L, Fraser M,
13. Holland IB, Schmitt L, Young J (2005) Type Hunter S, Lopez R, McAnulla C, McMenamin
1 protein secretion in bacteria, the ABC-­ C, Nuka G, Pesseat S, Sangrador-­ Vegas A,
transporter dependent pathway. Mol Membr Scheremetjew M, Rato C, Yong SY, Bateman
Biol 22(1–2):29–39 A, Punta M, Attwood TK, Sigrist CJ, Redaschi
14. Paulsen IT, Park JH, Choi PS, Saier MH Jr N, Rivoire C, Xenarios I, Kahn D, Guyot D,
(1997) A family of gram-negative bacterial Bork P, Letunic I, Gough J, Oates M, Haft
outer membrane factors that function in the D, Huang H, Natale DA, Wu CH, Orengo C,
export of proteins, carbohydrates, drugs and Sillitoe I, Mi H, Thomas PD, Finn RD (2015)
heavy metals from gram-negative bacteria. The InterPro protein families database: the
FEMS Microbiol Lett 156(1):1–8 classification resource after 15 years. Nucleic
15. Abby SS, Neron B, Menager H, Touchon M, Acids Res 43(Database issue):D213–D221
Rocha EP (2014) MacSyFinder: a program to 28. Shrivastava A, Johnston JJ, van Baaren JM,
mine genomes for molecular systems with an McBride MJ (2013) Flavobacterium johnso-
application to CRISPR-Cas systems. PLoS One niae GldK, GldL, GldM, and SprA are required
9(10):e110726 for secretion of the cell surface gliding motil-
16. Christie PJ (2004) Type IV secretion: the ity adhesins SprB and RemA. J Bacteriol
Agrobacterium VirB/D4 and related con- 195(14):3201–3212
jugation systems. Biochim Biophys Acta 29. McBride MJ, Zhu Y (2013) Gliding motility
1694(1–3):219–234 and Por secretion system genes are widespread
17. Franco IS, Shuman HA, Charpentier X (2009) among members of the phylum Bacteroidetes.
The perplexing functions and surprising origins J Bacteriol 195(2):270–278
of Legionella pneumophila type IV secretion 30. Zhu Y, McBride MJ (2014) Deletion of the
effectors. Cell Microbiol 11(10):1435–1443 Cytophaga hutchinsonii type IX secretion
18. Gillespie JJ, Brayton KA, Williams KP, Diaz system gene sprP results in defects in glid-
MA, Brown WC, Azad AF, Sobral BW (2010) ing motility and cellulose utilization. Appl
Phylogenomics reveals a diverse Rickettsiales Microbiol Biotechnol 98(2):763–775
 In Silico Identification of Secretion Systems 21

31. Kharade SS, McBride MJ (2015) Flavo­ 34. Finn RD, Tate J, Mistry J, Coggill PC, Sammut
bacterium johnsoniae PorV is required for SJ, Hotz HR, Ceric G, Forslund K, Eddy SR,
secretion of a subset of proteins targeted Sonnhammer EL, Bateman A (2008) The
to the type IX secretion system. J Bacteriol Pfam protein families database. Nucleic Acids
197(1):147–158 Res 36(Database issue):D281–D288
32. Altschul SF, Gish W, Miller W, Myers EW, 35. Haft DH, Selengut JD, Richter RA, Harkins
Lipman DJ (1990) Basic local alignment search D, Basu MK, Beck E (2013) TIGRFAMs and
tool. J Mol Biol 215:403–410 Genome Properties in 2013. Nucleic Acids Res
33. Boyer F, Fichant G, Berthod J, Vandenbrouck 41(Database issue):D387–D395
Y, Attree I (2009) Dissecting the bacterial 36. Quevillon E, Silventoinen V, Pillai S, Harte
type VI secretion system by a genome wide N, Mulder N, Apweiler R, Lopez R (2005)
in silico analysis: what can be learned from InterProScan: protein domains identi-
available microbial genomic resources? BMC fier. Nucleic Acids Res 33(Web Server
Genomics 10:104 issue):W116–W120
 Chapter 2

Protein Sorting Prediction

Henrik Nielsen

Abstract
Many computational methods are available for predicting protein sorting in bacteria. When comparing
them, it is important to know that they can be grouped into three fundamentally different approaches:
signal-based, global-property-based and homology-based prediction. In this chapter, the strengths and
drawbacks of each of these approaches is described through many examples of methods that predict secre-
tion, integration into membranes, or subcellular locations in general. The aim of this chapter is to provide
a user-level introduction to the field with a minimum of computational theory.

Key words Protein sorting, Subcellular location, Secretion, Transmembrane proteins, Prediction,
Machine learning

1  Introduction

Protein sorting prediction—in other words, inferring the subcel-

lular location (SCL) of proteins from their amino acid sequences—
has a long history in bioinformatics. The first attempts at predicting
the best known sorting signals, the transmembrane α-helix (TMH)
and the secretory signal peptide (SP), were published in 1982–1983,
long before bioinformatics was even established as a field [1, 2].
Since then, a plethora of methods for predicting sorting signals
and SCL have been published, and it can be a daunting task to
select the most relevant and reliable methods for analyzing a set of
sequences.
Of course, the development of algorithms and the growth in
available training data have led to an increase in the predictive per-
formance of the available methods. In 2005, some of the authors
of the PSORTb method for predicting SCL in bacteria [3] even
concluded that “on average, recent high-precision computational
methods such as PSORTb now have a lower error rate than labora-
tory methods” [4]. This conclusion should be taken with a grain of
salt; first, it applies only to high-throughput laboratory methods;
second, it should be remembered that computational methods will

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_2, © Springer Science+Business Media LLC 2017

23
24 Henrik Nielsen

never be better than the data used to train them. Nevertheless, the
authors had a point regarding the experimental sources of error
that can easily render a high-throughput experiment less reliable
than a well-trained computational method.
It can be difficult, however, to decide what to believe when the
authors of every computational method tend to describe their per-
formance as being superior to all others. There are different ways
of defining the problem, different ways of measuring the perfor-
mance, and different prerequisites used for prediction. The aim of
this chapter is not to provide a definitive answer to which method
is best for which problem—such a checklist would quickly become
outdated—but instead to give the reader a toolbox for critically
evaluating bioinformatics algorithms. This will involve a number of
examples of computational methods selected for their relevance
for bacteria, with the main focus being on Gram-negative bacteria.
A similar chapter with focus exclusively on Gram-positive bacteria
has been published elsewhere [5]. In general, prediction methods
will be mentioned only if they either provide publicly available web
servers or have strong historical relevance.

2  Three Approaches to Prediction

It is crucial to understand that there are basically three different

approaches to protein sorting prediction. The first approach is
recognition of the actual sorting signals. The aforementioned
early methods for TMH and SP recognition [1, 2] were examples
of this. A number of more recent examples will be given in
Subheadings 5 and 7.
The second approach is prediction based on global proper-
ties of the proteins, for example, their amino acid composition.
This approach was first used to discriminate between intracellu-
lar and extracellular proteins in both prokaryotic and eukaryotic
proteins in 1994 [6]. It has been shown that the main part of the
differences in amino acid composition between intracellular and
extracellular proteins resides in surface-exposed amino acids,
which makes sense since the surfaces should be adapted to the
varying physicochemical environments of the different SCLs [7].
This analysis was done for eukaryotic proteins only, but it would
be fair to assume that the observation holds true also for bacte-
rial proteins.
Two early SCL prediction methods that used only the amino
acid composition were NNPSL in 1998 [8] and SubLoc in 2001
[9] (see Table 1), based on artificial neural networks (ANNs) and
support vector machines (SVMs), respectively (see Subheading 3).
They were limited in their applicability because their data set did
not include any membrane proteins, and they did not distinguish
between Gram-positive and Gram-negative bacteria.
Table 1
Web addresses of servers reviewed in this chapter

Name Website address References

SubLoc http://www.bioinfo.tsinghua.edu.cn/SubLoc/ [9]
PROSITE http://prosite.expasy.org/prosite.html [28]
Pfam http://pfam.xfam.org/ [29]
TIGRFAMs http://www.jcvi.org/cgi-bin/tigrfams/index.cgi [30]
InterPro http://www.ebi.ac.uk/interpro/ [31]
SignalP http://www.cbs.dtu.dk/services/SignalP/ [48–51]
PrediSi http://www.predisi.de/ [56]
SOSUIsignal http://harrier.nagahama-i-bio.ac.jp/sosui/sosuisignal/sosuisignal_submit.html [57]
Signal-BLAST http://sigpep.services.came.sbg.ac.at/signalblast.html [58]
LipoP http://www.cbs.dtu.dk/services/LipoP/ [60]
SPEPlip http://gpcr.biocomp.unibo.it/cgi/predictors/spep/pred_spepcgi.cgi [62]
PRED-LIPO http://bioinformatics.biol.uoa.gr/PRED-LIPO/ [63]
PROSITE profile PROKAR_LIPOPROTEIN http://prosite.expasy.org/PS51257 and http://prosite.expasy.org/PDOC00013
TatFind http://signalfind.org/tatfind.html [65]
TatP http://www.cbs.dtu.dk/services/TatP/ [66]
PRED-TAT http://www.compgen.org/tools/PRED-TAT/ [67]
PROSITE profile TAT http://prosite.expasy.org/PS51318 and http://prosite.expasy.org/PDOC51318
Pfam profile TAT_signal http://pfam.xfam.org/family/PF10518
Protein Sorting Prediction

(continued)
25
Table 1
26

(continued)

Name Website address References

TIGRFAMs profile TAT_signal_seq http://www.jcvi.org/cgi-bin/tigrfams/HmmReportPage.cgi?acc=TIGR01409
SecretomeP 2.0 http://www.cbs.dtu.dk/services/SecretomeP/ [70]
Henrik Nielsen

SecretP 2.0 http://cic.scu.edu.cn/bioinformatics/secretPV2/ [71]

SecretP 2.1 http://cic.scu.edu.cn/bioinformatics/secretPV2_1/ [72]
T4EffPred http://bioinfo.tmmu.edu.cn/T4EffPred/ [78]
T4SEpre Web server at http://biocomputer.bio.cuhk.edu.hk/T4DB/T4SEpre.php, [79]
downloadable version at http://biocomputer.bio.cuhk.edu.hk/softwares/
T4SEpre/
SIEVE http://www.sysbep.org/sieve/ [82]
EffectiveT3 http://www.effectors.org/ [83]
Löwer and Schneider’s method http://gecco.org.chemie.uni-frankfurt.de/T3SS_prediction/T3SS_prediction.html [84]
BPBAac http://biocomputer.bio.cuhk.edu.hk/T3DB/BPBAac.php [85]
T3_MM http://biocomputer.bio.cuhk.edu.hk/T3DB/T3_MM.php [86]
BEAN http://systbio.cau.edu.cn/bean/ [87, 88]
pEffect http://services.bromberglab.org/peffect/ [89]
TMHMM http://www.cbs.dtu.dk/services/TMHMM/ [95]
HMMTOP http://www.enzim.hu/hmmtop/ [96]
Phobius & PolyPhobius http://phobius.sbc.su.se/ [101, 108]
Philius http://www.yeastrc.org/philius/ [102]
MEMSAT3 and MEMSAT-SVM Available through the PSIPRED Protein Sequence Analysis Workbench, http://bioinf. [103, 104]
cs.ucl.ac.uk/psipred/
Name Website address References
OCTOPUS and SPOCTOPUS http://octopus.cbr.su.se/ [105, 106]
SCAMPI http://scampi.cbr.su.se/ [109]
BPROMPT http://www.ddg-pharmfac.net/bprompt/BPROMPT/BPROMPT.html [111]
TOPCONS http://topcons.cbr.su.se/ or http://topcons.net/ [112, 113]
TOPCONS-single http://single.topcons.net/ [114]
PRED-TMBB http://biophysics.biol.uoa.gr/PRED-TMBB/ [117, 118]
ProfTMB https://rostlab.org/owiki/index.php/Proftmb [119, 120]
B2TMPRED http://gpcr.biocomp.unibo.it/cgi/predictors/outer/pred_outercgi.cgi [122]
TBBpred http://www.imtech.res.in/raghava/tbbpred/ [123]
ConBBPRED http://bioinformatics.biol.uoa.gr/ConBBPRED/input.jsp [121]
BOCTOPUS http://boctopus.bioinfo.se/ [124, 125]
BOMP http://services.cbu.uib.no/tools/bomp [126]
HHomp http://toolkit.tuebingen.mpg.de/hhomp [127]
BetAware http://betaware.biocomp.unibo.it/BetAware/ [128, 129]
transFold http://bioinformatics.bc.edu/clotelab/transFold/ [130, 131]
TMBpro http://tmbpro.ics.uci.edu/ [132]
PSORTb http://www.psort.org/psortb/ [3, 42,
134]
Proteome Analyst http://pa.wishartlab.com/pa/pa/ Note: the website requires login, but registration is [15, 33]
free.
Gneg-PLoc http://www.csbio.sjtu.edu.cn/bioinf/Gneg/ [19]
Gpos-PLoc http://www.csbio.sjtu.edu.cn/bioinf/Gpos/ [20]
Protein Sorting Prediction

(continued)
27
Table 1
28

(continued)

Name Website address References

Gneg-mPLoc http://www.csbio.sjtu.edu.cn/bioinf/Gneg-multi/ [21]
Gpos-mPLoc http://www.csbio.sjtu.edu.cn/bioinf/Gpos-multi/ [22]
Henrik Nielsen

iLoc-Gneg http://www.jci-bioinfo.cn/iLoc-Gneg [23]

iLoc-Gpos http://www.jci-bioinfo.cn/iLoc-Gpos [24]
PSLpred http://www.imtech.res.in/raghava/pslpred/ [137]
LocTree3 https://rostlab.org/services/loctree3/ [138]
CELLO http://cello.life.nctu.edu.tw/ [13]
SOSUI-GramN http://harrier.nagahama-i-bio.ac.jp/sosui/sosuigramn/sosuigramn_submit.html [140]
MetaLocGramN http://iimcb.genesilico.pl/MetaLocGramN/ [135]
 Protein Sorting Prediction 29

Using only the amino acid composition for prediction, of

course, means throwing away all sequence information, including
possible signatures of actual sorting signals. One way to retain
some of this information while still keeping a fixed number of
parameters is to count the occurrences of amino acid pairs, either
adjacent or separated by a small distance. Nakashima and Nishikawa
in 1994 [6] thus found that including the composition of amino
acid pairs with a separation distance of up to five positions improved
predictive performance.
The third approach is prediction by sequence homology. When
trying to predict functional aspects of an unknown protein, the
standard procedure is to do a BLAST search [10] and then infer
such aspects from the functional annotations of the found homo-
logs. Therefore, the intuitive expectation is that such a procedure
will also work for SCL—in other words, that a protein tends to stay
in the same compartment in the course of evolution. Indeed, a
significant part of the so-called subcellular location annotations of
bacteria in Swiss-Prot (the manually annotated part of UniProt
[11]), are found with “sequence similarity” as the evidence (more
than five times as many as the corresponding annotations with
experimental evidence).
However, it is not trivial to determine how similar a pair of
proteins must be in order to draw an inference about SCL. Nair
and Rost [12], working with eukaryotic proteins only, concluded
that more than 70% identical residues in a pairwise BLAST search
are needed to correctly infer SCL for 90% of the query proteins.
On the other hand, the authors of the CELLO method for both
eukaryotes and bacteria [13] found that SCL prediction by a sim-
ple BLAST search was better than a machine learning method
above a pairwise identity cutoff as low as 30%.
The simplest possible homology-based prediction is the direct
transfer of annotation from the best BLAST hit, i.e., the query pro-
tein is used to search a database of proteins with experimentally
known SCLs, and then the SCL of the best hit is assigned to the
query. However, more advanced approaches to homology-based
prediction are also possible, using indirect means to infer SCL from
the annotation of homologs that do not necessarily have experi-
mentally known SCLs. This annotation could be derived from key-
words or functional descriptions [14, 15], titles or abstracts of
literature references [16, 17], or Gene Ontology terms [18–24].
In this context, it should be mentioned that many signal-based
and global-property-based methods use a BLAST search to build a
profile of related sequences to enhance the prediction. This does
not make these methods homology-based since they do not use
the annotations of the found hits.
In addition to the three approaches described here, it is of
course possible to construct hybrids of them. Most of the multicat-
egory methods described in Subheading 8 are of the hybrid type.
30 Henrik Nielsen

When comparing methods based on sorting signals, global

properties, or homology, it is important to realize that each
approach has its strengths and weaknesses. Homology-based meth-
ods, or hybrid methods containing homology-based components,
often present the best measured performances; but the perfor-
mance depends critically on the source of the query protein.
Organisms that have been subjected to intense research will natu-
rally tend to have more high-quality annotations, so proteins from
those and their close relatives will find more close homologs with
richer annotations from which to make predictions, while predic-
tions for less-well-studied organisms will suffer from a lack of
annotations of close homologs. This is typically not taken into con-
sideration when reporting on the predictive performances of such
methods. Signal-based and global property methods should be
expected to be less sensitive to the source of the query protein,
unless the signals in the training data are very organism specific.
There are two advantages to using global-property-based or
homology-based methods. First, they can be used also for those
compartments where the actual sorting signals are not known, or
are too poorly characterized to support a prediction method.
Second, they may work for sequences that are fragments from
which the actual sorting signal may be missing or for amino acid
sequences derived from genomic or metagenomic sequence where
the start codon of the protein has not been correctly predicted,
thus obscuring any N-terminal sorting signals. On the downside,
global-property-based or homology-based methods do not pro-
vide the same degree of insight into the information processing in
the cell since they ignore which parts of the sequence are actually
important for sorting. Another drawback is that such methods will
not be able to distinguish between very closely related proteins
that differ in the presence or absence of a sorting signal, and they
will not be able to predict the effects of small mutations that
destroy or create a sorting signal.

3  Algorithms for Prediction

A rich variety of computational algorithms have been used in the

prediction of SCL from amino acid sequences. Common to all of
them is that they take a number of sequence-derived inputs and
produce an output that can be the presence or absence of a sorting
signal (for signal-based predictors) or an assignment of the protein
into one of a number of possible SCL classes (for multicategory
predictors). For Gram-negative bacteria, the number of SCL classes
is most often defined as five (cytoplasm, inner membrane, peri-
plasm, outer membrane, and extracellular), while for Gram-positive
bacteria, the corresponding number is four (cytoplasm, membrane,
cell wall, and extracellular).
 Protein Sorting Prediction 31

Some algorithms, such as sequence alignment and hidden

Markov models (HMMs), are naturally designed to work with
sequences, while others, such as ANNs and SVMs, take only a fixed
number of input values. When working with the latter category,
one can either input the sequence as a series of overlapping win-
dows of fixed length (typical for signal-based predictors) or extract
a fixed number of features from the sequence (typical for global
property predictors).
Numerical prediction algorithms can roughly be divided into
two groups, statistical and machine learning, although it can some-
times be a matter of definition where to draw the distinction. Both
classes of methods have a number of free parameters that must be
estimated from the data, but while the parameters in statistical
methods can be calculated directly, machine learning methods
depend on an iterative optimization process where parameters are
gradually changed until the classification error has reached a
minimum.
The simplest sequence pattern recognition method is the con-
sensus sequence or regular expression, for example, “RR.[FGAVML]
[LITMVF]” for SPs following the twin-arginine translocation
(Tat) pathway. It is interpreted like this: there should be two con-
secutive arginines, followed by any amino acid, then one amino
acid from the FGAVML group, and then one from the LITMVF
group. It is easy to check whether such a pattern is present in a
sequence, but it is also a very crude method because it defines
absolute requirements for certain amino acids at certain positions
and only provides yes or no answers. The preceding pattern, for
example, ignores the fact that not all amino acids from the
FGAVML and LITMVF groups are equally probable at positions
4 and 5 (cf. heights of individual letters in positions 17 and 18 in
Fig. 1).
An alternative to the consensus sequence or regular expression is
the position-weight matrix (PWM) [25], a statistical window-­based

Fig. 1 Sequence logo of Tat (twin-arginine translocation) signal peptides from both Gram-positive and Gram-­
negative bacteria, aligned to the PROSITE profile PS51318/TAT. The height of each stack of letters corresponds
to the information (conservation) at that position, while the height of each individual letter is proportional to the
fraction of that amino acid at that position. Note that individual sequences may be shorter or longer than 40
amino acids; in the logo, they have been stretched or shortened to fit the model. Picture from PROSITE [28]
made with WebLogo [146]
32 Henrik Nielsen

method that is very useful for characterizing and predicting short

sequence motifs. The procedure for constructing a PWM involves
using a set of examples of the motif of interest (the training set) to
estimate the probability of each amino acid at each position and then
convert those probabilities into weights. The score for a new
sequence window can then be calculated by looking up the weights
for each amino acid in each position in the window and adding them
up. In this way, the weight matrix can give a quantitative answer to
how well a sequence window fits the pattern.
A graphical counterpart to the PWM is the sequence logo [26],
where each position is summarized by a stack of letters. The height
of each stack is equivalent to the information content—a measure
of the conservation—of that position, while the height of each
letter is proportional to the probability of the corresponding
amino acid at that position. Figure 1 presents an example of a
sequence logo.
A straightforward extension of the PWM is the sequence profile,
which allows for insertions and deletions in the sequence and
therefore can model motifs of variable length. It is possible to for-
mulate a profile in probabilistic terms, in which case it becomes a
profile HMM [27]. An HMM is a machine learning algorithm,
since the probabilities it contains are found by an iterative optimi-
zation process from a set of training data. After training, new
sequences can be evaluated in terms of the probability that the
sequence was generated by the model (a process known as HMM
decoding). It should be emphasized that not all HMMs are profile
HMMs—any grammar that can be described as a diagram of con-
nected states can be modeled as an HMM. For example, a cyclic
HMM can describe a repeating pattern, and a branched HMM can
describe a choice between alternative patterns. Examples of cyclic
HMMs are given in Subheading 7.
Several publicly available databases specialize in creating and
storing profiles for protein families or domains. These include
PROSITE [28] (see Table 1), which contains both regular expres-
sion patterns and PWM-like profiles, and Pfam [29] (see Table 1)
and TIGRFAMs [30] (see Table 1), which are both databases of
profile HMMs. InterPro [31] (see Table 1) is a special case since it
does not create its own profiles but rather collects profiles from a
number of contributing databases, including PROSITE, Pfam, and
TIGRFAMs. Most profiles in these databases are evolutionarily
related families or domains, but there are also instances of func-
tional motifs that are similar owing to common selection pressure
rather than common descent. Among these are a few protein sort-
ing motifs, which can be used as prediction tools; examples will be
given in Subheading 5.
Among the methods that use a fixed set of numbers as input,
the simplest possible one is the Naïve Bayes classifier, which assumes
that all the input variables are independent. It can show surprisingly
 Protein Sorting Prediction 33

good performance also in cases where the assumption of indepen-

dence is known to be violated [32], and it is sometimes preferred
over more advanced machine learning methods because it offers
the opportunity to explain exactly which input variables were
important for each prediction [33, 34]. Two very popular machine
learning algorithms that are widely used in biological sequence
analysis are the ANN [35] and the SVM [36]. This is not the
place to go into technical details about the details of ANNs and
SVMs, but they both have large numbers of parameters that must
be estimated via a set of training data, and they are both potentially
able to model situations where there are correlations between
input features.

4  Performance of Prediction Methods

After having trained a statistical or machine learning method, it is

crucial to test its predictive performance on another data set. This is
a very important point: it is not enough that a trained method can
reproduce its input examples exactly—in fact, it is not even interest-
ing, since a database can do the same. What is interesting is whether
a model can generalize from the examples in the training set and
produce useful output for sequences it has not “seen” before.
There is often a certain degree of trade-off between training
set and test set performance. If a model reproduces its training
examples in too much detail, it uses its parameters to fit not only
the common pattern in the data but also the individual noise in
each data point. When this happens, the performance on the test
set declines, and the model is said to be overfitted—colloquially
speaking, it cannot see the forest for the trees.
Avoiding overfitting can be tricky; it may involve limiting the
number of free parameters in the model, adding some regularizing
terms to the parameters, or—especially in the case of ANNs—ter-
minating the training early. In some cases, this is done using the
test set performance as a criterion for choosing the optimal num-
ber of free parameters or the best point to stop the training; but in
fact this is cheating, since the test set in such a procedure has been
part of the training process. Instead, three data sets should be used:
a training set, a validation set for optimizing the model architec-
ture and training process, and a true test set (also known as evalu-
ation set) for measuring the performance.
Instead of using a fixed part of the data as the test set, perfor-
mance evaluation is often done by cross-validation, where the
data set is divided into a number of folds, and each fold is in turn
used as a test set while the others are used as the training set. The
final performance is then calculated as an average of the test set
performances. The number of folds can vary; most often, five- or
tenfold cross-validation is used, but some authors prefer N-fold
34 Henrik Nielsen

cross-­validation, where N is the number of data points—in other

words, just one example at a time is held out, while the training
is performed on all other examples. This is also known as leave-
one-out cross-validation or jackknife test.
The necessity for splitting the data into training and test is not
particular to bioinformatics; it applies to all prediction tasks.
However, bioinformatics has an added complication: sequences
are related by descent. If there are sequences in the test set that are
closely related to sequences in the training set, the measured per-
formance is arguably not a true test performance. This can be
taken into account by reducing homology in the data set before
splitting it into folds (homology reduction) or by ensuring that no
pairs of sequences that are too closely related end up in different
folds (homology partitioning). Two widely used algorithms for
homology reduction were published early in the history of bioin-
formatics [37].
There are diverging views concerning exactly how closely
related two sequences should be allowed to be in order to be
separated into different folds. Some authors arbitrarily set a rather
high cutoff, for example, 80% or 90% identity in a pairwise align-
ment [8, 38]. One approach to a nonarbitrary definition is to
identify a cutoff above which the problem could be better solved
by alignment than by machine learning [39, 40]. Another
approach is to use a cutoff in the alignment score above which
there is a statistical significance of homology [41]. These
approaches tend to result in much lower cutoff values, typically
corresponding to ≈25% identity in long alignments [39]. When
comparing reported performances of different methods, it is
important to take into account which type of homology reduc-
tion or partitioning was used (if any).
When reporting performances of prediction methods, a variety
of measures may be used, potentially confusing the untrained
reader. The conceptually simplest performance measure, the frac-
tion or percentage of correct answers (also known as accuracy), can
be misleading if the classes are not the same size. As an example,
consider a data set with 99 negative examples for each positive
example. If a prediction method consistently returns an answer of
no, it will be correct 99% of the time, even though the supposed
prediction is completely noninformative. Instead, a number of
alternative measures are often used. When discriminating between
two classes, the most important performance measures can be
defined in terms of the numbers of true positives (TPs), true nega-
tives (TNs), false positives (FPs) (type I errors or overpredictions),
and false negatives (FNs) (type II errors or misses):
●● Sensitivity (also known as recall or the TP rate—how many
positive examples are found?):
TP
Sn = ;
TP + FN
 Protein Sorting Prediction 35

●● Specificity (also known as the TN rate—how many negative

examples are found?):
TN
Sp = ;
TN + FP
●● Precision (also known as positive predictive value—how many
positive predictions are true?):
TP
Pr = ;
TP + FP
●● Matthews correlation coefficient (MCC)—a measure that takes
values between −1 and 1, where 1 is a perfect prediction, 0 is
a random guess or noninformative prediction, and −1 is a pre-
diction that is consistently wrong:
TP ´ TN - FP ´ FN
MCC = .
(TP + FP ) (TP + FN ) (TN + FP ) (TN + FN )
It should be mentioned that the term specificity is not unequiv-
ocal; it has sometimes been used to denote what is here referred to
as precision (e.g., in ref. 42).
Whenever a prediction method gives a quantitative output,
there is a trade-off between sensitivity and specificity, controlled
by the threshold (also known as cutoff) above which a predic-
tion is considered positive. Lowering the threshold reduces the
number of FNs, thereby increasing the sensitivity, but it also
increases the number of FPs, thereby reducing the specificity
(and the precision). It is possible to plot the sensitivity as a
function of the FP rate (1 minus the specificity) for varying
threshold values; such a plot is known as a receiver operating
characteristic (ROC) curve (see Fig. 2). The area under the
ROC curve (usually referred to as AUC or AROC) can be used
as a threshold-independent performance measure; it will be 1
for a perfect prediction, 0.5 for random guesses, and 0 for a
consistently wrong prediction.
When predicting more than two classes, for example, a num-
ber of SCLs, the maximum information about the prediction is
provided by the so-called confusion matrix: a table showing, for
each observed class, how many examples were predicted to
belong to each class. This can be used to see not only how well
each class was predicted, but also which classes were particularly
difficult to ­distinguish. From the confusion matrix, the sensitivity,
specificity, precision, and MCC can be calculated for each class.
There are also measures that summarize a whole confusion matrix
in one number, such as the Gorodkin correlation coefficient,
which is a generalization of the MCC to more than two classes,
or the normalized mutual information coefficient [43, 44]. In
practice, these are rarely calculated, and the percentage of correct
answers is often used instead, despite the shortcomings of this
measure.
36 Henrik Nielsen

Gram-negative data
1
0.98
0.96
0.94
0.92 SignalP 3.0 default

sensitivity
0.9 SignalP 4.0 default
0.88
0.86
0.84 SignalP 3.0 no TM
SignalP 3.0 all data
0.82 SignalP 4.0 no TM
SignalP 4.0 all data
0.8
0 0.05 0.1 0.15 0.2
FP rate

Fig. 2 ROC curve showing performance of SignalP versions 3 and 4 as sensitivity

versus FP rate. “No TM” means performance when the negative set did not con-
tain transmembrane segments; “all data” means that sequences with trans-
membrane segments were included in the negative data. Observe that although
SignalP 4 by default has a lower sensitivity than SignalP 3, this is only a question
of cutoff; the curve for SignalP 4 when using “all data” is consistently closer to
the upper left corner, showing that SignalP 4 is a better method. Note that the
sensitivity and FP rate values depicted here are not cross-validation perfor-
mances but measured by applying the finished method to the whole data set

5  Recognition of Signal Peptides

The secretory SP is among the earliest prediction targets for bio-

informatics algorithms. The oldest SP prediction methods used a
simple PWM for the SP cleavage site, first with a reduced alphabet
[2] and later with weights for all amino acids [45]. Another very
early SP prediction method used two simple sequence-derived fea-
tures, peak hydrophobicity and length of the uncharged region, to
discriminate SPs, but it did not predict the cleavage site [46].
SPs are present in all domains of life, but it was early discovered
that there are differences between broadly defined systematic
groups [47]. SPs of Gram-positive bacteria are longer than those
of Gram-negative bacteria, which in turn are longer than those of
eukaryotes. A sequence logo of SPs from Gram-negative bacteria is
shown in Fig. 3.
In 1997, the SP predictor SignalP (see Table 1) was among the
first to use ANNs to predict sorting signals [48]. Later, in versions
2 and 3, an HMM was added to the method [49, 50], while ver-
sion 4 is again purely ANN-based [51]. SignalP is among the most
cited prediction servers in bioinformatics, and it has performed
well in comparative studies [52–55].
 Protein Sorting Prediction 37

Fig. 3 Sequence logo of signal peptides from Gram-negative bacteria, aligned after their cleavage site (between
positions −1 and 0). The visible features are the cleavage site specifying residues in −3 and −1 (strong prefer-
ence for alanine), the hydrophobic region that approximately stretches from −16 to −7, and a preference for
the positively charged lysine in the N-terminal region. Note that no stretching or shortening of the sequences
has been performed; they have simply been aligned by the cleavage site; this is why no completely conserved
methionine is seen on the left-hand side. Picture made with WebLogo [146]

Other SP prediction methods worth mentioning are the PWM-­

based PrediSi [56] (see Table 1) and SOSUIsignal (see Table 1),
which is based on amino acid propensities in regions [57]. There is
also Signal-BLAST [58] (see Table 1), which, quite unusually for a
sorting signal prediction method, is homology-based. It uses
BLAST [10] with some customized settings to search a reference
set of SPs and non-SPs and returns the class of the best hit as its
prediction. In addition, some TMH prediction programs also offer
SP prediction; see Subheading 7 for details.
The performance of SP prediction in Gram-negative bacteria is
fairly high, with SignalP 4.0 reporting an MCC of 0.85 in
­distinguishing between SPs and non-SPs and a cleavage site preci-
sion of 71%. Note that these performances are cross-validation per-
formances on a strictly homology-reduced data set, so they reflect
the performance you would expect if you submitted sequences that
were completely unrelated to any in the SignalP 4.0 data set. The
performance measured by applying the finished method (where
the outputs of the different data set partitions are averaged) to the
whole data set is considerably higher, with an MCC of 0.96. Note
that SignalP 4.0, despite a higher MCC, has a lower sensitivity than
SignalP 3.0; the cutoff had simply been placed at a higher value in
order to maximize the MCC. In the slightly modified SignalP 4.1,
there is an option to select a cutoff that reproduces the sensitivity
of SignalP 3.0. Of course, this comes at a price of a higher FP rate,
38 Henrik Nielsen

but it is still lower than that of SignalP 3.0, as can be seen from the
ROC curves in Fig. 2.
It should be stressed that the presence of an SP does not neces-
sarily mean that the protein is secreted. First, it may be periplasmic
or integrated into the outer membrane; second, there may be
downstream TMHs keeping the protein integrated in the cytoplas-
mic membrane. It has been reported that cleavable SPs are rarely
found in bacterial cytoplasmic membrane proteins [59], but a
quick search in UniProt [11] revealed that they are not that rare
after all, so a prediction of SPs should always be combined with a
search for TMHs (see Subheading 7) before drawing conclusions
about the SCL.
SignalP and the other SP predictors mentioned so far only pre-
dict classical SPs, translocated by the Sec system and cleaved by
type I signal peptidases. For lipoproteins cleaved by lipoprotein
signal peptidase, there are other prediction methods. LipoP [60]
(see Table 1) is an HMM-based method (although an ANN was
also trained during the development of the method). Even though
LipoP has been trained on sequences from Gram-negative bacteria
only, both the original paper and a later study [61] reported that it
demonstrated good performance on sequences from Gram-positive
bacteria also. Other methods include the ANN-based SPEPlip
[62] (see Table 1), which has separate options for Gram-negative
and Gram-positive bacteria, and the HMM-based PRED-LIPO
[63] (see Table 1), which is specific to Gram-positive bacteria. In
addition to these methods, there is also a profile in PROSITE [28],
dedicated to lipoproteins from both Gram-negative and Gram-­
positive bacteria, called PROKAR_LIPOPROTEIN (see Table 1).
A sequence logo of lipoprotein SPs aligned to this model is shown
in Fig. 4. It should be noted that LipoP, SPEPlip, and PRED-­
LIPO are all able to predict classical SPs as well, differentiating
between the two types of SP. Additionally, LipoP differentiates
between SPs and N-terminal TMHs.

Fig. 4 Sequence logo of lipoprotein signal peptides from both Gram-positive and Gram-negative bacteria,
aligned to the PROSITE profile PS51257/PROKAR_LIPOPROTEIN. Lipid attachment occurs at the completely
conserved cysteine in position 35. Note that individual sequences may be shorter or longer than 35 amino
acids; in the logo, they have been stretched or shortened to fit the model. Picture from PROSITE [28] made with
WebLogo [146]
 Protein Sorting Prediction 39

For SPs translocated by the Tat pathway, a few dedicated

prediction methods are also available. In addition to the twin-argi-
nine motif in the N-terminal region that gave them their name,
they also differ from Sec SPs by being on average longer and less
hydrophobic [64]. The available servers are TatFind [65] (see
Table 1), which is based on a regular expression combined with a
set of simple rules concerning hydrophobicity and charge, TatP [66]
(see Table 1), which is based on a regular expression combined
with two ANNs, and the newer HMM-based PRED-TAT [67] (see
Table 1). In addition, three motifs are available in the family and
domain databases: the PROSITE profile TAT (see Table 1), the
Pfam profile TAT_signal (see Table 1), and the TIGRFAMs profile
TAT_signal_seq (see Table 1). A logo of sequences aligned to the
PROSITE profile is shown in Fig. 1. Note that none of these
methods makes any distinction between Gram-positive and Gram-­
negative bacteria—if the Tat SPs indeed differ between the two
bacterial groups, there should be room for improvement in the
prediction.

6  Prediction of Secretion Without Signal Peptides

In Gram-negative bacteria, secretion without an N-terminal,

cleaved SP happens to proteins belonging to secretion systems of
types I, III, IV, and VI [68, 69]. For Gram-positive bacteria, the
phenomenon appears to be less important, but there are some
examples of proteins exported via, for example, the Wss, holin, and
SecA2 pathways [69, 70]. This has sometimes been referred to as
nonclassical secretion [70–72], but at least in Gram-negative bacte-
ria, this term should properly be reserved for secretion that hap-
pens independently of any of the numbered secretion systems, for
example, by membrane vesicles [73].
SecretomeP 2.0 [70] (see Table 1) is a general secretion predic-
tor from 2005 designed to handle secretion without SPs. At the
time, it was not easy to locate experimentally confirmed examples
of this, so the SecretomeP authors took a different approach, based
on the idea that secreted proteins must be expected to share cer-
tain features independent of the pathway used to secrete them.
Thus, the positive training data set simply consisted of classically
secreted proteins with the SP removed. A large number of struc-
tural and functional features calculated from the amino acid
sequence were then tested for their predictive power for SCL, and
the most promising features were used to train an ANN. For Gram-­
negative bacteria, only four features were selected: amino acid
composition, arginine content, instability index, and predicted dis-
order. These features led to a sensitivity of 88% (on the truncated
examples) and a specificity of 96%. Unfortunately, the ability of
SecretomeP to predict type I, III, IV, and VI secretion has never
40 Henrik Nielsen

been tested. It is possible that effectors of types III, IV, and VI

secretion systems, which get injected directly into the host cyto-
plasm, do not share the extracellular characteristics that SecretomeP
is trained to recognize.
The competing method SecretP 2.0 [71] (see Table 1), which
is also based on feature selection, just using SVMs instead of ANNs,
used another, more problematic approach to the problem of data
set generation: the positive training set for so-called nonclassical
secretion consisted of proteins that were annotated for secretion
but had no annotated SP in UniProt [11]. The problematic aspect
of this is that a lack of annotated SP may simply reflect an incom-
plete annotation rather than a real absence of SP, meaning that
SecretP in fact could be based on classical SP-containing proteins.
This suspicion is confirmed by the fact that SignalP was found to
predict seven out of nine supposedly nonclassically secreted pro-
teins in Gram-negative bacteria [71].
Despite its name, SecretP 2.1 [72] (see Table 1) is not an
updated version of SecretP 2.0; it attempts to answer a completely
different question: Given that a protein from a Gram-negative bac-
terium is secreted, it yields a prediction about which secretion sys-
tem is responsible for it. SecretP 2.1 distinguishes between proteins
of the type I, II, III, IV, V, or VII secretion system—type VI and
type VIII data were also collected during data set construction, but
too few were found for prediction. Before the actual prediction
method was developed, the authors constructed sequence similar-
ity networks for each of the six groups and found that proteins of
the type I, V, and VII secretion systems formed a few large clusters
with a high average number of links, while the others formed many
small clusters or singletons. This can be taken to mean that
homology-­based prediction would be well suited for predicting
type I, V, and VII secretion. However, the authors chose to make
a global-property-based method, based on amino acid composi-
tion and autocorrelation of physicochemical parameters as inputs
to a system of SVMs, to make the multicategory classification. The
reported overall accuracy is approximately 90%. Note that the
server is not suited for genomic-scale analysis; it only accepts one
sequence at a time.
When it comes to secretion-system-specific predictors, very
little attention has been directed toward type I and VI secretion.
One method for machine-learning-based prediction of type I secre-
tion has been published [74], but it is not available as a web server,
and furthermore the prediction is limited to those type I secreted
proteins that contain RTX repeats. I am not aware of any attempt
to specifically predict type VI secretion; maybe the number of
known examples is too small to develop a prediction method.
Type IV secretion has been somewhat better investigated, with
several bioinformatics analyses published specifically concerning
the phenomenon in two species, Legionella pneumophila and
 Protein Sorting Prediction 41

Coxiella burnetii [75–77]. Two SVM-based machine learning

methods with a broader scope are available as web servers,
T4EffPred [78] (see Table 1) and T4SEpre [79] (see Table 1).
T4EffPred is a global-property-based predictor that uses entire
sequences as input, while T4SEpre uses the C-terminal 100 amino
acids (shorter window lengths were also tested but did not per-
form as well). Within this window, T4SEpre uses a PWM-like
method for encoding sequences, together with predicted second-
ary structures and solvent accessibility. Authors of both methods
reported the performance on type IVa and type IVb effectors sepa-
rately, but the T4SEpre authors additionally reported that the
method trained on IVa and tested on IVb or vice versa did work to
some degree, showing that the signals for the two pathways are not
completely different.
The prediction of type III effectors, on the other hand, has
received a lot of attention in recent years, and the rest of this sec-
tion will be dedicated to type III secretion. Several experimental
results point to the signal for secretion being within the N-terminal
part of the protein [80], but it is not clear whether the signal is
read at the mRNA level or the protein level. The mRNA hypoth-
esis is predominantly based on the observation that some proteins
retained their ability to be secreted after two balanced frameshift
mutations completely changed the amino acid sequence of the
N-terminal part of the protein [81].
Two methods published in the same journal issue in 2009
introduced the use of machine learning for type III effector predic-
tion: the SVM-based SIEVE [82] (see Table 1) and the Naïve
Bayes-based EffectiveT3 [83] (see Table 1). The EffectiveT3
authors tested several machine learning methods, including SVM,
but found that Naïve Bayes gave the best performance. In both of
these methods, a feature selection procedure was carried out, but
the feature sets were quite different—SIEVE used evolutionary
conservation, phylogenetic profiles, G + C content of the encoding
gene, and the sequence of the N-terminal part of the protein, while
EffectiveT3 used amino acid composition and occurrence of short
degenerate motifs such as “polar–hydrophobic–polar.” Later the
same year, an unnamed method was published by Löwer and
Schneider [84] (see Table 1) using a moving window input, which
was processed by an ANN or an SVM. The website gives the user
the opportunity to choose between ANN and SVM, but the
authors estimated that the ANN had the best generalization
capability.
Four newer methods have implemented web servers. These
are BPBAac [85] (see Table 1), which uses a PWM-like approach,
T3_MM [86] (see Table 1), which is based on first-order Markov
models, and the hybrid methods BEAN [87, 88] and pEffect [89]
(see Table 1), both of which use homology to known effectors to
enhance prediction.
42 Henrik Nielsen

Performances are difficult to compare, especially since the

authors disagree on which negative set to use. SIEVE used all pro-
teins in each represented organism that were not in the positive set,
yielding a ratio of positive to negative examples of approximately
1:120, while the authors of EffectiveT3, BPBAac, T3_MM, and
BEAN constructed balanced training sets where the ratio was 1:2
by sampling annotated proteins without annotation about type III
secretion. Löwer and Schneider used an even more balanced set,
where the ratio was approximately 1:1. While the SIEVE approach
will tend to underestimate performance because there could be
undiscovered positive examples in the negative data, the balanced-­
data-­set approach measures performance in an unrealistic setting.
Furthermore, as the SIEVE authors correctly remark, “any method
of filtering negative examples to provide a more confidently nonse-
creted set might introduce significant biases in the data set that
could render classification trivial” [80]. The pEffect authors took a
radically different approach and included eukaryotic as well as bac-
terial data in their negative set, which makes sense if you imagine
the method used in a metagenomics setting. The resulting ratio of
positive to negative examples was 1:30 after homology reduction.
One caveat about measuring performance is illustrated by the
BPBAac and T3_MM methods, which originate from the same
group. Although they both reported reasonably high performances
(sensitivity 91% and 90%, specificity 97% and 91%), they disagreed
wildly about predictions in genomes of Salmonella strains that had
not been part of the training set: Only around 25% of the predic-
tions by the more conservative method, BPBAac, were shared by
T3_MM [86].
The pEffect authors reported a sensitivity of 95% at a precision
of 87%, which is better than EffectiveT3, BPBAac, T3_MM, and
BEAN 2.0 on a data set that was not used to train any of the
methods.
An interesting aspect of these methods—perhaps more inter-
esting than their predictive performances—is what they can tell us
about the actual signal for type III secretion. The first lesson is that
the signal is apparently universal across different systematic groups
within Gram-negative bacteria. Several of the authors of these
methods tested this by a phylogenetically informed cross-­validation,
i.e., testing the performance on one systematic group of bacteria
with a method trained on the others [82, 83, 85, 86].
The fact that it is possible to predict type III secretion to some
degree from amino acids speaks against the mRNA hypothesis,
although nobody actually tested the hypothesis by training a
nucleotide-­ sequence-based predictor and comparing its perfor-
mance to an amino-acid-sequence-based predictor. As mentioned
earlier, the first version of SIEVE used G + C content as an input
feature, but the authors later found that this feature could be
 Protein Sorting Prediction 43

removed without a major impact on performance, and the current

version of SIEVE uses only amino acid sequences as input [80].
An alternative explanation for the observations that led to the
mRNA hypothesis could be that the signal is so unspecific that a
spurious reading frame has a relatively high chance of producing a
type III secretion signal. This was verified by the authors of
EffectiveT3 and BPBAac, who tested their predictors on artificially
created frameshifted sequences and found that 10–14% of frame-
shifted positive examples were predicted positive [83, 85].
Since the signal is so weakly defined, it is reasonable to ask
whether sequence order of amino acids plays any role or whether it
is just the composition of the N-terminal region. SIEVE and
BPBAac both use the actual sequence as input, while EffectiveT3,
T3_MM, and BEAN use the composition of amino acids or amino
acid pairs. The better performance of BPBAac over T3_MM using
the same data set could be taken as a sign that the positions of
amino acids do have a role to play.
Another question is whether the signal is actually N-terminal,
as many of the methods assume. The EffectiveT3 authors tested
this by training on the 15 C-terminal residues instead of N-terminal
and found no performance above the random expectation [83].
The EffectiveT3 and SIEVE authors and Löwer and Schneider all
investigated how many N-terminal residues were required for pre-
diction and found no improvement by moving beyond 30 (except
in the case of plant pathogens predicted by EffectiveT3, where the
limit seemed to be 50) [82–84]. In contrast, the BPBAac authors
used 100 N-terminal positions and found that to be superior to 50
[85]. BEAN version 1 used 51 N-terminal positions, but in BEAN
2.0 two additional windows were added comprising positions
52–121 from the N-terminus and 1–50 from the C-terminus [87,
88]. The authors state that their results show that C-terminal sig-
nals sometimes play a role, but since the two windows were added
simultaneously, it might as well be the 52–121 window that made
the difference (which would be in accordance with the BPBAac
result). The pEffect method, by contrast, uses the entire sequence
as input, and the authors write: “Our work reveals that signals for
recognition and transport of effectors are distributed over the
entire protein sequence instead of being confined to the
N-terminus” [89]. However, the evidence for this conclusion is
not very strong, since it is no surprise that homology between
known effectors is distributed across the entire sequence. To draw
conclusions on the placement of the signal, it is necessary to con-
sider the nonhomology part of pEffect, the so-called de novo pre-
diction, and that declined when fragments were used instead of the
whole sequence. To obtain a real answer to the question, it would
be necessary to train an N-terminal version of the de novo part of
pEffect to see whether it performed equally well.
44 Henrik Nielsen

7  Prediction of Transmembrane Topology

In Gram-negative bacteria, transmembrane proteins come in

two flavors: α-helix proteins, almost exclusively in the cytoplasmic
membrane, and β-barrel proteins, exclusively in the outer mem-
brane. The prediction of TMHs has a long history in bioinformat-
ics. Initially, the basis for the prediction was simply a plot of the
hydrophobicity, averaged in a sliding window over the sequence
[1, 90]. A slightly more advanced approach was represented by
TOP-PRED in 1992 [91], which combined hydrophobicity analy-
sis with a count of the number of positively charged residues in
each loop to choose the topological model that best conformed to
the “positive-inside rule” [92].
The prediction of transmembrane β-strands is more compli-
cated, not only because of the smaller number of known examples,
but also because of their shorter length and lower hydrophobicity.
Typically, only those amino acid side chains facing the lipid phase
are hydrophobic, while those facing the pore of the β-barrel tend
to be polar. The first attempt at a transmembrane β-barrel (TMBB)
topology prediction method, published in 1985 [93], focused not
on hydrophobicity patterns but instead on predicting the turns
separating the strands. However, the following year a method for
structural prediction that accounted for amphipathic β-strands was
published [94].
Later, machine learning methods were used to predict mem-
brane protein topology, i.e., which parts of a sequence are inside,
transmembrane, and outside. In particular, the HMM technology
has been popular in this area because it makes it possible to model
the so-called grammar of a problem: if a TMH or a strand follows
an inside loop, it must be followed by an outside loop, and vice
versa. This is typically modeled by a cyclic HMM, having submod-
els for helices/strands, inside loops, and outside loops. In the rest
of this section, TMH prediction will be described first, followed by
TMBB prediction.
The best known HMM for TMH prediction is TMHMM [95]
(see Table 1), but also HMMTOP [96] (see Table 1) has enjoyed
widespread usage. A comparative analysis in 2001 found TMHMM
to be the best performing TMH predictor [97]. Newer surveys
covering more recently published predictors unfortunately do not
provide quantitative performance comparisons [98–100].
Since hydrophobicity is a feature of both SPs and TMHs, the
two are easily confused by prediction methods. TMHMM often
falsely predicts an SP as a TMH, and versions 1–3 of SignalP would
often predict a TMH close to the N-terminus as an SP. Newer topol-
ogy prediction methods, such as the HMM-based Phobius [101],
Philius, which is based on Dynamic Bayesian Networks [102], the
ANN-based MEMSAT3 [103], the SVM-based MEMSAT-SVM
 Protein Sorting Prediction 45

[104], and the ANN + HMM-based SPOCTOPUS [105] (see

Table  1), deal with this problem by modeling both these signals.
However, the paper on version 4 of SignalP [51] reports a better
discrimination between SPs and TMHs than all these methods, and
it is worth noting that the performance difference is larger for bacte-
rial sequences than for eukaryotic sequences. This probably reflects
the fact that the signal peptide models in Phobius, Philius, MEMSAT,
and SPOCTOPUS are not divided into organism types, which
causes the results to be biased toward the organism group with the
most data (eukaryotes).
Another confounding factor is the fact that multispanning
membrane proteins sometimes have so-called reentrant loops—
segments of the sequence that dip into the membrane but do not
span it, leaving the membrane on the same side at which they
entered. Reentrant loops are not very frequent; only 137 examples
from bacteria are currently reported in UniProt, one of them with
experimental evidence. OCTOPUS [106] (see Table 1) and
SPOCTOPUS make an attempt at predicting reentrant loops.
The use of profiles of homologous sequences generated by
BLAST or PSI-BLAST (see Subheadings 2 and 3) in the training
and prediction of TMH recognition methods has been shown to
enhance predictive performance by approximately ten percentage
units [107]. Methods that use profiles include PRODIV-
TMHMM [107], PolyPhobius [108], MEMSAT3, MEMSAT-
SVM, OCTOPUS, and SPOCTOPUS.
An interesting alternative method is SCAMPI [109] (see
Table 1) that does not use machine learning or statistics on a train-
ing set to calculate its parameters; instead, the parameters are based
on a series of experiments where all 20 possible amino acids have
been inserted at various positions into a model TMH [110]. These
experiments have been used to calculate an apparent free energy
contribution, ΔGapp, that is used as an analog to a hydrophobicity
scale. The overall ΔGapp for each sequence window is calculated
and used as input to an HMM-like model with only two free
parameters to be estimated from the training data. The SCAMPI
authors reported a performance comparable to the best machine
learning methods.
Consensus methods for TMH prediction have been shown to
perform better than any of the constituent methods. An early effort
in this direction was BPROMPT from 2003 [111] (see Table 1).
The newer server TOPCONS [112, 113] (see Table 1) offers a
consensus prediction of both TMHs and SPs based on OCTOPUS,
SPOCTOPUS, PolyPhobius, Philius, and SCAMPI. TOPCONS
reports 83% correctly predicted topologies on a benchmark set.
The downside of TOPCONS is the running time, increased by the
fact that four of the five predictors are based on profiles that first
must be constructed from a database search. An alternative con-
sensus server, based only on methods that do not require profiles,
46 Henrik Nielsen

is TOPCONS-single [114] (see Table 1), which does approximately

six percentage units worse than TOPCONS, but 70 times faster.
In TMBB topology prediction, the first machine learning
method was an ANN published in 1998 [115], but the early
2000s saw a surge in the publication of HMMs, including HMM-
B2TMR [116], PRED-TMBB [117, 118] (see Table 1), and
ProfTMB [119, 120] (see Table 1). A comparative study in 2005
[121] found that HMM-based predictors performed better than
ANN- and SVM-based predictors but achieved the best perfor-
mance by constructing a consensus of HMM-B2TMR, PRED-
TMBB, and ProfTMB, together with the ANN-based methods
B2TMPRED [122] (see Table 1) and TBBpred [123] (see Table 1).
The resulting method, ConBBPRED (see Table 1), is available as
a server; however, it is very impractical to use since it does not
launch the constituent methods but requires the user to obtain
the various predictions and input those to the server manually.
BOCTOPUS [124, 125] (see Table 1) is a newer hybrid
method, where four window-based SVMs calculate scores for each
residue in the inner loop, outer loop, pore-facing transmembrane,
and lipid-facing transmembrane, respectively. These scores are fed
to a filter that delineates the barrel domain (if any) and then to an
HMM that calculates the final topology.
Some other methods have focused on the task of discriminat-
ing between outer membrane TMBB proteins and other proteins
rather than predicting the correct topology, and here, the cyclic
HMM is not necessarily the best approach. BOMP [126] (see
Table  1) from 2004 is one example that does not use machine
learning but simple statistical measures, including a C-terminal
pattern (regular expression) that is found in many outer membrane
proteins. HHomp [127] (see Table 1) uses a collection of profile
HMMs, built from outer membrane proteins of known structure,
to search for matches to the query sequence. It is thus based on the
idea that most TMBB proteins are evolutionarily related, and for
predicted TMBB proteins, it additionally provides a classification
into a number of functional subgroups. BetAware [128, 129] (see
Table 1) uses an ANN to scan entire sequences and predict whether
they are TMBB proteins. The second version of BetAware includes
a probabilistic model, a so-called Grammatical-Restrained Hidden
Conditional Random Field, for predicting the topology, but it is
only invoked when the ANN output indicates that the query is a
TMBB protein.
Another direction is represented by the methods transFold
[130, 131] (see Table 1) and TMBpro [132] (see Table 1), which
aim to deliver more information than just the TMBB topology
(1D structure). They both predict which residues will pair with
each other in the β-strand (2D structure), and TMBpro addition-
ally predicts a full set of coordinates of the protein (3D structure).
However, they cannot be used to screen a data set for outer membrane
 Protein Sorting Prediction 47

proteins since they both assume that the submitted protein (only
one at a time is accepted) is a TMBB protein.
Methods that are best at one task are not necessarily the same
ones that are best at another task. In the paper on BOCTOPUS
version 2 [125], the authors compare their performance to a num-
ber of other methods and find that BOCTOPUS 2 is better at
predicting the correct topology than a number of other recent
methods (PRED-TMBB, ProfTMB, and BetAware; transFold and
TMBpro are not included in the benchmark), but HHomp and
BetAware yield better discrimination. Note that TMBB prediction
is still more difficult than TMH prediction; BOCTOPUS 2 reports
69% correct TMBB topologies, in contrast with the 83% correct
TMH topologies reported by TOPCONS.

8  Multicategory Predictors

The first software to attempt a classification of proteins into mul-

tiple SCLs was PSORT [133]. It was basically a signal-based
method, incorporating the previously mentioned early methods
for prediction of SPs [45, 46] and TMHs [90], but it also used
amino acid composition, especially for recognizing outer mem-
brane proteins.
For bacteria, PSORT has been superseded by PSORTb [3, 42,
134] (see Table 1), which is now in version 3. Version 1 was for
Gram-negative bacteria only, but version 2 included Gram-positive
bacteria. Version 3 additionally offers predictions for Archaea and
the so-called problematic bacteria, which stain either Gram-­
positive, although they have an outer membrane (such as genus
Deinococcus), or Gram-negative, although they have no outer
membrane (phylum Tenericutes).
PSORTb is a hybrid method, incorporating signal-based,
global-property-based, and homology-based predictions. The
signal-­based component comprises recognition of SPs and TMHs
and a database of motifs (regular expressions) derived from
PROSITE, which are found to be exclusive to specific SCLs. The
global properties component is SVM-based; in version 1 its input
consisted of amino acid composition only, but in versions 2 and 3,
a collection of overrepresented subsequences is used. The
homology-­based component is a simple BLAST with direct anno-
tation transfer. Finally, a Bayesian network is used to integrate the
outputs from the components and arrive at a final prediction.
The final prediction, however, may be “unknown.” PSORTb
values precision over recall, so it prefers to deliver no prediction
rather than a prediction with weak evidence. It may also arrive
at two SCLs, signifying that the protein is predicted to function
in both compartments or belong to the interface between the
compartments.
48 Henrik Nielsen

The SCLs predicted by PSORTb 3 in some cases extend

beyond the standard five categories for Gram-negative bacteria;
there are new subcategory SCLs, such as fimbrial, flagellar, and
host-associated. The reported precision of PSORTb 3 on Gram-­
negative bacteria (the five main categories only) is 97%, with a
recall of 94%. This was tested by fivefold cross-validation with a
data set that was homology reduced, but only down to 80% iden-
tity. Note that this high recall only applies to bacterial species that
are well represented in the data set or closely related to such spe-
cies; a benchmark paper reported that in Campylobacter jejuni,
PSORTb 3 returned “unknown” for as many as 47% of the protein
sequences [135].
Another predictor that owes its high performance to homol-
ogy is the SCL predictor built into the prediction workbench
Proteome Analyst [15, 33] (see Table 1). It uses a combination of
direct and indirect annotation transfer by retrieving up to three
hits from the Swiss-Prot part of UniProt by BLAST and then pars-
ing the “subcellular location” field, the keywords, and the cross-­
referenced InterPro entries. The retrieved words are then processed
by a Naïve Bayes classifier. Other machine learning methods (ANN
and SVM) were also tried, and although they could enhance per-
formance by a few percent, the authors decided to stick with Naïve
Bayes in order to be able to provide explanations for the individual
predictions. The PSORTb 3 paper [134] reports that PSORTb 3.0
and Proteome Analyst 3.0 have comparable precisions but make
complementary predictions, so that a combined analysis with both
methods has the highest coverage overall.
Kuo-Chen Chou’s group has published a long series of predic-
tors for protein SCL (see, e.g., [18]). They prefer to publish one
website per organism group instead of providing one website with
an option for selecting organism group, and furthermore they tend
to change the name for each new version instead of adding a ver-
sion number. For bacteria, the relevant predictors are named
Gneg-­PLoc [19], Gpos-PLoc [20], Gneg-mPLoc [21], Gpos-
mPLoc [22], iLoc-Gneg [23], and iLoc-Gpos [24] (see Table 1).
The PLoc/mPLoc/iLoc servers are hybrid methods, mostly rely-
ing on indirect homology annotation through the Gene Ontology
(GO) terms of database hits. GO [136] is an ordered system (a
directed acyclic graph) of controlled terms that describe the bio-
logical process, molecular function, and cellular component of
proteins. In the prediction servers, GO terms are extracted from all
database hits with a pairwise identity above a certain cutoff, and
then a k-nearest neighbor classifier is applied to the high-dimen-
sional vectors of occurrences of GO terms. If no hits are found, or
if the found hits have no GO annotation, a profile-based global
property approach is used. The nature of this approach varies
between PLoc, mPLoc, and iLoc. However, the corresponding
papers contain no information about how often the global property
 Protein Sorting Prediction 49

approach is needed, and the performance of this approach has

never been reported separately. The overall accuracy, measured by
jackknife (leave-one-out cross-validation) is reported to be 87% for
Gneg-­PLoc (with eight categories; the standard five plus fimbrium,
flagellum, and nucleoid), 86% for Gneg-mPLoc, and 91% for iLoc-
Gneg. Note that the PLoc and mPLoc servers only accept one
sequence per submission, while iLoc accepts up to 50.
The new feature of the mPLoc and iLoc servers relative to the
PLoc servers is the ability to predict multiple SCLs, i.e. predict
whether a protein can exist in more than one cellular compartment.
This can be of importance in eukaryotes, where many proteins, for
example, may shuttle between the cytoplasm and the nucleus, but it
is of somewhat minor importance in bacteria. The data set for
Gneg-mPLoc and iLoc-Gneg contained only 64 such examples out
of 1,392 homology-reduced proteins.
PSLpred [137] (see Table 1) is a hybrid method specific to
Gram-negative bacteria. It uses amino acid composition, dipeptide
composition, physicochemical parameters, and PSI-BLAST against
a database of proteins with known SCL and integrates the features
by SVM. Interestingly, the authors found that the PSI-BLAST
module on its own performed worse than the other modules. The
final reported overall accuracy was 91% on the PSORTb 2 data.
Another hybrid approach is LocTree3 [138] (see Table 1),
which has two components: a PSI-BLAST [10] homology search
with direct transfer of SCL annotation and a global property
approach corresponding to LocTree2 [139]. If no homology hits
are found with an E-value better than a specified cutoff, the
LocTree2 method is applied. It consists of a decision tree of SVMs
trained with a so-called profile kernel, basically using the occur-
rence of short substrings in profiles made by PSI-BLAST as input.
When delivering a prediction, LocTree3 reports whether the evi-
dence is based on a homology or on LocTree2. For bacteria
LocTree2 has a reported performance (overall accuracy) of 86%,
and LocTree3 has 90%. Interestingly, in the LocTree3 paper, the
measured accuracy for PSORTb 3.0 was only 57%. This huge dif-
ference from PSORTb’s own reported performance is hard to
explain, but it may reflect different views on the exactly correct way
to parse UniProt’s SCL annotations.
It is claimed that LocTree2/3 can predict SCLs for all domains
of life, but it seems less well suited for Gram-positive bacteria since
it offers no opportunity to choose between Gram-positives and
Gram-negatives. Thus, it may predict categories like periplasm and
outer membrane for Gram-positive bacteria, while totally failing to
predict cell wall.
The methods described so far in this section have all been
wholly or partly homology-based. However, there is also the
global-property-based CELLO [13] (see Table 1), an SVM-
based predictor for eukaryotes, Gram-negative bacteria, and
50 Henrik Nielsen

Gram-­positive bacteria. The SVMs are organized in a two-level

system, where the first level contains a number of SVMs trained
on various sequence encodings, and the second layer is a so-called
jury SVM that decides on the prediction based on the outputs of
the first-­layer SVMs. The sequences are encoded by total amino
acid composition, dipeptide composition, and amino acid compo-
sition (in some cases with a reduced alphabet) in a number of parti-
tions of each sequence. The performance for Gram-negative
bacteria is reported to be 95% overall accuracy without homology
reduction and 83% with homology reduction (down to 30%
identity).
Another predictor that is not homology-based is SOSUI-­
GramN [140] (see Table 1), which, as the name suggests, is specific
to Gram-negative bacteria. It is not based on machine learning but
on various physicochemical properties measured over an entire
sequence and in the N- and C-terminal regions. Tests for these
measured parameters are then arranged in a complex decision tree
that takes into account that, for example, secretion can happen
through various pathways.
The authors of the consensus method MetaLocGramN [135]
(see Table 1) benchmarked four methods, PSORTb 3, CELLO,
SOSUI-GramN, and PSLpred, and found PSORTb 3 to be the
best in overall performance. However, PSORTb 3 had a very low
sensitivity for the extracellular category, where PSLpred was found
to be better. The consensus method MetaLocGramN is based not
only on these four predictors but also on a number of signal-based
predictors focusing on SPs, TMHs, TMBBs, and type III secre-
tion signals. To integrate all the predictors, an ANN was tried, but
it did not work consistently better than the constituent methods.
Instead, the finished method is based on statistical feature selec-
tion and logistic regression, and the result is better than in
PSORTb 3, especially for the extracellular category, measured on
a new data set. Unfortunately, the MetaLocGramN web server
depends on all the constituent methods being up and alive. At the
time of writing, SOSUI-GramN was temporarily unavailable,
which made MetaLocGramN hang indefinitely.

9  Discussion

As is apparent from this chapter, the many possible ways of

approaching the SCL prediction problem has resulted in a large
number of available prediction servers. Comparing their perfor-
mances can be complicated, and all their authors tend to claim
superior performance for their particular method. Add to this
that usability is sometimes limited (some web servers allow only
one or a few sequences in each submission), that response times
vary considerably, and that there are almost as many different
 Protein Sorting Prediction 51

output formats as there are servers, and you get a rather frustrating
situation. Even the definitions of SCLs may vary from server to
server; for example, a peripheral membrane protein may be
defined as belonging to the membrane or to the compartment it
protrudes into.
This situation is clearly not ideal for the user, who might prefer
a “one-stop shop” to go to for all sequence-based prediction needs,
an equivalent of UniProt or InterPro. But this kind of confusion is
probably inevitable in a field that is evolving so fast. Scientific com-
petition is basically beneficial, and competing groups should cer-
tainly not be discouraged from publishing their predictors
independently. That being said, prediction servers ought to follow
certain standards concerning usability, definitions, and formats.
Personally, I must admit to having added to the complexity
through my involvement in the servers SignalP, LipoP, and TatP
(see Subheading 5). In hindsight, we should have published LipoP
and TatP not as separate servers but as functionalities within the
SignalP server. Hopefully, the next version of SignalP will be able
to predict all these types of SP in one user interface.
The multicategory prediction methods report quite impressive
performances, and, as described in Subheading 1, the error rates
for prediction may now be lower than the error rates for high-­
throughput experiments. However, it is important to keep in mind
that these performances are achieved by analyzing the annotations
of homologs found by sequence similarity searches. I see three
problems with this. First, predictions for novel organisms and
metagenomics samples with few known homologs will necessarily
be harder than for the organisms the training and test sets were
built from, so coverage and precision for such organisms will be
considerably lower than the reported performances. Second, the
annotations used for prediction are themselves error prone and not
necessarily derived from experiments. Especially when relying on
keywords and GO terms, there is a real danger of circular reason-
ing, where annotations based on predictions are used as a basis for
new predictions, which then may enter the databases as annota-
tions. Third, homology-based predictions do not reflect real bio-
logical knowledge about the protein sorting process in the way a
successful signal-based predictor does.
As machine learning algorithms continue to evolve, new classes
of algorithms should also be expected to be applied to predictions
of SCL. In fields such as image processing and speech recognition,
novel types of ANNs—deep and recurrent neural networks—have
been used extensively in recent years [141, 142], and they are
beginning to be employed in bioinformatics as well, for example,
to predict protein secondary structures [143] or alternative splic-
ing [144]. The previously mentioned TMBpro (see Subheading 7)
is an example of a method based on recurrent ANNs. The advan-
tage of recurrent ANNs is that they are naturally designed to handle
52 Henrik Nielsen

sequential data, so sequences are not chopped up into apparently

unrelated windows, and they can potentially learn long-range cor-
relations. A first attempt at using a recurrent ANN in multicate-
gory SCL prediction was published recently [145]—so far, only for
eukaryotic data, and without a web server, but the results seem
promising. Coupled with a so-called convolutional layer—basically
a series of PWMs of varying width through which the input
sequences were presented—the network was able to learn from the
data where in each sequence to focus its attention. Performance
was much better than that of other methods working only on
sequence, and on the same level as advanced homology-based
methods. This technology represents a new kind of compromise
between signal-based and global-property-based methods since it
is apparently able to find sorting signals in sequences, even though
it is given the sequences and their SCL categories only during
training. It will be very interesting to see where this and other
novel technologies will take SCL prediction in the coming years.

References
1. Kyte J, Doolittle RF (1982) A simple method location of proteins. Nucleic Acids Res
for displaying the hydropathic character of a 26:2230–2236
protein. J Mol Biol 157:105–132 9. Hua S, Sun Z (2001) Support vector machine
2. von Heijne G (1983) Patterns of amino acids approach for protein subcellular localization
near signal-sequence cleavage sites. Eur prediction. Bioinformatics 17:721–728
J Biochem 133:17–21 10. Altschul SF, Madden TL, Schaffer AA et al
3. Gardy JL, Laird MR, Chen F et al (2005) (1997) Gapped BLAST and PSI-BLAST: a
PSORTb v.2.0: expanded prediction of bacte- new generation of protein database search
rial protein subcellular localization and programs. Nucleic Acids Res 25:3389
insights gained from comparative proteome 11. The UniProt Consortium (2015) UniProt: a
analysis. Bioinformatics 21:617–623 hub for protein information. Nucleic Acids
4. Rey S, Gardy J, Brinkman F (2005) Assessing Res 43:D204–D212
the precision of high-throughput computa- 12. Nair R, Rost B (2002a) Sequence conserved
tional and laboratory approaches for the for subcellular localization. Protein Sci
genome-wide identification of protein subcel- 11:2836–2847
lular localization in bacteria. BMC Genomics 13. Yu C-S, Chen Y-C, Lu C-H, Hwang J-K
6:162 (2006) Prediction of protein subcellular local-
5. Nielsen H (2016) Predicting subcellular ization. Proteins 64:643–651
localization of proteins by bioinformatic algo- 14. Nair R, Rost B (2002b) Inferring sub-cellular
rithms. In: Bagnoli F, Rappuoli R (eds) localization through automated lexical analy-
Protein export in gram-positive bacteria. sis. Bioinformatics 18(Suppl 1):S78–S86
Current topics in microbiology and immunol-
ogy. Springer, Berlin, Heidelberg 15. Lu Z, Szafron D, Greiner R et al (2004)
Predicting subcellular localization of proteins
6. Nakashima H, Nishikawa K (1994) using machine-learned classifiers. Bioinformatics
Discrimination of intracellular and extracellu- 20:547–556
lar proteins using amino acid composition and
residue-pair frequencies. J Mol Biol 16. Shatkay H, Höglund A, Brady S et al (2007)
238:54–61 SherLoc: high-accuracy prediction of protein
subcellular localization by integrating text
7. Andrade MA, O’Donoghue SI, Rost B (1998) and protein sequence data. Bioinformatics
Adaptation of protein surfaces to subcellular 23:1410–1417
location. J Mol Biol 276:517–525
17. Briesemeister S, Blum T, Brady S et al (2009)
8. Reinhardt A, Hubbard T (1998) Using neu- SherLoc2: a high-accuracy hybrid method for
ral networks for prediction of the subcellular
 Protein Sorting Prediction 53

predicting subcellular localization of proteins. 31. Mitchell A, Chang H-Y, Daugherty L et al
J Proteome Res 8:5363–5366 (2015) The InterPro protein families data-
18. Chou K-C, Shen H-B (2010) Cell-PLoc 2.0: base: the classification resource after 15 years.
an improved package of web-servers for pre- Nucleic Acids Res 43:D213–D221
dicting subcellular localization of proteins in 32. Rish I (2001) An empirical study of the naive
various organisms. Nat Sci 2:1090–1103 Bayes classifier. In: IJCAI 2001 workshop
19. Chou K-C, Shen H-B (2006) Large-scale pre- Empir Methods Artif Intell. IBM, New York,
dictions of gram-negative bacterial protein pp 41–46
subcellular locations. J Proteome Res 33. Szafron D, Lu P, Greiner R et al (2004)
5:3420–3428 Proteome analyst: custom predictions with
20. Shen H-B, Chou K-C (2007) Gpos-PLoc: an explanations in a web-based tool for high-­
ensemble classifier for predicting subcellular throughput proteome annotations. Nucleic
localization of gram-positive bacterial pro- Acids Res 32:W365–W371
teins. Protein Eng Des Sel 20:39–46 34. Briesemeister S, Rahnenführer J, Kohlbacher
21. Shen H-B, Chou K-C (2010) Gneg-mPLoc: a O (2010) Going from where to why—inter-
top-down strategy to enhance the quality of pretable prediction of protein subcellular
predicting subcellular localization of gram-­ localization. Bioinformatics 26:1232–1238
negative bacterial proteins. J Theor Biol 35. Hertz JA, Krogh AS, Palmer RG (1991)
264:326–333 Introduction to the theory of neural compu-
22. Shen H-B, Chou K-C (2009) Gpos-mPLoc: a tation. Westview Press, Redwood City, CA
top-down approach to improve the quality 36. Noble WS (2006) What is a support vector
of predicting subcellular localization of machine? Nat Biotechnol 24:1565–1567
gram-­ positive bacterial proteins. Protein 37. Hobohm U, Scharf M, Schneider R, Sander C
Pept Lett 16:1478–1484 (1992) Selection of representative protein
23. Xiao X, Wu Z-C, Chou K-C (2011) A multi-­ data sets. Protein Sci 1:409–417
label classifier for predicting the subcellular 38. Höglund A, Dönnes P, Blum T et al (2006)
localization of gram-negative bacterial pro- MultiLoc: prediction of protein subcellular
teins with both single and multiple sites. PLoS localization using N-terminal targeting
One 6:e20592 sequences, sequence motifs and amino acid
24. Wu Z-C, Xiao X, Chou K-C (2012) iLoc-­ composition. Bioinformatics 22:1158–1165
Gpos: a multi-layer classifier for predicting the 39. Sander C, Schneider R (1991) Database of
subcellular localization of singleplex and mul- homology-derived protein structures and
tiplex gram-positive bacterial proteins. Protein the structural meaning of sequence align-
Pept Lett 19:4–14 ment. Proteins 9:56–68
25. Stormo GD, Schneider TD, Gold L, 40. Nielsen H, Engelbrecht J, von Heijne G,
Ehrenfeucht A (1982) Use of the “percep- Brunak S (1996) Defining a similarity thresh-
tron” algorithm to distinguish translational old for a functional protein sequence pattern:
initiation sites in E. coli. Nucleic Acids Res the signal peptide cleavage site. Proteins
10:2997–3011 24:165–177
26. Schneider TD, Stephens RM (1990) Sequence 41. Nielsen H, Wernersson R (2006) An over-
logos: a new way to display consensus abundance of phase 0 introns immediately
sequences. Nucleic Acids Res 18:6097–6100 after the start codon in eukaryotic genes.
27. Krogh A, Brown M, Mian IS et al (1994) BMC Genomics 7:256
Hidden Markov models in computational 42. Gardy JL, Spencer C, Wang K et al (2003)
biology: applications to protein modeling. PSORT-B: improving protein subcellular
J Mol Biol 235:1501–1531 localization prediction for gram-negative bac-
28. Sigrist CJA, de Castro E, Cerutti L et al (2013) teria. Nucleic Acids Res 31:3613–3617
New and continuing developments at 43. Baldi P, Brunak S, Chauvin Y et al (2000)
PROSITE. Nucleic Acids Res 41:D344–D347 Assessing the accuracy of prediction algo-
29. Finn RD, Bateman A, Clements J et al (2014) rithms for classification: an overview.
Pfam: the protein families database. Nucleic Bioinformatics 16:412–424
Acids Res 42:D222–D230 44. Gorodkin J (2004) Comparing two
30. Haft DH, Selengut JD, Richter RA et al K-category assignments by a K-category cor-
(2013) TIGRFAMs and genome properties in relation coefficient. Comput Biol Chem
2013. Nucleic Acids Res 41:D387–D395 28:367–374
54 Henrik Nielsen

45. von Heijne G (1986) A new method for pre- peptides in gram-negative bacteria. Protein
dicting signal sequence cleavage sites. Nucleic Sci 12:1652–1662
Acids Res 14:4683–4690 61. Rahman O, Cummings SP, Harrington DJ,
46. McGeoch DJ (1985) On the predictive rec- Sutcliffe IC (2008) Methods for the bioinfor-
ognition of signal peptide sequences. Virus matic identification of bacterial lipoproteins
Res 3:271–286 encoded in the genomes of gram-positive
47. von Heijne G, Abrahmsén L (1989) Species-­ bacteria. World J Microbiol Biotechnol
specific variation in signal peptide design: 24:2377–2382
implications for protein secretion in foreign 62. Fariselli P, Finocchiaro G, Casadio R (2003)
hosts. FEBS Lett 244:439–446 SPEPlip: the detection of signal peptide and
48. Nielsen H, Brunak S, Engelbrecht J, von lipoprotein cleavage sites. Bioinformatics
Heijne G (1997) Identification of prokaryotic 19:2498–2499
and eukaryotic signal peptides and prediction 63. Bagos PG, Tsirigos KD, Liakopoulos TD,
of their cleavage sites. Protein Eng 10:1–6 Hamodrakas SJ (2008) Prediction of lipopro-
49. Nielsen H, Krogh A (1998) Prediction of sig- tein signal peptides in gram-positive bacteria
nal peptides and signal anchors by a hidden with a hidden Markov model. J Proteome Res
Markov model. Proc Int Conf Intell Syst Mol 7:5082–5093
Biol 6:122–130 64. Cristóbal S, de Gier J-W, Nielsen H, von
50. Bendtsen JD, Nielsen H, von Heijne G, Heijne G (1999) Competition between Sec-
Brunak S (2004) Improved prediction of sig- and TAT-dependent protein translocation in
nal peptides: SignalP 3.0. J Mol Biol Escherichia coli. EMBO J 18:2982–2990
340:783–795 65. Rose RW, Brüser T, Kissinger JC, Pohlschröder
51. Petersen TN, Brunak S, von Heijne G, M (2002) Adaptation of protein secretion to
Nielsen H (2011) SignalP 4.0: discriminating extremely high-salt conditions by extensive
signal peptides from transmembrane regions. use of the twin-arginine translocation pathway.
Nat Methods 8:785–786 Mol Microbiol 45:943–950
52. Menne KML, Hermjakob H, Apweiler R 66. Bendtsen JD, Nielsen H, Widdick D et al
(2000) A comparison of signal sequence pre- (2005a) Prediction of twin-arginine signal
diction methods using a test set of signal pep- peptides. BMC Bioinformatics 6:167
tides. Bioinformatics 16:741–742 67. Bagos PG, Nikolaou EP, Liakopoulos TD,
53. Klee E, Ellis L (2005) Evaluating eukaryotic Tsirigos KD (2010) Combined prediction of
secreted protein prediction. BMC Tat and Sec signal peptides with hidden
Bioinformatics 6:1–7 Markov models. Bioinformatics 26:
54. Choo K, Tan T, Ranganathan S (2009) A 2811–2817
comprehensive assessment of N-terminal sig- 68. Binnewies TT, Bendtsen JD, Hallin PF et al
nal peptides prediction methods. BMC (2005) Genome update: protein secretion
Bioinformatics 10:S2 systems in 225 bacterial genomes.
55. Zhang X, Li Y, Li Y (2009) Evaluating signal Microbiology 151:1013–1016
peptide prediction methods for gram-positive 69. Desvaux M, Hébraud M, Talon R, Henderson
bacteria. Biologia (Bratisl) 64:655–659 IR (2009) Secretion and subcellular localiza-
56. Hiller K, Grote A, Scheer M et al (2004) tions of bacterial proteins: a semantic aware-
PrediSi: prediction of signal peptides and ness issue. Trends Microbiol 17:139–145
their cleavage positions. Nucleic Acids Res 70. Bendtsen JD, Kiemer L, Fausbøll A, Brunak S
32:W375–W379 (2005b) Non-classical protein secretion in
57. Gomi M, Sonoyama M, Mitaku S (2004) bacteria. BMC Microbiol 5:58
High performance system for signal peptide 71. Yu L, Guo Y, Li Y et al (2010a) SecretP: iden-
prediction: SOSUIsignal. Chem-Bio Inform tifying bacterial secreted proteins by fusing
J 4:142–147 new features into Chou’s pseudo-amino acid
58. Frank K, Sippl MJ (2008) High-performance composition. J Theor Biol 267:1–6
signal peptide prediction based on sequence 72. Yu L, Luo J, Guo Y et al (2013) In silico iden-
alignment techniques. Bioinformatics 24: tification of gram-negative bacterial secreted
2172–2176 proteins from primary sequence. Comput
59. Broome-Smith JK, Gnaneshan S, Hunt LA Biol Med 43:1177–1181
et al (1994) Cleavable signal peptides are 73. Lloubes R, Bernadac A, Houot L, Pommier S
rarely found in bacterial cytoplasmic mem- (2013) Non classical secretion systems. Res
brane proteins. Mol Membr Biol 11:3–8 Microbiol 164:655–663
60. Juncker AS, Willenbrock H, von Heijne G 74. Luo J, Li W, Liu Z et al (2015) A sequence-­
et al (2003) Prediction of lipoprotein signal based two-level method for the prediction of
 Protein Sorting Prediction 55

type I secreted RTX proteins. Analyst 140: 88. Dong X, Lu X, Zhang Z (2015) BEAN 2.0:
3048–3056 an integrated web resource for the identifica-
75. Burstein D, Zusman T, Degtyar E et al (2009) tion and functional analysis of type III
Genome-scale identification of Legionella secreted effectors. Database 2015:bav064
pneumophila effectors using a machine learn- 89. Goldberg T, Rost B, Bromberg Y (2016)
ing approach. PLoS Pathog 5:e1000508 Computational prediction shines light on
76. Chen C, Banga S, Mertens K et al (2010) type III secretion origins. Sci Rep 6:34516
Large-scale identification and translocation of 90. Klein P, Kanehisa M, DeLisi C (1985) The
type IV secretion substrates by Coxiella bur- detection and classification of membrane-­
netii. Proc Natl Acad Sci U S A spanning proteins. Biochim Biophys Acta
107:21755–21760 815:468–476
77. Lifshitz Z, Burstein D, Peeri M et al (2013) 91. von Heijne G (1992) Membrane protein
Computational modeling and experimental structure prediction: hydrophobicity analysis
validation of the Legionella and Coxiella and the positive-inside rule. J Mol Biol 225:
virulence-­ related type-IVB secretion signal. 487–494
Proc Natl Acad Sci U S A 110:E707–E715 92. von Heijne G, Gavel Y (1988) Topogenic
78. Zou L, Nan C, Hu F (2013) Accurate predic- signals in integral membrane proteins. Eur
tion of bacterial type IV secreted effectors J Biochem 174:671–678
using amino acid composition and PSSM pro- 93. Paul C, Rosenbusch JP (1985) Folding pat-
files. Bioinformatics 29:3135–3142 terns of porin and bacteriorhodopsin. EMBO
79. Wang Y, Wei X, Bao H, Liu S-L (2014) J 4:1593–1597
Prediction of bacterial type IV secreted effec- 94. Vogel H, Jähnig F (1986) Models for the
tors by C-terminal features. BMC Genomics structure of outer-membrane proteins of
15:50 Escherichia coli derived from raman spec-
80. McDermott JE, Corrigan A, Peterson E et al troscopy and prediction methods. J Mol
(2011) Computational prediction of type III Biol 190:191–199
and IV secreted effectors in gram-negative 95. Krogh A, Larsson B, von Heijne G,
bacteria. Infect Immun 79:23–32 Sonnhammer EL (2001) Predicting trans-
81. Anderson DM, Schneewind O (1997) A membrane protein topology with a hidden
mRNA signal for the type III secretion of Yop Markov model: application to complete
proteins by Yersinia enterocolitica. Science genomes. J Mol Biol 305:567–580
278:1140–1143 96. Tusnády GE, Simon I (2001) The HMMTOP
82. Samudrala R, Heffron F, McDermott JE transmembrane topology prediction server.
(2009) Accurate prediction of secreted sub- Bioinformatics 17:849–850
strates and identification of a conserved puta- 97. Möller S, Croning MDR, Apweiler R (2001)
tive secretion signal for type III secretion Evaluation of methods for the prediction of
systems. PLoS Pathog 5:e1000375 membrane spanning regions. Bioinformatics
83. Arnold R, Brandmaier S, Kleine F et al (2009) 17:646–653
Sequence-based prediction of type III 98. Elofsson A, von Heijne G (2007) Membrane
secreted proteins. PLoS Pathog 5:e1000376 protein structure: prediction versus reality.
84. Löwer M, Schneider G (2009) Prediction of Annu Rev Biochem 76:125–140
type III secretion signals in genomes of gram-­ 99. Punta M, Forrest LR, Bigelow H et al (2007)
negative bacteria. PLoS One 4:e5917 Membrane protein prediction methods.
85. Wang Y, Zhang Q, Sun M, Guo D (2011) Methods 41:460–474
High-accuracy prediction of bacterial type III 100. Tusnády GE, Simon I (2010) Topology

secreted effectors based on position-specific prediction of helical transmembrane pro-
amino acid composition profiles. teins: how far have we reached? Curr
Bioinformatics 27:777–784 Protein Pept Sci 11:550–561
86. Wang Y, Sun M, Bao H, White AP (2013) 101. Käll L, Krogh A, Sonnhammer EL (2004) A
T3_MM: a Markov model effectively classi- combined transmembrane topology and signal
fies bacterial type III secretion signals. PLoS peptide prediction method. J Mol Biol 338:
One 8:e58173 1027–1036
87. Dong X, Zhang Y-J, Zhang Z (2013) Using 102. Reynolds SM, Käll L, Riffle ME et al (2008)
weakly conserved motifs hidden in secretion Transmembrane topology and signal peptide
signals to identify type-III effectors from bac- prediction using dynamic Bayesian networks.
terial pathogen genomes. PLoS One 8:e56632 PLoS Comput Biol 4:e1000213
56 Henrik Nielsen

103. Jones DT (2007) Improving the accuracy of 117. Bagos P, Liakopoulos T, Spyropoulos I,

transmembrane protein topology prediction Hamodrakas S (2004a) A hidden Markov
using evolutionary information. model method, capable of predicting and
Bioinformatics 23:538–544 discriminating beta-barrel outer membrane
104. Nugent T, Jones DT (2009) Transmembrane proteins. BMC Bioinformatics 5:29
protein topology prediction using support 118. Bagos PG, Liakopoulos TD, Spyropoulos IC,
vector machines. BMC Bioinformatics 10:159 Hamodrakas SJ (2004b) PRED-TMBB: a
105. Viklund H, Bernsel A, Skwark M, Elofsson A web server for predicting the topology of
(2008) SPOCTOPUS: a combined predictor β-barrel outer membrane proteins. Nucleic
of signal peptides and membrane protein Acids Res 32:W400–W404
topology. Bioinformatics 24:2928–2929 119. Bigelow HR, Petrey DS, Liu J et al (2004)
106. Viklund H, Elofsson A (2008) OCTOPUS: Predicting transmembrane beta-barrels in
improving topology prediction by two-track proteomes. Nucleic Acids Res 32:2566–2577
ANN-based preference scores and an 120. Bigelow H, Rost B (2006) PROFtmb: a web
extended topological grammar. Bioinformatics server for predicting bacterial transmembrane
24:1662–1668 beta barrel proteins. Nucleic Acids Res
107. Viklund H, Elofsson A (2004) Best α-helical 34:W186–W188
transmembrane protein topology predictions 121. Bagos P, Liakopoulos T, Hamodrakas S

are achieved using hidden Markov models (2005) Evaluation of methods for predicting
and evolutionary information. Protein Sci the topology of beta-barrel outer membrane
13:1908–1917 proteins and a consensus prediction method.
108. Käll L, Krogh A, Sonnhammer EL (2005) An BMC Bioinformatics 6:7
HMM posterior decoder for sequence feature 122. Jacoboni I, Martelli PL, Fariselli P et al (2001)
prediction that includes homology information. Prediction of the transmembrane regions of
Bioinformatics 21:i251–i257 β-barrel membrane proteins with a neural
109. Bernsel A, Viklund H, Falk J et al (2008) network-based predictor. Protein Sci
Prediction of membrane-protein topology 10:779–787
from first principles. Proc Natl Acad Sci 105: 123. Natt NK, Kaur H, Raghava GPS (2004)

7177–7181 Prediction of transmembrane regions of
110. Hessa T, Meindl-Beinker NM, Bernsel A et al β-barrel proteins using ANN- and SVM-based
(2007) Molecular code for transmembrane-­ methods. Proteins 56:11–18
helix recognition by the Sec61 translocon. 124. Hayat S, Elofsson A (2012) BOCTOPUS:

Nature 450:1026–1030 improved topology prediction of transmem-
111. Taylor PD, Attwood TK, Flower DR (2003) brane β barrel proteins. Bioinformatics
BPROMPT: a consensus server for membrane 28:516–522
protein prediction. Nucleic Acids Res 125. Hayat S, Peters C, Shu N et al (2016) Inclusion
31:3698–3700 of dyad-repeat pattern improves topology pre-
112. Bernsel A, Viklund H, Hennerdal A, Elofsson diction of transmembrane β-barrel proteins.
A (2009) TOPCONS: consensus prediction Bioinformatics 32:1571–1573
of membrane protein topology. Nucleic 126. Berven FS, Flikka K, Jensen HB, Eidhammer
Acids Res 37:W465–W468 I (2004) BOMP: a program to predict inte-
113. Tsirigos KD, Peters C, Shu N et al (2015) gral β-barrel outer membrane proteins
The TOPCONS web server for consensus encoded within genomes of gram-negative
prediction of membrane protein topology and bacteria. Nucleic Acids Res 32:W394–W399
signal peptides. Nucleic Acids Res 127. Remmert M, Linke D, Lupas AN, Söding

43:W401–W407 J (2009) HHomp—prediction and classifica-
114. Hennerdal A, Elofsson A (2011) Rapid mem- tion of outer membrane proteins. Nucleic
brane protein topology prediction. Acids Res 37:W446–W451
Bioinformatics 27:1322–1323 128. Savojardo C, Fariselli P, Casadio R (2011)

115. Diederichs K, Freigang J, Umhau S et al
Improving the detection of transmembrane
(1998) Prediction by a neural network of β-barrel chains with N-to-1 extreme learning
outer membrane β-strand protein topology. machines. Bioinformatics 27:3123–3128
Protein Sci 7:2413–2420 129. Savojardo C, Fariselli P, Casadio R (2013)

116. Martelli PL, Fariselli P, Krogh A, Casadio R BETAWARE: a machine-learning tool to
(2002) A sequence-profile-based HMM for detect and predict transmembrane beta-barrel
predicting and discriminating β barrel mem- proteins in prokaryotes. Bioinformatics
brane proteins. Bioinformatics 18:S46–S53 29:504–505
 Protein Sorting Prediction 57

130. Waldispühl J, Berger B, Clote P, Steyaert J-M 140. Imai K, Asakawa N, Tsuji T et al (2008)

(2006a) transFold: a web server for predicting SOSUI-GramN: high performance prediction
the structure and residue contacts of trans- for sub-cellular localization of proteins in
membrane beta-barrels. Nucleic Acids Res gram-negative bacteria. Bioinformation
34:W189–W193 2:417–421
131. Waldispühl J, Berger B, Clote P, Steyaert J-M 141. Krizhevsky A, Sutskever I, Hinton GE (2012)
(2006b) Predicting transmembrane β-barrels ImageNet classification with deep convolu-
and interstrand residue interactions from tional neural networks. In: Pereira F, Burges
sequence. Proteins 65:61–74 CJC, Bottou L, Weinberger KQ (eds)
132. Randall A, Cheng J, Sweredoski M, Baldi P Advances in neural information processing
(2008) TMBpro: secondary structure, systems, vol. 25, Curran Associates, Inc., Red
β-contact and tertiary structure prediction of Hook, NY, pp 1097–1105
transmembrane β-barrel proteins. 142. Dahl GE, Yu D, Deng L, Acero A (2012)
Bioinformatics 24:513–520 Context-dependent pre-trained deep neural
133. Nakai K, Kanehisa M (1991) Expert system networks for large-vocabulary speech recog-
for predicting protein localization sites in nition. IEEE Trans Audio Speech Lang
gram-negative bacteria. Proteins 11:95–110 Process 20:30–42
134. Yu NY, Wagner JR, Laird MR et al (2010b) 143. Magnan CN, Baldi P (2014) SSpro/ACCpro
PSORTb 3.0: improved protein subcellular 5: almost perfect prediction of protein sec-
localization prediction with refined localization ondary structure and relative solvent accessi-
subcategories and predictive capabilities for all bility using profiles, machine learning and
prokaryotes. Bioinformatics 26:1608–1615 structural similarity. Bioinformatics
135. Magnus M, Pawlowski M, Bujnicki JM
30:2592–2597
(2012) MetaLocGramN: a meta-predictor of 144. Xiong HY, Alipanahi B, Lee LJ et al (2015)
protein subcellular localization for gram-­ The human splicing code reveals new insights
negative bacteria. Biochim Biophys Acta into the genetic determinants of disease.
1824:1425–1433 Science 347:1254806
136. Ashburner M, Ball CA, Blake JA et al (2000) 145. Sønderby SK, Sønderby CK, Nielsen H,

Gene ontology: tool for the unification of Winther O (2015) Convolutional LSTM
biology. Nat Genet 25:25–29 networks for subcellular localization of pro-
137. Bhasin M, Garg A, Raghava GPS (2005)
teins. In: Dediu A-H, Hernández-Quiroz F,
PSLpred: prediction of subcellular localiza- Martín-Vide C, Rosenblueth DA (eds)
tion of bacterial proteins. Bioinformatics Algorithms for computational biology,
21:2522–2524 Lecture notes in computer science, vol
9199. Springer International Publishing,
138. Goldberg T, Hecht M, Hamp T et al (2014) New York, pp 68–80
LocTree3 prediction of localization. Nucleic
Acids Res 42:W350–W355 146.
Crooks GE, Hon G, Chandonia J-M,
Brenner SE (2004) WebLogo: a sequence
139. Goldberg T, Hamp T, Rost B (2012)
logo generator. Genome Res 14:1188–1190
LocTree2 predicts localization for all domains
of life. Bioinformatics 28:i458–i465
 Chapter 3

Cell Fractionation
Melissa Petiti, Laetitia Houot, and Denis Duché

Abstract
Protein function is generally dependent on its subcellular localisation. In Gram-negative bacteria such as
Escherichia coli, a protein can be targeted to five different compartments: the cytoplasm, the inner mem-
brane, the periplasm, the outer membrane and the extracellular medium. Different approaches can be used
to determine the protein localisation within a cell such as in silico identification of protein signal sequences
and motifs, electron microscopy and immunogold labelling, optical fluorescence microscopy, and bio-
chemical technics. In this chapter, we describe a simple and efficient method to isolate the different com-
partments of Escherichia coli by a fractionation method and to determine the presence of the protein of
interest. For inner membrane proteins we propose a method to discriminate between integral and periph-
eral membrane proteins.

Key words Spheroplast, Peptidoglycan, Osmotic shock, Freeze and thaw, Protein solubilisation,
Membrane, Subcellular localisation

1  Introduction

Many Gram-negative bacteria secrete extracellular proteins such

as hydrolytic enzymes or toxins. Secretion can occur through
specific macrocomplex systems composed of a more or less large
number of proteins located in the cell envelope. Identifying the
localisation of these proteins is therefore an important task to
address the assembly and the molecular mechanism of these
secretion systems.
Four subcellular compartments compose Gram-negative bacte-
ria, five if we consider the extracellular medium in which effectors
are delivered. These different compartments are the cytoplasm, the
inner membrane (IM), the periplasm in which the peptidoglycan
layer extends, and the outer membrane (OM) [1, 2]. Isolation an
characterization of effectors from the extracellular medium will be
described in Chapter 31 of this book. In this chapter, we will first
describe a simple and efficient method of recovering proteins from
the periplasm and to generate spheroplasts from Escherichia coli cells.

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_3, © Springer Science+Business Media LLC 2017

59
60 Melissa Petiti et al.

Then we will present a method to recover the cytoplasmic and the

membrane fractions from the spheroplasts by several cycles of
freezing and thawing. Finally, we describe how treatments with
specific buffers can give insight into protein–membrane
associations.

2  Materials

2.1  Cell 1. TES buffer: 200 mM Tris–HCl, pH 8.0, 0.5 mM EDTA

Fractionation (ethylenediaminetetraacetic acid), 0.5 M sucrose.
2. Lysosyme 10 mg/mL (freshly prepared solution).
3. DNase 1 10 mg/mL.
4. MgCl2 1 M (stock solution).
5. 100× phenylmethylsulfonyl (PMSF) 0.1 M in absolute ethanol.
Store at −20 °C.
6. Beckman Coulter (Brea, CA) Optima TLX ultracentrifuge
with TLA 55K rotor or equivalent.

2.2  Protein 1. Urea 2 M.

Solubilisation 2. NaCl 0.5 M.
3. Triton X-100 1% (v/v).
4. Sodium carbonate 100 mM pH 11.5, ice cold.
5. Trichloroacetic acid (TCA) 10% (v/v). The stock solution
[TCA 100% (w/v)] is stored at 4 °C in a brown bottle.
6. Acetone 90% (v/v) in ultrapure water stored at −20 °C in a
brown bottle.

3  Methods

3.1  Cell In this section, we detail step by step spheroplast preparation from
Fractionation/ Escherichia coli cells (see Note 1) using a method based on lysozyme/
Spheroplast Formation EDTA treatment [3] and a mild osmotic shock [4–6] (see Note 2).
1. Grow a 3 mL starter culture overnight in a lysogeny broth
medium at 37 °C with required antibiotics.
2. Inoculate a 20 mL culture at OD600 = 0.05 and incubate at 37
°C until the optical density of the culture is around 0.8. If
­necessary, induce protein production under the required con-
ditions (see Note 3).
3. Take 1 mL of culture and centrifuge for 5 min at 5000 × g to
pellet the cells. Discard the supernatant. Resuspend the pellet
in an appropriate volume of sodium dodecyl sulphate poly-
acrylamide gel electrophoresis (SDS-PAGE) loading buffer.
This fraction will be referred as the total cell fraction (T).
 Cell Fractionation 61

The next steps will be performed at 4 °C, and all buffers must
be cooled in advance on ice before use.
4. Centrifuge the remaining culture for 5 min at 5000 × g at 4 °C
(see Note 4). Discard the supernatant.
5. Gently resuspend the cell pellet in 200 μL of TES buffer
(described in materials section, see Note 5). Do not vortex
and do not pipette, only resuspend the cell pellet by inverting
the tube.
6. Add 8 μL of a freshly prepared solution of lysozyme (10 mg/mL
in TES buffer) and mix gently by shaking the tube.
7. Add 720 μL TES buffer diluted twice in water (v/v) and incu-
bate for 30 min on ice. Gently mix the suspension to perform
the osmotic shock by gently inverting and rolling the tube (see
Note 6).
8. Centrifuge at 5000 × g for 5 min at 4 °C. Keep the pellet as the
spheroplast fraction, IM + cytoplasm (+OM), and the superna-
tant as the periplasmic fraction, P.
9. Resuspend the spheroplast fraction in 1 mL of TES buffer
diluted twice in water (v/v) containing 2 mM PMSF, 2 mM
MgCl2 and 10 μg/mL DNase 1 (see Notes 7 and 8).
10. Lyse the spheroplasts by performing four cycles of freezing and
thawing, from −273 °C (liquid azote) to 37 °C (see Note 9).
11. Remove unbroken cells and cell debris by centrifugation at
2000 × g for 5 min. Keep the supernatant as cytoplasmic and
membrane fraction.
12. Centrifuge the supernatant at 120,000 × g, 4 °C, for 45 min.
Use a Beckman Coulter Optima TLX ultracentrifuge and a
TLA55 fixed-angle rotor for small-volume ultracentrifugation
or something similar. The pellet is kept as the membrane frac-
tion and the supernatant is conserved as the cytoplasmic
fraction.
13. The membrane fraction is suspended in 1 mL TES buffer diluted
twice in water (v/v) or in the desired buffer (see Note 10).
Separation of IM and OM is described in Chapter 6 of this
book.
14. At this step, fractions (OD600 = 0.2–0.4) could be tested for
the presence of the protein of interest by SDS-PAGE and west-
ern blot analysis with required antibodies. As a control, the
same fractions can be tested for the presence of specific IM,
OM, cytoplasm, or periplasm markers.

3.2  Protein 1. Prepare 1 mL membrane fractions as previously described.

Solubilisation 2. Aliquot the membrane fractions into five samples, 200 μL
(See Note 11) each.
62 Melissa Petiti et al.

3. Centrifuge at 120,000 × g for 45 min at 4 °C to pellet the

membranes as described previously (see Subheading 3.1,
step 12).
4. Resuspend each pellet in 200 μL 0.5 M NaCl, 2 M urea, 100
mM sodium carbonate, pH 11.5 ice cold, or 1% (v/v) of Triton
X-100 to compare the five extraction conditions.
5. Incubate at least 1 h at 4 °C with agitation.
6. Centrifuge the suspensions at 120,000 × g for 45 min at
4 °C. Carefully collect the different supernatants and transfer
to new tubes.
7. Resuspend each pellet in SDS-PAGE loading buffer and keep
as the membrane-associated protein fractions.
8. Add 10% TCA (final concentration) to supernatant samples,
and incubate for at least 1 h at 4 °C to allow protein precipita-
tion (see Notes 12 and 13).
9. Centrifuge at 18,000 × g for 30 min at 4 °C.
10. Wash the pellets with 200 μL of acetone 90% (pre-chilled
solution).
11. Centrifuge at 18,000 × g for 10 min at 4 °C.
12. Carefully discard the supernatant and air-dry the pellet con-
taining the extracted membrane proteins at room temperature
for 5–10 min (see Note 14).
13. Resuspend pellets in SDS-PAGE loading buffer and keep as
extracted membrane protein fractions.
14. Perform a western blot analysis to identify the extraction con-
dition that is suitable for the protein of interest.

4  Notes

1. Spheroplasts are cells resulting from a loss of the bacterial cell

wall. The OM has been altered but the cytoplasm remains
delimited by the IM [7].
2. Escherichia coli cells are first incubated in a concentrated
sucrose solution containing EDTA. Sucrose makes the medium
hypertonic, while EDTA chelates divalent cations and destabi-
lises the OM. Then lysozyme is added to cleave the periplasmic
peptidoglycan layer. However, peptidoglycan hydrolysis is not
total, and a mild osmotic shock is required to maximize the
procedure. Then the periplasmic content of the cell is sepa-
rated from the spheroplasts by centrifugation.
3. The volume of the cell culture can be adapted according to the
downstream application.
4. Pre-cool centrifuge before use.
 Cell Fractionation 63

5. The TES buffer is responsible for OM destabilisation. 0.5 M

sucrose makes the medium hypertonic, 0.5 mM EDTA, 200
mM Tris–HCl, and pH 8 affect the membrane structure by
removing the lipopolysaccharide coat from the cells [8].
6. This mild osmotic shock provokes a sudden influx of water in
the periplasmic space and increases the distance between poly-
saccharide chains of the peptidoglycan. This facilitates the lyso-
zyme binding and the degradation of the peptidoglycan [8].
7. PMSF is a serine protease inhibitor. However, it has been
shown that the addition of trypsin inhibitor after proteolysis is
not required to prevent further digestion because trypsin
digestion is very specific.
8. During spheroplast lysis, DNA is released in the medium,
adheres to membranes and makes the preparation difficult to
handle. To circumvent this issue, DNase 1 is added to lysates.
As DNase I activity requires magnesium, Mg2+ is added in
excess in order to overtake chelation by the EDTA present in
the TES buffer.
9. Three to five cycles of freezing and thawing are an efficient and
simple method to disrupt spheroplasts [9]. However, sphero-
plasts can also be disrupted by sonication. In this case, sonicate
spheroplast suspension twice for 30 s. Keep the suspension
cold during sonication. A Branson Microtip Sonifier 450
(BRANSON Ultrasonics Corp., Danbury, CT) can be used
with a microtip probe.
10. Membrane fraction resuspension can be difficult. Passing the
sample through the needle of a syringe several times can opti-
mise this step.
11. When studying a poorly characterised protein, it is important
to compare different extraction conditions to optimise protein
solubilisation. The use of an appropriate solubilisation buffer
can provide information about protein localisation in cells and
even be used to differentiate between integral and peripheral
membrane proteins. Thus, a high salt buffer allows the extrac-
tion of peripheral proteins associated with membranes by
­electrostatic interactions. It is common to use 2 M urea to
extract peripheral proteins that associate with membranes by
hydrophobic bonds [10]. Triton-X100 is the most commonly
used detergent for the solubilisation of integral IM proteins
[11]. It is worth noting that, during the preparation of mem-
brane fractions, membranes tend to re-anneal by an unknown
mechanism, resulting in the formation of closed membrane
vesicles that might trap some proteins of interest. In this case,
alkaline carbonate buffer can be added to convert membrane
vesicles to open membrane sheets and release trapped proteins
into the supernatant [12]. Membranes can subsequently be
cleared from the sample by centrifugation.
64 Melissa Petiti et al.

12. This step can be done overnight in a cold room under rotary
agitation.
13. Be careful; pellets are not always visible after centrifugation.
Carefully place the tubes in the centrifuge to determine the
location of the future pellet.
14. Never let the pellet air-dry completely because this will dra-
matically impede resuspension.

Acknowledgments

This work was supported by the Centre National de la Recherche

Scientifique and the Agence National de la Recherche
(ANR-14-CE09-0023).

References
1. Kaback HR (1972) Transport across isolated 7. Kaback HR (1971) Bacterial membranes.
bacterial cytoplasmic membranes. Biochim Methods Enzymol 22:99–120
Biophys Acta 265:367–416 8. Witholt B, Heerikhuizen HV, De Leij L (1976)
2. Kellenberger E, Ryther A (1958) Cell wall and How does lysozyme penetrate through the
cytoplasmic membrane of Escherichia coli. bacterial outer membrane. Biochim Biophys
J Biophys Biochem Cytol 25:323–326 Acta 443:534–544
3. Neu HC, Heppel LA (1964) The release of 9. Mowbray J, Moses V (1976) The tentative
Ribonuclease into the medium when Escherichia identification in Escherichia coli of a multien-
coli cells are converted to spheroplasts. J Biol zyme complex with glycolytic activity. Eur
Chem 239:3893–3900 J Biochem 66:25–36
4. French C, Keshavarz-Moore E, Ward JM 10. Schook W, Puszkin P, Bloom W, Ores C, Kochwa
(1996) Development of a simple method for S (1979) Mechanochemical properties of brain
the recovery of recombinant proteins from the clathrin: interactions with actin and alpha-actinin
Escherichia coli periplasm. Enzym Microb and polymerization into basketlike structures or
Technol 19:332–338 filaments. Proc Natl Acad Sci U S A 76:116–120
5. Skerra A, Plückthun A (1991) Secretion and in 11. Schnaitman CA (1971) Solubilization of the
vivo folding of the Fab fragment of the antibody cytoplasmic membrane of Escherichia coli by
McPC603 in Escherichia coli: influence of triton X-100. J Bacteriol 108:545–552
disulphides and cis-prolines. Protein Eng 12. Fujiki Y, Fowler S, Shio H, Hubbard AL,
4:971–979 Lazarow PB (1982) Polypeptide and phospho-
6. Nossal NG, Heppel LA (1966) The release of lipid composition of the membrane of rat liver
enzymes by osmotic shock from Escherichia peroxisomes: comparison with endoplasmic
coli in exponential phase. J Biol Chem reticulum and mitochondrial membranes.
241:3055–3062 J Cell Biol 93:103–110
 Chapter 4

Defining Lipoprotein Localisation by Fluorescence

Microscopy
†
Maria Guillermina Casabona, Mylène Robert-Genthon, Didier Grunwald ,
and Ina Attrée

Abstract
In recent years it has become evident that lipoproteins play crucial roles in the assembly of bacterial
envelope-­embedded nanomachineries and in the processes of protein export/secretion. In this chapter we
describe a method to determine their precise localisation, for example inner versus outer membrane, in
Gram-negative bacteria using human opportunistic pathogen Pseudomonas aeruginosa as a model. A fusion
protein between a given putative lipoprotein and the red fluorescent protein mCherry must be created and
expressed in a strain expressing cytoplasmic green fluorescent protein (GFP). Then the peripheral localisa-
tion of the fusion protein in the cell can be examined by treating cells with lysozyme to create spheroplasts
and monitoring fluorescence under a confocal microscope. Mutants in the signal peptide can be engi-
neered to study the association with the membrane and efficiency of transport. This protocol can be
adapted to monitor lipoprotein localisation in other Gram-negative bacteria.

Key words Lipoprotein, Localisation, Cell envelope, Spheroplasts, Bacterial secretion, Fluorescence
microscopy

1  Introduction

Lipoproteins are involved in a variety of processes, such as biogen-

esis of the cell envelope and signalling [1, 2], and they can play
important roles in virulence [3]. They are usually hydrophilic pro-
teins present in both Gram-positive and Gram–negative bacteria
that are anchored to membranes thanks to a lipid moiety attached
to their invariable cysteine residue in their N-terminal region [4].
In Gram-negative bacteria, lipoproteins are present in both inner
and outer membranes (IM and OM, respectively). The OM-localised
lipoproteins are transported through the periplasm by the well-­
characterised lipoprotein outer membrane localization (Lol) sys-
tem (reviewed in [5] and [6]). Lipoproteins can be predicted by

†
To the memory of Didier Grunwald, who passed away recently.

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_4, © Springer Science+Business Media LLC 2017

65
66 Maria Guillermina Casabona et al.

bioinformatics tools, such as DOLOP [7, 8] (see also Chapter 1 of

this book) and through their characteristic N-terminus signal
sequence composed of hydrophobic and uncharged residues called
a lipobox, which has a consensus (V/L)XXC motif. The lipobox is
both the lipidation site and the maturation site recognised by the
lipoprotein signal peptidase II, which cleaves the signal peptide just
upstream of the conserved cysteine [7, 8].
The Pseudomonas aeruginosa envelope is a dynamic multi-layer
structure that harbours many multi-protein assemblies, such as
extracellular appendages and secretion systems, crucial for bacterial
survival in different environments, bacterial pathogenesis and bac-
terial adaptation. In P. aeruginosa there are 175 predicted lipopro-
teins, most of them of unknown function and mostly predicted to
localise to the OM [9]. A recent study using shotgun proteomics
also identified many lipoproteins as being anchored to the IM [10].
The literature contains many examples of lipoproteins that are
essential in the assembly and function of bacterial secretion systems
(SSs). For example, in the Type IV SS (T4SS), VirB7 is a 5.5 kDa
lipoprotein that is anchored to the OM and exposed to periplasmic
space [11, 12]. VirB7 is essential for the stability of other Vir pro-
teins during the assembly of functional T-complex transport
machinery in Agrobacterium tumefaciens through dimerisation
with VirB9 [12, 13].
Another example is a family of lipoproteins, the so-called pilo-
tins, found in both T2SS and T3SS machineries [14]. The pilotin
ExsB is an OM lipoprotein of the T3SS of P. aeruginosa that plays
an important role in secretion as well as in virulence in vivo by
stabilising the β-barrel porin-secretin in the OM [15, 16]. ExsB
counterparts in Salmonella and Yersinia strains also display a
decrease in T3 protein secretion and T3SS-associated virulence.
In the T2SS, it has been shown that pilotin lipoproteins have
chaperone-like properties since they can protect proteins from pro-
teolytic degradation. Their correct localisation and insertion are
needed for survival under stress conditions [17] and for correct
functioning of the SS [18].
More recently, two lipoproteins were described as being cru-
cial in the T6SS-1 of P. aeruginosa: TssJ1 [19] and TagQ [20].
TssJ1 is conserved throughout the T6SS and is localised to the
OM, where it interacts with the TssM-L complex embedded within
the IM and stabilises the membrane-bound complex essential for
secretion [10, 21–23]. TagQ, on the other hand, is unique to P.
aeruginosa T6SS-1 and is also vital for competing activity of T6SS-­
1-­expressing strains. TagQ, together with three other envelope-­
localised proteins, TagS, TagT and TagR, plays a role in
post-translational regulation of T6SS-1 activity. It has been pro-
posed that the Tag module senses the signal that triggers the
assembly and activation of the T6SS-1. Moreover, it has been
shown that correct localisation and anchoring to the OM of TagQ
 Lipoprotein Localisation 67

are indispensable for the localisation of the periplasmic protein

TagR, which directly promotes signal transduction through a
phosphorylation pathway [20, 24] and modulates the activity of
the T6S machinery. Therefore, central to understanding the role of
lipoproteins in a bacterial SS is the determination of their localisa-
tion in the cell envelope and interplay with other components of
the given secretion apparatus.
In the past, mostly biochemical approaches such as cell frac-
tionation have been used to determine the localisation of these
proteins in cell membranes [18, 20]. In recent years, fluorescence
microscopy using superfolder green fluorescent protein (sfGFP)
and mCherry has been widely developed for monitoring the behav-
iour of bacterial proteins in vivo [25–28]. Here we describe a
method for defining P. aeruginosa lipoprotein localisation in vivo
by constructing a fusion protein between the protein of interest
and mCherry and monitoring its localisation by means of confocal
microscopy. In Gram-negative bacteria, the localisation of the lipo-
protein can then be studied by creating spheroplasts by means of
lysozyme treatment. As shown in Fig. 1, three different possibili-
ties, such as IM, periplasm and OM, will be distinguishable under
a confocal microscope analysis.

2  Materials

All solutions must be prepared using ultrapure water (prepared

by purifying deionised water to attain a sensitivity of 18 MΩ cm
at 25 °C). Prepare and store all reagents at room temperature
unless indicated otherwise. Waste disposal regulations should be
followed carefully when disposing of waste materials.

2.1  Preparation 1. P. aeruginosa strain producing cytoplasmic GFP (see Note 1).

of Bacteria 2. A plasmid producing a fusion of mCherry to the target protein
of interest (see Note 2).
3. 1% (w/v) agarose in phosphate saline buffer (PBS), sterile.
4. Luria–Bertani (LB) broth (Becton, Dickinson and Co.,
Franklin Lakes, NJ). Autoclave. Supplement with appropriate
antibiotics.
5. Inducer for gene expression: 20% (w/v) arabinose stock solu-
tion. Sterilise by filtration. Store at 4 °C.
6. Incubator shaker.
7. Slides and coverslips.
8. Benchtop centrifuge.

2.2  Preparation 1. 1% (w/v) agarose in PBS, sterile.

of Spheroplasts 2. LB broth, sterile, supplemented with appropriate antibiotics.
68 Maria Guillermina Casabona et al.

Fig. 1 Outer membrane localisation of T6SS signalling lipoprotein TagQ assessed by confocal microscopy. (a)
P. aeruginosa PAO1 expressing cytoplasmic GFP and TagQ-mCherry were imaged directly from exponentially
grown cultures (upper panel) or after treatment with lysozyme (lower panel). Note the red labelling in the
periphery of bacteria and in partially detached OM in spheroplast preparations. Inserts represent a plot
 Lipoprotein Localisation 69

3. TM buffer: 10 mM Tris-acetate, pH 8.2, 200 mM MgSO4.

Sterilise by filtration. Store at 4°C.
4. TSM buffer: 50 mM Tris-acetate, pH 8.2, 8% (w/v) sucrose,
10 mM MgSO4. Sterilise by filtration. Store at 4°C.
5. Lysozyme: 20 mg/mL and 2 mg/mL stock solutions. Sterilise
by filtration. Store at −20°C. Avoid freezing and thawing of
sample.
6. Incubator shaker.
7. Slides and coverslips.
8. Benchtop centrifuge.

2.3  Imaging 1.
Confocal microscope: TCS-SP2 (Leica Microsystems,
Mannheim, Germany) using a DMRE (Leica) upright micro-
scope. For confocal acquisitions, fluorescence is collected
through a spectral detection mode using mechanical filtering.
The different channels are acquired sequentially to avoid fluo-
rescence leakage.
2. Software for image processing like FiJi free software [29].

3  Methods

The following protocols have been developed for the analysis of

lipoproteins of P. aeruginosa, such as TagQ. The preparation of
spheroplasts is adapted from a study on sorting signals of lipopro-
teins in the P. aeruginosa strain PAO1 [28], in combination with
our previous expertise in fluorescent confocal imaging [30]. These
studies are often further validated by biochemical approaches, such
as separation of bacterial membranes by centrifugation on sucrose
gradients followed by immunodetection [10, 18, 20] (see also
Chapter 6).

3.1  Preparation of 1. Start an overnight culture of freshly transformed GFP-­

Bacteria for Confocal expressing P. aeruginosa producing the mCherry-tagged
Microscopy protein of interest in 3 mL LB broth supplemented with
appropriate antibiotics for plasmid maintenance at 37 °C with
shaking at 300 rpm/min.

Fig. 1 (continued) profile of the fluorescence intensity on selected objects obtained by FiJi free software. (b) P.
aeruginosa PAO1 expressing cytoplasmic GFP and the TagQ-mCherry protein lacking cysteine residue within
lipobox motif. The growth conditions, spheroplast preparation, imaging and image treatment were done as for
the wild-type TagQ, represented in a. Note the periphery localisation of TagQΔC-mCherry in bacteria and
absence of labelling in spheroplasts, indicating the loss of the protein during lysozyme treatment. (c) Schema
showing putative localisation of bacterial lipoproteins assessed by analysing mCherry protein fusions by fluo-
rescent microscopy before and after lysozyme treatment. Three options are represented: IM, OM and periplas-
mic localisation
70 Maria Guillermina Casabona et al.

2. Use the overnight culture to inoculate 3 mL LB broth supple-

mented with antibiotics to an OD600 of 0.15.
3. Culture up to mid-log phase of growth at 37 °C with shaking
and induce for 2 h by adding arabinose to a final concentration
to be determined for each protein (0.01–0.25%) (see Note 3).
4. Harvest 1 mL of culture by a centrifugation step at 6000 × g
for 5 min at room temperature.
5. Wash cells with 100 μL fresh LB broth without antibiotics.
6. Depose 5 μL bacteria on the centre of a clean glass slide.
7. Mix by pipetting gently with 5 μL 1% warm agarose (do not
exceed 60°C; see Note 4) and immediately cover the sample
using a coverslip and apply gentle and uniform pressure to
have a monolayer of bacteria. The sample is now ready to be
analysed under the microscope.

3.2  Preparation 1. Follow steps 1–3 in Subheading 3.1.

of Spheroplasts 2. Harvest 1 mL culture by a centrifugation step at 6000 × g for
5 min at room temperature.
3. Resuspend the bacterial pellet in 30 μL ice-cold TM buffer.
4. Add lysozyme to a final concentration of 500 μg/mL (see Note 3).
5. Incubate 30 min at room temperature without shaking.
6. Centrifuge at 1000 × g for 5 min at room temperature.
7. Discard supernatant and gently resuspend spheroplasts in 50
μL ice-cold TSM buffer and let stand on ice. Be careful not to
shake the preparation.
8. To analyse the spheroplasts under the microscope, gently
depose 5 μL on a clean slide, add 2 μL warm 1% agarose, and
cover with a coverslip. At this step, avoid pressing hard on the
coverslip, which may lead to an explosion of spheroplasts.
Observe images shortly after the mounting of the slide
(see Notes 4 and 5).

3.3  Imaging 1. Start the microscope and let it warm up.

and Image Analysis 2. Observe the specimens with an oil immersion objective (×63,
NA: 1.4), and analyse them by confocal laser scanning micros-
copy (CLSM) using a TCS-SP2 operating system (Leica)
working with an upright microscope.
3. Adjust the pinhole to Airy 1 (around 550 nm optical
sections).
4. Excite GFP and mCherry fluorescence and collect sequentially
(400 Hz line by line) using 488 nm for GFP and 543 nm for
mCherry excitation. Collect fluorescence emissions between
 Lipoprotein Localisation 71

500 and 535 nm for GFP and from 575 to 650 nm for mCherry
(see Note 6). Images (512 × 512 pixels) are obtained with vari-
ous electronic zooms (×8, ×32). The optical sections are
acquired at a focal position corresponding to a median position
in the analysed objects.
5. To determine the relative position for green and red fluores-
cence, display a plot profile of the fluorescence intensity along
the line in the images. Those histograms can be obtained using
FiJi free software (see Fig. 1).

4  Notes

1. To generate P. aeruginosa expressing cytoplasmic GFP, consti-

tutive promotor driving the gfp expression pX2-gfp [31] is
transferred to pmini-CTX1 using EcoRI-HindIII digestion
[20]. pmini-CTX1 is an integrative plasmid that carries an oriT
for conjugation-mediated plasmid transfer [32].
2. To generate fusions to mCherry, the gene encoding mCherry
is amplified and cloned into XbaI and SacI sites of pJN105
[33], giving pJN-mCherry. Then the DNA fragment contain-
ing the ribosome binding site and the gene of interest without
the stop codon is amplified by polymerase chain reaction and
cloned upstream of mCherry using the EcoRI and XbaI sites
[20]. Site-­directed mutations within the lipobox sequence are
created using the QuikChangeII Site-Directed Mutagenesis
kit. pJN105-derived plasmids are introduced in P. aeruginosa
by transformation [34].
3. As stated in the previous section, the methods presented here
were developed and optimised for P. aeruginosa. Note that dif-
ferent concentrations of inducer and lysozyme might be neces-
sary for other bacteria, and this will have to be adjusted
accordingly. Notably, if the concentration of inducer is too
high, there might be aggregation of the fusion protein, and
this can be spotted when imaging: the protein will form clus-
ters mostly in the pole of the bacterial cell. In some cases, the
growth rate of bacteria overexpressing the lipoprotein-
mCherry fusion protein could be greatly reduced due to the
toxic effect of the fusion protein. Concentration of lysozyme
below optimal conditions (to be determined experimentally
for each bacterium) might result only in partial destruction of
peptidoglycan and consequently in a heterogeneous popula-
tion of objects, leading to incomplete OM detachment  (con-
centrations tested from 75 to 500 μg/mL; data not shown).
Magnesium and sucrose also affect spheroplast stability and
must be present during imaging.
72 Maria Guillermina Casabona et al.

4. The agarose for fixing bacteria needs to be melted using an

incubator at 60–65 °C with shaking. Avoid boiling the solu-
tion as this will result in a more concentrated agarose solution.
Do not reuse this solution more than twice, and make sure to
keep it sterile. A 1% agarose solution cannot be too warm since
it causes fragility of the OM, which will result in leaking of the
green cytoplasmic fluorescence (red ghost bacteria).
5. In the same way, spheroplasts are fragile, and to avoid explo-
sion with warm agarose, microscope slide preparation must be
done immediately before observations. To avoid drying of the
spheroplast sample: first add pre-warmed 1% agarose and then
samples. A small volume of agarose (2 μL) is added with 5 μL
of spheroplast sample to preserve their delicate shape. If the
bacteria or spheroplasts are left on the slide for too long, they
might dry out and get damaged, which will affect fluorescence
detection and image quality. Cover with the coverslip immedi-
ately to avoid solidification of agarose. Apply gentle and uni-
form pressure to obtain a monolayer of bacteria. Use a tissue
to avoid marks on the coverslip.
6. It is recommended, while adjusting the focus and looking for
an interesting field with the oculars, to use GFP fluorescence
or the transmitted light. Indeed, mCherry is especially photo-
sensitive, and the intensity of the mercury arc lamp (HBO-
50W) at this wavelength excitation, which is largely more
phototoxic than the laser excitation used for the confocal
acquisition, can induce an important photobleaching.

Acknowledgements

We thank Dr. K. M. Sall for initiating studies on TagQ and TssJ1
fusion proteins and Dr. S. Elsen for help in plasmid generation.
MGC was supported by a PhD grant from the French Cystic
Fibrosis Association Vaincre la Mucovisidose. The microscopy
facility is supported by the Biosciences and Biotechnology Institute
of Grenoble (BIG), CEA-Grenoble and the grant to Laboratoire
of Excellence, LabEx GRAL (ANR-10-LABX-49-01).

References

1. Farris C, Sanowar S, Bader MW, Pfuetzner tein RcsF, a new substrate for the periplasmic
R, Miller SI (2010) Antimicrobial peptides protein-disulfide isomerase DsbC. J Biol Chem
activate the Rcs regulon through the outer 286(19):16734–16742
membrane lipoprotein RcsF. J Bacteriol 3. Aliprantis AO, Yang RB, Mark MR, Suggett
192(19):4894–4903 S, Devaux B, Radolf JD et al (1999) Cell
2. Leverrier P, Declercq JP, Denoncin K, activation and apoptosis by bacterial lipopro-
Vertommen D, Hiniker A, Cho SH et al (2011) teins through toll-like receptor-2. Science
Crystal structure of the outer membrane pro- 285(5428):736–739
 Lipoprotein Localisation 73

4. Pugsley AP (1993) The complete general for correct assembly of the Pseudomonas
secretory pathway in gram-negative bacteria. aeruginosa type III secretion apparatus in the
Microbiol Rev 57(1):50–108 bacterial membrane and full virulence in vivo.
5. Zuckert WR (2014) Secretion of bacterial lipo- Infect Immun 83(5):1789–1798
proteins: through the cytoplasmic membrane, 17. Guilvout I, Chami M, Engel A, Pugsley AP,
the periplasm and beyond. Biochim Biophys Bayan N (2006) Bacterial outer membrane
Acta 1843(8):1509–1516 secretin PulD assembles and inserts into the
6. Konovalova A, Silhavy TJ (2015) Outer mem- inner membrane in the absence of its pilotin.
brane lipoprotein biogenesis: Lol is not the EMBO J 25(22):5241–5249
end. Philos Trans R Soc Lond Ser B Biol Sci 18. Viarre V, Cascales E, Ball G, Michel GP,
370(1679) Filloux A, Voulhoux R (2009) HxcQ lipose-
7. Babu MM, Priya ML, Selvan AT, Madera M, cretin is self-piloted to the outer membrane
Gough J, Aravind L et al (2006) A database by its N-terminal lipid anchor. J Biol Chem
of bacterial lipoproteins (DOLOP) with func- 284(49):33815–33823
tional assignments to predicted lipoproteins. 19. Aschtgen MS, Bernard CS, De Bentzmann S,
J Bacteriol 188(8):2761–2773 Lloubes R, Cascales E (2008) SciN is an outer
8. Madan Babu M, Sankaran K (2002) membrane lipoprotein required for type VI
DOLOP--database of bacterial lipoproteins. secretion in enteroaggregative Escherichia coli.
Bioinformatics 18(4):641–643 J Bacteriol 190(22):7523–7531
9. Remans K, Vercammen K, Bodilis J, Cornelis 20. Casabona MG, Silverman JM, Sall KM, Boyer
P (2010) Genome-wide analysis and literature-­ F, Coute Y, Poirel J et al (2013) An ABC trans-
based survey of lipoproteins in Pseudomonas porter and an outer membrane lipoprotein par-
aeruginosa. Microbiology 156(Pt 9):2597–2607 ticipate in posttranslational activation of type
10. Casabona MG, Vandenbrouck Y, Attree I, VI secretion in Pseudomonas Aeruginosa.
Coute Y (2013) Proteomic characterization of Environ Microbiol 15(2):471–486
Pseudomonas aeruginosa PAO1 inner mem- 21. Durand E, Nguyen VS, Zoued A, Logger
brane. Proteomics 13(16):2419–2423 L, Pehau-Arnaudet G, Aschtgen MS et al
11. Fernandez D, Dang TA, Spudich GM, Zhou (2015) Biogenesis and structure of a type VI
XR, Berger BR, Christie PJ (1996) The secretion membrane core complex. Nature
Agrobacterium tumefaciens virB7 gene prod- 523(7562):555–560
uct, a proposed component of the T-complex 22. Felisberto-Rodrigues C, Durand E, Aschtgen
transport apparatus, is a membrane-associated MS, Blangy S, Ortiz-Lombardia M, Douzi B
lipoprotein exposed at the periplasmic surface. et al (2011) Towards a structural comprehen-
J Bacteriol 178(11):3156–3167 sion of bacterial type VI secretion systems:
12. Fernandez D, Spudich GM, Zhou XR, Christie characterization of the TssJ-TssM complex
PJ (1996) The Agrobacterium tumefaciens of an Escherichia coli pathovar. PLoS Pathog
VirB7 lipoprotein is required for stabiliza- 7(11):e1002386
tion of VirB proteins during assembly of the 23. Rao VA, Shepherd SM, English G, Coulthurst
T-complex transport apparatus. J Bacteriol SJ, Hunter WN (2011) The structure of
178(11):3168–3176 Serratia marcescens Lip, a membrane-bound
13. Christie PJ, Cascales E (2005) Structural component of the type VI secretion system.
and dynamic properties of bacterial type IV Acta Crystallogr Sect D 67(Pt 12):1065–1072
secretion systems (review). Mol Membr Biol 24. Basler M, Ho BT, Mekalanos JJ (2013) Tit-­
22(1–2):51–61 for-­
tat: type VI secretion system counterat-
14. Collin S, Guilvout I, Nickerson NN, Pugsley tack during bacterial cell-cell interactions. Cell
AP (2011) Sorting of an integral outer mem- 152(4):884–894
brane protein via the lipoprotein-specific Lol 25. Alcock F, Baker MA, Greene NP, Palmer T,
pathway and a dedicated lipoprotein pilotin. Wallace MI, Berks BC (2013) Live cell imag-
Mol Microbiol 80(3):655–665 ing shows reversible assembly of the TatA
15. Izore T, Perdu C, Job V, Attree I, Faudry component of the twin-arginine protein
E, Dessen A (2011) Structural character- transport system. Proc Natl Acad Sci U S A
ization and membrane localization of ExsB 110(38):E3650–E3659
from the type III secretion system (T3SS) 26. Guillon L, El Mecherki M, Altenburger S,
of Pseudomonas aeruginosa. J Mol Biol Graumann PL, Schalk IJ (2012) High cel-
413(1):236–246 lular organization of pyoverdine biosynthe-
16. Perdu C, Huber P, Bouillot S, Blocker A, sis in Pseudomonas aeruginosa: clustering of
Elsen S, Attree I et al (2015) ExsB is required PvdA at the old cell pole. Environ Microbiol
14(8):1982–1994
74 Maria Guillermina Casabona et al.

27. Imperi F, Visca P (2013) Subcellular localiza- 31. Thibault J, Faudry E, Ebel C, Attree I, Elsen
tion of the pyoverdine biogenesis machinery S (2009) Anti-activator ExsD forms a 1:1
of Pseudomonas aeruginosa: a membrane-­ complex with ExsA to inhibit transcription
associated “siderosome”. FEBS Lett of type III secretion operons. J Biol Chem
587(21):3387–3391 284(23):15762–15770
28. Lewenza S, Mhlanga MM, Pugsley AP 32. Hoang TT, Kutchma AJ, Becher A, Schweizer
(2008) Novel inner membrane retention sig- HP (2000) Integration-proficient plasmids for
nals in Pseudomonas aeruginosa lipoproteins. Pseudomonas aeruginosa: site-specific integra-
J Bacteriol 190(18):6119–6125 tion and use for engineering of reporter and
29. Schindelin J, Arganda-Carreras I, Frise E expression strains. Plasmid 43(1):59–72
et al (2012) Fiji: an open-source platform 33. Newman JR, Fuqua C (1999) Broad-host-­
for biological-­
image analysis. Nat Methods range expression vectors that carry the
9(7):676–682 L-arabinose-inducible Escherichia coli araBAD
30. De Bentzmann S, Giraud C, Bernard CS, promoter and the araC regulator. Gene
Calderon V, Ewald F, Plesiat P et al (2012) 227(2):197–203
Unique biofilm signature, drug susceptibility 34. Chuanchuen R, Narasaki CT, Schweizer HP
and decreased virulence in Drosophila through (2002) Benchtop and microcentrifuge prepa-
the Pseudomonas aeruginosa two-component ration of Pseudomonas aeruginosa competent
system PprAB. PLoS Pathog 8(11):e1003052 cells. Biotechniques 33(4):760. 2–3
 Chapter 5

Identification of Lipoproteins Using Globomycin

and Radioactive Palmitate
Nienke Buddelmeijer

Abstract
Bacterial lipoproteins are characterized by fatty acids that are covalently attached to their amino terminus
via posttranslational modification in the cytoplasmic membrane. Three enzymatic steps are involved in the
synthesis of mature triacylated lipoprotein: prolipoprotein converts into diacylglyceryl-prolipoprotein that
in turn converts into apolipoprotein, which is finally converted into mature triacylated lipoprotein. Here
we describe the detection of one of these intermediate forms of lipoprotein, diacylglyceryl-prolipoprotein,
using 3H-palmitate labeling and inhibition by globomycin and detection by fluorography.

Key words Tris-tricine gel electrophoresis, Globomycin, 3H-palmitate labeling, Fluorography

1  Introduction

Lipoproteins are abundant proteins of the bacterial cell envelope

[1]. They are modified at their amino terminus by fatty acids that
are derived from membrane phospholipids [2]. The mature part of
lipoproteins is highly variable, resulting in a wide variety of biologi-
cal functions. In proteobacteria and actinobacteria, the modifica-
tion pathway is composed of two acyltransferases and one peptidase
[3]. Lipoproteins are synthesized in the cytoplasm as prolipopro-
teins that are inserted into the membrane via The general secretion
(Sec) or twin-arginine translocation (Tat) machineries. The enzyme
phosphatidylglycerol:prolipoprotein diacylglyceryl transferase
(Lgt) adds a diacylglyceryl moiety from phosphatidylglycerol onto
the invariant cysteine in the so-called lipobox located in the amino
terminal region of the protein, resulting in the formation of
diacylglyceryl-­prolipoprotein. In the second step, lipoprotein signal
peptidase (Lsp) cleaves the signal peptide from diacylglyceryl-­
prolipoprotein, resulting in a free α-amino group on diacylglyceryl-
cysteine. The third and last step in lipoprotein m ­ odification is

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_5, © Springer Science+Business Media LLC 2017

75
76 Nienke Buddelmeijer

N-acylation catalyzed by apolipoprotein N-acyltransferase (Lnt).

The acyl donor is phosphatidylethanolamine, of which the sn-1
acyl group is transferred onto the α-amino group of apolipopro-
tein, resulting in mature triacylated protein. The intermediate
forms of small lipoproteins or peptides can be analyzed using gel
electrophoresis techniques in combination with specific inhibitors
or mutant strains. Globomycin and its derivatives specifically inhibit
Lsp leading to an accumulation of diacylglyceryl-prolipoprotein in
the cytoplasmic membrane [4]. This intermediate carries a diacyl-
glyceryl moiety that in case of Escherichia coli is composed of C16:0
and C18:cis-11 fatty acids and still has the signal peptide attached
[5, 6]. The migration of this form of lipoprotein on a sodium
dodecyl sulfate (SDS)-polyacrylamide gel is slower than prolipo-
protein, apolipoprotein, and mature lipoprotein.

2  Materials

2.1  3H-Palmitate 1. Luria Broth Miller (LB) medium: 5 g yeast extract, 10 g peptone,
Labeling of E. coli 10 g NaCl per liter in ultrapure water. Sterilize by autoclaving
Cultures for 20 min at 120 °C.
2. E. coli strains (see Note 1).
3. 5 mCi/mL of 9,10-3H(N)-palmitate in ethanol. Specific activity:
30–60 Ci(1,11–2,22 TBq)/mmol.
4. Incubator.

2.2  Inhibition of Lsp 1. Luria Broth Miller (LB) medium: 5 g yeast extract, 10 g peptone,
by Globomycin 10 g NaCl per liter in ultrapure water. Sterilize by autoclaving
for 20 min at 120 °C.
2. E. coli strains (see Note 1).
3. Globomycin: use final concentration of 160 μg/mL for E. coli
K12 strains and 5 μg/mL for B strains (see Note 2).
4. Incubator.

2.3  Immune 1. Antibodies specific to lipoprotein of interest (see Note 3).

Precipitation 2. 100% tricloroacetic acid (TCA).
of Lipoproteins
3. Acetone.
4. Solubilization buffer: 25 mM Tris–HCl, pH 8.0, 1% SDS, 1 mM
ethylenediaminetetraacetic acid (EDTA).
5. Immune precipitation buffer: 50 mM Tris–HCl, pH 8.0, 150
mM NaCl, 1 mM EDTA, 2% Triton X-100.
6. Protein G-Sepharose.
7. Wash buffer I: 50 mM Tris–HCl, pH 8.0, 1 M NaCl, 1%
Triton-X-100.
 Lipoprotein Labeling 77

8. Wash buffer II: 10 mM Tris–HCl pH 8.0.

9. Tabletop microcentrifuge.

2.4  Tris-Tricine Gel 1. Mini-gel caster system and SDS-polyacrylamide gel electro-
Electrophoresis phoresis (PAGE) migration apparatus.
2. Cathode buffer (top, 10×): 1 M Tris, 1 M Tricine, 1% SDS pH
8.25. Do not adjust pH.
3. Anode buffer (bottom, 10×): 1 M Tris–HCl, pH 8.9.
4. Gel buffer (3×): 3 M Tris, 1 M HCl, 0.3% SDS.
5. Acrylamide (49.5%T, 3%C or 6%C). Store at 4 °C.
6. Ammonium persulfate (APS): 10% solution in water. Store at
−20 °C.
7. N, N, N, N′-tetramethyl-ethylenediamine (TEMED). Store
at 4 °C.
8. SDS-PAGE running buffer: 25 mM Tris, 250 mM glycine,
0.1% SDS. Do not adjust pH.
9. SDS loading buffer (3×): 150 mM Tris–HCl, pH 6.8, 6% SDS,
0.3% bromophenol blue, 30% glycerol.
10. Water bath at 100 °C.
11. i-Butanol.
12. Vacuum gel-drying system.
13. Amplify solution.
14. X-ray film.

3  Methods

3.1  3H-Palmitate 1. Grow cultures of E. coli in LB Miller medium at 37 °C.

Labeling of E. coli 2. Add 100 μCi/mL 3H-palmitate to the cell culture at early
Cultures exponential phase (OD600 of 0.2), and let cultures grow for
2 h (see Note 4).

3.2  Inhibition of Lsp Add globomycin to the cell culture after 1 h of growth in the pres-
by Globomycin ence of 3H-palmitate and let cultures grow for an additional 1 h
(see Note 5).

3.3  Immune This protocol is based on ref. [7].

Precipitation
1. Add a final concentration of 10% TCA to 1 mL of cell culture
of Lipoproteins
to precipitate all proteins.
2. Centrifuge-precipitated proteins for 1 min in a tabletop
centrifuge.
3. Wash the pellets twice with 1 mL ice-cold (−20 °C) acetone.
78 Nienke Buddelmeijer

4. Briefly dry protein pellets.

5. Resuspend pellets in 50 μL of solubilization buffer and boil
samples for 2 min. Let cool down.
6. Add 450 μL immune precipitation buffer (see Note 6).
7. Centrifuge samples for 10 min in a tabletop centrifuge.
8. Take 200 μL from the top and add 300 μL immune precipita-
tion buffer.
9. Add antibodies and incubate on ice overnight.
10. Add 100 μL Protein G-Sepharose slush and incubate on ice
for 20 min.
11. Centrifuge 1 min at 4 °C in tabletop centrifuge and wash pel-
let twice with Wash Buffer I (see Note 7).
12. Wash once with Wash Buffer II.
13. Resuspend slush in 100 μL SDS loading buffer and boil for
2 min to release proteins from Protein G-Sepharose.
14. Centrifuge samples for 5 min in tabletop centrifuge and use
supernatant for gel electrophoresis (see Note 8).

3.4  Tris-Tricine Gel Tris-Tricine gels are especially useful for separating small proteins
Electrophoresis and peptides (less than 30 kDa) [8].
1. Prepare a mini-gel format separating gel (16%) by mixing
5 mL gel buffer, 6 mL acrylamide solution, and 4 mL water
(total volume 15 mL). Add 5 μL TEMED and 50 μL APS and
cast gel in an 8.6 × 6.8 × 0.75 cm gel holder. Allow space for
stacking gel and overlay with water or i-butanol.
2. Prepare stacking gel (4%) by mixing 3.3 mL gel buffer, 1 mL
acrylamide solution, and 5.7 mL water (total volume 10 mL).
Add 7.5 μL TEMED and 75 μL APS and cast gel and insert
comb immediately.
3.
Load samples on gel along with protein standard.
Electrophorese at 30 V until the samples have entered the
stacking gel and continue at 200 V till the dye front reaches
the bottom of the gel (see Note 9).
4. After migration of samples, pry gel plates open and briefly
wash gel in water.
5. Transfer to amplify solution and let soak for 10 min under
agitation.
6. Dry gel in gel dryer under vacuum for 60 min at 80 °C
(see Note 10).
7. Transfer the gel into a cassette and put an X-ray film over it.
Expose the film for 10 days at −80 °C. Let the cassette warm
up to room temperature before developing the X-ray film
(Fig. 1).
 Lipoprotein Labeling 79

Fig. 1 Accumulation of diacyl-prolipoprotein (diacyl-PLP) in envelope of E. coli

B cells labeled with 3H palmitate upon treatment with globomycin (5 μg/mL). Without
globomycin (lanes 1 and 3), with globomycin (lanes 2 and 4), anti-Lpp immunoprecipi-
tates in lane 3 (sample from lane 1) and lane 4 (sample from lane 2). Figure adapted
from Fig. 5 of [10] corresponding to lanes 7 through 10, with permission

4  Notes

1. Various wild-type strains of E. coli can be used.

2. Globomycin is commercially produced and derivatives have
been described [9]. The concentration of these derivatives
needs to be determined empirically.
3. Braun’s lipoprotein (Lpp) of E. coli (78 amino acids) has been
the reference lipoprotein to study modification and cellular
localization. Antibodies against other bacterial lipoproteins
have been used. It is recommended to use proteins of small
size or peptide fragments (10 kDa) to facilitate identification
of intermediate forms of lipoprotein modification.
4. For E. coli K12 strains such as MC4100 or MG1655, a final
OD600 of 0.6–0.8 is obtained.
5. Extensive exposure to globomycin leads to cell lysis owing to
the inhibition of essential enzyme Lsp and, as a result, the
accumulation of Lpp in the cytoplasmic membrane while still
being cross-linked to the peptidoglycan.
6. Triton X-100 solubilizes lipoproteins from membranes.
7. Pellet is antigen–antibody–sepharose slush.
8. The amount of sample to charge needs to be determined
empirically and depends on the antigen and antibody used.
9. Electrophoresis of 16% Tris-Tricine gels takes longer than
regular SDS-PAGE gels. Calculate 2–3 h for a mini-gel
format.
10. Remove vacuum from gel before switching off pump to avoid
cracking of gel.
80 Nienke Buddelmeijer

References
1. Kovacs-Simon A, Titball RW, Michell SL 6. Hantke K, Braun V (1973) Covalent binding
(2010) Lipoproteins of bacterial pathogens. of lipid to protein. Diglyceride and amide-­
Infect Immun 79:548–561 linked fatty acid at the N-terminal end of the
2. Lai J-S, Philbrick WM, Wu HC (1980) Acyl murein-lipoprotein of the Escherichia coli outer
moieties in phospholipids are the precursors membrane. Eur J Biochem 34:284–296
for the fatty acids in murein lipoprotein in 7. Kumamoto CA, Gannon PM (1988) Effects of
Escherichia coli. J Biol Chem 255: Escherichia coli secB mutations on pre-maltose
5384–5387 binding protein conformation and export
3. Buddelmeijer N (2015) The molecular mecha- kinetics. J Biol Chem 263:11554–11558
nism of bacterial lipoprotein ­modification--how, 8. Schagger H (2006) Tricine-SDS-PAGE. Nat
when and why? FEMS Microbiol Rev Protoc 1:16–22
39:246–261 9. Kiho T et al (2004) Structure-activity relation-
4. Inukai M, Takeuchi K, Shimizu K, Arai M ships of globomycin analogues as antibiotics.
(1978) Mechanism of action of globomycin. Bioorg Med Chem 12:337–361
J Antibiot 31:1203–1205 10. Hussain M, Ichihara S, Mizushima S (1980)
5. Cronan JE Jr, Rock CO (1996) Biosynthesis of Accumulation of glyceride-containing precursor
membrane lipids. In: Neidhardt FC (ed) of the outer membrane lipoprotein in the cyto-
Escherichia coli and Salmonella: molecular and plasmic membrane of Escherichia coli treated with
cellular biology. ASM, Washington, DC globomycin. J Biol Chem 255:3707–3712
 Chapter 6

Defining Membrane Protein Localization by Isopycnic

Density Gradients
Rhys A. Dunstan, Iain D. Hay, and Trevor Lithgow

Abstract
In many bacteria, membrane proteins account for around one-third of the proteome and can represent
much more than half of the mass of a membrane. Classic techniques in cell biology can be applied to char-
acterise bacterial membranes and their membrane protein constituents. Here we describe a protocol for
the purification of outer and inner membranes from Escherichia coli. The procedure can be applied with
minor modifications to other bacterial species, including those carrying capsular polysaccharide attached
to the outer membrane.

Key words Sucrose density gradient, Membrane biogenesis, Beta-barrel proteins, Cytoplasmic mem-
brane, Lipoproteins

1  Introduction

Gram-negative bacteria are characterised by two membranes. The

cytoplasmic (inner) membrane is a phospholipid bilayer into which
alpha-helical transmembrane proteins are integrated and onto
which peripheral membrane proteins are attached by lipid-­mediated
or protein–protein interactions [1, 2].
The lipid phase of the outer membrane is a bilayer with an
outer leaflet predominantly composed of lipopolysaccharides and
an inner leaflet of phospholipids [3, 4]. This, together with a mas-
sively high ratio of protein:lipid [5, 6], gives the outer membrane
a far greater buoyant density than the inner membrane. In 1975
Yamato et al. reported a reproducible procedure to disrupt
Escherichia coli cells in a French press, recover a relatively pure
membrane fraction, and segregate it into inner and outer mem-
brane fractions by sucrose isopycnic gradient ultracentrifugation
[7]. This methodology was validated using marker enzyme assays:
measuring partial reactions of oxidative phosphorylation for the
inner membrane and finding that phospholipase A activity
segregated into a higher-density fraction. These fractions were
­

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_6, © Springer Science+Business Media LLC 2017

81
82 Rhys A. Dunstan et al.

therefore deemed to represent the inner and outer membranes,

and electron microscopy was used to assess their relative morphol-
ogy and the homogeneity of the membrane preparations [7].
We now have a mature knowledge of the protein and lipid
composition of the inner and outer membranes for a range of bac-
terial species. By fractionating sucrose gradients and analysing the
fractions by SDS-PAGE and immunoblotting with antibodies, it is
possible to follow the co-purification of specific proteins of interest
with either membrane. Proteomics has been applied to assess and
validate the purity of membrane fractions, so that the method has
been applied to membrane purification from bacterial species
including Campylobacter [8], Caulobacter [9], Citrobacter [10],
Cornebacterium [11], Neisseria [12], Proteus [13], Pseudomonas
[14] and Salmonella [15]. The method has been applied to dem-
onstrate effects on the outer membrane protein composition when
the beta-barrel assembly machinery is diminished [16], when LPS
biosynthesis is diminished [17], or to dissect the assembly path-
ways for beta-barrel and alpha-helical membrane proteins into the
outer membrane [18].
Here we detail an optimised protocol for the purification of
outer and inner membrane fractions from E. coli.

2  Materials

2.1  Membrane 1. Lysogeny broth (LB) medium: 10 g tryptone, 5 g yeast extract

Purification and 5 g NaCl, make up to 1 L with distilled water. Autoclave
and store at room temperature.
2. 1 M Tris stock solution: Add 121.1 g Tris to 800 mL distilled
water. Adjust pH to 7.5 with HCl, then dilute to 1 L with
distilled water. Autoclave and store at room temperature.
3. 500 mM ethylenediaminetetraacetic acid (EDTA) stock solu-
tion: Add 186.1 g EDTA (disodium ethylenediamine tetraac-
etate, 2H2O) to 800 mL distilled water. Adjust pH to 8.0 with
NaOH, then dilute to 1 L with distilled water. Autoclave and
store at room temperature (see Note 1).
4. Ultra-pure sucrose.
5. Tris buffer: 10 mM Tris–HCl, pH 7.5 (dilute 10 mL 1 M Tris
stock solution to 1 L with distilled water).
6. 100 mg/mL lysozyme: Add 1 g lysozyme to 10 mL distilled
water, prepare 500 μL aliquots, and store at −20 °C.
7. 100 mM phenylmethylsulfonyl fluoride (PMSF): Dissolve
174 mg PMSF in 10 mL isopropanol. Prepare 500 μL aliquots
and store at −20 °C (see Note 2).
8. EDTA buffer: 1.65 mM EDTA, pH 7.5 (dilute 660 μL EDTA
stock solution to 200 mL of distilled water). Store at room
temperature.
 Membrane Isolation on Sucrose Gradients 83

9. Tris-Sucrose (TS) buffer: 10 mM Tris–HCl, pH 7.5, 0.75 M

sucrose (dissolve 51.3 g sucrose in 200 mL Tris buffer). Store
at 4 °C.
10. TES buffer: 3.3 mM Tris–HCl, pH 7.5, 1.1 mM EDTA,
0.25 M sucrose (add 1 volume TS buffer to 2 volumes 1.65
mM EDTA buffer). Approximately 40 mL for each sample will
be required. Store at 4 °C.
11. 5 mM EDTA, pH 7.5 (dilute 5 mL EDTA stock solution to
500 mL with distilled water).
12. 25% sucrose solution: 25% (w/w) sucrose, 5 mM EDTA, pH
7.5 (add 7.5 g sucrose to 22.5 mL of EDTA). Pass through a
0.45 μM filter and store at 4 °C.
13. Use Emulsiflex (AVESTIN, Ottawa, ON, Canada) or other
cell disruptor or similar.
14. Centrifuge with Sorvall SS34 (Thermo Fisher Scientific, Inc.,
Waltham, MA) tubes and rotor (or something similar with
ability to spin up to approximately 50 mL at up to 15,000 × g).
15. Ultracentrifuge with Beckman 70.1 Ti tubes (Beckman Coulter
Inc., Brea, CA) (open-top thick-wall polycarbonate tubes) and
rotor (or something similar with ability to spin approximately
100,000 × g).
16. Wheaton teflon tissue grinder/dounce (Millville, NJ).

2.2  Sucrose Density 1. 5 mM EDTA, pH 7.5 (see earlier steps). Store at room
Fractionation temperature.
2. Ultra-pure sucrose.
3. Sucrose EDTA fractions: 35–60% (w/w) sucrose, 5 mM
EDTA, pH 7.5 solutions. For example, to make 50% (w/w),
add 15 g sucrose to 15 mL 5 mM EDTA. Pass through a 0.45
μM filter and store at 4 °C.
4. Displacing sucrose solution: 70% (w/w) sucrose, 5 mM EDTA,
pH 7.5 (add 140 g sucrose to 60 mL EDTA) (see Note 3).
5. Beckman Coulter SW 40 Ti tubes (disposable plastic tubes)
and rotor.
6. Density Gradient Fractionation System (Teledyne Isco, Lincoln,
NE, USA)

2.3  Isolation of 1. TES buffer (see earlier steps).

Membranes after 2. 25% (w/w) sucrose 5 mM EDTA, pH 7.5 solution (see
Density Fractionation earlier).
3. Beckman 70.1 Ti tubes (open-top thick-wall polycarbonate
tubes) and rotor.
84 Rhys A. Dunstan et al.

3  Methods

3.1  Membrane 1. Grow 5 mL O/N starter culture in LB at 37 °C from a single

Purification colony.
2. Dilute culture 1:100 in 400 mL LB with antibiotics as required
and grow until OD600 = ~1.0 (see Note 4).
3. Pellet cells by centrifugation for 5 min at 5000 × g 4 °C (see
Note 5).
4. Resuspend cells in 10 mM Tris–HCl, pH 7.5 (~200 mL).
5. Repeat centrifugation and resuspend pellet in 10 mL TS.
6. Add 50 μg/mL lysozyme (5 μL stock) and 2 mM PMSF (200
μL stock) to break down peptidoglycan layer and inhibit host
serine proteases respectively.
7. Slowly add 2 volumes (20 mL) 1.65 mM EDTA, pH 7.5, to
destabilise outer membrane for lysis.
8. Incubate 10 min on ice.
9. Lyse cells using an AVESTIN Emulsiflex; 2–3 passes at ~15,000
psi will be required to fully lyse cells.
10. Centrifuge cell lysate at 15,000 × g, 20 min at 4 °C, to remove
the cell debris.
11. Collect the supernatant and pellet total membranes by ultra-
centrifugation at 38,000 revolutions per minute (rpm) (132,000
× g), 45 min, 4 °C (70.1 Ti rotor—use ~8 mL in each tube).
12. Resuspend membrane pellets in 1 mL TES using a dounce.
13. Pool membranes and make up to ~8 mL with TES and centri-
fuge 38,000 rpm, 45 min, 4 °C.
14. Resuspend membrane pellet in a small volume (~400 μL) of
25% sucrose in 5 mM EDTA, pH 7.5, using a dounce, and
either snap freeze in liquid nitrogen for storage at −80 °C or
continue to sucrose density fractionation.

3.2  Sucrose Density 1. Immediately before use carefully prepare a six-step sucrose gra-
Fractionation dient from 60–35% w/w (1.9 mL of 60%, 55%, 50%, 45%,
40%, 35%) in SW40 tubes. A sharp interphase between layers
should be clearly visible (see Note 6).
2. Layer 400 μL of total membranes on top of a 60–35% w/w
gradient.
3. Spin in an ultracentrifuge using SW40 rotor for 17 h. at 34,000
rpm (205,000 × g), 4 °C (see Note 7).
4. Isolate 1 mL fractions (use ISCO fractionator with 70%
sucrose, 5 mM EDTA, pH 7.5 as displacing fluid, or carefully
pipette 1 mL fractions off top of gradient). Store each fraction
at −80 °C until use (see Note 8).
 Membrane Isolation on Sucrose Gradients 85

5. To visualise on SDS-PAGE with coomassie brilliant blue

staining, load 30 μL of each fraction, or ~15 μL, for analysis
by western immunoblotting with antibodies against known
inner and outer membrane protein (e.g., F1β and BamA
respectively).

3.3  Isolation 1. To isolate membranes from specific fractions, add TES buffer
of Membranes (to a final volume of ~8 mL) to each fraction of interest or to
after Density pooled fractions and pellet by ultracentrifugation, 1.5 h, 4 °C,
Fractionation 38,000 rpm (70 Ti .1 rotor).
2. Resuspend each fraction in ~100 μL 25% (w/w) sucrose, 5
mM EDTA, pH 7.5, with a dounce, and store membranes at
−80 °C (see Note 9).

4  Notes

1. EDTA will not be soluble until pH reaches 8.0. Use vigorous

stirring and heat (if needed).
2. PMSF may crystallise in solution, vortex thoroughly before use.
3. Vigorous stirring and heat may be required to fully dissolve
sucrose.
4. Protocol can be adjusted for optimal conditions for protein
expression or conditional shutdowns, for example, if needed.
5. Everything (buffers, tubes, etc.) should be at 4 °C from now on.
6. Carefully pipette the sucrose gradient solutions to the edge of
the tube as close to the meniscus layer as possible. This will
prevent disruption between sucrose fractions’ interface.
7. Owing to the different nature of the outer membrane between
bacterial strains (e.g., capsulated, non-capsulated), the dura-
tion and speed of centrifugation may need to be adjusted. For
example, when performing sucrose gradients on total mem-
branes from the capsulated Klebsiella pneumonia we routinely
centrifuge for 40 h at 33,300 rpm (196,000 × g).
8. When collecting by drops using the fractionator, approximately
20 drops is equivalent to 1 mL. Alternatively, fractions can be
isolated by carefully piercing the bottom of the tube and allow-
ing the fraction to drip out the bottom; or, if visible, individual
layers can be directly isolated from the tube by piercing the
side of the tube with a syringe and sucking the corresponding
layers out. Depending on the amount of protein added to the
gradient, the denser outer membrane should predominate in
the bottom section of the gradient and may be present as a
white band; the lighter inner membrane will be present in the
top section of the gradient and will generally be more diffuse
and may have a reddish appearance.
86 Rhys A. Dunstan et al.

9. The volume of TES used will vary with the amount of mem-
branes isolated or desired membrane concentration for later
experiments.

References

1. Dalbey RE, Wang P, Kuhn A (2011) Assembly D, Lithgow T (2012) Discovery of an arche-
of bacterial inner membrane proteins. Annu typal protein transport system in bacte-
Rev Biochem 80:161–187 rial outer membranes. Nat Struct Mol Biol
2. Okuda S, Tokuda H (2011) Lipoprotein 19(5):506–510
sorting in bacteria. Annu Rev Microbiol 11. Marchand CH, Salmeron C, Bou Raad R,
65:239–259 Meniche X, Chami M, Masi M, Blanot D,
3. Kamio Y, Nikaido H (1976) Outer mem- Daffe M, Tropis M, Huc E, Le Marechal P,
brane of salmonella typhimurium: acces- Decottignies P, Bayan N (2012) Biochemical
sibility of phospholipid head groups to disclosure of the mycolate outer membrane
phospholipase c and cyanogen bromide of Corynebacterium glutamicum. J Bacteriol
activated dextran in the external medium. 194(3):587–597
Biochemistry 15(12):2561–2570 12. Masson L, Holbein BE (1983) Physiology of
4. Smit J, Kamio Y, Nikaido H (1975) Outer sialic acid capsular polysaccharide synthesis in
membrane of salmonella typhimurium: serogroup B Neisseria meningitidis. J Bacteriol
chemical analysis and freeze-fracture studies 154(2):728–736
with lipopolysaccharide mutants. J Bacteriol 13. Siegmund-Schultze N, Kroll HP, Martin
124(2):942–958 HH, Nixdorff K (1991) Composition of the
5. Osborn MJ, Gander JE, Parisi E, Carson outer membrane of Proteus mirabilis in rela-
J (1972) Mechanism of assembly of the tion to serum sensitivity in progressive stages
outer membrane of salmonella typhimurium. of cell form defectiveness. J Gen Microbiol
Isolation and characterization of cytoplas- 137(12):2753–2759
mic and outer membrane. J Biol Chem 14. Jagannadham MV, Abou-Eladab EF, Kulkarni
247(12):3962–3972 HM (2011) Identification of outer mem-
6. Schnaitman CA (1970) Protein composition brane proteins from an Antarctic bacte-
of the cell wall and cytoplasmic membrane of rium Pseudomonas syringae Lz4W. Mol Cell
Escherichia coli. J Bacteriol 104(2):890–901 Proteomics 10(6):M110 004549
7. Yamato I, Anraku Y, Hirosawa K (1975) 15. Bishop RE, Gibbons HS, Guina T, Trent MS,
Cytoplasmic membrane vesicles of Escherichia Miller SI, Raetz CR (2000) Transfer of pal-
coli. A simple method for preparing the cyto- mitate from phospholipids to lipid a in outer
plasmic and outer membranes. J Biochem membranes of gram-negative bacteria. EMBO
77(4):705–718 J 19(19):5071–5080
8. Hobb RI, Fields JA, Burns CM, Thompson 16. Charlson ES, Werner JN, Misra R (2006)
SA (2009) Evaluation of procedures for outer Differential effects of yfgL mutation on
membrane isolation from Campylobacter Escherichia coli outer membrane pro-
jejuni. Microbiology 155(Pt 3):979–988 teins and lipopolysaccharide. J Bacteriol
9. Anwari K, Webb CT, Poggio S, Perry AJ, 188(20):7186–7194
Belousoff M, Celik N, Ramm G, Lovering 17. Steeghs L, de Cock H, Evers E, Zomer B,
A, Sockett RE, Smit J, Jacobs-Wagner C, Tommassen J, van der Ley P (2001) Outer
Lithgow T (2012) The evolution of new membrane composition of a lipopolysaccharide-­
lipoprotein subunits of the bacterial outer deficient Neisseria meningitidis mutant.
membrane BAM complex. Mol Microbiol EMBO J 20(24):6937–6945
84(5):832–844 18. Dunstan RA, Hay ID, Wilksch JJ, Schittenhelm

10. Selkrig J, Mosbahi K, Webb CT, Belousoff RB, Purcell AW, Clark J, Costin A, Ramm G,
MJ, Perry AJ, Wells TJ, Morris F, Leyton Strugnell RA, Lithgow T (2015) Assembly of
DL, Totsika M, Phan MD, Celik N, Kelly M, the secretion pores GspD, Wza and CsgG into
Oates C, Hartland EL, Robins-Browne RM, bacterial outer membranes does not require
Ramarathinam SH, Purcell AW, Schembri the Omp85 proteins BamA or TamA. Mol
MA, Strugnell RA, Henderson IR, Walker Microbiol 97(4):616–629
 Chapter 7

Cell Surface Exposure

Anna Konovalova

Abstract
Surface-exposed proteins of Gram-negative bacteria are represented by integral outer membrane beta-­
barrel proteins and lipoproteins. No computational methods exist for predicting surface-exposed lipopro-
teins, and therefore lipoprotein topology must be experimentally tested. This chapter describes three
distinct but complementary methods for the detection of surface-exposed proteins: cell surface protein
labeling, accessibility to extracellular protease and antibodies.

Key words Biotinylation, PEGylation, Surface proteolysis, Whole-cell dot blot, Protein topology

1  Introduction

Cells of Gram-negative bacteria are surrounded by an additional

membrane, known as the outer membrane (OM) [1]. The outer
surface of the OM is composed of lipopolysaccharide (LPS) and
decorated with proteins. Until recently, it was thought that only
integral beta-barrel proteins (referred to as outer membrane pro-
teins or OMPs) were surface exposed as they often display long
extracellular loops. In contrast, OM lipoproteins, peripheral pro-
teins tethered to the OM by N-terminal lipids, were thought to be
found exclusively in the inner leaflet of the OM facing the aqueous
periplasm [2]. However, in recent years a number of surface-­
exposed lipoproteins that face the cell exterior instead of the peri-
plasm have been identified (see for reviews [3–5]).
In the case of OMPs, topology and extracellular loops can be
easily predicted computationally based on the presence of hydro-
phobic ß strands [6–8]. In contrast, lipoproteins are a very diverse
group of proteins that do not share sequence or structure similari-
ties. Many surface-exposed lipoproteins have no obvious trans-
membrane domains and are assembled on the cell surface by novel
mechanisms [3–5]. Therefore, lipoprotein surface exposure or

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_7, © Springer Science+Business Media LLC 2017

87
88 Anna Konovalova

topology in the OM cannot be predicted and must be experimen-

tally tested.
Several methods have been developed for the detection of
surface-­exposed proteins. All of them are based either on the avail-
ability of a functional group for protein modification/labeling or
protein accessibility to extracellular proteases or antibodies.
Protein modification utilizes reagents that are able to react effi-
ciently with certain functional groups and form a covalent bond
[9]. Reagents with N-hydroxysuccinimide (NHS)-esters are often
the first choice. They target primary amines (available in side chains
of lysine residues or the N-terminus of the protein if not modified).
Lysines are relatively abundant within protein sequences and often
surface exposed and, thus, accessible in protein structures. The
other commonly used reagents are based on maleimides, which
react with the cysteine sulfhydryl groups. Cysteines are not com-
monly found in protein sequences, a fact that can be taken advan-
tage of using genetically introduced cysteine codons to study
detailed protein topology (see Chapter 8 for more information).
Protein labeling reagents vary greatly in their properties, such
as size, hydrophobicity, and detection methods (see Table 1 for
examples). Reagents for selective surface labeling should be cell
impermeable. When working with Gram-negative bacteria, the
unique permeability properties of the OM must be taken into con-
sideration [10]. The OM is an asymmetric membrane with phos-
pholipids in the inner leaflet and LPS in the outer leaflet. LPS is
negatively charged and bridged with divalent cations, such as Mg2+
and Ca2+ (see Note 1). These lateral interactions seal the OM and
make it impermeable to hydrophobic compounds. On the other
hand, the OM also contains protein channels that allow diffusion
of nutrients and small molecules. Therefore, small hydrophilic
reagents can enter the periplasm and label proteins on both sides of
the membrane. Because of these properties of the OM, the criteria
for the selection of reagents for cell surface labeling are often the
opposite of those given in the product instruction manuals that
were most often developed for the labeling of eukaryotic cells.

Table 1
Protein labeling reagents for selective surface labeling

Functional group to
be labeled Name Polar Molecular weight, Da
Primary amine NHS-LC-LC-­Biotin − 567.70
NHS-PEG(n)-Biotin + Available in 1–10
kDa range
Sulfhydryl Mal-PEG(n)-Biotin + Available in 1–10
kDa range
 Cell Surface Exposure 89

In our lab we have used hydrophobic NHS-LC-LC-biotin to

selectively label surface-exposed lipoproteins in Escherichia coli
[11, 12]. In addition, hydrophilic reagents that are significantly
larger than the diffusion limit (600 Da for E. coli [13]) will also
preferentially label cell surfaces. The OM in other bacteria may
have different properties, so reagents should be validated every
time for their cell surface selectivity. For example, not all Gram-­
negative bacteria will have such a highly asymmetric OM and hence
are not as resistant to hydrophobic compounds as E. coli. The sen-
sitivity to detergents serves as a good indication of the presence of
phospholipids in the outer leaflet. If this is the case, using hydro-
phobic reagents is not recommended. This rule also applies to
E. coli mutants with defects in OM biogenesis and maintenance. In
addition, when the permeability properties of the OM are
unknown, it is better to use high-molecular-weight reagents to
avoid the generation of false positive results.
Protein labeling reagents allow the detection of modified pro-
teins by the addition of a biotin group, a long-chain polyethylene
glycol (PEG) linker, or a combination of these. Protein biotinyl-
ation allows for immunoblot detection using anti-biotin antibodies
or streptavidin conjugates. However, because all the cell surface
proteins are labeled, the specific protein of interest must often be
purified (or at least enriched) prior to detection. Alternatively, bio-
tinylated proteins can be affinity purified using a streptavidin resin
and then probed with protein-specific antibodies. Working with
high-molecular-weight PEG linkers provides an advantage for the
direct detection of the labeled protein in a cell lysate based on the
size shift during immunoblot analysis using protein-specific
antibodies.
Protein accessibility to extracellular proteases, also known as
surface proteolysis, is another common method of studying pro-
tein surface exposure [14–16]. It is based on the addition of pro-
teases with broad specificity, such as trypsin or proteinase K, to
intact cells. Because proteases cannot enter cells, only surface-­
accessible proteins or domains of proteins will be cleaved.
Antibodies that recognize the protein of interest are then used to
detect cleavage using immunoblotting.
Proteins can be inherently protease resistant because of their
tight folding or lack of protease cleavage sites or because they are
protected by interactions with other proteins. Therefore, negative
results of protease shaving experiments are hard to interpret. One
way to address this problem is to test whether the protein of inter-
est is protease sensitive in the cell lysate done under nondenaturing
conditions, for example, using a mild detergent lysis solution (see
Note 2).
It is crucial to use proper controls to ensure that proteases do
not target periplasmic proteins for the following reasons. First,
proteins are important components of the OM and contribute
90 Anna Konovalova

significantly to its stability. Complete proteolysis of surface domains

can destabilize the OM, an event made apparent by the degrada-
tion of periplasmic proteins. For this reason, titration experiments
are advisable to find the optimal protease concentration. Second,
both trypsin and proteinase K retain their activity in sodium
dodecyl sulfate (SDS) [17, 18] and therefore they can digest pro-
teins in cell lysates during preparation of the samples for immu-
noblotting. This also leads to the generation of false positives. To
avoid this, protease inhibitors should be added and excess protease
removed prior to cell lysis in SDS loading buffer.
A number of assays that utilize antibodies added extracellularly
to whole cells are used to study protein surface exposure [19–21].
These include dot blots, whole-cell enzyme-linked immunosor-
bent assay (ELISA), immunofluorescence, and flow-cytometry.
Just like proteases, antibodies cannot enter intact cells and, hence,
will bind to only surface-accessible epitopes.
Antibodies to be used in these assays should satisfy two require-
ments. First, they should be able to recognize the native protein.
Many antibodies that bind to denatured protein during immunob-
lot procedures cannot bind to native proteins because binding epi-
topes are hidden in the protein structure. This is especially
important when antibodies were raised against denatured proteins
or a peptide. Second, antibodies should be polyclonal. For exam-
ple, if a transmembrane protein contains a surface-exposed and
periplasmic domain, then the use of monoclonal antibodies that
recognize the periplasmic domain will lead to false negatives.
However, experiments with polyclonal and monoclonal (or
epitope-­specific) antibodies can provide valuable insights into pro-
tein topology [11, 22, 23]. A protocol for a dot blot assay is
described in what follows. This assay is easy and inexpensive and
requires no dedicated equipment.
Each of the aforementioned methods has its own advantages
and disadvantages, and ideally, a combination of methods should
be used. One of the common limitations is that the ability to detect
proteins on cell surfaces depends not only on protein localization
but also on sequence and structure because these features deter-
mine the presence and accessibility of groups for labeling, protease
sites, or epitopes for antibody binding. In addition, surface pro-
teins can be physically occluded from detection by interactions
with other proteins, hidden within long sugar chains of LPS, an S
layer, or extracellular matrix. On the other hand, these methods
can also generate false positives. For example, surface labeling and
proteolysis can destabilize the OM, giving the reagent and prote-
ases access to the periplasm. Many immunodetection techniques
require cell fixation, which can also lead to a number of artifacts
[24]. Therefore, it is very important to incorporate careful controls
for OM integrity when designing these experiments.
 Cell Surface Exposure 91

As a general recommendation, proteins with known periplas-

mic topology should be used as negative controls. These include
either soluble periplasmic proteins or lipoproteins with experimen-
tally verified periplasmic localization. Ideally, such proteins should
not be a part of a bigger protein complex and should be readily
detectable with specific antibodies. If such protein-specific anti-
bodies are not available, it is possible to use heterologously
expressed proteins as a controls, for example periplasmically local-
ized fluorescent proteins (such as mCherry or superfolding GPF
[25]), maltose binding protein, glutathione S-transferase, or other
proteins for which antibodies are commercially available. However,
it is important to remember that protein overexpression can also
lead to aberrant results. Therefore, when using such heterologous
proteins, one should verify that they do not have a negative impact
on cell growth and OM permeability. In addition, a variant of a
protein of interest that would localize it to different compartments
(e.g., by swapping signal sequences) can also serve as a negative
control.

2  Materials

2.1  Cell Surface 1. NHS reagents (see Table 1 for product information). NHS
Labeling Based reagents are moisture sensitive. Store desiccated at 4 °C, and
on Modification equilibrate to room temperature (RT) before opening. Prepare
of Primary Amines 25 mM stock solution according to product instructions
immediately before use.
2. Labeling buffer containing no primary amines, such as phos-
phate buffered saline (PBS): 10 mM Na2HPO4, 1.8 mM
KH2PO4, 2.7 mM KCl, 137 mM NaCl, pH 8.0.
3. Quenching solution: 1 M glycine or 1 M Tris–HCl, pH 8.0.

2.2  Cell Surface 1. Maleimide reagents (see Table 1 for product information).
Labeling Based Maleimide reagents are moisture sensitive. Store desiccated at
on Modification 4 °C, and equilibrate to RT before opening. Prepare 25 mM
of Sulfhydryls stock solution according to product instructions immediately
before use.
2. Labeling buffer containing no sulfhydryls, such as Tris-buffered
saline (TBS): 50 mM Tris–HCl, 150 mM NaCl, pH 7.0 or
PBS: 10 mM Na2HPO4, 1.8 mM KH2PO4, 2.7 mM KCl, 137
mM NaCl, pH 7.0.
3. Tris(2-carboxyethyl)phosphine (TCEP) solution: 500 mM
(optional).

2.3  Cell Surface 1. Proteinase K solution: 20 mg/mL.

Proteolysis 2. Reaction buffer, TBS: 50 mM Tris–HCl, 150 mM NaCl,
5 mM CaCl2, pH 8.0.
92 Anna Konovalova

3. Phenylmethylsulfonyl fluoride (PMSF): 500 mM in ethanol.

4. SDS loading buffer 1×: 50 mM Tris–HCl, pH 6.8, 2% SDS,
10% glycerol, 0.002% bromophenol blue.

2.4  Whole-Cell Dot 1. Nitrocellulose membrane.

Blot Assay 2. PBS: 10 mM Na2HPO4, 1.8 mM KH2PO4, 2.7 mM KCl, 137
mM NaCl, pH 7.0.
3. EDTA: 0.5 M.
4. Blocking buffer: PBS with 2% nonfat dried milk.
5. Antibodies for detection of protein of interest as well as a nega-
tive control.
6. Appropriate horseradish peroxidase (HRP)-conjugated sec-
ondary antibodies.
7. Chemiluminescent substrate for HRP detection.

3  Methods

3.1  Cell Surface This protocol is valid for any NHS-based reagent.
Labeling Based
on Modification 1. Collect exponentially growing cells by centrifugation (see Note 1).
of Primary Amines 2. Wash cells three times with ice-cold PBS to remove amine-­
containing culture medium.
3. Resuspend cells to 1010 cells/mL.
4. Add NHS reagent to a final concentration of 2.5 mM.
5. Incubate at RT for 30 min.
6. Add one-tenth of the volume of quenching solution.
7. Collect the cells by centrifugation.
8. Wash cells twice with PBS supplemented with 100 mM glycine
or directly in 100 mM Tris–HCl to quench and remove excess
reagent.
9. Analyze by immunoblotting or follow with protein purification
if needed (see Note 2).

3.2  Cell Surface This protocol is valid for any maleimide-based reagent.
Labeling Based
on Modification 1.
Collect exponentially growing cells by centrifugation
of Sulfhydryls (see Note 1).
2. Wash cells three times with ice-cold TBS or PBS.
3. Resuspend cells to 1010 cells/mL.
4. (Optional) If protein contains oxidized (disulfide bonded) cys-
teines, treat cells with 5 mM TCEP in TBS or PBS, pH 7.0, for
30 min at RT. Wash cells twice with TBS or PBS to remove
excess TCEP (see Note 3).
 Cell Surface Exposure 93

5. Add reagent to a final concentration of 2.5 mM.

6. Incubate at RT for 30 min.
7. Collect cells by centrifugation.
8. Wash cells twice with PBS or TBS to remove excess reagent.
9. Analyze by immunoblotting or follow with protein purification
if needed (see Note 2).

3.3  Cell Surface 1. Prepare 2× dilutions of proteinase K in range of 20–1.25 mg/

Proteolysis mL in the reaction buffer.
2.
Collect exponentially growing cells by centrifugation
(see Note 1).
3. Resuspend cells to 1010 cells/mL in reaction buffer. Use 90 μL
cell suspension for each reaction.
4. Add 10 μL corresponding proteinase K solution or 10 μL reac-
tion buffer (untreated control). Incubate at RT for 30 min.
5. Preheat SDS loading buffer in a 96 °C thermoblock or boiling
water bath.
6. Add 1 μL PMSF stock solution to inactivate proteinase K.
7. Collect cells by centrifugation and wash twice with reaction
buffer supplemented with 5 mM PMSF to remove excess pro-
teinase K.
8. Resuspend cells in 100 μL preheated SDS loading buffer. Boil
immediately for at least 10 min.
9. Analyze by immunoblotting (see Note 2).

3.4  Whole-Cell Dot 1.

Collect exponentially growing cells by centrifugation
Blot Assay (see Note 1).
2. Resuspend cells to 109 cells/mL in PBS. Split into two tubes.
3. Add EDTA to final concentration of 10 mM to one of the
tubes and sonicate on ice four times for 30 s to prepare cell
lysate (see Note 4).
4. Spot 2 μL of cell suspension or cell lysate on a nitrocellulose
membrane and air-dry (approx. 5 min).
5. Place membrane in blocking solution. Incubate with gentle
shaking for 30 min at RT.
6. Add appropriate amount of primary antibodies (see Note 5).
Incubate with gentle shaking for 1 h at RT.
7. Wash membrane five times for 3 min with PBS.
8. Add a blocking buffer containing secondary antibodies.
Incubate with gentle shaking for 1 h at RT.
9. Wash membrane five times for 3 min with PBS.
10. Use chemiluminescent substrate and develop according to
standard immunoblot procedure.
94 Anna Konovalova

4  Notes

1. Using a culture medium supplemented with cations helps to

reinforce the OM and prevent permeability. If using media with
low cation concentration (e.g., lysogeny broth (LB)), add 10
mM MgSO4 and 5 mM CaCl2. In addition, cations can be
added to the reaction buffers of any procedures described with-
out interference.
2. If negative results are obtained, analyzing the protein accessibil-
ity for labeling or protease cleavage might be necessary. To pre-
pare gentle cell lysate, add BugBuster® 10× protein extraction
reagent (EMD Millipore) to the cell suspension. Unlike the
original BugBuster, this reagent does not add salts or buffer
components and can be used with labeling/protease assays
without interference.
3. β-mercaptoethanol and dithiothreitol (DTT) contain sulfhydryl
groups and are incompatible with maleimide labeling. TCEP
does not contain sulfhydryl groups and therefore can be used to
reduce disulfide bonds prior to labeling.
4. Using detergents for cell lysis (such as BugBuster reagent) will
interfere with protein binding to nitrocellulose membrane.
Prepare lysate by sonication. Adding EDTA helps to disperse
LPS and form membrane vesicles with mixed orientation.
Sometimes it is necessary to readjust the concentration of the
primary antibodies for a dot blot assay. As a general recommen-
dation, start by using a concentration three times higher than
that used for an immunoblot procedure.

References

1. Silhavy TJ, Kahne D, Walker S (2010) The 6. Freeman TC Jr, Wimley WC (2010) A highly
bacterial cell envelope. Cold Spring Harb accurate statistical approach for the prediction
Perspect Biol 2(5):a000414 of transmembrane beta-barrels. Bioinformatics
2. Okuda S, Tokuda H (2011) Lipoprotein sort- 26(16):1965–1974
ing in bacteria. Annu Rev Microbiol 7. Singh NK, Goodman A, Walter P, Helms V,
65:239–259 Hayat S (2011) TMBHMM: a frequency pro-
3. Zuckert WR (2014) Secretion of bacterial lipo- file based HMM for predicting the topology
proteins: through the cytoplasmic membrane, of transmembrane beta barrel proteins and
the periplasm and beyond. Biochim Biophys the exposure status of transmembrane resi-
Acta 1843(8):1509–1516 dues. Biochim Biophys Acta
4. Konovalova A, Silhavy TJ (2015) Outer mem- 1814(5):664–670
brane lipoprotein biogenesis: lol is not the end. 8. Hayat S, Elofsson A (2012) BOCTOPUS:
Philos Trans R Soc Lond Ser B Biol Sci improved topology prediction of transmem-
370(1679) brane beta barrel proteins. Bioinformatics
5. Wilson MM, Bernstein HD (2015) Surface-­ 28(4):516–522
exposed lipoproteins: an emerging secretion 9. Hermanson GT (2013) Bioconjugate tech-
phenomenon in gram-negative bacteria. Trends niques, 3rd edn. Academic press, London,
Microbiol pp 1–1146
 Cell Surface Exposure 95

10. Nikaido H (2003) Molecular basis of bacterial 18. Hilz H, Wiegers U, Adamietz P (1975)
outer membrane permeability revisited. Stimulation of proteinase K action by denatur-
Microbiol Mol Biol Rev 67(4):593–656 ing agents: application to the isolation of
11. Konovalova A, Perlman DH, Cowles CE, nucleic acids and the degradation of 'masked'
Silhavy TJ (2014) Transmembrane domain of proteins. Eur J Biochem 56(1):103–108
surface-exposed outer membrane lipoprotein 19. Pinne M, Haake D (2011) Immuno-­
RcsF is threaded through the lumen of fluorescence assay of leptospiral surface-­exposed
­beta-­barrel proteins. Proc Natl Acad Sci U S A proteins. J Vis Exp 53
111(41):E4350–E4358 20. Blom K, Lundin BS, Bolin I, Svennerholm A
12. Cowles CE, Li Y, Semmelhack MF, Cristea IM, (2001) Flow cytometric analysis of the localiza-
Silhavy TJ (2011) The free and bound forms of tion of helicobacter pylori antigens during dif-
Lpp occupy distinct subcellular locations in ferent growth phases. FEMS Immunol Med
Escherichia coli. Mol Microbiol Microbiol 30(3):173–179
79(5):1168–1181 21. Matsunaga J, Werneid K, Zuerner RL, Frank
13. Rosenbusch JP (1990) Structural and func- A, Haake DA (2006) LipL46 is a novel surface-­
tional properties of porin channels in E. coli exposed lipoprotein expressed during lepto-
outer membranes. Experientia 46(2): spiral dissemination in the mammalian host.
167–173 Microbiology 152(Pt 12):3777–3786
14. Wilson MM, Anderson DE, Bernstein HD 22. Moeck GS, Bazzaz BS, Gras MF, Ravi TS,
(2015) Analysis of the outer membrane pro- Ratcliffe MJ, Coulton JW (1994) Genetic
teome and secretome of Bacteroides fragilis insertion and exposure of a reporter epitope in
reveals a multiplicity of secretion mechanisms. the ferrichrome-iron receptor of Escherichia
PLoS One 10(2):e0117732 coli K-12. J Bacteriol 176(14):4250–4259
15. Pugsley AP, Kornacker MG, Ryter A (1990) 23. Newton SM, Klebba PE, Michel V, Hofnung
Analysis of the subcellular location of pullula- M, Charbit A (1996) Topology of the mem-
nase produced by Escherichia coli carrying the brane protein LamB by epitope tagging and a
pulA gene from Klebsiella pneumoniae strain comparison with the X-ray model. J Bacteriol
UNF5023. Mol Microbiol 4(1):59–72 178(12):3447–3456
16. Pinne M, Haake DA (2009) A comprehensive 24. Schnell U, Dijk F, Sjollema KA, Giepmans BN
approach to identification of surface-exposed, (2012) Immunolabeling artifacts and the need
outer membrane-spanning proteins of for live-cell imaging. Nat Methods 9(2):
Leptospira interrogans. PLoS One 4(6):e6071 152–158
17. Porter WH, Preston JL (1975) Retention of 25. Dinh T, Bernhardt TG (2011) Using
trypsin and chymotrypsin proteolytic activity in s­
uperfolder green fluorescent protein for
sodium dodecyl sulfate solutions. Anal periplasmic protein localization studies.
­
Biochem 66(1):69–77 J Bacteriol 193(18):4984–4987
 Chapter 8

Probing Inner Membrane Protein Topology by Proteolysis

Maxence S. Vincent and Eric Cascales

Abstract
Inner membrane proteins are inserted into the membrane via α-helices. These helices do not only consti-
tute membrane anchors but may mediate specific interactions with membrane protein partners or partici-
pate in energetic processes. The number, location, and orientation of these helices is referred to as topology.
Bitopic membrane proteins that consist of a single membrane-embedded domain connecting two soluble
domains are distinguished from polytopic ones that consist of multiple membrane-spanning helices con-
nected by extramembrane domains. Defining inner membrane protein topology could be achieved by
different methods. Here we describe a protease accessibility assay that makes it possible to define topology
based on digestion profiles.

Key words Membrane protein, Inner membrane, Insertion, Topology, Transmembrane segment,
Bitopic, Polytopic, Proteolysis, Protease, Proteinase K, Carboxypeptidase Y

1  Introduction

Bacterial secretion systems are multiprotein machines that catalyze

the traffic of protein substrates across the cell envelope [1]. Most
secretion systems described so far assemble large channels com-
posed of inner and outer membrane proteins [1]. In secretion sys-
tems, inner membrane proteins are crucial for the assembly of
platforms for pilus polymerization, substrate recruitment and selec-
tion, or energetic purposes [1]. Defining inner membrane protein
topology, a term referring to the number, position, and orientation
of transmembrane helices (TMHs), is therefore an important step
to characterize these proteins. Depending on the number and posi-
tion of these TMHs, inner membrane proteins are categorized into
bitopic and polytopic proteins (Fig. 1). Bitopic membrane proteins
consist of a single membrane-embedded domain connecting two
soluble domains located in two different compartments. The TMH
of bitopic proteins could be located at the N- or C-terminus. By
contrast, polytopic membrane proteins consist of multiple TMHs
connected by extramembrane domains called loops. Bitopic proteins

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_8, © Springer Science+Business Media LLC 2017

97
98 Maxence S. Vincent and Eric Cascales

Fig. 1 Nomenclature of inner membrane proteins with selected topologies. Shown are the topologies of a
bitopic protein with an N-terminal TMH (a), a bitopic protein with a C-terminal TMH (b) and polytopic proteins
with different numbers of TMH (c). Representative examples of inner membrane (IM) proteins with these
topologies associated with bacterial secretion systems are listed. For polytopic proteins, the number of trans-
membrane segments are indicated in brackets

with N-terminal TMHs are relatively common, and this category

includes GcpC, YscD, VirB10 and PorM, subunits associated with
the Type II (T2SS), Type III (T3SS), Type IV (T4SS) and Type IX
(T9SS) secretion systems [2–5]. Bitopic proteins with C-terminus
TMHs, also called C-tail proteins, are rare. In secretion systems,
only the Type VI secretion system (T6SS) TssL protein has been
demonstrated to adopt this topology [6, 7]. Polytopic membrane
proteins are also commonly associated with secretion systems, and
this category includes the HlyB, YscU, VirB6, TssM and PorL pro-
teins that are respectively associated with the T1SS, T3SS, T4SS,
T6SS and T9SS [5, 8–12].
Inner membrane protein TMH position and orientation can
be predicted using computational methods based on hydrophobic-
ity patterns and the “positive-inside” rule (see Chapter 2). Several
approaches have also been developed to experimentally define pro-
tein topology [13, 14], including the pho-lac dual reporter system
(see Chapter 10) and the substituted cysteine accessibility method
(see Chapter 9). In this chapter, we will describe a third approach
based on the accessibility of extramembrane, soluble domains to
exogenous proteases. In addition to assessing inner membrane
topology, protease accessibility assays are also of interest for testing
the in vitro translocation of proteins [7, 15] and for testing whether
proteins are subjected to conformational changes in vivo (see
Chapter 22) [16–18].
 Protease Accessibility 99

2  Material

2.1  Cell Growth 1. Lysogeny broth (LB) or recommended medium to grow the
and Spheroplast strain of interest.
Preparation 2. TNS buffer: 20 mM Tris–HCl, pH 8.0, 100 mM NaCl, 30%
sucrose: Dissolve 0.243 g Tris(hydroxymethyl) aminometh-
ane, 0.684 g NaCl, and 30 g sucrose in sterile distilled water
(final volume of 100 mL). Adjust pH to 8.0 with 1 M HCl.
3. TN buffer: 20 mM Tris–HCl, pH 8.0, 100 mM NaCl: Dissolve
0.243 g Tris(hydroxymethyl) aminomethane and 0.684 g
NaCl in 100 mL sterile distilled water. Adjust pH to 8.0 with
1 M HCl.
4. 0.5 M ethylenediaminetetraacetic acid (EDTA), pH 8.0:
Dissolve 1.86 g EDTA (disodium salt) in 10 mL sterile dis-
tilled water. Adjust pH to 8.0 with 10 M NaOH.
5. Lysozyme stock solution (100×), 10 mg/mL lysozyme:
Dissolve 10 mg goose egg lysozyme in 1 mL sterile distilled
water. Store at −20 °C.
6. Incubator.
7. Spectrophotometer to measure bacterial density.
8. Labtop centrifuge.

2.2  Protease 1. Triton X-100 stock solution, 10% Triton X-100: Mix 1 mL 100%
Accessibility Assay Triton X-100 with 9 mL sterile distilled water (see Note 1). Store
at room temperature.
2. Carboxypeptidase Y stock solution (100×), 10 mg/mL car-
boxypeptidase Y: Dissolve 10 mg purified carboxypeptidase Y
in 1 mL sterile distilled water. Store at −20 °C.
3. Proteinase K stock solution (100×), 10 mg/mL proteinase K:
Dissolve 10 mg purified proteinase K in 1 mL sterile distilled
water. Store at −20 °C.
4. Cocktail of protease inhibitors (Complete, F. Hoffmann-La
Roche AG, Basel, Switzerland, or equivalent).
5. Phenylmethylsulfonyl fluoride (PMSF) stock solution (100×),
100 mM PMSF: Dissolve 17.4 mg PMSF in 1 mL absolute
ethanol (see Note 2). Store at −20 °C.
6. 50% trichloroacetic acid (TCA) solution: Dissolve 50 g TCA
in 30 mL distilled water. Complete to 100 mL with distilled
water (see Note 3).
7. Acetone.
8. Vortex.
100 Maxence S. Vincent and Eric Cascales

2.3  Sample Analysis 1. Sodium dodecyl sulfate (SDS)-polyacrylamide gel electropho-

by SDS-PAGE resis (PAGE) loading buffer: 60 mM Tris–HCl, pH 6.8, 2%
and Immunodetection SDS, 10% glycerol, 5% β-mercaptoethanol, 0.01% bromophe-
nol blue.
2. Water bath at 96 °C.
3. Mini-gel caster system and SDS-PAGE apparatus.
4. Protein blotting apparatus.
5. Antibodies for protein immunodetection.

3  Method

3.1  Cell Growth 1. Grow 30 mL culture in appropriate medium to allow cell

and Spheroplast growth and production of protein of interest (see Note 5).
Preparation 2. Collect cells by centrifugation at 5000 × g for 10 min at 4 °C.
(See Note 4)
3. Discard supernatant and gently resuspend cell pellet in optical
density at 600 nm (OD600) of 12 in ice-cold TNS buffer.
Incubate on ice for 10 min.
4. Add EDTA at 1 mM final concentration (see Note 6). Incubate
on ice for 5 min.
5. Add lysozyme at final concentration of 100 μg/mL and incu-
bate on ice for 15–40 min (see Note 7).
6. Dilute sample twice with ice-cold TN buffer, mix by gently
inverting the tube, and keep on ice for 10 min.
7. Collect spheroplasts by centrifugation at 10,000 × g for 5 min
at 4 °C.
8. Gently resuspend spheroplasts to an OD600 of 6 in ice-cold TN
buffer.

3.2  Protease 1. Divide cell suspension into five samples, numbered 1–5.
Accessibility Sample 1 will remain untreated.
(See Notes 8 and 9) 2. Add 1% (final concentration) of Triton X-100 in samples 3 and
5 to lyse spheroplasts (see Note 9). Mix by vortexing and incu-
bation 10 min on ice.
3. Add carboxypeptidase Y (100 μg/mL final concentration from
10 mg/mL stock solution) in tubes 2 and 3. Incubate for
30 min on ice.
4. Add proteinase K (100 μg/mL final concentration from 10
mg/mL stock solution) in tubes 4 and 5. Incubate for 30 min
on ice.
5. Quench the proteolysis reaction by adding PMSF and inhibi-
tor cocktail in tubes 1–5. Incubate for 5 min on ice.
 Protease Accessibility 101

6. Add 0.5 volume 50% TCA in tubes 1–5. Mix by vortexing and
incubate for 20 min on ice.
7. Collect precipitated material by centrifugation at 20,000 × g
for 20 min at 4 °C.
8. Discard supernatant and add 500 μL acetone. Vortex.
9. Collect precipitated proteins by centrifugation at 20,000 × g
for 20 min at 4 °C.
10. Discard supernatant and keep tubes open until pellet is dry
(see Note 10).

3.3  Sample Analysis 1. Resuspend pellet in SDS-PAGE loading buffer by throughout

by SDS-PAGE vortexing.
and Immunodetection 2. Boil samples in water bath for 5–10 min (see Note 11).
3. Proceed to SDS-PAGE and immunoblotting using your favor-
ite protocol.
A schematic example of expected results for topology mapping
using proteolysis is shown in Fig. 2.

Fig. 2 Schematic representation of expected results. The expected immunoblot results for inner membrane
proteins with the topology shown below are schematically represented. Samples 1–5 are shown (1, untreated
sample; 2, carboxypeptidase Y; 3, carboxypeptidase Y on Triton X-100-lysed spheroplasts; 4, proteinase K; 5,
proteinase K on Triton X-100-lysed spheroplasts). The representation of the protein degradation products cor-
responding to the immunodetection bands are shown on the right of each “blot”
102 Maxence S. Vincent and Eric Cascales

4  Notes

1. Triton X-100 is a detergent used to lyse cells and solubilize a

subset of membrane proteins. It is a viscous solution and there-
fore should be pipetted slowly and with care.
2. PMSF is a serine protease inhibitor with a short half-life. Owing
to its instability in solution, it is recommended to prepare fresh
solution on the spot.
3. TCA is highly irritating. Thus, it should be handled with care
(gloves, laboratory suit, and glasses).
4. For Gram-negative bacteria, spheroplasts should be prepared
to provide access of protease to periplasmic side of inner mem-
brane. For Gram-positive bacteria, grow, harvest, and resus-
pend cells as specified in steps 1–2 of Subheading 3.1 and then
proceed to step 8 of Subheading 3.1.
5. Use the appropriate medium to grow the cells. If the expres-
sion of the gene coding the protein of interest needs to be
induced, add the inducer at the appropriate concentration.
6. This concentration of EDTA is commonly used for disturbing
the lipopolysaccharide layer of the outer membrane in E. coli
cells. Other bacterial strains may need higher concentrations of
EDTA.
7. Lysozyme concentration and incubation time should be
adapted to the bacterial strain used in the assay. Efficient sphe-
roplast preparation of most Gram-negative bacteria requires
incubation on ice for 15–40 min.
8. Protease accessibility should be tested with two proteases: one
processive exopeptidase hydrolyzing from the C-terminus of
the protein (e.g., carboxypeptidase Y) and one endopeptidase
with low or broad specificity (e.g., trypsin, papain, proteinase
K). When using the calcium-dependent proteinase K, add 0.1
mM CaCl2 to the TN buffer. For simplification, this protocol
describes an assay with carboxypeptidase Y and proteinase K.
9. Appropriate controls include protease accessibility assays with
lysed spheroplasts. Spheroplasts are lysed by the addition of 1%
Triton X-100. The presence of Triton X-100 in the assay buf-
fer does not interfere with most proteases.
10. If available, the pellet could be dried using a SpeedVac (Thermo
Fisher Scientific, Inc., Waltham, MA) vacuum concentrator (or
equivalent).
11. A number of highly hydrophobic polytopic inner membrane
proteins precipitate in SDS-PAGE loading buffer when boiled.
For the first assay, keep the concentrating gel during the immu-
noblot to verify that the protein is not retained in the well.
 Protease Accessibility 103

Acknowledgements

Work in EC laboratory is supported by the Centre National de la

Recherche Scientifique, the Aix-Marseille Université, and grants
from the Agence Nationale de la Recherche (ANR-­
14-­CE14-0006-02 and ANR-15-CE11-0019-01). MSV is a recip-
ient of a doctoral fellowship from the French Ministère de
l’Enseignement Supérieur et de la Recherche.

References

1. Costa TR, Felisberto-Rodrigues C, Meir A, 10. Jakubowski SJ, Krishnamoorthy V, Cascales E,

Prevost MS, Redzej A, Trokter M, Waksman G Christie PJ (2004) Agrobacterium tumefaciens
(2015) Secretion systems in gram-negative VirB6 domains direct the ordered export of a
bacteria: structural and mechanistic insights. DNA substrate through a type IV secretion sys-
Nat Rev Microbiol 13:343–359 tem. J Mol Biol 341:961–977
2. Bleves S, Lazdunski A, Filloux A (1996) 11. Ma LS, Lin JS, Lai EM (2009) An IcmF family
Membrane topology of three Xcp proteins protein, ImpLM, is an integral inner mem-
involved in exoprotein transport by Pseudomonas brane protein interacting with ImpKL, and its
aeruginosa. J Bacteriol 178:4297–4300 walker a motif is required for type VI secretion
3. Ross JA, Plano GV (2011) A C-terminal region system-mediated Hcp secretion in agrobacte-
of Yersinia pestis YscD binds the outer mem- rium tumefaciens. J Bacteriol 191:4316–4329
brane secretin YscC. J Bacteriol 12. Logger L, Aschtgen MS, Guérin M, Cascales
193:2276–2289 E, Durand E (2016) Molecular dissection of
4. Das A, Xie YH (1998) Construction of trans- the interface between the type VI secretion
poson Tn3phoA: its application in defining the TssM cytoplasmic domain and the TssG base-
membrane topology of the Agrobacterium plate component. J Mol Biol 428:4424–4437
tumefaciens DNA transfer proteins. Mol 13. Traxler B, Boyd D, Beckwith J (1993) The
Microbiol 27:405–414  topological analysis of integral cytoplasmic
5. Vincent MS, Canestrari MJ, Leone P, membrane proteins. J Membr Biol 132:1–11
Stathopulos J, Ize B, Zoued A, Cambillau C, 14. van Geest M, Lolkema JS (2000) Membrane
Kellenberger C, Roussel A, Cascales E. (2017) topology and insertion of membrane proteins:
Characterization of the Porphyromonas gingi- search for topogenic signals. Microbiol Mol
valis Type IX Secretion trans-envelope Biol Rev 64:13–33
PorKLMNP core complex. J Biol Chem. 15. Cunningham K, Lill R, Crooke E, Rice M,
292:3252–3261.  Moore K, Wickner W, Oliver D (1989) SecA
6. Aschtgen MS, Zoued A, Lloubès R, Journet L, protein, a peripheral protein of the Escherichia
Cascales E (2012) The C-tail anchored TssL coli plasma membrane, is essential for the func-
subunit, an essential protein of the enteroag- tional binding and translocation of
gregative Escherichia coli Sci-1 type VI secre- proOmpA. EMBO J 8:955–959
tion system, is inserted by YidC. Microbiology 16. Larsen RA, Thomas MG, Postle K (1999)
1:71–82 Protonmotive force, ExbB and ligand-bound
7. Pross E, Soussoula L, Seitl I, Lupo D, Kuhn A FepA drive conformational changes in
(2016) Membrane targeting and insertion of TonB. Mol Microbiol 31:1809–1824
the C-tail protein SciP. J Mol Biol 17. Germon P, Ray MC, Vianney A, Lazzaroni JC
428:4218–4227. (2001) Energy-dependent conformational
8. Gentschev I, Goebel W (1992) Topological change in the TolA protein of Escherichia coli
and functional studies on HlyB of Escherichia involves its N-terminal domain, TolQ, and
coli. Mol Gen Genet 232:40–48 TolR. J Bacteriol 183:4110–4114
9. Allaoui A, Woestyn S, Sluiters C, Cornelis GR 18. Cascales E, Christie PJ (2004) Agrobacterium
(1994) YscU, a Yersinia enterocolitica inner VirB10, an ATP energy sensor required for
membrane protein involved in Yop secretion. type IV secretion. Proc Natl Acad Sci U S A
J Bacteriol 176:4534–4542 101:17228–17233
 Chapter 9

Mapping of Membrane Protein Topology by Substituted

Cysteine Accessibility Method (SCAM™)
Mikhail Bogdanov

Abstract
A described simple and advanced protocol for the substituted-cysteine accessibility method as applied to
transmembrane (TM) orientation (SCAM™) permits a topology analysis of proteins in their native state
and can be universally adapted to any membrane system to either systematically map an uniform topology
or identify and quantify the degree of mixed topology. In this approach, noncritical individual amino acids
that are thought to reside in the putative extracellular or intracellular loops of a membrane protein are
replaced one at a time by cysteine residue, and the orientation with respect to the membrane is evaluated
using a pair of membrane-impermeable nondetectable and detectable thiol-reactive labeling reagents.

Key words Membrane protein, Topology, Cysteine, Maleimides, SCAM™

1  Introduction

1.1  Membrane It has been estimated that the vast majority of membrane proteins
Protein Topology adopt α-helical bundle (Fig. 1a) structure, which contains trans-
and Topogenesis membrane domains (TMDs) that span the membrane in a zigzag
but not in a “one-by-one” fashion (Fig. 1b). A fundamental aspect
and primary structural element of the structure of integral mem-
brane proteins is membrane protein topology. Membrane protein
topology refers to the two-dimensional structural information of a
membrane protein and describes the way a polypeptide chain is
arranged in the membrane, i.e., the number of TMDs and their
orientation in the membrane that indicates the sidedness of extra-
membrane domains (EMDs) [1–4]. Although the topology of
membrane proteins provides low-resolution structural information,
it can be a starting point for different biochemical experiments or
modeling of three-dimensional structures. Membrane protein
topology and assembly are governed by structural principles and
topological rules and directed by topogenic signals and sequences
in the nascent polypeptide chain that are recognized and decoded

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_9, © Springer Science+Business Media LLC 2017

105
106 Mikhail Bogdanov

Fig. 1 Three-dimensional view of membrane protein topology of E. coli lactose

permease (LacY) probed by SCAM™. (a) A highly resolved α-helical bundle of
lactose permease reproduced from [44] with permission from American
Association for the Advancement of Science. (b) X marks diagnostic single cys-
teines introduced strategically one at a time to probe sidedness of this protein by
SCAM™

not only by the protein insertion and translocation machineries

(translocon) but also by the given lipid profile [4]. Thus each mem-
brane protein may contain different combinations of topogenic sig-
nals (positively and negatively charged residues) and sequences (the
number of charged flanking residues, the hydrophobicity and length
of the TMDs, EMDs with potential phosphorylation and glycosyl-
ation sites) that cooperate sequentially or differently with the trans-
locon components, the protein itself, and a given lipid profile to
finalize its membrane topology [4].

1.2  Method Topology studies offer guidance on membrane protein structure

of Choice and function. Given the enormous number of sequences available
in genome-sequencing projects, it is not realistic to assume that the
structures of all encoded proteins will be generated by crystallo-
graphic approaches, especially for membrane proteins. Moreover,
purification, crystallization, and structure determination of hydro-
phobic membrane proteins remain a challenge. Also the exact
boundaries between TMD ends and EMDs remain largely
unknown. For most highly resolved membrane proteins, the
 Determination of Membrane Protein Topology 107

hydrophobic thicknesses of TMDs does not seem to match the

lipid bilayer thickness expected or experimentally determined from
the acyl chain length of the surrounding lipids. Fortunately,
substituted-­cysteine accessibility method as applied to transmem-
brane (TM) orientation (SCAM™) can be used to define the
boundaries between membrane-embedded regions and the loop
regions exposed to the aqueous phase of membrane proteins in
their native environment, thereby supplementing high-resolution
structural information [5]. Although X-ray crystallography pro-
duces highly detailed structural information on membrane pro-
teins, crystal structures may be distorted owing to purification and
crystallization constraints. Information on interactions with other
proteins and the lipid environment are also lost during purification
since these crucial interactions are substituted by protein–deter-
gent ones. Heterologous expression in a host strain with a lipid
composition different from that of the native host can also result in
a loss of proper lipid–protein interactions, which can affect topo-
logical organization and function. Crystal structures themselves
are also static, and a structural basis for dynamic TM events is not
approachable either by crystallography or even nuclear magnetic
resonance (NMR) [6, 7]. Therefore, biochemical topological anal-
yses at low resolution can be invaluable for characterizing mem-
brane proteins, for which high-resolution structures are not yet
available, and for building a reliable mechanistic model even in the
absence of highly resolved protein structures [5, 8]. For all these
reasons and limitations, noncrystallographic approaches have been
developed and employed to determine the lower-resolution topo-
logical arrangement of TMDs in full-length membrane proteins
[5, 9–15] and understand its relationship to the mechanism of the
membrane protein insertion process or biological function [2–4,
15]. To verify predicted membrane protein topology models, the
existence of all putative TMDs and EMDs must be verified and the
hydrophilic loops must be localized to one or the other side of the
membrane. These strategies are quite varied but utilize the imper-
meability of the membrane bilayer to hydrophilic molecules, the
difference in properties between the compartments separated by
the membrane, and the incorporation into proteins of a wide vari-
ety of reporter groups whose orientation is presumed to reflect the
topology of the protein. Experimental topology mapping tech-
niques include but are not limited to cysteine scanning, glycosyl-
ation mapping, insertion of proteolytic sites, foreign antigenic
epitopes, and glycosylation motifs, or they can be as complex as
fusions of C-terminally truncated proteins to enzymatic and
­fluorescent topology reporters [5, 13, 16]. Together, these tools
document EMD residue or inserted tag accessibility and, there-
fore, the topology of a membrane protein. Reporter domains
should ideally lack intrinsic topogenic information, be readily iden-
tified, and passively and efficiently follow topogenic information
108 Mikhail Bogdanov

presented by the nascent target protein fragment. Nevertheless,

translational gene fusion approaches assume that the folding of the
C-terminally truncated proteins does not depend on C-terminal
sequence and therefore could not always faithfully assign the pre-
dicted TM topology for many polytopic membrane proteins [17].
In this context, SCAM™ has emerged as the method of choice
because of its relative simplicity, reliability, and feasibility [5]. In this
approach SCAM [18] was adapted to map and assign the TM topol-
ogy of polytopic membrane proteins (denoted by SCAM™) [19].
SCAM™ is still relatively labor intensive, but it is the most informa-
tive and least invasive topology mapping method and the most use-
ful technique thus far developed for topology studies. The method
demonstrates that reporter groups can be as simple as a single amino
acid substitution. Aside from its simplicity, the advantage of this
approach is that the topology is documented in the context of full-
length membrane protein molecules and chemical modification can
be carried out using whole cells, thereby avoiding problems related
to the conversion of cells into membrane vesicles with a uniform
orientation. This contrasts with genetic methods that infer the
topology from the disposition of reporter molecules fused to frag-
ments of the target protein and therefore completely ignore long-
range interhelical and interloop interactions. SCAM™ was also
further developed to map a uniform, dual, mixed, or unusual mem-
brane protein topology in intact cells, isolated membrane vesicles,
or liposomes using a two-step labeling protocol [10, 19–24].

1.3  Justifying The reasons for the broad application of SCAM™ are both concep-
SCAM™ Legacy tual and practical. The strategic use of SCAM™ for mapping mem-
and Advantages brane topology has made it possible to circumvent the many
limitations of alternative methods used to map a topology of inte-
gral membrane proteins:
1. Since only single amino acid changes are made, cysteine chem-
istry has the highest resolution in that water accessibility of
individual cysteine residues can be determined.
2. Cysteine has no or little preference for a particular secondary
structure. Owing to only minmal changes in the primary
sequence, structural perturbation of introduced cysteine muta-
tions is essentially absent or much milder than in other meth-
ods commonly used to determine topology.
3. Detection of engineered cysteine modifications is simple, and
analysis is done by chemical modification using a broad range
of commercially available reagents that differ in charge, size,
mass, and hydrophilicity.
4. These reagents can be used to detect protein sulfhydryl groups
with sensitivities in the femtomole range.
5. Chemical modification can be done using intact cells.
 Determination of Membrane Protein Topology 109

6. This method is capable of distinguishing the accessibility of

residues separated by only three or four residues and is useful
in the precise fine-structure mapping of TMD ends in poly-
topic membrane proteins.
7. This method is generally applicable to different membrane
systems.

1.4  Application In most cases SCAM™ provides topological information after the
of SCAM™ orientation of proteins within membranes is established [5, 19].
The application of this approach allowed either detailed mapping
or significant refinement of the topology of a variety of integral
membrane proteins, including a more accurate mapping of the
ends of TMDs of protein topology that was been established by
other methods [12–14]. However SCAM™ is not only an alterna-
tive approach to low-resolution determination of membrane pro-
tein structure; it also constitutes an attractive independent approach
to dynamic studies of membrane proteins. The dynamic aspects of
protein structure as a function of the physiological state of a cell is
best probed in whole cells or membranes. SCAM™ has been used
to monitor dynamic conformational and topological changes
accompanied by substrate binding and release during enzyme
turnover and function [25, 26]. The labeling of single-cysteine
replacements of a major component (SecG) of the SecYEG trans-
locon with 4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid
(AMS) either at rest or during adenosine triphosphate (ATP)-
dependent preprotein translocation clearly demonstrated that a
cytoplasmic region of SecG undergoes topology inversion [27].
Although the labeling patterns derived from SCAM™ assays
usually reflect a steady-state topology of a membrane protein,
semiquantitative analysis of the surface accessibility of individual
cysteines introduced into extramembrane loops can be carried out
at various stages of protein assembly. In this case, SCAM™ can be
used to provide topological information during membrane inser-
tion, folding, and assembly of proteins. Cysteine accessibility dur-
ing bacteriorhodopsin translation was monitored by pulse-chase
radiolabeling and modification by AMS to determine the order
and timing of insertion of TM segments into the membrane of
Halobacterium salinarum [28]. In this in vivo assay, the rate of
insertion of TMDs into the H. salinarum cytoplasmic membrane
was monitored by rapid modification of unique cysteines in extra-
cellular EMDs of the protein with AMS, resulting in a shift in
mobility of the protein in sodium dodecyl sulfate (SDS)-
polyacrylamide gel electrophoresis (PAGE). SCAM™ has also been
utilized to establish a packing geometry of a pilin VirB2 subunit
and its ATP-dependent in and out of membrane dynamics and
reorganization within the T-pilus and T4SS secretion channel [29].
110 Mikhail Bogdanov

The applicability of SCAM™ has been succesfully extended to

the study of TM toplogy of membrane proteins assembled in dif-
ferent lipid environments [19, 20, 30]. By combining SCAM™
with mutants of E. coliin which membrane phospholipid composi-
tion can be systematically controlled, the role of phospholipids as
determinants of membrane protein topological organization was
established [3, 4, 10]. This approach was essential to test TM pro-
tein conformation that is sensitive to lipid composition [22] and
directly monitor any conformational changes that are associated
with changes in phospholipid composition either in vivo [10, 19,
21, 31] or in vitro [23, 32]. The use of cysteine-specific probes in
combination with “lipid” mutants makes this approach a powerful
means of deriving a molecular understanding of a highly dynamic
topogenesis process from relatively static experimental data such as
endpoint topologies of membrane proteins [2–4]. The ability to
change lipid composition after the assembly of a membrane pro-
tein (by either resupplying or diluting a desired lipid) demonstrated
the potential for polytopic membrane proteins to change their
topological organization after insertion and assembly in the mem-
brane [10, 19, 21, 31]. Thus, SCAM™ became a unique technique
to establish a detailed mechanistic understanding of how lipid–pro-
tein interactions [10] and interactions with protein itself [22] con-
tribute to overall TM topogenesis [4].
Thus, SCAM™ is a powerful, well-tested, and popular tech-
nique for examining static and dynamic membrane protein
topologies.

1.5  Overview This approach is based on the introduction of cysteine residues one
and General Rationale at a time in the putative EMDs of otherwise cysteine-less mem-
of Topology Mapping brane proteins of interest followed by chemical modification with
Using SCAM™ a membrane-impermeable thiol-specific probes either before or
after compromising cell membrane integrity to determine cysteine
1.5.1  SCAM™ membrane sidedness. Accessibility in whole cells establishes extra-
cellular location, while accessibility only after cell disruption estab-
lishes intracellular location (see Notes 1 and 2). The accessibility of
EMDs flanking a TMD then establishes the orientation of the
TMD with respect to the plane of the membrane bilayer.

1.5.2  Developing SCAM™ is based on the generation of a library of independent

of a Working Topology single-cysteine mutants in which unique reactive residues are stra-
Model and Selection tegically “implanted” at desired positions to probe their sidedness.
of Diagnostic Cysteines The process of choosing suitable residues for replacement by a cys-
for SCAM™ teine is often empirically determined, and the rationale for decid-
ing which residues to substitute is aided by different
machine-learning topology predictors. The physicochemical con-
straints imposed by the lipid environment and the known hydro-
phobicity of individual amino acids provide a method using
hydropathy plots to predict the topology of a membrane protein.
 Determination of Membrane Protein Topology 111

Secondary structure predicted by computer-aided hydropathy

analysis (which so far is 60–70% reliable) is an initial starting point
for the likelihood that a particular residue is in an EMD. The avail-
able algorithms utilize also all available structure and topology
data, information from aligned homologous sequences, topologi-
cal rules, and bioinformatical evidence for topology prediction in a
probabilistic framework provided by the hidden Markov model.
The most recent state-of-the-art topology predictor, the Scale-­
Based Method for Prediction of Integral Membrane Proteins
(SCAMPI) performed simultaneously with the best statistics-based
Hydrophobicity Plot-Based Topology Predictor (TopPredΔG)
shows the highest accuracy (85%) [33]. However, still in many
cases different algorithms produce different predictions, and very
often such predictive methods generate a misleading topology. A
misassigned or misoriented TMDs and EMDs are still occurred
because the primary sequence and overall hydrophobicity of TMDs
are not the only determinants of membrane integration. Although
simple hydrophobicity is the predominant factor determining the
insertion efficiency, analysis of whole genomic data by the TM ten-
dency scale revealed an overlap of TMs and soluble sequences in
the so-called semihydrophobic range [34]. This raises the possibil-
ity that a significant number of proteins will have sequences that lie
close to an equilibrium between membrane insertion and exclusion
that can result in switching between TM and non-TM states
depending on environmental and physiological conditions or
assuming a dual or mixed topologies [3, 21, 22]. Ideally predictors
should exploit all molecular interactions of a nascent membrane
protein (water–protein, translocon–protein, lipid bilayer–protein).
Despite the fact that some topogenic signals in polytopic mem-
brane proteins have been identified, the cellular mechanisms (e.g.,
translocon, lipid bilayer, membrane potential) for response to these
signals are not fully understood. The difference in response may
reflect differences in the mechanisms determining TM orientation
[4]. Obviously, besides interactions of the nascent polypeptide
chain with itself, water, membrane interface, translocon, and
neighboring lipids, the electrophoretic and electrostatic properties
such as the asymmetry of the lipid bilayer itself are important.
Although SCAMPI accommodates a biological ΔG scale that
reflects the thermodynamic cost of translocon-to-bilayer and
water-to-membrane partitionings of TMDs and EMDs during
membrane integration, long-range interactions between TMDs,
unanticipated intersubunit and intraprotein (salt bridges between
charged residues within the hydrophobic core) interactions, post-
translational phosphorylation and glycosylation, retrograde trans-
location and insertion events, specific lipid–protein interactions,
and the unique membrane lipid composition of different organ-
isms and intracellular organells are some of the variables that are
not always addressed by predictive methods. Therefore,
112 Mikhail Bogdanov

hydropathy analysis of the sequence of a polytopic membrane may

only reveal potential TMDs and EMDs and their relative orienta-
tion and sidedness as a starting point for designing biochemical
experiments to establish topological organization. The choice of
sites, i.e., where to introduce a cysteine, can be further guided by
various other available algorithms and refined by a comparison of
predictions with experimentally determined structures. Several data-
bases of TM topologies are available and can be used to evaluate the
reliability of predicted topologies. The recently developed Consensus
Constrained TOPology prediction (CCTOP; http:// cctop.enzim.
ttk.mta.hu) server provides TM topology prediction and utilizes ten
different topology prediction methods and incorporates topology
information from different experimental sources. The TM topolo-
gies in these databases were determined experimentally by means of
X-ray crystallography, NMR, gene fusions, SCAM™, glycosylation
scanning, and other biochemical methods [35].

1.5.3  Mutation Strategy, The pre-requisite for the method is the generation of the cysteine-­
Host and Vector Selections, free template of a protein of interest. The expression of a func-
Construction of Plasmids tional “cysteine-less” template has permitted the use of scanning
Expressing Single-Cysteine cysteine mutagenesis and thiol modification techniques for map-
Derivatives ping membrane topology. All amino acid substitutions should be
verified by DNA sequencing and functional analysis and expres-
sion level by Western blotting of each derivative should be carried
out if possible. Ideally target gene expression should be under
control of an inducible promoter such as OPtac or repressor (TetR)
regulating the promoter (PLtetO-1) to minimize a drastic overex-
pression or continuous expression of potentially disruptive gene
products (see Note 3).
A prerequisite for each cysteine replacement is retention of
function that provides assurance of retention of near native struc-
ture. The native cysteine residues are usually changed into alanine
or serine residues which are small, commonly found in membrane
proteins and appear to be tolerated at most positions thus render-
ing an active protein. A residue that does not tolerate substitution
by Cys makes a crucial contribution to maintaining the structure of
the site, and/or to the folding and function of target protein.
Replacement of charged residues is generally not advised because
these have a high probability of being topogenic signals or may be
involved in long-range charge-charge intramolecular or intermo-
lecular (lipid-protein) interactions. If the protein contains stretches
of residues of intermediate hydrophobicity that cannot unambigu-
ously be identified as membrane spanning TMD, substitutions
should be made approximately every ten residues. The cysteine-­
less protein serves also as a negative labeling control to assure that
residues such as lysine and histidine are not labeled by the reagents
(see Note 1). Ideally the protein under study should be devoid of
all native cysteine residues because these residues may also react
 Determination of Membrane Protein Topology 113

with thiol-modifying reagents or they may form disulfide bonds

with the engineered cysteines and prevent their interaction with
modifying thiol-specific reagents. However the templates contain-
ing endogenous natural cysteines can be still utilized in SCAM™ if
they do not react with the thiol-specific reagents due to membrane
residency [31]. Some of which may be unreactive toward sulfhy-
dryl reagents because they are in disulfide-bonded pairs within
EMDs. However these “silent” thiols can be “awaken” by reduc-
ing agents and even utilized successfully as diagnostic residues in
topology mapping [19].
To obtain a minimal topological map, a single-cysteine replace-
ment in each of the putative EMDs should be expressed from a
plasmid and analyzed in an appropriate host. The host strain for
plasmid expression should be deleted from the target protein gene
if it contains native cysteines and is expressed at levels high enough
to be detected in the assay. Since the target protein is expressed
from a multicopy plasmid, it is often possible to analyze a protein
in its normal host without deletion of the native protein. Since
SCAM™ is based on the controlled membrane permeability of
sulfhydryl reagents, the results of SCAM™ analysis are valid only if
the modifying thiol-specific reagent is membrane impermeable and
cells are intact. Significant differences in the membrane permeabil-
ity of different host strains toward different maleimides [5, 31]
should be considered and tested during host selection procedure
(see Note 4).

1.5.4  Cell Growth Cells are first grown overnight at 37 °C in Luria–Bertani (LB)
and Regulated Expression medium supplemented with appropriate antibiotic and then sub-
of Single-Cysteine cultured in the morning to an OD600 of around 0.05 in the LB
Derivatives medium supplemented with appropriate antibiotic to maintain
plasmid-encoding single-cysteine replacements in cysteineless pro-
tein. Carrying plasmids encoding target protein under appropriate
promoters control are induced usually by growth in the presence
of inducer for at least several generations (OD600 reading ~0.5–
0.6) to reach a logarithmic growth phase of cells.

1.5.5  General Protocol Following the expression of monocysteine mutants, cells are sub-
for SCAM™ jected to SCAM™ analysis by in vivo labeling with different detect-
able and nondetectable thiol-reactive reagents [5, 9].
Maleimide-based thiol reagents, which are available in a wide vari-
ety of forms, are particularly suited for SCAM™ [5]. Maleimide
reacts with the ionized form of a thiol group (thiolate anion)
(Fig. 2a), and this reaction requires a water molecule as a proton
acceptor. In most cases the unreactive cysteine residues are located
within the membrane hydrophobic core or in a sterically hindered
environment [5]. The reaction rate of different thiols is controlled
primarily by their surface exposure and proximal environment.
Most experiments utilize biotinylated, radioactive, fluorescent [5],
114 Mikhail Bogdanov

Fig. 2 A chemical structure of thiol-modifying reagents and their reaction with a

thiol. (a) Reaction of a maleimide with the thiolate of a protein cysteine to form a
covalent adduct via nucleophilic addition to the double bond of the maleimide
ring. Maleimides are virtually unreactive until they encounter an available ionized
thiol group. For most water-exposed cysteine residues in proteins, the pKa of
the thiol of cysteine lies in the range of 8–9 and the formation of cysteinyl thio-
late anions is optimum in aqueous rather in a nonpolar environment, where the
pKa of the thiol of cysteine is around 14. Therefore, the reaction rate of different
sulfhydryls is controlled primarily by their water exposure, making the residues
that reside in regions of TMDs unfavorable for the generation of thiolate anions.
Thus, the labeling characteristics of intramembrane (unreactive) and extra-
membrane (reactive) cysteines should be consistent with their localization in
either a nonpolar or polar environment, respectively. (b) Structure of biotin-
containing labeling reagent MPB. (c) Structure of blocking nondetectable
reagent AMS. Figure is reproduced from [5] with permission from Elsevier

or mass-tagged derivatives of maleimide as alkylating reagents

[5, 11]. For the mapping of membrane topologies, these maleimides
may be used in combination with highly impermeant, nontagged
maleimide derivatives, which are used as blocking agents.
Nondetectable reagents, such as AMS, do not cross the cytoplas-
mic bacterial membrane but penetrate the outer membrane and
 Determination of Membrane Protein Topology 115

therefore only modify cysteines exposed outside of the cytoplasmic

membrane [5] (see Note 4).
Impermeable thiol reagents that can be easily detected after
modification of target proteins are essential for the successful appli-
cation of SCAM™. Biotin-linked maleimides, such as
3-(N-maleimido-propionyl) biocytin (MPB) (Fig. 2b) is particu-
larly useful owing to its low membrane permeability properties.
Following the SDS-PAGE of target proteins isolated by immuno-
precipitation or affinity tags, biotinylated proteins are easily
detected using avidin-horseradish peroxidase (HRP) and chemilu-
minescence. MPB essentially constitutes a universal, multipurpose,
thiol-specific probe capable of detecting protein SH groups within
the femtomole range [36]. SCAM™ using thiol-specific mem-
brane–impermeable MPB (Fig. 2) has been extensively employed
to probe the topological organization of many membrane proteins
[5]. A maleimide-based thiol reagent is added and the modifica-
tion reaction terminated by adding a 50- to 100-fold excess of
β-mercaptoethanol (β-ME), dithiothreitol, or cysteine to inactivate
the unreacted maleimide [5].
The general design of labeling experiments to distinguish
between cysteines located in an extracellular or intracellular EMD
is outlined in Fig. 3. Single-cysteine replacements are expressed in
the appropriate host, and the cells are harvested and suspended in
modification buffer. In this assay, sonication of cells is used to dis-
rupt cell membranes, making both extracellular (periplasmic) and
cytoplasmic cysteines accessible to MPB, whereas cysteines located
within a TM domain are still protected from labeling [10, 19–21].
In this approach, monocysteines are expressed, and reactivity with
MPB in intact cells (extracellular exposure) or only after cell dis-
ruption by sonication (cytoplasmic exposure) was used to establish
TM orientation (Fig. 3). Derivatization of cysteines in whole cells
will indicate extracellular exposure, while derivatization only dur-
ing sonication will indicate a cytoplasmic exposure since
­desintegration allows access of thiol-specific reagent to both sides
of a membrane. Indeed cysteine residing within extracellular EMDs
was labeled whether or not cells were disrupted by sonication,
while cysteine residing within cytoplasmic EMDs was protected
from labeling and is labeled only after cell disruption (Fig. 3, two
top panels), which exposes previously inaccessible cysteine resi-
dues. It is important to note that the extent of biotinylation should
be the same before and after sonication for an extracellular cyste-
ine. If sonication results in an increase in biotinylation, this may
indicate a mixed topology (please see following section). A major
problem is the variability between samples because of mechanical
loss during the work-up or because of differences in the expression
level of individual cysteine replacements. Although single-cysteine
replacements could affect protein expression, conclusions are based
on a comparison of the extent of labeling in whole cells and
116 Mikhail Bogdanov

Fig. 3 General strategy for SCAM™ using impermeable MPB and impermeant transparent AMS to probe the
sidedness of EMDs. A target membrane protein (only one TMD hairpin is shown) containing a single cysteine
exposed either to the extracellular (blue, periplasmic) or intracellular (red, cytoplasmic) side of the membrane
is expressed in host cells. Half of the cells are reacted with a detectable thiol reagent MPB to specifically label
the externally exposed cysteine (first row from top) and the other half are reacted with a nondetectable thiol
reagent AMS (third and fourth rows) to protect external cysteines in subsequent labeling steps. Both halves of
the cells are either kept intact (−) or disintegrated (+) by sonication to expose and label previously inaccesible
cytoplasmic cysteine (second and fourth rows). Labeling by MPB that can be blocked completely by pretreatment
with AMS is independent indicative of a periplasm-facing residue (third row). Labeling by MPB that cannot be
blocked by such AMS treatment is independent indicative of a residue that is facing the cytoplasmn (fourth
row). A target protein was immunoprecipitated and resolved by SDS-PAGE, and biotinylated protein was
detected using avidin-HRP and chemiluminescence (right panel)

disrupted cells for the same protein. This approach simplifies the
interpretation of data obtained with a series of protein derivatives
that may express at different levels since conclusions about topol-
ogy are based on the relative reactivity of cysteines in the same
sample before and after cell disruption. Since sample pairs are ana-
lyzed on the same western blot, no signal intensity normalization
is required. The level of expression of any given derivative will
affect the absolute intensity of labeling but not the ratio of the
labeling between the sample pairs.
Although central to the method is the use of detectable thiol-­
specific reagents to differentiate intracellular from extracellular
EMDs, confirmation of labeling of external water-exposed cyste-
ines by MPB can be achieved by first blocking putative external
cysteines in intact cells with a thiol-specific reagent that is transpar-
ent in the detection phase of the procedure. Such a preblocking
step also allows selective labeling of luminal (exposed to
 Determination of Membrane Protein Topology 117

cytoplasm) cysteines after cell disruption and, therefore, a detec-

tion of previously inaccessible cytoplasmic cysteine residues. A set
of impermeable blocking reagents that effectively react with thiols
exposed to solvent but are transparent in the detection phase of
the procedure is available for SCAM™ [5] (see Note 4). One such
reagent is AMS (Fig. 2c), which is membrane-impermeable due to
its size, two charged sulfonate groups, and high solubility in water.
As a blocking reagent AMS is widely used because of its demon-
strated inability to cross the cytoplasmic membrane of bacteria or
the plasma membrane of mammalian cells. Intact cells are treated
either with or without AMS followed by labeling with MPB of
intact cells and cells during disruption (Fig. 3 two bottom panels).
Excess AMS is removed prior to any subsequent treatments by
several pelletings and washing with buffer.
Thiol reagents react with cysteine residues present in all other
proteins in the membrane. Immunoprecipitation of the mem-
brane protein of interest or a rapid purification step is necessary to
eliminate other labeled proteins. A biotinylated protein can be
recovered from cell lysates directly with streptavidin-agarose beads
and then detected by western blotting using a target-specific anti-
body [5]. Antigen–antibody complexes can be isolated using
Pansorbin (Staphylococcus aureus cells), protein A agarose, or pro-
tein A/G Sepharose beads [5, 22]. If antibodies specific to the
protein under study are not available, then epitope tags, such as
Myc, or affinity tags, such as 6× His [5], can be incorporated at
the C-terminus of the target protein (see Note 5) for either immu-
noprecipitation or isolation by Ni2+ chelated affinity resin packed
into microcolumns or attached to agarose beads [5, 37]. Of
course, protein function or topology should not be compromised
by the presence of the tag (see Note 5). Following modification
with AMS and MPB and isolation, the target protein is resolved
by SDS-PAGE, blotted to a nitrocellulose membrane, and
detected by western blotting. Biotinylation of exposed cysteine
residues of whole cells ­(periplasmic exposure) confirmed in AMS-
pretreated cells or only during sonication (both periplasmic and
cytoplasmic exposure) is detected using a Fluor-S MaxTM
MultiImager (Bio-Rad)  or compatible imaging system and sig-
nals are quantified using available software.

1.5.6  Application Topological protein heterogeneity generated cotranslationally for

of SCAM™ to Identification native proteins has been observed [38, 39]. Manipulation of pro-
of Mixed and Dual tein domains has resulted often in the coexistence of multiple
Topologies topological arrangements for the same protein within the same
membrane with different ratios of properly oriented and inverted
topological isoforms [40]. Membrane proteins can also display
dual topologies dependent on membrane lipid composition,
thereby providing molecular insight into how some proteins might
118 Mikhail Bogdanov

exhibit multiple topological organizations within the same mem-

brane or some alternative organization in different membranes
[21]. How does one discriminate extracellular and intracellular
EMDs of membrane proteins that adopt mixed or dual TM topol-
ogies within the same membrane?
A preblocking step allows selective labeling of luminal (exposed
to cytoplasm) cysteines after cell permeabilization or disruption
and therefore the detection of mixed topologies coexisting within
the same cell membrane [21, 39]. The dual topology or degree of
mixed topology can be assessed with a two-step protocol as shown
on Fig. 3. Intact cells are treated either with or without AMS fol-
lowed by labeling with MPB during disruption by sonication. AMS
treatment will either prevent any biotinylation with whole cells
expressing protein with periplasmic-facing monocysteine residue
or only reduce the amount of biotinylation observed in disrupted
cells. Labeling by MPB that can be blocked completely by pretreat-
ment with AMS is indicative of a periplasmic-facing residue
(Fig. 3). Labeling by MPB that cannot be blocked by AMS pre-
treatment is indicative of a residue that is facing the cytoplasm and
uniform topology if biotinylated intensities of either periplasmic or
cytoplasmic monocysteines in intact and preblocked cells are equal
to the total intensity of biotinylation in unblocked cells labeled
during desintegration. If a protein adopts a dual topology, only
50% of diagnostic cysteine at periplasmic EMD would be protected
from labeling by MPB (Fig. 4, panel A). AMS pretreatment can
also reduce the amount of biotinylation observed in disrupted

Fig. 4 Detection of dual topologies and mixed topologies by SCAM™. Topology of

EMD containing a single-cysteine replacement exposed to the extracellular (peri-
plasmic) side of the membrane was analyzed using a two-step labeling protocol.
Intact cells were either labeled with MPB (1) or pretreated (2) or not with AMS (3)
followed by treatment of sonicated cells with MPB (2 and 3). The varoius labeling
of periplasmic monocysteine during sonication after pretreatment of intact cells
with AMS indicates a dual (panel A) or mixed (panels B and C) topology. The dif-
ference in the extent of biotinylation before and after preblocking with AMS can
be utilized to measure the percentages of oppositely oriented (In and Out) popu-
lations of membrane protein, as shown on right
 Determination of Membrane Protein Topology 119

cells. Reduced biotinylation of periplasmic cysteine by MPB in

whole cells not pretreated with AMS (Fig. 4, first row of panels
A–C), along with the biotinylation of monocysteine during sonica-
tion after pretreatment of intact cells with AMS (Fig. 4, second
row of panels B and C), indicates that some population of the
protein molecules is inserted in an inverted orientation. Thus, in
the case of mixed topology, biotinylation will occur with whole
cells not pretreated with AMS and to a greater or lesser extent after
cell disruption of cells pretreated with AMS, indicating a different
ratio of conversely oriented topological isoforms. The sum of bio-
tinylated intensities in intact (first row of panels A–C) and pre-
blocked cells (second row of panels A–C) should be equal to the
total intensity of biotinylation in unblocked cells labeled during
desintegration (third row of panels A–C). The protective effect of
AMS is almost complete [19, 22] and has been succesfully used to
quantify the degree of mixed topology [21, 22].
To confirm the versatility of this method, single-cysteine
variants of polytopic membranes were directly labeled with MPB
and their accessibility to AMS was scored in intact [19], permeabi-
lized [19, 31, 39], disintegrated cells [10, 20–22, 30], oriented
membrane vesicles [19], or proteoliposomes [23, 24].

2  Materials

2.1  Construction Quickchange site-directed mutagenesis kits (Stratagene) or

of Plasmids equivalent.
Expressing Single-­
Cysteine Derivatives

2.2  Growth of E. coli 1. Luria–Bertani (LB) medium.

Strains 2. Stock of appropriate antibiotic at desired concentration.
3. Stock of approproate inducer at desired concentration (either
500 mM isopropyl-ß-d-thiogalactoside (IPTG) or 20% arabi-
nose or anhydrotetracycline (aTc) (1 mg/mL)).

2.3  SCAM™ 1. Buffer A: 100 mM HEPES-KOH buffer, 250 mM sucrose, 25

mM MgCl2, 0.1 mM KCl, adjusted to pH 7.5 or the same
buffer adjusted to pH 9.0.
2. 10 mM MPB freshly dissolved in dimethyl sulfoxide (DMSO).
Final concentration of DMSO used to dissolve MPB should
never exceed 0.5% (see Note 6).
3. 100 mM aqueous stock solution of AMS.
4. 2 M β-ME.
5. Ultrasonic sonifier.
6. TLA-100 ultracentrifuge (Beckman Coulter, Indianapolis, IN)
equipped with TLA-55 rotor or equivalent.
120 Mikhail Bogdanov

7. Microfuge polyallomer tubes (natural tint, capacity 1.5 mL).

8. Pierce™ spin columns with screw cap (Thermo Fisher Scientific,
Waltham, MA) or equivalent.
9. Tabletop centrifuge.

2.4  Membrane 1. Solubilization buffer: 50 mM Tris–HCl, pH 8.1, 2% SDS, 1

Protein Solubilization mM ethylenediaminetetraacetic acid (EDTA).
2. Vortex equipped with microtube foam rack for multiple polyal-
lomer tubes.

2.5  Immunopre-­ 1. IP1 buffer: 50 mM Tris–HCl, pH 8.1, 0.15 M NaCl, 1 mM

cipitation (IP) EDTA, 2% Lubrol-PX, 0.4% SDS (see Note 7).
2. IP2 buffer: 50 mM Tris–HCl, pH 8.1, 1 M NaCl, 1 mM
EDTA, 2% Lubrol-PX, 2%, 0.4% SDS.
3. Protein A/G-agarose affinity resin.
4. Purification buffer: 20 mM Tris–HCl, pH 7.4, 300 mM NaCl,
25 mM imidazole, 10% (v/v) glycerol, supplemented with 2%
Lubrol-PX or other appropriate nonionic detergent (see Notes
5 and 7).

2.6  SDS-PAGE 1. 2× SDS gel loading (sample) buffer: 10 mM Tris–HCl, pH

and Western Blotting 6.8, 5.6% (w/v) SDS, 200 mM dithiothreitol, 10% (w/v) glyc-
erol, 0.01% bromophenol blue.
2. Precast gels for SDS-PAGE, 12.5% polyacrylamide.
3. 0.45 μm nitrocellulose transfer membranes.
4. Blocking buffer: 5% bovine serum albumin (BSA) in Tris-­
buffered saline (TBS: 10 mM Tris–HCl, pH 7.4, 0.9% NaCl).
5. Avidin linked to horseradish peroxidase (avidin-HRP) reconsti-
tuted to concentration of 2 mg/mL according to manufacturer.
6. Chemiluminescent substrates for detection of HRP.
7. Semidry blotting system.
8. Imaging system such as Fluor-S Max™ MultiImager (Bio-Rad
Laboratories, Hercules, CA) equipped with CCD camera and
Nikon 50 mm 1:1.4 AD (F 1.4) lens at ultrasensitive chemilu-
minescence setting to cool camera to −33 °C. Alternatively,
autoradiography films can be used.

3  Methods

3.1  Labeling 1. Harvest 100 mL mid-log phase cells expressing a single-­

with Maleimide cysteine derivative of protein of interest by centrifugation and
Derivatives suspend cell pellets in 3 mL buffer A adjusted to pH 7.5.
Divide sample into three equal aliquots (0.75 mL) in Microfuge
polyallomer tubes.
 Determination of Membrane Protein Topology 121

2. One cell aliquot is incubated for 30 min by rotating in dark at

room temperature at 25 °C with AMS at a final concentration
of 5 mM (37.5 μL 100 mM aqueous stock solution) to block
periplasmic water-accessible cysteine residues from outside of
cells. Remove excess unreacted AMS by two cycles of centrifu-
gation and resuspension in 0.75 mL buffer A.
3. Treat one set of samples with MPB at final concentration of
100 μM (7.5 μL 10 mM stock solution) (see Note 4) for 5 min
at room temperature to label cysteines exposed to extracellular
(periplasmic) side of inner membrane. To increase reactivity of
diagnostic cysteine residues (particularly those that might be in
sterically hindered EMDs), the reaction with MPB can be car-
ried out at pH 9 (see Note 1). Quench reaction by addition of
β-ME to 20 mM (7.5 μL 2 M stock solution). After labeling,
cells are sonicated for 1 min using amplitude of 15%.
4. To simultaneously label cysteines exposed to both sides of cell
membrane, subject third sample to sonication for 1 min in
presence of MPB at final concentration of 100 μM on ice.
Incubate for 4 min at room temperature and quench reaction
by addition of β-ME to 20 mM (7.5 μL 2 M stock solution).
5. To label previously inaccessible unblocked cytoplasmic cyste-
ines, remaining sample pretreated with AMS from step 2 is
biotinylated by adding MPB at final concentration of 100 μM
(7.5 μL 10 mM stock solution) during sonication for 1 min
followed by incubation for another 4 min at room temperature
before quenching reaction by addition of β-ME to 20 mM (7.5
μL 2 M stock solution).
6. All sonicated samples are centrifuged at 4 °C at 65,000 × g for
10 min followed by resuspension of isolated membranes in
100 μL Buffer A containing 20 mM β-ME by vigorous vortex-
ing for 2 h at room temperature.

3.2  Sample Solubilize isolated membranes with the appropriate detergent or

Solubilization detergent mixture, such as SDS alone, Triton X-100 alone, SDS
and Triton-X-100, CHAPS, octylglucoside, deoxycholate, cholate
and Tween 20, β-D-dodecylmaltoside, nonidet P-40, or sodium
deoxycholate [5] (see Note 7). Resuspended pellets derived from
sonicated cells are solubilized by addition of an equal volume (100
μL) of solubilization buffer followed by vigorous vortexing for
15 min at room temperature, incubation at 37 °C for 15 min, and
an additional 15 min vortexing at room temperature. If solubiliza-
tion buffer with SDS is used, dilute the sample with 0.3 mL cold
IP1 buffer containing nonionic detergent to neutralize the dena-
turing properties of SDS, vortex resulting 0.5 mL sample for 1
min, centrifuge at 4 °C at 20,800 × g for 10 min, and clear by
centrifugation in a prechilled (4 °C) tabletop centrifuge at 20,800
× g for 10 min.
122 Mikhail Bogdanov

3.3  Isolation 1. Transfer supernatants to cups of spin columns.

of Derivatized Target 2. Add appropriate target-protein-specific polyclonal or mono-
Proteins. clonal antibodies (see Note 5).
3. Incubate overnight at 4 °C with rocking.
4. Add 30 μL of a suspension of protein A/G-agarose affinity resin.
5. Incubate at 4 °C on a rocking platform for 90 min.
6. Wash agarose resin with 0.5 mL IP1 by vortexing spin column
for 1 min and subsequent centrifugation in prechilled (4 °C)
tabletop centrifuge at 20,800 × g for 1 min. Discard
flow-through.
7. Repeat washing step with 0.5 mL IP2.
8. Repeat washing step with 0.5 mL 10 mM Tris–HCl, pH 8.1.
9. Add SDS sample buffer.
10. Vortex vigorously for 15 min at room temperature.
11. Incubate at 37 °C for 15 min.
12. Vortex vigorously for 15 min at room temperature.
13. Centrifuge spin column on a tabletop centrifuge at 20,800 × g
for 1 min to collect flow-through with solubilized proteins
into microfuge polyallomer tube.

3.4  SDS-PAGE, 1. Subject immunoprecipitated samples to SDS-PAGE.

Western Blot Analysis, 2. Transfer proteins to nitrocellulose membranes by electroblot-
and Staining ting as described in detail in previously published SCAM ™
with Avidin-HRP protocol [9].
3. Incubate nitrocellulose membrane overnight with blocking
buffer.
4. Wash membrane with TBS buffer containing 0.3% BSA for
10 min.
5. Add Avidin–HRP at a final dilution of 1:5000–10,000 from a
2 mg/mL stock solution, in TBS buffer containing 0.3% BSA.
6. Incubate for at least 1 h.
7. Wash membrane twice with TBS buffer containing 0.3% BSA
for 15 min each.
8. Wash membrane twice with TBS/Nonidet P40 buffer.
9. Wash membrane once with TBS buffer.

10. Incubate membrane for 3 min with chemiluminescent
substrates.
11. Visualize biotinylated proteins using imaging system.

3.5  Data Analysis The criteria used for determining the location of an introduced
and Interpretation cysteine are as follows. Labeling of a cysteine residue with a
membrane-­impermeable sulfhydryl reagent before disruption of
 Determination of Membrane Protein Topology 123

whole cells is indicative of an extracellular (periplasmic) cysteine

residue provided accessibility to a cytoplasmically localized control
protein (see Note 4) and a cysteineless derivative of the target pro-
tein are negative (see Note 1). An absence of labeling in whole cells
but labeling during cell disruption indicates a cytoplasmic location
for a cysteine-containing EMD. The only valid comparison in
intensity is between whole-cell and sonicated sets (images treated
identically) run on the same gel.
No labeling with a sulfhydryl reagent before or during cell dis-
ruption implies localization to a hydrophobic membrane environ-
ment or unfavorable local orientation/positioning of introduced
thiol groups, which may prevent access by the reagent or result in
an increase of the pKa of the thiol group as discussed subsequently
(see Note 1). The percentage change in alkylation between samples
biotinylated with MPB only and those labeled in an AMS-protected
manner during disintegration should be used to detect and quan-
titate an amount of oppositely oriented membrane proteins adopt-
ing dual or mixed topologies (Figs. 3 and 4).

4  Notes

1. Caution must be used in assigning an intramembrane location

to a cysteine residue because it is unreactive to hydrophilic
thiol reagent in both intact and disrupted cells. No definitive
conclusion can be drawn for the location of a domain based on
a lack of reaction of the cysteine. Lack of or low levels of label-
ing may result from any of the following reasons: (1) steric
hindrance due to local secondary structure, (2) internalization
into the compact fold of the protein, (3) lack of ionization of
the thiol group owing to a hydrophobic environment, (4)
local environment with the same charge as the thiol reagent,
or (5) increased pKa of the thiol due to the high negative
charge density of neighboring residues or anionic lipids [5, 9].
Periplasmic EMDs tend to be shorter (sometimes only three
amino acids in length) than cytoplasmic EMDs. Therefore,
there may be little or no protrusion of these loops into the
extracellular space, thereby preventing reaction of the cysteine
residues in these locations with relatively bulky reagents [30].
Alkylating reagents appear to react better with cysteines
toward the middle of extended hydrophilic loops than near
the TMD interfacial domain. Cysteine scanning across a
domain or determining the accessibility of neighbors is an
effective means for identifying useful replacement sites and
differentiating between local effects and unreactive TMDs.
Scanning can be coupled with alkaline treatment. Since the
formation of cysteinyl thiolate anions is favored by increasing
the solution pH (optimum pH 8.0–9.0), increasing the pH
124 Mikhail Bogdanov

during labeling will favor the reaction [5, 9, 10, 30] (Fig. 2a).
However, maleimides are known to react also with primary
amines at pH values above 7.5. An effective control to rule out
nonthiol residue modification is to use a cysteineless template
of the target protein. Unfavorable orientation of a thiol group
owing to a local secondary structure may restrict or prevent
access by large thiol reagents. Increasing the reaction buffer
pH would not only favor alkylation of an extramembrane cys-
teine but also disrupt the local restrictive secondary structure
while truly membrane-embedded cysteines would not be
expected to react. EMDs that are sterically hindered or exhibit
an elevated pKa can be derivatized by increasing the pH up to
10.5 without compromising membrane integrity [10, 30].
However, appropriate controls, such as making known cyto-
plasmically exposed domains inaccessible and removing any
cysteineless target protein label, should be used.
2. Conclusions based on the full reactivity of diagnostic residues
should also be drawn with caution since cysteine residues fac-
ing a hydrophilic pore or near a substrate-binding site may be
within a TM segment but chemically reactive because of water
channels or pockets.
3. The araB system suffers from cell-to-cell heterogeneity when
expression levels vary greatly within the cell population,
thereby making physiological interpretations problematic
because of the induction of promoter yielded mixed popula-
tions of uninduced and fully induced cells. Contrary PLtetO-1
is tightly repressible by the Tet repressor in the absence of
inducer and can be induced with the nontoxic inducer aTc,
which has an increased affinity for repressor TetR and thus can
provide tight and homogeneous cell-to-cell expression.
Overexpression of membrane proteins under the control of
araB or T7 promoter is not recommended or should be carried
out with cation. If the araB system is still utilized a short induc-
tion time (1 h) at a low concetration (0.05%) of arabinose is
recommended. The pET expression systems is widely used
because of its ability to produce large quantities of a desired
protein when activated. However, the use of a pET vector in
SCAM™ is not recommended for the expression of target
membrane proteins because the overloading of cells with a
highly overexpressed target protein can “jam” a translocon and
trigger the accumulation of membrane protein into newly
made intracytoplasmic membranes (Lu, Zheng, and Bogdanov,
unpublished).
4. SCAM™ is based on the controlled membrane permeability of
sulfhydryl reagents. Membranes, either in their native state or
due to experimental manipulation, can be slightly permeable
to labeling reagents. Therefore, optimal labeling conditions
must be established for each reagent and host. The results of
 Determination of Membrane Protein Topology 125

SCAM™ analysis are valid only if the modifying reagent is

thiol-specific, the membrane is impermeable, cells are intact,
and cell disruption does not expose sterically hindered or water
inaccessible cysteine residues [5]. Various reagents, including
MPB, will cross membranes in a concentration-, time-, and
temperature-­dependent manner, and permeability varies with
the genetic background of the host cells [5, 30, 31]. Therefore,
conditions must be empirically determined to minimize the
derivatization of intracellular cysteines. The membrane perme-
ability of a thiol-­specific labeling reagent can be tested, and
labeling conditions (concentration, time, and temperature)
can be established by quantification of the degree of labeling of
an abundant cytoplasmic protein that is rich in surface-exposed
cysteine residues. E. coli ß-galactosidase and other cytosolic
bacterial markers, such as glutathione or elongation factor Tu,
have been used to access membrane permeability [5]. In such
cases, a labeling experiment with both intact and permeabi-
lized cells is carried out, except that soluble proteins rather
than the membrane fraction are retained by immunoprecipita-
tion and analysis. Whole and disrupted cells are treated with
various concentrations of reagent, from 10 μM to 1 mM, at
temperatures ranging from 0 to 25 °C, and for various lengths
of time, from 5 min to 1 h. In most cases, a low concentration
of MPB (100 μM) and relatively short incubation period (5
min) at room temperature favor biotinylation of extramem-
brane thiol groups. Significant differences in the permeability
of different host strains emphasize the need to screen host
strains for reagent permeability prior to initiating experiments
[5, 31]. Thiol reagents are available that contain a biotin
group, a fluorescent group, or a radiolabel allowing the detec-
tion of labeled proteins by avidin-HRP and indirect chemilu-
miscence detection, fluorescence, or autoradiography [5, 37].
A methylpolyethylene glycol-maleimide 5000 (Mal-PEG) adds
5 kDa to the target protein and therefore can be used success-
fully as a mobility shift reagent [41, 42]. In this type of assay,
intact cells are pretreated first with impermeant sulfhydryl
reagents AMS or sodium (2-sulfonatoethyl) methanethiosulfo-
nate (MTSES) that can modify only extracellular thiol groups.
The thiol groups of cytoplasmic or membrane-­embedded cys-
teine residues are not accessible to AMS or MTSES, and pro-
teins can be alkylated with Mal-PEG after full denaturation by
SDS/Urea/EDTA, which renders all unprotected cysteine
residues accessible to modification. The application of fluores-
cent maleimides (UV-excitable Oregon Green 488 maleimide
carboxylic acid (OGM) in SCAM™ eliminates the western
blotting procedure [37]. Infrared fluorescent dye IRDye800-
maleimide (LI-COR) delivers enhanced sensitivity, dissolves
freely in water, and modifies free thiols efficiently at physiologi-
cal pH and therefore can be utilized in SCAM™ for visualization
126 Mikhail Bogdanov

in target proteins by a LI-COR system [43]. Detectable thiol-

reactive labeling reagents with high membrane permeabilities
can also be used to label a test protein. Reagents that can cross
membranes, such as N-ethylmaleimide (used as 14C–labeled
form), modify all extramembrane cysteines irrespective of their
sidedness. These detectable maleimides added to samples after
preincubation with AMS or MTSES cannot modify prelabeled
surface-exposed cysteine residues, resulting in a different label-
ing pattern. However, detectable maleimides with limited per-
meabilities, biotin-linked (MPB) or UV-­ excitable (OGM)
maleimides, are the most commonly used thiol-specific reagents
owing to their low membrane permeability and simple detec-
tion [5, 9, 37].
5. The introduction of 6× or 10× His tags should be used with
caution because such tags may affect the topology of small
integral membrane proteins with weak topological determi-
nants because their intrinsic flexibility makes them prone to
different rearrangements [11, 41]. SCAM™ should map the
topology of such proteins in their untagged form [41].
Negatively charged residues become potent translocation sig-
nals when they are present in high numbers, flank a marginally
hydrophobic TMD, or lie within a window of six residues from
the end of a highly hydrophobic TMD [3, 4]. Therefore, Myc
or FLAG tags may affect the topology of small-membrane pro-
teins owing to the presence of four and five negatively charged
residues in their sequences (EQKLISEEDL and DYKDDDDK,
respectively). The topology of these proteins can be affected by
the insertion of tags or by their position (Gordon and
Bogdanov, unpublished).
6. Maleimide-based reagents are often sensitive to hydrolysis, and
so reagents that have not been stored or handled properly may
no longer be reactive. Therefore, reagents should be prepared
immediately prior to use and should be protected from light
where possible.
7. Prechilled (4 °C) 50 mM Tris–HCl (pH 8.1) should be used
to prepare IP buffer with an SDS:nonionic detergent ratio of
1:5, which is instrumental for immunoprecipitation of highly
hydrophobic integral membrane proteins. Lubrol-PX can be
substituted by ThesitR (Honeywell Fluka™ USA) or lauryldi-
methylamine-oxide (LDAO) at exactly the same concentra-
tion. Triton X-100 is not recommended for immunoprecipiation
or affinity purification of very hydrophobic multispanning
membrane proteins due to their severe aggregation after solu-
bilization by this detergent alone or even in mixture with SDS
(Bogdanov, unpublished observation). LDAO can be advanta-
geous in the purification of small integral membrane proteins
that span membranes two or four times (Bogdanov, unpub-
lished observation).
 Determination of Membrane Protein Topology 127

References

1. von Heijne G (2006) Membrane-protein experimental tools. Biochem Biophys Res

topology. Nat Rev Mol Cell Biol 7:909–918 Commun 453:268–276
2. Bogdanov M, Xie J, Dowhan W (2009) Lipid-­ 15. Liapakis G (2014) Obtaining structural and
protein interactions drive membrane protein functional information for GPCRs using the
topogenesis in accordance with the positive substituted-cysteine accessibility method
inside rule. J Biol Chem 284:9637–9641 (SCAM). Curr Pharm Biotechnol 15:980–986
3. Dowhan W, Bogdanov M (2009) Lipid-­ 16. van Geest M, Lolkema JS (2000) Membrane
dependent membrane protein topogenesis. topology and insertion of membrane proteins:
Annu Rev Biochem 78:515–540 search for topogenic signals. Microbiol Mol
4. Bogdanov M, Dowhan W, Vitrac H (2014) Biol Rev 64:13–33
Lipids and topological rules governing mem- 17. van Geest M, Lolkema JS (1999) Transmembrane
brane protein assembly. Biochim Biophys Acta segment (TMS) VIII of the Na(+)/citrate trans-
1843:1475–1488 porter CitS requires downstream TMS IX for
5. Bogdanov M, Zhang W, Xie J, Dowhan W insertion in the Escherichia coli membrane. J Biol
(2005) Transmembrane protein topology Chem 274:29705–29711
mapping by the substituted cysteine accessibil- 18. Karlin A, Akabas MH (1998) Substituted-­
ity method (SCAM™): application to lipid-­ cysteine accessibility method. Methods
specific membrane protein topogenesis. Enzymol 293:123–145
Methods 36:148–171 19. Bogdanov M, Heacock PN, Dowhan W (2002)
6. Fleishman SJ, Unger VM, Ben-Tal N (2006) A polytopic membrane protein displays a
Transmembrane protein structures without reversible topology dependent on membrane
X-rays. Trends Biochem Sci 31:106–113 lipid composition. EMBO J 21:2107–2116
7. Lacapere JJ, Pebay-Peyroula E, Neumann JM, 20. Bogdanov M, Heacock P, Guan Z, Dowhan W
Etchebest C (2007) Determining membrane (2010) Plasticity of lipid-protein interactions in
protein structures: still a challenge! Trends the function and topogenesis of the membrane
Biochem Sci 32:259–270 protein lactose permease from Escherichia coli.
8. Bochud A, Ramachandra N, Conzelmann A Proc Natl Acad Sci U S A 107:15057–15062
(2013) Adaptation of low-resolution methods 21. Bogdanov M, Dowhan W (2012) Lipid-­
for the study of yeast microsomal polytopic dependent generation of a dual topology for a
membrane proteins: a methodological review. membrane protein. J Biol Chem
Biochem Soc Trans 41:35–42 287:37939–37948
9. Bogdanov M, Heacock PN, Dowhan W (2010) 22. Vitrac H, Bogdanov M, Heacock P, Dowhan
Study of polytopic membrane protein topologi- W (2011) Lipids and topological rules of mem-
cal organization as a function of membrane lipid brane protein assembly: balance between long-
composition. Methods Mol Biol 619:79–101 and short-range lipid-protein interactions.
10. Bogdanov M, Xie J, Heacock P, Dowhan W J Biol Chem 286:15182–15194
(2008) To flip or not to flip: lipid-protein 23. Vitrac H, Bogdanov M, Dowhan W (2013) In
charge interactions are a determinant of final vitro reconstitution of lipid-dependent dual
membrane protein topology. J Cell Biol topology and postassembly topological switch-
182:925–935 ing of a membrane protein. Proc Natl Acad Sci
11. Nasie I, Steiner-Mordoch S, Gold A, Schuldiner U S A 110:9338–9343
S (2010) Topologically random insertion of 24. Vitrac H, Bogdanov M, Dowhan W (2013)
EmrE supports a pathway for evolution of Proper fatty acid composition rather than an
inverted repeats in ion-coupled transporters. ionizable lipid amine is required for full trans-
J Biol Chem 285:15234–15244 port function of lactose permease from
12. Zhu Q, Casey JR (2007) Topology of trans- Escherichia coli. J Biol Chem 288:5873–5885
membrane proteins by scanning cysteine acces- 25. Tang XB, Casey JR (1999) Trapping of
sibility mutagenesis methodology. Methods inhibitor-­induced conformational changes in
41:439–450 the erythrocyte membrane anion exchanger
13. Islam ST, Lam JS (2013) Topological mapping AE1. Biochemistry 38:14565–14572
methods for alpha-helical bacterial membrane 26. Hu YK, Kaplan JH (2000) Site-directed chem-
proteins--an update and a guide. Microbiology ical labeling of extracellular loops in a mem-
2:350–364 brane protein. The topology of the
14. Lee H, Kim H (2014) Membrane topology of Na,K-ATPase alpha-subunit. J Biol Chem
transmembrane proteins: determinants and 275:19185–19191
128 Mikhail Bogdanov

27. Nagamori S, Nishiyama K, Tokuda H (2002) 35. Dobson L, Remenyi I, Tusnady GE (2015)
Membrane topology inversion of SecG CCTOP: a consensus constrained TOPology
detected by labeling with a membrane-­ prediction web server. Nucleic Acids Res
impermeable sulfhydryl reagent that causes a 43:W408–W412
close association of SecG with SecA. J Biochem 36. Bayer EA, Zalis MG, Wilchek M (1985)
132:629–634 3-(N-Maleimido-propionyl)biocytin: a versa-
28. Dale H, Angevine CM, Krebs MP (2000) tile thiol-specific biotinylating reagent. Anal
Ordered membrane insertion of an archaeal Biochem 149:529–536
opsin in vivo. Proc Natl Acad Sci U S A 37. Berezuk AM, Goodyear M, Khursigara CM
97:7847–7852 (2014) Site-directed fluorescence labeling
29. Kerr JE, Christie PJ (2010) Evidence for reveals a revised N-terminal membrane topol-
VirB4-mediated dislocation of membrane-­ ogy and functional periplasmic residues in the
integrated VirB2 pilin during biogenesis of the Escherichia coli cell division protein FtsK. J Biol
agrobacterium VirB/VirD4 type IV secretion Chem 289:23287–23301
system. J Bacteriol 192:4923–4934 38. Moss K, Helm A, Lu Y, Bragin A, Skach WR
30. Xie J, Bogdanov M, Heacock P, Dowhan W (1998) Coupled translocation events generate
(2006) Phosphatidylethanolamine and mono- topological heterogeneity at the endoplasmic
glucosyldiacylglycerol are interchangeable in reticulum membrane. Mol Biol Cell
supporting topogenesis and function of the 9:2681–2697
polytopic membrane protein lactose permease. 39. Woodall NB, Yin Y, Bowie JU (2015) Dual-­
J Biol Chem 281:19172–19178 topology insertion of a dual-topology mem-
31. Zhang W, Bogdanov M, Pi J, Pittard AJ, brane protein. Nat Commun 6:8099
Dowhan W (2003) Reversible topological 40. Gafvelin G, von Heijne G (1994) Topological
organization within a polytopic membrane “frustration” in multispanning E. coli inner
protein is governed by a change in membrane membrane proteins. Cell 77:401–412
phospholipid composition. J Biol Chem 41. Nasie I, Steiner-Mordoch S, Schuldiner S
278:50128–50135 (2013) Topology determination of untagged
32. Wang X, Bogdanov M, Dowhan W (2002) membrane proteins. Methods Mol Biol
Topology of polytopic membrane protein sub- 1033:121–130
domains is dictated by membrane phospholipid 42. Gelis-Jeanvoine S, Lory S, Oberto J,
composition. EMBO J 21:5673–5681 Buddelmeijer N (2015) Residues located on
33. Bernsel A, Viklund H, Falk J, Lindahl E, von membrane-embedded flexible loops are essen-
Heijne G, Elofsson A (2008) Prediction of tial for the second step of the apolipoprotein
membrane-protein topology from first princi- N-acyltransferase reaction. Mol Microbiol
ples. Proc Natl Acad Sci U S A 105: 95:692–705
7177–7181 43. Liu Y, Basu A, Li X, Fliegel L (2015)
34. Zhao G, London E (2006) An amino acid Topological analysis of the Na+/H+ exchanger.
“transmembrane tendency” scale that Biochim Biophys Acta 1848:2385–2393
approaches the theoretical limit to accuracy for 44. Abramson J, Smirnova I, Kasho V, Verner G,
prediction of transmembrane helices: Kaback HR, Iwata S (2003) Structure and
relationship to biological hydrophobicity.
­ mechanism of the lactose permease of
Protein Sci 15:1987–2001 Escherichia coli. Science 301:610–615
 Chapter 10

Defining Membrane Protein Topology

Using pho-lac Reporter Fusions
Gouzel Karimova and Daniel Ladant

Abstract
Experimental determination of membrane protein topology can be achieved using various techniques. Here
we present the pho-lac dual reporter system, a simple, convenient, and reliable tool to analyze the topology
of membrane proteins in vivo. The system is based on the use of two topological markers with complemen-
tary properties, the Escherichia coli β-galactosidase LacZ, which is active in the cytoplasm, and the E. coli
alkaline phosphatase PhoA, which is active in the bacterial periplasm. Specifically, in this pho-lac gene system,
the reporter molecule is a chimera composed of the mature PhoA that is in frame with the β-galactosidase
α-peptide, LacZα. Hence, when targeted to the periplasm, the PhoA-LacZα dual reporter displays high
alkaline phosphatase activity but no β-galactosidase activity. Conversely, when located in the cytoplasm,
PhoA-LacZα has no phosphatase activity but exhibits high β-galactosidase activity in E. coli cells expressing
the ω fragment of LacZ, LacZω (via the α-complementation phenomenon). The dual nature of the PhoA-
LacZα reporter allows a simple way to normalize both enzymatic activities to obtain readily interpretable
information about the subcellular location of the fusion site between the membrane protein under study
and the reporter. In addition, the PhoA-LacZα reporter permits utilization of dual-indicator agar plates to
easily discriminate between colonies bearing cytoplasmic fusions, periplasmic fusions, or out-­ of-­frame
fusions. In total, the phoA-lacZα fusion reporter approach is a straightforward and rather inexpensive
method of characterizing the topology of membrane proteins in vivo.

Key words Membrane proteins, Membrane topology, Dual reporter system, Phosphatase,
β-galactosidase

1  Introduction

Membrane proteins are key players in the vast majority of cellular

processes [1–3]. Understanding how these proteins perform their
functions often starts by determining their topology, i.e., the num-
ber of transmembrane segments (TMSs), their location, and their
orientation relative to the membrane [1]. In this chapter, only pro-
teins employing α-helices to span the membrane are discussed.
Many accurate computational methods are now available for
predicting membrane protein topology, the best ones achieving a
prediction accuracy exceeding 80% [4–7] (see also Chapter 1).

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_10, © Springer Science+Business Media LLC 2017

129
130 Gouzel Karimova and Daniel Ladant

Yet predicted topologies need to be experimentally validated.

Here we present the pho-lac dual reporter system, a simple, conve-
nient, and reliable tool to analyze the topology of membrane
proteins in vivo.
Manoil and Beckwith were the first to describe a reporter gene
fusion approach to studying the topological organization of integral
membrane proteins [8]. In a practical way, the membrane protein
under study is genetically fused in frame to the N-terminus of a
reporter molecule that displays a characteristic phenotype depend-
ing on its subcellular (i.e., cytosolic or extracellular) location. Here
we will focus on two Escherichia coli proteins, alkaline phosphatase
(PhoA) and β-galactosidase (LacZ), that are widely used for reporter
fusion technology as a pair of topological markers with complemen-
tary properties. PhoA is a dimeric, Zn2+-dependent enzyme, which
is active only when located in the periplasm (reviewed in [9, 10]).
In contrast, LacZ is a large homotetrameric enzyme that is active in
the bacterial cytoplasm [10–12].
Initially, the PhoA and LacZ reporters were used separately to
determine the topological organization of many proteins expressed
in E. coli cells (reviewed in [10]), but several important drawbacks
were reported [13–18]. A main problem is that in such separate
fusion approaches, the topological location of a given residue of
the target protein is determined by comparing two enzymatic
activities, the alkaline phosphatase activity of the PhoA fusion and
the β-galactosidase activity of the LacZ fusion. To enable a straight-
forward comparison, the reported enzymatic activities should be
normalized to the rate of fusion protein synthesis [13, 15, 16].
This requires time-consuming experiments, for example, pulse
labeling, immunodetection, and quantification of incorporated
radioactivity [13, 15].
To overcome this limitation, Alexeyev and Winkler designed a
dual reporter system in which the PhoA and LacZ reporters are
combined [19]. In their pho-lac gene system, the reporter molecule
is a chimeric protein composed of the α-peptide of LacZ (aa 4–60),
which is fused in frame to the C-terminus of the mature PhoA
(aa 22–472). This PhoA-LacZα chimeric protein displays high
phosphatase activity when targeted to the periplasm and has no
β-galactosidase activity. In contrast, when the PhoA-LacZα reporter
is directed to the cytoplasm, it has no phosphatase activity but dis-
plays high β-galactosidase activity because the α-fragment of LacZ
is able to interact with a truncated, inactive variant of LacZ (the
so-called ω fragment, such as LacZΔM15) to restore its enzymatic
activity (due to the phenomenon of α-complementation) [20, 21].
Importantly, the authors of the system suggested a simple way to
normalize PhoA and LacZ enzymatic activities without determin-
ing protein synthesis rates [19]. Their assumption was that the
expression level of a particular membrane ­protein/PhoA-­LacZα
fusion would affect the activity levels of both phosphatase and
 pho-lac Reporter Fusions 131

β-galactosidase enzymes but should not change their relative ratio.

Therefore, when the PhoA-LacZα reporter is fused to various
residues of a target protein to generate a set of fusion molecules,
the phosphatase and β-galactosidase enzymatic activities of each
hybrid protein can be normalized (usually to the highest value
observed within the set of fusions) to obtain a so-called normalized
activity ratio (NAR). Thus, this ratio provides readily interpretable
information about the subcellular location of the particular fusion
point (i.e., the residue of the membrane protein after which the
PhoA-LacZα is fused) [19].
In this system, both enzymatic activities of the PhoA-LacZα
reporter can be measured quantitatively using standard colorimetric
substrates, ortho-nitrophenyl-β-D-galactoside (ONPG) for
β-galactosidase and p-nitrophenyl phosphate (pNPP) for phospha-
tase [22, 23]. Remarkably, the PhoA-LacZα reporter activities can be
directly visualized on indicator plates containing compatible PhoA-
and LacZ-specific chromogenic substrates. On such agar plates, the
cells expressing high levels of phosphatase activity are able to convert
the X-Pho (5-bromo-4-chloro-3-indolyl-phosphate) substrate into a
blue-colored, precipitated compound, while the cells expressing high
levels of β-galactosidase activity can be revealed with the Red-Gal
(6-chloro-3-indolyl-β-D-galactoside) substrate that is enzymatically
converted into an insoluble red chromophore.
In the laboratory, to apply the dual PhoA-LacZα reporter sys-
tem, we constructed a low copy number vector, pKTop (4279 bp,
ori p15A), that expresses the PhoA-LacZα reporter under the tran-
scriptional control of a lac promoter and carries a kanamycin resis-
tance selectable marker [24]. Briefly, the phoA-lacZα cassette
designed as in [19] was constructed by a polymerase chain reaction
(PCR) overlapping technique and inserted into the vector pKNT25
(by replacing the T25 ORF). A multicloning site located upstream
of phoA-lacZα makes it possible to create in-frame fusions at the
N-terminus of the PhoA-LacZα dual reporter (Fig. 1) (for more
details, see ref. [24]).
The phoA-lacZα reporter system has been successfully applied
to the analysis of the topology of polytopic membrane proteins
involved in a wide range of cellular functions [19, 25–30]. In our
hands, the phoA-lacZα dual system has been used to experimentally
validate the topological organization of various E. coli cell division
proteins [24, 28], the Staphylococcus aureus multicomponent
GraXSR-VraFG signal transduction system [29], and the Pil pro-
teins involved in the biogenesis of type IV pilus from Neisseria
meningitidis [30].
To illustrate the methodology of utilization of the phoA-lacZα
dual reporter system, we describe here the use of the pKTop v­ ector,
which expresses the phoA-lacZα dual reporter, to characterize the
topology of the E. coli YmgF protein, a 72-residue-long membrane
polypeptide that associates to the cell division machinery [24].
132 Gouzel Karimova and Daniel Ladant

Fig. 1 Schematic representation of pKTop vector used for studying membrane

protein topology with PhoA-LacZα dual reporter method. The figure shows the
position of the phoA-lacZα reporter gene, a lac promoter, kanamycin resistance
gene, origin of replication (p15A), and multicloning site (MCS) that makes it possible
to create fusions at the N-terminus of PhoA-LacZα dual reporter. The plasmid can
also be used to create an Exo III-generated library of 3′-truncated variants of
target protein gene fused to phoA-lacZα, as described in Subheading 3.3. The map
is created with SnapGene software (from GSL Biotech)

2  Materials

2.1  Bacterial Growth 1. Luria–Bertani (LB) broth.

Media, Strain 2. LB agar plates (see Note 1).
and Plasmid
3. Kanamycin: 50 mg/mL in water. Store at −20 °C.
Construction
4. Glucose: 20% w/v in water. Store at room temperature (RT).
5. Isopropyl β-d-1-thiogalactopyranoside (IPTG): 100 mM in
water. Store at −20 °C.
 pho-lac Reporter Fusions 133

6. 6-chloro-3-indolyl-β-d-galactoside (Red-Gal, synonyms: Rose-

Gal, Salmon-Gal): 25%w/v in DMSO (dimethyl sulfoxide)
or DMF (dimethylformamide). Store in the dark at −20 °C
(see Note 2).
7. 5-bromo-4-chloro-3-indolyl-phosphate (X-Pho): 100 μg/mL
in water. Store in dark at −20 °C.
8. E. coli K12 strains capable of α-complementation of LacZ:
XL1-blue, TG1, DH5α, DH10B, etc. (see Notes 3 and 4).
9. A plasmid carrying the phoA-lacZα dual reporter cassette, such
as pKTop [24].
10. Common molecular biology enzymes: restriction enzymes,

DNA modification enzymes, ligases, PCR polymerases, etc.
11. Molecular biology kits for purification of plasmid DNA, PCR,
and DNA fragments, etc.
12. Equipment for growth of bacterial culture: thermostat,

incubation shaker.
13. PCR thermal cycler.

2.2  β-Galactosidase 1. M63 medium: 100 mM KH2PO4, 15 mM (NH4)2SO4, 1.7

Assay mM Fe2SO4, 1 mM MgSO4, pH 7.0 (see Note 5).
2. Chloroform.
3. 20% sodium dodecyl sulfate (SDS).
4. β-galactosidase assay buffer (PM2): 70 mM Na2HPO4, 30 mM
NaHPO4, 1 mM MgSO4, 0.2 mM MnSO4, pH 7.0 (see Note 6).
5. ONPG: 0.4% w/v in PM2 buffer without 2-mercaptoethanol.
Store at −20 °C (see Note 7).
6. 1 M Na2CO3.
7. Microplate reader (or microphotometer).
8. Microtiter plates.

2.3  Phosphatase 1. Wash buffer for phosphatase assay (WB): 10 mM Tris–HCl,

Assay pH 8.0, 10 mM MgSO4.
2. Iodoacetamide: 500 mM in water. Prepare a fresh solution
and store in dark until ready to use at 2–8 °C (see Note 8).
3. Phosphatase assay buffer (PM1): 1 M Tris–HCl, pH 8.0, 0.1
mM ZnCl2, 1 mM iodoacetamide.
4. p-nitrophenyl phosphate (pNPP) (see Note 9).
5. 1 M Tris–HCl, pH 8.0.
6. 2N NaOH.
7. Microfuge.
8. Microplate reader (or microphotometer).
9. Microtiter plates.
134 Gouzel Karimova and Daniel Ladant

3  Methods

Readers are expected to be familiar with experimental tools that are

generally used in classical molecular biology, for example, PCR
amplification, DNA digestion, ligation, transformation, and plasmid
DNA purification [23, 31].

3.1  Selection 1. Predict in silico the membrane topology of the protein of

of pho-lac Fusion interest. For consensus prediction of membrane protein topo-
Sites Within Target logical organization, it is important to explore methods based
Membrane Protein on various prediction algorithms [32] (see Note 10 and
Chapter 2).
For example, as shown in Fig. 2a, all four predictors present
YmgF as a protein with two TMSs separated by a short periplasmic
loop and with both the N- and C-terminal extremities in the cytosol.
2. Choose the fusion points for the PhoA-LacZα reporter. At
least one reporter fusion is required for each extramembra-
nous domain of the protein under study. The PhoA-LacZα
reporter should be preferentially fused to the C-terminal ends
of the extramembranous loops between the potential TMSs
[10, 33, 34].
For experimental validation of the predicted YmgF topol-
ogy, we selected seven distinct codons (D8, M21, K32, E39,
N48, L57, and Q72) as sites for the PhoA-LacZα reporter
insertion based on the predicted topology models (see Fig. 2a).
3. Construct the recombinant plasmids expressing the target
membrane protein/phoA-lacZα fusions following the three
commonly used methods described in Subheadings 3.2, 3.3,
and 3.4 (see Note 11).

3.2  C-Terminal 1. Amplify by PCR the DNA fragments carrying the predeter-
Fusion Approach mined 3′-truncated target protein gene with specific PCR
primers harboring appropriate restriction sites for subcloning
in pKTop plasmid. These PCR primers should be designed in
such a way that the amplified DNA fragment is fused in frame
to the downstream phoA-lacZα reporter gene (i.e., to generate
the expected C-terminal translation fusion).
2. Digest the PCR-amplified DNA fragments with appropriate
restriction enzymes.
3. Ligate the digested PCR-amplified DNA fragments into the
phoA-lacZα reporter vector linearized with the same restric-
tion enzymes.
4. Transform the ligation mixture into E. coli competent cells
(see Notes 3, 4, and 12).
 pho-lac Reporter Fusions 135

(A) D8 M21 K32 E39 N48 L57 Q72

QQ

(1)
(2)
(3)
(4)

(B) (C)
100
C
D8

Arbitary units(%)
Q72
75

M21 50
E57

25
K32
N48
E39 0
Fusion point: D8 M21 K32 E39 N48 L57 Q72
NAR
(PhoA/LacZ): 0.17 1.3 3.0 2.5 2.0 0.5 0.2
Location: cyt tm1 per per tm2 tm2 cyt

Fig. 2 YmgF topology analysis. (a) In silico predicted topological models of YmgF. Predictions were made using
four different methods: (1) PSIPRED, (2) Topcons, (3) Phobius, and (4) TopPred. The red lines indicate the cytosolic
domains, light blue TMSs (TMS1 and TMS2), and the deep blue line the periplasmic part. The small black arrow-
heads at the top of the sequences indicate the positions of the different sites selected to construct fusions with
the Pho-LacZα reporter. (b) Experimental determination of YmgF membrane topology. E. coli DH5α cells express-
ing different YmgF/Pho-LacZα fusions (position of insertion indicated in label) were plated on an indicator medium
containing the two chromogenic substrates Red-Gal (for β-galactosidase activity) and X-Pho (for phosphatase
activity). Blue coloration of the colonies (high phosphatase activity) indicates a membrane or periplasmic location
of the fusion point. Red coloration of the colonies (high β-galactosidase activity) indicates cytosolic location of
fusion point. Control cells (i.e., E. coli DH5α/pKTop) are indicated by the C label. (c) Quantitative enzymatic assays
of various YmgF/PhoA-LacZα fusions. The bar chart (blue bar, phosphatase activity; red bar, β-galactosidase
activity) represents the relative PhoA and LacZ enzymatic activities, measured on liquid cultures of DH5α express-
ing the indicated YmgF/Pho-LacZα fusions (position of insertion indicated on the abscissa). The normalized
activities (NAR) are plotted below as is the deduced subcellular localization of the residues at the position of the
Pho-LacZα insertion (cyt: cytoplasmic; per: periplasmic; tm: TMS)

3.3  Nested Deletion This method makes it possible to obtain a library of randomly
Approach generated 3′-truncations of the target protein gene to be fused to
the phoA-lacZα reporter cassette [19, 25–27, 35].
1. Clone the gene encoding the target membrane protein
upstream of the phoA-lacZα reporter gene (e.g., between the
PstI and XbaI sites of pKTop).
136 Gouzel Karimova and Daniel Ladant

2. Digest the resulting plasmid with a restriction enzyme that

cuts between the target gene and phoA-lacZα cassette to gen-
erate a 5′-overhang end (or blunt end) at the 3′-end of the
target gene sequence (e.g., XbaI, BamHI, or SmaI of pKTop),
and then with a second enzyme that generates a 3′-overhang
end at the 5′-end of the phoA-lacZα gene (e.g., KpnI or SacI
of pKTop).
3. Add Exo III nuclease to progressively digest the target gene
from its 3′-end [36]. Remove aliquots at regular intervals to
yield a pool of random 3′-target gene truncations.
4. After removing single-stranded regions by a treatment with
the Mung Bean nuclease, treat with the Klenow fragment of
DNA polymerase I (in the presence of deoxynucleotides), and
recircularize the plasmids with T4 DNA ligase.
5. Transform mixture into competent cells (see Notes 3, 4, and 12)
and plate on dual-indicator medium as described in
Subheading 3.5 (see Note 13).

3.4  Sandwich Fusion This method allows one to insert the PhoA-LacZα reporter into
Approach various loops (cytosolic or periplasmic) of the otherwise intact
protein [19, 25–27].
1. Design PCR primers to allow for an insertion of the phoA-
lacZα cassette in frame with both the upper and downstream
portions of the target gene.
2. Clone the PCR-amplified phoA-lacZα cassette at selected
restriction sites within the target gene.
3. The selected restriction sites in the target gene may already
preexist within the native sequence or, alternatively, they must
be introduced by site-directed mutagenesis. Caution should
be exercised during construction of recombinant plasmids that
express a hybrid protein between the target membrane protein
(or its fragment) and the PhoA-LacZα reporter (see Notes 12
and 14).

3.5  Analyzing Clones Screening for topological locations can be done directly on agar
on Dual Substrate plates that contain the specific chromogenic substrates for both
Plates PhoA- and LacZ-enzymatic activities (see Notes 15 and 16).
1. Select the E. coli cells (e.g., DH5α or other suitable strains; see
Notes 3, 4, and 12) transformed with the constructed plas-
mids on LB agar plates containing the appropriate antibiotic
(e.g., kanamycin for pKTop) and glucose (0.1–0.2%) and
incubate for 20–24 h at 30 °C to reduce the expression of the
recombinant phoA-lacZα fusions (see Note 12).
2. Streak a single clone of each phoA-lacZα fusion under study
on a fresh dual-indicator plate, containing kanamycin
 pho-lac Reporter Fusions 137

(50  μg/mL), Red-Gal (80 μg/mL), X-Pho (100 μg/mL),

IPTG (1 mM).
3. Incubate plate for 20–24 h at 30°–37 °C (for an example,
see Fig. 2b).

3.6  Growth Many different protocols for PhoA and LacZ activity assays have
of Bacterial Culture been described in the literature [22, 23, 31, 37]. We use a simpli-
for Enzymatic Assays fied version of these protocols, in which the E. coli cells are per-
meabilized with chloroform and SDS. The protocols can be easily
adapted for use in a 96-well microtiter plate format.
1. Pick a single colony from a fresh LB/kanamycin/glucose plate
that has been inoculated with DH5α (pKTop-x) cells. Transfer
the colony into 5 mL LB broth with kanamycin (50 μg/mL)
and glucose (0.1%). Grow overnight at 37 °C with shaking
(150–200 rpm).
2. The next day, dilute overnight culture 1:100 in LB broth fresh
medium with kanamycin and let it grow at 37 °C with shaking
(150–200 rpm) for 2.5–3 h (to reach mid-exponential phase
of growth).
3. Add 1 mM IPTG to each culture, to induce expression of
hybrid X/PhoA-LacZα protein, and incubate for an additional
hour with aeration at 37 °C (see Note 17).

3.7  Assay 1. Centrifuge 1.2 mL of the bacterial culture in an Eppendorf

of β-Galactosidase tube (e.g., for 5 min at 7000 rpm (4500 g) at RT in a tabletop
Activity microfuge) and resuspend the pellet in 1.2 mL M63 medium.
Transfer 200 μL of bacterial suspension to microtiter plate
well and measure optical density (OD595) of cells at 595–
600 nm (see Note 18).
2. To permeabilize cells, add 100 μL chloroform and 100 μL
0.05% SDS to 1 mL of the washed cells, vortex for 10 s, and
incubate for 5 min at 37 °C. Then place the tubes on ice for
5 min.
3. After the chloroform has settled, transfer 50 μL of the upper
phase of the bacterial suspension to a microtiter plate well.
4. To start the reaction, add 100 μL of the reaction mixture,
which contains PM2 buffer and ONPG (0.15%), to the bacte-
rial suspension, and incubate at RT until a yellow color devel-
ops. To stop the reaction, add 50 μL 1 M Na2CO3. Record
incubation time. Record OD600 and OD405 for each sample.
5. Calculate enzymatic activity in relative units (A) according to
the following formula:
A = 1000 × (OD405sample − OD405control well)/(OD595
sample − OD595control well)/t (min) of incubation.
138 Gouzel Karimova and Daniel Ladant

3.8  Assay 1. Centrifuge 1.2 mL of the bacterial culture in Eppendorf tube

of Phosphatase (e.g., for 5 min at 7000 rpm at RT in a tabletop microfuge).
Activity 2. Wash cells in cold WB and resuspend pellet in 1.2 mL cold
PM1 buffer. Transfer 200 μL of the bacterial suspension to a
microtiter plate well and measure optical density (OD595) of
cells at 595–600 nm.
3. To permeabilize the cells, add 100 μL chloroform and 100 μL
0.05% SDS to 1 mL of the washed cells, vortex for 10 s, and
incubate for 5 min at 37 °C. Then place tubes on ice for 5 min.
After the chloroform has settled, transfer 100 μL of the upper
phase of the bacterial suspension to a microtiter plate well.
4. To start the reaction, add 50 μL of the pNPP solution (0.15%
in 1 M Tris–HCl, pH 8.0) to the bacterial suspension and
incubate at RT until yellow color develops. Add 50 μL 2 N
NaOH to stop the reaction. Record incubation time. Record
OD405 for each sample.
5. Calculate enzymatic activity in relative units (A) according to
the following formula: A = 1000 × (OD405sample −
OD405control well)/(OD595 sample − OD595control well)/t
(min) of incubation.
6. After obtaining both enzymatic activities for each tested
x/phoA-lacZα fusion (from steps 5 in Subheadings 3.7 and 3.8),
the NAR is calculated as follows:
NAR = (PhoA activity/Highest PhoA activity)/(LacZ
activity/Highest LacZ activity), where Highest PhoA or Highest
LacZ are the corresponding maximum activities measured within
the set of analyzed fusions (see Fig. 2c for an example).
As shown in Fig. 2c, the dual PhoA-LacZα experimental
approach confirmed that YmgF is a polypeptide possessing two
TMSs separated by a short periplasmic loop and with both termini
exposed to the cytosol. The membrane association of YmgF was
further corroborated by subcellular fractionation of cells ­expressing
YmgF-GFP fusion, which was found to be fully associated with the
bacterial membrane fraction [24]. Altogether, these results showed
that YmgF is an integral membrane protein.

4  Notes

1. LB agar plates are prepared by adding, just before pouring and

when necessary, kanamycin (50 μg/mL), Red-Gal (80 μg/mL),
X-Pho (100 μg/mL), 1 mM IPTG, or glucose (0.1–0.2%).
2. Magenta-Gal (synonym: Red-β-D-Gal, 5-Bromo-6-chloro-
3-­­indolyl-β-d-galactopyranoside) can be used instead of Red-
Gal at the same final concentration (80 μg/mL). Stock
 pho-lac Reporter Fusions 139

solution 25% w/v in DMF or DMSO. Store at −20 °C and

protect from light. Importantly, Magenta-Gal is generally
more expensive than Red-Gal.
3. In E. coli, the expression of the endogenous phoA gene is inhib-
ited by moderate concentrations of inorganic phosphate in
growth media. In LB broth, a high-phosphate medium (ca 4–6
mM) [38], endogenous phosphatase activity of various widely
used E. coli strains (DH5, JM101, JM109, etc.) is shown to be
quite low (about 3–5 U) [39]. Therefore, the phoA-lacZα dual
reporter system can be used in any E. coli strain that carries the
wild-type phoA gene and is capable of β-galactosidase
α-complementation (i.e., it should express the ω fragment of
LacZ such as the common variant produced by E. coli
LacZΔM15).
4. One study revealed that E. coli strain DH5α is indeed phoA-­
deficient (ΔphoA) [39]. This makes the strain especially useful
for studies in which the phoA-lacZα dual reporter technology
is applied.
5. In the laboratory, we prepare 2× stock solution.
6. 2-mercaptoethanol (100 mM) can be added to PM2 buffer to
approximately double the level of the β-galactosidase enzy-
matic activity. But 2-mercaptoethanol is considered toxic,
causing irritation to the respiratory tract, nasal passageways,
skin, etc., so it can be omitted.
7. The ONPG stock solution can be frozen and thawed several
times.
8. Iodoacetamide is included in the buffers to block activation of
cytoplasmic PhoA by oxidation during cell lysis [16, 22].
9. In the laboratory, we use SIGMAFAST pNPP tablets. To pre-
pare a stock solution (0.5%), add one tablet to 1 mL 1 M Tris–
HCl, pH 8.0. Store at −20 °C.
10. Diverse accurate predictors for potential α-helical transmem-
brane proteins are freely available online. We routinely use
PSIPRED [40], Topcons [41], Phobius [42], and TopPred
1.10 [43].
11. If the target protein is a relatively large polypeptide (e.g., poly-
topic membrane proteins), the nested deletion approach and
the sandwich fusion approach may be more suitable.
12. Many factors can affect the success of a recombinant plasmid
construction, including, for example, the potential toxicity of
the expressed gene product, the plasmid copy number, and the
genotype of the bacterial host strains. In the laboratory, to mini-
mize problems during the construction of recombinant plas-
mids, in which the expression of a cloned gene is driven by a
relatively strong lac promoter (such as in pKTop), the
140 Gouzel Karimova and Daniel Ladant

corresponding ligation mixtures are transformed into LacIq E.

coli competent cells (i.e., overexpressing the LacI repressor),
routinely, into XL1 blue. The transformed cells are grown at 30
°C, for 24–32 h, on LB agar plates containing appropriate
antibiotic(s) and supplemented with glucose (0.1–0.2%) that,
because of the catabolic repression phenomenon, will reduce
the basal transcription of the lac promoter and, therefore,
diminish the expression of the membrane protein/PhoA-­LacZα
fusions. Any standard transformation protocol may be used [23,
31, 37]. Routinely, we use a simple CaCl2 procedure to prepare
E. coli competent cells [31, 37]. This method, which yields a
competency level > 106 cfu/mg, is adequate for most needs.
13. Statistically, only one-third of recombinant plasmids are

expected to encode in-frame fusions between target protein
truncations and the PhoA-LacZα reporter. The cells bearing
plasmids with out-of-frame fusions can be easily detected on
dual-indicator agar plates since they should stay colorless.
14. It is important to note that the PhoA-LacZα reporter could
influence the folding of the target protein when inserted inter-
nally. Often, C-terminal PhoA-LacZα fusions have higher
enzymatic activities than the corresponding sandwich fusions,
indicating lower expression levels or more steric problems in
the latter [19].
15. Bacteria expressing high levels of phosphatase activity, i.e., when
a PhoA-LacZα reporter is localized in periplasm, will turn blue
as a result of conversion of the X-Pho substrate into a blue-
colored precipitated product. Conversely, cells that express
high levels of β-galactosidase activity, i.e., when the PhoA-
LacZα reporter is localized in the cytoplasm, will become red
owing to the conversion of the Red-Gal substrate into an
insoluble red compound. Fusions of the PhoA-LacZα reporter
within TMSs usually result in a purple pigmentation of colo-
nies (i.e., a combination of both red and blue colorations as a
mixture of cytoplasmic and periplasmic fusions). In addition,
for ExoIII-generated phoA-lacZα fusion library constructions,
utilization of such dual-indicator agar plates allows for easy
detection (and elimination) of colorless colonies that harbor
noninformative, out-of-frame fusions.
16. It should be noted that, whereas true in-frame fusions to
cytoplasmic domains develop red coloration in 12–16 h with
E. coli TG1, out-of-frame fusions are also able to develop red
coloration after a longer incubation period (> in 30–48 h).
The precise timing is likely to be different for different E. coli
strains and would depend on the growth rate [19].
17. It is possible to grow subcultures from overnight inoculations
to the mid-exponential phase in the presence of IPTG, i.e.,
increasing time of induction (3.5–4 h instead of 1 h).
 pho-lac Reporter Fusions 141

18. Adjust the aliquot volume for OD measurements depending

on the equipment used. The volume of cells used in a reaction
may depend on the level of expected enzymatic activity.

Acknowledgment

This work was supported by Institut Pasteur and Centre National

de la Recherche Scientifique (CNRS UMR 3528, Biologie
Structurale et Agents Infectieux).

References
1. von Heijne G (2006) Membrane-protein 12. Silhavy TJ, Shuman HA, Beckwith J et al
topology. Nat Rev Mol Cell Biol 7:909–918 (1977) Use of gene fusions to study outer
2. Islam ST, Lam JS (2013) Topological mapping membrane protein localization in Escherichia
methods for α-helical bacterial membrane pro- coli. Proc Natl Acad Sci U S A 74:5411–5415
teins - an update and a guide. Microbiologyopen 13. Bibi E, Beja O (1994) Membrane topology of
2:350–364 multidrug resistance protein expressed in
3. Dobson L, Remenyi I, Tusnady GE (2015a) The Escherichia coli. N-terminal domain. J Biol
human transmembrane proteome. Biol Direct Chem 269:19910–19915
10:1–18 14. Boyd D, Manoil C, Beckwith J (1987)
4. Chen CP, Rost B (2002) State-of-the-art in Determinants of membrane protein topology.
membrane protein prediction. Appl Bioinforma Proc Natl Acad Sci U S A 84:8525–8529
1:21–35 15. Boyd D, Manoil C, Froshauer S et al (1990)
5. Tusnady GE, Simon I (2010) Topology pre- Use of gene fusions to study membrane-­protein
diction of helical transmembrane proteins: how topology. In: Gierash LM, King J (eds) Protein
far have we reached? Curr Protein Pept Sci folding: deciphering the second half of the
11:550–561 genetic code. AAAS Books, Washington
6. Dobson L, Remenyi I, Tusnady GE (2015b) 16. Manoil C (1990a) Analysis of protein localiza-
CCTOP: a consensus constrained TOPology tion by use of gene fusions with complemen-
prediction web server. Nucleic Acids Res tary properties. J Bacteriol 172:1035–1042
43:W408–W412 17. San Millan JL, Boyd D, Dalbey R et al (1989)
7. Peters C, Konstantinos D, Shu N et al (2016) Use of phoA fusions to study the topology of
Improved topology prediction using the termi- the Escherichia coli inner membrane protein
nal hydrophobic helices rule. Bioinformatics leader peptidase. J Bacteriol 171:5536–5541
32:1158–1162 18. Silhavy TJ, Beckwith JR (1985) Uses of lac
8. Manoil C, Beckwith J (1986) A genetic fusions for the study of biological problems.
approach to analyzing membrane protein Microbiol Rev 49:398–418
topology. Science 233:1403–1408 19. Alexeyev MF, Winkler HH (1999) Membrane
9. Manoil C, Mekalanos JJ, Beckwith J (1990) topology of the Rickettsia prowazekii ATP/
Alkaline-phosphatase fusions - sensors of sub- ADP translocase revealed by novel dual pho-lac
cellular location. J Bacteriol 172:515–518 reporters. J Mol Biol 285:1503–1513
10. van Geest M, Lolkema JS (2000) Membrane 20. Langley KE, Villarejo MR, Fowler AV et al
topology and insertion of membrane proteins: (1975) Molecular basis of beta-galactosidase
search for topogenic signals. Microbiol Mol alpha-complementation. Proc Natl Acad Sci U
Biol Rev 64:13–33 S A 72:1254–1257
11. Lee C, Inouye H, Brickman ER et al (1989) 21. Ullmann A, Jacob F, Monod J (1967)
Genetic studies on the inability of beta-­ Characterization by in vitro complementation of
galactosidase to be translocated across the a peptide corresponding to an operator-­proximal
Escherichia coli cytoplasmic membrane. segment of the beta-galactosidase structural
J Bacteriol 171:4609–4616 gene of Escherichia coli. J Mol Biol 24:339–343
142 Gouzel Karimova and Daniel Ladant

22. Manoil C (1990b) Analysis of membrane

32. Nilsson J, Persson B, von Heijne G (2000)
protein topology using alkaline phosphatase Consensus predictions of membrane protein
and beta-galactosidase gene fusions. Methods topology. FEBS Lett 486:267–269
Cell Biol 34:35–47 33. Boyd D, Traxler B, Beckwith J (1993) Analysis
23. Miller JH (1992) A short course in bacterial of the topology of a membrane protein by
genetics: a laboratory manual and handbook using a minimum number of alkaline phospha-
for Escherichia coli and related bacteria. Cold tase fusions. J Bacteriol 175:553–556
Spring Harbor Laboratory Press, Cold Spring 34. Cassel M, Seppala S, von Heijne G (2008)
Harbor, New York Confronting fusion protein-based membrane
24. Karimova G, Robichon C, Ladant D (2009) protein topology mapping with reality: the
Characterization of YmgF, a 72-residue inner Escherichia coli ClcA H+/Cl- exchange trans-
membrane protein that associates with the porter. J Mol Biol 381:860–866
Escherichia coli cell division machinery. 35. Sugiyama JE, Mahmoodian S, Jacobson GR
J Bacteriol 191:33–46 (1991) Membrane topology analysis of
25. Islam ST, Taylor VL, Qi M et al (2010) Escherichia coli mannitol permease by using a
Membrane topology mapping of the O-antigen nested-deletion method to create mtlA-phoA
flippase (Wzx), polymerase (Wzy), and ligase fusions. Proc Natl Acad Sci U S A 88:
(WaaL) from Pseudomonas aeruginosa PAO1 9603–9607
reveals novel domain architectures. MBio 36. Henikoff S (1987) Unidirectional digestion
1:e00189–e00110 with exonuclease III in DNA sequence analy-

26. Korres H, Verma NK (2004) Topological sis. Methods Enzymol 155:156–165
analysis of glucosyltransferase GtrV of Shigella 37. Sambrook J, Russell DW (2006) The condensed
flexneri by a dual reporter system and identifi- protocols from molecular cloning : a laboratory
cation of a unique reentrant loop. J Biol Chem manual. Cold Spring Harbor Laboratory Press,
279:22469–22476 Cold Spring Harbor, New York
27. Nair AH, Korres H, Verma NK (2011) 38. Schurig-Briccio LA, Farias RN, Rintoul MR
Topological characterisation and identification et al (2009) Phosphate-enhanced stationary-­
of critical domains within glucosyltransferase phase fitness of Escherichia coli is related to
IV (GtrIV) of Shigella flexneri. BMC Biochem inorganic polyphosphate level. J Bacteriol
12:1–14 191:4478–4481
28. Karimova G, Davi M, Ladant D (2012) The 39. Rodriguez-Quinones F, Benedi VJ (2003)
beta-lactam resistance protein Blr, a small Escherichia coli strain DH5α is a suitable host
membrane polypeptide, is a component of the for the study of phoA insertions. Focus
Escherichia coli cell division machinery. 15:110–112
J Bacteriol 194:5576–5588 40. Jones DT (2007) Improving the accuracy of
29. Falord M, Karimova G, Hiron A et al (2012) transmembrane protein topology prediction
GraXSR proteins interact with the VraFG ABC using evolutionary information. Bioinformatics
transporter to form a five-component system 23:538–544
required for cationic antimicrobial peptide sens- 41. Tsirigos KD, Peters C, Shu L et al (2015) The
ing and resistance in Staphylococcus aureus. TOPCONS web server for combined mem-
Antimicrob Agents Chemother 56:1047–1058 brane protein topology and signal peptide pre-
30. Georgiadou M, Castagnini M, Karimova G diction. Nucleic Acids Res 43:W401–W407
et al (2012) Large-scale study of the interac- 42. Käll L, Krogh A, Sonnhammer ELL (2007)
tions between proteins involved in type IV Advantages of combined transmembrane
pilus biology in Neisseria meningitidis: charac- topology and signal peptide prediction - the
terization of a subcomplex involved in pilus Phobius web server. Nucleic Acids Res
assembly. Mol Microbiol 84:857–873 35:W429–W432
31. Green MR, Sambrook J (2012) Molecular 43. Claros MG, von Heijne G (1994) TopPred II:
cloning : a laboratory manual, 4th edn. Cold an improved software for membrane protein
Spring Harbor Laboratory Press, Cold Spring structure predictions. Comput Appl Biosci
Harbor, New York 10:685–686
 Chapter 11

In Vivo and In Vitro Protein–Peptidoglycan Interactions

Gang Li and S. Peter Howard

Abstract
Bacteria have developed a number of trans-envelope systems to transport molecules or assemble organelles
across bacterial envelopes. However, bacterial envelopes contain a rigid netlike peptidoglycan structure
that protects cells from osmotic lysis. Trans-envelope systems thus must interact with the peptidoglycan
barrier to generate gaps or anchor structures to the peptidoglycan scaffold. Here we describe methods to
use in vivo cross-linking and in vitro co-sedimentation to study protein–peptidoglycan interactions in
Gram-negative bacteria. In particular, we address important considerations to ensure the specificity of the
interactions in question.

Key words Trans-envelope systems, Peptidoglycan, Cross-linking, Co-sedimentation, Muramic acid assay

1  Introduction

Bacterial cells have a unique peptidoglycan layer in the cell enve-

lope [1]. The rigid netlike peptidoglycan structure determines cell
shape and protects bacteria from osmotic lysis. However, it also
acts as a barrier for transport of proteins or assembly of large
envelope-­spanning complexes [2]. Local hydrolysis or remodeling
of peptidoglycan is therefore necessary to generate gaps to assem-
ble trans-envelope structures. In addition, macromolecular com-
plexes may use peptidoglycan as a structural extension to anchor
securely to a cell envelope [3]. To date, peptidoglycan-interacting
components have been identified in a wide range of trans-envelope
systems including type II secretion (T2SS), type IV pilus (T4P),
flagella, type III secretion (T3SS), type IV secretion (T4SS), and
type VI secretion (T6SS).
Here we describe an in vivo cross-linking approach to studying
interactions between a protein component of the T2SS (ExeA) and
peptidoglycan in Aeromonas hydrophila [4]. Bacterial cells are
in­cubated with the cleavable cross-linking reagent
­3,3′-dithiobis[sulfosuccinimidylpropionate] (DTSSP), which has
two amine-reactive groups at the ends of an eight-atom spacer

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_11, © Springer Science+Business Media LLC 2017

143
144 Gang Li and S. Peter Howard

arm. DTSSP can covalently link two primary amines (from ExeA
and peptidoglycan in this case) if they are in close proximity. After
cross-linking, peptidoglycan sacculi are isolated by a modified
small-volume version of the sodium dodecyl sulfate (SDS)-boiling
method to remove noncovalently associated proteins [5]. Purified
peptidoglycan samples are then treated with β-mercaptoethanol to
release cross-linked proteins to be analyzed by SDS-PAGE (poly-
acrylamide gel electrophoresis) and immunoblot.
We also describe a co-sedimentation (pulldown) assay that uses
purified peptidoglycan sacculi to investigate the binding of pro-
teins of interest to peptidoglycan [6]. This is a direct and straight-
forward method to detect protein–petidoglycan interactions
in vitro. We provide a protocol to prepare highly pure peptidogly-
can sacculi using SDS-boiling and enzymic treatments [5]. A colo-
rimetric method is used to quantitate peptidoglycan by measuring
lactic acids released from the muramic acid residues of peptidogly-
can upon acidic and alkaline hydrolysis [7]. We pay particular
attention to the hydrophobic nature of purified peptidoglycan sac-
culi (presumably originating from lipid-linked disaccharide precur-
sors that remain at the ends of glycan strands) [8]. Special
experimental procedures are followed to ease peptidoglycan han-
dling and overcome nonspecific interactions in co-sedimentation
experiments.

2  Materials

2.1  In Vivo 1. Phosphate buffered saline (PBS): 150 mM NaCl, 40 mM

Cross-Linking sodium phosphate, pH 7.5.
2. Sodium citrate buffer: 5 mM sodium citrate, pH 5.0.
3. DTSSP solution: 10 mM DTSSP in sodium citrate buffer.
Dissolve DTSSP in sodium citrate buffer. Prepare fresh prior to
each experiment.
4. Stop solution (20×): 1 M Tris–HCl, pH 8.0.
5. SDS solution (2×): 8% SDS in water.
6. Boiling-water bath.
7. Ultracentrifuge capable of 130,000 × g.

2.2  Purification 1. SDS solution: 8% SDS in water.

of Peptidoglycan 2. Tris–HCl buffer (100×): 1 M Tris–HCl, pH 7.0.
3. α-amylase stock solution (100×): 10 mg/mL α-amylase in 10
mM Tris–HCl, pH 7.0. Store at −20 °C.
4. Pronase stock solution (100×): 20 mg/mL pronase in 10 mM
Tris–HCl, pH 7.0. Store at −20 °C. Incubate stock solution at
60 °C for 2 h to inactivate possible muramidase contamination
before use.
 Protein-Peptidoglycan Interactions 145

5. Magnetic stirrer with incorporated hotplate.

6. Ultracentrifuge capable of 130,000 × g.

2.3  Muramic Acid 1. H2SO4 hydrolysis solution: 5 M H2SO4 in water.

Assay of 2. NaOH neutralization solution: 10 M NaOH in water.
Peptidoglycan
3. Concentrated H2SO4 (18.8 M).
4. CuSO4 solution: 4% (w/v) CuSO4∙5H2O in water.
5. 4-phenylphenol solution: 1.5% (w/v) 4-phenylphenol in etha-
nol. Store at −20 °C.
6. Muramic acid standard solutions: 0–1 mM muramic acid in
water. Store at −20 °C.
7. Spectrophotometer capable of 570 nm wavelength.
8. Sulfuric-acid-resistant cuvettes (glass or quartz).
9. Boiling-water bath.
10. Fume hood and personal protective equipment.

2.4  Co-sedimenta- 1. Tween 20 stock solution (10×): 0.5% Tween 20 in water. Store
tion Assay at 4 °C.
2. Binding buffer (10×): 400 mM sodium phosphate, pH 6.5.
3. Bovine albumin stock solution (100×): 1 mg/mL bovine albu-
min in water. Store at −20 °C.
4. Refrigerated microcentrifuge.

3  Methods

3.1  In Vivo 1. Grow bacterial strains under conditions in which the proteins
Cross-Linking of interest are produced and functional. For each cross-linking
experiment, 5–10 mL culture is generally sufficient. A. hydroph-
ila strains cultured in buffered Luria-Bertani (LB) medium are
used in this protocol [4]. In addition to the wild-­ type A.
hydrophila strain, cells expressing ExeA variants that contain
substitution mutations in the putative peptidoglycan binding
domain are included as controls (see Note 1).
2. Pellet cells by centrifugation at 6000 × g for 5 min. Wash twice
with PBS and resuspend cells in PBS. Adjust cell suspension to
an OD600 of 2.0. Transfer 1 mL cells to microcentrifuge tubes
for cross-linking. To avoid cold shock that may affect cell enve-
lope structure and physiology, perform the preceding steps at
room temperature.
3. Add fresh 10 mM DTSSP solution to a final concentration of
0.5 mM (or a range of 0.1–1 mM in initial experiments). Mix
immediately by inverting the tube several times. Incubate the
146 Gang Li and S. Peter Howard

mixtures at room temperature for 5 min or for a range of

2–10 min in initial experiments (see Note 2).
4. Add 1 M Tris–HCl, pH 8.0 solution to a final concentration of
50 mM to stop the cross-linking reaction and quench excessive
cross-linker. Incubate the mixtures at room temperature for
15 min.
5. Take aliquots of each cross-linked sample (whole-cell samples)
for later SDS-PAGE analysis. Store at −20 °C.
6. Add the rest of the cross-linked samples dropwise to an equal
volume of 8% SDS solution in glass tubes preincubated in a
boiling-water bath. Incubate the samples for 15 min with
intervals of vigorous vortexing. Cool the samples to room
temperature (may leave on bench overnight).
7. Pellet peptidoglycan by ultracentrifugation at 130,000 × g at
room temperature for 1 h (see Note 3).
8. Resuspend pellets in 0.5 mL of water by vigorous vortexing.
Repeat SDS-boiling and centrifugation steps two additional
times.
9. Resuspend final pellets in 0.1 mL water (peptidoglycan
samples).
10. Mix aliquots of peptidoglycan samples with 2× SDS-PAGE
sample buffer containing 0 or 10% β-mercaptoethanol. Heat
samples to 95 °C for 5 min.
11. Analyze whole-cell samples and peptidoglycan samples by rou-
tine SDS-PAGE and immunoblot to detect proteins of interest
(see Note 4).

3.2  Purification 1. Grow A. hydrophila or E.coli cells to late exponential phase in

of Peptidoglycan LB or other appropriate medium. Typically 1 L culture yields
from Gram-Negative 2–5 mg purified peptidoglycan.
Bacteria 2. Pellet bacterial cells in 250–500 mL centrifuge bottles at 6000
× g at 4 °C for 10 min. Resuspend cells in 20 mL ice-cold water
per liter bacterial culture.
3. Slowly pour resuspended cells into an equal volume of 8% SDS
solution stirred in a beaker in a boiling-water bath. Incubate
mixture with continuous stirring for 1 h. Cool sample to room
temperature (may leave stirring overnight at room
temperature).
4. Pellet the peptidoglycan by ultracentrifugation at 130,000 × g
at room temperature for 1 h (see Note 3).
5. Resuspend peptidoglycan in 20 mL water (room temperature)
with vigorous vortexing. Pour sample into boiling 8% SDS
solution and incubate for 15 min.
 Protein-Peptidoglycan Interactions 147

6. Wash the peptidoglycan four times with room temperature

water, centrifuging 1 h at 130,000 × g for each wash to remove
residual SDS (see Note 5).
7. Resuspend peptidoglycan in 10 mL water. Add 10 mM Tris–
HCl, pH 7.0, and 0.1 mg/mL α-amylase. Incubate at 37 °C
for 2 h to hydrolyze glycogen trapped in peptidoglycan
sacculi.
8. Add 0.2 mg/mL preincubated pronase and incubate at 60 °C
for 90 min to hydrolyze proteins associated with
peptidoglycan.
9. Add mixture to an equal volume of boiling 8% SDS solution
and incubate for 15 min in a boiling-water bath.
10. Wash peptidoglycan four times with water at room tempera-
ture as described earlier.
11. Resuspend purified peptidoglycan in 2 mL water. Store at 4 °C.
Do not freeze (see Note 6).

3.3  Muramic Acid Warning: This colorimetric method uses strong acid and alkaline
Assay of solutions. Wear suitable personal protective equipment and follow
Peptidoglycan laboratory safety guidelines. Alternative methods using high-­
performance liquid chromatography or quantitative aminosugar
analysis are described elsewhere [5, 9].

1. Add 80 μL peptidoglycan in water (see Note 7) to an equal

volume of 5 M H2SO4 solution and incubate at 90 °C for 2 h
to hydrolyze the peptidoglycan. Include muramic acid solu-
tions (0–1 mM) in parallel to generate a standard curve.
2. Add 360 μL water and 140 μL 10 M NaOH solution. Incubate
at 37 °C for 30 min to release lactic acids from muramic acid
residues of peptidoglycan.
3. Transfer 300 μL of the hydrolyzed samples to clean glass tubes
in duplicate. Add 2 mL concentrated H2SO4 (18.8 M). Cap
tubes and vortex to mix (see Note 8).
4. Incubate glass tubes in a boiling-water bath for 5 min. Cool
tubes in a room-temperature water bath.
5. Add 20 μL 4% CuSO4 solution and 40 μL 1.5% 4-phenylphe-
nol solution. Cap and vortex.
6. Incubate samples at 30 °C for 30 min to develop a blue color
that is stable for at least 1 h.
7. Measure absorbance at 570 nm in glass or quartz cuvettes
(see Note 9). Average the readings of duplicate samples and
determine the muramic acid concentration using the standard
curve. The peptidoglycan preparation generally contains 0.5–2
mM muramic acid.
148 Gang Li and S. Peter Howard

3.4  Co-sedimenta- 1. For co-sedimentation of peptidoglycan and purified ExeA pro-

tion of Peptidoglycan tein, the following binding conditions are used: 100 μM
and Proteins muramic acid units of peptidoglycan sacculi, 0.05% Tween 20,
of Interest 40 mM sodium phosphate buffer, pH 6.5, 10 μg/mL bovine
albumin, and variable amounts of ExeA in a reaction volume of
150 μL. The co-sedimentation procedure is optimized to over-
come hydrophobic aggregation of peptidoglycan sacculi and
nonspecific binding of proteins. Add each component in the
following order.
2. Add peptidoglycan and water to 1.5 mL microcentrifuge tubes.
3. Add Tween 20 stock solution (10×) and vortex.
4. Add sodium phosphate binding buffer (10×) and mix.
5. Add bovine albumin stock solution (100×) and mix.
6. Add purified proteins of interest and mix (see Note 10).
7. Incubate the mixtures at 4 °C for 1 h.
8. Pellet the peptidoglycan sacculi and associated proteins in a
microcentrifuge at top speed (21,000 × g) at 4 °C for 1 h.
9. Add 1× SDS-PAGE sample buffer to the peptidoglycan pellets.
Incubate the samples at 95 °C for 5 min. Vortex vigorously to
resuspend peptidoglycan and release associated proteins.
10. Analyze the mixture samples (before centrifugation), supernatant
samples, and pellet samples by SDS-PAGE and immunoblot.

4  Notes

1. A general consideration in cross-linking experiments is that the

cross-linking reaction may result in artefactual linkage of non-
complexed components. It is therefore critical to include suit-
able controls to exclude such a possibility. The most d­ esirable
controls are mutant proteins that contain deletion or substitu-
tion mutations in the peptidoglycan binding motif.
2. The cross-linker concentration and reaction time will need to
be optimized for different experiments.
3. Store and centrifuge peptidoglycan–SDS mixtures at room tem-
perature to avoid precipitation of SDS at low temperatures.
4. Cross-linked proteins migrate as higher-molecular-weight
complexes in SDS-PAGE. Proteins cross-linked to peptidogly-
can sacculi are not able to enter gels unless the cross-linkers are
cleaved by β-mercaptoethanol treatment prior to electrophore-
sis. See [4] for analysis of protein–peptidoglycan cross-­linking
in detail.
5. Resuspend the peptidoglycan pellets first in a small volume of
water by vortexing. Add water to full volume before
ultracentrifugation.
 Protein-Peptidoglycan Interactions 149

6. Purified peptidoglycan sacculi form large aggregates when fro-

zen and thawed. It is difficult to break up the aggregates by
vortexing without using sonication, which in turn fragments
the peptidoglycan and reduces yields following centrifugation.
The sacculi also aggregate in the presence of buffers or salts. It
is thus preferable to store peptidoglycan sacculi in pure water
at 4 °C.
7. It is important that the peptidoglycan samples do not contain
chloride salts (e.g., NaCl), which can result in the release of
hydrogen chloride gas during heating with concentrated
H2SO4 [10].
8. The tubes do not need to be sealed in an airtight manner.
However, exercise caution when handling concentrated
H2SO4.
9. Check if the cuvettes are sulfuric-acid compatible. The cuvettes
need to be dry. If the cuvettes are wet, precipitation may occur
and interfere with accurate absorbance readings. If sufficient
numbers of cuvettes are not available, rinse the cuvettes with 1
mL concentrated H2SO4 (18.8 M) with a pipette before add-
ing the next sample.
10. We do not suggest adding crude bacterial lysate because the
lysate may contain peptidoglycan hydrolases.

References

1. Höltje JV (1998) Growth of the stress-bearing 6. Li G, Howard SP (2010) ExeA binds to pepti-
and shape-maintaining murein sacculus of doglycan and forms a multimer for assembly of
Escherichia coli. Microbiol Mol Biol Rev the type II secretion apparatus of Aeromonas
62:181–203 hydrophila. Mol Microbiol 76:772–781
2. Dijkstra AJ, Keck W (1996) Peptidoglycan as a 7. Hoijer MA, Melief MJ, van Helden-Meeuwsen
barrier to transenvelope transport. J Bacteriol CG, Eulderink F, Hazenberg MP (1995)
178:5555–5562 Detection of muramic acid in a carbohydrate
3. Scheurwater EM, Burrows LL (2011) fraction of human spleen. Infect Immun
Maintaining network security: how 63:1652–1657
­macromolecular structures cross the peptido- 8. Typas A, Banzhaf M, Gross CA, Vollmer W
glycan layer. FEMS Microbiol Lett 318:1–9 (2011) From the regulation of peptidoglycan
4. Howard SP, Gebhart C, Langen GR, Li G, synthesis to bacterial growth and morphology.
Strozen TG (2006) Interactions between pep- Nat Rev Microbiol 10:123–136
tidoglycan and the ExeAB complex during 9. Clarke AJ (1993) Compositional analysis of
assembly of the type II secretin of Aeromonas peptidoglycan by high-performance anion-­
hydrophila. Mol Microbiol 59:1062–1072 exchange chromatography. Anal Biochem
5. Glauner B (1988) Separation and quantifica- 212:344–350
tion of muropeptides with high-performance
10. Sulfuric Acid (2015) The Columbia
liquid chromatography. Anal Biochem Encyclopedia, 6th edn. http:// www.encyclo-
172:451–464 pedia.com. Accessed 26 Jan 2016
 Chapter 12

Measure of Peptidoglycan Hydrolase Activity

Yoann G. Santin and Eric Cascales

Abstract
Most gene clusters encoding multiprotein complexes of the bacterial cell envelope, such as conjugation
and secretion systems, Type IV pili, and flagella, bear a gene encoding an enzyme with peptidoglycan
hydrolase activity. These enzymes are usually glycoside hydrolases that cleave the glycan chains of the pep-
tidoglycan. Their activities are spatially controlled to avoid cell lysis and to create localized rearrangement
of the cell wall. This is assured by interaction with the structural subunits of the apparatus. Here we
describe protocols to test the peptidoglycan hydrolase activity of these proteins in vitro and in solution.

Key words Cell wall, Localized degradation, Peptidoglycan, Lytic transglycosylase, Remazol blue

1  Introduction

The peptidoglycan is a mesh-like structure that provides the shape

and protection against external pressure to bacterial cells. It is
composed of glycan chains resulting from the polymerization of
N-acetylmuramic acid (MurNAc)-N-acetylglucosamine (GlcNAc)
disaccharides. These chains are linked by peptide stems that differ
from one species to another. With pores of approximately 2 nm,
the cell wall constitutes a physical barrier for the passage of mac-
romolecules and for the assembly of cell-envelope-spanning com-
plexes [1–3]. Most trans-envelope multiprotein machineries
therefore have evolved dedicated enzymes that locally degrade the
cell wall to provide sufficient space for their assembly and inser-
tion without compromising the bacterial shape and survival [3, 4].
These enzymes usually cleave the β-1,4 bond between the
N-acetylmuramic acid and the N-acetylglucosamine of the glycan
chains and form nonreducing 1,6-anhydromuropeptides charac-
teristic of lytic transglycosylases (LTGs) [4–7]. Genes encoding
these enzymes are found associated in Type III secretion, Type IV
secretion, or flagellum gene clusters [3, 4, 6]. The best-studied
specialized LTGs are FlgJ and SltF, which are associated with fla-
gellar assembly [8–10], and EtgA, VirB1, and TagX/MltE,

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_12, © Springer Science+Business Media LLC 2017

151
152 Yoann G. Santin and Eric Cascales

which are necessary for the biogenesis of Type III (T3SS), Type
IV (T4SS), and Type VI (T6SS) secretion systems, respectively [6,
8–18]. However, these enzymes could be deleterious for bacterial
cells, and therefore their activity needs to be restricted to the site
of assembly in order to avoid breaches in the cell wall. Studies
have provided evidence that LTGs are recruited via specific inter-
actions to subunits of the machine [18–21] and, in a few cases,
that these interactions stimulate LTG activity [18, 21, 22].
Methods have been developed and used to test whether puta-
tive LTGs have peptidoglycan hydrolase activities. An indirect
approach is to clone the gene encoding the putative LTG to a signal
sequence in order to address the protein to the periplasm of E. coli
and follow the cell growth after induction as overproduction of the
LTG causes cell lysis [23, 24]. More direct protocols have been
developed using purified LTGs, including zymogram [25, 26].
However, this technique, which consists of subjecting purified
LTGs to sodium dodecyl sulfate-polyacrylamide gel electrophoresis
(SDS-PAGE) in a gel supplemented with purified peptidoglycan,
has limits, such as the refolding of the protein after migration.
Additional approaches, performed in solution, do not require the
denaturation and refolding steps. These turbidometric assays,
detailed in what follows, are methods for following the activity of
purified LTGs on peptidoglycan or peptidoglycan labeled with
remazol brilliant blue (RBB) dye [27, 28]. The peptidoglycan assay
relies on the decrease of absorbance of the peptidoglycan solution
[27], whereas the RBB assay relies on the release of the dye cap-
tured in the peptidoglycan net [28] in the presence of the LTG. In
addition, more precise approaches, such as the analysis of peptido-
glycan degradation products released after incubation of the pepti-
doglycan with the purified protein by reverse-phase high-­performance
liquid chromatography coupled to mass spectrometry [29, 30],
make it possible to define the site of cleavage of the enzyme.

2  Material

2.1  Peptidoglycan 1. 8% SDS solution: Resuspend 8 g SDS resuspended in 100 mL

Purification sterile distilled water.
2. 20 mM Tris–HCl, pH 8.0, 100 mM NaCl: Dissolve 2.43 g
Tris(hydroxymethyl) aminomethane and 5.84 g NaCl in 1 L
sterile distilled water. Adjust pH to 8.0 with 1 M HCl.
3. 20 mM Tris–HCl, pH 7.2, 50 mM NaCl: Dissolve 2.43 g
Tris(hydroxymethyl) aminomethane and 2.92 g NaCl in 1 L
sterile distilled water. Adjust pH to 7.2 with 1 M HCl.
4. 0.5 M NaCl: Dissolve 29.22 g NaCl in 1 L sterile distilled
water.
 Peptidoglycan Remodelling Enzymes 153

5. α-amylase stock solution (100×): 20 mg/mL α-amylase in 20

mM Tris–HCl, pH 7.2. Store at −20 °C.
6. Pronase stock solution (100×): 20 mg/mL pronase in 20 mM
Tris–HCl, pH 7.2. Incubate pronase stock solution for 1 h at
56 °C. Store at −20 °C.
7. French press, Emulsiflex apparatus or any apparatus to disrupt
bacterial cells.
8. Vortex.
9. Water bath at 96 °C.
10. Incubate at 37 °C.
11. Ultracentrifuge (Beckman Coulter (Brea, CA) with TLA100.3
and TLA100.4 rotors, or equivalent).

2.2  Turbidometric 1. 60 mM MES, pH 6.0, 180 mM NaCl buffer: Dissolve 11.71 g

Analyses 2-(N-morpholino)ethanesulfonic acid and 10.52 g NaCl in
of Peptidoglycan 1 L sterile distilled water.
Degradation 2. Purified protein to be tested.
3. Lysozyme stock solution: 10 mg/mL egg-white lysozyme in
sterile distilled water.
4. Incubate at 37 °C.
5. Spectrophotometer.

2.3  Peptidoglycan 1. 400 mM NaOH: Dissolve 16 g NaOH in 1 L sterile distilled

Labeling with Remazol water.
Brilliant Blue 2. RBB stock solution (10×): Dissolve 1.566 g RBB R (Sigma-­
Aldrich, St. Louis, MO) in 10 mL sterile distilled water.
3. 1 M HCl: Dilute 10 mL HCl 37% solution (10 M) with 90 mL
distilled water.
4. Phosphate buffered saline (PBS) buffer: Dissolve 1.44 g
Na2HPO4, 0.24 g KH2PO4, 0.2 g KCl, and 8 g NaCl in 1 L
sterile distilled water. Adjust pH to 7.4 with 1 M HCl.
5. Incubate at 37 °C.
6. Vortex.
7.
Ultracentrifuge (Beckman with TLA100.3 rotor, or
equivalent).

2.4  RBB-Labeled 1. PBS buffer: Dissolve 1.44 g Na2HPO4, 0.24 g KH2PO4, 0.2 g
Peptidoglycan KCl, and 8 g NaCl in 1 L sterile distilled water. Adjust pH to
Degradation Assay 7.4 with 1 M HCl.
2. Purified protein to be tested.
3. Ethanol 96° or absolute.
4. Lysozyme stock solution: 10 mg/mL egg-white lysozyme in
sterile distilled water.
154 Yoann G. Santin and Eric Cascales

5. Incubate at 37 °C.
6.
Ultracentrifuge (Beckman with TLA100.3 rotor, or
equivalent).
7. Spectrophotometer.

3  Methods

3.1  Peptidoglycan The peptidoglycan purification protocol is adapted from [31, 32].
Purification 1. Grow cells in 400 mL of appropriate medium until culture
reaches an A600 of about 1–1.2.
2. Harvest cells by centrifugation at 10,000 × g for 20 min at
4 °C. Resuspend cells in 20 mL 20 mM Tris–HCl, pH 8.0,
100 mM NaCl. Break cells by three passages at French press or
using an Emulsiflex apparatus.
3. Pellet cell envelopes by centrifugation at 400,000 × g
(90,000 rpm in a Beckman TLA-100.4 rotor) for 45 min at
4 °C. Resuspend cells in 10 mL 0.5 M NaCl.
4. Add 10 mL 8% SDS and incubate for 1 h at 96 °C.
5. Leave solution at room temperature overnight.
6. Pellet peptidoglycan by ultracentrifugation at 400,000 × g at
25 °C for 45 min (see Note 1).
7. Resuspend peptidoglycan fraction in 10 mL 0.5 M NaCl and
add 10 mL 8% SDS. Incubate for 30 min at 96 °C.
8. Pellet peptidoglycan by ultracentrifugation at 400,000 × g at
25 °C for 30 min and resuspend peptidoglycan in 10 mL water.
9. Repeat step 8 twice.
10. Resuspend peptidoglycan in 10 mL 20 mM Tris–HCl, pH 7.2,
50 mM NaCl supplemented with 200 μg/mL α-amylase and
200 μg/mL pronase. Incubate overnight at 37 °C.
11. Add 10 mL 8% SDS and incubate for 1 h at 96 °C.
12. Pellet peptidoglycan by ultracentrifugation at 400,000 × g at
25 °C for 30 min and resuspend peptidoglycan in 10 mL water.
13. Repeat step 12 twice.
14. Resuspend peptidoglycan pellet in 1 mL water. Store at 4 °C.

3.2  Turbidometric 1. Dilute 125 μL of the purified peptidoglycan suspension

Analyses of obtained at step 14 in Subheading 3.1 with 875 μL 60 mM
Peptidoglycan MES, pH 6.0, 180 mM NaCl, and incubate at 37 °C for
Degradation 30 min. Use three tubes for each reaction to measure peptido-
glycan hydrolysis in triplicate.
2. Measure the A600 for each tube (see Note 2).
 Peptidoglycan Remodelling Enzymes 155

Fig. 1 LTG activity measured by peptidoglycan hydrolysis assay. A representative

example of peptidoglycan degradation is shown. Purified peptidoglycan was incubated
with buffer (open square) or purified LTG (closed circles), and the absorbance at
600 nm (A600) was measured every 20 min. The difference of absorbance at time
zero (t0 ) minus the absorbance at time t (∆A600) was plotted against time (in minutes)

3. Add 2–5 nmol of protein to be tested to each tube and incu-

bate at 37 °C (see Note 3).
4. Measure the A600 every 10 min and plot the difference of
absorbance (initial absorbance subtracted from the absorbance
at time t) against time (see Note 4).
A typical example of the turbidometric peptidoglycan assay is
shown in Fig. 1.

3.3  Peptidoglycan The peptidoglycan labeling protocol is adapted from [28].

Labeling with Remazol
1. Mix 250 μL of the purified peptidoglycan fraction obtained at
Brilliant Blue
step 14 in Subheading 3.1 with 250 μL 400 mM NaOH and
incubate for 30 min at 37 °C.
2. Add RBB dye to mixture at final concentration of 25
mM. Vortex and incubate mixture overnight at 37 °C.
3. Add 500 μL 1 M HCl and mix by vortexing.
4. Pellet peptidoglycan by ultracentrifugation at 400,000 × g at
25 °C for 30 min and resuspend peptidoglycan in 2 mL water.
5. Repeat step 3 twice.
6. Resuspend peptidoglycan pellet in 250 μL PBS buffer. Store at
4 °C.

3.4  RBB-Labeled 1. Dilute 10 μL RBB-labeled peptidoglycan obtained at step 5 in

Peptidoglycan Subheading 3.3 with 90 μL PBS buffer and incubate at 37 °C
Degradation Assay for 30 min. Use nine tubes for each reaction to measure pepti-
doglycan hydrolysis in triplicate, at three different times.
156 Yoann G. Santin and Eric Cascales

Fig. 2 LTG activity measured by RBB release assay. A representative example of

peptidoglycan degradation is shown. (a) RBB-labeled peptidoglycan was incubated
with buffer (open bars) or purified LTG (+LTG, blue bars) and the absorbance at
595 nm (A595) of the supernatant was measured after 0.5, 1, and 4 h of incubation.
(b) Photographs of supernatant fractions of RBB-labeled peptidoglycan incubated
with buffer (left tube) or purified LTG (+LTG) (right tube) after 4 h of incubation

2. Add 0.2–0.5 nmol of protein to be tested to mixture and incu-

bate at 37 °C (see Note 3). This step corresponds to time zero.
3. Add 100 μL ethanol in three tubes 30 min after time zero to
quench reaction.
4. Pellet peptidoglycan by ultracentrifugation at 400,000 × g at
25 °C for 30 min.
5. Measure A595 of the supernatant.
6. At 1 and 4 h after time zero, repeat steps 3–5.
A typical example of the dye release assay is shown in Fig. 2.

4  Notes

1. Do not incubate at 4 °C to avoid SDS precipitation.

2. Typically, an A600 ~ 0.4–0.7 is measured from peptidoglycan
purified from E. coli.
3. Control assays include incubation of the peptidoglycan sus-
pension with (i) buffer and (ii) purified lysozyme. Ideally, addi-
tional controls include incubation of the peptidoglycan with
(a) the protein to be tested but bearing amino-acid substitu-
tions in the catalytic site (if known or predicted) and (b) the
wild-type protein in presence of 100 μM of bulgecin A, an
inhibitor of lytic transglycosylases [33].
4. The initial rate of the hydrolysis reaction (in AU/min/mol)
can be calculated from the slope of the initial linear curve.
 Peptidoglycan Remodelling Enzymes 157

Acknowledgements

Work in EC laboratory is supported by the Centre National de la

Recherche Scientifique, the Aix-Marseille Université, and grants
from the Agence Nationale de la Recherche (ANR-­
14-­CE14-0006-02 and ANR-15-CE11-0019-01).

References

1. Demchick P, Koch AL (1996) The perme- 12. Kohler PL, Hamilton HL, Cloud-Hansen
ability of the wall fabric of Escherichia coli and K, Dillard JP (2007) AtlA functions as a
Bacillus subtilis. J Bacteriol 178:768–773 peptidoglycan lytic transglycosylase in the
2. Scheurwater E, Reid CW, Clarke AJ (2008) Neisseria gonorrhoeae type IV secretion system.
Lytic transglycosylases: bacterial space-making J Bacteriol 189:5421–5428
autolysins. Int J Biochem Cell Biol 40:586–591 13. Zhong Q, Shao S, Mu R, Wang H,
3. Scheurwater EM, Burrows LL (2011) Huang S, Han J, Huang H, Tian S (2011)
Maintaining network security: how macromo- Characterization of peptidoglycan hydrolase in
lecular structures cross the peptidoglycan layer. Cag pathogenicity island of Helicobacter pylori.
FEMS Microbiol Lett 318:1–9 Mol Biol Rep 38:503–509
4. Koraimann G (2003) Lytic transglycosyl- 14. García-Gómez E, Espinosa N, de la Mora J,
ases in macromolecular transport systems Dreyfus G, González-Pedrajo B (2011) The
of gram-­negative bacteria. Cell Mol Life Sci muramidase EtgA from enteropathogenic
60:2371–2388 Escherichia coli is required for efficient type III
5. Höltje JV (1996) Lytic transglycosylases. EXS secretion. Microbiology 157:1145–1160
75:425–429 15. Arends K, Celik EK, Probst I, Goessweiner-­
6. Zahrl D, Wagner M, Bischof K, Bayer M, Mohr N, Fercher C, Grumet L, Soellue C,
Zavecz B, Beranek A, Ruckenstuhl C, Zarfel Abajy MY, Sakinc T, Broszat M, Schiwon K,
GE, Koraimann G (2005) Peptidoglycan deg- Koraimann G, Keller W, Grohmann E (2013)
radation by specialized lytic transglycosylases TraG encoded by the pIP501 type IV secre-
associated with type III and type IV secretion tion system is a two-domain peptidoglycan-­
systems. Microbiology 151:3455–3467 degrading enzyme essential for conjugative
transfer. J Bacteriol 195:4436–4444
7. van Heijenoort J (2011) Peptidoglycan hydro-
lases of Escherichia coli. Microbiol Mol Biol 1 6. Laverde Gomez JA, Bhatty M, Christie PJ
Rev 75:636–663 (2014) PrgK, a multidomain peptidoglycan
hydrolase, is essential for conjugative transfer
8. de la Mora J, Ballado T, González-Pedrajo B, of the pheromone-responsive plasmid pCF10.
Camarena L, Dreyfus G (2007) The flagel- J Bacteriol 196:527–539
lar muramidase from the photosynthetic bac-
terium Rhodobacter sphaeroides. J Bacteriol 1 7. Weber BS, Hennon SW, Wright MS, Scott NE,
189:7998–8004 de Berardinis V, Foster LJ, Ayala JA, Adams MD,
Feldman MF (2016) Genetic dissection of the
9. de la Mora J, Osorio-Valeriano M, González-­ Type VI secretion system in Acinetobacter and
Pedrajo B, Ballado T, Camarena L, Dreyfus G identification of a novel peptidoglycan hydro-
(2012) The C terminus of the flagellar muram- lase, TagX, required for its biogenesis. MBio
idase SltF modulates the interaction with
FlgJ in Rhodobacter sphaeroides. J Bacteriol 1 8. Santin YG, Cascales E (2016) Domestication
194:4513–4520 of a housekeeping transglycosylase for assem-
bly of a type VI secretion system. EMBO Rep

10. Nambu T, Minamino T, Macnab RM, 18(1):138–149
Kutsukake K (1999) Peptidoglycan-­
hydrolyzing activity of the FlgJ protein, essen- 1 9. Höppner C, Carle A, Sivanesan D, Hoeppner
tial for flagellar rod formation in Salmonella S, Baron C (2005) The putative lytic trans-
typhimurium. J Bacteriol 181:1555–1561 glycosylase VirB1 from Brucella suis interacts
with the type IV secretion system core compo-
11. Mushegian AR, Fullner KJ, Koonin EV,
nents VirB8, VirB9 and VirB11. Microbiology
Nester EW (1996) A family of lysozyme-like 151:3469–3482
virulence factors in bacterial pathogens of
plants and animals. Proc Natl Acad Sci U S A 2 0. Creasey EA, Delahay RM, Daniell SJ, Frankel
93:7321–7326 G (2003) Yeast two-hybrid system survey
158 Yoann G. Santin and Eric Cascales

of interactions between LEE-encoded pro- 27. Fibriansah G, Gliubich FI, Thunnissen AM

teins of enteropathogenic Escherichia coli. (2012) On the mechanism of peptidoglycan
Microbiology 149:2093–2106 binding and cleavage by the endo-specific lytic
21. Burkinshaw BJ, Deng W, Lameignère E, transglycosylase MltE from Escherichia coli.
Wasney GA, Zhu H, Worrall LJ, Finlay BB, Biochemistry 51:9164–9177
Strynadka NC (2015) Structural analy- 28. Uehara T, Parzych KR, Dinh T, Bernhardt TG
sis of a specialized type III secretion system (2010) Daughter cell separation is controlled
peptidoglycan-­cleaving enzyme. J Biol Chem by cytokinetic ring-activated cell wall hydroly-
290:10406–10417 sis. EMBO J 29:1412–1422
22. Herlihey FA, Osorio-Valeriano M, Dreyfus G, 29. Scheurwater EM, Clarke AJ (2008) The
Clarke AJ (2016) Modulation of the lytic activ- C-terminal domain of Escherichia coli YfhD
ity of the dedicated autolysin for flagellum for- functions as a lytic transglycosylase. J Biol
mation SltF by flagellar rod proteins FlgB and Chem 283:8363–8373
FlgF. J Bacteriol 198:1847–1856 30. Clarke AJ (1993) Compositional analysis of
23. Engel H, Kazemier B, Keck W (1991) Murein-­ peptidoglycan by high-performance anion-­
metabolizing enzymes from Escherichia coli: exchange chromatography. Anal Biochem
sequence analysis and controlled overexpression 212:344–350
of the slt gene, which encodes the soluble lytic 31. Leduc M, Joseleau-Petit D, Rothfield LI

transglycosylase. J Bacteriol 173:6773–6782 (1989) Interactions of membrane lipopro-
24. Lommatzsch J, Templin MF, Kraft AR, teins with the murein sacculus of Escherichia
Vollmer W, Höltje JV (1997) Outer membrane coli as shown by chemical crosslinking stud-
localization of murein hydrolases: MltA, a third ies of intact cells. FEMS Microbiol Lett
lipoprotein lytic transglycosylase in Escherichia 51:11–14
coli. J Bacteriol 179:5465–5470 32. Cascales E, Lloubès R (2004) Deletion analy-
25. Leclerc D, Asselin A (1989) Detection of ses of the peptidoglycan-associated lipopro-
bacterial cell wall hydrolases after denatur- tein Pal reveals three independent binding
ing polyacrylamide gel electrophoresis. Can sequences including a TolA box. Mol Microbiol
J Microbiol 35:749–753 51:873–885
26. Bernadsky G, Beveridge TJ, Clarke AJ (1994) 33. Imada A, Kintaka K, Nakao M, Shinagawa
Analysis of the sodium dodecyl sulfate-sta- S (1982) Bulgecin, a bacterial metabolite
ble peptidoglycan autolysins of select gram-­ which in concert with beta-lactam antibiotics
negative pathogens by using renaturing causes bulge formation. J Antibiot (Tokyo)
polyacrylamide gel electrophoresis. J Bacteriol 35:1400–1403
176:5225–5232
 Chapter 13

Protein–Protein Interaction: Bacterial Two-Hybrid

Gouzel Karimova, Emilie Gauliard, Marilyne Davi, Scot P. Ouellette,
and Daniel Ladant

Abstract
The bacterial two-hybrid (BACTH, for “Bacterial Adenylate Cyclase-Based Two-Hybrid”) system is a
simple and fast genetic approach to detecting and characterizing protein–protein interactions in vivo. This
system is based on the interaction-mediated reconstitution of a cyclic adenosine monophosphate (cAMP)
signaling cascade in Escherichia coli. As BACTH uses a diffusible cAMP messenger molecule, the physical
association between the two interacting chimeric proteins can be spatially separated from the transcription
activation readout, and therefore it is possible to analyze protein–protein interactions that occur either in
the cytosol or at the inner membrane level as well as those that involve DNA-binding proteins. Moreover,
proteins of bacterial origin can be studied in an environment similar (or identical) to their native one. The
BACTH system may thus permit a simultaneous functional analysis of proteins of interest—provided the
hybrid proteins retain their activity and their association state. This chapter describes the principle of the
BACTH genetic system and the general procedures to study protein–protein interactions in vivo in E. coli.

Key words Two-hybrid system, Protein interaction assay, Membrane protein, cAMP signaling,
Chimeric proteins

1  Introduction

Two-hybrid systems are genetic assays that permit the detection

and characterization of protein–protein interactions in vivo. This
approach was pioneered by Fields and Song, who described the
original yeast two-hybrid system [1]. All of the two-hybrid tech-
niques discussed in what follows are based on the co-expression, in
the same cell, of two hybrid proteins that, upon interaction, pro-
duce a phenotypic or selective trait [2]. In the bacterial two-hybrid
(BACTH, for “bacterial adenylate cyclase-based two-hybrid”) sys-
tem, the readout of the interactions relies on the complementation
between two fragments from the adenylate cyclase of Bordetella
pertussis to reconstitute a cyclic adenosine monophosphate (cAMP)
signaling cascade in Escherichia coli [3]. Because it exploits a cAMP
signaling cascade, the BACTH system can be easily applied to

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_13, © Springer Science+Business Media LLC 2017

159
160 Gouzel Karimova et al.

study interactions between membrane proteins [4], and it has

indeed been widely used to characterize the assembly of bacterial
secretion systems. These specialized nanomachines, which bacteria
use to secrete a wide variety of compounds (e.g., small molecules,
sugars, proteins, DNA), consist of up to tens of proteins that
assemble in the bacterial membranes in multimolecular complexes.
The BACTH system has been instrumental in characterizing the
molecular interactions between these different components of vari-
ous secretion systems [5–9]. This chapter describes the principle of
this genetic system and outlines the main procedures to study pro-
tein–protein interactions in vivo in E. coli.

1.1  Principle The BACTH bacterial two-hybrid system is a simple and fast
of Bacterial Adenylate approach to detecting and characterizing protein–protein interac-
Cyclase-Based tions in vivo. It offers all the advantages of working with E. coli and
Two-Hybrid System is readily accessible to many researchers having basic knowledge in
standard microbiological and molecular biology techniques (e.g.,
plasmid preparation, bacterial transformation, polymerase chain
reaction (PCR)).
The BACTH system is based on the interaction-mediated
reconstitution of adenylate cyclase enzyme activity in an E. coli cya
mutant, defective in its endogenous adenylate cyclase [3, 10]. It
exploits the fact that the catalytic domain of adenylate cyclase
(CyaA) from B. pertussis [11] consists of two complementary frag-
ments, T25 and T18, that are not active when physically separated
(Fig. 1a). When these two fragments are fused to interacting poly-
peptides, X and Y, heterodimerization of the hybrid proteins results
in functional complementation between the T25 and T18 frag-
ments and, therefore, in cAMP synthesis (Fig. 1b). Cyclic AMP
produced by the reconstituted chimeric enzyme binds to the catab-
olite activator protein (CAP). The cAMP/CAP complex is a pleio-
tropic regulator of gene transcription in E. coli [12]. It turns on the
expression of several resident genes, including genes of the lac and
mal operons involved in lactose and maltose catabolism (Fig. 1c).
Consequently, bacteria become able to utilize lactose or maltose as
the unique carbon source and can be easily distinguished on indi-
cator or selective media [3, 10].

1.2  General Detection of in vivo interactions between two proteins of interest

Procedure with the BACTH system requires the co-expression of these pro-
teins as fusions with the T25 and T18 fragments in bacteria that
are lacking its endogenous adenylate cyclase activity (E. coli cya).
This is achieved by using two compatible vectors, one expressing
the T25 fusion (pKT25 or pKNT25), the other expressing the T18
fusion (pUT18 or pUT18C) [10, 13]. The bacteria are cotrans-
formed with the two recombinant plasmids and plated on either
indicator or selective media to reveal the resulting Cya+ phenotype
(Fig.  2). The efficiency of complementation between the two
 Bacterial Two-Hybrid Interaction Assays 161

Fig. 1 Principle of BACTH system. (a) When the two fragments of B. pertussis
adenylate cyclase, T25 and T18, are coexpressed as separate polypeptides, they
cannot assemble and there is no enzyme activity. (b) When the T25 and T18 frag-
ments are coexpressed as fusions with polypeptides X and Y that can interact,
the association of the T25-X and T18-Y hybrid proteins reconstitutes the adenyl-
ate cyclase activity. (c) Cyclic AMP synthesized by the reconstituted enzyme
binds to the catabolite activator protein (CAP), and the cAMP/CAP complex can
associate with specific promoter DNA and activates transcription of catabolite
operons (such as lac operon or mal regulon)

Fig. 2 Analysis of protein–protein interactions with the bacterial two-hybrid

­system. See text for detailed explanations
162 Gouzel Karimova et al.

hybrid proteins can be further quantified by measuring cAMP lev-

els (a direct measure of the reconstituted adenylate cyclase enzy-
matic activity) or by assaying the β-galactosidase enzymatic activities
in bacterial extracts [3, 10], an easy and robust assay that is directly
correlated with the cAMP produced in the cells since the expres-
sion of β-galactosidase is positively regulated by cAMP/CAP. The
hybrid proteins expressed in E. coli can also be characterized using
diverse biochemical approaches, for example, immunodetection,
immunoprecipitation, and copurification.
The BACTH system has been used by many different labora-
tories to detect and characterize interactions between a wide vari-
ety of bacterial, eukaryotic, or viral proteins [3, 13–16]. An
attractive aspect of this genetic assay is that, because it uses a cAMP
signaling cascade, the interaction between the hybrid proteins does
not need to take place near the transcription machinery, as is the
case with the yeast two-hybrid system or many other bacterial
­two-­hybrid systems [1, 2]. For this reason, the BACTH system is
particularly appropriate for studying interactions between mem-
brane proteins because these interactions cannot be easily tested
with transcription-based two-hybrid systems [4, 14, 17].

2  Materials

2.1  Equipment 1. Equipment for DNA cloning and bacterial transformation.

2. Incubator for plates and shaking liquid cultures.
3. 2.2 mL 96-well storage plate or deep-well storage plate,
sterile.
4. 1.2 mL polypropylene 96-well storage block or glass tubes,
sterile.
5. Microporous tape sheet, e.g., AirPore (Qiagen Co., Helden,
Germany).
6. Multichannel pipettor.
7. Shaker (for shaking deep-well 96-well blocks).
8.
Microplate reader, e.g., by Tecan Co. (Männedorf,
Switzerland), or equivalent plate reader.
9. Equipment and reagents for western blotting (optional).

2.2  Bacterial Media 1. Luria–Bertani (LB) broth: 10 g NaCl, 10 g tryptone, and 10 g

yeast extract, adjust pH to 7.0 with NaOH, add deionized
H2O to a final volume of 1 L, and autoclave.
2. LB plates: add 15 g agar per liter of LB broth and autoclave.
Allow the medium to cool to less than 45 °C, then add the
antibiotics and pour the plates.
 Bacterial Two-Hybrid Interaction Assays 163

3. LB/X-Gal plates: To prepare LB/X-Gal plates, the LB/agar

medium (above) is autoclaved, allowed to cool to less than
45 °C, and supplemented, just before pouring plates, with
40  μg/mL of the X-Gal (5-bromo-4-chloro-3-indolyl-β-d-­
galactopyranoside) chromogenic substrate and appropriate
antibiotics. Isopropyl-β-d-thiogalactopyranoside (IPTG) (final
concentration of 0.5 mM) is usually also added to the medium
to induce full expression of hybrid proteins as well as that of
the β-galactosidase reporter enzyme.
4. MacConkey/maltose medium: 40 g MacConkey agar are
­dissolved in 1 L distilled water and autoclaved (see Note 1).
A stock solution of glucose-free maltose (20% in water) is ster-
ilized by filtration. Maltose (1% final concentration) as well as
antibiotics (ampicillin at 100 μg/mL and kanamycin at 50 μg/
mL) are added to the autoclaved MacConkey medium just
before pouring plates. IPTG (final concentration of 0.5 mM)
is usually added to the medium to induce full expression of
hybrid proteins.
5. 5× M63/maltose minimal medium: 10 g (NH4)2SO4, 68 g
KH2PO4, 2.5 mg FeSO4·7H2O, add deionized H2O to final
volume of 1 L, adjust pH to 7.0 with KOH and autoclave.
When necessary, vitamin B1 is added to a final concentration of
1 μg/mL and casamino acids at 50 μg/mL.
6. M63/maltose plates: Autoclave 15 g agar in 800 mL
H2O. Then add 200 mL sterile 5× M63 medium, 0.2–0.4%
maltose, and the appropriate antibiotics at half the usual con-
centrations (i.e., 50 μg/mL ampicillin, 25 μg/mL kanamycin)
just before pouring plates.

2.3  Solutions 1. β-galactosidase assay medium (PM2): 70 mM Na2HPO4,

for β-Galactosidase 30 mM NaH2PO4, 1 mM MgSO4, 0.2 mM MnSO4, pH 7.0.
Assays Add 100 mM β-mercaptoethanol just before use (see Note 2).
2. Substrate solution: ONPG, o-nitrophenol-β-galactoside, solu-
tion of 4 mg/mL in PM2 medium without β-mercaptoethanol
(store at −20 °C).
3. Stop solution: 1 M Na2CO3.
4. Chloroform.
5. Sodium dodecyl sulfate (SDS) 0.1%: Dissolve 0.1 g SDS into
100 mL H2O.

2.4  BACTH Reporter 1. E. coli reporter strain carrying a deletion of the cya gene (see
Strains, Plasmids, Note 3).
and Antibodies 2. Set of compatible vectors allowing genetic fusions of proteins
of interest at either the N- or the C-terminus of the T25 frag-
ment (pKT25 and pKNT25) or of the T18 fragment (pUT18
and pUT18C) (see Note 4).
164 Gouzel Karimova et al.

3. Anti-CyaA monoclonal antibody (3D1, sc-13582; Santa Cruz

Biotechnology) for T18 fragment detection.
4. Rabbit polyclonal antiserum directed against purified B. pertus-
sis CyaA protein (serum L24023, DL unpublished) for T25
fragment detection.

3  Methods

3.1  General The general methodology to analyze interactions between two

Methodology proteins of interest with the BACTH system is diagrammed in
Fig. 2.
–– In a first step, clone the genes encoding the two proteins of
interest (e.g., X and Y) into the two sets of BACTH vectors
(pKT25 or pKNT25 and pUT18C or pUT18) using standard
molecular biology techniques [18] or with the Gateway®
recombineering technique [19].
–– In a second step, cotransform the recombinant plasmids encod-
ing the T25-X (or X-T25) and T18-Y (or Y-T18) hybrid pro-
teins into competent BACTH cells (DHM1, DHT1, or
BTH101), and plate the transformed cells on indicator plates
(i.e., LB-X-Gal or MacConkey media supplemented with malt-
ose) or on selective plates (synthetic medium supplemented
with maltose as unique carbon source) [3, 4, 10, 20, 21] (see
Note 1). Complementation is usually detected within 1–3 days
of incubation at 30 °C (or 37 °C, although it is usually less effi-
cient at this temperature). If no interaction occurs, colonies will
remain colorless on indicator plates or will not grow on selec-
tive plates.

3.2  Construction This section assumes that the reader has background knowledge in
of BACTH Plasmids basic molecular biology techniques. Additional protocols for
Encoding Hybrid molecular cloning, PCR, DNA analysis, and transformation can be
Proteins found in many textbooks (e.g., [18]) or on the Internet.

3.2.1  Standard Cloning 1. Design specific primers to amplify the genes encoding the pro-
of Genes Encoding Proteins teins of interest. The primers should include restriction sites
of Interest into BACTH (e.g., BamHI on 5′ primer and KpnI on 3′ primer) for allow-
Vectors ing oriented cloning of the amplified genes into the BACTH
vectors. Be careful to correctly position these restriction sites
so that the genes of interest will be in frame with the T25 and
T18 open reading frames.
2. PCR amplify the genes encoding the proteins of interest using
a standard protocol [18].
3. Purify the PCR-amplified DNA fragments using a standard
PCR purification kit (available from various companies) and
 Bacterial Two-Hybrid Interaction Assays 165

digest them with the appropriate restriction enzymes (e.g.,

BamHI and KpnI or others depending on the restriction sites
introduced into the primers). Digest the BACTH vectors with
the same restriction enzymes.
4. Ligate the digested fragments and vectors with T4 DNA ligase
[18]. Transform the ligation mixtures into competent XL1-­
Blue cells (Stratagene) and plate transformants on LB plates
supplemented with appropriate antibiotics. Incubate plates at
30 °C for 24–36 h.
5. Pick 6–12 colonies for each cloning experiment and grow
them overnight at 30 °C in 4 mL LB medium plus antibiotics
(see Note 5). Purifiy plasmid DNA using standard protocol or
commercial kit (e.g., QIAprep Spin Miniprep Kit from
Qiagen). Check recombinant plasmids by restriction analysis
and DNA sequencing to verify that no mutation was intro-
duced during PCR amplification.

3.2.2  Gateway™ Cloning The Gateway® cloning technology (Life Technologies, Thermo
of Genes Encoding Proteins Fisher Scientific) is used to transfer genes of interest to the BACTH-­
of Interest to BACTHGW Gateway destination vectors, pST25-DEST, pSNT25-DEST, and
Vectors pUT18C-DEST [17]. For detailed descriptions of the Gateway®
cloning techniques, the reader may refer to the manufacturer’s
guidelines.
1. PCR-amplify the genes of interest (from genomic DNA or
from other appropriate sources) using appropriate primers that
also contain specific attB sites (see Note 6) and purify the PCR
products as described earlier.
2. Mix the purified PCR products with the pDONR™221 plas-
mid [19], add the BP Clonase™ II enzyme and incubate 2 h at
room temperature to allow the BP recombination reaction.
Add 2 μg proteinase K to terminate recombination reaction
and transform mixture into E. coli XL1 competent cells. Select
transformants on LB plates supplemented with 50 μg/mL
kanamycin [17].
3. Purify plasmid DNA from 3 to 4 independent clones from each
cloning as described earlier, and check the recombinant plas-
mids by restriction analysis and DNA sequencing.
In the resulting plasmids (pDONR™-gene X) the genes of
interest are flanked by attL recombination sites and can be eas-
ily transferred into Gateway® destination vectors by a so-called
LR reaction [19].
4. Mix the entry pDONR™-gene X plasmid with the appropriate
BACTHGW destination vectors, pST25-DEST, pSNT25-
DEST, or pUT18C-DEST. Add the LR Clonase™ enzyme mix
166 Gouzel Karimova et al.

(see manufacturer’s guidelines) and incubate 1 h at

25 °C. Longer incubation times may be necessary for larger
inserts. Add proteinase K as described earlier to terminate
reaction.
5. Transform mixture in E. coli XL1 competent cells. Select trans-
formants on LB plates supplemented with appropriate antibi-
otic (spectinomycin or ampicillin).
6. Purify plasmid DNA from two to three independent clones
from each cloning as earlier and check recombinant plasmids
by restriction analysis or DNA sequencing.
The resulting plasmids encode the T25 or T18 fragments fused
in frame to the gene of interest (gene X).

3.3  Analysis 1. Prepare chemically competent or electro-competent DHT1,

of Interactions DHM1, or BTH101 cells by using standard procedures ([18];
by Screening see Note 7).
Procedure 2. Cotransform the BACTH competent cells with one of the
on Indicator Plates recombinant plasmids encoding the T25-fusions (pKT25,
pKNT25, pST25-DEST, or pSNT25-DEST derivatives) and
one of the recombinant plasmids encoding the T18-fusions
(pUT18, pUT18C, or pUT18C-DEST derivatives).
3. In parallel, cotransform a separate aliquot of cells with plasmids
pKT25 and pUT18C (coding for the unfused T25 and T18
fragments) to serve as a negative control. For a positive con-
trol, cotransform another aliquot of cells with plasmids pKT25-
zip and pUT18C-zip (encoding the T25 and T18 fragments
fused to a leucine-zipper dimerization motif).
4. Plate different amounts of the transformation mixture (in
order to have no more than 2–500 colonies per plate) on
LB-X-Gal or MacConkey-maltose indicator plates (plus antibi-
otics) and incubate at 30 °C for 24–48 h.
The results of typical phenotypic assays on LB-X-Gal or
MacConkey-­maltose plates are shown in Fig. 2. DHM1 (or other
BACTH strains) transformants, expressing the T25-zip and T18-­
zip hybrid proteins that can heterodimerize through their leucine
zipper motif, form blue colonies on LB X-Gal medium and red
colonies on MacConkey/maltose, while cells expressing the
unfused T25 and T18 remain colorless.

3.4  BACTH Screening The BACTH system can be used to screen libraries to isolate part-
of Interacting ners of a protein of interest (e.g., protein X, classically designated
Partners: Selection as “bait”) as follows:
Procedure 1. Construct a library of genomic DNA (or cDNA) fragments in
on Minimal Medium one of the BACTH vectors, e.g., pKT25, using standard pro-
cedures [18]. Obviously the quality (i.e., complexity) of the
 Bacterial Two-Hybrid Interaction Assays 167

library is critical for the success in isolating putative partners. A

brief summary of a procedure used in our laboratory to con-
struct a library of genomic E. coli chromosomal DNA frag-
ments in the pKT25 vector is provided in Note 8 (further
experimental details can be found in [20, 21]). Clone the gene
encoding protein X into one of the BACTH vectors (e.g.,
pUT18C) to generate so-called bait plasmid pUT18C-X that
codes for the T18-X hybrid protein.
2. Transform pUT18C-X into a BACTH reporter strain (e.g.,
DHM1).
3. Prepare electrocompetent cells from the resulting transfor-
mants DHM1/pUT18C-X (see Note 9).
4. Transform electrocompetent DHM1/pUT18C-X cells with
50–100 ng of DNA from the BACTH DNA library con-
structed in plasmid pKT25. Add 1 mL LB medium and incu-
bate 90 min at 30 °C. Collect the cells by centrifugation, wash
them four to five times with M63 medium, and plate them
(approximately 1 × 106 transformants/plate) on M63 minimal
medium agar supplemented with maltose (0.2%), as the sole
carbon source, kanamycin, ampicillin, IPTG, and X-Gal (to
facilitate the detection of Cya+ clones that are Mal+ and Lac+).
5. Incubate plates at 30 °C for 4–8 days until appearance of blue
Cya+ colonies. Reisolate these colonies on fresh plates, purify
their pKT25 plasmids, and further characterize the DNA
inserts by sequencing.
This procedure (and related ones) has been used in our labora-
tory to isolate several novel components of the E. coli cell division
machinery [20, 21].

3.5  Quantification Quantification of the functional complementation mediated by

of Functional interaction between the different hybrid proteins is performed by
Complementation measuring β-galactosidase activities in bacterial liquid cultures [3,
Between Hybrid 10]. These β-galactosidase activity assays are conveniently carried
Proteins out in 96-well microtiter plate format as it allows performing many
by β-Galactosidase assays in parallel [17, 21]. Other methods for β-galactosidase assays
Assays can be found elsewhere [16, 18, 22].
1. Pick eight independent colonies from each set of transforma-
tion (i.e., expressing a given couple of T25 and T18 hybrid
proteins), and use them to inoculate 300–400 μL sterile LB
broth supplemented with 0.5 mM IPTG and appropriate anti-
biotics and distributed to individual wells of a 96-well microti-
ter plate (2.2 mL 96-well storage plate or deep-well storage
plate). Seal the plate with a microporous tape sheet to allow
gas exchange and incubate overnight at 30 °C on a rotary
shaker.
168 Gouzel Karimova et al.

2. Dilute the cultures fivefold by adding appropriate volume of

M63 medium to the same microplate.
3. Transfer 175 μL of the diluted cultures into a flat-bottom
microtiter plate and record the OD595 nm absorbance data with
a microplate reader.
4. Transfer 200 μL of the diluted bacterial suspensions into a new
microtiter plate (1.2 mL polypropylene 96-well storage block)
and add 7 μL 0.05% SDS and 10 μL chloroform to permeabi-
lize the cells. Mix vigorously and then leave the plate under a
fume hood at room temperature for 30–40 min to allow chlo-
roform evaporation.
5. In a new microtiter plate, distribute 105 μL/well PM2 reac-
tion buffer containing 100 mM β-mercaptoethanol, and 0.1%
o-nitrophenol-β-galactoside (ONPG). Start the enzymatic
reactions by adding 20 μL aliquots of the permeabilized cells
and incubate the plate at room temperature for 20–30 min or
until sufficient yellow color has developed. In parallel, perform
control assays with 20 μL of M63 medium instead of cells.
6. Stop the reaction by adding 50 μL 1 M Na2CO3 and record the
OD405 absorbance data with a microplate reader.
7. Analyze data with an appropriate software (e.g., Microsoft
Excel or other spreadsheet program). For each well, calculate
the enzymatic activity, A (in relative units), according to

A = 1000 ´ (OD 405 - OD 405 in control wells ) / (OD595 - OD595 in control wells )
/ t ( min ) of incubation.

Results are given in relative units (RUs) of β-galactosidase

activity. It is important to include in the assay negative and positive
controls, i.e., bacteria that express noninteracting (e.g., T25 and
T18 only or fused to proteins that do not interact or should not
interact with the protein of interest) and interacting (e.g., T25-zip
and T18-zip) hybrid proteins, respectively. Under routine condi-
tions, the β-galactosidase activities measured with the positive con-
trols (T25-zip/T18-zip) is defined as 100% activity, while the
β-galactosidase activities of the negative controls (T25/T18)
should be below 2–3% of positive control activity. The
β-galactosidase activities in cells expressing the hybrid proteins of
interest should be at least four to five times higher than the back-
ground level to demonstrate a positive interaction in the BACTH
assay [17, 20, 21, 23].

3.6  Characterization In many cases, it is important to characterize immunologically or

of Hybrid Proteins biochemically the hybrid proteins and eventually to quantify their
by Western Blot level of expression in the complementing cells. For this, western
blot analysis of hybrid proteins can be carried out using standard
 Bacterial Two-Hybrid Interaction Assays 169

procedures [18]. The T25 fragment can be detected with a rabbit

polyclonal antiserum directed against the purified B. pertussis CyaA
protein (serum L24023, DL unpublished) while the T18 fragment
is revealed by an anti-CyaA monoclonal antibody (3D1, sc-13582)
that reacts specifically with the C-terminal region of T18 [20, 24].
Alternatively, it is also possible to append to the T25 or T18 frag-
ments different epitope tags that can be detected with specific
monoclonal antibodies (e.g., myc, HA, or T7 tags) or a 6× histi-
dine tag that permits purification of the complex of hybrid proteins
by chromatography on Ni-NTA-agarose resin [25]. These modi-
fied fragments can be used to perform immunoprecipitation exper-
iments or pull-down assays to demonstrate by direct biochemical
means the physical association of the hybrid proteins [18].

4  Notes

1. Two types of indicator plates are commonly used to reveal pro-

tein interaction with the BACTH assay:
LB-X-Galplates: In E. coli, expression of the lacZ gene encoding
β-galactosidase is positively controlled by cAMP/CAP. Hence,
bacteria expressing interacting hybrid proteins form blue colonies
on rich LB medium in the presence of the chromogenic substrate
X-Gal (see Fig. 2), while cells expressing noninteracting proteins
remain white (pale blue).
MacConkeymedium: E. coli cya bacteria are unable to ferment
lactose or maltose [15, 18]; they form white (or pale pink) colo-
nies on MacConkey indicator media containing maltose (see
Fig. 2). In contrast, Cya+ bacteria form red colonies on the same
media (fermentation of the sugar results in the acidification of the
medium and induces a color change of the phenol red dye). Note
that not all MacConkey agar base media are of equal quality.
MacConkey from Difco Laboratories (cat # 216830) is strongly
recommended.
Cells expressing interacting proteins can be selected by plating
transformants on a selective medium consisting of a synthetic mini-
mal medium supplemented with maltose as a unique carbon source
[4, 20, 21]: as the mal regulon (involved in maltose catabolism)
expression is under a strict cAMP/CAP dependency, only Cya+
bacteria can utilize maltose as a carbon source. Hence, only the
cells that express interacting hybrid proteins will be able to grow
on this minimal medium (Fig. 2). X-Gal and IPTG are also com-
monly added to the selective medium to facilitate the early visual-
ization of the Cya+ colonies (these cells should also be Lac+ and
therefore exhibit a blue phenotype on X-Gal). Note that when
using the DHT1 as a reporter strain [23, 26], casamino acids
should be added to the minimal medium/maltose plates to allow
growth, as this strain is ilv− (i.e., unable to synthetize isoleucine
and valine).
170 Gouzel Karimova et al.

2. β-mercaptoethanol is considered toxic, causing irritation to the

skin and respiratory tract upon inhalation and should be
manipulated under a fume hood. In fact, it can be easily omit-
ted from the PM2 buffer: the β-galactosidase activities will be
reduced by a factor of 2, which is not problematic because only
relative enzymatic activities will be considered.
3. Several adenylate-cyclase-deficient(cya) E. coli reporter strains,
DHT1, DHM1, and BTH101 (see genotypes below), can be
used as host organisms for the detection of protein–protein
interactions in a BACTH assay [4, 13, 27]. Other E. coli cya
strains (see E. coli strain collection at http://cgsc.biology.yale.
edu) may also be used. The different genetic backgrounds of
these strains provide different complementation efficiencies
-
and different reporter gene stringencies. DHT1 [F , cya-854,
ilv 691::Tn10, recA1, endA1, gyrA96(Nalr), thi1, hsdR17,
spoT1, rfbD1, glnV44(AS)] is a recA strain that displays a high
BACTH complementation efficiency and fast growth, but it
requires casamino acid supplementation for growth on minimal
-
medium as it carries an ilv mutation. DHM1 [F , cya-854,
recA1, endA1, gyrA96(Nalr), thi1, hsdR17, spoT1, rfbD1,
glnV44(AS)] is an ilv + DHT1 derivative able to grow on mini-
mal media plus sugars, but it displays a lower complementation
efficiency and slower growth than the parental DHT1. BTH101
-
[F , cya-99, araD139, galE15, galK16, rpsL1 (Strr), hsdR2,
mcrA1, mcrB1] also d ­ isplays a good BACTH efficiency and fast
growth, but some instability of plasmids may be observed
owing to the Rec+ character of the strain. The frequencies of
spontaneous Lac+ revertants (owing to cAMP/CAP indepen-
dent promoter mutations) of these different strains range from
10−7 to 10−8, while frequencies of spontaneous Mal+ revertants
are below the detection threshold (i.e., <10−10).
4. The BACTH technology requires coexpression of two hybrid
proteins within the same recipient cya bacteria. For this, two
sets of compatible vectors allowing genetic fusions of the pro-
teins of interest at either the N- or the C-terminus of the T25
fragment (pKT25 and pKNT25) or of the T18 fragment
(pUT18 and pUT18C) are available: their schematic maps are
shown in Fig. 3a, and their nucleotide sequences are available
upon request [4, 13].
Plasmid pKT25 expresses under a lac promoter control, the
T25 fragment (corresponding to the first 224 amino acids of
CyaA). It is a low-copy-number pSU40 plasmid derivative harbor-
ing a kanamycin resistance selectable marker. It contains a multi-
cloning site sequence (MCS) at the 3′ end of T25 to allow
construction of in-frame fusions at the C-terminal end of the T25
polypeptide. Plasmid pKNT25 is similar to pKT25 except that
MCS is located at the 5′ end of T25 coding region to allow fusions
of proteins to the N-terminus of T25.
 Bacterial Two-Hybrid Interaction Assays 171

Plasmid pUT18 is an ampicillin-resistant pUC19 derivative

that expresses the T18 fragment (amino acids 225 to 399 of CyaA)
under the transcriptional control of the lac promoter. The T18
open reading frame is located downstream of the pUC19 MCS,
and therefore pUT18 is used to express chimeric proteins in which
the polypeptide of interest is fused to the N-terminal end of T18.
In plasmid pUT18C, the same MCS is located at the 3′ end of the
T18 open reading frame to allow fusions of proteins to the
C-­terminus of T18.
In addition two plasmids, pKT25-zip and pUT18C-zip, are
commonly used as positive controls for BACTH complementa-
tion. They are derivatives of pKT25 and pUT18C, respectively,
that code for the T25 and T18 fragments fused to the leucine zip-
per of GCN4 [10, 11].
Another set of vectors was recently designed to be compatible
with the Gateway® recombineering technique (Thermo Fisher
Scientific). The Gateway® technique allows for facile recombinase-­
mediated transfer of an open reading frame (ORF), flanked by
recombination sites, from an “entry” vector into a wide variety of
“destination” vectors [19]. The Gateway®-compatible destination
vectors were constructed by the insertion of a recombination cas-
sette encoding a chloramphenicol resistance marker and the toxin
CcdB and flanked by attR bacteriophage lambda recombination
sites (Fig. 3b) in the BACTH vectors. The resulting plasmids
pST25-DEST and pUT18C-DEST are suitable for fusing proteins
of interest to the C-terminus of the T25 and T18 fragments,
respectively, while pSNT25-DEST is used for fusing ORFs to the
N-terminus of the T25 fragment [17]. Importantly, the pST25-­
DEST and pSNT25-DEST BACTHGW vectors contain a
spectinomycin-­resistant (instead of kanamycin-resistant) marker to
be compatible with the popular Gateway® entry vector pDONR221
that harbors a kanamycin resistance gene. The pST25-DEST,
pSNT25-DEST, and pUT18C-DEST plasmids must be propa-
gated in E. coli strains that are resistant to the lethal effect of the
CcdB toxin, such as the DB3.1™ E. coli strain (harboring the
CcdB-resistant, gyrA462 gyrase mutation; see manufacturer’s
guidelines).
5. Vectors and recombinant plasmids are commonly propagated
at 30 °C in standard E. coli K12 recA strains (such as XL1-
Blue). To avoid any problems during construction of the plas-
mids, it is wise to grow the cells in LB medium containing
0.2% glucose or to use an E. coli host strain that overproduces
the LacI repressor to prevent expression of the hybrid proteins
(e.g., XL1-­Blue). Plasmid DNA is routinely purified with com-
mercial kits used for minipreparation of DNA, according to the
manufacturer’s instructions.
172 Gouzel Karimova et al.

Fig. 3 Schematic representation of BACTH plasmids. (a) Standard BACTH plasmids. Yellow and green rect-
angles represent the open reading frames of T25 and T18 fragments, respectively, under the control of the
lac promoter (small arrow). Pink and orange arrows indicate the antibiotic selectable markers and the direc-
tion of transcription. The red and blue boxes indicate the plasmid origin of replication. The hatched boxes
represent the multicloning sequences (MCS) that allow insertion of foreign genes: some unique restriction
sites are displayed above the nucleotide sequence, and the encoded polypeptide sequences are shown
below. (b) Gateway™-compatible, BACTHGW plasmids
 Bacterial Two-Hybrid Interaction Assays 173

6. During the design of primers to amplify the genes of interest,

the following sequences (red, underlined) corresponding to
the specific attB sites (necessary for recombination reaction)
should be appended to the gene-specific sequences (indicated
by XXX…):
Direct primer (the bold ATG corresponds to the initiation
codon of the open reading frame): 5′-GCCGCACAAGTTTGTA
CAAAAAAGCAGGCTTTATGXXXXXXXX
Reverse primer: (the stop codon can be removed if desired):
5 ′- G C G G A C C A C T T T G TA C A A G A A A G C T G G G T T
XXXXXXXX
Refer to the manufacturer’s guidelines for more precise expla-
nations regarding the design of primers for Gateway® cloning.
7. Before use, the strains (DHT1, DHM1, or BTH101) from the
LB-DMSO stock should be restreaked on either MacConkey/
maltose or LB/X-Gal/IPTG plates and grown overnight at
37 °C. White colonies (i.e., cya) should be picked up to start the
overnight liquid preculture. Any red (on MacConkey/maltose)
or blue colonies (on LB/X-Gal/IPTG) that may appear should
be avoided (they likely correspond to Lac+ or Mal+ revertants or
contaminants). If too many contaminants are present upon
restreaking of the stock, a selective antibiotic may be added to
the MacConkey/maltose or the LB/X-Gal/IPTG plates:
DHT1 and DHM1 are resistant to nalidix acid (30 μg/mL),
whereas BTH101 is resistant to streptomycin (100 μg/mL).
DHT1, DHM1, or BTH101 competent cells can be prepared
by the classical CaCl2 technique [18], which yields a competency
level (>106 cfu/μg) sufficient for most routine transformations.
Briefly, freshly reisolated cells are grown in 1 L LB medium at 37
°C to OD 0.25–0.3, cooled on ice, and pelleted by centrifugation.
Cells are washed twice in 100 mL ice-cold 0.1 M CaCl2 solution.
Cells are finally resuspended in 30–40 mL ice-cold 0.1 M CaCl2
and incubated overnight at 4 °C (it is critical to keep cells, buffers,
and vessels well chilled at all stages of the process).
For transformation, 50 μL of chemically competent DHM1
cells are mixed in a chilled microcentrifuge tube with 5–10 ng of
each plasmid, incubated 30 min at 4 °C, and then heat-shocked at
42 °C for 2 min. Then 1 mL LB is added, and the cell suspension
is further incubated at 30 °C for 60–90 min before being plated.
Different volumes of the transformation mixture should be plated
to obtain about 100–200 colonies per plate. It is important that
the number of colonies not exceed 500; otherwise, the detection
of positive clones might be difficult. It should be noted that after
prolonged incubation (4–5 days), negative colonies (i.e., cya−) will
show a weak red (on MacConkey-maltose) or blue spot (on LB-­X-­
Gal) in the center but will remain colorless at the periphery. It
might also be worth testing the complementation at 37 °C,
174 Gouzel Karimova et al.

although in many cases complementation at 37 °C is less efficient

than at 30 °C.
8. The genomic DNA (≈50 μg) from a ∆cya derivative of the
E. coli strain MG1655 was randomly fragmented by sonication
(size range of 500–1500 bp). The fragments were end-repaired
by Mung Bean nuclease and treated with a mixture of T4 DNA
polymerase and Klenow fragment (with dNTP). In parallel,
the pKT25 vector (10 μg) was digested with SmaI and dephos-
phorylated with shrimp alkaline phosphatase, and the linear-
ized vector was gel purified. The blunt-ended DNA fragments
were then ligated with the SmaI-digested pKT25 vector and
transformed into electrocompetent ElectroMAX DH10B cells
(Thermo Fisher Scientific). About 5 × 105 independent clones
were thus obtained. All these colonies were pooled and their
plasmid DNA was purified and used as a stock for the BACTH
DNA library [21].
9. Efficient (>108 cfu/μg) electrocompetent DHM1/pUT18C-
X cells can be prepared as follows [18]: freshly reisolated cells
are grown at 37 °C in 1 L LB containing 100 μg/mL ampicil-
lin until OD600 of 0.5–0.7. Cells are chilled on ice and pelleted
by centrifugation at 4 °C. Cells are washed at least three times
with ice-cold water and resuspended in 10 mL 10% glycerol (in
water). For transformation, 50 μL are transferred into an elec-
troporation cuvette (1 mm wide) previously equilibrated on
ice, and 50–100 ng DNA from the BACTH plasmid DNA
library are added. After mixing and a few minutes of incuba-
tion at 4 °C, the cuvette is placed in an electroporator (e.g.,
BioRad) set at 2.5 KV, 100 Ohms capacitance, and electro-
poration is carried out. One milliliter of LB media is immedi-
ately added to the cuvette, and cells are further incubated at
30 °C for 60–90 min. Cells are then collected by centrifuga-
tion (5 min at 6000 rpm, or 4500 × g) and washed several
times with M63 medium (to remove all nutrients from the rich
medium) before being plated on M63 minimal medium agar
(approximately 1 × 106 transformants/plate).

Acknowledgments

This work was supported by Institut Pasteur and the Centre

National de la Recherche Scientifique (CNRS UMR 3528, Biologie
Structurale et Agents Infectieux). E.G. was supported by Ph.D.
funding from the Université Paris Diderot, Sorbonne Paris Cité,
Cellule Pasteur, Paris, France.
 Bacterial Two-Hybrid Interaction Assays 175

References

1. Fields S, Song O (1989) A novel genetic sys- transcription activation. Curr Opin Struct Biol
tem to detect protein-protein interactions. 14:10–20
Nature 340:245–246 13. Karimova G, Ullmann A, Ladant D (2001)
2. Stynen B, Tournu H, Tavernier J, Van Dijck P Protein-protein interaction between Bacillus
(2012) Diversity in genetic in vivo methods for stearothermophilus tyrosyl-tRNA synthetase
protein-protein interaction studies: from the subdomains revealed by a bacterial two-hybrid
yeast two-hybrid system to the mammalian system. J Mol Microbiol Biotechnol 3(1):
split-luciferase system. Microbiol Mol Biol Rev 73–82
76:331–382 14. Fransen M, Brees C, Ghys K, Amery L,
3. Karimova G, Pidoux J, Ullmann A, Ladant D Mannaerts GP, Ladant D, Van Veldhoven PP
(1998) A bacterial two-hybrid system based on (2002) Analysis of mammalian peroxin interac-
a reconstituted signal transduction pathway. tions using a non-transcription-based bacterial
Proc Natl Acad Sci U S A 95:5752–5756 two-hybrid assay. Mol Cell Proteomics 1:
4. Karimova G, Dautin N, Ladant D (2005) 243–252
Interaction network among Escherichia coli 15. Dautin N, Karimova G, Ladant D (2003)
membrane proteins involved in cell division as Human immunodeficiency virus (HIV) type 1
revealed by bacterial two-hybrid analysis. transframe protein can restore activity to a
J Bacteriol 187:2233–2243 dimerization-deficient HIV protease variant.
5. Jack RL, Buchanan G, Dubini A, Hatzixanthis J Virol 77:8216–8226
K, Palmer T, Sargent F (2004) Coordinating 16. Battesti A, Bouveret E (2012) The bacterial
assembly and export of complex bacterial pro- two-hybrid system based on adenylate cyclase
teins. EMBO J 23:3962–3972 reconstitution in Escherichia coli. Methods
6. Paschos A, den Hartigh A, Smith MA, Atluri 58:325–334
VL, Sivanesan D, Tsolis RM, Baron C (2011) 17. Ouellette SP, Gauliard E, Antosova Z, Ladant
An in vivo high-throughput screening approach D (2014) A Gateway((R)) -compatible bacte-
targeting the type IV secretion system compo- rial adenylate cyclase-based two-hybrid system.
nent VirB8 identified inhibitors of Brucella Environ Microbiol Rep 6:259–267
abortus 2308 proliferation. Infect Immun 18. Sambrook J, Russell DW (2006) The con-
79:1033–1043 densed protocols from molecular cloning: a
7. Cisneros DA, Bond PJ, Pugsley AP, Campos laboratory manual. Cold Spring Harbor
M, Francetic O (2012) Minor pseudopilin self-­ Laboratory Press, Cold Spring Harbor, NY
assembly primes type II secretion pseudopilus 19. Hartley JL, Temple GF, Brasch MA (2000)
elongation. EMBO J 31:1041–1053 DNA cloning using in vitro site-specific recom-
8. Georgiadou M, Castagnini M, Karimova G, bination. Genome Res 10:1788–1795
Ladant D, Pelicic V (2012) Large-scale study 20. Karimova G, Robichon C, Ladant D (2009)
of the interactions between proteins involved Characterization of YmgF, a 72-residue inner
in type IV pilus biology in Neisseria meningiti- membrane protein that associates with the
dis: characterization of a subcomplex involved Escherichia coli cell division machinery.
in pilus assembly. Mol Microbiol 84:857–873 J Bacteriol 191:333–346
9. Zoued A, Durand E, Brunet YR, Spinelli S, 21. Karimova G, Davi M, Ladant D (2012) The
Douzi B, Guzzo M, Flaugnatti N, Legrand P, beta-lactam resistance protein Blr, a small
Journet L, Fronzes R et al (2016) Priming and membrane polypeptide, is a component of the
polymerization of a bacterial contractile tail Escherichia coli cell division machinery.
structure. Nature 531:59–63 J Bacteriol 194:5576–5588
10. Karimova G, Ullmann A, Ladant D (2000) A 22. Griffith KL, Wolf REJ (2002) Measuring beta-­
bacterial two-hybrid system that exploits a galactosidase activity in bacteria: cell growth,
cAMP signaling cascade in Escherichia coli. permeabilization, and enzyme assays in 96-well
Methods Enzymol 328:59–73 arrays. Biochem Biophys Res Commun
11. Ladant D, Ullmann A (1999) Bordatella per- 290:397–402
tussis adenylate cyclase: a toxin with multiple 23. Ouellette SP, Rueden KJ, Gauliard E, Persons
talents. Trends Microbiol 7:172–176 L, de Boer PA, Ladant D (2014) Analysis of
12. Lawson CL, Swigon D, Murakami KS, Darst MreB interactors in Chlamydia reveals a RodZ
SA, Berman HM, Ebright RH (2004) homolog but fails to detect an interaction with
Catabolite activator protein: DNA binding and MraY. Front Microbiol 5:279
176 Gouzel Karimova et al.

24. Robichon C, Karimova G, Beckwith J, Ladant D 26. Ouellette SP, Karimova G, Subtil A, Ladant D
(2011) Role of leucine zipper motifs in associa- (2012) Chlamydia co-opts the rod shape-­
tion of the Escherichia coli cell division proteins determining proteins MreB and Pbp2 for cell
FtsL and FtsB. J Bacteriol 193:4988–4992 division. Mol Microbiol 85:164–178
25. Battesti A, Bouveret E (2008) Improvement of 27. Dautin N, Karimova G, Ullmann A, Ladant D
bacterial two-hybrid vectors for detection of fusion (2000) Sensitive genetic screen for protease
proteins and transfer to pBAD-tandem affinity activity based on a cyclic AMP signaling cas-
purification, calmodulin binding peptide, or 6-his- cade in Escherichia coli. J Bacteriol 182:
tidine tag vectors. Proteomics 8:4768–4771 7060–7066
 Chapter 14

Protein–Protein Interactions: Yeast Two-Hybrid System

Jer-Sheng Lin and Erh-Min Lai

Abstract
The yeast two-hybrid system is a powerful and commonly used genetic tool to investigate interactions
between artificial fusion proteins inside the nucleus of yeast. Here we describe how to use the Matchmaker
GAL4-based yeast two-hybrid system to detect the interaction of the Agrobacterium type VI secretion
system (T6SS) sheath components TssB and TssC41. The bait and prey gene are expressed as a fusion to
the GAL4 DNA-binding domain (DNA-BD) and GAL4 activation domain (AD, prey/library fusion pro-
tein) respectively. When bait and prey fusion proteins interact in yeast nucleus, the DNA-BD and AD are
brought into proximity, thereby activating the transcription of reporter genes. This technology can be
widely used to identify interacting partners, confirm suspected interactions, and define interacting domains.

Key words Protein–protein interaction, Yeast two-hybrid, Gal4 transcriptional activation domain
(AD), Gal4 DNA-binding domain (BD), Saccharomyces cerevisiae AH109, Type VI secretion system,
TssB, TssC

1  Introduction

The yeast two-hybrid system (Y2H) was first developed in 1989 and
revolutionized the process of searching for and identifying interact-
ing proteins [1]. To date, the Y2H system has proven to be a useful
and sensitive method for detecting not only stable interacting pro-
teins but also weak and transient protein interactions [2]. Because
Y2H is performed in vivo, the great advantage of the system is that
the testing proteins are more likely to be in their native conforma-
tions, which may lead to increased sensitivity and accuracy of detec-
tion [1, 3, 4]. Importantly, the Y2H system is complementary to
biochemical methods such as co-immunoprecipitation/pulldown
followed by western blotting or mass spectrometry analysis to
increase accuracy and dynamics for a more complete and reliable
map of interactions [2]. Notably, the Y2H method has been modi-
fied and improved greatly in recent years, including applications in
protein–DNA interactions and yeast three-hybrid, and has proven to
be amenable to interaction studies of membrane proteins, DNA

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_14, © Springer Science+Business Media LLC 2017

177
178 Jer-Sheng Lin and Erh-Min Lai

binding proteins, and RNA binding proteins [5–8]. Using the

Matchmaker GAL4-based Y2H system (Clontech, Mountain View,
CA) as an example, the principle of the Y2H system is illustrated in
Fig. 1a. Based on the properties of the yeast GAL4 transcription fac-
tor it consists of separable domains responsible for DNA binding
and transcriptional activation [3]. Bait proteins are expressed as a
fusion to the GAL4 DNA binding domain (DNA-BD), while prey
proteins are expressed as fusions to the GAL4 activation domain
(AD). When bait and prey fusion proteins interact in yeast nucleus,
the DNA-BD and AD are brought into proximity to restore to a
functional GAL4 transcriptional activator, which binds onto an
upstream activating sequence (UAS) of reporter genes (such as

(A)

AD

Bait Prey
protein Transcriptional activation
protein

DNA-BD

GAL UAS Minimal promoter Reporter genes

(ADE2, HIS3, and MEL1)

Saccharomyces cerevisiae AH109

(B)

Co-transformation Selecting
AH109 Yeast transformant

(PEG/LiAc) 2-3 days AH109 with pGBKT7 and pGADT7

SD-Trp-Leu
( SD-WL medium)
(SD-WL medium)
Gene construction in pGBKT7
and pGADT7 vectors

Yeast protein extraction Testing for protein-protein

and western blot analysis interaction
3-6 days
To comfirm the proper SD-Trp-Leu-His-Ade
expression of tested proteins (SD-WLHA medium)

Fig. 1 Principle and experimental flowchart of yeast two-hybrid system. (a) Schematic
diagram of principle of Y2H system. Two testing proteins are each fused with two
different Gal4 domains, with the bait protein fused to the Gal4 DNA-binding domain
(DNA-BD, 1-147 a.a.) and the prey protein fused to the Gal4 transcriptional activation
domain (AD, 768-881 a.a.). In yeast strain AH109, transcriptional activation of the
reporters (ADE2, HIS3, and MEL1) only occurs in a cell that bait protein interact with
prey protein to restore functional Gal4 transcription factor binding to the Gal4-
responsive promoter GAL UAS [3]. (b) Experimental flowchart of Y2H system.
Cotransformation was performed using PEG/LiAc-mediated transformation method.
SD-WL medium is representative of synthetic defined (SD) minimal medium lacking
tryptophan (Trp) and leucine (Leu). SD-WLHA medium is representative of synthetic
dextrose minimal medium lacking Trp, Leu, adenine (Ade), and histidine (His)
 Yeast Two-Hybrid 179

ADE2 and HIS3) for transcriptional activation. The Y2H system has
been widely used to detect the interactions of a wide range of pro-
teins from yeast, bacteria, animal, and plant systems. Y2H has been
successfully used to study the interaction of proteins involved in bac-
terial protein secretion from type IV [9–12] and type VI secretion
systems (T4SS, T6SS) in Agrobacterium tumefaciens [13, 14].
Here, the Y2H protocol describes the use of Matchmaker Y2H
system to detect the interaction of the Agrobacterium T6SS sheath
components TssB and TssC41 according to the instructions of the user
manual (Clontech, Mountain View, CA), with minor modifications.
TssB and TssC interact to form a cogwheel-like tubular structure,
which is analogous to the outer sheath structure of a contractile phage,
and wraps around the T6SS tail tube to propel the tail tube toward the
target cell interior upon infection [15, 16]. In A. tumefaciens, we
showed the interaction of the T6SS sheath components TssB and
TssC41 by Y2H assay, copurification in E. coli, and co-IP in A. tumefa-
ciens [14]. For the Y2H assay, each bait and prey plasmid pair was
cotransformed into the Saccharomyces cerevisiae strain AH109. The
transformants were selected by their growth on synthetic dextrose
(SD) minimal medium lacking tryptophan (Trp) and leucine (Leu)
(SD-WL medium), which are the nutritional selection markers for
pGBKT7 and pGADT7, respectively. The positive interaction of
expressed fusion proteins was then determined by their growth on SD
lacking Trp, Leu, adenine (Ade), and histidine (His) (SD-WLHA
medium) at 30 °C for at least 3 days (Fig. 1b). The positive interac-
tions wwere observed only for plasmid pairs expressing TssB and
TssC41 but not when each of them coexpressed with vector only, sug-
gesting the specific interactions of TssB and TssC41 (Fig. 2) [14].
BD-TssC41
BD-TssB
vector

+
vector
AD-TssC41 SD-WLHA
AD-TssB

vector
AD-TssC41 SD-WL
AD-TssB

Fig. 2 TssB and TssC41 interact with each other in yeast strain AH109. SD-WL
medium (SD minimal medium lacking Trp and Leu) was used for the selection of
plasmids. SD-WLHA medium (SD minimal medium lacking Trp, Leu, His, and Ade)
was used for the auxotrophic selection of bait and prey protein interactions. The
positive interaction was determined by the growth on SD-WLHA medium at
30 °C for at least 2 days. The positive control (+) showing interactions of SV40
large T-antigen and murine p53 and negative control (vector) are indicated
(reproduced from [14]; no permission is required for reuse of the content pub-
lished from Public Library of Science, PLoS)
180 Jer-Sheng Lin and Erh-Min Lai

2  Materials

All growth media and solutions are prepared using Milli-Q purified
water and analytical or molecular biology grade reagents.

2.1  Yeast Strain 1. The yeast Saccharomyces cerevisiae strain AH109: The com-
and Vectors (the plete genotype of AH109 is provided in what follows.
Following Information
MATa, trp1–901, leu2–3, 112, ura3–52, his3–200, gal4Δ,
Is According to [3]) gal80Δ, LYS2:: GAL1UAS-GAL1TATA-HIS3, GAL2UAS-GAL2TATA-
ADE2, URA3:: MEL1UAS-MEL1TATA-lacZ.
AH109 strain is gal4 − and gal80 −; this prevents interfer-
ence of native regulatory proteins with the regulatory elements
in the two-hybrid system. AH109 features three reporters,
ADE2, HIS3, and MEL1 (or lacZ), under the control of distinct
GAL4 upstream activating sequences (UASs) and TATA boxes.
2. pGBKT7 vector: The pGBKT7 vector contains a multiple
cloning site (MCS) for cloning to express proteins with
N-terminal fusion to amino acids 1–147 of the GAL4 DNA
binding domain (DNA-BD). In yeast, fusion proteins are
expressed at high levels from the constitutive ADH1 promoter
(PADH1). Transcription is terminated by the T7 and ADH1
transcription termination signals (TADH1). The pGBKT7 vector
can replicate autonomously in both E. coli and S. cerevisiae
from the pUC and 2 μ ori, respectively. The vector carries the
kanamycin-­resistant gene for selection in E. coli and the TRP1
nutritional marker for selection in yeast. In addition, pGBKT7
also contains the T7 promoter and a c-Myc epitope tag for
in vitro transcription and translation of the c-Myc-tagged
fusion protein without GAL4 DNA-BD.
3. The pGADT7 vector: The pGADT7 vector contains MCSs
for cloning to express protein with N-terminal fusion to
amino acids 768–881 of the GAL4 activation domain (AD).
In yeast, fusion proteins are expressed at high levels from the
constitutive ADH1 promoter (PADH1). Transcription is ter-
minated at the ADH1 transcription termination signal
(TADH1). The fusion protein is targeted to the yeast nucleus
by the SV40 nuclear localization sequences that have been
added to the activation domain sequence. pGADT7 also
contains the T7 promoter and an HA epitope tag for in vitro
transcription and translation of the HA-tagged fusion pro-
tein without GAL4 AD. The pGADT7 vector can replicate
autonomously in both E. coli and S. cerevisiae from the pUC
and 2 μ ori, respectively. The vector carries the ampicillin-
resistant gene for selection in E. coli and the LEU2 nutri-
tional marker for selection in yeast.
 Yeast Two-Hybrid 181

2.2  Yeast Cultures 1. Yeast peptone dextrose adenine (YPDA) medium: 20 g Bacto
and Yeast peptone, 10 g yeast extract, 20 g glucose, 40 mg adenine, 15 g
Transformation [17] agar (For plate use only), add water to 1 L, autoclave.
2. Minimal synthetic defined (SD) plate: 1.675 g yeast nitrogen
base without amino acid, 5 g glucose, 3.75 g Agar, add water
to 250 mL, autoclave. Dropout (DO) supplements (such as
-Trp-­Leu or -Trp-Leu-Ade-His) can be added to the Minimal
SD Base to make a synthetic, defined medium lacking the spec-
ified nutrients (see Note 1).
3. Carrier DNA: 10 mg/mL salmon sperm DNA (ssDNA)
(UltraPure™ Salmon Sperm DNA Solution, ThermoFisher),
store at −20 °C (see Note 2).
4. 10× LiAc: 1 M Lithium acetate, pH 7.5 (see Note 3), autoclave
and store at room temperature (RT).
5. 40% polyethylene glycol (PEG) solution: 22 g Polyethylene
glycol (Molecular weight is 6000 or 3350 Da), add 31 mL of
water, autoclave and store at room temperature.
6. Plasmid DNA: ~200 ng per plasmid for co-transformation
(see Note 4).
7. Laminar flow.

2.3  Selective Media 1. Selective medium for transformants: Minimal synthetic

defined (SD) plate with -Leu/-Trp DO supplement (con-
taining every essential amino acids except for leucine and
tryptophan) (see Note 5).
2. Selective medium for Protein–Protein interactions: Minimal
synthetic defined (SD) plate with -Leu/-Trp/-His/-Ade DO
supplement (containing all essential amino acids except for leu-
cine, tryptophan, histidine, and adenine).

2.4  Preparation 1. Yeast peptone dextrose adenine (YPDA) medium: See item 1
of Yeast Cultures in Subheading 2.2.
for Protein Extraction 2. 2× minimal synthetic defined (SD) medium: 40 g Bacto pep-
and Western Blot tone, 20 g yeast extract, 20 g glucose, 80 mg adenine, with
2X-Leu/-Trp dropout (DO) supplement (containing every
essential amino acids except for leucine and tryptophan), add
water to 1 L, autoclave.

2.5  Preparation 1. 1.5 mL microtube.

of Yeast Protein 2. Acid-washed glass beads (425–600 μm).
Extracts
3. Protease inhibitor cocktail solution.
4. Phenylmethylsulfonyl fluoride (PMSF) stock solution: 0.1 M
5. Yeast protein extraction buffer (see Note 6): 0.1% NP-40,
250 mM NaCl, 50 mM Tris–HCl, pH 7.5, 5 mM ethylenedi-
aminetetraacetic acid (from a 0.5 M, pH 8.0 stock solution),
182 Jer-Sheng Lin and Erh-Min Lai

mix well, and place on ice. Before use, add 1 mM dithiothreitol
(DTT), 2× protease inhibitor cocktails (stock: 50×), 4 mM
PMSF, mix well, then ready for use.

3  Methods

3.1  Gene The constructs used for Y2H analysis are generated based on the
Construction information of the vector map and MCS provided by the user man-
in pGBKT7 ual (Clontech) [3]. Briefly, bait and prey coding sequence (without
and pGADT7 Vectors stop codon) are PCR-amplified with appropriate primers, digested
with appropriate enzymes, and cloned to the same sites of pGBKT7
or pGADT7 [14].

3.2  Preparation 1. Inoculate 3 mL YPDA with a colony of AH109 (see Note 7)
of Yeast Cultures and incubate at 30 °C overnight (>16 h) with shaking
for Yeast (250 rpm) to the stationary phase (see Note 8).
Transformation [17] 2. Subculture by adding 1 mL AH109 overnight culture into
50 mL fresh YPDA medium.
3. Incubate at 30 °C for 4 h with shaking (250 rpm) (see Note 9).
4. Pour cells into 50 mL tubes and pellet cells at 450 × g for
3 min at 4 °C or RT (see Note 10).
5. Discard supernatant and resuspend cell pellet with 10 mL ster-
ile water by vortexing, and repellet cells at 450 × g for 3 min at
4 °C or RT (see Note 11).
6. Resuspend cell pellet in 100 μL 10× LiAc and 900 μL sterile
water (final concentration is 1× LiAc) (see Note 12). Incubate
cell suspension at 30 °C for 1 h with gentle shaking (150 rpm)
(see Note 13).
7. The suspended yeast competent cells are ready to use for trans-
formation (see Note 14).

3.3  PEG/LiAc-­ 1. Pretreat ssDNA by heating at 100 °C for 10 min and then put
Mediated on ice for 5–10 min before use (see Note 16).
Transformation 2. Add 80 μL heat-treated ssDNA (10 μg/μL) to 1 mL yeast
of Yeast (Small-Scale competent cell (final concentration ~ 0.8 mg/mL) and mix
Transformation of Bait well (see Note 17).
and Prey Plasmids) 3. Aliquot 100 μL of cell mixture into 1.5 mL microtube, and
(see Note 15) add an approximate amount of plasmid DNA (about 3–5 μL)
(see Note 18). Mix well by vortexing and incubate at 30 °C for
30 min (see Note 19).
4.
Freshly prepare LiAc-PEG solution (10× LiAc: 40%
PEG = 1:10, mix 1 mL 10× LiAc with 10 mL 40% PEG) and
add 700 μL LiAc-PEG solution to cell mixture after 30 min
 Yeast Two-Hybrid 183

incubation (see Note 20). Resuspend cell mixture immediately

by vortexing (see Note 21) before incubation at 30 °C for 1 h.
5. Heat shock at 42 °C for 5 min (see Note 22).
6. Pellet cells by centrifugation at 14,500 × g for 1 min at RT
(see Note 23).
7.
Discard supernatant as much as possible to remove
PEG. Resuspend cells in 300 μL sterile water (see Note 24).

3.4  Selection 1. Streak cells on SD/-Trp-Leu selective plate and incubate at

of Transformants 30 °C for 2–3 days (see Note 25).
2. Patch single colony on SD/-Trp-Leu selection plate and incu-
bate at 30 °C for 2 days (see Note 26).

3.5  Testing 1. Patch cells on both SD/-Trp-Leu (control) and SD/-Trp-

for Protein–Protein Leu-­His-Ade selection plates for 3–6 days (see Note 27).
Interactions 2. Photograph the plate to record the final protein–protein inter-
action results (For example, TssB and TssC41 can interact
strongly in yeast, Fig. 2) [14].

3.6  Preparation 1. Grow culture from a single colony (see Note 29) in 3 mL
of Yeast Cultures YPDA or 2× SD selection medium (see Note 30) overnight at
for Protein Extraction 30 °C.
(see Note 28) 2. Add 100 μL of overnight culture (OD600 should reach 1.5) in
fresh 5 mL 2× SD selection medium. Incubate at 30 °C with
shaking (about 250 rpm) until OD600 reaches 0.4–0.6.
Depending on tested proteins, it may take 4–5 h to reach
desired cell number (see Note 31).
3. Place cells in 15 mL tubes and pellet cells at 1000 × g for 5 min
at 4 °C.
4. Discard supernatant and resuspend cell pellet with 10 mL ster-
ile water by vortexing (see Note 32).
5. Repellet cells at 1000 × g for 5 min at 4 °C.
6. Repeat steps 4 and 5.
7. Discard supernatant. You can continue to extract yeast protein
or immediately freeze cell pellet by placing tube in liquid nitro-
gen, then store cells at −80 °C until western blot analysis.

3.7  Preparation 1. Keep protein samples on ice, add 100 μL freshly prepared yeast
of Yeast Protein protein extraction buffer to tube followed by addition of 50 μL
Extracts and Western acid-washed glass beads (see Note 33).
Blot Analysis 2. Vortex tubes at maximum speed for 30 s, then place tubes on
ice for 30 s (see Note 34). Repeat this step six times.
3. Transfer supernatant above settled glass beads to a new 1.5 mL
microtube using P200 PIPETMAN, and place tubes on ice.
The supernatant is the first cell extract.
184 Jer-Sheng Lin and Erh-Min Lai

4. Add 50 μL yeast protein extraction buffer to tube containing

glass beads and vortex tubes at highest speed for 30 s, then
transfer supernatant (second cell extract) above settled glass
beads to 1.5 mL microtube containing first cell extract.
5. Centrifuge combined cell extract at 14,500 × g for 5 min at
4 °C (see Note 35).
6. Transfer supernatant to new 1.5 mL microtube and measure
protein concentration (see Note 36), and prepare protein sam-
ples for western blot analysis using appropriate antibodies.

4  Notes

1. Yeast nitrogen base without amino acid and DO supplements

is very hygroscopic for curdling. Both them need to be stored
in a humidity cabinet.
2. Aliquot stock carrier DNA in 100 μL working stock aliquots to
avoid quality changes by repeated heating.
3. The pH of 1 M lithium acetate must be adjusted to 7.5 by
acetic acid.
4. In general, we routinely obtain 100–200 colonies for a suc-
cessful cotransformation using 200 ng per plasmid DNA.
5. It is not necessary to prepare the stock solution for DO supple-
ment. Add DO supplement directly to minimal SD medium
before autoclaving.
6. Yeast protein extraction buffer must be freshly prepared prior
to use.
7. To grow overnight culture of yeast strain AH109, use colonies
freshly streaked out (less than 2 months old). We also recom-
mend refreshing yeast AH109 strain on YPDA plate every
2 months.
8. Grow yeast AH109 strain to stationary phase, which corre-
sponds to OD600 > 1.5.
9. After 4 h incubation, the value of OD600 is about 0.3–0.4.
Note that yeast cells may be precipitated. It is recommended to
take out the flask and shake the culture a few times during
incubation.
10. Centrifugation at 4 °C or RT does not significantly affect the
transformation efficiency.
11. All steps are carried out in a laminar flow under aseptic
conditions.
12. Discard supernatant as much as possible. Add 900 μL sterile
water, then add 100 μL of 10× LiAc.
 Yeast Two-Hybrid 185

13. This is a very critical step as the excessive speed may cause yeast
cell breakage and reduce the transformation efficiency.
14. Yeast competent cells must be freshly prepared to maintain
high transformation efficiency.
15. All steps of PEG/LiAc-mediated transformation of yeast
should be carried out in a laminar flow under aseptic
conditions.
16. We recommend pretreating ssDNA at 100 °C for only
10–15 min. Prolonged heating may cause instability of ssDNA.
17. This step should be carried out on ice to maintain low
temperature.

18. We recommend using 200 ng/plasmid for PEG/LiAc-­
mediated cotransformation of yeast.
19. We recommend vortexing mixture for only 1 s before incuba-
tion in incubator at 30 °C for 30 min. LiAc-PEG solution can
be prepared during incubation time.
20. LiAc-PEG solution is very sticky, so it is better to use blunt-­
end tip by cutting the end of a regular pipette tip to draw the
solution. This step is very critical and must be done within
2 min. Otherwise, the following resuspending step will be dif-
ficult to perform. Therefore, avoid handling more than ten
samples at the same time.
21. Resuspend cell mixture immediately by vortexing using maxi-
mum speed for 2–3 s.
22. Before heat shock, we recommend gently shaking microtube
several times to mix cell mixture well.
23. Pellet cells directly by centrifugation at 14,500 × g for 1 min.
It is not necessary to incubate cells on ice before
centrifugation.
24. Final cell suspension can be stored overnight at 4 °C before
use.
25. In general, we recommend picking six to eight single colonies
for further analysis. If possible, choose relatively large colonies,
which usually correlate with high protein expression levels.
26. We highly recommend using a flat toothpick (750 flat tooth-
picks, Diamond Brands) to patch single colony on selection
plate. The use of a sharp toothpick will often cause the break-
age of the agar surface. The yeast cells are ready for further
protein–protein interactions test when the yeast cells have
grown nicely after 2–3 days of incubation.
27. Use a flat toothpick to patch cells on selection plate. In gen-
eral, it is recommended to patch three individual colonies for
186 Jer-Sheng Lin and Erh-Min Lai

protein–protein interaction analysis in each tested interacting

pair. In many cases, the growth rate of colonies on selection
plate correlates with the binding strength of the two tested
proteins.
28. It is highly recommended to perform western blot analysis to
confirm the proper expression of tested proteins using com-
mercially available antibody for tagged epitope of fusion
proteins.
29. To prepare the yeast protein extraction, use yeast cells freshly
grown on plates within a week.
30. It is recommended to use appropriate SD minimal medium
with selection to maintain extrachromosomal plasmid(s) of
transformants. The use of 2× SD minimal medium with
more nutrients can facilitate faster and better growth of
transformants.
31. During late log phase, the ADH1 promoter shuts down and
the expression level of endogenous yeast proteases is increased.
Therefore, do not grow yeast cells beyond saturation.
32. Resuspend and wash cell pellets by vortexing at maximum
speed for 2–3 s after adding 10 mL sterile water.
33. Because it is very difficult to take accurate amount of acid-­
washed glass beads using a pipet, we recommend to use a small
spatula instead.
34. We use the Vortex-Genie 2 mixer (Scientific Industries, Inc.)
to vortex the tubes at maximum speed for 30 s. Wear thick
gloves to prevent your hands from being temporarily paralyzed
during vortex.
35. The purpose of this step is to remove cell debris and glass beads
by centrifugation.
36. To minimize the amount of protein extract used to determine
protein concentration, it is recommended to use the NanoDrop
1000 Spectrophotometer (Thermo Fisher Scientific Inc.) for
the measurement.

Acknowledgements

This work was supported by a research grant from the Ministry of

Science and Technology (MOST 104-2311-B-001-025 -MY3) to
E.M. Lai. J. S. Lin is the recipient of postdoctoral fellowships from
Academia Sinica.
 Yeast Two-Hybrid 187

References
1. Fields S, Song O (1989) A novel genetic sys- VirB9 in the membranes of Agrobacterium
tem to detect protein–protein interactions. tumefaciens. J Bacteriol 179:1211–1218
Nature 340:245–246 11. Das A, Anderson LB, Xie YH (1997)
2. Stasi M, De Luca M, Bucci C (2015) Two-­ Delineation of the interaction domains of
hybrid-­based systems: powerful tools for inves- Agrobacterium tumefaciens VirB7 and VirB9
tigation of membrane traffic machineries. by use of the yeast two-hybrid assay. J Bacteriol
J Biotechnol 202:105–117 179:3404–3409
3. Clontech (2007) Matchmaker™ GAL4 two-­ 12. Das A, Xie YH (2000) The Agrobacterium
hybrid system 3 & libraries user manual. T-DNA transport pore proteins VirB8, VirB9,
http://www.clontech.com/images/pt/ and VirB10 interact with one another.
PT3247-1.PDF J Bacteriol 182:758–763
4. Chien CT, Bartel PL, Sternglanz R, Fields S 13. Ma LS, Lin JS, Lai EM (2009) An IcmF family
(1991) The two-hybrid system: a method to protein, ImpLM, is an integral inner mem-
identify and clone genes for proteins that inter- brane protein interacting with ImpKL, and its
act with a protein of interest. Proc Natl Acad walker a motif is required for type VI secretion
Sci U S A 88:9578–9582 system-mediated Hcp secretion in
5. Causier B, Davies B (2002) Analysing protein– Agrobacterium tumefaciens. J Bacteriol
protein interactions with the yeast two-hybrid 191:4316–4329
system. Plant Mol Biol 50:855–870 14. Lin JS, Ma LS, Lai EM (2013) Systematic dis-
6. Petschnigg J, Groisman B, Kotlyar M, Taipale section of the Agrobacterium type VI secretion
M, Zheng Y et al (2014) The mammalian-­ system reveals machinery and secreted compo-
membrane two-hybrid assay (MaMTH) for nents for subcomplex formation. PLoS One
probing membrane–protein interactions in 8:e67647
human cells. Nat Methods 11:585–592 15. Bonemann G, Pietrosiuk A, Diemand A,
7. Reece-Hoyes JS, Barutcu AR, McCord RP, Zentgraf H, Mogk A (2009) Remodelling of
Jeong JS, Jiang L et al (2011) Yeast one-hybrid VipA/VipB tubules by ClpV-mediated thread-
assays for gene-centered human gene regulatory ing is crucial for type VI protein secretion.
network mapping. Nat Methods 8:1050–1052 EMBO J 28:315–325
8. Reece-Hoyes JS, Marian Walhout AJ (2012) 16. Lossi NS, Manoli E, Forster A, Dajani R,

Yeast one-hybrid assays: a historical and techni- Pape T et al (2013) The HsiB1C1 (TssB-
cal perspective. Methods 57:441–447 TssC) complex of the Pseudomonas aerugi-
9. Tsai YL, Chiang YR, Narberhaus F, Baron C, nosa type VI secretion system forms a
Lai EM (2010) The small heat-shock protein bacteriophage tail sheathlike structure. J Biol
HspL is a VirB8 chaperone promoting type IV Chem 288:7536–7548
secretion-mediated DNA transfer. J Biol Chem 17. Ito H, Fukuda Y, Murata K, Kimura A

285:19757–19766 (1983) Transformation of intact yeast cells

10. Baron C, Thorstenson YR, Zambryski PC treated with alkali cations. J Bacteriol
(1997) The lipoprotein VirB7 interacts with 153:163–168
 Chapter 15

Protein–Protein Interactions: Cytology Two-Hybrid

Krishnamohan Atmakuri

Abstract
Identifying protein–protein interactions between the machine components of bacterial secretion systems
and their cognate substrates is essential. Establishing which component and substrate interactions are
direct or indirect further facilitates (1) advancing the architecture and assembly of the machines and (2)
understanding the substrates’ translocation mechanistics. Currently, though biochemical means exist for
identifying such direct interactions, they primarily remain in vitro and are quite labor intensive. Thus,
adopting genetic approaches to help visualize these interactions in vivo is quick and advantageous. Here I
describe bimolecular fluorescence complementation and cytology-based two-hybrid assays that could easily
be adopted to understand the bacterial secretions systems.

Key words Bimolecular fluorescence complementation (BiFC), Cytology-based two-hybrid (C2H),

Nonfluorescing halves, Protein–protein interactions, Retargeting fluorescence

1  Introduction

Evaluating protein–protein interactions among machine components

of protein secretion systems and their cognate substrates is neces-
sary and essential. To visualize them directly in bacterial cells, in
the last decade, bimolecular fluorescence complementation (BiFC)
and cytology-based two-hybrid (C2H) assays have gained significant
momentum [1–4].
Bimolecular fluorescence complementation. BiFC primarily
requires the functional reconstitution of a fluorescent protein, such
as green fluorescent protein (GFP) and yellow fluorescent protein
(YFP). When complementary but nonfluorescing portions of a flu-
orescent protein that are genetically fused to putative protein part-
ner pairs are brought together to restore fluorescence, the protein
partners under evaluation are said to interact [5]. In recent times,
improved versions of BiFC have evolved to help identify protein–
protein interactions in varied environments and diverse systems
[6–10]. Since BiFC operates through reconstitution, it is largely
immune to influence by cellular conditions. Consequently, it

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_15, © Springer Science+Business Media LLC 2017

189
190 Krishnamohan Atmakuri

supersedes fluorescence resonance energy transfer (FRET)-based

in vivo visualization of protein–protein interactions [11], which is
hypersensitive to interference from the same cellular conditions [5].
Further, BiFC assays for a given protein partner complex is not
hampered by other interacting proteins because such interactions
remain largely invisible [5, 8]. Finally, though several reporter pro-
teins, such as ubiquitin, β-galactosidase, and dihydrofolate reduc-
tase, have also been fragmented and assembled back together
through interacting partners, the reconstitution of fluorescence
with fluorescent fragments eliminates requirements such as, for
example, additional stains and reagents or stoichiometric protein-­
level constraints to visualize protein complexes.
Despite several advantages of BiFC, under some conditions
and in a few experimental model systems, without being fused to
interacting partners, the nonfluorescent halves intrinsically come
together to fluoresce. Therefore, before setting up several BiFC
assays, it is always useful to first test, with a couple of fluorescent
proteins’ halves, for the specificity of fluorescence complementa-
tion (in the system under study) and then move forward.
Cytology-based two-hybrid: In contrast to BiFC, C2H primarily
involves the targeting of interacting protein partners by cell divi-
sion proteins such as DivIVA (from Bacillus subtilis) or FtsZ (from
Escherichia coli) to their native localizing sites, i.e., poles and mid-
cell, respectively. Thus, when either of the cell division proteins
fused to a bait protein target the interacting GFP/YFP-fused prey
protein to the midcell/poles, the protein partners in question are
said to interact [2]. While BiFC efficiently establishes interactions
between soluble machine components [1, 4], C2H helps explore
interactions between soluble and membrane-associated machine
components of any secretion system [1, 2]. Recent advances in
BiFC also make it possible to detect such interactions and deter-
mine membrane protein topology [5]. Since C2H involves the
targeting of partner proteins to the midcell or poles, this assay
works best when prey and bait proteins per se do not exhibit simi-
lar localization patterns. FtsZ fused to a bait protein can sometimes
lead to cell filamentation, especially when the fusion protein domi-
nates native FtsZ during cell division.

2  Materials

2.1  Bimolecular 1. Plasmid vectors (either regular or Gateway-based) for the

Fluorescence expression of potential interacting partners as fusions to non-
Complementation fluorescing halves. The plasmids must be of (i) different
incompatibility groups, (ii) different antibiotic markers, and
(iii) similar copy numbers and (iv) preferably harbor identical
promoters for expression.
 Cytology Two-Hybrid 191

2. DNA encoding of either full length or N- and C-terminal

nonfluorescing halves of fluorescent protein under consider-
ation. (Also refer to Table 2 in ref. [12] for fragment lengths
of each half of fluorescent protein(s) under consideration).
3. DNA encoding proteins of interest, i.e., the interacting
partners.
4. Appropriate cloning primers (with required restriction sites
designed into them) for polymerase chain reaction (PCR)-based
amplification followed by directed cloning of the required
halves and interacting proteins. Alternatively, Gateway-­based
pDONR vectors could be used for cloning and then the poten-
tial interacting proteins moved to appropriate pDEStination
vectors harboring the two nonfluorescing halves.
5. DNA encoding mutated proteins of interest OR site-­directed
mutagenesis kit for generating mutant proteins to test that
indeed the interaction of protein partners is driving nonfluo-
rescent halves to interact.
6. Competent cells of E. coli and other bacterial systems under
consideration (if any). If using Gateway technology, use E. coli
DH5α for selection and E. coli DB3.1 for cloning and main-
taining pDONR and pDEStination vectors.
7. Appropriate growth media for in vitro growing of bacteria
under consideration.
8. Additional reagents to confirm fusions by immunoblotting.
9. Electroporator (for electroporating constructs into competent
cells) OR a water bath or dry heat block (for heat shock-for
the transformation of constructs into chemical-based compe-
tent cells).
10. Shaking incubator.
11. Fluorescence microscope equipped with 20× to 100× objectives,
a 100× oil-immersion phase-contrast objective, a charge-­coupled
device (CCD) camera, appropriate filters to help visualize fluo-
rescent proteins [13], and accompanying software for image
capture, image analysis, and instrument control.

2.2  Cytology-Based 1. Plasmid vectors (either regular or Gateway-based) for the

Two-Hybrid expression of potential interacting partners as fusions to either
the cell division protein or fluorescent reporter. Again, the
plasmids must be of (i) different incompatibility groups, (ii)
different antibiotic markers, and (iii) similar copy numbers and
(iv) harbor an identical promoter for expression.
2. DNA encoding fluorescent protein and cell division proteins.
3. DNA encoding proteins of interest, i.e., the interacting partners
under study.
192 Krishnamohan Atmakuri

4. Appropriate cloning primers (with required restriction sites

designed into them) for PCR-based amplification followed by
directed cloning of interacting proteins, cell division proteins,
and fluorescent reporters. Alternatively, Gateway-based
pDONR vectors could be used for cloning and then the
potential interacting proteins moved to appropriate pDESti-
nation vectors harboring fluorescing reporter or cell division
proteins.
5. DNA encoding mutated proteins of interest OR site-directed
mutagenesis kit for generating mutant proteins to test that
indeed the interaction of protein partners is driving the fluo-
rescent reporter to midcell/poles in cells.
6. Competent cells of E. coli and other bacterial systems under
consideration (if any). If using Gateway technology, use E. coli
DH5α for selection and E. coli DB3.1 for cloning and main-
taining pDONR and pDEStination vectors.
7. Appropriate growth media for in vitro growing of bacteria
under consideration.
8. Additional reagents to confirm fusions by immunoblotting.
9. Electroporator (for electroporating constructs into competent
cells) OR a water bath or dry heat block (for heat shock-for
the transformation of constructs into chemical-based compe-
tent cells).
10. Shaking incubator.
11. Fluorescence microscope equipped with 20× to 100× objectives,
a 100× oil immersion phase-contrast objective, a CCD camera,
appropriate filters to help visualize fluorescent proteins [13],
and accompanying software for image capture, image analysis,
and instrument control.

3  Methods

Unless otherwise noted, all steps may be performed at room

temperature.

3.1  Bimolecular 1. Select fluorescent protein, its halves, and appropriate fusion
Fluorescence sites (see Notes 1 and 2).
Complementation 2. Select appropriate controls (see Note 3).
3. Amplify required DNA fragment by PCR (see Note 4).
4. Use standard restriction-ligation procedures to clone PCR
amplicons into appropriate expression vectors (see Note 5).
5. Transform ligation mixture into electro- or chemically competent
bacteria (see Notes 6 and 7).
 Cytology Two-Hybrid 193

6. Inoculate four to five colonies from freshly transformed bacte-

ria separately into required growth media and grow under
required conditions to approximately an OD (optical density
at A600 nm) of 0.1.
7. Induce the expression of the proteins under study with appro-
priate inducers for required duration (see Note 8).
8. Wash a few hundred cells with fresh medium (to stop
induction).
9. Observe cells under fluorescent microscope (see Notes 9–11).
10. Perform image analyses using ImageJ or commercially available
software that comes with the fluorescent microscopes of most
companies (see Note 12).

3.2  Cytology-Based 1. Select fluorescent protein, its halves, and appropriate fusion
Two-Hybrid sites (see Notes 13 and 14).
2. Select appropriate controls (see Note 15).
3. Amplify the required DNA fragment by PCR (see Note 4).
4. Use standard restriction-ligation procedures to clone PCR
amplicons into appropriate expression vectors (see Note 5).
5. Transform ligation mixture into electro- or chemically competent
bacteria (see Notes 6 and 7).
6. Inoculate four to five colonies from freshly transformed bacteria
separately into required growth media and grow under
required conditions to approximately an OD (optical density
at A600 nm) of 0.1.
7. Induce expression of proteins under study with appropriate
inducers for required duration (see Note 8).
8. Wash a few hundred cells with fresh media (to stop induction).
9. Observe cells under fluorescent microscope (see Notes 9–11).
10. Perform image analyses using ImageJ or commercially available
software that comes with the fluorescent microscopes of most
companies (see Note 16).

4  Notes

1. For interaction studies in Agrobacterium, we have used GFP

halves, N’GFP (1–154 amino-acid residues), and GFP’C
(153—end) [1, 4]. YFP and CFP halves can also be evaluated
[12]. While generating fusion proteins, no linkers/spacers
were used while connecting protein partners to nonfluoresc-
ing halves. However, recently, linkers/spacers of 5–17 amino
acid residues have been used with better results in eukaryotic
model systems [12].
194 Krishnamohan Atmakuri

2. The success of BiFC also largely relies on the end of the

protein partners to which (N- or C-terminus) the nonfluores-
cent fragments gets fused. Primarily, we have fused the
N-terminus end of the N’GFP half to the C-terminal end of a
protein partner and the C-terminus end of the C’GFP half to
the N-­terminal end of another protein partner. However, it is
important to evaluate other fusion ends (eight combinations
in total—either ends of the partner proteins and nonfluoresc-
ing halves) to narrow down the suitable fusion ends. It is also
important to ensure that the fused protein partners do not
show altered localization patterns (e.g., cytosolic to mem-
brane bound and vice versa). All fused protein partners need
to be evaluated for their expression kinetics and accumulation
by western blot analysis.
3. To make sure that BiFC works in a given model system under
study, appropriate controls are to be evaluated first: (i) clone
and express nonfluorescing halves alone (under the same pro-
moter used for experimental study) to confirm that the halves
per se do not interact. If they do, then switch to test several
other fluorescent protein halves [12]; (ii) clone and express
noninteracting protein partners (from established studies)
fused to the same nonfluorescing halves to make sure that the
noninteracting partners do not bring the nonfluorescing
halves together to exhibit fluorescence; (iii) clone and express
nonfluorescing halves fused to protein partner (under study)
mutants (point mutants—at their site of interaction); this also
helps to confirm/evaluate various residues at interaction sites.
However, if the interacting protein pair is fairly uninvestigated
or their site of interaction is undetermined, this negative con-
trol could easily be skipped. Alternatively, BiFC could be per-
formed with these constructs to determine the site of
interaction; (iv) clone and express either protein partners fused
to a nonfluorescing half together with another construct
expressing the other nonfluorescing half alone; this eliminates
spurious interactions driving fluorescence; (v) clone and
express positively interacting protein partners (from e­ stablished
studies) fused to same nonfluorescing halves to make sure that
the assay works under the conditions being used.
4. Any high-fidelity proofreading polymerase could be used for
PCR-based amplification of the required DNA fragments.
5. Standard cloning techniques could be applied to move frag-
ments of interest to expression vectors chosen for the study. For
Gateway-based cloning, the required kits are available with Life
Technologies (now associated with Thermo Fisher Scientific).
6. Commercially available electro- or chemically competent
bacteria or those generated by standard methods can be employed
for transformations. We generally transform with 10–25 ng
plasmid DNA to obtain several hundred colonies.
 Cytology Two-Hybrid 195

7. Better fluorescence levels are usually obtained with freshly

transformed cells than when using cells stored at 4 °C or
stocked at −80 °C.
8. Proper gene induction requires standardization. This depends
on the model system under study, the type of inducer, the
copy number of plasmids, and toxicity issues.
9. Generally, good images can be obtained by observing cells
under a 100× oil-immersion phase-contrast objective.
10. As a standard practice, it is recommended to perform immu-
noblotting to check the level of expression and monitor the
fusion proteins in the bacterial cultures under study.
11. If nonfluorescing halves of either YFP or CFP (BiFC) or YFP
and CFP (C2H) are used, the cells might need to be briefly
incubated at 30 °C for the fluorescent proteins to mature and
generate fluorescence of high intensity.
12. If the two nonfluorescing halves interact in the model system
under results study, alternative fluorescent protein halves must
be evaluated. If point mutations at interaction sites of protein
partners do not abolish fluorescence, then complementation
of the nonfluorescing halves could be nonspecific. At this
juncture, it might be important to determine whether higher
protein levels would affect the outcome. If so, the concentration
or time of induction could be modified. If not, an alternate
promoter could be evaluated. Otherwise, alternatives to BiFC
have been explored (e.g., C2H). However, if after evaluating
all controls (see Note 3) fluorescence is observed only when
both protein partners’ fusions are available, then the two
proteins under study are said to interact.
13. For C2H-based protein–protein interaction studies in

Agrobacterium and E. coli, for retargeting we have used two
cell division proteins DivIVA (from B. subtilis) and FtsZ (from
E. coli), and for fluorescence GFP [1, 2, 4]. YFP and CFP can
also be explored as alternate fluorescent proteins. While gen-
erating fusion proteins, no linkers/spacers were used and seem
unnecessary. However, for the success of C2H, it is important
to evaluate all possible fusion sites, i.e., the fusion of cell division
proteins and fluorescent protein to either ends of protein
partners under study.
14. We have fused the C-terminal end of FtsZ or DivIVA to the
N-terminal end of protein partners [2, 4]. As a standard prac-
tice, it is recommended to routinely perform immunoblotting
to evaluate the levels of expression and monitor protein fusion
stability in the bacterial cultures under study.
15. To make sure C2H works in model systems under study,

appropriate controls must be evaluated first: (i) clone and
express interacting partners alone to confirm that neither of
them localizes to the midcell and poles. If one of them does
196 Krishnamohan Atmakuri

localize to these locations, while it must be fused to either of

the cell division proteins, the other partner in question must
be fused to the fluorescent protein under consideration. If,
however, both potential interacting partners localize to the
midcell/poles, C2H cannot be utilized as the method of study
for interactions; (ii) clone and express GFP/YFP/CFP fused
protein partners (under study) to confirm that neither fusion
localizes to the midcell/poles. If any fusions localize to the
poles/midcell, it is important to determine whether the local-
ization is natural or merely an artifact of inclusion bodies.
Thus, expression and localization could be evaluated with
alternate promoter(s), by altering the inducer concentrations,
or by modifying the inducing time or temperature; (iii) clone
and express fluorescent protein/cell division proteins fused to
protein partner (under study) mutants (point mutants—at
their site of interaction); this also helps confirm/evaluate vari-
ous residues at interaction sites. However, if the interacting pro-
tein pair is fairly uninvestigated or its site of interaction
undetermined, this negative control could easily be skipped; (iv)
clone and express either protein partners fused to a fluorescent
protein together with another construct expressing either cell
division protein alone-this eliminates spurious interactions driv-
ing retargeting; (v) clone and express positively interacting pro-
tein partners (from established studies) fused to retargeting cell
division proteins/fluorescent protein to make sure that the
assay works under the conditions being used.
16. If in the model system under study the two protein partners
localize to the midcell/poles, C2H cannot be employed. If
point mutations at the interaction sites of the protein partners
do not abolish retargeting, then localization could be nonspe-
cific or a consequence of inclusion bodies. At this juncture, it
might be important to determine whether higher protein lev-
els would affect the outcome. If so, the concentration or time
of induction could be modified. If not, an alternate promoter
could be evaluated. Otherwise, alternatives to C2H must be
explored (e.g., BiFC). However, if after evaluating all controls
(see Note 15) fluorescence is observed only at the midcell/
poles, then the two proteins under study are said to interact.

Acknowledgments

I thank Peter (Prof. Peter J. Christie) and Bill (Prof. William

Margolin), both from the Department of Microbiology and
Molecular Genetics, University of Texas Health Science Center,
Houston, Texas, USA. Peter was instrumental in mentoring and
providing an excellent opportunity to train as a post-doc in his lab
and Bill provided a lot of help, guidance, and training in fluores-
cent microscopy.
 Cytology Two-Hybrid 197

References
1. Ding Z, Atmakuri K, Christie PJ (2003) The 7. Kodama Y, Hu CD (2012) Biomolecular fluo-
outs and ins of bacterial type IV secretion sub- rescence complementation (BiFC): a 5-year
strates. Trends Microbiol 11:527–535 update and future perspectives. BioTechniques
2. Ding Z, Zhao Z, Jakubowski SJ, Atmakuri K, 53:285–298
Margolin W, Christie PJ (2002) A novel 8. Kerppola TK (2008) Bimolecular Fluorescence
cytology-­based, two-hybrid screen for bacteria Complementation (BiFC) analysis as a probe of
applied to protein–protein interaction studies protein interactions in living cells. Annu Rev
of a type IV secretion system. J Bacteriol Biophys 37:465–487
184:5572–5582 9. Hu CD, Chinenov Y, Kerppola TK (2002)
3. Taylor KW, Kim JG, Su XB, Aakre CD, Roden Visualization of interactions among bZIP and
JA, Adama CM, Mudgett MB (2012) Tomato Rel proteins in living cells using bimolecular
TFT1 is required for PAPMP-triggered immu- fluorescence complementation. Mol Cell
nity and mutations that prevent T3S effector 9:789–798
XopN from binding to TFT1 attenuate 10. Hu CD, Kerppola TK (2003) Simultaneous
Xanthomonas virulence. PLoS Pathog visualization of multiple protein interactions in
8:e1002768 living cells using multicolor fluorescence com-
4. Atmakuri K, Ding Z, Christie PJ (2003) VirE2, plementation analysis. Nat Biotechnol 21:
a type IV secretion substrate, interacts with the 539–545
VirD4 transfer protein at cell poles of 11. Jares-Erijman EA, Jovin TM (2003) FRET
Agrobacterium tumefaciens. Mol Microbiol imaging. Nat Biotechnol 21:1387–1395
49:1699–1713 12. Kerppola TK (2006) Design and
5. Kerppola TK (2006) Visualization of molecu- Implementation of Bimolecular Fluorescence
lar interactions by fluorescence complementa- Complementation (BiFC) assays for the visual-
tion. Nat Rev Mol Cell Biol 7:449–456 ization of protein interactions in living cells.
6. Zhang XE, Cui Z, Wang D (2016) Sensing of Nat Protoc 1:1278–1286
biomolecular interactions using fluorescence 13. Shaner NC, Steinbach PA, Tsein RY (2005)
complementing systems in living cells. Biosens A guide to choosing fluorescent proteins. Nat
Bioelectron 76:243–250 Methods 2:905–909
 Chapter 16

Fusion Reporter Approaches to Monitoring Transmembrane

Helix Interactions in Bacterial Membranes
Laureen Logger, Abdelrahim Zoued, and Eric Cascales

Abstract
In transenvelope multiprotein machines such as bacterial secretion systems, protein–protein interactions not
only occur between soluble domains but might also be mediated by helix–helix contacts in the inner mem-
brane. Here we describe genetic assays commonly used to test interactions between transmembrane α-helices
in their native membrane environment. These assays are based on the reconstitution of dimeric regulators
allowing the control of expression of reporter genes. We provide detailed protocols for the TOXCAT and
GALLEX assays used to monitor homotypic and heterotypic transmembrane helix–helix interactions.

Key words Membrane protein, Protein-protein interaction, Transmembrane segment, Helix–helix

interaction, One-hybrid, Two-hybrid, cI repressor, TOXCAT, GALLEX

1  Introduction

The proper assembly of multiprotein complexes such as bacterial

secretion systems requires specific interactions between the differ-
ent subunits. While most of the interactions involve contacts
between soluble domains of these subunits, the transmembrane
helices (TMHs) of inner membrane proteins are also key players in
membrane protein complex formation. For examples, the Type II
secretion (T2SS)-associated GspC, GspL, and GspM proteins inter-
act with each other via their TMHs [1]. A similar situation has been
evidenced for the Type VI secretion system (T6SS) TssLM complex
[2–4]. The TMH could be involved in homotypic interaction, i.e.,
participate in the formation of dimers such as the Type IV secretion
(T4SS) and T6SS-associated VirB10 and TssL inner membrane
proteins [4, 5] or in heterotypic interactions with other subunits
[1–3]. Monitoring interactions between TMHs is not an easy task
because mutations within or swapping of the TMH could interfere
with the conformation of the soluble domains and ­therefore may
indirectly affect protein–protein interactions. Thus, genetic one- or
two-hybrid approaches based on fusion to transcriptional

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_16, © Springer Science+Business Media LLC 2017

199
200 Laureen Logger et al.

reporters, such as the λcI repressor, TOXCAT, GALLEX, and

Bacterial Adenylate Cyclase-Based Two-Hybrid (BACTH) assays,
have been developed. While cI-repressor and TOXCAT can only be
used for testing homotypic interactions, the GALLEX and BACTH
approaches can also be used to monitor interactions between differ-
ent TMHs. This chapter provides protocols to monitor homotypic
and heterotypic transmembrane helix–helix interactions using
TOXCAT and GALLEX. We refer the reader to excellent reviews
summarizing the forces exerted to catalyze TMH folding and inser-
tion, as well as the different methods to analyze TMH interactions
in bacteria [6, 7].

1.1  Monitoring TMH Methods of testing the homodimerization of TMHs, such as the λ cI
Homotypic repressor and TOXCAT assays, are based on the one-hybrid reporter
Interactions fusion approach.
The cI transcriptional regulator represses the expression of
early promoters of the bacteriophage λ genome. Repression only
occurs when cI dimerizes, a behavior conferred by the C-terminal
domain. The λcI repressor assay is therefore based on the reconsti-
tution of a dimeric λcI repressor by two interacting fragments [8–10].
The construct consists of a fusion between the monomeric
N-terminal DNA-binding domain of λcI (called cI’) with the TMH
(Fig. 1a). TMH-mediated cI’ dimerization induces binding of cI
to its operator sequence, allowing repression of phage λ early
genes, hence conferring protection against superinfection by phage
λ (Fig. 1a). The cI repressor assay has been successfully used to
demonstrate that the T2SS XcpR and T4SS VirB4 and VirB11
proteins oligomerize [11–13].
The TOXCAT assay is based on the characteristics of the
Vibrio cholerae ToxR regulator: a strict dimerization-dependent
transcriptional activator consisting of an N-terminal helix-turn-
helix DNA-­ binding domain and a C-terminal dimerization
domain. The construct consists of a fusion in which the TMH is
inserted between the monomeric ToxR DNA-binding domain
and the MalE periplasmic protein (Fig. 1b). By supporting growth
on maltose-­minimal media, MalE makes it possible to verify that
the TMH is properly inserted. TMH-mediated ToxR dimeriza-
tion induces binding of ToxR on its operator sequence, allowing
transcription of a reporter gene. In the initial ToxR system, the
reporter gene is lacZ [14], while the TOXCAT assay uses the cat
gene [15] (Fig. 1b). Hence, dimerization of the TMH could then
be assessed by measuring the β-galactosidase and chloramphenicol
acetyltransferase (resistance to chloramphenicol) activities [16].
The TOXCAT assay has been successfully used to provide evi-
dence that the TMH of the T4SS VirB10 subunit oligomerizes
[5]. Further improvements of the ToxR and TOXCAT assays have
been published [17–19].
 Protein-Protein Interaction in the Membrane 201

Fig. 1 Schematic representation of the assays for monitoring TMH homotypic interactions. (a) cI repressor
assay. The TMH of interest (orange) is fused to the cI DNA-binding domain (green). Reconstitution of the cI
dimer results in the repression of early phage λ genes (blue). Repression of phage λ genes confers protection
against phage λ infection. (b) ToxR and TOXCAT assays. The TMH of interest (orange) is fused between the
Vibrio cholerae ToxR DNA-binding domain (red) and the MalE protein (blue). Reconstitution of the ToxR dimer
results in the expression of the reporter genes (blue, lacZ for the ToxR assay, cat for the TOXCAT assay)

1.2  Monitoring TMH Methods of testing the heterodimerization of TMHs, such as the
Heterotypic GALLEX and BACTH assays, are based on the two-hybrid
Interactions reporter fusion approach.
The GALLEX assay is based on the reconstitution of a dimeric
LexA transcriptional repressor by two interacting TMHs. The con-
struct consists of a fusion in which each TMH is inserted between
the monomeric LexA N-terminal DNA-binding domain and the
MalE periplasmic protein. The elegant improvement is that one of
the two TMHs is fused to the wild-type LexA N-terminal
domain (LexAWT), whereas the second TMH is fused to a LexA
N-terminal domain variant bearing a mutation in the DNA-binding
motif (LexA408), allowing recognition of a different operator
sequence (op408). Formation of helix heterodimers induces binding
of LexA/LexA408 on a dual operator sequence (opWT/op408), allow-
ing repression of a reporter gene [20–22] (Fig. 2a).
The BACTH assay is based on the reconstitution of the adenyl-
ate cyclase activity conferred by the T18 and T25 domains of the
Bordetella pertussis Cya protein [23–25] (Fig. 2b). Widely
202 Laureen Logger et al.

Fig. 2 Schematic representation of assays for monitoring TMH heterotypic interactions. (a) GALLEX assay. The
first TMH of interest (orange) is fused between the wild-type LexA DNA-binding domain (WT LexA) and MalE,
whereas the second TMH (blue) is fused between the LexA408 variant (LexA 408) and MalE. Reconstitution of
the LexAWT/LexA408 dimer results in the repression of the reporter gene (blue). (b) BACTH assays. The first TMH
of interest (orange) is fused to the T18 domain of the B. pertussis adenylate cyclase, whereas the second TMH
(blue) is fused to the T25 of adenylate cyclase. Reconstitution of the T18/T25 adenylate cyclase results in the
production of cyclic adenosine monophosphate (cAMP). Binding of cAMP to the catabolite activator protein
(CAP) induces the expression of the reporter gene (blue)

used for testing interactions between soluble domains or proteins in

multiprotein complexes such as the divisome or secretion systems
[26–34], it has only rarely been used for the study of transmem-
brane helix–helix interactions [35–37]. A detailed protocol for the
bacterial two-hybrid assay is described in Chapter 13. In this chap-
ter, we provide protocols for the TOXCAT and GALLEX assays.

2  Material

2.1  Monitoring TMH 1. pcckan vector [15] (see Note 1).

Homotypic Interactions: 2. Escherichia coli NT326 or MM39 bacterial strains [15] (see
The TOXCAT Assay Note 2).
3. Lysogeny broth (LB) medium: Dissolve 10 g tryptone, 5 g
yeast extracts, and 10 g NaCl in 1 L distilled water. Autoclave
for 15 min at 121 °C. For LB agar plates, add 15 g bacto-agar
prior to autoclaving.
 Protein-Protein Interaction in the Membrane 203

4. M9-maltose medium: Dissolve 0.6 g Na2HPO4·12H2O, 0.3 g

KH2PO4·H2O, 50 mg NaCl, 100 mg NH4Cl, and 1.5 g bacto-­
agar in 90 mL distilled water. Autoclave. Add 100 mg casa-
mino acids, 400 mg maltose, 25 mg MgSO4·7H2O, and 1 mg
CaCl2.
5. Ampicillin stock solution (250×): 25 mg/mL ampicillin.
Dissolve 250 mg ampicillin in 10 mL distilled water. Filter to
sterilize. Store at 4 °C.
6. Chloramphenicol stock solution: Dissolve 90 mg chloram-
phenicol in 1 mL absolute ethanol. Store at −20 °C.
7. 2.5 mM chloramphenicol solution: Dissolve 8.1 mg chloram-
phenicol in 10 mL ethanol.
8. Sodium dodecyl sulfate (SDS)-polyacrylamide gel electropho-
resis (PAGE) loading buffer: 60 mM Tris–HCl, pH 6.8, 2%
SDS, 10% glycerol, 5% β-mercaptoethanol, 0.01% bromophe-
nol blue.
9. Lysis buffer: 25 mM Tris–HCl, 2 mM EDTA, pH 8.0. Dissolve
303 mg Tris(hydroxymethyl) aminomethane and 58 mg ethyl-
enediaminetetraacetic acid (EDTA, disodium salt) in 100 mL
sterile distilled water. Adjust pH to 8.0.
10. Reaction buffer: 100 mM Tris–HCl, pH 7.8, 0.1 mM acetyl-­
CoA, 0.4 mg/mL 5,5’-dithiobis-(2 nitrobenzoic acid) (dTNB).
Dissolve 121 mg Tris, 0.81 mg acetyl-CoA, and 4 mg dTNB in
10 mL sterile distilled water. Adjust pH to 7.8 with HCl.
11. 10 mm filter paper disk.
12. 96-well microplates.
13. Anti-maltose-binding protein (MBP) antibodies for MalE
immunodetection.
14. Incubator.
15. Spectrophotometer.
16. Benchtop centrifuge.
17. Water bath at 96 °C.
18. Mini-gel caster system and SDS-PAGE apparatus.
19. Protein blotting apparatus.
20. Sonifier.
21. Microplate reader.

2.2  Monitoring TMH 1. pALM148 and pBML100 vectors [20].

Heterotypic 2. E. coli NT326 or MM39 bacterial cells [15] (see Note 2).
Interactions:
3. E. coli SU202 bacterial strain [20, 38] (see Note 3).
The GALLEX Assay
4. LB medium: see Subheading 2.1.
5. M9-maltose medium: see Subheading 2.1.
204 Laureen Logger et al.

6. Ampicillin stock solution (250×): see Subheading 2.1.

7. Tetracyclin stock solution (1000×): 12 mg/mL tetracyclin.
Dissolve 120 mg tetracyclin in 10 mL ethanol. Filter to steril-
ize. Store at 4 °C.
8. Isopropyl-β-d-thiogalactopyranoside (IPTG) stock solution
(500×). 0.1 M IPTG. Dissolve 238 mg IPTG in 10 mL sterile
distilled water. Filter to sterilize. Store at 4 °C.
9. X-Gal stock solution (1000×): 40 mg/mL X-Gal. Dissolve
40 mg 5-Bromo-4-chloro-3-indolyl β-d-galactopyranoside
(X-Gal) in 1 mL dimethylformamide. Prepare fresh and do not
store.
10. 0.1% SDS: Dissolve 50 mg SDS in 50 mL distilled water.
11. Chloroform.
12. Ortho-nitrophenyl-β-d-galactopyranoside (ONPG) stock
solution. 4 mg/mL ONPG: Dissolve 20 mg ONPG in 5 mL
buffer Z.
13. SDS-PAGE loading buffer: 60 mM Tris–HCl, pH 6.8, 2%
SDS, 10% glycerol, 5% β-mercaptoethanol, 0.01% bromophe-
nol blue.

14. Buffer Z: Dissolve 2.15 g Na2HPO4·12 H2O, 0.29 g
Na2HPO4·H2O, 75 mg KCl, and 25 mg MgSO4·7H2O in 100
mL distilled water. Adjust pH to 7.0. Add 270 μL
β-mercaptoethanol. Prepare fresh and do not store.
15. Anti-MBP antibodies for MalE immunodetection.
16. 96-well microplates.
17. Incubator.
18. Spectrophotometer.
19. Benchtop centrifuge.
20. Water bath at 96 °C.
21. Mini-gel caster system and SDS-PAGE apparatus.
22. Protein blotting apparatus.
23. Microplate reader.

3  Methods

3.1  Monitoring TMH 1. Clone the DNA fragment corresponding to the TMH to be stud-
Homotypic ied into the pcckan vector to yield a plasmid producing the ToxR’-
Interactions: TMH-MalE fusion protein. Before testing the homodimerization
The TOXCAT Assay of the TMH, verify that your fusion protein is properly produced
(steps 3–8) and inserted in the inner membrane (steps 9 and
10). The dimerization of the TMH is then assessed by the disk
diffusion assay (steps 11–16) and quantified by measuring the
chloramphenicol acetyltransferase activity (steps 17–26).
 Protein-Protein Interaction in the Membrane 205

2. Transform the empty pcckan vector and your pcckan construct

into NT326 or MM39 E. coli competent cells. Select on LB-­
ampicillin plates (see Note 1).
3. Pick a single colony of each transformation and grow cells in 20
mL LB medium supplemented with ampicillin (100 μg/mL)
until an optical density at 600 nm (OD600) of 0.8 is reached.
4. Harvest 2 mL of cells by centrifugation at 4000 × g for 5 min.
5. Discard supernatants and resuspend cell pellets into 20 μL
SDS-PAGE loading buffer.
6. Boil samples for 10 min at 96 °C.
7. Separate proteins by SDS-PAGE and transfer onto nitrocellu-
lose membrane using your favorite protocol.
8. Use western blotting to immunodetect your fusion protein
using commercial anti-MalE (anti-MBP) antibodies.
9. Streak 20 μL of the bacterial culture obtained at step 3 in
Subheading 3.1 onto M9-maltose medium.
10. After incubation for 48 h at 37 °C, verify that your strain grew
on M9-maltose medium.
11. Drop a 10 mm filter paper disk in center of LB-ampicillin plate
(see Note 4).
12. Add 60 μL chloramphenicol stock solution (90 mg/mL) on
filter paper disk.
13. Incubate LB plates with chloramphenicol disks for 6 h at 37 °C.
14. Remove disk.
15. Spread 2 mL of the culture obtained at step 3 in Subheading
3.1 on the LB-ampicillin plate to make a lawn. Eliminate excess
culture.
16. After incubation for 16 h at 37 °C, measure the halo of chlor-
amphenicol sensitivity (see Note 5).
17. Centrifuge 3 mL of the culture obtained at step 3 in Subheading
3.1 at 4000 × g for 5 min (in triplicate).
18. Discard supernatant and resuspend cell pellets in 500 μL lysis
buffer. Vortex.
19. Lyse cells by sonication using a sonifier.
20. Clear lysate by centrifugation at 10,000 × g for 15 min.
21. In a 96-well microplate, mix 15 μL of the cleared lysate with
220 μL reaction buffer.
22. Measure absorbance at 412 nm (A412) (see Note 6) and at
550 nm (A550; cell debris) every 20 s for 4 min using a micro-
plate reader.
23. Inject 15 μL 2.5 mM chloramphenicol in each well.
206 Laureen Logger et al.

24. Measure absorbance at 412 nm (see Note 6) and at 550 nm

(cell debris) every 20 s for 10 min using a microplate reader.
25. Divide each A412 value by the corresponding A550 value and
plot these values against time.
26. Calculate the chloramphenicol acetyltransferase activity based
on the slope in the linear part of the curve (initial rate).

3.2  Monitoring TMH 1. Clone the DNA fragment corresponding to the first TMH to
Heterotypic be studied (TMH1) into the pBLM100 vector to yield a
Interactions: pBR322 derivative plasmid producing the LexAWT’-TMH1-­
The GALLEX Assay MalE fusion protein. Clone the DNA fragment corresponding
to the second TMH to be studied (TMH2) into the pALM148
vector to yield a pACYC184 derivative plasmid producing the
LexA408’-TMH2-MalE fusion protein. Before testing the het-
erodimerization of the TMH, verify that your fusion protein is
properly produced (steps 3–8) and inserted in the inner mem-
brane (steps 9 and 10). The dimerization of the TMH is then
assessed on LB-X-Gal plates (steps 11–14) and quantitated by
measuring the β-galactosidase activity (steps 15–22).
2. Transform the empty pBLM100 and pALM148 vectors as well
as the pBLM100-TMH1 and pALM148-TMH2 constructs
into NT326 or MM39 E. coli competent cells. Select on LB
plates supplemented with ampicillin (pBLM100 derivatives) or
tetracyclin (pALM148 derivatives).
3. Pick a single colony of each transformation and grow cells in
3 mL LB medium supplemented with IPTG and ampicillin or
tetracyclin until an OD600 of 0.8 is reached.
4. Harvest 2 mL of cells by centrifugation at 4000 × g for 5 min.
5. Discard supernatants and resuspend cell pellets into 20 μL of
SDS-PAGE loading buffer.
6. Boil samples for 10 min at 96 °C.
7. Separate proteins by SDS-PAGE and transfer onto nitrocellu-
lose membrane using your favorite protocol.
8. Use western blotting to immunodetect your fusion protein
using commercial anti-MalE (anti-MBP) antibodies.
9. Streak 20 μL of the bacterial culture obtained at step 3 in
Subheading 3.2 onto M9-maltose medium.
10. After incubation for 48 h at 37 °C, verify that your strain grew
on M9-maltose medium.
11. Cotransform pBLM100 and pBLM100-TMH1 vectors in
combination with pALM148 and pALM148-TMH2 vectors
into SU202 E. coli competent cells (see Note 7). Select on LB
plates supplemented with ampicillin and tetracyclin.
 Protein-Protein Interaction in the Membrane 207

12. Pick a single colony of each transformation and grow cells in

3 mL LB medium supplemented with IPTG, ampicillin, and
tetracyclin until an OD600 of 0.8 is reached.
13. Drop 15 μL of the bacterial culture obtained at step 12 in
Subheading 3.2 onto LB plates supplemented with IPTG,
ampicillin, tetracyclin, and X-Gal.
14. After 6, 14, and 24 h of incubation at 37 °C, observe the col-
oration of the spots. White spots correspond to strains with no
β-galactosidase activity (i.e., interaction between the two
TMHs), whereas blue spots correspond to strains with
β-galactosidase activity (i.e., no interaction between the two
TMHs) (see Note 8).
15. Mix 200 μL of the bacterial culture obtained at step 12 in
Subheading 3.2 with 800 μL of buffer Z into a 1.5 mL
Eppendorf tube. Vortex.
16. Add one drop of 0.1% SDS and two drops of chloroform to
lyse cells. Vortex for 10 s.
17. In a 96-well microplate, mix 50 μL of the cleared lysate with
150 μL buffer Z.
18. Measure the absorbance at 420 nm (absorption wavelength of
ortho-nitrophenol, the product of degradation of ONPG) and
at 550 nm (A550; cell debris) every 30 s for 2 min using a
microplate reader.
19. Inject 40 μL ONPG solution in each well.
20. Measure the absorbance at 420 nm and at 550 nm every 30 s
for 20 min using a microplate reader.
21. Divide each A420 value by the corresponding A550 value and
plot these values against time.
22. Calculate the β-galactosidase activity based on the slope in the
linear part of the curve (initial rate).

4  Notes

1. pcckan is a vector comprising the sequence corresponding to

the ToxR N-terminal domain and that corresponding to MalE
separated by a multiple cloning site allowing insertion of the
sequence corresponding to the TMH of interest. Positive and
negative controls have been developed by Russ and Engelman
corresponding to the wild-type and mutated TMH of the gly-
cophorin A, respectively [15].
2. NT326 and MM39 strains do not produce the MBP and
therefore could be used as reporters to verify the proper inser-
tion of the ToxR’-TMH-MalE and LexA-TMH-MalE fusions.
208 Laureen Logger et al.

3. Strain SU202 is a reporter for the GALLEX assay. It has a

chromosomally integrated fragment corresponding to a hybrid
operator sequence (opWT/op408) controlling the expression of
the lacZ reporter gene. Transformed SU202 cells are not stable
and therefore transformations should be made fresh and colo-
nies should not be stored at 4°C.
4. Use three LB-ampicillin plates per strain to be tested.
5. The diameter of the halo reflects the ability of the strain to resist
chloramphenicol and therefore is directly and inversely linked
to the expression of the cat gene that is induced by TMH
dimerization. If the TMH dimerizes, the expression level of cat
will be high and, hence, the diameter of the halo small.
6. The reaction catalyzed by the chloramphenicol acetyltransfer-
ase consists in the acetylation of the chloramphenicol and the
release of free coenzyme A. Coenzyme A then reacts with the
5,5’-dithiobis-(2-nitrobenzoic acid), resulting in an increase of
the absorbance at 412 nm.
7. You should obtain the combinations pBLM100 + pALM148,
pBLM100 + pALM148-TMH2, pBLM100-TMH1 +
pALM148, and pBLM100-TMH1 + pALM148-TMH2.
8. MacConkey/maltose could be used as reporter medium
instead of LB-X-Gal plates. If MacConkey/maltose plates are
used, the coloration of the spots differs: yellow spots corre-
spond to strains with no β-galactosidase activity (i.e., interac-
tion between the two TMHs), whereas red spots correspond
to strains with β-galactosidase activity (i.e., no interaction
between the two TMHs).

Acknowledgements

Work in the EC laboratory is supported by the Centre National de

la Recherche Scientifique, the Aix-Marseille Université, and grants
from the Agence Nationale de la Recherche (ANR-­
14-­CE14-0006-02 and ANR-15-CE11-0019-01). LL and AZ are
recipients of doctoral fellowships from the French Ministère de
l’Enseignement Supérieur et de la Recherche and end-of-thesis fel-
lowships from the Fondation pour la Recherche Médicale
(FDT20160435498 and FDT20140931060).

References

1. Lallemand M, Login FH, Guschinskaya N, 2. Ma LS, Lin JS, Lai EM (2009) An IcmF family
Pineau C, Effantin G, Robert X, Shevchik VE protein, ImpLM, is an integral inner membrane
(2013) Dynamic interplay between the peri- protein interacting with ImpKL, and its walker
plasmic and transmembrane domains of GspL a motif is required for type VI secretion system-
and GspM in the type II secretion system. mediated Hcp secretion in Agrobacterium
PLoS One 8:e79562 tumefaciens. J Bacteriol 191:4316–4329
 Protein-Protein Interaction in the Membrane 209

3. Aschtgen MS, Gavioli M, Dessen A, Lloubès 15. Russ WP, Engelman DM (1999) TOXCAT: a
R, Cascales E (2010) The SciZ protein anchors measure of transmembrane helix association in
the enteroaggregative Escherichia coli Type VI a biological membrane. Proc Natl Acad Sci U S
secretion system to the cell wall. Mol Microbiol A 96:863–868
75:886–899 16. Joce C, Wiener A, Yin H (2011) Transmembrane
4. Durand E, Zoued A, Spinelli S, Watson PJ, domain oligomerization propensity determined
Aschtgen MS, Journet L, Cambillau C, Cascales by ToxR assay. J Vis Exp 51
E (2012) Structural characterization and oligo- 17. Lindner E, Langosch D (2006) A ToxR-based
merization of the TssL protein, a component dominant-negative system to investigate het-
shared by bacterial type VI and type IVb secre- erotypic transmembrane domain interactions.
tion systems. J Biol Chem 287:14157–14168 Proteins 65:803–807
5. Garza I, Christie PJ (2013) A putative trans- 18. Lindner E, Unterreitmeier S, Ridder AN,
membrane leucine zipper of agrobacterium Langosch D (2007) An extended ToxR
VirB10 is essential for T-pilus biogenesis but not POSSYCCAT system for positive and negative
type IV secretion. J Bacteriol 195:3022–3034 selection of self-interacting transmembrane
6. Schneider D, Finger C, Prodöhl A, Volkmer domains. J Microbiol Methods 69:298–305
T (2007) From interactions of single trans- 19. Lis M, Blumenthal K (2006) A modified,
membrane helices to folding of alpha-helical dual reporter TOXCAT system for monitor-
membrane proteins: analyzing transmembrane ing homodimerization of transmembrane
helix-helix interactions in bacteria. Curr segments of proteins. Biochem Biophys Res
Protein Pept Sci 8:45–61 Commun 339:321–324
7. Fink A, Sal-Man N, Gerber D, Shai Y (2012) 20. Schneider D, Engelman DM (2003) GALLEX,
Transmembrane domains interactions within a measurement of heterologous association of
the membrane milieu: principles, advances transmembrane helices in a biological mem-
and challenges. Biochim Biophys Acta brane. J Biol Chem 278:3105–3111
1818:974–983 21. Cymer F, Sanders CR, Schneider D (2013)
8. Hu JC (1995) Repressor fusions as a tool to Analyzing oligomerization of individual trans-
study protein-protein interactions. Structure membrane helices and of entire membrane
3:431–433 proteins in E. coli: a hitchhiker’s guide to
9. Leeds JA, Beckwith J (1998) Lambda repressor GALLEX. Methods Mol Biol 932:259–276
N-terminal DNA-binding domain as an assay 22. Tome L, Steindorf D, Schneider D (2013)
for protein transmembrane segment interac- Genetic systems for monitoring interactions
tions in vivo. J Mol Biol 280:799–810 of transmembrane domains in bacterial mem-
10. Leeds JA, Beckwith J (2000) A gene fusion branes. Methods Mol Biol 1063:57–91
method for assaying interactions of protein 23. Karimova G, Pidoux J, Ullmann A, Ladant D
transmembrane segments in vivo. Methods (1998) A bacterial two-hybrid system based on
Enzymol 2327:165–175 a reconstituted signal transduction pathway.
11. Turner LR, Olson JW, Lory S (1997) The Proc Natl Acad Sci U S A 95:5752–5756
XcpR protein of Pseudomonas aeruginosa 24. Ladant D, Karimova G (2000) Genetic systems
dimerizes via its N-terminus. Mol Microbiol for analyzing protein-protein interactions in
26:877–887 bacteria. Res Microbiol 151:711–720
12. Dang TA, Zhou XR, Graf B, Christie PJ (1999) 25. Battesti A, Bouveret E (2012) The bacterial
Dimerization of the agrobacterium tumefaciens two-hybrid system based on adenylate cyclase
VirB4 ATPase and the effect of ATP-­binding reconstitution in Escherichia coli. Methods
cassette mutations on the assembly and func- 58:325–334
tion of the T-DNA transporter. Mol Microbiol 26. Karimova G, Dautin N, Ladant D (2005)
32:1239–1253 Interaction network among Escherichia coli
13. Rashkova S, Zhou XR, Chen J, Christie PJ membrane proteins involved in cell division
(2000) Self-assembly of the Agrobacterium as revealed by bacterial two-hybrid analysis.
tumefaciens VirB11 traffic ATPase. J Bacteriol J Bacteriol 187:2233–2243
182:4137–4145 27. Sivanesan D, Hancock MA, Villamil Giraldo
14. Langosch D, Brosig B, Kolmar H, Fritz HJ AM, Baron C (2010) Quantitative analy-
(1996) Dimerisation of the glycophorin A sis of VirB8-VirB9-VirB10 interactions pro-
transmembrane segment in membranes probed vides a dynamic model of type IV secretion
with the ToxR transcription activator. J Mol system core complex assembly. Biochemistry
Biol 263:525–530 49:4483–4493
210 Laureen Logger et al.

28. Cisneros DA, Bond PJ, Pugsley AP, Campos platform for tail tube/sheath polymerization.
M, Francetic O (2012) Minor pseudopilin self-­ PLoS Genet 11:e1005545
assembly primes type II secretion pseudopilus 34. Zoued A, Durand E, Brunet YR, Spinelli S, Douzi
elongation. EMBO J 31:1041–1053 B, Guzzo M, Flaugnatti N, Legrand P, Journet L,
29. Georgiadou M, Castagnini M, Karimova G, Fronzes R, Mignot T, Cambillau C, Cascales E
Ladant D, Pelicic V (2012) Large-scale study (2016) Priming and polymerization of a bacterial
of the interactions between proteins involved contractile tail structure. Nature 531:59–63
in type IV pilus biology in Neisseria meningiti- 35. Llosa M, Zunzunegui S, de la Cruz F (2003)
dis: characterization of a subcomplex involved Conjugative coupling proteins interact with cog-
in pilus assembly. Mol Microbiol 84:857–873 nate and heterologous VirB10-like proteins while
30. Zoued A, Durand E, Bebeacua C, Brunet YR, exhibiting specificity for cognate relaxosomes.
Douzi B, Cambillau C, Cascales E, Journet L Proc Natl Acad Sci U S A 100:10465–10470
(2013) TssK is a trimeric cytoplasmic protein 36. Segura RL, Aguila-Arcos S, Ugarte-Uribe B,
interacting with components of both phage-­ Vecino AJ, de la Cruz F, Goñi FM, Alkorta I
like and membrane anchoring complexes of the (2013) The transmembrane domain of the
type VI secretion system. J Biol Chem T4SS coupling protein TrwB and its role in
288:27031–27041 protein-protein interactions. Biochim Biophys
31. Pais SV, Milho C, Almeida F, Mota LJ (2013) Acta 1828:2015–2025
Identification of novel type III secretion 37. Sawma P, Roth L, Blanchard C, Bagnard D,
chaperone-­ substrate complexes of Chlamydia Crémel G, Bouveret E, Duneau JP, Sturgis JN,
trachomatis. PLoS One 8:e56292 Hubert P (2014) Evidence for new homotypic
32. Pineau C, Guschinskaya N, Robert X, Gouet P, and heterotypic interactions between trans-
Ballut L, Shevchik VE (2014) Substrate recog- membrane helices of proteins involved in
nition by the bacterial type II secretion system: receptor tyrosine kinase and neuropilin signal-
more than a simple interaction. Mol Microbiol ing. J Mol Biol 426:4099–4111
94:126–140 38. Dimitrova M, Younès-Cauet G, Oertel-­
33. Brunet YR, Zoued A, Boyer F, Douzi B, Buchheit P, Porte D, Schnarr M, Granger-­
Cascales E (2015) The type VI secretion Schnarr M (1998) A new LexA-based genetic
TssEFGK-VgrG phage-like baseplate is system for monitoring and analyzing protein
recruited to the TssJLM membrane complex heterodimerization in Escherichia coli. Mol
via multiple contacts and serves as assembly Gen Genet 257:205–212
 Chapter 17

Protein–Protein Interactions: Co-Immunoprecipitation

Jer-Sheng Lin and Erh-Min Lai

Abstract
Proteins often do not function as single substances but rather as team players in a dynamic network.
Growing evidence shows that protein–protein interactions are crucial in many biological processes in living
cells. Genetic (such as yeast two-hybrid, Y2H) and biochemical (such as co-immunoprecipitation, co-IP)
methods are the methods commonly used at the beginning of a study to identify the interacting proteins.
Immunoprecipitation (IP), a method using a target protein-specific antibody in conjunction with Protein
A/G affinity beads, is a powerful tool to identify molecules that interact with specific proteins. Therefore,
co-IP is considered to be one of the standard methods of identifying or confirming the occurrence of
protein–protein interaction events in vivo. Co-IP experiments can identify proteins via direct or indirect
interactions or in a protein complex. Here, we use Agrobacterium type VI secretion system (T6SS) sheath
components TssB-TssC41 interaction as an example to describe the principle, procedure, and experimental
problems of co-IP.

Key words Protein–protein interaction, Immunoprecipitation (IP), Co-immunoprecipitation (Co-­

IP), Immobilization, Protein A/G Sepharose, Physical interaction

1  Introduction

The earliest concept of immunoprecipitation (IP) was used to trace

protein turnover by pulse labeling of total proteins during transla-
tion using radioactive amino acids added in the cell culture [1, 2].
The use of antibodies for IP can cause the spontaneous precipita-
tion of antigen–antibody complexes formed by the interaction of
certain polyclonal antibodies with their antigens. As a consequence,
in the study discussed here, the antigen was purified from the pro-
tein mixture using a specific antibody immobilized on beads
directly or precipitated by affinity beads conjugated by protein
A/G, which binds the conserved region of the antibody. Purified
antigens (proteins) were then visualized by sodium dodecyl sulfate
(SDS)-polyacrylamide gel electrophoresis (PAGE) followed by
autoradiography [3]. Co-IP adapts the concept of IP to identify
interacting partners and has become one of the most popular
methods for protein–protein interaction studies in recent years.

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_17, © Springer Science+Business Media LLC 2017

211
212 Jer-Sheng Lin and Erh-Min Lai

In a typical experiment, co-IP consists of several steps, including

preparation of protein extract (usually a cell lysate), coupling a spe-
cific antibody to beads, purification of specific protein complexes,
and analysis of the co-IP complexes (Fig. 1) [4]. The unbound
proteins are washed away while the antibody, bait protein, and pro-
teins associated to the bait are eluted. Purified protein complexes
can then be identified by mass spectrometry or western blot analy-
sis. Depending on the specificity and quality of antibody and
experimental conditions, co-IP experiments may generate signifi-
cant background noise owing to nonspecific binding to the anti-
body or beads. Thus, negative controls of samples without bait
protein or antibody run in parallel are critical to identifying specific
interacting proteins. The increased sensitivity of mass spectrometry
instrumentation has also greatly reduced the quality and quantity

Agarose
Bead

Protein
A/G

Agarose
Bead
Protein
A/G
Add antibody Add Protein
A/G beads

Protein sample
(Usually a cell lysate)
Discard
Centrifugation

Supernatant
Agarose
Bead

Protein
A/G

Immunoblot analysis Other

with specific antibodies biochemical analysis Collection of Protein A/G beads

Wash

Non-bound sample
components
Sample buffer
Elute
Agarose
Bead

Protein
A/G

SDS-PAGE sample Purified complexes

Fig. 1 Schematic diagram of principle of co-IP. Antigen-containing protein sample (usually a cell lysate), spe-
cific antibody, and affinity beads (usually protein A/G, which can specifically bind to conserved region of anti-
body) are added sequentially for binding reaction. The affinity beads with bound proteins are collected by
centrifugation. The supernatant containing unbound proteins is discarded and further washed away during
washing steps. Antibody and antigen are eluted with a buffer that dissociates proteins from affinity beads.
Purified protein complexes can be further used for immunoblot or other biochemical analysis
 Co-Immunoprecipitation 213

Co-IP Ab : RpoA TssC41 RpoA TssB

I IP I IP I IP I IP
TssC41 55

TssB 15

Fig. 2 Co-IP analysis of TssB and TssC41 in A. tumefaciens. Total protein extracts
isolated from A. tumefaciens wild-type strain C58 treated with DTBP cross-linker
were solubilized by buffer containing 1% SDS, then diluted in Triton X-100-­
containing solution for IP. Coprecipitated proteins were identified by western
blotting. Co-IP was also performed with antibody against RNA polymerase alpha
subunit (RpoA) as a negative control. The proteins analyzed and sizes of molecu-
lar weight standards are indicated on the left and right, respectively, and by
arrows when necessary (I input, IP immunoprecipitation) (reproduced from Ref.
[10]; no permission is required for reuse of content published from Public Library
of Science, PLOS

of starting protein samples required for successful protein identifi-

cations, which allows for a more complete and reliable map of
interactions [5, 6].
Co-IP has been successfully used to study the interaction of proteins
involved in bacterial type IV and type VI secretion systems (T4SS and
T6SS) in Agrobacterium tumefaciens [7–11]. An example is presented in
Fig. 2, which illustrates a protocol for a co-IP experiment adapted from
“Immunoprecipitation (IP) technical guide and protocols” [3] and
“Detection of Protein–Protein Interactions by Coprecipitation” [12],
with minor modifications. The cleavable and membrane-permeable
cross-linker dimethyl 3,3′-dithiobispropionimidate (DTBP) is used to
cross-link interacting proteins before cell lysis to ensure the identification
of both stable and weak interacting proteins of type VI secretion compo-
nents, including the T6SS sheath components TssB and TssC41 in A.
tumefaciens [10]. Coprecipitated proteins are further identified by west-
ern blotting. Using this protocol, we found that TssB and TssC41 were
respectively coprecipitated with each other in A. tumefaciens (Fig. 2). In
contrast, TssC41 and TssB were not precipitated by control antibody
(Fig.  2). Together with the interacting data obtained by yeast two-
hybrid, copurification in E. coli, we concluded that TssB and TssC41
could interact strongly to form a protein complex [10].

2  Materials

2.1  Cross-Linking 1. Bacterial cells (see Note 2).

of the Bacterial Cells 2. DTBP: 0.5 M (see Note 3).
(see Note 1)
3. Phosphate buffer: 20 mM sodium phosphate, pH 7.6, 20 mM
sodium chloride.
4. 1 M Tris–HCl buffer, pH 7.6.
214 Jer-Sheng Lin and Erh-Min Lai

2.2  Preparation 1. TES buffer: 50 mM Tris–HCl, pH 6.8, 2 mM ethylenediami-

of Bacterial Cell netetraacetic acid (EDTA), 1% SDS.
Extracts (see Note 4) 2. NP1 buffer: 150 mM Tris–HCl, pH 8.0, 0.5 M sucrose, 10
mM EDTA.
3. Lysozyme (see Note 5).
4. Triton X-100.
5. Rotating wheel.
6. Protease inhibitor cocktail.

2.3  Protein Sample 1. Protein A-Sepharose™ CL4B (GE Healthcare Life Sciences)
Preclearing or equivalent (see Note 6).
2. 2 mL microtube.
3. Rotating wheel.

2.4  Coupling 1. Specific antibody and control antibody.

of Antibody to Protein 2. Protein A-Sepharose™ CL4B.
A-Sepharose Beads
3. Rotating wheel.

2.5  Purification 1. NP1 buffer supplemented with 1% Triton X-100 (see Note 7).
and Isolation 2. NP1 buffer supplemented with 0.1% Triton X-100.
of Protein Complexes
3. Elution buffer: 0.1 M glycine–HCl, pH 2.5.
4. 2× SDS sample buffer: 100 mM Tris–HCl, pH 6.8, 4% SDS,
20% glycerol, 5% 2-mercaptoethanol, 2 mM EDTA, 0.1 mg/
mL bromophenol blue (see Note 8).

2.6  TrueBlot 1. Minigel caster system and SDS-PAGE apparatus.

for Protein Detection 2. Transblot apparatus for western blot transfer.
of Co-IP Complexes
3. Rabbit TrueBlot®: Anti-Rabbit IgG horseradish peroxidase
(HRP) conjugated secondary antibody, which enables unhin-
dered detection of molecules (eBioscience Inc.).

3  Methods

All procedures are performed in a cold room or on ice unless oth-

erwise indicated. For example, we perform the cross-linking and
preclearing steps at room temperature.

3.1  Cross-Linking 1. Grow bacterial cells (ex: A. tumefaciens) under appropriate

of Sample culture conditions.
2. Centrifuge bacterial cell culture at 6000 × g for 10 min at 4 °C,
wash the cells by resuspending the cell pellets with 12 mL
phosphate buffer, followed by centrifugation at 6000 × g for
 Co-Immunoprecipitation 215

10 min at 4 °C. Repeat this washing step twice, then the cell

pellets in the same buffer are adjusted to OD600 4 (see Note 9).
3. Add cross-linker DTBP in the cell suspension to a final concen-
tration of 5 mM (see Note 10).
4. Incubate mixture at room temperature for 45 min (see Note 11).
5. Stop cross-linking reaction by adding Tris–HCl, pH 7.6, to a
final concentration of 20 mM for 15 min (see Note 12).
6. Collect cells by centrifugation at 6000 × g for 10 min at 4 °C,
and wash the cells twice by resuspending the cell pellets with
12 mL 50 mM Tris–HCl, pH 7.6, followed by centrifugation
at 6000 × g for 10 min at 4 °C before cell lysis (see Note 13).

3.2  Preparation 1. Pellet the cross-linked cells by centrifugation at 6000 × g for

of Bacterial Cell 10 min at 4 °C and resuspend the cell pellets in 4 mL TES buf-
Extracts fer to OD600 20 (see Note 14).
2. Incubate the cell resuspension for 30 min at 37 °C with shak-
ing at 200 rpm.
3. Add 18 mL NP1 buffer supplemented with 1.5 mg/mL lyso-
zyme (see Note 15) and incubate for 2 h on ice.
4. Incubate mixture for 30 min at 37 °C with shaking at 200 rpm.
5. Add Triton X-100 to a final concentration of 4% and incubate
for 20 min at room temperature with rotation (see Note 16).
6. Add protease inhibitor cocktail to working concentration (1×)
(see Note 17) and incubate for 15 min at 37 °C with shaking
at 200 rpm.
7. Store the sample mixture for at least 3 h at 4 °C with rotation
(see Note 18).
8. Add 64 mL NP1 buffer to the mixture (see Note 19); the final
concentration of SDS and Triton X-100 are about 0.05% and
1% respectively. The insoluble material is removed by
­centrifugating twice for 15 min at 14,000 × g. The resulting
supernatant is the detergent-solubilized solution (see Note 20).
9. The choice of detergents and the appropriate concentration
used for co-IP is made according to the user’s manual of pro-
tein A-Sepharose.

3.3  Protein Sample The preclearing step will reduce the background noise caused by the
Preclearing adhesion of some protein components to the protein A-Sepharose.
and Coupling
of Antibody to Protein 1. For each 2 mL of the detergent-solubilized solution, add a 60
A/G Beads μL bed volume of protein A-Sepharose and incubate for 60 min
at room temperature with rotation (see Note 21).
2. Remove protein A-Sepharose with nonspecifically bound proteins
by centrifugation for 5 min at 5000 × g at 4 °C (see Note 22).
216 Jer-Sheng Lin and Erh-Min Lai

3. After the preclearing step, the supernatant (protein sample)

will serve as the “starting material” for co-IP. The supernatant
(about 1.5 mL) is directly incubated with antibody with opti-
mized titer (see Note 23) and protein A-Sepharose (about 60
μL) overnight at 4 °C with slow rotation (see Note 24).

3.4  Purification 1. After overnight incubation, pellet the beads by centrifugation

and Isolation at 5000 × g at 4 °C. The supernatant is designated as the “flow
of Protein Complexes through” for co-IP.
2. Wash the beads twice with 1 mL NP1 buffer supplemented
with 1% Triton X-100 and once with 1 mL NP1 buffer supple-
mented with 0.1% Triton X-100 by centrifugation at 5000 × g
at 4 °C. Discard each wash solution (supernatant) or collect
them for SDS-PAGE analysis to examine the washing effi-
ciency. Typically repeat this washing step four to eight times to
remove nonspecific binding proteins. In general, the washing
steps are carried out until no signal can be detected in the
negative control. Then proceed to step 3 or step 4 to recover
the co-IP complexes.
3. Sample-buffer elution: add 100 μL of 2× SDS sample buffer to
the microtube containing beads and place the tube at 96 °C for
20 min. After centrifugation at 10,000 × g at room tempera-
ture for 5 min, the supernatant is ready to use for western blot
analysis (see Note 25).
4. Low-pH elution: add 100 μL elution buffer to microtube con-
taining beads and place tube at room temperature with low-­
speed rotation for 20 min to elute the proteins. After
centrifugation at 10,000 × g at room temperature for 5 min, the
co-IP proteins should be eluted in supernatant (see Note 26).

3.5  TrueBlot 1. Analyze the samples by western blot.

for Protein Detection 2. When the antibody used for co-IP is generated from rabbit, we
of Co-IP Complexes recommend using the Rabbit TrueBlot® as a secondary antibody
with 5000× dilution to minimize interfering signals caused by
immunoglobulin heavy and light chains (see Note 27).

4  Notes

1. The purpose of cross-linking of the bacterial cells is to fix the

protein interactions before cell lysis and detergent treatment,
especially for weak and dynamic interactions.
2. You must use fresh, live bacterial cells (without freezing) for
cross-linking reaction.
3. Always freshly prepare DTBP before use. It is not feasible to
dissolve DTBP directly in the chemical bottle by pipetting.
 Co-Immunoprecipitation 217

Thus, it is easier to handle by weighing DTBP powder on

weight paper and transfer the powder to a microtube and then
add a metered volume of buffer to slowly dissolve the powder
by vortex.
4. Several methods can be used to prepare bacterial cell extracts
(such as sonication or French press). Here, we use a
lysozyme/detergent-solubilized method for our co-IP sample
preparation [13].
5. We recommend freshly preparing the lysozyme stock solution
in NP1 buffer at a concentration of 1 M before use.
6. Both protein A and protein G can associate to rabbit serum
with high affinity [4]. Here we chose protein A-Sepharose for
our co-IP experiment.
7. Because of the viscous property of Triton X-100, be sure to stir
and mix the NP1 buffer with Triton X-100 well before use. We
recommend preparing the 10% Triton X-100 stock solution
prior to use.
8. 2-mercaptoethanol should be added freshly to 2× sample buf-
fer prior to use.
9. The excess concentration of bacterial cells will cause poor
cross-link efficiency.
10. DTBP should be freshly prepared as 0.5 M stock with buffer,
in which phosphate buffer is used so as to be consistent with
the cell suspension buffer.
11. Gently mix the mixture once every 10 min.
12. Directly add the 1 M Tris–HCl (pH 7.6) stock solution to the
mixture until the final concentration of 20 mM is reached;
then mix well to stop the cross-linking reaction.
13. The cross-linked cells can be frozen at −80 °C until use.
However, we recommend conducting the experiments within
2 weeks after storage.
14. The concentration of cross-linked cells can be reduced.
However, if the concentration of bacterial cells is higher than
OD 20, the efficiency of cell lysis will become poor.
15. Use NP1 buffer to dissolve lysozyme first before use.
16. Use rotating wheel for this step.
17. The protease inhibitor cocktail is 100× stock.
18. Place the sample mixture on the rotating wheel in a cold room.
19. After adding NP1 buffer to the mixture, carefully mix the mix-
ture well by vortex.
20. The detergent-solubilized solution can be directly used for the
co-IP experiment.
218 Jer-Sheng Lin and Erh-Min Lai

21. Generally, the preclearing efficiency is better when incubation

is performed at room temperature than in a cold room.
22. Transfer the supernatant to a new 2 mL microtube. The treated
protein sample can be used directly for the subsequent co-IP
experiment.
23. The titer of antibody for co-IP is based on the reference of
antibody titer used for western blotting. In general, we use 5-
to 10-fold higher titer than the titer used for western
blotting.
24. Place the sample mixture on the rotating wheel in a cold room
with gentle rotation (about 10–12 rpm). This is a critical step
because an excessive rotation speed will affect the binding effi-
ciency of the antibody and protein A-Sepharose.
25. The sample-buffer elution method is ideal for western blot
analysis.
26. A low-pH elution is ideal for enzymatic or functional assays
after the low pH is neutralized. In general, the efficiency of
elution is lower using the low-pH elution method compared
with the sample-buffer elution method.
27. Rabbit IgG TrueBlot® is a unique anti-rabbit IgG immunob-
lotting reagent (used as the secondary antibody). Rabbit IgG
TrueBlot enables detection of immunoblotted target protein
bands, with reduced interfering IP immunoglobulin heavy (55
kDa) and light (23 kDa) chains. The Rabbit TrueBlot®: Anti-­
Rabbit IgG HRP can be reused at least three to five times.

Acknowledgements

This work was supported by a research grant from the Taiwan

Ministry of Science and Technology (MOST 104-2311-B-001-
025 -MY3) to E.M. Lai. J.S. Lin is the recipient of postdoctoral
fellowships from Academia Sinica.

References
1. Kessler SW (1975) Rapid isolation of anti- tools.thermofisher.com/content/sfs/bro-
gens from cells with a staphylococcal protein chures/TR0064-Immunoprecipitation-guide.
A-antibody adsorbent: parameters of the inter- pdf.
action of antibody-antigen complexes with 4. Lee C (2007) Coimmunoprecipitation assay.
protein A. J Immunol 115:1617–1624 Methods Mol Biol 362:401–406
2. Kessler SW (1976) Cell membrane anti- 5. Aebersold R, Mann M (2003) Mass
gen isolation with the staphylococcal pro- spectrometry-­ based proteomics. Nature
tein A-antibody adsorbent. J Immunol 422:198–207
117:1482–1490 6. Gevaert K, Vandekerckhove J (2000) Protein
3. ThermoFisher (2009) Immunoprecipitation identification methods in proteomics.
(IP) technical guide and protocols. https:// Electrophoresis 21:1145–1154
 Co-Immunoprecipitation 219

7. Atmakuri K, Cascales E, Christie PJ (2004) 10. Lin JS, Ma LS, Lai EM (2013) Systematic dissec-
Energetic components VirD4, VirB11 and tion of the Agrobacterium type VI secretion sys-
VirB4 mediate early DNA transfer reactions tem reveals machinery and secreted components
required for bacterial type IV secretion. Mol for subcomplex formation. PLoS One 8:e67647
Microbiol 54:1199–1211 11. Ma LS, Narberhaus F, Lai EM (2012) IcmF
8. Atmakuri K, Cascales E, Burton OT, Banta family protein TssM exhibits ATPase activity
LM, Christie PJ (2007) Agrobacterium ParA/ and energizes type VI secretion. J Biol Chem
MinD-like VirC1 spatially coordinates early 287:15610–15621
conjugative DNA transfer reactions. EMBO 12. Elion EA (2007) Detection of protein-protein
J 26:2540–2551 interactions by coprecipitation. Curr Protoc
9. Anderson LB, Hertzel AV, Das A (1996) Immunol Chapter 8:Unit 8.7
Agrobacterium tumefaciens VirB7 and VirB9 13. Cascales E, Christie PJ (2004) Definition of a
form a disulfide-linked protein complex. Proc bacterial type IV secretion pathway for a DNA
Natl Acad Sci U S A 93:8889–8894 substrate. Science 304:1170–1173
 Chapter 18

Protein–Protein Interaction: Tandem Affinity

Purification in Bacteria
Julie P.M. Viala and Emmanuelle Bouveret

Abstract
The discovery of protein–protein interaction networks can lead to the unveiling of protein complex(es)
forming cellular machinerie(s) or reveal component proteins of a specific cellular pathway. Deciphering
protein–protein interaction networks therefore contributes to a deeper understanding of how cells func-
tion. Here we describe the protocol to perform tandem affinity purification (TAP) in bacteria, which
enables the identification of the partners of a bait protein under native conditions. This method consists in
two sequential steps of affinity purification using two different tags. For that purpose, the bait protein is
translationally fused to the TAP tag, which consists of a calmodulin binding peptide (CBP) and two immu-
noglobulin G (IgG) binding domains of Staphylococcus aureus protein A (ProtA) that are separated by the
tobacco etch virus (TEV) protease cleavage site. After the first round of purification based on the binding
of ProtA to IgG coated beads, TEV protease cleavage releases CBP-tagged bait-protein along with its
partners for a second round of purification on calmodulin affinity resin and leaves behind protein contami-
nants bound to IgG. Creating the TAP-tag translational fusion at the chromosomal locus allows detection
of protein interactions occurring in physiological conditions.

Key words Protein–protein interaction, Protein complex, Affinity purification, Tandem affinity puri-
fication (TAP), Calmodulin binding peptide (CBP), ProtA, Tobacco etch virus (TEV), Escherichia
coli, Salmonella

1  Introduction

At the end of the 1990s, mass spectrometry combined with genome

sequencing rendered possible the rapid and systematic identifica-
tion of all the proteins present in a purified sample. However, a
protocol amenable to standardized and systematic purification of
protein complexes without any prior knowledge was missing. In
1999, the laboratory of B. Séraphin at the European Molecular
Biology Laboratory in Heidelberg, Germany, proposed such a
generic procedure for the identification of protein complexes in
yeast [1]. This permitted the subsequent description of the full
interactome of yeast [2, 3]. This method has since been used in a
variety of organisms. We first described its use in bacteria [4], and

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_18, © Springer Science+Business Media LLC 2017

221
222 Julie P.M. Viala and Emmanuelle Bouveret

soon after this, it was used to obtain the first interactome of

Escherichia coli [5].
One general principle of the tandem affinity purification (TAP)
method involves the use of two successive steps of affinity purifica-
tion to lower as much as possible the amount of contaminants,
together with an elution preserving the interactions (without signifi-
cantly changing the buffer’s chemical properties) between these two
steps. Specifically, the original TAP tag consists of two repeats of the
immunoglobulin G (IgG) binding domain of protein A (ProtA)
from Staphylococcus aureus and a calmodulin binding peptide (CBP),
separated by a TEVprotease cleavage site (Fig. 1). However, it must
be noted that any combination of affinity tags is potentially usable.
Published examples are the GS-TAP (protein G and Strep tag), the
sequential peptide affinity (SPA) tag (CBP and 3Flag), the SF-TAP
(Strep-tag II and Flag tag), and the HB tag (6Histidine and Biotin)
(see [6] for specific references). The second general principle of the
TAP procedure is to use physiological expression of the recombinant
tagged protein. This needs to be adapted to each organism of inter-
est. For E. coli and closely related bacteria, lambda Red-based recom-
bination [7] combined with specific dedicated SPA and TAP cassettes
[8], makes it very easy to introduce the tag at the 3′ extremity of the
gene on the chromosome to obtain the physiological production of
a recombinant protein tagged at its C-terminus (Fig. 2). If more
convenient, however, TAP tag translational fusion can also be
expressed from a plasmid (Fig. 3).
We present here the TAP protocol that has been successfully
used in our institute to purify protein complexes of E. coli,
Salmonella, and Bacillus subtilis [4, 9–11]. A detailed protocol for
the SPA purification has been published [12]. To isolate a protein
complex by TAP, a strain producing a recombinant bait protein
tagged with the TAP tag must be constructed first (Fig. 1, step 1).
Then a soluble extract is prepared from a sufficient volume of bac-
teria (about 500 mL). The complex is enriched by a first step of
affinity chromatography on IgG beads (Fig. 1, step 5). After
washes, TEVprotease is added, which cleaves the specific site
located between the CBP and ProtA domains, resulting in the elu-
tion of the specifically bound material (Fig. 1, step 6). This mate-
rial is purified a second time by affinity of the CBP tag with
calmodulin beads (Fig. 1, step 7). After washes, the purified com-
plex is eluted by adding ethylene glycol-bis(β-­aminoethyl ether)-
N,N,N',N'-tetraacetic acid (EGTA) that chelates the calcium
required for the CBP/calmodulin interaction (Fig. 1, step 8). The
totality of the purified material is analyzed on sodium dodecyl sul-
fate (SDS)-polyacrylamide gel electrophoresis (PAGE). Bands
detected by coomassie blue or silver staining are cut from the gel
and analyzed by mass spectrometry.
This is the basic TAP procedure. Note that the procedure can
be amenable to adaptation or improvements depending on the
 TAP in Bacteria 223

Fig. 1 Guidelines for overall TAP procedure

specific needs. For example, the extensive washes and the duration
of the procedure only allow for the recovery of relatively stable com-
plexes. For the detection of more transient or unstable interactions,
a cross-linking procedure can be applied before purification [13].
224 Julie P.M. Viala and Emmanuelle Bouveret

Fig. 2 Creation of a C-terminal TAP-tag translational fusion at chromosomal locus. Scheme of TAP-tag trans-
lational fusion at chromosomal locus (a) and corresponding nucleotidic and protein TAP-tag sequences (b).
CBP sequence is in clear purple; uppercase at the beginning of the nucleotidic CBP sequence corresponds to
the primer sequence mentioned in Subheading 3.1, TEVprotease cleavage site is in yellow, and ProtA sequence
is in dark purple, with stop codon in red

This might be helpful also for the purification of membrane com-

plexes, where modifications must be made in the protocol for the
solubilization of membranes [3]. Finally, it is possible to play with
the two tags to obtain information on the organization of the
Fig. 3 Creation of plasmidic N-terminal TAP-tag translational fusion. Map of plasmid pEB587 used to create an N-terminal
TAP-tag translational fusion under the control of the PBAD arabinose-inducible promoter (a) and the correspond-
ing nucleotidic and protein TAP-tag sequences (b). ProtA sequence is in dark purple, TEVprotease cleavage site is in
yellow, CBP sequence is in clear purple, the multicloning site is underlined, and restriction enzyme sites are indicated
226 Julie P.M. Viala and Emmanuelle Bouveret

complexes. Indeed, in some cases, one bait protein might partici-

pate in the formation of several types of complexes. To purify one
specific type of complex, it is therefore possible to put the two tags
on two distinct proteins, both of which are members of the desired
type of complex (split tag method [9, 14]). Alternatively, it is pos-
sible to perform the subtraction method that consists in eliminating
the unwanted complex(es) during the first purification step by leav-
ing it, for example, on IgG beads thanks to a partner protein of the
bait that belongs to the unwanted complex and bears a noncleavable
ProtA tag. The desired complex, made of untagged partner pro-
teins, will elute with the bait after TEVprotease cleavage [14, 15].
To our knowledge, the TAP procedure has not been used very
much for the characterization of secretion systems in bacteria, cer-
tainly owing to the difficulty of working with integral envelope
components [11]. However, it has proved to be powerful in iden-
tifying the target of effectors of Legionella T4SS or Pseudomonas
T6SS in eukaryotic host cells [16, 17]. In addition, as mentioned
earlier, it is amenable to several improvements that might make it
possible to identify unsuspected partners of the secretion machin-
eries in the bacterium.

2  Materials

2.1  Creation 1. Bacterial strain with translational fusion between protein of

of a TAP-Tag interest and TAP tag at chromosomal locus (see Note 1).
Translational Fusion 2. Alternatively, translational fusion between protein of interest
and Verification and TAP tag on an appropriate plasmid (see Note 2).
of Production 3. Yeast extract and tryptone media (2YT): 16 g yeast extract, 10
of Hybrid Protein g tryptone, 10 g NaCl, make up to 1 L with distilled water.
by Western Blot Autoclave and store at room temperature.
4. Lysogeny broth (LB): 5 g yeast extract, 10 g Bacto Tryptone,
10 g NaCl, make up to 1 L with distilled water. Autoclave and
store at room temperature.
5. Minigel caster system and SDS-PAGE apparatus.
6. Transblot apparatus for western blot.
7. Peroxidase–antiperoxidase antibody (PAP) (Sigma).

2.2  Protein 1. Phosphate buffered saline (PBS): 8 g NaCl, 0.2 g KCl, 0.2 g
Cytoplasmic Extract KH2PO4, 2.9 g Na2HPO4, make up to 1 L with distilled water.
Autoclave and store at room temperature.
2. 10% Nonidet P-40 (NP-40 or Igepal): Mix 10 mL NP-40 in
90 mL distilled water, pass through a 0.2 μm filter, and store at
room temperature (see Note 3).
3. ProtA binding buffer: 10 mM Tris–HCl, pH 8, 150 mM NaCl,
0.1% NP-40. Approximately 50 mL will be required per sample
 TAP in Bacteria 227

per experiment. Prepare 500 mL containing 5 mL 1 M Tris–

HCl, pH 8, 15 mL 5 M NaCl, 5 mL 10% NP-40, and 475 mL
distilled water. Store at 4 °C.
4.
0.1 M phenylmethylsulfonyl fluoride (PMSF): Dissolve
87.1 mg PMSF in 5 mL isopropanol. Prepare 1 mL aliquots
and store at −20 °C (see Note 4).
5. Liquid nitrogen.
6. Sonicator, French press, or cell disruptor.
7. Centrifuge tubes and rotor, compatible with spinning volumes
of 250 mL, 10 mL, and 50 mL, at approximately 5000 × g and
25,000 × g, respectively.

2.3  Tandem Affinity 1. IgG Sepharose 6 fast flow (GE Healthcare).

Purification 2. ProtA binding buffer: See Subheading 2.2.
3. 0.5 M EDTA (C10H14N2Na2O8·2H2O): Dissolve 18.6 g EDTA
in 80 mL distilled water, make up to 100 mL with distilled
water once pH has been adjusted to 8 with 10 N NaOH (see
Note 5). Autoclave and store at room temperature.
4. TEV cleavage buffer: 10 mM Tris–HCl, pH 8, 150 mM NaCl,
0.1% NP-40, 0.5 mM EDTA, 1 mM dithiothreitol (DTT) (see
Note 6).
5. AcTEV™ protease (Invitrogen).
6. Calmodulin binding buffer: 10 mM Tris–HCl, pH 8, 150 mM
NaCl, 0.1% NP-40, 1 mM magnesium acetate, 1 mM imidaz-
ole, 2 mM CaCl2, 10 mM β-mercaptoethanol. Approximately
40 mL will be required per sample per experiment. Prepare
500 mL containing 5 mL 1 M Tris–HCl, pH 8, 15 mL 5 M
NaCl, 5 mL 10% NP-40, 500 μL 1 M magnesium acetate,
500 μL 1 M imidazole, 1 mL 1 M CaCl2, 348.5 μL 14.3 M
β-mercaptoethanol (see Note 7), and 473 mL distilled water.
Store at 4 °C.
7. 1 M CaCl2: Dissolve 11.1 g in 100 mL distilled water. Autoclave
and store at room temperature.
8. Calmodulin affinity resin (Agilent).
9. 1 M EGTA: Dissolve 19 g in 40 mL distilled water, make up to
50 mL with distilled water once pH has been adjusted to 8
with 10 N NaOH (see Note 5), 0.2 μm filter, and store at 4 °C.
10. Calmodulin elution buffer: 10 mM Tris–HCl, pH 8, 150 mM
NaCl, 0.1% NP-40, 1 mM magnesium acetate, 1 mM imidaz-
ole, 2 mM EGTA, 10 mM β-mercaptoethanol. Approximately
1 mL will be required per sample per experiment. Prepare
100 mL containing 1 mL 1 M Tris–HCl, pH 8, 3 mL 5 M
NaCl, 1 mL 10% NP-40, 100 μL 1 M magnesium acetate,
100 μL 1 M imidazole, 200 μL 1 M EGTA, 69.7 μL 14.3 M
228 Julie P.M. Viala and Emmanuelle Bouveret

β-mercaptoethanol (see Note 7), and 94.5 mL distilled water.

Store at 4 °C.
11. Disposable chromatography columns of 10 mL with narrow
bottom, for example, Poly-Prep chromatography column from
Biorad.
12. Rotating wheel.
13. Centrifuge tubes and rotor compatible with spinning volumes
of 10 mL at approximately 25,000 × g.

2.4  Trichloroacetic 1. 16 mg/mL sodium deoxycholate: Dissolve 160 mg sodium

Acid Precipitation deoxycholate in 10 mL water. Pass through a 0.2 μm filter and
store at room temperature.
2. Liquid trichloroacetic acid (TCA) (stock is 100%).
3. TCA washing buffer: Mix 70 mL acetone, 20 mL ethanol,
5 mL 1 M Tris–HCl, pH 8, and 5 mL distilled water. Store at
4 °C.
4. SDS-PAGE loading buffer.

3  Methods

A translational fusion between the protein of interest and TAP tag,

either at the chromosomal locus or on an appropriate plasmid,
must be constructed (see Notes 1 and 2). Translational fusion at
the chromosomal locus will allow a physiological expression, while
constructing the translational fusion on a plasmid may be more
amenable.

3.1  Verification 1. Prepare a cytoplasmic or a crude protein extract (see Note 8).
of Expression of TAP-­ 2. Load 10 μg of protein extract (or proteins corresponding to a
Tag Translational bacterial sample of 0.3 OD600 unit) on a SDS-PAGE and pro-
Fusion by Western ceed to transfer and western blot to verify production of hybrid
Blot protein (see Note 9).
3. Perform a one-step western blot using PAP antibody (see
Note 10) and using an appropriate substrate to detect horse-
radish peroxidase activity (see Note 11).

3.2  Protein 1. Day 1—Inoculate 10 mL 2YT media with a bacterial colony

Cytoplasmic Extract and grow overnight at 37 °C with shaking (see Note 8).
2. Day 2—Dilute culture 100-fold in 500 mL LB and grow for
5 h 30 min at 37 °C with shaking until OD600 ≈ 2–3.
3. Pellet bacteria by centrifugation for 20 min, 5000 × g at 4 °C.
4. Wash once with cold PBS, transfer to 50 mL centrifuge tubes,
centrifuge again for 10 min, 5000 × g at 4 °C, discard superna-
tant, and freeze bacterial pellets with liquid nitrogen.
 TAP in Bacteria 229

Maintain frozen bacterial pellets at −80 °C until you are

ready to prepare cytosolic protein extract and proceed to TAP.
5. Day 3—Resuspend frozen bacterial pellets with 10 mL ProtA
binding buffer containing 0.5 mM PMSF (see Note 4).
6. Use sonication, French press, or cell disruptor to break bacte-
rial cells (see Note 12).
7. Centrifuge for 30 min, 25,000 × g at 4 °C, and save superna-
tant, which is the cytoplasmic protein extract.

3.3  Tandem Affinity From here, carry out all procedures with gloves to avoid contami-
Purification nation of your sample(s) with keratine.
1. Put 200 μL IgG Sepharose beads in a disposable chromatog-
raphy column and wash by gravity with 5 mL ProtA binding
buffer.
2. Binding of ProtA tag to IgG Sepharose beads. After washing the
beads, close the bottom of the chromatography column, and,
using a pipette, transfer 9 mL of the cytoplasmic protein extract.
Close the top of the column and put on a wheel for 2 h at 4 °C.
3. Remove first the top plug of the column and then the bottom
one. Leave the unbound material flow by gravity and discard.
4. Wash three times the IgG beads with 10 mL of ProtA binding
buffer.
5. TEVprotease cleavage. Close bottom of column, fill it with 1 mL
TEV cleavage buffer and 100 units of AcTEV™ protease. Close
top of column and put on wheel at room temperature for 1 h.
6. Remove top and bottom plugs and recover elution fraction by
gravity. Add additional 200 μL TEV cleavage buffer in column
to recover as much material as possible from sides of column.
7. Add 3 mL calmodulin binding buffer and 3 μL 1 M CaCl2
(see Note 13) to elution fraction.
8. Binding by the CBP tag part on calmodulin affinity resin. Put
200 μL calmodulin affinity resin in a new disposable chroma-
tography column, and wash it with 5 mL calmodulin binding
buffer. Then close bottom of column.
9. Add the 4.2 mL of the elution fraction (obtained at steps 6
and 7). Close top of column and put on wheel for 1 h at 4 °C.
10. Remove first the top plug of the column and then the bottom
one. Leave unbound material to flow by gravity and discard.
11. Wash three times calmodulin affinity resin with 10 mL calmod-
ulin binding buffer.
12. Elution. Elute with five times 200 μL of calmodulin elution
buffer.
13. Pool fractions 2, 3, and 4 and proceed to TCA precipitation of
elution fraction 1, pooled fractions 2, 3, and 4, and fraction 5.
230 Julie P.M. Viala and Emmanuelle Bouveret

3.4  TCA Precipitation 1. To each of the eluted protein samples add 1/100th of 16 mg/
mL sodium deoxycholate. Vortex and leave on ice 30 min.
2. Add TCA to 10% final. Vortex and leave on ice 30 min.
3. Centrifuge 15 min, 15,000 × g at 4 °C.
4. Wash twice with TCA washing buffer.
5. Leave pellets to dry on bench and resuspend in 20 μL protein
loading buffer 1×.

3.5  Analysis 1. Load totality of samples on SDS-PAGE (see Note 14) and
by SDS-PAGE stain with coomassie blue.
and Mass 2. Unstain and then rinse with distilled water.
Spectrometry
3. Cut bands to identify partner proteins by mass spectrometry
(see Note 15).

4  Notes

1. A C-terminal TAP tag translational fusion can be introduced

at the chromosomal locus using the λ Red recombination sys-
tem [7]. To prepare the appropriate polymerase chain reaction
(PCR) product, use a pJL72 plasmid as template (this latter
harbors a cassette made of the TAP tag and the kanamycin
resistance gene, Fig. 2a) [8], design a forward primer that
contains, in the 5′-end, the 45 nucleotides that are immedi-
ately upstream of the stop codon of the gene of interest, fol-
lowed by the sequence 5′-TCCATGGAAAAGAGAAG-3′
(this sequence will form a hybrid to the CBP tag, Fig. 2b), and
design a reverse primer that contains at its 5′-end the reverse
complement of 45 nucleotides that are immediately down-
stream of the stop codon of the gene of interest, followed by
the sequence 5′-CATATGAATATCCTCCTTAG-3′ (Fig. 2a).
2. Alternatively, the sequence corresponding to the open reading
frame of the gene of interest can be cloned in the plasmid
pEB587 [18] (Fig. 3a), which allows an N-terminal TAP tag
translational fusion (Fig. 3b) under the control of the arabinose-­
inducible promoter PBAD.
3. Gently agitate the solution for complete dissolution of NP-40
if necessary.
4. PMSF crystallizes at −20 °C, so heat the PMSF aliquot to
37 °C to redissolve the PMSF before use. We use PMSF as
generic protease inhibitor, but a protease inhibitor cocktail can
be used as well.
5. EDTA and EGTA may not be soluble until the pH is adjusted
to 8 with 10 N NaOH.
 TAP in Bacteria 231

6. Add DTT to the volume of buffer you will need when starting
the experiment. DTT is necessary for TEV activity.
7. Add β-mercaptoethanol to the volume of buffer you will need
when starting the experiment.
8. Also plan to prepare a protein extract of an untagged strain as
the negative control of the experiment.
9. Translational TAP tag fusion adds 20 kDa to the mass of the
protein of interest; 3 kDa corresponds to the CBP tag and
15 kDa to the ProtA tag.
10. Immunoglobulins will bind the ProtA fragment of the TAP tag.
11. In our experience, the detection of a tagged protein in crude
extracts using this PAP antibody is mandatory for a successful
TAP purification.
12. A French press or cell disruptor might be more gentle for pre-
serving protein complexes.
13. The addition of extra CaCl2 is required to quench the EDTA
that was previously necessary for TEV protease activity.
14. Usually 12% SDS-PAGE allows visualization of low- and high-­
molecular-­weight proteins.
15. Use one blade per band.

References

1. Rigaut G, Shevchenko A, Rutz B, Wilm M, 7. Datsenko KA, Wanner BL (2000) One-step

Mann M et al (1999) A generic protein purifi- inactivation of chromosomal genes in
cation method for protein complex character- Escherichia coli K-12 using PCR products.
ization and proteome exploration. Nat Proc Natl Acad Sci U S A 97:6640–6645
Biotechnol 17:1030–1032 8. Zeghouf M, Li J, Butland G, Borkowska A,
2. Gavin AC, Bosche M, Krause R, Grandi P, Canadien V et al (2004) Sequential Peptide
Marzioch M et al (2002) Functional organiza- Affinity (SPA) system for the identification of
tion of the yeast proteome by systematic analy- mammalian and bacterial protein complexes.
sis of protein complexes. Nature 415:141–147 J Proteome Res 3:463–468
3. Gavin AC, Aloy P, Grandi P, Krause R, Boesche M 9. Gully D, Bouveret E (2006) A protein network
et al (2006) Proteome survey reveals modularity for phospholipid synthesis uncovered by a vari-
of the yeast cell machinery. Nature 440:631–636 ant of the tandem affinity purification method
4. Gully D, Moinier D, Loiseau L, Bouveret E in Escherichia coli. Proteomics 6:282–293
(2003) New partners of acyl carrier protein
10. Pompeo F, Luciano J, Galinier A (2007)
detected in Escherichia coli by tandem affinity Interaction of GapA with HPr and its homo-
purification. FEBS Lett 548:90–96 logue, Crh: novel levels of regulation of a key
5. Butland G, Peregrin-Alvarez JM, Li J, Yang W, step of glycolysis in Bacillus subtilis? J Bacteriol
Yang X et al (2005) Interaction network con- 189:1154–1157
taining conserved and essential protein com-
11. Viala JP, Prima V, Puppo R, Agrebi R,
plexes in Escherichia coli. Nature 433:531–537 Canestrari MJ, Lignon S et al (2017) Acylation
6. Collins MO, Choudhary JS (2008) Mapping of the type 3 secretion system translocon using
multiprotein complexes by affinity purification a dedicated acyl carrier protein. PLoS Genet
and mass spectrometry. Curr Opin Biotechnol 13(1):e1006556
19:324–330
232 Julie P.M. Viala and Emmanuelle Bouveret

12. Babu M, Butl G, Pogoutse O, Li J, Greenblatt JF plex that participates in mRNA degradation.
et al (2009) Sequential peptide affinity purifica- EMBO J 19:1661–1671
tion system for the systematic isolation and iden- 16. So EC, Schroeder GN, Carson D, Mattheis C,
tification of protein complexes from Escherichia Mousnier A et al (2016) The Rab-binding pro-
coli. Methods Mol Biol 564:373–400 files of bacterial virulence factors during infec-
13. Stingl K, Schauer K, Ecobichon C, Labigne A, tion. J Biol Chem 291:5832–5843
Lenormand P et al (2008) In vivo interactome 17. Sana TG, Baumann C, Merdes A, Soscia C, Rattei
of Helicobacter pylori urease revealed by tan- T et al (2015) Internalization of Pseudomonas
dem affinity purification. Mol Cell Proteomics aeruginosa strain PAO1 into epithelial cells is
7:2429–2441 promoted by interaction of a T6SS effector with
14. Puig O, Caspary F, Rigaut G, Rutz B, the microtubule network. MBio 6:e00712
Bouveret E et al (2001) The tandem affinity 18. Battesti A, Bouveret E (2008) Improvement of
purification (TAP) method: a general proce- bacterial two-hybrid vectors for detection of
dure of protein complex purification. Methods fusion proteins and transfer to pBAD-tandem
24:218–229 affinity purification, calmodulin binding pep-
15. Bouveret E, Rigaut G, Shevchenko A, Wilm tide, or 6-histidine tag vectors. Proteomics
M, Seraphin B (2000) A Sm-like protein com- 8:4768–4771
 Chapter 19

Site-Directed and Time-Resolved Photocrosslinking

in Cells Metabolically Labeled with Radioisotopes
Raffaele Ieva

Abstract
To efficiently transport proteins into and across cellular membranes, specialized transport machineries
engage in recognition events with different domains of their client proteins, forming sequential intermedi-
ate complexes. The short-lived nature of these interactions poses a big challenge in the identification of the
key factors involved in transport reactions and their mechanism of action. Site-directed photocrosslinking
is a powerful method for the detection and accurate mapping of interacting protein domains. This chapter
describes a protocol that combines site-directed photocrosslinking to metabolic labeling of proteins and
lipids as a method to characterize, with temporal and spatial resolution, the interactions of a secretory
protein as it transverses the bacterial envelope.

Key words Site-directed photocrosslinking, Transient protein–protein interaction, Protein–lipid

interaction, Protein secretion, Escherichia coli, Autotransporter, BAM complex

1  Introduction

Efficient and correct sorting of proteins, from the site of synthesis

to the compartment where they function, is vital for cells. This task
is particularly challenging for proteins that must cross one or more
cellular membranes before reaching their destination, such as pro-
teins secreted across the bacterial envelope. In bacteria, protein
secretion is aided by specialized molecular machineries, which
insert or transport their client proteins into and across lipid bilayers
[1]. Capturing interactions between secretory proteins and their
transport machineries “at work” is a key approach to elucidate the
molecular events that govern these sophisticated reactions.
Several biochemical methods have been developed to describe
the interactions of proteins within multisubunit complexes in their
cellular native environment, including native complex isolation by
affinity purification or immunoprecipitation, blue native polyacryl-
amide gel electrophoresis of assembly intermediates, and chemical
crosslinking. Recognition events between secretory proteins and

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_19, © Springer Science+Business Media LLC 2017

233
234 Raffaele Ieva

their transport machineries, however, are often too short-lived to

be detected by these methods. Different strategies can be designed
to “trap” client proteins in a complex with their transport machin-
eries. This can be achieved, for instance, by stabilizing structural
motifs within the client protein or by fusing it to a bulkier protein
moiety, generating an obstruction that prevents the completion of
the transport reaction. Although such “trapping” strategy may
help to capture a transient intermediate complex that forms at a
specific stage of the transport reaction, it may not provide informa-
tion on the sequence of molecular events that precede and follow
the formation of the captured intermediate.
A further limitation of methods based on native complex isola-
tion or chemical crosslinking concerns the mapping of protein–
protein interactions. In order to map interaction sites, a method
for the incorporation of unnatural photoactivatable amino acids
into newly synthesized proteins was originally developed using an
amber suppression approach in cellfree translation systems [2]. A
subsequent modification of the method has made it possible to
express proteins with site-specific photoprobes in vivo. To this end,
cells that coexpress an engineered orthogonal aminoacyl tRNA
synthetase, which loads a cognate amber suppressor tRNA with a
photoreactive amino acid analog, such as p-benzoyl-l-­phenylalanine
(Bpa), are used [3]. The photoprobe can be introduced at an
amber codon engineered at a specific position of the open reading
frame of a gene of interest. Upon irradiation with ultraviolet (UV)
light (350–365 nm), Bpa forms a highly reactive radical that can
crosslink to C–H bonds in the vicinity [4]. Note that Bpa is a small
amino acid of approximately 4 Å in diameter, thus it can crosslink
only proteins that are in close proximity, allowing precise mapping
of protein interaction sites.
The method described in this chapter exploits the in vivo site-­
directed photocrosslinking approach to resolve consecutive interac-
tions of a client protein with its transport machinery. Site-directed
photocrosslinking is combined with transient metabolic labeling of
cells with radioactive amino acids. Labeling is performed using an
in vivo pulse-chase methodology, thus the fate of radiolabeled pro-
teins can be followed in time, helping to distinguish the order of the
detected interactions. The example protocol described here analyzes
the biogenesis of EspP, an autotransporter expressed by Escherichia
coli O157:H7. Autotransporters are a class of virulence factors pro-
duced by Gram-negative bacteria. After transport across the inner
membrane, the autotransporter carboxy-terminal “β domain” inte-
grates into the outer membrane by folding into a β-barrel structure,
while its amino-terminal “passenger domain” is ultimately translo-
cated into the extracellular space. The mechanism by which the
passenger domain is transported across the outer membrane has
been debated for some time [5, 6]. In a series of studies, the
described approach revealed that passenger domain secretion and
 In vivo Site-Directed Photocrosslinking 235

outer membrane integration of the β domain are both mediated by

the β-barrel assembly machinery (BAM complex) in a coordinated
reaction. BAM is a multisubunit complex consisting of BamA, an
integral outer membrane protein, and four lipoproteins, BamBCDE,
that associate to the inner leaflet of the outer membrane lipid bilayer
[7, 8]. Consecutive interactions of distinct EspP domains, first with
periplasmic chaperones and then with subunits of the BAM machin-
ery, were identified, leading to a detailed model of autotransporter
biogenesis [9–11]. Finally, this chapter provides an example of how
site-directed photocrosslinking can be combined with the metabolic
labeling of cells using radioactive inorganic phosphate [12] to reveal
the time of insertion of the EspP β domain into the lipid bilayer of
the bacterial outer membrane [10, 11].

2  Materials

2.1  Plasmid 1. Site-directed mutagenesis kit or a similar set of reagents for

Construction and high-fidelity polymerase chain reaction (PCR).
Transformation of  2. Chemically ultracompetent E. coli cells.
E. coli Cells
3. Plasmid miniprep kit.
4. Cuvettes for electroporation.
5. Electroporator.
6. Lysogenic broth (LB) agar plates containing selected antibiotics.

2.2  Expression of a 1. 10× M9 salt (67.8 g/L Na2HPO4, 30 g/L KH2PO4, 5 g/L
EspP Variant NaCl, 10 g/L NH4Cl).
Containing Bpa and  2. M9 complete medium containing 1× M9 salt, 1 mM MgSO4,
Pulse-Chase Metabolic 0.1 mM CaCl2, 0.2% w/v glycerol, 40 μg/mL l-amino acids
Labeling of Cells Using (except methionine and cysteine).
35
S-Labeled Amino 3. One disposable 125 mL Erlenmeyer flask.
Acids
4.
A high-specific activity mixture of 35
S-cysteine and
35
S-­methionine (1075–1175 Ci/mmol).
5. A stock solution of non-radiolabeled methionine and cysteine
(100 mM methionine, 100 mM cysteine).
6. 100 mM isopropyl-β-d-thio-galactoside (IPTG) stock solution.
7. Bpa (Bachem).
8. Water bath shaker.
2.3  Metabolic
1. One disposable 125 mL Erlenmeyer flask.
Labeling
with 32P-Labeled 2. Modified G56 medium (45 mM MES, pH 7.0, 10 mM KCl,
10 mM MgCl2, 15 mM (NH4)2SO4, 5 μg/L thiamine, 0.2%
Inorganic Phosphate
glycerol, 40 μg/mL l-amino acids).
and Expression
of an EspP Variant 3. A stock solution of 0.5 M KH2PO4.
Containing Bpa 4. Radioactive KH232PO4 (900–1100 mCi/mmol).
236 Raffaele Ieva

5. 100 mM IPTG stock solution.

6. Bpa (Bachem).

2.4  Photocross 1. High-intensity UV lamp such as Spectroline SB-100P

linking (Spectronics Corporation).
2. Six-well tissue culture plates.
3. Disposable pipettes.

2.5  Immunoprecipita 1. Antisera specific for EspP, BamA, and BamB or other proteins
tion and Sodium of interest.
Dodecyl Sulfate 2. Protein solubilization buffer (15% glycerol, 200 mM Tris base,
(SDS)-Polyacrylamide 15 mM EDTA, 4% SDS, 2 mM phenylmethylsulfonyl fluoride
Gel Electrophoresis (PMSF)).
(PAGE) Analysis of 3. Radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris–
Photocrosslinked HCl, pH 8, 150 mM NaCl, 1% Igepal CA-630, 0.5% sodium
Proteins deoxycholate, 0.1% SDS).
4. Staphylococcus aureus protein A-Sepharose beads.
5. SDS-polyacrylamide gels.
6. 4× SDS-PAGE sample buffer (8% SDS, 40% glycerol, 240 mM
Tris–HCl, pH 6.8, 1.6% β-mercaptoethanol, 0.04% bromo-
phenol blue).
7. Gel drying system.
8. Storage phosphor screen.

3  Method

3.1  Strategy Design 1. Clone a construct encoding the selected protein of interest
and Plasmid under the control of an inducible promoter in an expression
Construction plasmid that can be propagated in the desired bacterial model
organism. In this example, RI23, an RB11-derived plasmid,
harbors a construct encoding EspP(586TEV) (see Note 1)
under the control of the IPTG-inducible lac promoter. The
selected model organism is the laboratory E. coli strain AD202,
an MC4100-derived strain that lacks the gene encoding the
outer membrane protease OmpT [13].
2. Use a PCR-based site-specific mutagenesis approach to incor-
porate an amber codon at a specific position of the protein of
interest. In this example, the mutagenesis approach is con-
ducted on pRI23 to replace with an amber codon the EspP
Trp 1149 codon, generating pRI23-1149Bpa (see Note 2).
3. Digest the parental DNA used as template in the PCR with the
restriction enzyme DpnI.
4. Transform ultracompetent E. coli cells using a small aliquot of
the amplified DNA product.
 In vivo Site-Directed Photocrosslinking 237

5. Select single colonies growing on LB containing 100 μg/mL

of ampicillin, extract plasmidic DNA, and verify the correct
incorporation of the amber codon by plasmid sequencing.
6. Cotransform by electroporation the E. coli strain AD202 with
pDULE-pBpa, a plasmid encoding an engineered amino acyl
tRNA synthetase and a cognate amber suppressor tRNA of
Methanococcus jannaschii (see Note 3), and the generated
­plasmid (pRI23-1149Bpa). Plate cells on LB-agar containing
100 μg/mL ampicillin and 5 μg/mL tetracycline.
7. Select a single colony of transformed cells.

3.2  Preparation of Two methodologies of metabolic labeling are used (Fig. 1). In one
Cell Cultures for EspP culture 35S labeling is conducted following a “pulse-chase” proce-
Expression and dure, in which all newly synthesized proteins become radiolabeled by
Metabolic Labeling exposing cells for a short time period to 35S-methionine and
35
S-cysteine (pulse phase). Subsequently, the addition of an excess of
non-radiolabeled (“cold”) methionine and cysteine prevents further
incorporation of radiolabeled amino acids into proteins, thereby end-
ing the pulse phase. The fate of radiolabeled proteins can be chased
over time (chase phase). By combining 35S-labeling with site-specific

Fig. 1 Flow diagram of the illustrated experimental procedure. Two cultures are conducted in parallel. One
culture is subjected to pulse-chase labeling using 35S-methionine and 35S-cysteine in order to label all newly
synthesized proteins (left). The other culture is subjected to labeling with 32P-inorganic phosphate in order to
label cellular phospholipids (right)
238 Raffaele Ieva

photocrosslinking and protein immunoprecipitation, the interactions

of EspP at sequential steps of its biogenesis can be monitored over
time. In a parallel culture, 32P-labeled inorganic phosphate is incor-
porated into phospholipids and lipopolysaccharide (LPS) and allows
for the identification of interactions between the EspP β domain and
lipids of the outer membrane.
1. Start two 5 mL overnight cultures of AD202 cells transformed
with pRI23-1149Bpa and pDULE-pBpa, one in M9 medium
and one in G56 medium containing 0.13 mM KH2PO4.
Supplement both cultures with 100 μg/mL ampicillin and
5 μg/mL tetracycline.
2. The following day, centrifuge cells from both cultures and
wash them once with fresh medium (see Note 4). Finally resus-
pend cells in 2 mL fresh M9 and 2 mL fresh G56 media,
respectively, and measure the optical density of the cell suspen-
sion at a wavelength of 550 nm (OD550) in a
spectrophotometer.
3. Use M9 resupended cells to inoculate a 50 mL culture in M9
medium.
4. Use G56 resuspended cells to inoculate a 12 mL culture in
G56 medium w/o phosphate in a disposable flask. The start-
ing OD550 of both cultures is set to 0.03 (Fig. 1).

3.3  Expression When the 50 mL culture in M9 medium reaches an OD550 of 0.2,

of Photoprobed EspP induce EspP expression by adding 200 μM IPTG (see Note 5).
and Preparation Immediately after, add 1 mM Bpa (see Note 6). Incubate the cul-
of Cells for 35S-Pulse-­ ture for an additional 30 min at 37 °C (Fig. 1).
Chase Labeling 1. During this incubation time, label six 15 mL tubes as follows
(this step is preparatory for the next experimental phase of
metabolic labeling):
–– S Time 1 min −UV;
35

–– S Time 7 min −UV;

35

–– S Time 15 min −UV;

35

–– S Time 1 min +UV;

35

–– S Time 7 min +UV;

35

–– S Time 15 min +UV.

35

2. Label three wells of a six-well plate as follows:

–– S Time 1 min +UV;
35

–– S Time 7 min +UV;

35

–– S Time 15 min +UV.

35

3. Fill the tubes with ice chips, approximately to the 5 mL volume

line. Place the ice collected in the +UV tube into the correspond-
ing labeled wells of the tissue culture plate. Keep the −UV tubes
in one ice bucket and the cell culture plate in another ice bucket.
 In vivo Site-Directed Photocrosslinking 239

4. After 30 min from IPTG addition, perform pulse-chase label-

ing with 35S-methionine and 35S-cysteine, as described in
Subheading 3.5.

3.4  Cell Labeling 1. Immediately after inoculation, supplement the 12 mL culture

with 32P-Inorganic in G56 medium with 133 μCi/mL KH232PO4 (see Note 7).
Phosphate, Expression 2. When the culture reaches an OD550 of 0.2, induce expression
of Photoprobed EspP, of EspP by adding 200 μM IPTG (see Note 5). Immediately
and Photocrosslinking after, add 1 mM Bpa (see Note 6). Incubate the culture for an
additional 30 min at 37 °C (Fig. 1).
3. During this incubation time, label two 15 mL tubes as follows:
–– P −UV;
32

–– P +UV.
32

4. Fill both tubes with ice chips, approximately up to the 5 mL

volume mark line. Place the ice collected in the “32P +UV”
tube into a well of a six-well tissue culture plate. Label the well
accordingly. Keep the “32P −UV” tube in an ice bucket and the
cell culture plate in another ice bucket.
5. After 45 min from IPTG induction place 5 mL of the culture
in the “32P −UV”-labeled tube and 5 mL in the “32P +UV”-
labeled plate well containing ice chips.
6. Immediately move the multiwell plate under the lamp to expose
the “32P +UV” sample to UV light for 5 min (see Notes 8
and 9). After this time, place the sample into the corresponding
15 mL tube previously labeled “32P +UV” (see Note 10).
7. Collect cells in both 15 mL tubes by centrifugation at 2000 ×
g for 5 min at 4 °C.
8. Resuspend the cells pelleted in each tube in 1 mL M9 salt, place
each sample in 1.5 mL Eppendorf tubes. Precipitate the whole
protein content in both samples by adding trichloracetic acid.
9. Collect precipitated proteins by centrifugation at 16,000 × g
for 15 min.
10. Wash samples with ice-cold acetone. Dry protein pellets.

3.5  35S-Pulse-Chase 1. Move 35 mL of the M9 culture into a disposable 125 mL flask.

Labeling of Cells 2. Place the flask in the water bath shaker.
and Photocrosslinking
3. Start a timer counting up (see Note 11).
4. Time 30″: add 11 μCi/mL of 35S-Met/35S-Cys protein label-
ing mix. Close the flask and swirl rapidly by hand. Place
the flask in the water bath to prevent excessive cooling of the
culturing medium.
5. Time 1′00'': Add 350 μL of cold methionine and cysteine 
stock solution. Close flask, swirl rapidly by hand. Place the flask
in the water bath.
240 Raffaele Ieva

6. Time 2′00″: Remove 10 mL using a disposable pipette. Place

5 mL into the well labeled “35S Time 1 min +UV” containing ice
chips and 5 mL into the 15 mL “35S Time 1 min −UV” tube
containing ice chips. Close the flask and place it in the water bath.
7. Immediately move the multiwell plate under the lamp to expose 
sample “35S Time 1 min +UV” to UV light for 5 min (see Notes 8
and 9). After this time, place the sample in the corresponding 15
mL tube labeled “35S Time 1 min +UV” (see Note 10).
8. Time 8′00″: Repeat steps 5 and 6 using the well labeled “35S
Time 7 min +UV” and tubes labeled “35S Time 7 min −UV”
and “35S Time 7 min +UV.”
9. Time 16′00″: Repeat steps 5 and 6 using the well labeled “35S
Time 15 min +UV” and tubes labeled “35S Time 15 min −
UV” and “35S Time 15 min +UV.”
10. Place all 15 mL tubes in a centrifuge and collect cells at 2000
× g for 5 min at 4 °C.
11. Resuspend the cells pelleted in each tube using 1 mL M9 salt,
place each sample in 1.5 mL Eppendorf tubes. Precipitate
the whole protein content in each sample by adding trichlor-
acetic acid.
12. Collect precipitated proteins by centrifugation at 16,000 × g
for 15 min at 4 °C.
13. Wash samples with ice-cold acetone. Dry protein pellets.

3.6  Immuno-­ 1. Solubilize precipitated proteins by adding 50 μL of protein

precipitation of EspP solubilization buffer (15% glycerol, 200 mM Tris base, 15 mM
and Analysis EDTA, 4% SDS, 2 mM PMSF) to each sample.
of Photocrosslinking 2. Place tubes in a thermoblock at 95 °C with agitation set to
Products 1000 rpm.
3. Add 1 mL RIPA buffer. Subject samples to a clarifying spin at
16,000 × g for 10 min at 95 °C to pellet nonsolubilized
proteins.
4. Move three 300 μL aliquots of supernatant from 35S-labeled
samples into new Eppendorf tubes and supplement them with
1 μL of anti-EspP, anti-BamA, or anti-BamB antisera, respec-
tively. Move a single 300 μL aliquot from each 32P-labeled
sample into a new tube and supplement it with 1 μL anti-EspP
antiserum.
5. Mix and incubate samples on ice for 2 h.
6. Add S. aureus protein A-Sepharose beads to sediment
antibodies.
7. Wash beads with high-salt RIPA buffer (containing 500 mM
NaCl) at least twice (see Note 12).
 In vivo Site-Directed Photocrosslinking 241

Fig. 2 Transient and stable interactions detected at a specific position of the EspP β domain. AD202 cells
transformed with plasmids RI23-1149Bpa and pDULE-pBpa are subjected either to pulse-chase labeling with
radioactive amino acids (lanes 1–18) or to labeling with radioactive phosphate (lanes 19, 20). Each sample is
divided into two equal aliquots, one of which is subjected to UV irradiation (lanes 10–18 and 20). All samples
are subjected to immunoprecipitation using the indicated antisera. The crystal structure of EspP (PDB: 2QOM)
is shown to highlight the position of W1149

8. Elute proteins using SDS-PAGE loading buffer.

9. Analyze eluted proteins by SDS-PAGE (see Note 13). Dry gels
and expose them for autoradiography using storage phosphor
screens. Phosphorimager-acquired autoradiography images are
shown in Fig. 2 (see Notes 14 and 15 for data analysis and
interpretation).
242 Raffaele Ieva

4  Notes

1. To slow down EspP biogenesis and increase the temporal reso-

lution of sequential EspP interactions, a modified EspP variant
is used. EspP(586TEV) harbors a short linker insertion in the
EspP passenger domain that delays (but does not permanently
block) both passenger domain secretion and β domain inser-
tion into the lipid bilayer [9, 10].
2. When available, use structural data to guide the strategy for
site-specific incorporation of photoprobes. Trp 1149 is situ-
ated in a β-strand of the β domain with its side chain projecting
into the outer leaflet of the outer membrane lipid bilayer [14].
Thus, a photoprobe at position 1149 would be predicted to be
in close proximity to the outer membrane LPS upon comple-
tion of EspP biogenesis.
3. A series of plasmids for efficient incorporation of unnatural
amino acids into proteins of interest using amber suppression
have been developed by Schultz and coworkers [15] and
deposited into the Addgene plasmid repository.
4. This washing step helps to remove any β-lactamase that might
have been released in the medium of overnight cultures. Thus,
the concentration of ampicillin in the new culture remains
optimal for plasmid maintenance throughout the entire time
of cell growth.
5. The level of induction of a given protein of interest must be
determined empirically. Excessive protein overexpression may
lead to the formation of aggregates, which can generate unde-
sired photocrosslinking reactions.
6. Bpa is highly insoluble at a neutral pH, thus a 1000× stock solu-
tion of 1 M Bpa is prepared in 1 M NaOH. To facilitate rapid
dilution of Bpa and prevent precipitation when added to the cul-
ture, slowly dispense the Bpa stock solution in the culture medium
using a micropipette while gently swirling the culture flask.
7. It is recommended to start a parallel identical culture in G56
containing 0.13 mM KH2PO4 (instead of 32P-labeled inorganic
phosphate) to monitor cell growth. This will limit the risk of
contaminating equipment with 32P while measuring the cul-
ture optical density to assess bacterial growth.
8. Turn on the UV lamp 10 min before sample irradiation. After
this warming-up time the lamp irradiates with maximal power.
Wear protective goggles.
9. The positioning of the UV lamp with respect to the samples will
depend on its wattage power. The described protocol is con-
ducted using a 100 W mercury lamp. The lamp is positioned
approximately 4 cm from the samples. To prevent excessive
 In vivo Site-Directed Photocrosslinking 243

heating of samples, it is important to adjust the position of the

lamp so that the added ice chips do not melt completely during
UV irradiation.
10. Do not switch off the UV lamp in between irradiations of differ-
ent samples, unless the waiting time is longer than 20 min. Once
switched off, it takes about 15 min before it can be reactivated.
11. To successfully conduct the pulse-chase technique, it is critical
that the several described steps are conducted in parallel and
coordinated in a timely fashion. Thus, the labeling and prepa-
ration of tubes where different samples will be collected
(described in the previous section) must be conducted before
proceeding with pulse-chase labeling. In addition, UV irradia-
tion of samples is performed in the interval times of the chase
phase. To minimize the time of handling and preparation of
the samples, make sure that the following reagents have been
thawed in advance and are ready for use on the bench: an ali-
quot of the 35S-Met/35S-Cys protein labeling mix, cold methi-
onine and cysteine, an automatic pipettor and disposable 10
mL pipettes, a P100 micropipette set to 35 μL, and a P1000
micropipette set to 350 μL.
12. With samples labeled with 32P-inorganic phosphate, four washing
steps with high-salt RIPA buffer should be conducted to reduce
the background of radioactivity detected after SDS-PAGE.
13. The type of gel and the concentration of polyacrylamide to
choose will depend on the sizes of the protein that contains the
photocrosslinker and the generated crosslinking products. In
an initial experiment, it might be necessary to analyze samples
with different gels of varying polyacrylamide concentrations to
increase the chance of resolving crosslinking products.
14. Analysis of pulse-chase labeled samples not irradiated with UV
light reveals that an approximately 135 kDa band, correspond-
ing to EspP, is progressively converted to an approximately 30
kDa band (Fig. 2, lanes 1–3). This conversion is the result of
EspP passenger domain cleavage by an autoproteolytic reaction
occurring in the membrane-integrated β domain [16]. Upon
exposure to UV irradiation, two high-molecular-weight prod-
ucts are immunoprecipitated with EspP antibodies (Fig. 2, lanes
10–12, products labeled “1” and “2”). In parallel reactions, an
antiserum against BamA precipitates product 1 (Fig. 2, lane 16),
while an antiserum against BamB precipitates product 2 (Fig. 2,
lane 13). The signal intensities of EspP-BamA and EspP-BamB
crosslinking products are maximal after 1 min chase and decrease
over time (lanes 10–18), indicating that EspP position 1149
interacts with BamA and BamB at an early stage of biogenesis
prior to cleavage of the secreted passenger domain. Another
crosslink product detected in samples exposed to UV irradiation
244 Raffaele Ieva

migrates 2–4 kDa higher than the EspP β domain (Fig. 2, lanes
10–12, product labeled “3”). This crosslinking product is gen-
erated with high efficiency, and its amount is proportional to the
amount of mature EspP β domain that forms over time (lanes
10–12). Thus crosslinking product 3 results from a stable inter-
action of the mature EspP β domain with a low-­ molecular-­
weight factor. In a parallel experiment, cellular phospholipids are
labeled with 32P. Upon UV irradiation, EspP-specific antibodies
precipitate a 32P-labeled product that runs with the same appar-
ent molecular weight of crosslinking product 3 (Fig. 2, lane 20).
Furthermore, this crosslinking product can be detected using
antibodies specific for E. coli LPS [10, 11]. Thus, product 3
reveals an interaction of EspP with lipids of the outer leaflet of
the outer membrane.
15. This protocol detects two types of interactions for EspP amino
acid position 1149: (i) transient interactions with BamA and
BamB at an early stage of EspP biogenesis, prior to passenger
domain cleavage; (ii) an interaction of the mature EspP β
domain with the LPS of the outer membrane, following secre-
tion and cleavage of the passenger domain, that remains stable
over time. Together with analyses of the interactions occurring
at other amino acids of EspP, the site-directed photocrosslink-
ing approach in metabolically labeled cells has helped to
describe consecutive intermediate steps of the autotransporter
assembly reaction in the bacterial outer membrane [9–11]. In
brief, distinct autotransporter segments interact with periplas-
mic chaperons such as Skp and SurA at an early stage of its
transport through the cellular envelope. At a later time, tran-
sient interactions with one of three subunits of the BAM com-
plex (BamA, BamB, and BamD) are mapped at amino acid
positions that are located approximately 120° from each other
at the periplasmic side of the folded EspP β-barrel, suggesting
the formation of an intermediate where the assembling EspP β
domain is at the center of the BAM complex. On the circum-
ference of the folded barrel, amino acid 1149 is positioned
between the BamA and the BamB interacting sites. At a subse-
quent stage, a stable interaction of position 1149 with the
outer membrane LPS indicates the release of the EspP β
domain by the BAM complex into the lipid bilayer.

Acknowledgements

This protocol was originally developed in the laboratory of Dr.

Harris Bernstein (National Institutes of Health, Bethesda, MD,
USA). Plasmid pDULE-pBpa was kindly provided by Dr. Peter
Schultz (Scripps Research Institute, La Jolla, CA, USA). R.I. is
supported by the CNRS-Inserm ATIP-Avenir program.
 In vivo Site-Directed Photocrosslinking 245

References

1. Holland IB (2010) The extraordinary diver- during its translocation across the bacterial
sity of bacterial protein secretion mechanisms. outer membrane. Proc Natl Acad Sci U S A
Methods Mol Biol 619:1–20 106:19120–19125
2. Ellman J, Mendel D, Anthony-Cahill S et al 10. Ieva R, Tian P, Peterson JH et al (2011)
(1991) Biosynthetic method for introducing Sequential and spatially restricted interac-
unnatural amino acids site-specifically into pro- tions of assembly factors with an autotrans-
teins. Methods Enzymol 202:301–336 porter beta domain. Proc Natl Acad Sci U S A
3. Chin JW, Martin AB, King DS et al (2002) 108:E383–E391
Addition of a photocrosslinking amino acid to 11. Pavlova O, Peterson JH, Ieva R et al (2013)
the genetic code of Escherichia coli. Proc Natl Mechanistic link between β barrel assembly and
Acad Sci U S A 99:11020–11024 the initiation of autotransporter secretion. Proc
4. Dormán G, Prestwich GD (1994) Natl Acad Sci U S A 110:E938–E947
Benzophenone photophores in biochemistry. 12. Ganong BR, Leonard JM, Raetz CR (1980)
Biochemistry 33:5661–5673 Phosphatidic acid accumulation in the mem-
5. Dautin N, Bernstein HD (2007) Protein secre- branes of Escherichia coli mutants defective
tion in gram-negative bacteria via the auto- in CDP-diglyceride synthetase. J Biol Chem
transporter pathway. Annu Rev Microbiol 255:1623–1629
61:89–112 13. Akiyama Y, Ito K (1990) SecY protein, a
6. Leyton DL, Rossiter AE, Henderson IR membrane-­ embedded secretion factor of E
(2012) From self sufficiency to dependence: coli, is cleaved by the ompT protease in vitro.
mechanisms and factors important for auto- Biochem Biophys Res Commun 167:711–715
transporter biogenesis. Nat Rev Microbiol 14. Barnard TJ, Dautin N, Lukacik P et al (2007)
10:213–225 Autotransporter structure reveals intra-barrel
7. Hagan CL, Silhavy TJ, Kahne D (2011) cleavage followed by conformational changes.
β-Barrel membrane protein assembly by the Nat Struct Mol Biol 14:1214–1220
Bam complex. Annu Rev Biochem 80:189–210 15. Young TS, Ahmad I, Yin JA et al (2010) An
8. Noinaj N, Rollauer SE, Buchanan SK (2015) enhanced system for unnatural amino acid
The β-barrel membrane protein insertase mutagenesis in E coli. J Mol Biol 395:361–374
machinery from Gram-negative bacteria. Curr 16. Dautin N, Barnard TJ, Anderson DE et al
Opin Struct Biol 31:35–42 (2007) Cleavage of a bacterial autotransporter
9. Ieva R, Bernstein HD (2009) Interaction of an by an evolutionarily convergent autocatalytic
autotransporter passenger domain with BamA mechanism. EMBO J 26:1942–1952
 Chapter 20

Protein–Protein Interactions: Pull-Down Assays

Arthur Louche, Suzana P. Salcedo, and Sarah Bigot

Abstract
Determining protein partners is an essential step toward understanding protein function and identifying
relevant biological pathways. Many methods exist for investigating protein–protein interactions. The pull-­
down assay is an in vitro technique used to detect physical interactions between two or more proteins and
an invaluable tool for confirming a predicted protein–protein interaction or identifying novel interacting
partners. This method typically involves the use of affinity purification with various wash and elution steps.
In this chapter, we describe how an interaction between two purified bacterial proteins or between bacte-
rial and eukaryotic proteins can be detected by pull-down experiments.

Key words Pull-down, Protein–protein interactions, Tagged protein, Affinity purification

1  Introduction

Pathogenic bacteria produce virulence factors that usually help the

pathogen to survive in an environmental niche, to promote coloni-
zation and invasion of host tissues, or to modulate the immune
system. Virulence factors are toxins or effector proteins than can be
transported by diverse secretion machineries in bacteria [1, 2].
Once secreted, these proteins can be assembled on the bacterial
cell surface, released in the extracellular space, or secreted directly
into a host cell or a neighboring bacterium. Once in host cells,
effectors often target key proteins to hijack the host cellular
machinery to remodel signaling cascades. The yeast two-hybrid
system is often used to screen a large number of host proteins that
potentially interact with bacterial effectors [3]. Regarding the
mechanism of the secretion systems, a bacterial two-hybrid system
is frequently employed to identify interaction networks between
components of the secretory apparatus, as well as interaction
between effectors and proteins of the machinery [4]. However,
protein–protein interactions that have been determined by two-­
hybrid assay must be confirmed by other methods [5].

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_20, © Springer Science+Business Media LLC 2017

247
248 Arthur Louche et al.

Pull-down is an in vitro method widely used to detect or con-

firm interactions among multiple proteins. This assay is similar in
methodology to co-immunoprecipitation experiments in its use of
an affinity ligand to capture interacting proteins. The difference
between these two methods is that while co-immunoprecipitation
uses immobilized antibodies to capture protein complexes, the
pull-down approach uses a purified and tagged protein as a “bait”
to bind any interacting proteins. The method consists of first
immobilizing the tagged protein (bait) on an affinity ligand specific
to the tag, creating an affinity support to capture and purify other
proteins (prey) that interact with the bait. The bait and prey pro-
teins can be obtained from multiple sources, such as cell lysates,
purified proteins, expression systems, and in vitro transcription/
translation systems. Once the prey proteins have been incubated
with an immobilized bait protein, interacting complexes are eluted
using an eluting buffer depending on the affinity ligand. Each
experiment needs proper controls to demonstrate that character-
ized interactions are not an artifact. For example, a positive control
consisting of an immobilized bait protein alone is necessary to
verify proper attachment of the tagged bait protein to the affinity
support. To identify and eliminate false positives caused by nonspe-
cific binding of prey proteins to the affinity support, cell lysates or
purified proteins can be analyzed after being passed through a
minus bait support. Following a pull-down experiment, protein
fractions are resolved by sodium dodecyl sulfate-polyacrylamide
gel electrophoresis (SDS-PAGE) and then visualized by gel stain-
ing or western-blotting detection.
In this chapter, we describe detailed pull-down assay proce-
dures that allow the identification of interacting proteins. First, we
focus on how to perform a pull-down experiment to identify an
interaction between a bacterial bait protein and eukaryotic prey
proteins expressed in host cells (Subheadings 3.1 and 3.2). Next,
we present how the interaction between two purified proteins can
be visualized by a pull-down assay (Subheading 3.3). In these pro-
cedures, pull-down experiments have been performed using spe-
cific bait proteins fused to a 6× histidine tag. As a consequence, we
selected Ni-NTA agarose beads as the affinity support used to
immobilize these recombinant proteins.

2  Materials

Prepare all solutions with distilled water at room temperature and

keep them at the indicated temperatures.

2.1  Preparation 1. Eukaryotic cells.

of Cell Lysate 2. Cell culture dish, treated for optimal cell attachment, with
growth surface area around 55 cm2, sterile.
 Pull-Down Assays 249

3. Plasmid containing the gene of interest fused to a specific tag

(obtained from a EndoFree maxipreparation).
4. Transfection reagent.
5. Phosphate buffered saline (PBS): Prepare a 10× solution with
bidistilled water (18.2 MΩ cm) containing 10.6 mM KH2PO4,
30 mM Na2HPO4, 2H2O, and 1.54 M NaCl, and sterilize with
a 0.2 μm filter. The 1× solution obtained following dilution
with bidistilled water will have a pH of around 7.4.
6. Radioimmunoprecipitation assay (RIPA) buffer: Ready-to-use
solution containing 150 mM NaCl, 1.0% IGEPAL® CA-630,
0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris, pH 8.0.
7. Antiprotease cocktail: Mix 1% (v/v) of protease inhibitor cock-
tail (Sigma-Aldrich), phosphatase inhibitor cocktail 2 (Sigma-­
Aldrich), phosphatase inhibitor cocktail 3 (Sigma-Aldrich),
and phenylmethylsulfonyl fluoride (PMSF).

2.2  Pull-­ 1. 1 M Tris–HCl, pH 7.5 stock solution. Weigh 121.1 g Tris base
Down Assays and transfer to a 1 L graduated cylinder. Add water to 800 mL,
mix, adjust pH with HCl, and make up to 1 L with water.
Store at room temperature (see Note 1).
2. 5 M NaCl stock solution. Weigh 292.2 g NaCl and transfer to
a 1 L graduated cylinder. Add water to 800 mL, stir, and adjust
volume to 1 L with water (see Note 1).
3. Equilibrium buffer (see Note 2): 20 mM Tris–HCl, pH 7.5,
250 mM NaCl. Mix 1 mL 1 M Tris–HCl, pH 7.5 stock solu-
tion with 2.5 mL 5 M NaCl stock solution in a 50 mL centri-
fuge tube, and add water to a volume of 50 mL. Keep at 4 °C
(see Note 3).
4. Elution buffer (see Note 2): 20 mM Tris–HCl, pH 7.5, 250
mM NaCl, 500 mM imidazole. Weigh 1.7 g imidazole in 50
mL solution of equilibrium buffer. Keep at 4 °C (see Note 3).
5. Purified His-tagged protein (bait).
6. Ni-NTA agarose beads: 6% beaded agarose (cross-linked), pre-
charged with Ni2+ (Protino® Ni-NTA Agarose, Macherey
Nagel, or equivalent). Store at 4 °C (see Note 4).
7. 0.8 mL empty columns for gravity flow (Pierce™ Centrifuge
Columns, Thermo Fisher Scientific, or equivalent).
8. Refrigerated microcentrifuge.

2.3  Sodium Dodecyl 1. Resolving gel: 1.5 M Tris–HCl, pH 8.8. Weigh 90.8 g, transfer
Sulfate (SDS) to 500 mL graduated cylinder, and add 300 mL water. Adjust
Polyacrylamide Gel pH with HCl and fill with water to 500 mL. Store at room
Components temperature.
2. Stacking gel buffer: 0.5 M Tris–HCl, pH 6.8. Weigh 30.275 g,
transfer to 500 mL graduated cylinder, and add 300 mL water.
250 Arthur Louche et al.

Adjust pH with HCl and fill with water to 500 mL. Store at

room temperature.
3. 30% acrylamide/Bis solution (37.5:1 acrylamide:Bis). Store at
4 °C.
4. Ammonium persulfate (APS): 20% solution in water. Store at
−20 °C (see Note 5).
5. N,N,N′,N′-tétraméthyléthylènediamine (TEMED). Store at
room temperature.
6. SDS-PAGE running buffer: 25 mM Tris–HCl, 192 mM gly-
cine, 0.1% SDS. Prepare 10× running buffer solution: Weigh
30 g Tris base, 144 g glycine, and 10 g SDS and add distilled
water to 1 L. Store at room temperature. Prepare fresh 1×
solution before gel electrophoresis.
7. Laemmli lysis buffer [6], 4× concentrate: 62.5 mM Tris–HCl
pH 6.8, 2% SDS, 10% glycerol, 0.01% bromophenol blue, 5%
β-mercaptoethanol. Store at −20 °C (see Note 6).
8. Protein ladder.

3  Methods

3.1  Preparation 1. Seed eukaryotic cells at 5.105 in 10 cm cell culture dish
of Cell Lysate (see Note 7) and incubate overnight at 37 °C in CO2.
2. Transfect cells with plasmid containing gene of interest fused
to a specific tag with appropriate transfection reagent for time
necessary for optimal expression of protein (16–24 h is usually
a good range).
3. Cool cells by placing plates on ice, wash cells with 1× PBS. Add
2 mL cold PBS and harvest cells using cell scraper.
4. Centrifuge 5 min at 80 × g at 4 °C.
5. Resuspend cells with 200 μL RIPA buffer supplemented with
antiprotease cocktail.
6. Incubate on ice 20 min and mix gently every 5 min with a
P200 micropipette.
7. Stock prepared cells at −80 °C (see Note 8).
8. Right before pull-down experiment, thaw prepared cell extract.
Centrifuge at 17,000 × g at 4 °C for 20 min. Use the superna-
tant as prey by following step 9 in Subheading 3.2 (see
Note 9).

3.2  Pull-Down Assay 1. Transfer 120 μL Ni-NTA agarose beads to gravity flow column
Using Cell Lysate (see Note 12).
as Prey (See Notes 10 2. Centrifuge column for 1 min at 1000 × g at 4 °C. Discard
and 11) flow-through.
 Pull-Down Assays 251

3. Add 400 μL distilled water to column (see Note 13).

4. Centrifuge column for 1 min at 1000 × g at 4 °C. Discard
flow-through.
5. Mix carefully 50 μg His-tagged protein (bait) with 400 μL equi-
librium buffer and load onto column (see Notes 14 and 15).
6. Incubate 1 h (see Note 16) with agitation at 4 °C (see Note
17) and 10 min on ice without agitation (see Note 18).
7. Centrifuge column for 1 min at 1000 × g at 4 °C and keep
flow-through.
8. Load flow-through to column, and centrifuge column for
1 min at 1000 × g at 4 °C (see Note 19). Keep flow-through at
4 °C for analysis.
9. Mix 200 μL cell extract (see Note 20) with 200 μL equilibrium
buffer and load onto column (see Note 21).
10. Incubate 1 h at 4 °C under agitation (see Note 22) then 10 min
on ice without agitation (see Note 18).
11. Centrifuge column for 1 min at 1000 × g at 4 °C. Keep flow-­
through for analysis.
12. Wash column by adding to column 400 μL equilibrium
buffer.
13. Centrifuge column for 1 min at 1000 × g at 4 °C. Discard
flow-through.
14. Wash column by adding to column 400 μL equilibrium buffer
containing 50 mM imidazole. Keep the first washing for
analysis.
15. Centrifuge column for 1 min at 1000 × g at 4 °C. Discard
flow-through.
16. Repeat steps 14 and 15 three times and go to step 17. Keep
last washing fraction at 4 °C for analysis.
17. Elute by loading 80 μL elution buffer to column and incubate
10 min at 4 °C (see Note 18).
18. Centrifuge column for 1 min at 1000 × g at 4 °C and keep
eluted fraction.
19. Repeat steps 17 and 18 with eluted fraction (see Note 22).
Keep eluted fraction at 4 °C for analysis.

3.3  Pull-Down Assay 1. Incubate 50 μg His-tagged bait protein with 50 μg purified prey
Using Purified Protein protein in total volume of 400 μL equilibrium buffer (see Note
as Prey (See Note 11) 23) 2 h 30 min at 4 °C under agitation (see Notes 17 and 24).
2. Add 80 μL Ni-NTA agarose beads to gravity flow column and
follow steps 1–4 of Subheading 3.2.
252 Arthur Louche et al.

3. Equilibrate column by adding 400 μL equilibrium buffer sup-

plemented with 20 mM imidazole.
4. Centrifuge column for 1 min at 1000 × g at 4 °C. Discard
flow-through.
5. Load 400 μL incubated bait and prey proteins onto column.
Incubate 10 min on ice without agitation (see Note 18).
6. Centrifuge column for 1 min at 1000 × g at 4 °C. Keep flow-­
through at 4 °C for analysis.
7. Wash by adding to column 400 μL equilibrium buffer supple-
mented with 20 mM imidazole.
8. Centrifuge column for 1 min at 1000 × g at 4 °C. Save first
washing at 4 °C for analysis.
9. Repeat washing steps 7 and 8 four times and keep last washing
fraction at 4 °C for analysis.
10. Add 200 μL elution buffer to column and incubate on ice
10 min.
11. Centrifuge column for 1 min at 1000 × g at 4 °C. Keep eluted
fraction at 4 °C for analysis.

3.4  SDS-PAGE 1. To 15 μL protein fraction add 5 μL Laemmli lysis buffer, 4×

and Analysis concentrate. Heat for 3 min at 100 °C and centrifuge 30 s
of Protein Fractions using a microcentrifuge to bring down condensate.
2. Load 10 μL protein fraction and 5 μL protein ladder on SDS-­
polyacrylamide gel.
3. Electrophorese proteins in running buffer at 100 V for 15 min
then 180 V until dye front has reached bottom of gel.
4. Identify interacting proteins by immunodetection or blue coo-
massie coloration (see Note 25).

4  Notes

1. We prefer not to use the solutions after 6 months of storage.

2. A different buffer, such as HEPES (4-(2-hydroxyethyl)-1-­
piperazineethanesulfonic acid), MES (2-(N-morpholino) eth-
anesulfonic acid), or phosphate buffers, may be required for
your specific protein–protein interaction. Additionally,
­different pH values may be tested as these are specific and
dependent on the interaction between proteins.
3. We found that pull-down experiments work better with fresh
equilibrium and elution buffers.
4. The bait proteins used in this protocol are tagged with 6× His
that bind the nickel agarose affinity support. The choice of the
matrix-associated antibody depends on the fusion tag. The His
 Pull-Down Assays 253

tag is composed of a peptide motif that consists of six histidine

residues with a high affinity towards metals like nickel that
composes the used Ni-NTA agarose but also the Ni-­NDA,
Ni-TED, or Ni-TALON resins. The 6× His tag is very small
(~1 kDa), which renders it less immunogenic than other larger
tags, is shown not to affect the native conformation of bait
proteins, and maintains its partner binding activity. Few natu-
rally occurring proteins also bind to Ni-NTA matrices, making
this tag the most commonly used affinity tag. In pull-­down
assays, the choice of the matrix-associated antibody depends
on the fusion tag. What follow are some examples of tags with
their advantages and disadvantages. The FLAG tag is an octa-
peptide that is likely located on the surface of the fusion pro-
tein due to the hydrophilic nature of amino acid residues and
has affinity to anti-FLAG resin. Like the His tag, the FLAG tag
is small, but a disadvantage is that the monoclonal antibody
matrix is not as stable as Ni-NTA. Glutathione S-transferase
(GST) tag binds to glutathione-associated support with high
affinity and specificity. This tag has the advantage that GST
isoforms are not normally found in bacteria, so purified bacte-
rial prey proteins normally do not have affinity with glutathi-
one resin. However, GST tag is large (26 kDa), exists as a
dimer, is prone to nonspecific interaction, is expensive, and
affinity to its support depends on certain reagents. The malt-
ose-binding protein (MBP) tag from an Escherichia coli peri-
plasmic protein has affinity for matrix consisting of sugars or
anti-MBP. This tag is used for the purposes of overcoming
problems associated with the expression and purification of
recombinant proteins [7]. However, the disadvantage of the
MBP tag is its large size, its immunogenicity, and the mild elu-
tion of MBP-tagged proteins, which complicate pull-­ down
experiments.
5. Make an aliquot of 1 mL before −20 °C storage. This will pre-
vent the degradation caused by repeated thawing.
6. Make an aliquot of 500 μL before −20 °C storage. The used
Laemmli lysis buffer can be kept at 4 °C for 1 month.
7. As negative control, prepare a cell lysate without expressing
bait protein (negative cell lysate). This will eliminate false posi-
tives resulting from nonspecific interactions of cell lysate
­proteins with the Ni-NTA agarose beads. Additional negative
controls can include an irrelevant protein with the same tag or
expression of the tag alone, as in the case of the GFP.
8. Before stocking the cells, remove an aliquot and control by
western blot the production of the prey protein.
9. Whole-cell lysate instead of the supernatant fraction can also
be used to test whether the prey protein of interest localizes in
the pellet fraction.
254 Arthur Louche et al.

10. Pull-down experiments using cell lysates will not demonstrate

that interaction between the bait and prey proteins is direct but
only determine that they are part of the same complex. To
prove a direct interaction, the prey protein must be purified
and used in pull-down experiments as described in Subheading
3.3.
11. Try to work mostly on ice or at 4 °C to prevent the degrada-
tion or the denaturation of the proteins.
12. Break the end cap of the gravity flow column and place it on a
1.5 mL Eppendorf tube. Thoroughly resuspend the Ni-NTA
resin by inverting the bottle several times to obtain a uniform
suspension. Pipette tips must be cut to allow the Ni-NTA aga-
rose beads to get into.
13. This step eliminates the left 30% ethanol present in the Ni-­
NTA resin.
14. Before loading the bait protein, plug the gravity flow column
using a piece of parafilm before replacing it on a 2 mL
Eppendorf tube.
15. Prepare a supplementary column by mixing 50 μg of a known
noninteracting bait fused to 6× His tag with 400 μL equilib-
rium buffer to an empty column. Additionally, prepare a col-
umn by adding 400 μL equilibrium buffer to an empty column.
These negative bait columns will be used in combination with
cell lysates to eliminate false positives resulting from nonspe-
cific interactions.
16. The incubation time can be increased from several hours to
overnight at 4 °C under agitation depending on the strength
of the interaction between bait and prey proteins.
17. Rotate on roller or rotating platform.
18. The column should stand straight on the ice. This step allows
the resin to flow by gravity before centrifugation.
19. We found that loading two times the flow-through increases
the capacity of the binding.
20. The volume is dependent on the protein concentration of the
cell extract. As a guide, 125–150 μg of protein of a cell extract
is usually incubated per microgram of bait protein. A
­ lternatively,
cell extract samples can be normalized by visualization of trans-
fected proteins to ensure equivalent expression of the prey and
the relevant controls (see Note 7).
21. Several controls should be added at this step. Load 400 μL
equilibrium buffer without prey protein to analyze the effi-
ciency of the immobilization of the bait protein. As negative
controls, load onto the negative column (see Note 12) 200 μL
cell lysate containing the prey protein or the negative cell lysate
(see Note 7) mixed with 200 μL equilibrium buffer. Additionally,
 Pull-Down Assays 255

load 200 μL negative cell lysate mixed with 200 μL equilib-

rium buffer onto the column associated with the bait protein.
22. We found that loading two times the eluted fraction increased
its quantity.
23. As negative control, incubate 50 μg bait protein (minus prey)
or prey protein alone (minus bait) in 400 μL equilibrium buf-
fer. The minus prey control will ensure that the Ni-NTA aga-
rose resin will correctly capture the His-tagged bait protein
alone. The minus bait control will eliminate false positives
resulting from an interaction between affinity support and prey
protein.
24. A different incubation temperature and time may be required
for your specific protein–protein interaction.
25. A prey protein that interacts with the bait protein will be found
in the eluted fraction. In contrast, a noninteracting protein will
not be retained by the bait protein, will pass through the col-
umn, and will be found in the flow-through protein fraction.

References
1. Costa TRD, Felisberto-Rodrigues C, Meir A, Cascales E (2014) Architecture and assembly of
Prevost MS, Redzej A, Trokter M, Waksman the Type VI secretion s­ystem. Biochim Biophys
G (2015) Secretion systems in Gram-negative Acta 1843:1664–1673
bacteria: structural and mechanistic insights. 5. Boucrot E, Henry T, Borg J-P, Gorvel J-P,
Nat Rev Microbiol 13:343–359 Méresse S (2005) The intracellular fate of
2. McBride MJ, Nakane D (2015) Flavobacterium Salmonella depends on the recruitment of
gliding motility and the type IX secretion sys- kinesin. Science 308:1174–1178
tem. Curr Opin Microbiol 28:72–77 6. Laemmli UK (1970) Cleavage of structural
3. Rodríguez-Negrete E, Bejarano ER, Castillo proteins during the assembly of the head of
AG (2014) Using the yeast two-hybrid sys- bacteriophage T4. Nature 227:680–685
tem to identify protein-protein interactions. 7. Di Guan C, Li P, Riggs PD, Inouye H (1988) Vectors
Methods Mol Biol 1072:241–258 that facilitate the expression and purification
4. Zoued A, Brunet YR, Durand E, Aschtgen M-S, of foreign peptides in Escherichia coli by fusion
Logger L, Douzi B, Journet L, Cambillau C, to maltose-binding protein. Gene 67:21–30
 Chapter 21

Protein–Protein Interactions: Surface Plasmon Resonance

Badreddine Douzi

Abstract
Surface plasmon resonance (SPR) is one of the most commonly used techniques to study protein–protein
interactions. The main advantage of SPR is it gives on the ability to measure the binding affinities and
association/dissociation kinetics of complexes in real time, in a label-free environment, and using relatively
small quantities of materials. The method is based on the immobilization of one of the binding partners,
called the ligand, on a dedicated sensor surface. Immobilization is followed by the injection of the other
partner, called the analyte, over the surface containing the ligand. The binding is monitored by subsequent
changes in the refractive index of the medium close to the sensor surface upon injection of the analyte.
During the last 10 years, SPR has been intensively used in the study of secretion systems because of its
ability to detect highly dynamic complexes that are difficult to investigate using other techniques.
This chapter will guide users in the setup of SPR experiments in order to identify protein complexes and
to assess their binding affinity or kinetics. It will include detailed protocols for (i) the immobilization of
proteins with the amine coupling capture method, (ii) analyte-binding analysis, (iii) affinity/kinetic mea-
surements, and (iv) data analysis.

Key words Surface Plasmon Resonance, Protein–protein interaction, Analyte, Ligand, Affinity,
Kinetics, BIAcore T200

1  Introduction

Secretion systems are multiprotein complexes allowing the trans-

port of a large number of effectors from the inside to the outside
of bacterial cells. The assembly of these supramolecular machiner-
ies is ensured by the formation of protein complexes with extremely
different times of stability, from transitory to stable interactions.
To understand the function of these machineries as well as their
modes of association, it is important to study their building blocks
by identifying the different interacting partners and assessing their
relative affinities and association/dissociation kinetics. For that
purpose, scientists combine genetic, biochemical, and biophysical
tools. During the last decade, the use of surface plasmon resonance
(SPR) in the study of secretion systems has increased spectacularly
[1–12]. This in vitro approach is the method of choice to study

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_21, © Springer Science+Business Media LLC 2017

257
258 Badreddine Douzi

such dynamic systems owing to its ability to detect both weak and
strong interactions ranging from the millimolar to the nanomolar
range [13, 14]. SPR can be used as a primary tool to screen inter-
acting partners or as a validation tool for interactions previously
identified by other methods (e.g., bacterial two-hybrid, co-­
immunoprecipitation, chemical crosslinking). The determination
of the affinity or kinetics of an interaction, as can be done by SPR,
is fundamental to understanding the nature of binding at the cel-
lular level.
In SPR terminology, the immobilized biomolecule is called the
ligand and the binding partner present in solution is called the
analyte. SPR uses an optical method based on the detection of very
tiny modifications of the refractive index of the medium in close
proximity to a metal surface. The binding of the analyte to the
ligand induces a change in the mass concentration at the metal
surface and, consequently, in the refractive index, which is con-
verted to resonance or response units (RUs). The metal surface
typically consists of a thin gold layer and contains flow cells with
very small volume (less than 100 nL). In a typical SPR experiment,
the first step consists in the immobilization of the ligand in one
flow cell by covalent or noncovalent capture methods. In the sec-
ond step, the analyte is injected into the flow cell containing the
immobilized ligand. The presence of a reference flow cell, which
can either be empty or contain a protein irrelevant for the studied
complex, is important to monitor the true binding of the analyte
to the ligand. The binding of the analyte on the ligand induces a
change on the mass concentration at the surface of the flow cell.
The signal induced by the binding of the analyte to the metal sur-
face in the reference flow cell is subtracted from the signal obtained
from the binding of the analyte to the ligand. Consequently, upon
binding, the curve that we observe, called a sensogram, is divided
into three distinct parts (Fig. 1). The first phase is the association
phase, during which the analyte molecules bind to the binding sites
of the ligand, leading to the increase in RUs. The curve obtained
can be used to calculate the rate of association (kon). The next phase
is a steady state, during which the association/dissociation events
are equal. The level of RUs obtained in this equilibrium phase is
called the response at equilibrium, Req, and can be used to calcu-
late the binding affinity (KD). Upon stopping the injection of the
analyte over the flow cell, the dissociation phase takes place: the
ligand dissociates from the surface as the running buffer flows over
the chip. The dissociation of the analyte from the ligand binding
site induces a decrease in RUs. The corresponding curve can be
used to measure the rate of dissociation (koff). The full removal of
the analyte from the ligand binding sites takes place during the
regeneration step, which is determinant for keeping the surface
intact and ready for new injection cycles.
 Surface Plasmon Resonance 259

Fig. 1 Different phases of an SPR sensogram. Initial baseline is obtained after the immobilization of the ligand.
Following injection of analyte over the ligand, the increase in the RUs corresponds to the association phase.
When the number of association and dissociation events is equal, the equilibrium level is maintained. The
dissociation phase starts when stopping the analyte injection: the decrease in the RUs corresponds to the
flowing out of the analyte from the ligand surface. The regeneration step aims at the full removal of the analyte
and at reaching the baseline level

Several SPR-based systems have been developed by different

manufacturers. The most widely used system is the BIAcore devel-
oped by GE Healthcare. Despite the high cost of the machines and
related products, this system offers many advantages, which has
encouraged scientists to acquire it. The most important advantages
are the capacity to detect weak interactions and the reproducibility
of the obtained results.
This chapter illustrates the setup procedures routinely used for
studying protein–protein interactions by SPR on a BIAcore T200.
It describes the amine coupling immobilization method as well as
analyte binding analysis, affinity/kinetic measurements, and data
analysis.

2  Materials

All buffers are prepared with ultrapure water and analytical-grade

reagents. All prepared buffers should be stored at 4 °C unless spe-
cifically required otherwise.
–– Series S sensor chip CM5 from GE Healthcare (see Note 1).
–– BIAcore T200 system from GE Healthcare Life Sciences (see
Note 2).
–– Amine coupling kit: the kit contains 750 mg 1-ethyl-3-(3-­
dimethylaminopropyl) carbodiimide hydrochloride (EDC),
115 mg N-hydroxysuccinimide (NHS), and 10.5 mL 1.0 M
ethanolamine–HCl, pH 8.5.
260 Badreddine Douzi

–– Immobilization buffers: 10 mM sodium acetate buffers at

different pHs: pH 4, pH 4.5, pH 5, and pH 5.5.
–– HBS-EP buffer: 0.01 M HEPES, 0.15 M NaCl, 3 mM EDTA,
0.05% v/v P20, pH 7.4.
–– Regeneration scouting kit or equivalent homemade solutions:
11 mL ethylene glycol (100 wt%), 11 mL 10 mM glycine–
HCl, pH 1.5, 11 mL 10 mM glycine–HCl, pH 2.0, 11 mL 10
mM glycine–HCl, pH 2.5, 11 mL 10 mM glycine–HCl, pH
3.0, 11 mL 4.0 M magnesium chloride, 11 mL 0.2 M sodium
hydroxide, 11 mL 0.5% sodium dodecyl sulphate (SDS), 11
mL 5.0 M sodium chloride, 20 mL surfactant P20 (10% (v/v)
solution of polysorbate 20).
–– Vials: 0.8 mL rounded polypropylene microvials.
–– Vial caps: Penetrable cap made of kraton G (SEBS).
–– Ligand and analyte (see Note 3).
–– Control ligand, e.g., Thioredoxin from Escherichia coli.
–– Tabletop centrifuge.

3  Methods

In a typical protein–protein interaction study using an SPR-based

system, many tasks should be undertaken in the following order:
1. Choice of protein to immobilize and protein to use as analyte
(see Note 4).
2. Choice of immobilization type and sensor surface (see Note 5).
3. Choice of immobilization level (see Note 6).
4. Preparation of ligand and analyte.
5. Preparation of material and buffers (see Subheading 2).
6. pH scouting.
7. Immobilization of ligand.
8. Immobilization of control ligand.
9. Analyte binding analysis.
10. Regeneration optimization.
11. Affinity and kinetic measurements.
12. Data analysis.
These steps will be developed in subsequent sections in more
detail.

3.1  Ligand 1. The purified proteins must be dialyzed ON (overnight) at 4 °C

and Analyte against the running buffer to be used in the SPR experiment
Preparation (see Notes 7 and 8).
 Surface Plasmon Resonance 261

2. Check the purity and the stability of the proteins at least 1 day
before doing the experiment. This can be done by sodium
dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis
(PAGE) and coomassie blue staining (see Note 9).
3. Thaw proteins on ice for 30 min.
4. Centrifuge proteins at 16,000 × g for 20 min at 4 °C.
5. Collect supernatant and keep it on ice on a new 1.5 mL
Eppendorf tube.
6. Measure concentration using NanoDrop or alternative spec-
troscopic method (Bradford or bicinchoninic acid assays). Try
to use the same method during the SPR experiment.

3.2  Material 1. Turn on BIAcore T200 system and prime system with filtered
and Buffer Preparation ultrapure water (see Note 10).
2. Take sensor chip CM5 from 4 °C and keep it at room tempera-
ture (RT) for at least 1 h before experiment.
3. Prepare 1 L running buffer by diluting stock solution ten times.
4. Filter running buffer using 0.22 μM membrane filter.
5. Prime system using running buffer at least three times to be
sure that all system tubing is properly flushed with running
buffer.
6. Insert CM5 sensor chip on sensor chip port and close sensor
port.
7. Prime system three times.
8. When priming is finished, the system will be automatically
shifted to a continuous standby flow.
9. Start a manual run by setting flow path 2 and flow rate of 30
μL min−1.
10. Prepare two vials, one containing 400 μL running buffer and
the other with 400 μL 50 mM NaOH regeneration solution.
11. Eject rack tray and place vials in rack.
12. The rack will be inserted after 1 min.
13. Inject running buffer three times for 1 min.
14. Inject regeneration solution for 30 s after each running buffer
injection until baseline becomes stable.

3.3  pH Scouting (See 1. Open Biacore T200 control software and go to File/Open/
Note 11) New wizard template/Immobilization scouting wizard/New.
2. The next steps consist in setting the different parameters:
–– Specify different solutions used for your experiments and
corresponding pH.
–– Set protein injection and dissociation times; injection time
can be fixed at 2 min, dissociation time at 1 min.
262 Badreddine Douzi

–– Specify regeneration solution, usually 50 mM NaOH

(see Note 12).
–– Set flow rate used for experiment at 10 μL min−1.
3. Dilute protein at different pHs (4, 4.5, 5, and 5.5) using immo-
bilization buffers (see Subheading 2) (see Notes 13 and 14).
4. Final protein concentration after dilution on immobilization
buffers should be in range of 5–200 μg mL−1.

3.4  Immobilization 1. In the Biacore T200 control software dialogue window, go to

of Ligand Using Amine File/Open/New wizard template/Immobilization/New.
Coupling 2. Specify the sensor surface (CM5), the flow path (flow cell 2)
3.4.1  Wizard and the amine coupling method.
Template Method 3. The wizard template offers the possibility to choose between
specifying the aim of your immobilization level on RUs or the
contact time of your ligand and the flow rate.
–– If you choose to specify the aim of your immobilization
level, you have to fix a target level. The system will inject 10
μL of the ligand at 5 μL min−1. This short injection aims to
estimate the rate of the preconcentration based on the level
reached and the slope of the sensogram. This step is fol-
lowed by the injection of the regeneration solution (50
mM NaOH to regenerate the surface before activation).
–– If you choose to specify the contact time and the flow rate,
the concentration of your protein and the contact time
should be estimated to obtain the desired level of Ri
(amount of the immobilized ligand in RUs). It is recom-
mended to use a low flow rate (5 μL min−1) to maximize
the ligand contact time to the surface.
4. Thaw EDC, NHS, and ethanolamine solutions from −20 °C at
RT for 10 min (solutions are provided in amine coupling kit)
(see Notes 15 and 16).
5. Thaw ligand and dilute it on corresponding pH solution at a
concentration between 10 and 50 μg mL−1 (the concentration
depends on the desired Ri and on the level of RUs obtained on
the pH scouting).
6. Eject the rack tray.
7. Place vials in right positions specified when you set up
experiment.
8. Insert rack.
9. Run immobilization program.

3.4.2  Manual Method 1. In Biacore T200 control software, open manual run.
2. Set flow rate to 10 μL min−1.
3. Select flow path 2 (for flow cell 2).
 Surface Plasmon Resonance 263

4. Select start.
5. Dilute your protein in 100 μL 10 mM sodium acetate with
suitable pH and at desired concentration (10–50 μg mL−1).
6. Mix 120 μL EDC (0.4 M in water) and 120 μL NHS (0.1 M
in water).
7. Prepare 120 μL ethanolamine (1 M ethanolamine–HCl at
pH 8.5).
8. Inject EDC/NHS mixture between 6 and 10 min over surface.
9. After activation step, set flow rate to 5 μL min−1.
10. Inject ligand (see Note 17).
11. Inject ethanolamine for 5 min.
12. To calculate the binding capacity, subtract RUs obtained after
activation from RUs obtained after ethanolamine deactivation.
13. Wait until signal is stabilized to perform analyte-binding analysis
(see Note 18).

3.5  Immobilization 1. Dissolve 1 mg Thioredoxin powder in 10 mM sodium acetate

of Control Ligand (See at pH 4 in 1.5 mL Eppendorf tube.
Note 19) 2. Centrifuge protein at 16,000 × g for 20 min at 4 °C.
3. Make aliquots of 80 μL and freeze using liquid nitrogen and
conserve tube at −80 °C.
4. In Biacore T200 control software, open manual run.
5. Set flow rate to 10 μL min−1.
6. Select flow cell number 1 (reference flow cell).
7. Perform tasks 4–13 from Subheading 3.4.2. The target level of
the Thioredoxin immobilized on the reference flow cell should
be the same as the Ri of the ligand.

3.6  Analyte Binding Despite the availability of wizard methods in the Biacore T200
Analysis (See Note 20) control software (wizard template), it is recommended to perform
a manual run (see Note 21).
1. In Biacore T200 control software, open manual run.
2. Set flow rate to 30 μL min−1.
3. Select flow path 2-1 (analyte will be injected over flow cells 1
and 2).
4. Select start.
5. Thaw analyte on ice for 30 min.
6. Measure analyte concentration (see Subheading 3.1).
7. Make dilution of analyte on running buffer. If you estimate
that the affinity is in the micromolar range, prepare 50 μM of
analyte.
264 Badreddine Douzi

8. Eject rack tray.

9. Place sample in selected position.
10. Insert rack.
11. Inject analyte for 1 min (see Note 22).

3.7  Regeneration If analyte-binding analysis is performed using manual run, it is rec-

Optimization ommended to test the regeneration solutions in the same cycle.
(See Note 23)
1. After analyte injection (step 11 in Subheading 3.6), estimate
strength of interaction.
2. Prepare 100 μL 1 M NaCl, 2 M MgCl2, 10 mM Glycine–HCl
(pH 3), 10 mM HCl (pH 2), and 10 mM HEPES-NaOH (pH 9).
3. Inject 30 μL 1 M NaCl. If no dissociation is observed, go to
next step.
4. Inject 30 μL 2 M MgCl2. If no dissociation is observed, go to
next step.
5. Inject 30 μL 10 mM Glycine–HCl. If no dissociation is
observed, go to next step.
6. Inject 30 μL 10 mM HCl, If no dissociation is observed, go to
next step.
7. Inject 30 μL 10 mM HEPES-NaOH.
8. After each injection step, measure amount of analyte remain-
ing bound (see Note 24).

3.8  Affinity 1. Go to T200 Biacore software control dialog interface and open
Measurements (See File/Open/New wizard template/Kinetics-affinity.
Notes 25–27) 2. Specify flow path (2-1) if ligand is immobilized.
3. Specify chip type (CM5).
4. Select regeneration, if needed.
5. Select three startup cycles with buffer to stabilize baseline sig-
nal before starting experiment.
6. Specify contact time, flow rate, and dissociation time (see Note 28).
7. Specify contact time and flow rate of regeneration solution.
8. Specify stabilization period (100 s).
9. Fill sample identity and specify serial twofold dilutions of ana-
lyte used for experiment.
10. Prepare analyte (see Subheading 3.1).
11. Prepare ten twofold dilutions of analyte starting from 10 × KD
as specified in wizard template (see Note 29). Dilutions must
be prepared on vials (see Note 30).
12. Place vials in corresponding positions.
13. Start experiment.
 Surface Plasmon Resonance 265

3.9  Affinity Data 1. Open Biacore evaluation software.

Analysis 2. Open Surface-bound kinetics/affinity from Kinetics/affinity at
bottom of toolbar.
3. Select data to analyze. All sensograms are shown in different
colors except for blanks (light gray) (see Note 31).
4. If you select Next, blank curves will be subtracted from other
sensograms automatically.
5. Select Affinity for steady-state evaluation. The top panel shows
a plot of the response at equilibrium (Req) against analyte con-
centrations (CA) based on the average response of the selected
region on the plateau of each sensogram (see Note 32).
6. Select Next.
7. Select the binding model (1:1 binding).
8. Select Fit.
The calculated KD is shown as a vertical line in the curve plot.
The KD corresponds to the concentration of the analyte at half
the Rmax.

3.10  Kinetics Perform steps 3–13 as described in Subheading 3.8.

Measurements (See
Notes 33–35)

3.11  Kinetics Data 1. Follow steps 1–4 as described in Subheading 3.9.

Analysis (See Note 36) 2. Select kinetics.
3. Select 1:1 binding model and then select Fit.
4. Results are displayed as fitted curves in black over original
sensograms.
5. The Quality control interface gives you the statistic quantify-
ing the quality of your fitting.
6. For access to the calculated values, go to Report window.

4  Notes

1. XanTec bioanalytics GmbH (www.Xantec.com) offers a large

number of different types of sensor surfaces. These chips are
cheaper and give the same results as the Biacore chips.
2. For more details, read “Getting Started BIAcore T200” and
BIAcore T200 instrument handbook, available for download
from the GE Healthcare website.
3. To obtain good data, pure and stable protein samples are criti-
cal. The purity of the ligand is very important to ensure bind-
ing specificity and capacity. Impurities can be immobilized
266 Badreddine Douzi

with the ligand and therefore lead to a nonspecific binding or

alter the accurate determination of the affinity or kinetic mea-
surements. The purity of the analyte is important for the deter-
mination of the affinity and kinetic parameters. In fact,
injections of impurities with the analyte over the ligand can
give false constants or rates owing to an incorrect estimation of
analyte concentrations. Analyte purity must exceed 95% and be
analyzed by SDS-­PAGE. It is important to check the quality of
the ligand and the analyte at least 1 day before experiments. It
is essential to apply rigorous purification and conservation pro-
tocols of protein samples.
4. When studying protein–protein interactions, generally each of
the two protein candidates should be used as ligand in one
experiment and as analyte in another experiment to validate the
interaction. However, many factors must be taken into account
in the choice of protein to be immobilized: the amount, size,
stability, solubility, and valency of the proteins. For example, if
you have a limitation in one of your proteins, you can use it as
ligand because immobilization requires a very low amount of
proteins (5–200 ng). If the protein is unstable, it should be
used as analyte and to immobilize the stable one. If one of your
proteins is multimeric in solution, it should be immobilized and
monomeric proteins should be used as analytes.
5. For protein coupling, two approaches can be used:
–– Covalent immobilization. This strategy uses amine, thiol,
or aldehyde functional groups on proteins. Covalent cou-
pling methods use free carboxymethyl groups exposed on
the surface of the sensor chip [15].
–– Noncovalent immobilization. Three capture approaches
are frequently used: capture of biotinylated molecules on
streptavidin sensor chip [16], nickel-chelated nitrilotriace-
tic acid (NTA) groups for His-tagged proteins [17], and
capture by immobilized specific antibodies [18].
Amine coupling is the more frequently used method. It
includes three-step reactions: EDC/NHS activation, followed by
ligand immobilization by its primary amine (Lysine), and finally
deactivation of the remaining free ester groups by ethanolamine.
The choice of immobilization method is critical in an SPR
experiment. Generally, the amine coupling method is tested first.
However, this approach has many limitations, and alternative
immobilization techniques, such as noncovalent immobilization,
must be adopted in some cases. Among these limitations, amine
coupling can induce ligand precipitation or aggregation owing to
the low pH used during immobilization, or the immobilization can
hold the ligand in inactive conformation.
 Surface Plasmon Resonance 267

6. The immobilization is optimized when you have on the surface

just enough ligand in active form and homogeneously ori-
ented. High-density surfaces (immobilizing a high concentra-
tion of ligand) must be avoided because they can introduce
mass transport issues during the association phase and a rebind-
ing effect of analyte to ligand during the dissociation phase.
Thus, a low level of ligand immobilization and the use of a
higher flow rate are recommended in the case of kinetic mea-
surements. A simple way to determine the level of the ligand to
be immobilized is to calculate a theoretical Rmax of the interac-
tion to be studied. The Rmax is the maximum analyte response
capacity of the surface expressed on RUs. It depends on the
molecular weight of both the ligand and the analyte and the
stoichiometry of the reaction. This theoretical Rmax is calcu-
lated assuming that 100% of the ligand molecules will be active
once immobilized (Fig. 2).
For example, to study the interaction between GspH (16 kDa)
and GspJ (26 kDa) from the Type II secretion system of
Pseudomonas aeruginosa, 180 RUs of GspJ should be immobilized
to obtain an Rmax of 300 assuming that GspJ has one binding site
on GspH (S = 1):
Ri = ( MW ligand / MW analyte ) ´ Rmax ´ (1 / S ) .

Ri = (16 / 26 ) ´ 300 ´ (1 / 1) => Ri = 180 RUs .

The experimental Rmax is always lower than the theoretical cal-
culated Rmax. In fact, it is difficult to obtain 100% active ligand on
the surface (having the same orientation, immobilized at the same
interface, where all binding sites are available and all molecules are
correctly folded). To avoid this heterogeneity, it is recommended
that between 20 and 50% of additional ligand be immobilized.
In general, kinetic measurements require a small amount of
ligand immobilization (Rmax between 50 and 150 RUs). For affin-
ity studies, the binding capacity can vary from low to moderate
levels (Rmax between 100 and 800 RUs). The important factor in
this last case is that the analyte should saturate the surface during
contact time.

Fig. 2 Determination of theoretical Rmax

268 Badreddine Douzi

7. If your proteins are not stable in the running buffer (HBS-­EP),

a running buffer with a different composition, like phosphate
buffer saline (PBS), should be used.
8. After purification, it is important to characterize the multimer-
ization state of your proteins. This can be done using size
exclusion chromatography (SEC), dynamic light scattering
(DLS), or multiple angle light scattering coupled to SEC
(SEC-MALS). If your protein forms multimers, it is recom-
mended to use it as ligand.
9. If the proteins are already purified in the running buffer and
conserved at −80 °C, check their quality after thawing by con-
centration quantification. Compare the concentration of the
protein sample before freezing and after thawing 1 day before
the experiment to be sure that thawing the protein did not lead
to precipitation.
10. Be sure that the maintenance procedures are properly followed
before starting the experiments.
11. The aim of the pH scouting experiment is to test the precon-
centration of the ligand on the sensor surface at different pHs,
and so to determine the pH at which the ligand can be adsorbed
at highest concentration on the dextran matrix by electrostatic
interaction. Typically, the optimal pH for preconcentration
should be at 1 pH unit below the isoelectric point (pI) of your
protein. At low pH (3.5 < pH < 5.5), the carboxylated dextran
is negatively charged and the ligand is positively charged (pI >
6). The choice of the pH depends on the “real” pI of your
protein. A low pH may be not suitable for many proteins. To
avoid protein precipitation or denaturation, dilute the protein
before performing the experiment. pH scouting is performed
only in one flow cell. It is recommended that the flow cell
planned for immobilization be used. The BIAcore T200 con-
trol software contains an immobilization wizard method to
help users find the optimum pH for ligand ­ immobilizing.
Nevertheless, you can easily set up a manual run in which you
inject the ligand diluted at different pHs.
12. The use of 50 mM NaOH solution for regeneration is some-
times sufficient. Nevertheless, the injection time may not
be sufficient to dissociate the ligand from the sensor surface.
In this case, it is recommended to use multiple injections
with short contact times (e.g., three injections with 30 s of
contact time).
13. If the theoretical pI of the ligand is around 5, start with a pH
range between 3.5 and 4.5. If your ligand is an acidic protein
(pI < 4), you can use the covalent immobilization using thiol
groups or a noncovalent immobilization method.
 Surface Plasmon Resonance 269

14. Once the protein injection is performed, the choice of the

optimal immobilization pH is based on the pattern of the sen-
sograms. The ligand should bind rapidly to the sensor surface
and completely dissociate after the end of the injection. If this
condition is satisfied by all pHs used, use the highest pH
because it is the least offensive to your protein. For example, in
the case of the immobilization of TssE from the Type VI secre-
tion system from enteroaggregative E. coli, and based on the
pH scouting sensograms (Fig. 3), the optimum pH use for its
immobilization is pH 5.
15. If you use the immobilization wizard template from the Biacore
control software, do not mix the EDC and NHS solution. The
software dialog box will ask you to prepare an empty vial to
mix the two solutions.
16. The solutions should be prepared just before performing the
experiments.
17. The injection of the ligand on the sensor surface is an irrevers-
ible step. Be careful when you inject the ligand; do not immo-
bilize over the desired level. To minimize this risk, perform a
short injection (5–10 μL of your ligand to estimate the immo-
bilizing level). Once the relationship between the injection
time and the reached RU values is established, perform a sec-
ond injection to reach the final Ri. Ligand contact should be
completed within 15 min after activation of the surface with
EDC/NHS to ensure coupling.

3500

30000 pH 4
pH 4.5
25000
Response Units

20000
pH 5
(RUs)

15000

10000

5000
pH 5.5
0

-5000
-50 0 50 100 150 200 250 300 350 400

Time (sec)

Fig. 3 pH scouting of TssE protein from Type VI secretion system over CM5 sensor surface. The pH of each
solution tested is indicated at the top of the corresponding sensogram
270 Badreddine Douzi

18. The baseline stability is an indication of the “good” quality of

the ligand after immobilization. It is important to wait until
the baseline becomes stable and no decrease in RUs is observed.
19. The nonspecific binding of the analyte to the sensor surface is
a common problem faced by SPR users. In some cases, the
activation/deactivation of the reference cell is sufficient to
remove nonspecific binding. However, frequently ­activation/
deactivation does not help and immobilization of an irrelevant
protein on the sensor surface is recommended. A large number
of proteins can be used to this end: for example Thioredoxin,
maltose-binding protein (MBP), or a homemade protein not
related to the ligand.
20. This step is performed after ligand immobilization and baseline
stability. The choice of the analyte concentration depends on
the strength of the interaction. If you are testing a new interac-
tion, it is recommended to start with micromolar concentra-
tion (10–50 μM). If this leads to a high signal with a very slow
dissociation, decrease your analyte concentration to the nano-
molar range. In the case of antibody–antigen interactions, it is
known that they are tight and you can start with a nanomolar
concentration (100 nM). The function of your proteins can
help you to estimate the strength of the interaction. For exam-
ple, in the case of haemolysin-coregulated protein Hcp and the
tail-sheath component TssB, in the T6SS, Hcp hexamers are
able to pack and form a tail tube wrapped by the tail sheath
composed of two proteins, TssB and TssC [19]. The contrac-
tion of the tail sheath leads to the expulsion of the Hcp tail
tube to the extracellular milieu. Based on the dynamic of such
a complex, you can estimate a weak interaction between Hcp
and TssB. Consequently, we started with TssB as analyte in the
micromolar range.
21. The use of the wizard template method is recommended if the
regeneration buffer is known or described in the literature. If
not, manual runs with different regeneration buffers should be
carried out to optimize the regeneration step after analyte
binding.
22. When examining the binding of the analyte to the ligand, the
subtraction of the signal of the reference flow cell from that
of the flow cell containing the ligand should give a typical
sensogram of protein–protein interaction (Fig. 1). If the
interaction is weak (micromolar range) or transient, low RU
values are obtained following injection of the analyte (e.g.,
less than 20 RUs) and the sensograms have fast association
and dissociation phases. In this case, it is recommended to
perform a binding analysis with increasing concentrations of
the analyte.
 Surface Plasmon Resonance 271

By contrast, if the interaction is strong (nanomolar range),

higher RU values are obtained and the association and dissociation
phases are slow. In this case, after the regeneration step, try a lower
analyte concentration.
If no change in the RU value is observed, either the analyte did
not bind to the ligand or the ligand immobilized to the sensor
surface was inactivated. In this case, the immobilization of the
ligand by the noncovalent method (see Note 5) may help solve the
problem. Another possibility may be to immobilize the analyte (if
the protein is suitable for immobilization) and test the opposite
interaction by injecting the ligand.
23. Once the binding of the analyte on the ligand is confirmed, the
regeneration step can be started. The goal of this step is to
totally dissociate the analyte from the ligand binding sites
without affecting the activity of the immobilized ligand. To
this end, many regeneration solutions can be tested depending
on the type of interaction and the nature of the ligand. If the
interaction is reversible and fast dissociation is observed, wash-
ing with low-ionic-strength solutions (1 M NaCl, 2 M MgCl2)
or ethylene glycol (from 10 to 100%) for hydrophobic interac-
tions may accelerate the dissociation. In the case of high-­
affinity interactions, stronger solutions may be necessary
(low- or high-pH solutions or highly hydrophobic solutions).
If the ligand is an antibody, it is recommended to use a strong
acid solution (10 mM phosphoric acid).
24. If the decrease in the signal is below 30%, move to the next
regeneration solution.
–– If the amount of the decrease is more than 30% but below
90%, try to repeat the injection.
–– If this procedure is not sufficient to remove more than 90%
of the analyte, try a higher concentration of the regenera-
tion solution.
–– If the decrease in the signal is significant (more than 100%),
the regeneration solution is not suitable and may affect the
ligand activity. If so, try to inject the analyte at the same
concentration used in step 3 in Subheading 3.6.
–– If the same binding level is obtained, use the same regen-
eration solution at a lower concentration.
–– If the binding level is lower compared to the first analyte
injection, try another type of regeneration buffer.
–– If the regeneration fails, try to combine two solutions
among those giving more than 30% surface regeneration.
–– If the residual activity of the ligand is less than 90% follow-
ing regeneration, the flow cell is no longer suitable for
binding analysis. It is recommended that the ligand be
272 Badreddine Douzi

immobilized by amine coupling in a different flow cell or

that a noncovalent immobilization method be used.
–– If an efficient regeneration buffer is not found, immobilize
the ligand in a new flow cell. Kinetic measurement could
be undertaken using single cycle kinetics (SCK) with no
regeneration between injections. In an SCK experiment,
increased concentrations of the analyte are injected sequen-
tially in the same cycle.
25. Observation of the pattern of the sensogram resulting from
ligand immobilization and the analyte binding gives you infor-
mation on the strength of the interaction. If the association
and the dissociation are fast (1–2 min), it is difficult to estimate
the interaction rates. This suggests that the experiment will
allow only for the estimation of the binding affinity of the
complex. On the other hand, if the sensogram shows slow
association/dissociation phases (5 min for association and
10–60 min or more for dissociation), it will be possible to esti-
mate the kinetics of the interaction.
26. The affinity of molecule A to molecule B is described by the
dissociation constant KD (Fig. 4). KD is expressed in molars
(M). The KD can be calculated using SPR data by the equilib-
rium binding analysis. The steady-state binding level is related
to the concentration of the analyte (Fig. 5).
27. To perform the experiment, the analyte concentrations must
be varied from 0.1 × KD to 10–100 × KD.
28. The association and the dissociation times are determined
based on the sensogram obtained after the analyte-binding
analysis. The dissociation time should be sufficient to go back
to the original level of the baseline. If it is time consuming, a
regeneration step can be added with a soft regeneration solu-
tion to avoid ligand inactivation.

Fig. 4 Representation of different binding parameters of an interaction between

two molecules A and B
 Surface Plasmon Resonance 273

Fig. 5 Determination of KD using response units at equilibrium and concentra-

tions of analyte

Fig. 6 Determination of rate of association kon of complex AB

Fig. 7 Determination of rate of dissociation koff of complex AB

29. The volume of the analyte dilution solution depends on the

contact times and the flow rate. These parameters are estimated
during the analyte-binding analysis test.
30. When making twofold dilutions, avoid forming bubbles when
mixing your analyte on the buffer. If bubbles are present, cen-
trifuge your sample and put it in a new vial.
31. The sensograms are adjusted to zero at the start of the analyte
injection on both the response and the time axis.
32. Adjust the region used for the calculation of Req by selecting
the setting bottom.
33. Using SPR you can determine the association and the dissocia-
tion rate constants. The first phase in the sensogram, corre-
sponding to the injection of the analyte over the ligand and the
binding, allows for the determination of the rate of formation
of the complex kon according to the equation described in
Fig. 6. The unit of kon is M−1 s−1.
The dissociation phase, when the analyte is removed from the
flow cell till a zero concentration is reached, makes it possible to
calculate the rate of dissociation koff using the equation presented
in Fig. 7. The unit of koff is s−1.
274 Badreddine Douzi

34. Mass transport is one of the more widely known limitations for
the determination of the association rate constant. Mass trans-
port takes place when the rate of analyte binding is higher than
the rate of diffusion of the analyte. By contrast, a limitation on
the determination of the dissociation rate constant comes from
the rebinding of the analyte to the ligand owing to the ineffi-
cient effusion of the free analyte from the ligand surface. These
issues can be avoided by the immobilization of a low amount
of ligand (50–150 RUs) and performing the binding analysis
at a high flow rate (30–100 μL min−1).
35. To perform the experiment, the analyte concentrations must
be varied from 0.1 × KD to 10 × KD.
36. The best way to analyze the kinetic data is to use the Biacore
T200 evaluation software. Kinetics data are interpreted in
terms of an interaction model, and the kinetic constants
obtained from the SPR data analysis are apparent constants,
which are valid in the context of the binding model adopted. A
simple model used to determine the rates of an interaction is
the Langmuir model, in which it is assumed that molecule A
binds B with a 1:1 stoichiometry and the binding events are
independent and equivalent.

Acknowledgments

I am grateful to Dr. Romé Voulhoux and Dr. Mariella Tegoni for

their constant training, encouragement, and support and to Dr.
Sawsan Amara and John Young for their careful reading of the
manuscript.

References
1. Barison N, Lambers J, Hurwitz R, Kolbe M role in the assembly of a quaternary complex
(2012) Interaction of MxiG with the cytosolic within the T2SS pseudopilus. J Biol Chem
complex of the type III secretion system 284:34580–34589
controls
­ Shigella virulence. FASEB 5. Douzi B, Spinelli S, Blangy S, Roussel A,
J 26:1717–1726 Durand E, Brunet YR, Cascales E, Cambillau C
2. Benabdelhak H, Kiontke S, Horn C, Ernst R, (2014) Crystal structure and self-interaction of
Blight MA, Holland IB, Schmitt L (2003) A the type VI secretion tail-tube protein from
specific interaction between the NBD of the enteroaggregative Escherichia coli. PLoS One
ABC-transporter HlyB and a C-terminal frag- 9:e86918
ment of its transport substrate haemolysin A. J 6. Felisberto-Rodrigues C, Durand E, Aschtgen
Mol Biol 327:1169–1179 MS, Blangy S, Ortiz-Lombardia M, Douzi B,
3. Douzi B, Ball G, Cambillau C, Tegoni M, Cambillau C, Cascales E (2011) Towards a
Voulhoux R (2011) Deciphering the Xcp structural comprehension of bacterial type VI
Pseudomonas aeruginosa type II secretion secretion systems: characterization of the TssJ-­
machinery through multiple interactions with TssM complex of an Escherichia coli pathovar.
substrates. J Biol Chem 286:40792–40801 PLoS Pathog 7:e1002386
4. Douzi B, Durand E, Bernard C, Alphonse S, 7. Girard V, Cote JP, Charbonneau ME, Campos
Cambillau C, Filloux A, Tegoni M, Voulhoux R M, Berthiaume F, Hancock MA, Siddiqui N,
(2009) The XcpV/GspI pseudopilin has a central Mourez M (2010) Conformation change in a
 Surface Plasmon Resonance 275

self-recognizing autotransporter modulates 13. Ohlson S, Strandh M, Nilshans H (1997)

bacterial cell–cell interaction. J Biol Chem Detection and characterization of weak affinity
285:10616–10626 antibody antigen recognition with biomolecu-
8. Schroder G, Lanka E (2003) TraG-like proteins lar interaction analysis. J Mol Recognit 10:
of type IV secretion systems: functional dissec- 135–138
tion of the multiple activities of TraG (RP4) and 14. Peess C, von Proff L, Goller S, Andersson K,
TrwB (R388). J Bacteriol 185:4371–4381 Gerg M, Malmqvist M, Bossenmaier B, Schraml
9. Swietnicki W, O’Brien S, Holman K, Cherry S, M (2015) Deciphering the stepwise binding
Brueggemann E, Tropea JE, Hines HB, Waugh mode of HRG1beta to HER3 by surface plas-
DS, Ulrich RG (2004) Novel protein-protein mon resonance and interaction map. PLoS One
interactions of the Yersinia pestis type III secre- 10:e0116870
tion system elucidated with a matrix analysis by 15. Fischer MJ (2010) Amine coupling through
surface plasmon resonance and mass spectrom- EDC/NHS: a practical approach. Methods
etry. J Biol Chem 279:38693–38700 Mol Biol 627:55–73
10. Zoued A, Durand E, Bebeacua C, Brunet YR, 16. Hutsell SQ, Kimple RJ, Siderovski DP, Willard
Douzi B, Cambillau C, Cascales E, Journet L FS, Kimple AJ (2010) High-affinity immobili-
(2013) TssK is a trimeric cytoplasmic protein zation of proteins using biotin- and GST-based
interacting with components of both phage-­ coupling strategies. Methods Mol Biol
like and membrane anchoring complexes of the 627:75–90
type VI secretion system. J Biol Chem 17. Khan F, He M, Taussig MJ (2006) Double-­
288:27031–27041 hexahistidine tag with high-affinity binding for
11. Zoued A, Durand E, Brunet YR, Spinelli S, protein immobilization, purification, and
Douzi B, Guzzo M, Flaugnatti N, Legrand P, detection on ni-nitrilotriacetic acid surfaces.
Journet L, Fronzes R, Mignot T, Cambillau C, Anal Chem 78:3072–3079
Cascales E (2016) Priming and polymerization 18. Della Pia EA, Martinez KL (2015) Single
of a bacterial contractile tail structure. Nature domain antibodies as a powerful tool for high
531(7592):59–63 quality surface plasmon resonance studies.
12. Pineau C, Guschinskaya N, Robert X, Gouet P, PLoS One 10:e0124303
Ballut L, Shevchik VE (2014) Substrate recog- 19. Cianfanelli FR, Monlezun L, Coulthurst SJ
nition by the bacterial type II secretion system: (2015) Aim, load, fire: the type VI secretion
more than a simple interaction. Mol Microbiol system, a bacterial nanoweapon. Trends
94:126–140 Microbiol 24(1):51–62
 Chapter 22

Assessing Energy-Dependent Protein

Conformational Changes in the TonB System
Ray A. Larsen

Abstract
Changes in conformation can alter a protein’s vulnerability to proteolysis. Thus, in vivo differential
proteinase sensitivity provides a means for identifying conformational changes that mark discrete states in
the activity cycle of a protein. The ability to detect a specific conformational state allows for experiments
to address specific protein–protein interactions and other physiological components that potentially
contribute to the function of the protein. This chapter presents the application of this technique to the
TonB-­dependent energy transduction system of Gram-negative bacteria, a strategy that has refined our
understanding of how the TonB protein is coupled to the ion electrochemical gradient of the cytoplasmic
membrane.

Key words Protein conformation, Ion electrochemical gradient, Proteinase sensitivity, TonB,
Spheroplast, Energy transduction

1  Introduction

Many proteins contain regions that are intrinsically disordered,

unable to achieve stable structures on their own. Rather than a
detriment, this conformational flexibility is an essential feature in
many protein-mediated processes, particularly those involving
protein–protein interactions [1]. A feature common to most pro-
teins with intrinsically disordered regions is an enhanced sensitivity
to proteolysis [2]. One such protein is the Gram-negative bacterial
cell envelope protein TonB [3].
Anchored to the cytoplasmic membrane by a single N-terminal
transmembrane domain, the bulk of TonB occurs in the periplasmic
space, which TonB spans to interact with outer membrane trans-
porters. TonB functions as an energy transducer, coupling the ion
electrochemical gradient of the cytoplasmic membrane to drive
active transport processes at the outer membrane (reviewed in
[4]). In this capacity, TonB appears to cycle through several ­distinct
conformations, as evidenced in vivo by differential sensitivity to

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_22, © Springer Science+Business Media LLC 2017

277
278 Ray A. Larsen

both intrinsic and extrinsic proteinases [5]. We have proposed that

these conformational changes in TonB involve regions of intrinsic
disorder [6]. Interactions with known proteins appear to drive at
least some of these conformational changes, and in doing so support
the energy transduction function of TonB [7, 8]. These and other
studies suggest a current working model of the TonB energy trans-
duction cycle, wherein the conformational flexibility of intrinsically
disordered regions is a key feature [9].
It had long been recognized that intrinsic proteolysis was
important in TonB and that this sensitivity correlated with apparent
interactions between TonB and other proteins [10]. The discovery
that TonB was differentially sensitive to exogenous proteinases,
and the correlation of this sensitivity to the energy state of the
cytoplasmic membrane was accidental. We were characterizing a
set of mutations in the signal anchor of TonB that rendered TonB
inactive [11]. Because these mutations occurred in a region of
TonB involved in the trafficking of TonB to the cytoplasmic mem-
brane, a trivial explanation for the TonB negative phenotype could
have been that the protein was simply not delivered to the cyto-
plasmic membrane. To determine whether the TonB derivatives
were properly oriented in the cytoplasmic membrane, we modified
a method initially developed to examine the delivery of maltose-
binding protein to the periplasmic space [12]. Here, the outer
membrane is permeabilized and the peptidoglycan layer disrupted,
converting the cells into spheroplasts. This allows large molecules,
in this case proteinase K, access to periplasmically exposed pro-
teins. Using the same general strategy, we found that both the
wild-type and the mutant TonB derivatives were readily degraded
by proteinase K in spheroplasts, but not in intact cells, indicating
that the mutant derivatives were properly trafficked to and oriented
in the cytoplasmic membrane.
In the experiments just described we included a control where
the spheroplasts were osmotically lysed to allow proteinase K access
to the cytoplasmic contents. Thus, if the mutant derivatives were
not susceptible to proteolysis in the intact spheroplasts, we could
confirm that the problem involved trafficking (and not simple resis-
tance to proteinase K). To our great surprise, we found that in lysed
spheroplasts, wild-type TonB was not fully degraded but rather
reduced to a smaller fragment. Further studies suggested that this
fragment represented the ~130 residue N-terminal domain of
TonB present in the cytoplasmic membrane. Interestingly, the
ability to form this proteinase K-resistant fragment was dependent
upon the presence of the cytoplasmic membrane protein ExbB,
with which TonB was known to associate. Also of interest, the
mutant derivatives did not form the proteinase K-resistant fragment
in either the presence or absence of ExbB, suggesting the mutations
disrupted that particular interaction. Together, these data suggested
that ExbB influences the conformation of TonB [11].
 Assessing Protein Conformational Changes 279

TonB+, ExbB+ TonB+, ExbB- TonB∆V17, ExbB+

WC S S-CCCP WC S S-CCCP WC S S-CCCP
Prot K - + - + - + - + - + - + - + - + - +
35.0 kDa
25.5 kDa

Fig. 1 Identification of an ion electrochemical responsive TonB conformation by differential proteinase K sus-
ceptibility. Western blots of whole-cell (WC), spheroplast (S), and CCCP-treated spheroplast (S-CCCP) samples,
treated with (+) or without (−) proteinase K (“Prot K”), are shown. The sample set on the left was generated
from the wild-type strain W3110 and is wild type for TonB, ExbB, and ExbD. The sample set in the middle was
generated from the W3110 derivative and is wild type for TonB but does not express ExbB and ExbD. The
sample set on the right was generated from a W3110 derivative bearing a tonB allele that encodes an inactive
TonB with a deletion of the valine residue from position 17 and is wild type for ExbB and ExbD. Samples were
developed on 11% SDS-polyacrylamide gels, transferred to polyvinylidene fluoride membranes and probed
with the monoclonal antibody 4H4 (specific for TonB residues 79–84). The positions and apparent molecular
masses of intact TonB and the proteinase K-resistant fragment are indicated on the left of the panel. Reproduced
in part from Larsen et al. [5] with permission from John Wiley & Sons

While it seems obvious in retrospect, it was initially unclear to us

how lysing the spheroplast could alter TonB conformation.
Eventually it occurred to us that spheroplast lysis collapses the
cytoplasmic membrane ion electrochemical gradient. We tested this
hypothesis by treating spheroplasts with protonophores and deter-
mined that indeed the proteinase K-resistant fragment represented a
conformation that energized TonB can become trapped in. Similar
to the results obtained with lysed spheroplasts, only functional TonB
in the presence ExbB could achieve this conformation and only
when the ion electrochemical gradient was collapsed (Fig. 1). This
was the first direct evidence that TonB conformation was coupled to
the ion electrochemical potential of the cytoplasmic membrane [5].
Soon thereafter it was demonstrated that the ability of the TonB
analog TolA to interact with an outer membrane lipoprotein was
dependent upon the cytoplasmic membrane proton gradient [13].
Subsequently, our strategy was used to demonstrate that TolA
undergoes conformational changes in response to the ion electro-
chemical gradient similar to that occurring in TonB [14].
This general strategy provides a means to examine conforma-
tional changes in periplasmically exposed proteins in response to
varied stimuli as well as through protein–protein interactions. We
have focused on the ion electrochemical gradient, but, depending
upon the protein and the system, the influence of a wide range of
physical and chemical conditions on protein conformation and,
consequently, their influence on function could be addressed.
Likewise, we have relied upon proteinase K as a probe; a myriad of
other proteases with a diversity of mechanisms and specificities
have potential as informative probes.
Ultimately, the readout of this assay requires visualization of
the protein of interest. Here, samples are developed on sodium
280 Ray A. Larsen

dodecyl sulfate (SDS) 11% polyacrylamide gels, transferred by

electrophoretic elution to polyvinylidine difluoride membranes,
probed with a monoclonal antibody specific for the protein, and
visualized by enhanced chemiluminescence. Description of this
methodology is beyond the scope of this chapter; it is presumed
that laboratories interested in adapting a strategy that uses differ-
ential sensitivity to proteolysis to detect conformational changes
will have a similar visualization technique optimized for their
protein of interest (see Note 1).

2  Materials

Prepare solutions and culture media with double-distilled water

and reagent-grade materials. The procedures described here use
the laboratory-adapted Escherichia coli K-12 strain W3110 [15]
and derivatives thereof (see Note 2).

2.1  Spheroplast 1. Luria–Bertani (LB) agar medium ([16]) (see Note 3).
Production 2. Supplemented M9 medium: Prepare a 10× stock solution con-
taining 60 g Na2HPO4, 30 g KH2PO4, 5 g NaCl, and 10 g
NH4Cl per liter. Sterilize the 10× solution by autoclaving and
store at room temperature. To make the culture medium, add
10 mL of the 10× solution to 85.6 mL sterile distilled water,
then supplement this mixture with 2 mL of a filter-sterilized
20% (w/v) carbohydrate source (see Note 4), 1 mL 20% (w/v)
vitamin-free casamino acids (autoclaved), 1 mL filter-sterilized
0.4% (w/v) tryptophan, 200 μL filter-sterilized 0.2% (w/v)
thiamin (see Note 5), 100 μL 1 M MgSO4, 100 μL 0.5 M
CaCl2, and 50 μL 0.1 (w/v) FeCl3·6H2O (see Notes 6 and 7).
3. Isopropyl β-d-1-thiogalactopyranoside (IPTG): 1 M (see
Note 8).
4. Buffer 1: 200 mM Tris–acetate, pH 8.2, 500 mM sucrose,
0.5 mM ethylenediaminetetraacetic acid (EDTA). Keep solution
at 4 °C.
5. Buffer 2: 200 mM Tris–acetate, pH 8.2. Keep solution at
4 °C.
6. Buffer 3: 200 mM Tris–acetate, pH 8.2, 250 mM sucrose,
20 mM MgSO4 (see Note 9). Keep solution at 4 °C.
7. Lysozyme: Stock solution of 2 mg lysozyme in 1 mL sterile
water (see Note 10). Keep solution at 4 °C.
8. Incubator.
9. Spectrophotometer (see Note 11).
10. Microcentrifuge and microfuge tubes.
11. Micropipette or a pulled Pasteur pipet and a rubber bulb.
 Assessing Protein Conformational Changes 281

2.2  Proteinase 1. Proteinase K: Stock solution of 2 mg proteinase K in 1 mL

Accessibility sterile water (see Note 10).
2. Phenylmethylsulfonyl fluoride (PMSF): Stock solution of
100 mM in dimethyl sulfoxide (DMSO) (see Note 12).
3. Carbonylcyanide m-chlorophenylhydrazone (CCCP): Stock
solution of 15 mM in DMSO (see Note 12).
4. Trichloroacetic acid (TCA): Stock solution of 10% (w/v)
(see Note 13).
5. 100 mM Tris–HCl, pH 6.8.
6. 1× Laemmli sample buffer: 63 mM Tris–HCl, pH 6.8, 2%
(w/v) sodium dodecyl sulfate, 10% (v/v) glycerol, 0.1% (v/v)
β-mercaptoethanol, and 0.0005% (w/v) bromophenol blue.

3  Methods

Bacterial strains are grown at 37 °C with shaking to provide aeration;

all subsequent steps are performed at 4 °C, except for β-galactosidase
assays and solubilization of samples for electrophoresis.

3.1  Production 1. Select an isolated colony from an LB agar stock plate (see Note 3)
of Spheroplasts and inoculate 5 mL of supplemented M9 medium in a stan-
dard culture tube. Incubate at 37 °C overnight with shaking
(~200 rpm).
2. Inoculate a fresh culture tube containing 5 mL of supple-
mented M9 medium with 25 μL of the overnight culture and
incubate as described earlier, with periodic spectrophotomet-
ric monitoring of growth (see Notes 11 and 14).
3. At or near an A550 of 0.2, add 5 μL 1 M IPTG (1 mM final
concentration) to each culture to induce β-galactosidase expres-
sion for subsequent measurement of spheroplast integrity.
4. At an A550 of 0.4, harvest cells as six 500 μL samples into pre-
chilled 1.5 mL Eppendorf microfuge tubes, and centrifuge at
~20,000 × g for 5 min at 4 °C (see Note 15).
5. Remove supernatant by manual aspiration using either a micro-
pipette with a disposable tip or a pulled Pasteur pipet and a
rubber bulb. Discard the supernatant and place tubes on ice.
6. Four of the six tubes are used to generate spheroplasts. To each
of these tubes, add 250 μL chilled (4 °C) buffer 1. Suspend
pellets by gently pipetting up and down with a micropipette.
When fully suspended, add 20 μL of the 2 mg/mL lysozyme
solution. Mix by gently flicking each tube, then immediately
add 250 μL chilled (4 °C) buffer 2. Mix gently as above, then
incubate on ice for 5 min (see Note 16). Proceed to step 7
while incubating these tubes, then proceed to step 8.
282 Ray A. Larsen

7. The two remaining pellets are used as whole-cell controls.

To each of these add 500 μL chilled (4 °C) buffer 3. Suspend
pellets by gently pipetting up and down with a micropipette,
then store on ice.
8. After 5 min on ice, add 10 μL 1 M MgSO4 to each of the four
tubes used to produce spheroplasts in step 6 (see Note 17).
Mix gently.
9. Centrifuge all six tubes (prepared in steps 7 and 8) at
~20,000 × g for 5 min at 4 °C. Remove the supernatant by
manual aspiration and discard, then add 500 μL buffer 3 to
each tube and gently suspend each pellet using a micropipette
(see Note 18).

3.2  Proteinase Preparations are divided into pairs for analysis. The whole-cell con-
Accessibility trol pair was produced separately at the same time the four tubes of
spheroplasts were produced. The spheroplast tubes are now divided
into two groups: The first is used to evaluate baseline proteinase
susceptibility, and the second pair of tubes is used to examine the
effect of collapsing the ion electrochemical gradient of the cyto-
plasmic (i.e., spheroplast) membrane on the conformation of
TonB. Take care to maintain all tubes at 4 °C and to handle tubes
gently through the following steps.
1. Add the protonophore CCCP to two spheroplast tubes as
3.4 μL of a 15 mM solution of CCCP in DMSO, for a final
concentration of ~50 μM (see Note 19). These two tubes
become the second pair. The two remaining spheroplast
tubes (pair 1) and the two whole-cell tubes receive 3.4 μL of
DMSO alone.
2. Add 6.3 μL of the 2.0 mg/mL solution of proteinase K to one
tube of the whole-cell pair and one tube of each spheroplast
pair (for a final concentration of 25 μg/mL). Add 6.3 μL
water to the other tube in each pair. Incubate all tubes for
15 min at 4 °C, with each tube gently flicked every several
minutes to maintain cells and spheroplasts in suspension (see
Note 20).
3. Add 5 μL 100 mM PMSF to each tube (for a final concentra-
tion of 1 mM), mix gently, and incubate for 2 min at 4 °C to
inactivate proteinase K.
4. Optional: Remove two 100 μL aliquots from each tube for the
determination of β-galactosidase activity. Store on ice while
step 5 is performed, then proceed to step 6.
5. Add 500 μL (300 μL if step 4 performed) of cold (4 °C) 10%
(w/v) TCA to both whole-cell tubes and to each tube in sphero-
plast pairs 1 and 2. Mix by inverting several times, then incubate
the samples on ice for 15 min to precipitate total protein
(see Note 21). Proceed to step 6 while incubating these tubes.
 Assessing Protein Conformational Changes 283

6. Optional: Centrifuge one of the two aliquots taken from each

tube at ~20,000 × g for 5 min at 4 °C. Transfer the supernatant
from each to a fresh tube. These samples can then be assayed
for the presence of β-galactosidase to determine spheroplast
membrane integrity (see Note 22).
7. Centrifuge the six TCA precipitation tubes at ~20,000 × g for
5 min at 4 °C (see Note 23). Remove supernatant by hand
aspiration, then gently rinse the tube with 200 μL 100 mM
Tris–HCl, pH 6.8. Add 50 μL 1× Laemmli sample buffer,
then incubate at 96 °C for 5 min to denature sample for
subsequent electrophoretic analysis.

4  Notes

1. Our system uses a monoclonal antibody [17] that recognizes

a linear epitope present in the proteinase K-resistant fragment
of TonB. Other fragments might occur that would be missed
using this antibody as a probe. In our case, no additional or
alternative proteinase K-resistant fragments were identified
using several other monoclonal antibodies specific for epitopes
located elsewhere in TonB. In the absence of a good mono-
specific antibody, expression of the protein of interest modified
to include an epitope tag or a small binding domain for a
detectable ligand might prove useful.
2. The strain W3110 is a commonly used so-called wild-type E. coli
K-12 strain, with a complete genome sequence available in
Genbank (NC_007779.1). The methods described in this
chapter should work well with other wild-type K-12 strains
and most of their mutant derivatives. Successful application to
non-laboratory-adapted strains with more robust outer mem-
branes may require some modifications, particularly in creating
spheroplasts.
3. Several different formulations of LB agar are available; there is
nothing particularly special about the Miller formulation, but
it is the one we first used long ago, and so we continue to use
it for the sake of consistency. Other formulations are not likely
to alter the results of the methods described here, nor should
any other rich medium commonly used to grow laboratory
E. coli stains.
4. Normally glucose is used; however, many of our studies involve
products of arabinose-regulated genes that are subject to
catabolite repression; therefore, we substitute glycerol as the
major carbon source. The solution is filter-sterilized to avoid
caramelization.
5. Thiamin and tryptophan are both heat-labile, so the stock
solution should be sterilized by filtration and stored at 4 °C.
284 Ray A. Larsen

Tryptophan is also light sensitive and should be stored in an

opaque container.
6. Autoclaving of phosphates with ionic metals (such as magnesium
and calcium) can result in precipitation. The metal cations will
preferentially precipitate with only one ionic form of phos-
phate, thereby altering the pH of the medium.
7. For strains carrying plasmids, the medium is supplemented
with the appropriate amount of antibiotic required to provide
selection for a given plasmid. Additional supplementation
with arabinose is made for cells carrying plasmids with
arabinose-­regulated genes of interest; the working concentra-
tion for a given strain is that which has been predetermined to
provide for the expression of the gene product at a normal
physiologic level.
8. IPTG stock solution is made at 1 M in water and should be
stored at −20 °C.
9. Buffers 1–3 are made fresh as 10 mL volumes from stocks of
1.0 M Tris–acetate, pH 8.2, 2.0 M sucrose, 1.0 M MgSO4,
and 500 mM EDTA. For buffer 1: to 2.0 mL 1.0 M Tris–ace-
tate, pH 8.2, add 2.5 mL 2.0 M sucrose, 10 μL 500 mM
EDTA, and add water to 10 mL final volume. For buffer 2: to
2.0 mL 1.0 M Tris–acetate, pH 8.2, add water to 10 mL final
volume. For buffer 3: to 2.0 mL 1.0 M Tris–acetate, pH 8.2,
add 1.25 mL 2.0 M sucrose, 200 μL 1.0 M MgSO4, and add
water to 10 mL final volume.
10. Store enzyme solutions at −20 °C.
11. Cultures are monitored using a Spectronic 20 spectrophotom-
eter fitted with an adaptor to receive our culture tubes, which
have an internal diameter of 1.5 cm. Culture tubes must con-
tain at least 5 mL of medium to insure that the light path goes
through the medium below the meniscus. Absorbance of light
at 550 nm is measured. Because the path length of light is
1.5 cm, we refer to our measure as an “A550,” as the standard
term optical density (OD550) is specifically defined for a 1.0 cm
path length.
12. PMSF is unstable in water, and CCCP has a low solubility in
water, hence the use of DMSO as a carrier. Both compounds
are toxic and should be handled with caution; a mask should
be worn to protect against inhalation when weighing the solid,
and gloves should be worn at all times, as the carrier solvent
(DMSO) facilitates absorption through the skin.
13. 10% (w/v) TCA is prepared by dilution in water from a 100%
(w/v) stock solution, made by adding 227 mL water to a new
reagent bottle containing 500 g TCA.
14. For W3110 grown under the conditions described, cells

experience an initial lag phase of ~45 min, transitioning into
 Assessing Protein Conformational Changes 285

exponential growth with a generation time of 45–50 min.

After four to five generations (about 4–4.5 h) cells reach an
A550 of 0.4, at which point they will be harvested. Grown
under these conditions, an A550 of 0.4 corresponds to
~1.0 × 108 colony forming units (CFU)/mL.
15. It is challenging to catch cells at exactly at A550 = 0.4, and when
dealing with more than one independently inoculated tube, it is
unlikely that they will all reach A550 = 0.4 at the same time.
What is important is to harvest an equivalent number of cells for
each culture. One approach is to allow each culture to grow past
A550 = 0.4 and then dilute back with fresh medium, good in
theory but logistically complex in practice. A more practical
solution is to vary the volume harvested relative to the A550
value. To determine how much to sample, we consider this in
terms of A550 equivalents (A550eq), where A550eq = A550 × vol-
ume (in milliliters). Here, the A550eq = 0.4 × 0.5 mL = 0.2
A550eq. Rearranged, A550eq/A550 = volume (mL) harvested.
For example, for a culture at A550 = 0.43: 0.2/0.43 = 0.465;
harvest 465 μL. For a second culture at A550 = 0.36:
0.2/0.36 = 0.556; harvest 556 μL. This approach does intro-
duce some error; the A550 value is a measure of light absorbance,
and as culture densities increase, the probability of any single
bacterium’s being in the shadow of another bacterium
increases. Thus, the correlation of the measured A550 value
and the CFU is not linear. However, working in the range of
A550 = 0.35–0.45, the amount of error introduced is lower
than the range of error in determining CFU.
16. The 500 mM sucrose in buffer 1 is hypertonic, creating turgor
pressure on the outer membrane. The EDTA chelates divalent
cations that stabilize the anionic inner core of lipopolysaccha-
ride. The addition of buffer 2 renders the solution relatively
isotonic. This rapid shift of pressure on the cation-depleted
outer membrane permeabilizes the barrier, allowing lysozyme
access to the periplasmic side, where it then catalyzes degradation
of the peptidoglycan layer.
17. In the absence of peptidoglycan the cytoplasmic membrane
becomes very fragile (which is why we are keeping everything
cold and treating the spheroplasts gently). The addition of
magnesium cations partially compensates for the absence of
peptidoglycan, stabilizing the cytoplasmic membrane of the
spheroplasts by ionic interactions with anionic phosphoryl
lipid head groups.
18. Cells are now ready for the in vivo assay of differential protein-
ase K susceptibility. The centrifugation and suspension of the
two whole-cell controls seems superfluous because they were
already in suspension in the final buffer, but they are included
in this step to minimize any handling-based differences between
samples.
286 Ray A. Larsen

19. We have also used the protonophore dinitrophenol (DNP),

with similar results. However, the effective working concentra-
tion of DNP is 10 mM, vs. 50 μM for CCCP. We therefore
chose not to use DNP because such higher concentrations
might allow other, unidentified physiological perturbations.
20. The amount of proteinase K used corresponds to roughly one
unit per reaction. Given the small sample size and length of
incubation, this seems excessive; however, at 4 °C the reaction
is occurring at well below the optimal of 37 °C for this enzyme.
Initial experiments using several different concentrations of
proteinase should be performed to optimize the concentration
for the system and the protein of interest.
21. Precipitation with TCA is routinely used to denature proteins,
including proteinases, to protect samples as they are prepared
for electrophoresis.
22. The enzyme β-galactosidase is a readily assayed cytoplasmic
protein, providing a means to evaluate spheroplast membrane
integrity. We determine β-galactosidase activity for 100 μL of
supernatant and uncentrifuged samples. The percentage of
lysed spheroplasts is calculated as = 100 × (supernatant)/
(uncentrifuged). We routinely find that the amount of
β-galactosidase activity in the supernatant samples is less than
15% of that found in the uncentrifuged samples. Because some
spheroplast lysis occurs with centrifugation, the percentage of
spheroplasts damaged during the proteolysis phase of the
experiment is less. We measure β-galactosidase activity using
the assay described by Miller [16]. This is a classic, widely used
assay whose description is beyond the scope of this chapter.
23. The pellet formed in a TCA precipitation is not always tightly
formed and in a fixed-angle rotor will form as a streak along
the upward-facing side of the tube. It is important to note
tube orientation in the centrifuge so that one can avoid inad-
vertent loss of sample when aspirating the supernatant.

References
1. Oldfield CJ, Dunker AK (2014) Intrinsically C-terminal domain of TonB and interaction
disordered proteins and intrinsically disordered studies with TonB box peptides. J Mol Biol
protein regions. Annu Rev Biochem 83: 345:1185–1197
553–584 4. Postle K, Larsen RA (2007) TonB dependent
2. Johnson DE, Xue B, Sickmeier MD, Meng J, energy transduction between outer and cyto-
Cortese MS, Oldfield CJ, Gall TL, Dunker AK, plasmic membranes. Biometals 20:453–465
Uversky VN (2012) High-throughput charac- 5. Larsen RA, Thomas MG, Postle K (1999)
terization of intrinsic disorder in proteins from Protonmotive force, ExbB and ligand-bound
the protein structure initiative. J Struct Biol FepA drive conformational changes in
180:201–215 TonB. Mol Microbiol 31:1809–1824
3. Peacock RS, Weijie AM, Howard SP, Price FD, 6. Larsen RA, Deckert G, Kastead K, Devanathan
Vogel HJ (2005) The solution structure of the S, Keller KL, Postle K (2007) His20 provides
 Assessing Protein Conformational Changes 287

the sole functionally significant side chain in 12. Randall LL, Hardy SJS (1986) Correlation of
the essential TonB transmembrane domain. competence for export with lack of tertiary
J Bacteriol 189:2825–2833 structure of the mature species: a study in vivo
7. Ollis AA, Postle K (2012) Identification of of maltose-binding protein in E. coli. Cell
functionally important TonB-ExbD periplas- 46:921–928
mic domain interactions in vivo. J Bacteriol 13. Cascales E, Gavioli M, Sturgis JN, Lloubés R
194:3078–3087 (2000) Proton motive force drives the interac-
8. Ollis AA, Kumar A, Postle K (2012) The ExbD tion of the inner membrane TolA and outer
periplasmic domain contains distinct functional membrane pal proteins in Escherichia coli. Mol
regions for two stages in TonB energization. Microbiol 38:904–915
J Bacteriol 194:3069–3077 14. Germon P, Ray MC, Vianney A, Lazzaroni JC
9. Gresock MG, Kastead KA, Postle K (2015) (2001) Energy-dependent conformational
From homodimer to heterodimer and back: change in the TolA protein of Escherichia coli
elucidating the TonB energy transduction involves its N-terminal domain, TolQ, and
cycle. J Bacteriol 197:3433–3445 TolR. J Bacteriol 183:4110–41104
10. Fischer E, Günter K, Braun V (1989) 15. Hill CW, Harnish BW (1981) Inversions
Involvement of ExbB and TonB in transport between ribosomal RNA genes of Escherichia
across the outer membrane of Escherichia coli: coli. Proc Natl Acad Sci U S A 78:7069–7072
phenotypic complementation of exbB mutants 16. Miller JH (1972) Experiments in molecular
by overexpressed tonB and physical stabiliza- genetics. Cold Spring Harbor Press, Cold
tion of TonB by ExbB. J Bacteriol 171: Spring Harbor
5127–5134 17. Larsen RA, Myers PS, Skare JT, Seachord CL,
11. Larsen RA, Thomas MG, Wood GE, Postle K Darveau RP, Postle K (1996) Identification of
(1994) Partial suppression of an Escherichia coli TonB homologs in the family Enterobacteriaceae
TonB transmembrane domain mutation and evidence for conservation of TonB-­
(∆V17) by a missense mutation in ExbB. Mol dependent energy transduction complexes.
Microbiol 13:627–640 J Bacteriol 178:1363–1373
 Chapter 23

Defining Assembly Pathways by Fluorescence Microscopy

Abdelrahim Zoued and Andreas Diepold

Abstract
Bacterial secretion systems are among the largest protein complexes in prokaryotes and display remarkably
complex architectures. Their assembly often follows clearly defined pathways. Deciphering these pathways
not only reveals how bacteria accomplish building these large functional complexes but can provide crucial
information on the interactions and subcomplexes within secretion systems, their distribution within bac-
teria, and even functional insights. The emergence of fluorescent proteins has provided a new powerful
tool for biological imaging, and the use of fluorescently labeled components presents an interesting
method to accurately define the biogenesis of macromolecular complexes. Here we describe the use of this
method to decipher the assembly pathway of bacterial secretion systems.

Key words Fluorescence microscopy, Biogenesis, Secretion systems, Fluorescently labeled proteins,
Macromolecular complexes, Epistasis experiments, Subcellular localization

1  Introduction

Bacterial secretion systems are macromolecular machines that

mediate the transport of proteins between bacteria or from bacte-
ria to eukaryotic cells [1, 2]. These complexes incorporate one or
multiple copies of a large number of different proteins that are
recruited in a hierarchical order. Important insights into the assem-
bly of secretion systems were obtained by the purification and visu-
alization of assembly intermediates, either in strains lacking certain
components of the system or upon overexpression of defined com-
ponents [3–10]. However, the often low number of stable inter-
mediates and the difficulties in obtaining and visualizing these have
limited the utilization of this approach. Increasingly, the assembly
of secretion systems is therefore deciphered based on the localiza-
tion of fluorescently labeled subunits. Variations of this approach
have been applied for a variety of secretion systems including the
Tat system [11, 12], the type II secretion system (T2SS) [13, 14],
the T3SS [15, 16], the T4SS [17], or, more recently, the T6SS
[18]. The rationale is that a fluorescently labeled component forms

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_23, © Springer Science+Business Media LLC 2017

289
290 Abdelrahim Zoued and Andreas Diepold

distinctive fluorescent foci where the secretion system resides, in

wild-type cells and in strains lacking components that are not
required for its assembly, but will have a diffuse fluorescent pattern
when components responsible for its recruitment are missing.
Prerequisites for this approach are the genetic amenability of the
bacterium to create fluorescent fusion proteins and a specific distri-
bution of the secretion system(s) within the bacterium (see Note
1). Beyond studying the kinetics of secretion system assembly in
live bacteria, this method can be used to obtain a detailed descrip-
tion of the assembly pathway by visualizing the labeled subunits in
strains lacking other components of the secretion system. In this
chapter, we describe a generally applicable approach to deciphering
the assembly pathway of secretion systems using fluorescently
labeled proteins.

2  Materials

2.1  Strains 1. Strain(s) expressing a chromosomal fusion of a fluorescent

protein to the target protein of interest, e.g., superfolder green
fluorescent protein (sfGFP) fused either to the N- or C-­terminus
of the target protein (see Notes 2 and 3).
2. Additional deletions of other components of the secretion sys-
tem in the strain background mentioned earlier to allow the
investigation of the order of assembly.
3. Recommended: Untagged strain as a control for autofluorescence.
4. Recommended: Strain expressing fluorescent protein in cyto-
sol from plasmid as a control (see Note 4).

2.2  Sample 1. Incubator shaker.

Preparation 2. Spectrophotometer.
3. Culture medium: Standard culture medium and appropriate
antibiotics for overnight incubation and growth of bacteria,
e.g., lysogeny broth (LB) or M9 minimal medium (see Note 5).
4. Microscopy buffer: Nonfluorescent minimal medium or imag-
ing buffer, e.g., phosphate buffered saline (PBS) (see Note 6).

2.3  Microscope Slide 1. Low-melting agarose (or agar), commercially available.

Preparation 2. Microscopy slides and cover slips compatible with used micro-
scope. Standard sizes include 75 × 25 × 1 mm glass slides and
22 × 22 mm cover slips, with No. 1 (0.13–0.16 mm) being the
most commonly used thickness.
3. Microwave oven to prepare agarose solution.
 Protein Dynamics by Fluorescence Microscopy 291

2.4  Image 1. Automated inverted epifluorescence microscope with 60× or

Acquisition 100× objective (see Note 7).
2. Optical filters for visualizing fluorescence, e.g., ET-GFP filter
set (Chroma 49002) for visualizing GFP fluorescence and ET-­
mCherry filter set (Chroma 49008) for visualizing mCherry
fluorescence.
3. Dichroic mirrors compatible with fluorophores and filter sets
used.
4. Incubation chamber for microscope stage, where required.

2.5  Software 1. Proprietary software often preinstalled on microscopy control-

for Image Processing ler, commercial software Like Adobe Photoshop, or open-­
source solutions like ImageJ, a widely used and adaptable
open-source image-processing program [19].

3  Methods

Because the protocols for propagation of bacteria and induction

of secretion systems vary greatly, we aim to provide a general pro-
tocol that can be adapted to the specificities of the studied secre-
tion system.
All buffers and solutions should be prepared using ultrapure
water at room temperature.

3.1  Preparation 1. Streak bacteria from −80 °C stock onto LB agar plates con-
of Bacteria and Setup taining the required additives and antibiotics and incubate at
of Microscopy the required temperature (e.g., 37 °C) until single colonies are
Equipment visible (usually 12–36 h).
2. Inoculate overnight cultures of bacteria with single colonies
from agar plates and grow in shaking incubator at required
temperature and agitation.
3. On the next day, determine the optical density at 600 nm
wavelength (OD600) of the overnight culture and inoculate a
main culture to OD600 that is suitable for the expression of the
analyzed secretion system (a culture volume of 5 mL is suffi-
cient, and an OD600 of 0.1 is a good starting point for many
systems).
4. Incubate bacteria in a shaking incubator until they reach the
early stationary phase (usually 1–2 h).
5. Induce secretion system according to standard conditions
(e.g., temperature shift, addition of inducer).
6. In the meantime, prepare a 1.5% solution of low-fluorescence
agarose in microscopy buffer (see Note 8). While less than 100
μL of agarose solution is required per strain, a higher volume
292 Abdelrahim Zoued and Andreas Diepold

Fig. 1 Methods for the preparation of agarose pads. (a) Approximately 50 μL of

agarose solution (beige) are transferred onto a microscope slide and quickly cov-
ered with a cover slip (gray), which is then gently and evenly pressed onto the
agarose to form an evenly distributed patch. This simple method is sufficient for
quick imaging but often leads to uneven or sloped agarose patches, which reduce
image quality. (b) To ensure a more level surface of the agarose patch, the covering
cover slip can be supported by two flanking cover slips. (c) Double-­sided tape or
commercially available adhesives (e.g. Gene Frame, Thermo Fisher) (rippled pat-
tern) can be used to permanently adhere the cover slip to the sample. This prevents
evaporation during imaging, which is very useful for longer experiments. However,
the decreasing oxygenation of the sample should be kept in mind in this case

(20–50 mL) is easier to prepare, and the agarose concentration

is less influenced by evaporation. Add the agarose to the buf-
fer, which is then carefully brought to a boil in a microwave
oven. Be aware of the possibility of delays in boiling and use
precaution. Check that the agarose is completely dissolved and
allow to cool to approximately 55 °C. After cooling, add any
required additives (antibiotics, inducers) (see Note 9).
7. Prepare a thin pad of 1.5% agarose on a microscopy slide
(Fig. 1) (see Note 10).
8. Harvest exponentially growing cells (OD600 ~ 0.8–1) by cen-
trifugation (2400 × g, 4 min; these values depend on the bac-
terium). Resuspend in imaging buffer and recentrifuge once,
using the same settings, before resuspending in imaging buffer
to an OD600 ~ 2 (see Note 11).
9. Remove cover slip from agarose patch prior to spotting the
bacteria, and wait until no more liquid areas are visible on the
surface of the agarose patch (usually 1–5 min).
10. Spot bacteria by either of the following two methods:
(a) Pipette 1–2 μL of resuspended bacteria into the center of
the agarose or agar patch without damaging the patch
itself, let dry for about 1–2 min (see Note 12), and care-
fully cover with a cover slip.
(b) Spot 1–2 μL of resuspended bacteria onto a cover slip and
carefully cover with an agarose or agar pad (Fig. 2).

3.2  Microscopy 1. Place a drop of immersion oil onto the center of the cover slip,
and flip the slide so that the cover slip faces the objective.
Carefully insert the slide into the microscope and bring the
lens into contact with the immersion oil.
 Protein Dynamics by Fluorescence Microscopy 293

Fig. 2 Schematic representation of bacteria loaded onto cover slip and covered
with agar pad

2. In phase contrast or differential interference contrast (DIC)

mode, slowly decrease the distance between the lens and the
cover slip until bacteria are visible (see Note 12 in case large
numbers of detached or swimming bacteria are visible).
3. Adjust Koehler illumination of microscope for optimal phase
contrast or DIC images [20].
4. To determine which phase contrast/DIC plane corresponds to
the best fluorescence plane, run an automated z stack of phase
contrast/DIC and fluorescence images (see Note 13). For bac-
teria with a diameter of about 1 μm, z stacks containing 10–20
planes with ∆z = 100 nm yield sufficient coverage.
5. Capture phase contrast and fluorescence micrographs:
(a) For kinetic studies: Every 30 s with a minimal exposure
time to minimize bleaching and phototoxicity effects (see
Note 14).
(b) For the determination of the assembly pathway: Single
micrograph or z stack image (see Note 15) in wild-type
and mutant strains.
294 Abdelrahim Zoued and Andreas Diepold

3.3  Image 1. Phase contrast and fluorescence images can be adjusted and
Processing merged using ImageJ or equivalent software (see Note 16).
2. Slight movements of the whole field during the time of the
experiment can be corrected by registering individual frames
using the StackReg plug-in in ImageJ [19].
3. Blurring of the image can be reduced by deconvolution, a
mathematical postimaging process that removes or reassigns
the fraction of detected photons caused by out-of-focus struc-
tures. This is especially useful when three-dimensional infor-
mation from z stacks is available. Care should be taken not to
mistake deconvolution artifacts for clustering, and negative
controls are mandatory when applying deconvolution.
4. Detection and quantification of bacteria and fluorescent foci
can be performed in ImageJ or using the Oufti software pack-
age (formerly known as MicrobeTracker [21]).

3.4  Determination 1. To build an assembly pathway from the aforementioned data,

of Assembly Pathway determine the presence and number of foci per cell in wild-
type and mutant strains. Nucleating component(s) of the sys-
tem are correctly localized in foci in strains lacking all other
components of the secretion system. Subsequently assembled
proteins correctly localize in foci in strains expressing all earlier
assembled proteins. Finally, proteins that require the presence
of all the other components of the system are recruited at the
end of the assembly. Proteins that are recruited at the same
time or that interact before being recruited to the system will
show the same localization in mutant strains and usually require
each other to form foci.

4  Notes

Notes 1–4 correspond to the prerequisites for the applicability of

this approach.
1. Most bacterial secretion systems have a sufficiently distinct
localization within the bacterium that differs from the distribu-
tion of the free component. Should this not be the case, stan-
dard fluorescence microscopy cannot distinguish between the
free and assembled state of a labeled component, and more
sophisticated methods, such as diffusion-based fluorescence
correlation spectroscopy (FCS) or interaction-based Förster
resonance energy transfer microscopy (FRET), must be applied.
2. While most fluorescent proteins are relatively inert to interac-
tions, their size of 25–30 kDa [22] can lead to cleavage or
degradation of the fusion protein and impede the assembly or
 Protein Dynamics by Fluorescence Microscopy 295

functionality of the tagged protein. Smaller alternatives, such

as tetracysteine tags [23, 24], require additional manipulation
[25] and may also disturb the functioning of the protein (own
unpublished observations). It is therefore essential to test the
expression level and stability of the fusion protein (by immu-
noblot), as well as the functionality of the secretion system in
the corresponding strains (by a functional assay). While fusions
that influence functioning may be perfectly fine tools for deci-
phering the assembly, this must be corroborated by indepen-
dent experiments. To maximize the chances of obtaining a
functional fusion protein, both termini of the protein as well as
internal flexible loops should be considered. Flexible linkers
between the fluorophore and the secretion system component
(e.g., a stretch of 6–15 amino acids with a high glycine con-
tent) have been shown to preserve the functionality of the
fusion protein. Chromosomal fusion proteins are preferred to
avoid mislocalization owing to overproduction of the protein
or wrong timing and order of the expression of subunits.
Moreover, chromosomal fusions make it possible to analyze
the secretion system under close-to-wild-type conditions.
However, especially for C-terminal fusions, care must be taken
not to disturb the expression of downstream genes in the same
operon, and it has proven helpful to repeat the genetic region
upstream of the following gene.
3. Concerning the choice of fluorescent protein, many variants of
GFP have been produced that vary in spectral properties,
degrees of multimerization, folding rates, and functionality in
oxidizing environments, so it can be helpful to try different
fusion proteins [22]. Owing to their fast folding, low multi-
merization tendency, and proper folding in the periplasm,
sfGFP and mCherry are good starting points. mCherry has
additionally been observed to retain functionality of the T3SS
in cases where GFP fusions were nonfunctional ([26] and
unpublished results). Most bacteria also display considerably
less autofluorescence in the red spectrum; however, mCherry
is less photostable than GFP, which might make it less suitable
for time course studies.
4. While these control strains are not absolutely required, espe-
cially in the presence of good controls for the protein of choice
itself (i.e., deletion of a protein required for its localization),
they are immensely valuable for setting up and testing the
microscopy pipeline.
5. M9 and similar buffers have the advantage that, owing to
their low autofluorescence, they can be directly used in
microscopy. This ensures constant external conditions for
the bacteria and may eliminate the need for the washing step
described in step 8.
296 Abdelrahim Zoued and Andreas Diepold

6. The choice of imaging buffer is crucial to obtain reproducible

results because it will influence bacterial metabolism and possi-
bly the state of the secretion system to be analyzed. While phos-
phate buffered saline (PBS) is a popular and easy-to-­ obtain
imaging buffer, some bacteria show visible alterations in cell
morphology in PBS within less than an hour. Preliminary experi-
ments can reveal whether cell morphology and the distribution
of secretion systems are affected in different imaging buffers.
7. The microscope must have a sufficient resolution and, most
importantly, a high sensitivity to visualize and resolve the assem-
bled proteins. A 100× objective is required for most distribu-
tions of secretion systems, although the formation of a polar
spot or the distribution of few membrane-bound foci in large
bacteria can be detected with a 60× objective. The sensitivity of
the microscope is crucial, especially for low-­stoichiometry com-
ponents. The labeled protein must be present in multiple copies
within the complex to be detectable. In our experience, sensitive
wide-field microscopes can detect approximately ten molecules
within a diffraction-­limited spot over low background. For sin-
gle-molecule detection, more sensitive methods, such as total
internal reflection microscopy (TIRF) [27] or photoactivated
localization microscopy/stochastic optical reconstruction
microscopy (PALM/STORM) [28], must be applied.
8. The pad can also be done with agar (instead of agarose). This
can be especially useful for longer time course experiments,
where bacteria can be incubated for 1 h at optimal growth
temperature prior to microscopy acquisition to allow cell divi-
sion on a plane surface.
9. The agarose solution should be prepared or redissolved freshly
before the experiment. The solution will stay liquid in a 55 °C
water bath; small aliquots can be kept in a tabletop incubator
shaker for 1.7 mL reaction tubes (vigorous shaking is required
to prevent solidification of the agarose in this case).
10. The depth of the pad can vary; however, the surface should
remain as smooth as possible. The pad can be prepared using a
microscopy slide and a cover slip, with spacers, using commer-
cially available systems (e.g. GeneFrame) (Fig. 1) or,
­alternatively, using two microscope slides or prewarmed pro-
tein gel chambers for larger patches. The pad should be bubble
free to facilitate observation. Let the agarose solidify and dry at
room temperature (>1 min; longer storage times are possible if
the pad remains covered).
11. An OD of 2 leads to about 5% of the area being covered with
bacteria (for E. coli; this obviously depends on the size of the
bacterium). Increasing the OD will increase confluence, lead-
ing to more cell–cell contact.
 Protein Dynamics by Fluorescence Microscopy 297

12. Depending on their surface and the properties of the agarose

patch, bacteria may take some time to settle at this point. If a
large part of the bacteria are still moving after some minutes,
the volume of bacterial resuspension should be reduced and
drying times increased. Do not dry the agarose pad at 4 °C to
avoid drifts during observation.
13. To avoid saturation, strong photobleaching, or phototoxicity
effects, all fluorescence images should be acquired with the
minimal exposure time required to reach a sufficient signal-to-­
noise ratio. The optimal exposure time must be determined for
each protein; depending on the sensitivity of the microscope,
exposure times of 20–100 ms for phase contrast or DIC and
100 ms to 2 s for fluorophores are good starting points.
Narrow-band microscopy filters can also reduce
photobleaching.
14. For time-lapse experiments, many microscopy systems allow
one to define fields of view (x, y, z, focus offset) that are stored
and then automatically accessed by a motorized stage. The
fields of view should be sufficiently far apart to avoid cross-­
photobleaching (run a preliminary experiment with long expo-
sures to determine the area of bleaching, if required). Ten
fields of view at an OD of about 2 usually yield a sufficient
number of bacteria for further analysis.
15. Z stacks allow a more complete coverage of the bacterium,
ensuring images that comprise the region of interest (often the
center of the bacterium). In addition, the three-dimensional
data yield information about the spatial distribution of the
secretion systems within the bacterium and allow better decon-
volution of the images. However, imaging z stacks leads to
stronger photobleaching and is therefore generally avoided in
kinetic experiments. For kinetic experiments, keeping the focus
is of particular importance, and hardware-based focusing sys-
tems or sealed plates can be advantageous.
16. To avoid the loss of raw data, the original image files should be
preserved. Keep the low and high boundary below the back-
ground value and above the highest measured intensity, respec-
tively, to prevent misinterpretation of the data. Within an
experiment, these values should be kept constant after back-
ground correction.

References

1. Tseng T-T, Tyler B, Setubal J (2009) Protein bacteria: structural and mechanistic insights.
secretion systems in bacterial-host associations, Nat Rev Microbiol 13:343–359
and their description in the Gene Ontology.
3.
Kimbrough TG, Miller SI (2000)
BMC Microbiol 9:S2 Contribution of Salmonella typhimurium type
2. Costa TRD, Felisberto-Rodrigues C, Meir A, III secretion components to needle complex
Prevost MS, Redzej A, Trokter M, Waksman G formation. Proc Natl Acad Sci U S A
(2015) Secretion systems in Gram-negative 97:11008–11013
298 Abdelrahim Zoued and Andreas Diepold

4. Sukhan A, Kubori T, Wilson J, Galán JE structural components and involves the for-
(2001) Genetic analysis of assembly of the mation of an YscV oligomer. Mol Microbiol
Salmonella enterica serovar Typhimurium 82:502–514
type III secretion-­associated needle complex. 17. Aguilar J, Zupan J, Cameron TA, Zambryski
J Bacteriol 183:1159–1167 PC (2010) Agrobacterium type IV secretion
5. Kimbrough TG, Miller SI (2002) Assembly system and its substrates form helical arrays
of the type III secretion needle complex of around the circumference of virulence-induced
Salmonella typhimurium. Microbes Infect cells. Proc Natl Acad Sci U S A 107:3758–3763
4:75–82 18. Durand E, Nguyen VS, Zoued A, Logger L,
6. Ogino T, Ohno R, Sekiya K, Kuwae A, Péhau-Arnaudet G, Aschtgen M-S, Spinelli
Matsuzawa T, Nonaka T, Fukuda H, Imajoh-­ S, Desmyter A, Bardiaux B, Dujeancourt A,
Ohmi S, Abe A (2006) Assembly of the type Roussel A, Cambillau C, Cascales E, Fronzes
III secretion apparatus of enteropathogenic R (2015) Biogenesis and structure of a type
Escherichia coli. J Bacteriol 188:2801–2811 VI secretion membrane core complex. Nature
7. Fronzes R, Schäfer E, Wang L, Saibil H, 523:555–560
Orlova E, Waksman G (2009) Structure of a 19. Schneider CA, Rasband WS, Eliceiri KW
type IV secretion system core complex. Science (2012) NIH Image to ImageJ: 25 years of
323:266–268 image analysis. Nat Methods 9:671–675
8. Schraidt O, Lefebre MD, Brunner MJ, Schmied 20. Köhler A (1893) Ein neues Beleuchtungsverfahren
WH, Schmidt A, Radics J, Mechtler K, Galán für mikrophotographische Zwecke. Z Wiss
JE, Marlovits TC (2010) Topology and organi- Mikrosk 10:433–440
zation of the Salmonella typhimurium type III 21. Paintdakhi A, Parry B, Campos M, Irnov I,
secretion needle complex components. PLoS Elf J, Surovtsev I, Jacobs-Wagner C (2015)
Pathog 6:e1000824 Oufti: an integrated software package for high-­
9. Reichow SL, Korotkov KV, Hol WGJ, Gonen accuracy, high-throughput quantitative micros-
T (2010) Structure of the cholera toxin secre- copy analysis. Mol Microbiol 99:767–777
tion channel in its closed state. Nat Struct Mol 22. Shaner NC, Steinbach PA, Tsien RY (2005)
Biol 17:1226–1232 A guide to choosing fluorescent proteins. Nat
10. Chandran Darbari V, Waksman G (2015) Methods 2:905–909
Structural biology of bacterial type IV secre- 23. Adams S, Campbell R, Gross L, Martin B,
tion systems. Annu Rev Biochem 84:603–629 Walkup G, Yao Y, Llopis J, Tsien RY (2002)
11. Rose P, Fröbel J, Graumann PL, Müller M New biarsenical ligands and tetracysteine
(2013) Substrate-dependent assembly of the motifs for protein labeling in vitro and in vivo:
Tat translocase as observed in live Escherichia synthesis and biological applications. J Am
coli cells. PLoS One 8:e69488 Chem Soc 124:6063–6076
12. Alcock F, Baker MAB, Greene NP, Palmer T, 24. Andresen M, Schmitz-Salue R, Jakobs S
Wallace MI, Berks BC (2013) Live cell imaging (2004) Short tetracysteine tags to beta-tubulin
shows reversible assembly of the TatA compo- demonstrate the significance of small labels for
nent of the twin-arginine protein transport sys- live cell imaging. Mol Biol Cell 15:5616–5622
tem. Proc Natl Acad Sci U S A 110:3650–3659 25. Enninga J, Mounier J, Sansonetti P, Tran Van
13. Lybarger S, Johnson TL, Gray M, Sikora A, Nhieu G, Van Nhieu GT (2005) Secretion of
Sandkvist M (2009) Docking and assembly of type III effectors into host cells in real time.
the type II secretion complex of Vibrio chol- Nat Methods 2:959–965
erae. J Bacteriol 191:3149–3161 26. Diepold A, Kudryashev M, Delalez NJ, Berry
14. Johnson TL, Sikora AE, Zielke RA, Sandkvist RM, Armitage JP (2015) Composition, forma-
M (2013) Fluorescence microscopy and pro- tion, and regulation of the cytosolic C-ring, a
teomics to investigate subcellular localization, dynamic component of the type III secretion
assembly, and function of the type II secretion injectisome. PLoS Biol 13:e1002039
system. Methods Mol Biol 966:157–172 27. Poulter NS, Pitkeathly WTE, Smith PJ,
15. Diepold A, Amstutz M, Abel S, Sorg I, Jenal U, Rappoport JZ (2015) In: Verveer PJ (ed)
Cornelis GR (2010) Deciphering the assembly Advanced fluorescence microscopy. Springer,
of the Yersinia type III secretion injectisome. New York
EMBO J 29:1928–1940 28. MacDonald L, Baldini G, Storrie B (2015)
16. Diepold A, Wiesand U, Cornelis GR (2011) Does super-resolution fluorescence micros-
The assembly of the export apparatus copy obsolete previous microscopic approaches
(YscR,S,T,U,V) of the Yersinia type III secre- to protein co-localization? Methods Mol Biol
tion apparatus occurs independently of other 1270:255–275
 Chapter 24

Large Complexes: Cloning Strategy, Production,

and Purification
Eric Durand and Roland Lloubes

Abstract
Membrane proteins can assemble and form complexes in the cell envelope. In Gram-negative bacteria, a
number of multiprotein complexes, including secretion systems, efflux pumps, molecular motors, and pilus
assembly machines, comprise proteins from the inner and outer membranes. Besides the structures of
isolated soluble domains, only a few atomic structures of these assembled molecular machines have been
elucidated. To better understand the function and to solve the structure of protein complexes, it is thus
necessary to design dedicated production and purification processes. Here we present cloning procedures
to overproduce membrane proteins into Escherichia coli cells and describe the cloning and purification
strategy for the Type VI secretion TssJLM membrane complex.

Key words Membrane protein complexes, Escherichia coli, T7 overexpression, Protein purification

1  Introduction

Protein overproduction results from cloning a gene of interest into

a plasmid vector, downstream of a tightly regulated promoter, and
from inducing its expression after plasmid transformation into a
bacterial strain. For large protein complexes containing multiple
subunits, the genes encoding the different subunits can be
expressed under the control of an inducible promoter either from
a single plasmid containing a cluster of genes or from different
compatible plasmids harboring single or multiple genes.

1.1  Cloning Vectors Several inducible promoters have been described and are available
to overexpress a gene of interest. These promoters are usually
cloned into vectors that also contain the gene encoding the cognate
regulatory protein and a transcriptional terminator to prevent non-
productive transcription from the downstream gene [1–7]. The tac
and trc promoters that contain the −35 and −10 sequences from
the trp and lacUV5 promoters, respectively, have been ­optimized
for high expression levels [1]. The most famous regulated E. coli

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_24, © Springer Science+Business Media LLC 2017

299
300 Eric Durand and Roland Lloubes

promoters are the tetA and araBAD promoters, regulated by the

TetR repressor and the AraC activator, respectively [3–6]. In addi-
tion, heterologous combinations of promoter/regulator binding
sequences can be used [2]. The last family of inducible promoters
gathers sequences that are not recognized by the E. coli RNA poly-
merase but are rather recognized by phage RNA polymerases such
as the SP6, T3, and T7 promoters. Vectors and strains that express
the T7 RNA polymerase (T7RNAp) have been extensively devel-
oped. Three independent methods are used to regulate the T7
expression systems. First, the expression of the chromosomally
encoded or plasmid-encoded T7 RNAp gene can be itself under the
control of an inducible promoter (see Note 1). Second, the inhibi-
tion of the T7 RNAp basal activity can be controlled by producing
the T7 lysozyme under constitutive or regulated conditions [8, 9].
Finally, the transcription by the T7 RNAp can be repressed by the
Lac repressor, adding the lac operator sequence [10], the lacI gene
being cloned on the expression vector.
Many cloning vectors are available and can be selected accord-
ing to the cellular location, toxicity, stability and folding rate of the
proteins to overproduce:
–– Production of proteins in the cytoplasm, the periplasm, or the
membrane of E. coli (using the addition of synthetic N-­terminal
signal sequences, such as that of the OmpA and PelB
proteins);
–– Compatible T7 expression vectors containing one or two T7
promoters to overproduce protein complexes with up to eight
subunits (see Duet vectors from Novagen);
–– Protein tagging sequences (6–10×His, Strep-Tag II…) or fusion
partners (protein G, glutathione-S-transferase (GST), calmodu-
lin-binding peptide (CBP), maltose-binding protein (MBP)) to
increase protein solubility or to simplify purification using affin-
ity chromatography techniques. In addition, these tag affinity
sequences can be removed using specific proteases (the most
commonly used proteases are Tobacco Etch Virus (TEV) pro-
tease, enterokinase, Thrombin, Xa Factor, PreScission). For this
purpose, the corresponding protease recognition sequence is
inserted either downstream (for N-terminal tagging) or
upstream (for C-terminal tagging) of the tag sequence. After
protease digestion, the purified protein complex is subjected to
a new purification step to remove the protease, the uncleaved
protein, and the tag peptide (e.g., affinity chromatography to
remove the His-tag peptide and the His-­tagged TEV).
Currently, cloning into these vectors is facilitated by poly-
merase chain reaction (PCR) techniques based on gene and plas-
mid amplifications, leading to restriction site/ligation-free cloning
methods. These techniques have proven efficient to carry out rapid
gene expression strategies. (see Subheading 3.1 and Note 2).
 Membrane Protein Complex Purification 301

If none of the overproduction strategies is sufficient to pro-

duce and purify protein complexes, in vitro alternatives, such as
cell-free transcription–translation systems using the T7RNAp and
E. coli cell extracts, have been found efficient at producing
milligram-­range amounts of some membrane proteins [11].

1.2  Membrane Membrane protein complexes have been successfully overproduced

Protein Complex and purified based on pBAD or T7 vectors. For example, the Tol
Overproduction proteins from the Tol cell envelope molecular motor were overpro-
and Purification duced and specifically radiolabeled with 35S–Methionine after clon-
ing into a pT7-driven vector using restriction sites (RS) and ligation
techniques (RSl) ([12, 13], see Notes 3 and 4). Subcomplexes
from the Ton system [14–17], the PomA-PomB flagellar rotor
[18], the AcrAB-TolC efflux pump [19], the Type IV and Type VI
secretion systems (T4SS and T6SS, respectively) [20, 21], and the
β-barrel assembly machinery [22] have also been successfully puri-
fied using similar approaches.
In addition to the cloning strategy and the production to levels
compatible with purification, studying these multiprotein machines
also requires extracting and solubilizing the protein complex with-
out disrupting the contacts between the subunits. Finally, since the
overproduction condition may induce stoichiometry artifacts, the
tagging of the minor subunit or the specific tagging of different
subunits should be performed to purify stable protein complexes
(see subsequent discussion and Subheading 3.2).
Here we describe the cloning strategy as well as the produc-
tion, extraction, and purification of the T6SS TssJLM membrane
complex [21].

2  Materials

2.1  Cloning 1. Desalted oligonucleotides of up to 35 bases, 5′ ends

of TssJLM Complex unphosphorylated.
2. Restriction enzymes (NdeI, XhoI, DpnI).
3. T4 DNA ligase.
4. Thermoblock (16–42 °C).
5. E. coli DH5α competent cells.
6. PCR thermocycler.
7. High-performance liquid chromatography purified Mega-­
primer pairs (≥50 bases), 5′ ends unphosphorylated.
8. Vector: pRSF-Duet1.
9. High-fidelity Taq DNA polymerase (Pfu turbo, Agilent).
10. dNTPs, 10 mM stock solution in water.
11. Agarose gel electrophoresis (AGE) system.
302 Eric Durand and Roland Lloubes

2.2  Production 1. Isopropyl β-d-1-thiogalactopyranoside (IPTG) stock solution:

and Purification 0.1 M in water.
of TssJLM Complex 2. BL21(DE3): E. coli strain B F− ompT gal dcm lon hsdSB(rB−mB−)
λ(DE3 [lacI lacUV5-T7 gene 1 ind1 sam7 nin5]) [malB+] (λS).
3. Lysozyme stock solution: 10 mg/mL in water.
4. DNase stock solution: 10 mg/mL in water.
5. Tris(2-carboxyethyl)phosphine hydrochloride (TCEP) extem-
poraneously added.
6. Ethylenediaminetetraacetic acid (EDTA) stock solution: 0.5 M
in water.
7. MgCl2 stock solution: 1 M in water.
8. Lysis buffer: 50 mM Tris–HCl, pH 8.0, 50 mM NaCl, 1 mM
EDTA.
9. Solubilization buffer: 50 mM Tris–HCl, pH 8.0, 50 mM NaCl,
1 mM EDTA, 0.5% (w/v) n-dodecyl-β-d-maltopyranoside
(DDM), 0.75% (w/v) decyl maltose neopentyl glycol
(DM-NPG), 0.5% (w/v) digitonin (Sigma-Aldrich).
10. EDTA-free protease inhibitor tablets.
11. Affinity buffer: 50 mM Tris–HCl, pH 8.0, 50 mM NaCl,
0.05% (w/v) DM-NPG.
12. Imidazole-HCl, pH 8.0: Stock solution 4 M.
13. Incubation shakers (16–37 °C).
14. Centrifugation (from 5000 to 100,000 × g).
15. Glass potter.
16. Emulsiflex-C5 (Avestin).
17. 5-mL StrepTrap HP and 5-mL HisTrap HP columns (GE
Healthcare).
18. Desthiobiotin.

3  Methods

3.1  Cloning TssJLM To overproduce the T6SS TssJLM membrane complex, the tssJ,
Membrane Complex tssL, and tssM genes were assembled in an artificial operon with
individual optimized ribosome-binding sites (RBS). In addition,
based on previous data regarding permissive positions, each sub-
unit was tagged with a specific tag (see Fig. 1).
The pRSF–TssJStrep-TssLFLAG-His6TssM plasmid was constructed
by both RSl (see Note 4) and RSl-free (see Note 2, [23]) cloning
methods (Fig. 1).
1. PCR amplify the tssJ gene (encoding the TssJ lipoprotein)
using high-fidelity Taq DNA polymerase and primers contain-
ing the NdeI (5′ primer) and the XhoI RS (3′ primer) extensions.
 Membrane Protein Complex Purification 303

Fig. 1 Cloning of synthetic tssJLM operon encoding the T6SS membrane core complex. (a) The tssL, tssM, and
tssJ genes were PCR amplified from the entero-aggregative sci-1 T6SS operon. DNA sequences were added
encoding an optimized RBS at the 5′ end of each gene and sequences encoding a C-terminal Strep-II
tag (STREP) for TssJ and N-terminal 6×His and Flag tags for TssM and TssL, respectively. (b) The tssJLM
artificial operon was cloned in the pRSF-Duet vector. (c) Schemes representing the protein constructs, the
subdomain boundaries, and some indicated characteristics (TM transmembrane segments, SP signal peptide,
and CYS acylated cysteine)

The Strep-tag II DNA sequence followed by a stop codon was

introduced in the 3′ primer (see Note 5). Check the correct
amplification by AGE.
2. Digest PCR product and pRSF-Duet vector with NdeI and
XhoI restriction enzymes (sites present in pRSF-Duet MCS2).
3. Mix the digested PCR amplified fragment with about 50 ng of
the digested pRSF-Duet vector, with a molar ratio of insert/
vector of 2/1 to 5/1.
4. Add 1 U of T4 DNA ligase to the DNA mixture plus its spe-
cific buffer in a total volume of 15 μL and incubate at 16–20 °C
for at least 2 h.
5. Transform E. coli DH5α competent cells with about 30% of
the ligation mix, plate on lysogeny broth (LB) agar containing
the appropriate antibiotic (in case of pRSF-Duet, use kanamy-
cin 50 μg/mL), and incubate overnight at 37 °C.
6. Select positive clones containing the pRSF–TssJStrep plasmid by
colony PCR screening using the same primers. Extract the
plasmid DNA and check the accuracy of the cloned sequence
by DNA sequencing.
7. PCR amplify the tssL gene using (1) a 5′ primer containing a 5′
extension (22 bp) corresponding to the Strep-TagII and the
stop codon followed by the ribosome binding site (RBS)
sequence (see Note 6) and the 5′-FLAG-tssL extension (63 bp),
and (2) a 3′ primer containing the 3′-tssL gene extension
304 Eric Durand and Roland Lloubes

(35 bp) including the stop codon followed by RBS, the ATG

start codon, and 5′-His6-tssM extension (14 bp). Check cor-
rect amplification by AGE.
8. PCR amplify the tssM gene using (1) the 5′ primer comple-
mentary to the 3′ end of tssL gene, RBS, His-tag (20 bp), and
5′ end of tssM (20 bp) and (2) the 3′ primer containing the
3′-tssM gene extension (35 bp) including the stop codon fol-
lowed by the pRSF-Duet sequence extension (35 bp). Check
correct amplification by AGE.
9. Mix the flag-tssL, his-tssM PCR products and the pRSF–TssJStrep
plasmid, and PCR amplify with high-fidelity DNA polymerase
in a single reaction, as described for the RSl-free technique (see
Note 2).
10. Digest the mixture with DpnI. Transform into E. coli DH5α
competent cells and plate on LB agar containing the appropri-
ate antibiotic and incubate overnight at 37 °C.
11. Select positive clones containing the pRSF–TssJStrep-FLAGTssL-­
6His
TssM plasmid by colony PCR screening using the same
primers. Extract the plasmid DNA and check the accuracy of
the cloned sequence by DNA sequencing.

3.2  TssJLM 1. Transform the expression vector (pRSF-TssJStrep-FLAGTssL-­

Membrane Complex, 6His
TssM) into the E. coli BL21(DE3) expression strain.
Extraction 2. Grow cells at 37 °C in 8 L of LB to an optical density at 600 nm
and Purification (OD600) ~ 0.7. Induce the expression of the tssJLM genes with
1.0 mM IPTG for 16 h at 16 °C (see Note 7).
3. Pellet cells by centrifugation at 7000 × g for 20 min. Resuspend
cell pellets in 300 mL of ice-cold lysis buffer supplemented
with 1 mM TCEP, 100 μg/mL DNase I, 100 μg/mL lyso-
zyme, and one tablet of EDTA-free protease inhibitor. Add
MgCl2 to the final concentration of 10 mM.
4. Break the cell suspension with an Emulsiflex-C5 by four pas-
sages at 15,000 psi (100 MPa). Pellet unbroken cells by cen-
trifugation at 7000 × g (see Note 8).
5. Pellet membranes by ultracentrifugation at 98,000 × g for
45 min.
6. Resuspend membranes in 120 mL of solubilization buffer sup-
plemented with 1 mM TCEP at 22 °C and homogenize mem-
branes mechanically with a potter (duration about 45 min) (see
Note 9).
7.
Clarify the membrane suspension by centrifugation at
98,000 × g for 20 min.
8. Load the supernatant onto a 5 mL StrepTrap HP column and
then wash with affinity buffer at 4 °C.
 Membrane Protein Complex Purification 305

9. Elute the TssJLM core complex in affinity buffer supplemented

with 2.5 mM desthiobiotin into a 5 mL HisTrap HP column.
10. Wash the HisTrap HP column in affinity buffer supplemented
with 20 mM imidazole and proceed to the elution of the
TssJLM core complex in the same buffer supplemented with
500 mM imidazole.
11. Pool the peak fractions and load onto a Superose 6 10/300
column equilibrated in 50 mM Tris–HCl, pH 8.0, 50 mM
NaCl, 0.025% (w/v) DM-NPG (see Note 10). Elute the
TssJLM complex as a single monodisperse peak close to the
void volume of the column.

4  Notes

1. E. coli T7RNAp expression systems result from (1) plasmid

(pGP1–2) expression: the T7RNAp under the control of the
Lambda PL promoter is regulated by the plasmid-encoded
temperature-sensitive C1–857 repressor [24]; (2) chromo-
somal expression: the T7RNAp gene is under the control of
the lacUV5 [25] or the ara promoter [26]; (3) besides tight
control of the chromosomal T7RNAp expression upon AraC
control [26], infection by phages (M13 mGP1–2 or Lambda
CE6) encoding the T7RNAp [27, 28] has proven to be effi-
cient for the expression of toxic gene products.
2. The RSl-free cloning strategy allows the insertion of a gene at
a precise position on a target plasmid but does not require the
presence of RS. The restriction free cloning method consists in
two sequential PCR amplifications using only two primers.
These primers (with ≥25 bp homology extension) are designed
to hybridize on the 5′ and 3′ ends of the gene of interest and
contain additional extension to hybridize on the expression
vector at a selected position downstream of the regulated pro-
moter sequence. First, the gene of interest is amplified by PCR,
then the amplified mega-primer pairs containing the gene are
annealed to the vector of interest. A new amplification with
high-fidelity polymerase produces a linear gene-­vector amplifi-
cation [29]. The gene is thus plasmid included in a nicked and
circular DNA molecule. The PCR is treated by DpnI restric-
tion enzyme to digest the unwanted Dam methylated plasmid
templates. Competent E. coli cells are further transformed with
the annealed DNA complex directly. An alternative method
has been developed [30]. It uses a PCR-­amplified gene con-
taining at least 15 bp extensions that are homologous to each
end of the linearized vector (PCR-­amplified or RS-digested).
PCR DNA and vector treated with T4 polymerase, to create 5′
overhangs, are annealed to form a recombinant plasmid.
306 Eric Durand and Roland Lloubes

3. Addition of a T7 promoter by an RS-dependent cloning strat-

egy. RS present in the upstream sequence of the gene of inter-
est can be used to insert a synthetic T7 promoter DNA
fragment that is formed by two overlapping oligonucleotides
containing the T7 consensus promoter sequence of 23 bases
corresponding to 5′-taatacgactcactatagggaga-3′. The T7 pro-
moter sequence is inserted upstream of the RBS of the gene of
interest (it is also possible to insert the T7 promoter upstream
of the natural promoter sequence). For this purpose, the syn-
thetic DNA fragment contains additional 5′- and 3′-end exten-
sions that are complementary to the sticky ends of the RS
present in the plasmid (blunt-ended RSs can also be used but
with lower ligation efficiency and random insertion). Two
complementary desalted oligonucleotides (ONs) of about 35
bases, 5′ end unphosphorylated, are hybridized following heat
denaturation in ultrapure water and further cooling at room
temperature. The annealed DNA fragment is further ligated
into an RS-digested plasmid. It is possible to favor the selec-
tion of positive clones using synthetic DNA that do not recre-
ate the initial RS. Then the heat-inactivated ligation mix is
digested with the RE corresponding to the RS that was
destroyed upon DNA fragment ligation (e.g., DNA fragment
insertion into EcoRI RS: after EcoRI digestion, the overhang
EcoRI: 5′-AATTC… should be filled with the synthetic
sequence 5′-AATTX…, where the X nucleotide does not cor-
respond to the C nucleotide). It is notable that this fast tech-
nique does not require plasmid sequencing. It should be used
for plasmids harboring resistance and regulatory genes in the
opposite orientation from that of T7 regulated genes [31].
4. The general RSl-dependent strategy is often used for cloning
genes of interest into the multiple cloning sites (MCS) of
expression vectors. The genes can be either purified from RS
digestion [32] or obtained from PCR amplification using
primers that contain additional RS extensions. Then the PCR-­
amplified DNA is RE digested and inserted into the cognate
RS of the MCS present in the vector [16].
5. Since lipoproteins undergo a post-translational modification at
their N-termini [33], the Strep-Tag II affinity tag sequence was
introduced at the C-terminal of TssJ. It is important to note
that the positions of affinity tags (Strep, FLAG, and His tags)
were rationally chosen to maintain functional proteins.
6. To optimize the production of the TssL and TssM proteins,
the endogenous RBS-ATG 5′ sequence in front of the genes
was replaced (in the 5′ primer sequence) by the consensual
sequence “AAGGAGATATACATATG” [34] (RBS and start
codon are presented, bold and italic letters, respectively).
7. Growth and induction at 37 °C for 3 h led to very low biomass
and protein yield. The culture was conducted at 37 °C until
 Membrane Protein Complex Purification 307

OD600 ~ 0.7–0.9 before shifting the incubator at 16 °C prior to

induction by IPTG. The final OD after the 16H induction was
around 1.4–1.6. It is important not to induce around 0.4 OD
since the cell would stop growing shortly thereafter.
8. It is important to note that whereas many cell-disrupting pro-
tocols have been tried, only lysis with Emulsiflex produced a
stable and homogeneous sample of the TssJLM membrane
core complex.
9. This specific combination of detergents (same recipe as that
used for the purification of the T4SS [20]) gave a higher extrac-
tion yield of the TssJLM membrane complex. Other isolated
detergents (Triton-X-100, n-Octyl-β-D-Glucoside, DDM, and
DM-NPG) gave poor extraction yields. All detergent buffers
were prepared the day of the purification. Digitonin tends to
precipitate in high salt buffers and without other detergents.
The addition of DDM and DM-NPG prevents digitonin
precipitation.
10. DM-NPG was chosen because the solubilized TssJLM com-
plex is stable and since this detergent has been used success-
fully for high-resolution structural biology studies, giving a
very clear and reproducible electron microscopy background
after negative staining. Using a low salt concentration (50 mM
maximum) was key to preventing the aggregation of the
TssJLM complex. The affinity purification steps were of para-
mount importance for isolating a stoichiometric complex.
Indeed, a large excess of the TssJ lipoprotein was eliminated
during the second His column.

Acknowledgements

We would like to thank M. Petiti, L. Houot, H. Célia, and D. Duché

for their careful reading of the text. E.D. and R.L. are funded by the
Centre National de la Recherche Scientifique, the Aix-Marseille
Université, and two grants from the Agence Nationale de la
Recherche (ANR-10-JCJC-1303-03 and ANR-­ 14-­CE09-0023,
respectively). ED is supported by the Institut National de la Santé Et
de la Recherche Médicale through a permanent research position.

References

1. Amann E, Brosius J, Ptashne M (1983) Vectors 3. Cagnon C, Valverde V, Masson JM (1991) A

bearing a hybrid trp-lac promoter useful for new family of sugar-inducible expression vectors
regulated expression of cloned genes in for Escherichia coli. Protein Eng 4:843–847
Escherichia coli. Gene 25:167–178 4. Skerra A (1994) Use of the tetracycline pro-
2. Lutz R, Bujard H (1997) Independent and moter for the tightly regulated production of a
tight regulation of transcriptional units in murine antibody fragment in Escherichia coli.
Escherichia coli via the LacR/O, the TetR/O Gene 151:131–135
and AraC/I1-I2 regulatory elements. Nucleic 5. Guzman LM, Belin D, Carson MJ, Beckwith
Acids Res 25:1203–1210 J (1995) Tight regulation, modulation, and
308 Eric Durand and Roland Lloubes

high-level expression by vectors containing the Escherichia coli as revealed by LILBID mass
arabinose PBAD promoter. J Bacteriol 177: spectrometry. Biochemistry 50:8950–8956
4121–4130 17. Sverzhinsky A, Fabre L, Cottreau AL, Biot-­
6. Haldimann A, Daniels LL, Wanner BL (1998) Pelletier DM, Khalil S, Bostina M, Rouiller I,
Use of new methods for construction of tightly Coulton JW (2014) Coordinated rearrange-
regulated arabinose and rhamnose promoter ments between cytoplasmic and periplasmic
fusions in studies of the Escherichia coli phos- domains of the membrane protein complex
phate regulon. J Bacteriol 180:1277–1286 ExbB-ExbD of Escherichia coli. Structure
7. Balzer S, Kucharova V, Megerle J, Lale R, 22:791–797
Brautaset T, Valla S (2013) A comparative 18. Yonekura K, Maki-Yonekura S, Homma M
analysis of the properties of regulated promoter (2011) Structure of the flagellar motor protein
systems commonly used for recombinant gene complex PomAB: implications for the torque-­
expression in Escherichia coli. Microb Cell Fact generating conformation. J Bacteriol 193:
12:26 3863–3870
8. Studier FW (1991) Use of bacteriophage T7 19. Kim JS, Jeong H, Song S, Kim HY, Lee K,
lysozyme to improve an inducible T7 expres- Hyun J, Ha NC (2015) Structure of the tripar-
sion system. J Mol Biol 219:37–44 tite multidrug efflux pump AcrAB-TolC sug-
9. Schlegel S, Löfblom J, Lee C, Hjelm A, Klepsch gests an alternative assembly mode. Mol Cells
M, Strous M, Drew D, Slotboom DJ, de Gier 38:180–186
JW (2012) Optimizing membrane protein 20. Low HH, Gubellini F, Rivera-Calzada A,
overexpression in the Escherichia coli strain Braun N, Connery S, Dujeancourt A, Lu F,
Lemo21(DE3). J Mol Biol 423:648–659 Redzej A, Fronzes R, Orlova EV, Waksman G
10. Dubendorff JW, Studier FW (1991) (2014) Structure of a type IV secretion system.
Controlling basal expression in an inducible T7 Nature 508:550–553
expression system by blocking the target T7 21. Durand E, Nguyen VS, Zoued A, Logger L,
promoter with lac repressor. J Mol Biol Péhau-Arnaudet G, Aschtgen MS, Spinelli S,
219:45–59 Desmyter A, Bardiaux B, Dujeancourt A,
11. Schwarz D, Junge F, Durst F, Frölich N, Roussel A, Cambillau C, Cascales E, Fronzes R
Schneider B, Reckel S, Sobhanifar S, Dötsch V, (2015) Biogenesis and structure of a type VI
Bernhard F (2007) Preparative scale expression secretion membrane core complex. Nature
of membrane proteins in Escherichia coli-based 523:555–560
continuous exchange cell-free systems. Nat 22. Bakelar J, Buchanan SK, Noinaj N (2016) The
Protoc 2:2945–2957 structure of the β-barrel assembly machinery
12. Guihard G, Boulanger P, Bénédetti H, Lloubes complex. Science 351:180–186
R, Besnard M, Letellier L (1994) Colicin A 23. Unger T, Jacobovitch Y, Dantes A, Bernheim
and the Tol proteins involved in its transloca- R, Peleg Y (2010) Applications of the
tion are preferentially located in the contact Restriction Free (RF) cloning procedure for
sites between the inner and outer membranes molecular manipulations and protein expres-
of Escherichia coli cells. J Biol Chem sion. J Struct Biol 172:34–44
269:5874–5880 24. Tabor S, Richardson CC (1985) A bacterio-
13. Cascales E, Lloubes R, Sturgis JN (2001) The phage T7 RNA polymerase/promoter system
TolQ-TolR proteins energize TolA and share for controlled exclusive expression of specific
homologies with the flagellar motor proteins genes. Proc Natl Acad Sci U S A 82:1074–1078
MotA-MotB. Mol Microbiol 42:795–807 25. Studier FW, Moffatt BA (1986) Use of bacte-
14. Celia H, Noinaj N, Zakharov SD, Bordignon E, riophage T7 RNA polymerase to direct selec-
Botos I, Santamaria M, Barnard TJ, Cramer tive high-level expression of cloned genes.
WA, Lloubes R, Buchanan SK (2016) Structural J Mol Biol 189:113–130
insight into the role of the Ton complex in 26. Narayanan A, Ridilla M, Yernool DA (2011)
energy transduction. Nature 538:60–65 Restrained expression, a method to overpro-
15. Pramanik A, Zhang F, Schwarz H, Schreiber F, duce toxic membrane proteins by exploiting
Braun V (2010) ExbB protein in the cytoplas- operator-repressor interactions. Protein Sci
mic membrane of Escherichia coli forms a stable 20:51–61
oligomer. Biochemistry 49:8721–8728 27. Stuchlík S, Turna J (1998) Overexpression of
16. Pramanik A, Hauf W, Hoffmann J, Cernescu M, the FNR protein of Escherichia coli with T7
Brutschy B, Braun V (2011) Oligomeric struc- expression system. Folia Microbiol (Praha)
ture of ExbB and ExbB-ExbD isolated from 43:601–604
 Membrane Protein Complex Purification 309

28. Doherty AJ, Connolly BA, Worrall AF (1993) and E1 interact with a component of their
Overproduction of the toxic protein, bovine translocation system. EMBO J 10:1989–1995
pancreatic DNaseI, in Escherichia coli using a 32. Derouiche R, Bénédetti H, Lazzaroni JC,
tightly controlled T7-promoter-based vector. Lazdunski C, Lloubes R (1995) Protein com-
Gene 136:337–340 plex within Escherichia coli inner membrane.
29. van den Ent F, Löwe J (2006) RF cloning: a TolA N-terminal domain interacts with TolQ
restriction-free method for inserting target and TolR proteins. J Biol Chem 270:
genes into plasmids. J Biochem Biophys 11078–11084
Methods 67:67–74 33. Konovalova A, Silhavy TJ (2015) Outer mem-
30. Jeong J, Yim H, Ryu J, Lee H, Lee J, Seen D, brane lipoprotein biogenesis: Lol is not the
Kang S (2012) One-step sequence- and end. Philos Trans R Soc Lond B Biol Sci
ligation-­independent cloning as a rapid and 370:1679
versatile cloning method for functional genom- 34. Ringquist S, Shinedling S, Barrick D, Green L,
ics studies. Appl Environ Microbiol 78: Binkley J, Stormo GD, Gold L (1992)
5440–5443 Translation initiation in Escherichia coli:
31. Bénédetti H, Lazdunski C, Lloubes R (1991) sequences within the ribosome-binding site.
Protein import into Escherichia coli: colicins A Mol Microbiol 6:1219–1229
 Chapter 25

Shearing and Enrichment of Extracellular Type IV Pili

Alba Katiria Gonzalez Rivera and Katrina T. Forest

Abstract
Pili are widespread among bacteria. Type IVa pili (T4aP) are associated with a variety of bacterial functions,
including adhesion, motility, natural transformation, biofilm formation, and force-dependent signaling.
In pathogenic bacteria, T4aP play a crucial role during infection and have been the subject of hundreds of
studies. Methods for the isolation and purification of T4aP were first described in the 1970s. Purified pili
have been used for studies of filament protein content, morphology, immunogenicity, post-translational
modifications, and X-ray crystallography. We detail a tried-and-true method of isolating large amounts of
native T4aP from bacterial surfaces. The method requires supplies and equipment that are available in
most microbiology labs.

Key words T4P, Fimbriae, Pili, Filament, Shear, Isolation

1  Introduction

Pili are hairlike appendages displayed by many pathogenic and

environmental bacteria. One well-studied class is the Type IV pili
(T4P). At approximately 6 nm in diameter, T4P filaments are thin-
ner than flagella and can reach many micrometers in length [1].
They are expressed in some species at cell poles (e.g., in Myxococcus
xanthus or Pseudomonas aeruginosa) [2, 3] and in others peritri-
chously (e.g., in Neisseria gonorrhoeae or Deinococcus geothermalis)
[4, 5]. Found in diverse Gram-negative and Gram-positive species
[6, 7], T4P are involved in various bacterial functions, including
adherence, microcolony formation, biofilm initiation, and long-­
range electron transfer [8]. One distinctive feature of these fila-
ments is the capacity of some to retract, a property that has been
associated with phage sensitivity, force generation, twitching motil-
ity, natural transformation, and virulence [8].
T4P from Gram-negative organisms can be subdivided into
two families; subtype a (T4aP) and subtype b (T4bP), with notable
differences both in gene organization and biochemical properties.
At the genome level, genes for T4aP subunits and assembly
machinery are located in multiple operons dispersed around the

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_25, © Springer Science+Business Media LLC 2017

311
312 Alba Katiria Gonzalez Rivera and Katrina T. Forest

chromosome, whereas those for T4bP cluster in a single genomic

region [9]. We describe a method for the isolation of T4aP that
capitalizes on their properties in two kinds of buffers [10]. T4aP
filaments can be isolated by disaggregation in high-pH, low salt
buffer and aggregation (due to bundling) in buffer with near-neu-
tral pH and physiological salt concentration. Cycles of these two
steps can increase the purity of a preparation because contaminants
that do not share these solubility properties are differentially lost.
This method is not ideal for the more hydrophobic T4bP, which in
some cases are successfully purified using an ammonium sulfate
precipitation procedure [11].
Depending on the application, pili may be further purified. For
example, to generate anti-pilus antibodies, it may be desirable to
remove lipopolysaccharides (LPS). This can be accomplished by
incubating the purified pili with polymixin B agarose. For high-­
resolution crystallographic studies of the full-length pilin mono-
mer, dissociation of filaments in nondenaturing detergent followed
by filtration is appropriate [12–15].

2  Materials

2.1  Growth 1. Bacterial strain: P. aeruginosa PAK/2Pfs or PAK ΔpilT

and Harvest of Piliated (see Notes 1 and 2).
Bacteria 2. 60 Tryptic soy agar (TSA) plates, 1.5% agar (see Notes 3 and 4).
3. Tryptic soy broth (TSB) (see Note 5).
4. 30 °C incubator (see Note 6).
5. Stereo microscope.
6. Glass spreader.
7. Inoculating turntable.
8. 50 mL disposable conical tube.
9. 70% v/v ethanol solution.
10. Bunsen burner.

2.2  Collection of Pili 1. High-speed floor model centrifuge and appropriate rotor
(e.g., Beckman JA25.50).
2. Oak Ridge centrifuge tubes with O-ring screw caps (maxi-
mum relative centrifugal force (rcf) > 17,500 × g). These are
needed to handle Biosafety Level-2 (BSL-2) organisms.
3. 250-mL beaker.
4. Magnetic stir bar and magnetic stir plate.
5. Parafilm.
6. 25 and 5 mL serological pipets.
 Isolation of T4P 313

7. Plastic transfer pipets.

8. Pilus disaggregation buffer (PDB): 1 mM dithiothreitol
(DTT), 150 mM ethanolamine, pH 10.5 (4 °C). DTT must
be added shortly before use.

2.3  Purification 1. Centrifuge tubes (maximum rcf > 20,000 × g). 50 mL dispos-

of Pili able conical centrifuge tubes are convenient since at this stage,
BSL-2 safety containment is not required. Standard Oak Ridge
tubes are fine, too.
2. Pilus bundling buffer (PBB): 150 mM NaCl, 0.02% NaN3,
50 mM Tris–HCl, pH 7.5 (4 °C).
3. Dialysis tank or large flask (at least 4 L).
4. Dialysis tubing (molecular weight cut-off >3.5 kDa, 29 mm
diameter).

2.4  Assessment 1. 16% Tricine-SDS-PAGE Gel (see Note 7).

of Results

3  Methods

3.1  Growth 1. To isolate single colonies of P. aeruginosa, streak bacteria from

and Harvest of Piliated frozen stock stored at −80 °C onto a TSA plate. Incubate
Bacteria (see Note 1) plate for 24 h at 30 °C.
2. Identify single colonies with round morphology and smooth
margins using a stereo microscope (Fig. 1, and see Note 8).
Streak four colonies onto fresh TSA plates. Incubate for 24 h
at 30 °C.
3. Select a patch of cells with a sterile swab and resuspend in TSB
to reach an optical density at a wavelength of 600 nm (OD600)
of 8.0. If at this stage there is already sufficient volume for

Fig. 1 Colony morphologies. P. aeruginosa strain K ΔpilT is plated on TSA plates.

Piliated colonies have a smooth, domed appearance (black arrow). Avoid flat,
spreading colonies (white arrow)
314 Alba Katiria Gonzalez Rivera and Katrina T. Forest

spreading 50 μL of this bacterial suspension onto 50 plates of

TSA (that is, 2.5 mL), skip steps 4–6.
4. Grow a robust lawn of bacterial cells by inoculating 100 μL of
a cell suspension of OD600 = 8.0 onto five TSA plates. Grow
bacteria for 24 h at 30 °C.
5. Using a glass spreader and a turntable, remove bacterial lawns
from all five TSA plates.
6. Transfer bacteria into a 50 mL conical tube with 5 mL
TSB. Use a sterile transfer pipet to scrape cells from the glass
spreader into the tube as well as to gently but thoroughly
resuspend bacteria. Adjust TSB volume so that the suspension
has an OD600 of 8.0.
7. Inoculate 50 μL of this cell suspension onto each of the 50
TSA plates and spread using a glass spreader and a turntable.
Incubate plates for 24 h at 30 °C.
8. Collect bacterial lawns as just described using glass spreader
and turntable. Ideally these lawns should be confluent and
sticky; they will come off the plate in a goopy clump. We find
it convenient to use the same glass spreader for approximately
four plates before transferring bacteria into a 250 mL beaker
with 25 mL ice-cold PDB. Use transfer pipet to remove bac-
teria from glass spreader and to resuspend cells in PDB. Keep
adding PDB to suspended bacteria from all 50 plates. A total
of 75 mL PDB should be sufficient. This will be suspension 0
(S0) (Fig. 2). During the following procedures, keep bacterial
suspension and pilus suspension on ice at all times.
9. To dissociate large bacterial clumps, use a transfer pipet and
then a 10 mL serological pipet to gently pull cells up and
down to achieve a uniform suspension without large clumps.
Be patient as bacterial resuspension will increase pilus yield;
however, overly vigorous mixing will lyse cells and contami-
nate the final pilus preparation with other proteins.
10. After dissociating large bacterial clumps, add a magnetic stir
bar and cover the beaker with parafilm. Then stir the bacterial
suspension S0 at medium-low speed in cold room for 1 h.

3.2  Collection of Pili 1. Transfer samples to centrifuge tubes with O-ring screw caps.
Fill only half of the centrifugation tube to allow good vortexing
action.
2. Shear pili from cells by vortexing the cell suspension three
times in 1 min bursts at maximum strength. Cool on ice for
2 min between vortexing steps. Collect sample for sodium
dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis
(PAGE) analysis. In choosing a volume to load into gel, keep
in mind this solution will have a high protein concentration.
 Isolation of T4P 315

Resuspend Bacteria in PDB

S0

Shear pili
by vortexing

Remove cells
by centrifugation

P1 S1

Dialyze against PBB

Collect pili
by centrifugation

S2
P2

Resuspend pili
in PDB

Optional
Centrifugation S3

P3
Dialyze against PBB

Collect pili
by centrifugation

S4 P4

Fig. 2 Workflow for T4aP isolation. Bacteria are suspended in PDB and pili are
sheared by vortexing (S0). Cell debris and bacteria are removed by centrifugation
(P1). The remaining solubilized pili (S1) are aggregated by dialyzing against PBB
and collected by centrifugation (P2). For purification purposes, pili in P2 are
resuspended in PDB (S3) and aggregated by another dialysis against PBB (P4).
Cycles of resuspension and aggregation can be repeated to improve pilus purity,
although at a loss of overall yield

3. Sheared pili are now in suspension. To remove cells and cell

debris, combine samples to fill and balance centrifuge tubes,
and centrifuge samples at 15,000 × g for 20 min (see Note 9).
316 Alba Katiria Gonzalez Rivera and Katrina T. Forest

4. Carefully remove supernatant using a serological pipet and

transfer to clean centrifuge tubes with O-ring screw caps. Do
not disrupt pellet (P1).
5. Centrifuge the supernatant a second time to remove residual
cells at 15,000 × g for 10 min (see Note 9). Transfer superna-
tant (S1) containing sheared pili to a clean disposable tube. Set
aside a sample of the supernatant for analysis.
6. Prepare a 16% Tricine-SDS-PAGE gel following ref. 16.
7. Run samples to confirm successful separation of pili from cell
fractions.

3.3  Purification 1. Prepare dialysis membrane according to manufacturer’s

of Pili instructions.
2. Load sheared pili sample S1 into dialysis tubing and dialyze
against 4 L cold PBB for 4 h at 4 °C. Stir at low speed to facili-
tate buffer exchange.
3. Change dialysis buffer and repeat dialysis procedure until pH
of sample reaches 7.5 (see Note 10).
4. Upon neutralization of pH during dialysis, the pilus filaments
will aggregate by bundling. Carefully remove sample from
dialysis tubing and transfer it to a centrifuge tube. For maxi-
mum yield, rinse the inside of the tubing with PBB and add to
sample. Keep an aliquot for analysis.
5. To collect aggregated pili, centrifuge sample at 20,000 × g for
40 min.
6. Remove supernatant (S2) using serological pipet. Invert tubes
and drain onto paper towel. At this step, the pellet (P2) con-
tains sheared pili, and its appearance should be white. (If this
pellet contains pink, it includes residual cells.) Resuspend pel-
leted pili (P2) in PDB, starting with ~3 mL. Add PDB as
needed to make a nonviscous solution.
7. Optionally, centrifuge to remove contaminants not yet in suspen-
sion, which will pellet (P3) while pili remain in supernatant (S3).
8. Repeat dialysis procedure, now dialyzing S3 against PBB.
9. Collect purified pili by centrifugation; remove supernatant S4,
and resuspend pellet (P4) in desired final buffer to reach
appropriate concentration (e.g., 5 mL) (see Notes 11 and 12).

3.4  Assessment 1. To monitor pilus purity, run pilus preparation samples in 16%
of Results Tricine-SDS-PAGE gel (Fig. 3a, Table 1) and visualize by
Coomassie Brilliant Blue or silver staining. The latter also
permits estimation of LPS contamination. Yields of ~0.2–
0.5 mg of pili per plate at the S3 step can be achieved;
although the overall yield of pilin falls with each cycle, the
purity increases (Fig. 3a, Table 1).
 Isolation of T4P 317

Table 1
Purification yields at each step

P3 S3 P4 S4
Total protein (mg) 1 0.9 16a 30.5 10 11.7 6.4 12.6
Purity (%) 84 92 71 88 81 98 70 82
Pilin yield/loss (mg) 1 0.5 11 26.8 8 11.5 4.5 14.5
Approximate purity was estimated by image analysis of the Coomassie-stained gel in Fig. 3a using ImageJ [23]. Protein
concentration was estimated using OD280 and a calculated extinction coefficient for pilin of 14,100/M/cm (as calcu-
lated using ExPasy ProtParam [24]). We present data for two representative examples of yield from the protocol
described here. The left-hand columns are based on a purification by a relatively inexperienced researcher using 44
plates with volumes exactly as described in protocol (and shown in Fig. 3a). The right-hand columns are based on a
purification by a more practiced investigator from 32 plates with correspondingly reduced volumes
a
Note in this case the S3 concentration was estimated from the dialysate after S3 dialysis against PBB rather than in the
initial S3. Thus, this value is an underestimate of the true yield at the S3 stage

Fig. 3 Pili isolated from P. aeruginosa strain K ΔpilT. (a) Proteins were separated
in 16% Tricine-SDS-PAGE gel (molecular mass markers are in kDa). (b) Whole
cells and (c) purified pili (plus contaminating flagella) analyzed by negative stain-
ing and transmission electron microscopy (scale bars: 100 nm)

2. Negative staining can be used for the visualization of pilus

filaments using transmission electron microscopy (Fig. 3b, c).

4  Notes

1. P. aeruginosa must be handled in a BSL-2 laboratory. These

procedures will produce aerosols that could contain patho-
gens. This method will yield high amounts of biohazardous
waste. All steps here should be carried out under BSL-2 condi-
tions with appropriate personal protection and caution against
aerosol release of bacteria.
2. PAK/2Pfs is a strain with nonretractile pili [17] caused by a
mutation in the pilT gene [18]. PAK/2Pfs is available from
318 Alba Katiria Gonzalez Rivera and Katrina T. Forest

ATCC as strain 53,308 and is also known as PAK2.2 [19].

Strains that lack the PilT pilus retraction motor express higher
amounts of pili than retraction-proficient strains and thus are
commonly used for pilus purification, with yields up to 10
times higher than wild-type strains [17, 20]. We have achieved
similar results with a PAK derivative carrying an in-frame dele-
tion in the pilT gene (a gift of Dr. Stephen Lory, Harvard
Medical School). In this chapter, data are presented for the
PAK ΔpilT strain.
3. Prepare TSA following manufacturer’s instructions or by mix-
ing 15 g tryptone, 5 g soytone, 5 g NaCl, and 15 g agar per
liter of purified water. Autoclave to sterilize. Pour 20–22 mL
into 100 × 15 mm Petri dishes. Allow plates to dry (with lids
on) at room temperature for 2–3 days.
4. In our experiments, TSA or LB in 1.5% agar plates yield equiv-
alent amounts of pili based on assessment by SDS-PAGE.
5. Prepare TSB media as directed by manufacturer or dissolve
and sterilize 17 g tryptone, 3 g soytone, 2.5 g d-glucose, 5.0 g
NaCl, and 2.5 g dipotassium hydrogen phosphate per liter of
purified water.
6. Avoid incubating plates in high humidity. If applicable, leave
out water tray.
7. To prepare Tricine-SDS-PAGE gel follow method of Schagger
[16]. We favor this recipe for separating low-molecular-weight
proteins; however, any standard SDS-PAGE protocol will
work.
8. This is true for the isolation of P. aeruginosa piliated but non-
twitching mutant strains. Colony morphology for piliated
strains varies among bacterial strains and species [21, 22].
9. The centrifugal force used for these cell-pelleting steps should
be high enough to pellet cells but not high enough to lyse cells
or pellet disaggregated pilus filaments. 15,000 × g is a good
guide; however, lower rcf (8–10,000 × g) for a longer time
(30 min) may be sufficient. It is worthwhile to follow pilin
protein by SDS-PAGE to ensure pili are not spinning down
with cells.
10. For best success, actually check the pH of the solution inside
the dialysis bag. The lowering of salt together with pH neu-
tralization is the fundamental basis of the method, and we
have found that failure to reach equilibrium at this stage leads
to very low pilus yield.
11. To achieve the desired purity, repeat resuspension and aggre-
gation cycles. One modification to this procedure is at the last
step, where the cold solution is brought to a final concentration
 Isolation of T4P 319

of 20% ammonium sulfate and stirred for 2 h to precipitate pili

while leaving contaminating flagellar filaments in solution [20].
12. Depending on the application, one can remove residual LPS
using polymixin B agarose beads (our unpublished result).

Acknowledgements

We are grateful to Dr. Nicole Koropatkin for purifying grams of pili

and optimizing this protocol in the process, and to Dr. Lisa Craig
for many years of collegial interactions and helpful suggestions for
this chapter.

References
1. Hansen JK, Forest KT (2006) Type IV pilin bacterial pili in disease, purification and proper-
­
structures: insights on shared architecture, ties of Gonococcal pilus vaccine for Gonorrhea.
fiber assembly, receptor binding and type II In: Brooks GF (ed) Immunobiology of
secretion. J Mol Microbiol Biotechnol 11: Neisseria gonorrhoeae: proceedings of a confer-
192–207 ence held in San Francisco, CA. American
2. MacRae TH, Dobson WJ, McCurdy HD Society for Microbiology, Washington, DC
(1977) Fimbriation in gliding bacteria. Can 11. Li J, Lim MS, Li S, Brock M, Pique ME,
J Microbiol 23:1096–1108 Woods VL Jr, Craig L (2008) Vibrio cholerae
3. Henrichsen J, Blom J (1975) Examination of toxin-coregulated pilus structure analyzed by
fimbriation of some gram-negative rods with hydrogen/deuterium exchange mass spec-
and without twitching and gliding motility. trometry. Structure 16:137–148
Acta Pathol Microbiol Scand B 83:161–170 12. Forest KT, Dunham SA, Koomey M, Tainer JA
4. Swanson J (1973) Studies on gonococcus (1999) Crystallographic structure reveals phos-
infection. IV. Pili: their role in attachment of phorylated pilin from Neisseria: phosphoserine
gonococci to tissue culture cells. J Exp Med sites modify type IV pilus surface chemistry and
137:571–589 fibre morphology. Mol Microbiol 31:743–752
5. Saarimaa C, Peltola M, Raulio M, Neu TR, 13. Parge HE, Bernstein SL, Deal CD, McRee DE,
Salkinoja-Salonen MS, Neubauer P (2006) Christensen D, Capozza MA, Kays BW, Fieser
Characterization of adhesion threads of TM, Draper D, So M (1990) Biochemical
Deinococcus geothermalis as type IV pili. purification and crystallographic characteriza-
J Bacteriol 188:7016–7021 tion of the fiber-forming protein pilin from
6. Imam S, Chen Z, Roos DS, Pohlschroder M Neisseria gonorrhoeae. J Biol Chem 265:
(2011) Identification of surprisingly diverse 2278–2285
type IV pili, across a broad range of gram-­ 14. Parge HE, Forest KT, Hickey MJ, Christensen
positive bacteria. PLoS One 6:e28919 DA, Getzoff ED, Tainer JA (1995) Structure
7. Melville S, Craig L (2013) Type IV pili in of the fibre-forming protein pilin at 2.6 Å reso-
Gram-positive bacteria. Microbiol Mol Biol lution. Nature 378:32–38
Rev 77:323–341 15. Craig L, Taylor RK, Pique ME, Adair BD,
8. Berry JL, Pelicic V (2015) Exceptionally wide- Arvai AS, Singh M, Lloyd SJ, Shin DS, Getzoff
spread nanomachines composed of type IV pil- ED, Yeager M, Forest KT, Tainer JA (2003)
ins: the prokaryotic Swiss Army knives. FEMS Type IV pilin structure and assembly: X-ray
Microbiol Rev 39:134–154 and EM analyses of Vibrio cholerae toxin-­
coregulated pilus and Pseudomonas aeruginosa
9. Pelicic V (2008) Type IV pili: e pluribus unum? PAK pilin. Mol Cell 11:1139–1150
Mol Microbiol 68:827–837
16. Schagger H (2006) Tricine-SDS-PAGE. Nat
10. Brinton CC, Bryan J, Dillon J-A, Guerina N, Protoc 1:16–22
Jen Jacobson L, Labik A, Lee S, McMichael J,
Polen S, Rogers K, ACC T, SCM T (1978) 17. Bradley DE (1974) The adsorption of
Uses of pili in Gonorrhea control: role of Pseudomonas aeruginosa pilus-dependent bac-
320 Alba Katiria Gonzalez Rivera and Katrina T. Forest

teriophages to a host mutant with nonretractile motility is essential for virulence in

pili. Virology 58:149–163 Dichelobacter nodosus. J Bacteriol 190:
18. Whitchurch CB, Hobbs M, Livingston SP, 3323–3335
Krishnapillai V, Mattick JS (1991) 22. Meng Y, Li Y, Galvani CD, Hao G, Turner JN,
Characterisation of a Pseudomonas aeruginosa Burr TJ, Hoch HC (2005) Upstream migra-
twitching motility gene and evidence for a spe- tion of Xylella fastidiosa via pilus-driven twitch-
cialised protein export system widespread in ing motility. J Bacteriol 187:5560–5567
eubacteria. Gene 101:33–44 23. Schneider CA, Rasband WS, Eliceiri KW
19. Bradley DE (1980) A function of Pseudomonas (2012) NIH Image to ImageJ: 25 years of
aeruginosa PAO polar pili: twitching motility. image analysis. Nat Methods 9:671–675
Can J Microbiol 26:146–154 24. Artimo P, Jonnalagedda M, Arnold K, Baratin
20. Frost LS, Paranchych W (1977) Composition D, Csardi G, de Castro E, Duvaud S, Flegel V,
and molecular weight of pili purified from Fortier A, Gasteiger E, Grosdidier A,
Pseudomonas aeruginosa K. J Bacteriol Hernandez C, Ioannidis V, Kuznetsov D,
131:259–269 Liechti R, Moretti S, Mostaguir K, Redaschi N,
21. Han X, Kennan RM, Davies JK, Reddacliff LA, Rossier G, Xenarios I, Stockinger H (2012)
Dhungyel OP, Whittington RJ, Turnbull L, ExPASy: SIB bioinformatics resource portal.
Whitchurch CB, Rood JI (2008) Twitching Nucleic Acids Res 40:W597–W603
 Chapter 26

Blue Native PAGE Analysis of Bacterial Secretion

Complexes
Susann Zilkenat, Tobias Dietsche, Julia V. Monjarás Feria,
Claudia E. Torres-Vargas, Mehari Tesfazgi Mebrhatu, and Samuel Wagner

Abstract
Bacterial protein secretion systems serve to translocate substrate proteins across up to three biological
membranes, a task accomplished by hydrophobic, membrane-spanning macromolecular complexes. The
overexpression, purification, and biochemical characterization of these complexes is often difficult, imped-
ing progress in understanding the structure and function of these systems. Blue native (BN) polyacryl-
amide gel electrophoresis (PAGE) allows for the investigation of these transmembrane complexes right
from their originating membranes, without the need for long preparative steps, and is amenable to the
parallel characterization of a number of samples under near-native conditions. Here we present protocols
for sample preparation, one-dimensional BN PAGE and two-dimensional BN/sodium dodecyl sulfate
(SDS)-PAGE, as well as for downstream analysis by staining, immunoblotting, and mass spectrometry on
the example of the type III secretion system encoded on Salmonella pathogenicity island 1.

Key words Bacterial secretion systems, Membrane proteins, Blue native polyacrylamide gel electro-
phoresis, Two-dimensional polyacrylamide gel electrophoresis, Sucrose gradients, Bacterial cell frac-
tionation, Type III secretion systems, Salmonella typhimurium, Escherichia coli

1  Introduction

Blue native (BN) polyacrylamide gel electrophoresis (PAGE) is an

electrophoretic method originally developed by Schägger and von
Jagow to investigate the composition of protein complexes of the
respiratory chain from isolated mitochondrial membranes [1].
Since its introduction, it has been widely adopted to characterize
individual membrane protein complexes like the bacterial Sec and
Tat translocons [2, 3] or mitochondrial and chloroplast import
complexes [4, 5], as well as to assess the compositions of global
complexomes in wild-type conditions and upon perturbation [6–
10]. Lately it has also been recognized as a suitable tool for the
elucidation of the composition and assembly of bacterial protein
secretion systems: T4SS [11, 12], T3SS [13–17], and T7SS [18].

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_26, © Springer Science+Business Media LLC 2017

321
322 Susann Zilkenat et al.

BN PAGE relies on the mild extraction of membrane proteins

and membrane protein complexes by nonionic detergents, which
preserve the native protein conformation [19]. Charging and elec-
trophoretic migration of the extracted proteins is facilitated by the
adsorption of the anionic, water-soluble blue dye coomassie G to
the hydrophobic regions of the extracted membrane proteins,
which are then separated based on their complex size in gradient
gels of a Tricine-based PAGE.
Here we explain the use of BN PAGE for the characterization
of bacterial secretion systems on the example of the needle com-
plex of the Salmonella typhimurium type III secretion system
encoded on Salmonella pathogenicity island 1, a >180 component
complex of 4.5 MDa [17, 20, 21]. We provide protocols for the
preparation of samples from whole bacterial cells, crude and puri-
fied membranes, and immunoprecipitated material, for BN PAGE–
based separation of complexes using one-dimensional BN PAGE
or two-dimensional BN/ sodium dodecyl sulfate (SDS)-PAGE,
and for the detection and analysis of separated complexes by coo-
massie and silver staining, immunoblotting, and mass spectrometry
(MS). The individual protocols are provided as modules that can
be combined freely according to individual needs.

2  Materials

2.1  Sample 1. Buffer K: 50 mM triethanolamin (TEA), 250 mM sucrose, 1

Preparation mM ethylenediaminetetraaceticacid (EDTA) (see Note 1), pH
7.5 (adjusted with acetic acid (HAc)). Store at 4 °C.
2.1.1  Sample
Preparation General 2. Lysozyme solution: 10 mg/mL in deionized distilled water
Materials (ddH2O). Store aliquots of 100 μL at −20 °C.
3. Phosphate buffered saline (PBS): 137 mM NaCl, 2.7 mM
KCl, 4.3 mM Na2HPO4, 1.47 mM KH2PO4, pH 7.4, adjusted
with 1 M NaOH.
4. DNase solution: 10 mg/mL in 1× PBS. Store aliquots of 100
μL at −20 °C.
5. 1 M MgSO4 in ddH2O. Store at room temperature.
6. Complete protease inhibitor cocktail (without EDTA).
7. ACA750: 750 mM aminocaproic acid in ddH2O.
8. Liquid nitrogen.
9. 10% (w/v) n-dodecyl-β-d-maltoside (DDM).
10. BN loading buffer: 5% Serva Blue G (see Notes 2 and 3), 250
mM aminocaproic acid, 25% glycerol in ddH2O.

2.1.2  Extraction 1. Common material from Subheading 2.1.

of Membrane Proteins 2. Glass beads, acid washed, 150–212 μm.
from Crude Bacterial
Membrane Preparations
 Blue Native PAGE 323

Table 1
Sucrose solutions for two gradients

Percentage of Buffer
sucrose (%) Buffer 2× M 57% sucrose 1× M
(w/w) Sucrose (g) (mL) solution (mL) (mL)
57 22.8 17.2
32 12.8 10.0

Percentage of Buffer
sucrose (%) Buffer 2× M 55% sucrose 1× M
(w/w) Sucrose (g) (mL) solution (mL) (mL)
55 19.25 13.5
50 4.55 0.45
45 4.1 0.91
40 3.6 1.36
35 3.2 1.8
30 2.7 2.3

3. Homogenizer, e.g., SpeedMill Plus (Analytik Jena) or

FastPrep-­24 (MPBio).
4. Ultracentrifuge.

2.1.3  Membrane 1. Common material from Subheadiing 2.1.

Fractionation by Sucrose 2. Buffer 2× M: 100 mM TEA, 2 mM EDTA (see Note 1), pH
Density Gradient 7.5 adjusted with acetic acid (HAc).
Centrifugation
3. Sucrose solutions (see Notes 4 and 5) for two gradients in 14
× 89 mm tubes; see Table 1.
4. Buffer L: 50 mM TEA, 250 mM sucrose (see Note 1), pH 7.5
(adjusted with HAc). Store at 4 °C.
5. French press.
6. SW 41 Ti rotor (Beckman) for ultracentrifuge.
7. Dounce homogenizer.
8. Seton 14 × 89 mm open-top polyclear ultracentrifuge tubes for
SW 41 Ti rotor (see Notes 6–8) (optional; see Subheading 3.1.4).
9. Syringes and hypodermic needles (optional; see Subheading
3.1.4).
10. Gradient station (Biocomp Instruments), including accessories
(optional; see Subheading 3.1.4).
11. Bicinchoninic acid (BCA) protein assay.
324 Susann Zilkenat et al.

2.1.4  Immuno-­ 1. Anti-FLAG M2 affinity gel (see Note 9).

precipitation of Membrane 2. 3× FLAG peptide (see Note 9).
Protein Complexes
3. Rotating wheel.

2.2  Blue Native PAGE 1. Precast native polyacrylamide gradient gels, e.g., Novex®
NativePAGE™ Bis–Tris or SERVAGel™ N gel systems.
2.2.1  One-Dimensional
BN PAGE Using Precast 2. 10× BN anode buffer: 500 mM Bis-Tris-HCl, pH 7.0. Store at
Mini Gels 4 °C.
3. 10× BN cathode buffer A: 500 mM Tricine, 150 mM Bis–Tris,
0.2% (w/v) Serva Blue G. Do not adjust pH and store at 4 °C.
4. 10× BN cathode buffer B: 500 mM Tricine, 150 mM Bis–Tris.
Do not adjust pH and store at 4 °C.
5. NativeMark™ Unstained Protein Standard (Thermo Fisher).

2.2.2  Two-Dimensional 1. Hoefer SE600 or SE660 vertical electrophoresis unit.

BN/SDS PAGE 2. GelBond PAG film (Lonza).
3. Double-sided Scotch tape (Tesa Photo® Film) (see Note 10).
4. Gel seal.
5. 3× BN gel buffer: 1.5 M ACA750, 150 mM Bis–Tris–HCl, pH
7.0. Store at 4 °C
6. Acrylamide 30% T, 3% C.
7. Ammonium persulfate (APS): 10% (w/v) in ddH2O. Store at
−20 °C.
8. N,N,N,N′-tetramethyl-ethylenediamine (TEMED). Store at
4 °C.
9. 80% (v/v) glycerol in ddH2O. Autoclave and store at room
temperature.
10. Gradient maker, e.g., Hoefer SG30 or SG50.
11. Peristaltic pump including tubing, e.g., GE Healtcare P-1.
12. 10× BN anode buffer: 500 mM Bis–Tris–HCl, pH 7.0. Store
at 4 °C.
13. 10× BN cathode buffer A: 500 mM Tricine, 150 mM Bis–Tris,
0.2% (w/v) Serva Blue G. Do not adjust pH and store at 4 °C.
14. 10× BN cathode buffer B: 500 mM Tricine, 150 mM Bis–Tris.
Do not adjust pH and store at 4 °C.
15. NativeMark™ Unstained Protein Standard (Thermo Fisher).
16. Acrylamide 30% T, 2.6% C.

17.
10% (w/v) SDS solution in ddH2O. Store at room
temperature.
18. SDS stacking gel buffer: 0.5 M Tris–HCl, pH 6.8.
19. SDS resolving gel buffer: 1.5 M Tris–HCl, pH 8.8.
 Blue Native PAGE 325

20. 10× SDS-PAGE running buffer: 0.25 M Tris, 1.92 M glycine,

1% (w/v) SDS.

21.
SDS equilibration buffer: 2% (w/v) SDS, 1% (v/v)
β-mercaptoethanol, 50 mM Tris–HCl, pH 6.8.
22. Low melting agarose solution: 1% (w/v) low melting agarose,
0.5% (w/v) SDS, a few grains of bromphenol blue, 50 mM
Tris–HCl, pH 6.8.
23. Razor blades, scissors.
24. Filter paper.
25. Protein standard.

2.3  Protein 1. Fixing solution: 50% (v/v) ethanol, 3% (w/v) phosphoric acid
Detection, in ddH2O.
and Analysis 2. Serva Blue G (see Note 2).
2.3.1  Colloidal 3. ddH2O.
Coomassie Staining 4. Neuhoff’s solution: 16% (w/v) ammonium sulfate, 25% (v/v)
methanol, 5% (v/v) phosphoric acid, fill up to 100% with ddH2O.
5. Storage solution: 5% (v/v) glacial acetic acid in ddH2O.

2.3.2  Silver Staining Because silver staining is extremely sensitive to trace impurities in
water, it is strongly recommended to use high-quality ddH2O in all
recipes and protocol steps.
1. Fixing solution: 45% (v/v) methanol and 5% (v/v) glacial ace-
tic acid in ddH2O.
2. Sensitizing solution: 0.02% (w/v) sodium thiosulfate in
ddH2O. Always prepare fresh.
3.
Silver nitrate solution: 0.1% (w/v) silver nitrate in
ddH2O. Always prepare fresh.
4. Developing solution: 2% (w/v) sodium carbonate and 0.04%
(v/v) formalin in ddH2O. Prepare this solution the day of use
and add formalin to the developer at most 1 h before use.
5. Stop solution: 1% (v/v) glacial acetic acid in ddH2O.
6. Plastic containers: Polyethylene trays are recommended (see
Note 11).
7. Rocking table.
8. Special waste container for silver nitrate waste.

2.3.3  Immunoblotting 1. 10× SDS-PAGE running buffer: 0.25 M Tris base, 1.92 M gly-
Using Dual-Color Detection cine, 1% (w/v) SDS.
2. 10× transfer buffer: 0.25 M Tris base, 1.92 M glycine, 0.05%
SDS.
3. Polyvinylidene fluoride (PVDF) membrane.
4. Wet blot unit.
326 Susann Zilkenat et al.

5. 100% methanol.
6. 10× Tris-buffered saline (TBS): 200 mM Tris-HCl, 1.5 M
NaCl, pH 8.0.
7. TBS-T: TBS/0.05% Tween 20.
8. 0.1% Ponceau S, 5% acetic acid in ddH2O.
9. Primary antibodies.
10. Secondary DyLight antibodies (Thermo Fisher).
11. LiCor Odyssey infrared scanner.
12. Image Studio software (LiCor) or Quantity One software
(Bio-Rad).

2.3.4  Preparation of BN 1. 5% HAc (v/v).

PAGE-Separated 2. 100% ethanol.
Complexes for Analysis
by Mass Spectrometry
3. Powder-free gloves (newly opened box).
4. Light table.
5. Razor blades.
6. Tweezers.

3  Methods

Methods are divided into sample preparation (Subheading 3.1),

BN PAGE (Subheading 3.2), and protein detection and analysis
(Subheading 3.3). Modules within these divisions can be com-
bined as needed to yield a complete protocol. Each module con-
tains a suggestion for suitable methodological combinations.

3.1  Sample This protocol serves to extract membrane protein complexes from
Preparation small amounts of whole bacterial cells. The extracted proteins are
best run on BN mini gels and analyzed by immunoblotting as
3.1.1  Extraction
described previously [13].
of Membrane Proteins
from Whole Bacterial Cells 1. All steps are to be performed at 4 °C or on ice unless otherwise
stated.
2. Harvest 0.4–0.6 optical density units (ODU, see Note 12) of
bacterial cells by centrifugation for 2 min at 5000 × g and 4 °C.
3. Aspirate supernatant.
4. Optional: Snap freeze cell pellet in liquid nitrogen and store
cell pellet at −80 °C (see Note 13). Thaw pellet on ice to
continue.
5. Resuspend bacterial cell pellet in 10 μL freshly prepared buffer
K, supplemented with protease inhibitor cocktail, 10 μg/mL
lysozyme, 10 μg/mL DNase, and 1 mM MgSO4 (see Note 14).
6. Incubate for 30 min on ice to allow for cell wall digestion.
 Blue Native PAGE 327

7. Add 70 μL buffer ACA750 and mix (see Note 15).

8. Freeze-thaw three times in liquid nitrogen 1 min/20 °C water
1 min.
9. Add 10 μL freshly prepared 10% DDM and mix to allow for
extraction of membrane proteins (see Note 16).
10. Incubate for 1 h on ice with occasional mixing. Alternatively,
place into shaker for 1.5 mL reaction tubes, cooled down to
4 °C, and mix at 1000 rpm.
11. Spin sample for 20 min at 20,000 × g and 4 °C to pellet unsol-
ubilized material (see Note 17).
12. Transfer 45 μL of supernatant to new 1.5 mL tube containing
5 μL BN loading buffer and mix.
13. Load 25 μL of the suspension per well of a BN mini gel (see
Subheading 3.2.1).

3.1.2  Extraction This protocol describes the preparation of small amounts of crude
of Membrane Proteins bacterial membranes and the extraction of membrane protein com-
from Crude Bacterial plexes thereof (see Note 18). The extracted proteins are best run on
Membrane Preparations BN mini gels and analyzed by immunoblotting as described previ-
ously [16]; however, we have also had good experience with the
analysis of these preparations by two-dimensional BN/SDS-PAGE.
1. All steps are to be performed at 4 °C or on ice unless otherwise
stated.
2. Harvest 2–10 ODU of bacterial cells by centrifugation for
2 min at 5000 × g and 4 °C (see Notes 19 and 20).
3. Aspirate supernatant.
4. Optional: Snap freeze cell pellet in liquid nitrogen and store
cell pellet at −80 °C (see Note 13). Thaw pellet on ice to
continue.
5. Resuspend bacterial cell pellet in 750 μL freshly prepared buf-
fer K supplemented with protease inhibitor cocktail, 10 μg/
mL lysozyme,10 μg/mL DNase (see Note 14).
6. Incubate for 30 min on ice to allow for cell wall digestion.
7. Meanwhile, prepare 2 mL screw-cap tubes with 500 μL glass
beads. Label tubes on lid and on sides. Place tubes on ice for
cooling (see Notes 21 and 22).
8. Add 0.8 μL 1 M MgSO4 to each sample (final concentration).
9. Transfer cell suspension to a bead-containing tube and close
properly.
10. Break up cells by bead milling (2 min continuous mode using
SpeedMill Plus, 2 × 20 s at 4 m/s using FastPrep-24). For
cooling of samples, cool down the SpeedMill Plus sample
holder to −20 °C or place the samples on ice for 5 min after
first 20 s cycle when using the FastPrep-24.
328 Susann Zilkenat et al.

11. Pellet glass beads by centrifugation for 1 min at 1000 × g and

4 °C.
12. Transfer supernatant to a fresh 1.5 mL tube. Take care to pipet
carefully and transfer as few glass beads as possible in this step.
13. Add 1 mL buffer K to glass beads and mix vigorously to wash
off remaining material.
14. Pellet glass beads again by centrifugation for 1 min at 1000 × g
and 4 °C.
15. Transfer supernatant to 1.5 mL tube containing supernatant of
step 12. Take care to pipet carefully and transfer as few glass
beads as possible in this step.
16. Pellet glass beads and cell debris by centrifugation for 10 min
at 10,000 × g and 4 °C.
17. Transfer 1.3 mL of supernatant to 1.5 mL ultracentrifugation
tubes (see Note 23). Take care to avoid the transfer of glass
beads or cell debris.
18. Balance tubes to the same weight using buffer K for subse-
quent ultracentrifugation.
19. Pellet crude membranes by ultracentrifugation for 45 min at
55,000 rpm (135,520 × g) and 4 °C (Beckman TLA-55 rotor).
20. Discard supernatant.
21. Optional: Store membrane pellet at −80 °C. Thaw pellet on
ice to continue.
22. Resuspend membrane pellet in ACA750 (see Note 15) by
carefully pipetting up and down 40 times using 100–200 μL
pipette tips. 8 μL buffer per ODU harvested bacterial cells is
recommended, e.g., 24 μL for 3 ODU bacteria.
23. Add 1 μL freshly prepared 10% DDM per ODU harvested bac-
teria (3 μL for 3 ODU) to allow for extraction of membrane
proteins (see Note 16).
24. Incubate for 1 h on ice with occasional mixing. Alternatively,
place into shaker for 1.5 mL reaction tubes, cooled down to
4 °C, and mix at 1000 rpm.
25. Spin sample for 20 min at 20,000 × g and 4 °C to pellet unsol-
ubilized material (see Note 17).
26. Transfer 18 μL supernatant to new 1.5 mL tube containing
2 μL BN loading buffer and mix.
27. Load 10–20 μL of the suspension per well of a BN mini gel (see
Note 24). Larger amounts of up to 50 μL must be used for the
analysis of membrane protein complexes by two-­dimensional
BN/SDS-PAGE (see dedicated protocol in Subheading 3.2.2).
 Blue Native PAGE 329

3.1.3  Preparation This protocol describes the large-scale preparation of crude mem-
of Crude Membranes branes that are used for further membrane fractionation using
for Sucrose Density sucrose density gradient centrifugation described in Subheadings
Gradient Centrifugation 3.1.4 and 3.1.5 (see Note 25).
1. All steps are to be performed at 4 °C or on ice unless otherwise
stated.
2. Harvest 500–2000 ODU of bacterial cells by centrifugation
for 15 min at 6000 × g and 4 °C, e.g., in Beckman JLA-8.1000
(see Note 26).
3. Pour off supernatant.
4. Resuspend bacterial pellet in 35 mL cold PBS.
5. Transfer suspension to 50 mL Falcon tube.
6. Pellet bacteria by centrifugation for 10 min at 6000 × g and
4 °C.
7. Resuspend bacterial pellet in 10–15 mL buffer K supplemented
with protease inhibitor cocktail, 1 mM EDTA, 10 μg/mL
DNase, and 10 μg/mL lysozyme (all final concentrations).
8. Pass the cell suspension twice through a French press at
124 MPa (high position, 1000–1100 units).
9. Add MgSO4 to a final concentration of 1 mM to activate the
DNase. Mix.
10. Pellet cell debris by centrifugation for 20 min at 24,000 × g
and 4 °C (see Notes 27 and 28).
11. Transfer supernatant to suitable ultracentrifugation tubes or
bottles, e.g., for Beckman Type 45 Ti. Fill up bottles with buf-
fer K.
12. Pellet crude membranes by centrifugation for 45 min at
234,000 × g and 4 °C (45,000 rpm in Beckman Type 45 Ti).
13. Discard supernatant. Wipe off residual supernatant with lint-­
free tissue.
14. Resuspend crude membranes in 500 μL buffer 1× M. Use a cut
1 mL pipette tip to detach membrane from tube wall. Transfer
crude suspension to 1 mL dounce homogenizer. Homogenize
with loose piston 15 times.
15. Store crude membrane suspension on ice until further use and
proceed to alternatives (see Subheading 3.1.4 or 3.1.5).

3.1.4  Membrane This protocol describes the preparation of separated inner and
Fractionation by Sucrose outer membranes of Gram-negative bacteria using sucrose density
Density Gradient gradient centrifugation and extraction of membrane protein com-
Centrifugation Using plexes thereof. Sucrose gradient formation and fractionation are
a Biocomp Gradient Station described in two ways: either supported by a Biocomp gradient
station (this section, see Note 29) or manually without the need
for specialized equipment (see Subheading 3.1.5).
330 Susann Zilkenat et al.

The extracted proteins of inner or outer membrane fractions

are suitable for any of the described downstream analyses. We pre-
viously described the use of purified inner membranes for the anal-
ysis of complete membrane complexomes by two-dimensional
BN/SDS-PAGE [6–10], one-dimensional BN PAGE followed by
quantitative immunoblotting [13], and immunoprecipitation fol-
lowed by one-dimensional BN PAGE or two-dimensional BN/
SDS-PAGE[13–15].
1. Prepare Seton 14 × 89 mm open-top polyclear centrifuge tubes
for SW 41 Ti rotor. Mark at “half-full, long cap,” as described
in the gradient station manual.
2. Place tubes in tube holder of gradient station. Fill up to
2–3 mm above half-full mark with the 32% sucrose solution
using a syringe and the needle provided with the gradient sta-
tion (Fig. 1a).
3. Carefully underlay with 57% sucrose solution using a syringe and
the needle provided with the gradient station (see Note 30).
Bring up 57% sucrose solution exactly to the half-full mark. While
underlaying, keep tip of syringe just below phase boundary.
Strictly avoid having air bubbles come out of needle (Fig. 1b).
4. When removing the syringe needle, do so swiftly and make
sure to fix the piston to prevent further leakage of the 57%
solution during the movement.
5. Close the tubes with the long caps provided by the gradient
station. Avoid entrapment of air bubbles by pushing the cap
down at a slight angle so that air can leave through the ventila-
tion whole in the cap (Fig. 1c).
6. Level gradient platform according to the gradient station
manual.
7. Place tube holder at center of gradient platform.
8. Run gradient protocol SW41 LONG-SUCR-32-57%-w/w-­
2ST (step 1: 4:00 min, 70°, 30 rpm; step 2: 0:25 min, 85°, 25
rpm, check gradient station manual for programming
instructions).
9. Remove excess sucrose solution from cap (Fig. 1d).
10. Carefully place gradients on ice until further use. Proceed to
step 11 for the ultracentrifugation of sucrose gradients.
11. Carefully layer membrane suspension from step 15 in

Subheading 3.1.3 on top of continuous 32–57% (see Note 31,
Fig. 1e, g).
12. Tare two opposing tubes with buffer 1× M to <0.01 g.
13. Centrifuge gradients for 14 h at 41,000 rpm and 4 °C in a
Beckman SW41 Ti swing-out rotor (287,000 × g) (see Note
32). Set to slow acceleration and slow brake (see Note 33).
 Blue Native PAGE 331

Fig. 1 Preparation of sucrose gradients (a–e) using Biocomp Instruments Gradient Station or (f–g) by hand. For
details see Subheadings 3.1.4 and 3.1.5

14. Carefully place tubes on ice once the run is finished. Proceed
to fractionation using Biocomp gradient station fractionator
following step 15.
15. Label 1.5 mL tubes for collecting gradient fractions (13 per
sample). Place tubes on ice.
16. Visually check if membranes are clearly separated and mem-
brane bands are approximately at the same height. Make note
if this is not the case (see Note 34). Take a photo of your
gradient.
17. Program fractionation protocol: Speed 0.3 mm/s, distance 6.6
mm, number of fractions 12, rinse #0 (no automatic washing
steps). Check manual for programming instructions.
332 Susann Zilkenat et al.

18. Fill buffer reservoir with buffer 1× M.

19. Rinse tubing with buffer 1× M, then blow out remaining buf-
fer from tubing.
20. Pour ice-cold water into tube holder of fractionator. Ice water
should not reach rim of tube. Avoid ice clumps in holder that
might dislocate tube.
21. Place tube in tube holder, lock lid, and place tube holder on
fractionator as described in manual of gradient station. Take
care that the correct piston tip (SW 41 Ti) is attached to the
piston and its edge is undamaged.
22. Dim light of fractionator as low as possible to avoid heating up
sample.
23. Move down position of piston until first drop appears at end of
sample tubing. Reset position to zero (see Note 35).
24. Choose your program, put sample tubing into first collection
tube, and start fractionation.
25. Collect twelve 6.6 mm fractions (see Note 36). Note which
fractions contain the desired membrane bands.
26. Move up piston and collect remaining material (fraction 13).
27. Clean fractionator soon after completion of fractionation as
dried sucrose solution will clog tubing of fractionator.
28. To prepare membrane fractions for downstream experiments,
proceed to Subheading 3.1.6.

3.1.5  Membrane 1. Carefully pour gradient layers on top of each other, starting
Fractionation by Sucrose with 55% sucrose solution. Use 5 mL pipettes and Peleus ball
Density Gradient or a cut 1 mL pipette tip (Fig. 1f) (see Note 37).
Centrifugation Using 2. Carefully place gradients on ice until further use. Proceed to
a Manual Sucrose Step step 3 for ultracentrifugation of sucrose gradients.
Gradient
3. Carefully layer membrane suspension from step 15 in
Subheading 3.1.3 on top of continuous 30–55% step gradient
(see Note 31, Fig. 1e, g).
4. Tare two opposing tubes with buffer 1× M to <0.01 g.
5. Centrifuge gradients for 14 h at 41,000 rpm and 4 °C in a
Beckman SW41 Ti swing-out rotor (287,000 × g) (see Note
32). Set to slow acceleration and slow brake (see Note 33).
6. Carefully place tubes on ice once run is finished. Proceed to
fractionation of sucrose gradient by hand following step 7.
7. Label 1.5 mL tubes for collecting gradient fractions (13 per
sample). Place tubes on ice.
8. Visually check if membranes are clearly separated and mem-
brane bands are approximately at same height. Make note if
this is not the case. Take a photo of your gradient.
 Blue Native PAGE 333

Fig. 2 Collection of fractions of sucrose six-step gradient by hand. (a) Labeling of

fraction size. (b) Taking fractions from top by pipetting. (c) Taking individual frac-
tions using syringe. For details see Subheading 3.1.5

9. Mark fractions on sides of tubes: use a fine marker, mark steps

starting from meniscus of highest filled sample. Mark other
tubes identically to first one, even if meniscus is slightly lower.
Note which fractions contain membrane bands (Fig. 2a).
10. Collect fractions from top by carefully aspirating 1 mL from
just underneath meniscus using a 1 mL pipette. Transfer frac-
tions to collection tube. Use fresh pipette tip for every fraction
(Fig. 2b).
11. Alternative: Discrete bands can be fractionated using syringe
and a long needle. Place tip of needle just underneath the
desired band and aspirate until the entire band is taken up.
Transfer fractions to collection tube. Use fresh syringe and
needle for every fraction (Fig. 2c).
12. Keep collected fractions on ice until further use.
13. To prepare membrane fractions for downstream experiments,
proceed to Subheading 3.1.6

3.1.6  Preparing 1. Dilute sucrose at least 1:3 with buffer 1× M. Mix well.
Membrane Fractions 2. Pellet membranes by centrifugation for 45 min at 230,000 × g
from Sucrose Gradient and 4 °C (e.g., in Beckman Type 70.1 Ti).
Fractionations
for Downstream
3. Aspirate supernatant.
Experiments 4. Resuspend each membrane pellet in 200 μL buffer L.
5. Measure protein concentration by BCA assay according to
manufacturer’s instructions (see Note 38).
334 Susann Zilkenat et al.

6. Adjust protein concentration of desired fractions. Extraction

of membrane proteins from Salmonella or E. coli inner mem-
branes by DDM is optimal at a protein equivalent of 3 mg/mL
bovine serum albumin (BSA); however, the optimal value must
be determined empirically for every combination of detergent
and protein or complex of interest (see Note 39).
7. Optional: Store membranes at −80 °C until further use and
thaw on ice when needed.
8. Add 10% DDM (w/v) to desired amount of sample at 3 mg/
mL protein to achieve final concentration of 1% DDM.
9. Incubate rotating in cold room for 1 h to extract proteins from
membranes.
10. Spin sample for 20 min at 20,000 × g and 4 °C to pellet unsol-
ubilized material (see Note 17).
11. Transfer supernatant to fresh tube.
12. For subsequent BN PAGE, add BN loading buffer to final con-
centration of 0.5% coomassie dye.
13. Load 5–10 μL of the suspension per well of a BN mini gel (see
Note 24). Larger amounts of up to 50 μL must be used for the
analysis of membrane protein complexes by t­wo-­dimensional
BN/SDS-PAGE (see dedicated protocol in Subheading 3.2.2).

3.1.7  Immuno-­ This protocol describes the purification of DDM-solubilized mem-

precipitation of Membrane brane protein complexes by immunoprecipitation of a 3× FLAG
Protein Complexes Using epitope-tagged fusion protein (see Note 40). Purified complexes
3× FLAG Epitope Tags can be analyzed by one-dimensional BN PAGE or two-dimensional
BN/SDS-PAGE, respectively, as described previously [13–15, 17].
1. All steps are to be performed at 4 °C or on ice unless otherwise
stated.
2. Continue with membrane samples from step 8 in Subheading
3.1.6.
3. Incubate rotating in cold room for 30 min to extract proteins
from membranes.
4. Prepare 5 μL packed beads of anti-FLAG M2 affinity gel per
1 mg of membrane sample. Wash beads according to the man-
ufacturer’s instructions, twice with PBS and once with
PBS/0.1% DDM.
5. Add washed beads to sample and incubate for 4 h at 4 °C (see
Note 41).
6. Pellet beads by centrifugation for 1 min at 500 × g and 4 °C.
7. Wash beads 3× 15 min with 10-fold bead volume PBS/0.1%
DDM.
8. Resuspend pelleted beads in 1 mL PBS/0.1% DDM, transfer
to 1.5 mL tube, and pellet again.
 Blue Native PAGE 335

9. Aspirate supernatant completely.

10. Resuspend beads in one packed bead volume of PBS/0.04%
DDM/150 μg/mL 3× FLAG peptide.
11. Incubate beads rotating 30 min at 4 °C to elute immunopre-
cipitated protein.
12. Pellet beads by centrifugation for 1 min at 500 × g and 4 °C.
13. Transfer supernatant to fresh 1.5 mL tube.
14. Repeat steps 10–12 and pool supernatant.
15. Store at 4 °C until further use (see Note 42).
16. For subsequent BN PAGE, add BN loading buffer to a final
concentration of 0.5% coomassie dye.
17. Load suspension into a well of a BN gel. For downstream anal-
ysis by MS, it is advised to load the maximum amount the well
can hold.

3.2  Blue Native PAGE One-dimensional BN PAGE using precast mini gels accommo-
dates all previously described sample preparations but is best suited
3.2.1  One-Dimensional
for analytical amounts of membrane protein complexes and down-
BN PAGE Using Precast
stream immunoblotting. If preparative amounts of membrane
Mini Gels
complexes are needed, or the investigation of entire complexomes
is sought, larger gel format may be better suited. Preparation and
run of larger gel formats are described in Subheading 3.2.2. 3–12%
gradient gels are suited for the analysis of protein complexes
between 100 kDa and 5 MDa, 4–16% gradient gels for the analysis
of protein complexes between 30 kDa and 1 MDa.

1. Assemble mini PAGE (10 × 10 cm gels) equipment according

to manufacturer’s instructions.
2. Completely fill anode buffer tank with cold BN anode buffer.
3. Pour cold BN cathode buffer A into wells of the precast gel.
Do not fill up entire cathode buffer tank because the blue stain
makes loading of sample very difficult.
4. Load suitable protein standard, e.g., 10 μL NativeMark from
Thermo Fisher, into one well of the gel.
5. Load solubilized and coomassie-treated samples from step 12
from Subheading 3.1.1, step 26 from Subheading 3.1.2, step
12 from Subheading 3.1.6, or step 16 from Subheading 3.1.7
into the desired wells of the gel.
6. Carefully fill up cathode buffer tank with cold BN cathode
buffer A. Take care not to wash out samples from wells.
7. Run electrophoresis at 4 °C as follows: 130 V and 300 mA for
50 min, 250 V and 300 mA for 2 h or until thick blue front
runs out of gel (see Note 43).
336 Susann Zilkenat et al.

8. Optional (see Note 44): For downstream analysis by immu-

noblotting, pause the run when the blue front has run one-
third into the gel (typically after 50 min). Mix one volume of
BN cathode buffer A with nine volumes of BN cathode buffer
B. Empty cathode buffer tank completely. Refill cathode buffer
tank with prepared BN cathode buffer 1A/9B mix.
9. Disassemble gel once the run has completed and continue with
desired downstream analysis.

3.2.2  Two-Dimensional Two-dimensional BN/SDS-PAGE allows for the electrophoretic

BN/SDS-PAGE resolution of individual components of protein complexes. The
protocol presented here uses a previously described approach with
plastic-backed first-dimension BN gels that facilitates the transfer
of BN lane strips onto second-dimension gels [6, 22, 23] (please
also refer to these publications for illustrations of the methodol-
ogy) (see Notes 45 and 46). An exemplary result of a coomassie-­
stained two-dimensional BN/SDS-PAGE gel of inner membranes
of S. typhimurium wt and spaS (T3SS export apparatus compo-
nent) mutant is shown in Fig. 3.
1. Cut polyacrylamide-coated plastic film to size of glass plates.
Take care that the length of the plastic film perfectly fits the
length of the glass plate; the width of the film is allowed to be
somewhat smaller than the glass plate.

Fig. 3 Two-dimensional BN/SDS-PAGE showing coomassie-stained inner membrane proteins of S. typhimurium.

Left: S. typhimurium wild type; T3SS components are indicated. Right: S. typhimurium ∆spaS mutant; other
abundant and clearly observed complexes are indicated
 Blue Native PAGE 337

Table 2
Solutions for casting of BN gradient gel

Resolving gel Stacking gel

3% 12% 4% 16% 3%
ddH2O 6.0 mL 1.05 mL 5.6 mL – 5.1 mL
3× BN gel buffer 3.5 mL 3.5 mL 3.5 mL 3.5 mL 3.0 mL
Acrylamide (30% 1.05 mL 4.2 mL 1.4 mL 5.6 mL 0.9 mL
T, 3% C)
Glycerol (80%) – 1.75 mL – 1.4 mL –
TEMED 5.25 μL 5.25 μL 5.25 μL 5.25 μL 4 μL
APS (10%) 52.5 μL 52.5 μL 52.5 μL 52.5 μL 40 μL

2. Fix the plastic film on the lower edge of one of the glass plates
using double-sided Scotch tape. Make sure that the hydro-
philic side of the plastic film is facing up.
3. Place a few drops of ddH2O between the glass plate and plastic
film. Wipe tightly along film with tissue to remove excess water
and trapped air.
4. Slightly grease 1.0 mm spacers on one side with gel seal. Place
spacers with greased side on plastic film.
5. Finish assembly of gel casting assembly according to manufac-
turer’s instructions.
6. Prepare solutions for casting of BN gradient gel according to
recipes in Table 2. Do not add APS at this point (see Note 47).
7. Assemble gradient maker including magnet stirrer and peristal-
tic pump. Squeeze end of outlet tubing between glass plates of
gel casting assembly.
8. Add APS to gel solution of lower percentage, mix well, and
pour solution into distal chamber of gradient maker. Add APS
to gel solution of higher percentage, mix well, and pour solu-
tion into proximal chamber of gradient maker (see Note 48).
9. Switch on peristaltic pump. Open outlet valve and then
chamber-­connecting valve. Pour gradient gel (see Note 49).
10. When all the gel solution has entered the gel, add APS to stack-
ing gel solution, mix, and pour solution into proximal cham-
ber of gradient maker (see Note 50). Continue pouring gel.
11. Carefully place comb (10 or 15 wells) into stacking gel (see
Note 51). Take care not to destroy gradient.
12. Allow gel to polymerize for at least 4 h at room temperature.
Gels can be stored after polymerization up to 1 week (4 °C and
moist).
338 Susann Zilkenat et al.

13. Remove comb from gel, rinse gel and wells with ddH2O, and
assemble gel running equipment in cold room (see Note 52).
Fill anode buffer tank with 2 L cold BN anode buffer.
14. Pour cold BN cathode buffer A into wells of gel. Do not fill up
entire cathode buffer tank because the blue stain will make
loading of sample very difficult.
15. Load suitable protein standard, e.g., 10 μL NativeMark from
Thermo Fisher, into one well of gel (see Note 43).
16. Load solubilized and coomassie-treated samples from step 12
from Subheading 3.1.1, step 26 from Subheading 3.1.2, step
12 from Subheading 3.1.6, or step 16 from Subheading 3.1.7
into desired wells of gel.
17. Carefully fill up cathode buffer tank with cold BN cathode
buffer A. Take care not to wash out samples from wells.
18. Run electrophoresis at 4 °C as follows: 130 V and 30 mA for 1
h, 150 V and 30 mA for 14 h, or until thick blue front runs out
of gel (see Note 43).
19. Proceed to second dimension—SDS-PAGE (see Note 53).
Pour as many polyacrylamide gels (resolving and stacking gel)
as needed to analyze each lane of BN PAGE. Use a polyacryl-
amide concentration for resolving gel that allows resolution of
all proteins expected to compose complex of interest (e.g.,
12% polyacrylamide).
20. After pouring stacking gel, place a preparative comb (one long
well matching length and width of one lane of first-dimension
BN gel) in gel and allow polymerization at room temperature
(see Note 54).
21. Disassemble first-dimension BN PAGE assembly (obtained
from step 15). Remove glass plate facing gel (not plastic film)
and spacers.
22. Using razor blades, scrape off partition walls of wells and thickly
stained coomassie front if not everything has run out of gel.
23. Lift up plastic-backed gel and remove double-sided Scotch tape.
24. Using scissors, cut out lanes of first-dimension run.
25. Place lane strips into SDS equilibration buffer to denature pro-
tein complexes and reduce disulfide bonds. Incubate for
20 min (see Note 55).
26. Remove comb of second-dimension gel and overlay top of gel
with hot low-melting agarose solution.
27. Place lane strips of first-dimension gel onto second-dimension
gel (see Notes 56 and 57).
28. Drop 5 μL of protein size ladder onto a small piece of filter
paper (3 × 3 mm). Place marker-soaked filter paper on top of
second-dimension gel next to BN gel lane strip.
 Blue Native PAGE 339

29. Run electrophoresis at 4 °C as follows: 150 V and 300 mA for

1 h, 300 V and 300 mA, until coomassie front runs out of gel
(see Note 43).
30. Disassemble gel assembly, take out gels, and proceed with
downstream analysis as needed (see Subheading 3.3).

3.3  Protein Detection The described protocol according to Neuhoff et al. [24] allows for
and Analysis the highly sensitive detection of proteins by coomassie G.
3.3.1  Colloidal
1. After electrophoresis, place gel in a clean plastic container and
Coomassie Staining
rinse briefly in ddH2O. Reduce gel manipulation to minimum
and always wear powder-free gloves.
2. Decant ddH2O. Add enough fixing solution to completely
immerse gel and incubate o/n at room temperature with gen-
tle agitation (see Note 58).
3. Decant fixing solution and wash gel three times in ddH2O,
30 min for each wash.
4. Decant ddH2O. Equilibrate gel in Neuhoff’s solution for 1 h.
The exact volume of Neuhoff’s solution must be known in
order to add the right amount of Serva Blue G in the next step.
5. Add Serva Blue G powder to a final concentration of 0.1%
(w/v) (e.g., 0.1 g/100 mL Neuhoff’s solution) and stain for
up to 48 h (see Note 59).
6. Transfer gel into a clean container and wash several times with
ddH2O. Gently remove coomassie particles from gel surface.
7. Decant ddH2O and store gel in 5% acetic acid solution at 4 °C
until further usage (see Note 60).

3.3.2  Silver Staining (MS The described protocol according to Shevchenko et al. [25] allows
Compatible) for a highly sensitive detection of proteins.
1. After electrophoresis, place gel in tray and rinse briefly with
ddH2O. Always manipulate gel wearing powder-free gloves
that have been rinsed with deionized water to avoid fingerprint
contamination and pressure marks.
2. Decant ddH2O. Immerse gel in fixing solution for 20–30 min;
gently agitate (see Note 61).
3. To remove acid, rinse gel with ddH2O for at least 20 min (see
Note 62).
4. Soak gel in sensitizing solution for 1–2 min (see Note 63).
5. Decant sensitizing solution. Rinse gel twice with ddH2O,
1 min for each wash.
6. Pour out ddH2O and impregnate gel in chilled silver nitrate
solution for 20–40 min, preferably at 4 °C (see Note 64).
340 Susann Zilkenat et al.

7. Discard solution and rinse gel twice in water, 1 min for each
wash (see Note 65).
8. Develop by adding a sufficient amount of developing solution
to completely cover gel. Agitate until intensity of staining is as
desired (see Note 66).
9. Once the adequate degree of staining is achieved, decant devel-
oping solution. Add stop solution and incubate for 30 min.
Agitate slowly.
10. Store silver-stained gel in stop solution at 4 °C until scanning
or preparation of bands for MS.

3.3.3  Immunoblotting BN PAGE facilitates the analysis of protein complexes that often
Using Dual-Color Detection contain more than one kind of protein component. Quantitative
analysis of the composition of protein complexes by BN PAGE is
very powerful when a simultaneous detection of two different pro-
teins within the same complex band is achieved by dual-color
immunoblotting. The protocol described here uses infrared-­
fluorescent secondary antibodies and a LiCor Odyssey scanner.
Although the protocol is described for immunoblotting of BN
gels, it can be easily adapted for two-dimensional BN/SDS-PAGE
or regular SDS-PAGE. Figure 4 shows an example of a dual-color
immunoblotting of a BN PAGE-separated T3SS needle complex.
1. Equilibrate BN gels in SDS-PAGE running buffer for
15–20 min (see Note 67).
2. Assemble wet blot sandwich according to manufacturer’s
instructions. Use only PVDF membranes because coomassie G
binds irreversibly to nitrocellulose.
3. Transfer proteins according to the instructions of the wet blot
equipment manufacturer.
4. When the transfer is finished, take out the PVDF membrane,
place it in a dark container (see Note 68), and rinse it three or
four times in 100% MetOH to remove all bound coomassie G
(see Note 69).
5. Decant MetOH and rinse PVDF membrane with TBS until
wetted.
6. Stain PVDF membrane with Ponceau S solution to visualize
unstained native protein standard.
7. When sufficient staining is achieved, rinse PVDF membrane
with ddH2O until bands become visible and mark bands with
a pencil.
8. Destain PVDF membrane completely using ddH2O.
9. Block unoccupied binding sites of PVDF membrane by incubat-
ing in 5% skim milk/TBS or 3% BSA/TBS for 1 h (see Note 70).
10. Incubate PVDF membrane with primary antibody diluted to
desired concentration TBST (1 × TBS plus 0.05% Tween-20)
for 1 h (see Note 71).
 Blue Native PAGE 341

Fig. 4 Dual-color immunoblotting of a BN PAGE-separated T3SS needle complex.

Crude membranes of wild-type and mutant S. typhimurium were extracted with
DDM and analyzed by BN PAGE and immunoblotting. Component A was probed with
a primary rabbit antibody and a secondary DyLight goat anti-rabbit 680 nm (red).
The FLAG-tagged component B was probed with a primary anti-FLAG M2 mouse
monoclonal antibody and a secondary DyLight goat anti-mouse 800 nm (green)

11. Wash the PVDF membrane three times 15 min with TBST.
12. Incubate PVDF membrane with secondary infrared-­fluorescent
antibody diluted to desired concentration in TBST for 1 h (see
Note 72).
13. Wash PVDF membrane three times for 15 min with TBST.
14. Scan PVDF membrane in LiCor Odyssey scanner according to
manufacturer’s instructions.
15. Analyze protein bands using western blot analysis tool of
Image Studio software (LiCor). A typical result of dual-color
immunoblotting of a type III secretion system is shown in
Fig. 4.

3.3.4  Preparation of BN MS is well suited to detect the composition of protein complexes
PAGE-Separated separated by BN PAGE; however, it can be difficult to distinguish
Complexes for Analysis a true complex component from an unrelated comigrating protein.
by Mass Spectrometry Distinction can be achieved using suitable controls and bioinfor-
matic analysis, as described by Fischer et al. [15], but also by the
mass spectrometrical profiling of complete lanes of BN gels [26],
as described in what follows. Figure 5 shows the MS-based analysis
of lane profiles of immunoprecipitated T3SS needle complexes.
342 Susann Zilkenat et al.

Fig. 5 Example of MS–based analysis of a lane profile of BN PAGE-separated T3SS needle complex. BN lane
of immunoprecipitated S. typhimurium SPI-1 type III secretion needle complex stained with colloidal coo-
massie, divided into 20 gel slices. Protein intensity profiles based on quantification by MS is shown for indi-
cated proteins. InvA, SpaS, SpaP, and SpaQ are components of the export apparatus of the type III secretion
needle complex. InvG, PrgH, and PrgK are components of the needle complex base. PrgI and PrgJ are secreted
substrates of the type III secretion system forming the filament structures of the needle complex. All these
components form the stable complex that was co-immunoprecipitated. The true components belonging to this
complex can be deduced from the similar protein intensity profiles in slices 2 and 3. It is also apparent that a
large fraction of non-needle-complex-associated SpaS and InvA are present in the sample (slices 15–20).
InvC, SpaO, and OrgA are cytosolic components of the type III secretion system that are not stably associated
with the needle complex and just coprecipitate during purification. This is reflected by the sharp peaking of the
intensity of these proteins in slice 2 but not slice 3. The profile of the unrelated outer membrane protein OmpA
is shown as a control
 Blue Native PAGE 343

1. Make a picture of your stained BN gel before cutting any

bands.
2. To avoid contamination, clean surface used to cut protein
bands out of gel with fresh 100% ethanol. Work with fresh,
clean, powder-free gloves. Avoid touching outside of gloves
while putting them on.
3. Measure total length of BN lane and decide on a number of
bands.
4. Prepare a 1.5 mL tube for storage of each band (see Note 73).
Fill tubes with 0.5 mL 5% acetic acid (v/v).
5. Remove most of the 5% acetic acid from the translucent tray
the gel is stored in. Leave just enough to keep gel covered.
Place tray on a light table (see Note 74).
6. Clean a razor blade and tweezers with ethanol.
7. Use razor blade to make horizontal cuts (see Fig. 6a). Cut all
replicates at once by pressing the razor blade down over full
length of replicate lanes. Clean razor blade with ethanol in
between cuts (see Note 75).
8. Use a new, cleaned razor blade for cutting vertical lines, start-
ing at top (see Note 76).
9. After each cut, collect bands with tweezers and place them in
prepared tubes containing 5% acetic acid (Fig. 6b, c) (see Note
77). Always clean razor blades and tweezers in between cuts
using 100 EtOH.
10. Store cut bands at 4 °C until further processing for MS.

Fig. 6 Preparation of samples for MS-based analysis of lane profiles of BN PAGE-­

separated samples. (a) Cut horizontal lines while leaving border on both sides
intact. (b, c) Cut vertical lines from top to bottom and collect gel bands subse-
quently in 1.5 mL tubes. For details see Subheading 3.3.4
344 Susann Zilkenat et al.

4  Notes

1. Also include 1 mM dithiothreitol if reducing conditions are

desired.
2. Serva Blue G guarantees the highest quality of the coomassie
dye. If replaced by a dye of another vendor, make sure it is of
high quality.
3. It is best to add the dye 30 min before use. Vortex vigorously
to dissolve dye. Just before use, spin briefly to pellet unsolubi-
lized dye.
4. The solutions can be prepared in advance and stored at 4 °C but
should be at room temperature when making the gradient.
5. Depending on the organism, the percentage of the sucrose
gradient may have to be adjusted for an optimal separation of
inner and outer membranes.
6. Biocomp Instruments offers certified Seton tubes to guarantee
fine sealing when using the Biocomp gradient station.
7. Before making gradients, check each tube for prominent edges
at seams, differences in length or wall thickness (happens
rarely), and fissures.
8. For smaller gradients 13 × 51 mm tubes (Seton) for, e.g., SW
55 Ti rotor (Beckman) can be used.
9. Or similar depending on protein tag.
10. Choosing the right double-sided Scotch tape is crucial. We use
double-sided Scotch tape from Tesa. Other double-sided Scotch
tapes lose their adhesive properties during electrophoresis.
11. When used for the first time, clean containers exhaustively with
acetone and afterward with alcohol to remove traces of plasti-
cizers that may interfere with mass spectrometric analysis.
Make sure the containers are used exclusively for coomassie or
silver staining, and wash them with soap and ultrapure water
after every staining round. To ensure free movement while
shaking and complete immersion in staining solutions, the
bottom area of the container should be at least 20% bigger
than the area of the gel to be stained.
12. 1 ODU corresponds to 1 mL of bacterial culture at A600 = 1.0.
13. Snap freezing in liquid nitrogen or dry-ice ethanol helps to
prevent damaging of complexes by ice crystal formation.
14. The basic buffer recipe used for cell wall digestion may require
empirical evaluation depending on the bacterium or protein
complex of interest. We have also successfully used PBS, pH
7.4, supplemented with protease inhibitor, lysozyme, and
DNase. Alternative cell-wall-degrading enzymes may be needed
for other bacteria, e.g., lysostaphin for Staphylococcus aureus.
 Blue Native PAGE 345

Fig. 7 BN PAGE analysis of T3SS needle complex using a set of different deter-
gents. Crude membranes of S. typhimurium were extracted with indicated deter-
gents and subsequently analyzed by BN PAGE and immunoblotting. The
FLAG-tagged T3SS export apparatus component SpaS was probed with a pri-
mary anti-FLAG M2 mouse monoclonal antibody and a secondary DyLight 800
nm goat anti-mouse antibody. The band of SpaS in the type III secretion needle
complex is indicated

15. Aminocapronic acid was reported to facilitate extraction of

proteins. Its benefit at this step may be empirically evaluated.
Use PBS or your buffer of choice.
16. The detergent used for extraction may require empirical evalu-
ation depending on the protein complex of interest. In addi-
tion to DDM, we have had good experience with Cymal 4,
Cymal 5, DM, UDM, DMNG, LMNG, LDAO, and Triton
X-100 (see Fig. 7).
17. The speed of this spin depends on the molecular weight of the
complex of interest. T3SS needle complexes are already pel-
leted at speeds of 100,000 × g; hence, it is not advisable to
remove unsolubilized material at this speed. If the focus is on
smaller complexes, spins up to 30 min at 100,000 × g are
recommended.
18. Bacterial lysis is based on cell wall digestion and glass bead
milling, an approach suitable for the preparation of several
samples in parallel.
19. The expected yield of crude membranes is 10 μg protein per
ODU of bacterial cells (E. coli, Salmonella). Of this, about 50%
can be extracted by DDM. Loading 10 μg of crude membranes
per well is sufficient for analysis by BN mini gels; however, a
346 Susann Zilkenat et al.

larger preparation ensures a clearly visible pellet after

ultracentrifugation.
20. The volumes for glass bead milling do not depend on the
amount of bacterial cells when working in a range of 2–10
ODU.
21. It is important to use screw-cap tubes because snap-cap tubes
may pop open during bead milling.
22. During bead milling, the label may be abraded. Check which
labeling position is most suitable.
23. Do not transfer a higher volume since excess liquid may leak
out of the tube during ultracentrifugation.
24. Suitable amount needs to be empirically evaluated.
25. Bacterial lysis in this protocol is based on French pressing but
can be tailored to specific needs and laboratory equipment.
26. The maximum loading capacity of one gradient is crude mem-
brane extracted from about 2000 ODU of culture. If mem-
branes from more material need to be separated, it is better to
divide the sample on several gradients. Samples can be pooled
after step 4 from Subheading 3.1.6.
27. This spin can be done in 50 mL Falcon tubes if using the
Fiberlite F15-8x50C rotor. Otherwise, transfer lysate to suit-
able centrifugation bottles.
28. This speed will also remove a considerable amount of outer
membranes. If these are needed, spin at 8000 × g instead.
29. Use of a Biocomp gradient station results in six identical con-
tinuous gradients, which can be fractionated in defined steps
with very limited cross contamination of fractions. Both gradi-
ent formation and fractionation are highly reproducible across
different experiments. Buffers and gradients described in this
protocol were optimized for separating membranes of E. coli
and Salmonella, but the protocol can easily be adapted for
other organisms. The protocols are described for use with a
Beckman SW 41 Ti swing-out rotor but can be adapted to
other rotors accordingly.
30. After filling the syringe, clean the outside of the needle with a
lint-free tissue to prevent mixing of 57 and 32% sucrose
solutions.
31. Overfilling the tubes (less than 3 mm left at the top) can cause
spilling during the centrifugation step, which makes breaking
of the tubes more likely.
32. The swinging buckets for the centrifuge should be cleaned
before use.
33. If a tube breaks during the overnight step, you can collect the
membrane–sucrose mixture, dilute it with the buffer used dur-
 Blue Native PAGE 347

ing the membrane extraction (at least 1:2), and spin down the
crude membrane pellet again to repeat the gradient.
34. If the membrane bands are not at an equal height in the differ-
ent samples, the most common problem is the pipetting of the
gradient. Either the half-full mark for the wrong cap type
(long/short) was used or the sucrose solutions were not prop-
erly dissolved before layering the gradient. Accidental stirring
of the solutions while underlaying can also lead to aberrations
in gradient formation.
35. For the following samples: Move down position of piston until
position value is exactly 0.00. If you move too far, make note of
the position and reset to 0 to continue; however, fractions will
not be identical to the first sample. Do NOT press “UP” at this
point. Pressing “UP” will move the piston all the way to the top
again, mixing at least the upper part of your gradient.
36. Fractionation of a continuous gradient does not require rins-
ing of sample tubing after each fraction. If a discrete fraction is
targeted, rinsing and blowing off tubing before start of frac-
tionation is recommended.
37. The phase boundaries should be visible from an angle. If you
cannot see any boundaries at all, you have mixed the layers and
should make new gradients.
38. The BCA assay is largely tolerant to lipids in the sample and
therefore preferred over Bradford or Lowry assays. For better
comparability, always use the same standard protein to make a
standard curve, e.g., BSA.
39. The extraction power of a given detergent toward a protein of
interest can be tested by extracting crude membranes with a
protein concentration equivalent of 1–10 mg/mL BSA in
1 mg increments with 1% of a given detergent. The highest
concentration that is still correlated to a linear increase in
extracted protein (testable, for example, by immunoblotting)
is the optimal concentration to use for extraction.
40. Immunoprecipitation can be performed using suitable anti-
bodies or any other epitope tags that tolerate the detergents
used. In an optimal case, the 3× FLAG epitope-tagged fusion
protein is encoded in its natural context, e.g., on the chromo-
some or on its virulence plasmid. However, also expression of
plasmid-based artificial operons encoding all or some complex
partners of interest (of which one is 3× FLAG-tagged) is
possible.

41. During immunoprecipitation with beads, choose a tube
according to your sample volume. It should be one-third to
two-thirds full and rotated over its own head to guarantee a
good distribution of beads in the sample over the whole incu-
bation time.
348 Susann Zilkenat et al.

42. Freezing of solubilized complexes is not recommended as this

may cause damage.
43. Voltage clamp limits the run.
44. The coomassie G contained in the BN gel efficiently binds to
PVDF membranes upon transfer and blocks protein binding
sites. To enable downstream immunoblotting, a BN cathode
buffer is used after one-third of the run that contains only one-
tenth of the initial coomassie concentration.
45. The first dimension described in what follows utilizes a polyes-
ter backing to facilitate the first- to second-dimension transfer
of BN lane strips but can in principle also be performed as a
single one-dimensional gel without the polyester backing.
46. Here we describe this approach using 16 cm long gels run in a
Hoefer SE600 system while the original references describe
BN gels of 24 cm length followed by a second dimension in
GE Ettan Dalt tanks.
47. The acrylamide concentration of the BN gradient gel should
be chosen to suit the molecular mass of the analyzed complex:
For complexes beyond 1000 kDa, 3–12% gels should be used;
smaller complexes are separated best in 4–16% gels.
48. Be aware that polyacrylamide polymerization can be very quick
and clog the gradient maker and tubing when casting gels at
warm room temperatures (>25 °C) in the summer. Consider
casting gradient gels in the cold room if this is a problem.
49. Make sure the levels of the two gel solutions are at an equal
height during the pouring process. If necessary, tilt the gradi-
ent maker to put more hydrostatic pressure on the solution of
lower percentage.
50. Do not wait for the resolving gel to polymerize before casting
the stacking gel. The function of the stacking gel is only to
hold the comb, not to stack the sample proteins.
51. Best results are obtained when the comb is released from its
support and pushed into the gel solution such that the gel’s
entire surface is covered by the comb and does not contact any
air. Air inhibits polyacrylamide polymerization, in particular
when very low polyacrylamide percentages are used.
52. Putting some gel seal on top of the spacers helps to prevent
leakage of cathode buffer during the run.
53. For the second dimension the choice of electrophoresis equip-
ment is not critical as long as its width allows the placement of
one BN lane on the gel. The thickness of the second-­dimension
gel needs to be 1.5 mm (0.5 mm thicker than the first-­
dimension BN gel). For gel recipes, please refer to your stan-
dard SDS-PAGE protocol and adjust the volumes accordingly.
Alternatively, Tricine SDS-PAGE can be used as a second
 Blue Native PAGE 349

dimension, which has been described in detail [23] and was

the basis for the gel shown in Fig. 3.
54. If no suitable comb is at hand, one may pour the stacking gel
up to one BN lane width below the edge of the glass plate and
overlay the acrylamide solution with isobutanol.
55. Equilibration for longer periods may lead to a loss of small
proteins due to diffusion out of the gel.
56. Take care that no air bubbles are trapped between the first-­
dimension lane strip and the second-dimension gel.
57. Take care not to damage the BN gel during this procedure and
push only on the plastic film, not on the gel itself.
58. Fixation can last from 4 h to 4 days.
59. Usually bands are visible after 24 h, but optimal staining lasts
for 3–4 days.
60. The gel can be stored in 5% acetic acid for several weeks.
However, for analysis of protein bands by MS, it is recom-
mended to immediately excise and process the bands of inter-
est. Alternatively, bands can be stored at −20 °C until further
processing.
61. Short fixation improves sequence coverage in subsequent MS
but results in poor detection of small proteins. Longer fixation
may increase staining sensitivity but lower sequence coverage.
62. Extensive washing increases sensitivity and reduced back-
ground staining.
63. Avoid the use of glutaraldehyde as sensitizing agent. Although it
increases sensitivity and uniformity of staining, glutaraldehyde
crosslinks lysine residues and prevents complete trypsinization.
64. Impregnation with silver nitrate can last from 20 min to 2 h
without affecting the quality of the results.
65. Collect used silver solution in a dedicated waste container con-
taining sodium chloride to precipitate the silver.
66. Replace developing solution as soon as it turns yellow.
67. It is critical that only BN gels be used for immunoblotting that
were finally run with BN cathode buffer containing 0.02%
coomassie G to prevent blocking of the PVDF membrane by
excess coomassie G.
68. The infrared-fluorescent secondary antibodies are sensitive to
light. To avoid bleaching, immunodetection is done in dark
containers.

69. Work in fume hood. Collect MetOH in special waste
container.
70. Do not use Tween 20 in the blocking buffer. When bound to
the membrane, it may give rise to an increased background
fluorescence in the 700 nm channel.
350 Susann Zilkenat et al.

71. Two different proteins can be detected in the same complex if

mouse antibodies are available against one protein (e.g., against
an epitope tag) and rabbit antibodies are available against a
second protein.
72. Dual-color detection is achieved by using a secondary infrared-­
fluorescent antibody with an emission at 680 nm against one
species, e.g., mouse, and another secondary infrared-­fluorescent
antibody with an emission at 800 nm against another species,
e.g., rabbit. It is recommended to use the antibody of higher
quality in the 700 nm channel and the antibody of lower qual-
ity in the 800 nm channel because the background is typically
lower at 800 nm than at 700 nm.
73. Ask your MS facility if they have special requirements/prefer-
ences for sample tubes.
74. Depending on your light table, the samples may be heated up.
Especially in the lower percentage part of the gel, the gel will
be very sticky; the warmer it is, the harder it will be to cut. To
avoid this, the gel can be kept at 4 °C prior to cutting.
75. While cutting, make sure that your hair is tied back if necessary
and do not inhale too much ethanol.
76. Starting at the top ensures that you cut the stickiest part of the
gel while it is still cool.
77. It is easier and cleaner if a second person, also working with
clean gloves, opens, handles, and closes the tubes and double
checks the lane and band numbers.

Acknowledgments

Work performed in the laboratory of SW was supported by the

Alexander von Humboldt Foundation in the framework of the Sofja
Kovalevskaja Award endowed by the Federal Ministry of Education
and Research (BMBF) and in the framework of the Georg Forster
Research Fellowships (to J.V.M.F.), and by the Deutsche
Forschungsgemeinschaft (DFG) as part of the Collaborative
Research Center (SFB) 766 Bacterial cell envelope, project B14.

References

1. Schägger H, von Jagow G (1991) Blue native coli twin-arginine translocation apparatus
electrophoresis for isolation of membrane pro- incorporates a distinct form of TatABC com-
tein complexes in enzymatically active form. plex, spectrum of modular TatA complexes and
Anal Biochem 199:223–231 minor TatAB complex. J Mol Biol 346:
2. Bessonneau P, Besson V, Collinson I, Duong F 295–305
(2002) The SecYEG preprotein translocation 4.
Wiedemann N, Kozjak V, Chacinska A,
channel is a conformationally dynamic and Schönfisch B, Rospert S, Ryan MT, Pfanner N,
dimeric structure. EMBO J 21:995–1003 Meisinger C (2003) Machinery for protein
3. Oates J, Barrett CML, Barnett JP, Byrne KG, sorting and assembly in the mitochondrial
Bolhuis A, Robinson C (2005) The Escherichia outer membrane. Nature 424:565–571
 Blue Native PAGE 351

5. Kikuchi S, Hirohashi T, Nakai M (2006) workflow for affinity purification mass spec-
Characterization of the preprotein translocon trometry data. J Proteome Res 13:2239–2249
at the outer envelope membrane of chloro- 16. Monjarás Feria JV, Lefebre MD, Stierhof YD,
plasts by blue native PAGE. Plant Cell Physiol Galán JE, Wagner S (2015) Role of autocleavage
47:363–371 in the function of a type III secretion specificity
6. Wagner S, Baars L, Ytterberg AJ, Klussmeier A, switch protein in Salmonella enterica serovar
Wagner CS, Nord O, Nygren PA, van Wijk KJ, Typhimurium. MBio 6:e01459–e01415
de Gier JW (2007) Consequences of mem- 17. Zilkenat S, Franz-Wachtel M, Stierhof YD,
brane protein overexpression in Escherichia Galán JE, Macek B, Wagner S (2016)
coli. Mol Cell Proteomics 6:1527–1550 Determination of the stoichiometry of the
7. Baars L, Wagner S, Wickström D, Klepsch M, complete bacterial type III secretion needle
Ytterberg AJ, van Wijk KJ, de Gier JW (2008) complex using a combined quantitative pro-
Effects of SecE depletion on the inner and teomic approach. Mol Cell Proteomics
outer membrane proteomes of Escherichia coli. 18. Houben ENG, Bestebroer J, Ummels R,
J Bacteriol 190:3505–3525 Wilson L, Piersma SR, Jiménez CR, Ottenhoff
8. Wagner S, Klepsch MM, Schlegel S, Appel A, THM, Luirink J, Bitter W (2012) Composition
Draheim R, Tarry M, Högbom M, van Wijk KJ, of the type VII secretion system membrane
Slotboom DJ, Persson JO, de Gier JW (2008) complex. Mol Microbiol 86:472–484
Tuning Escherichia coli for membrane protein 19. Wittig I, Braun HP, Schägger H (2006) Blue
overexpression. Proc Natl Acad Sci U S A native PAGE. Nat Protoc 1:418–428
105:14371–14376 20. Schraidt O, Marlovits TC (2011) Three-­
9. Wickström D, Wagner S, Baars L, Ytterberg dimensional model of Salmonella’s needle
AJ, Klepsch M, van Wijk KJ, Luirink J, de Gier complex at subnanometer resolution. Science
JW (2011) Consequences of depletion of the 331:1192–1195
signal recognition particle in Escherichia coli. 21. Loquet A, Sgourakis NG, Gupta R, Giller K,
J Biol Chem 286:4598–4609 Riedel D, Goosmann C, Griesinger C, Kolbe
10. Wickström D, Wagner S, Simonsson P, Pop O, M, Baker D, Becker S, Lange A (2012) Atomic
Baars L, Ytterberg AJ, van Wijk KJ, Luirink J, model of the type III secretion system needle.
de Gier JWL (2011) Characterization of the Nature 486:276–279
consequences of YidC depletion on the inner 22. Klepsch M, Schlegel S, Wickström D, Friso G,
membrane proteome of Escherichia coli using van Wijk KJ, Persson JO, de Gier JW, Wagner
2D blue native/SDS-PAGE. J Mol Biol S (2008) Immobilization of the first dimension
409:124–135 in 2D blue native/SDS-PAGE allows the rela-
11. Krall L, Wiedemann U, Unsin G, Weiss S, tive quantification of membrane proteomes.
Domke N, Baron C (2002) Detergent extrac- Methods 46:48–53
tion identifies different VirB protein subassem- 23. Schlegel S, Klepsch M, Wickström D, Wagner
blies of the type IV secretion machinery in the S, de Gier JW (2010) Comparative analysis of
membranes of Agrobacterium tumefaciens. cytoplasmic membrane proteomes of
Proc Natl Acad Sci U S A 99:11405–11410 Escherichia coli using 2D blue native/SDS-­
12. Kuroda T, Kubori T, Thanh Bui X, Hyakutake PAGE. Methods Mol Biol 619:257–269
A, Uchida Y, Imada K, Nagai H (2015) 24. Neuhoff V, Arold N, Taube D, Ehrhardt W
Molecular and structural analysis of Legionella (1988) Improved staining of proteins in poly-
DotI gives insights into an inner membrane acrylamide gels including isoelectric focusing
complex essential for type IV secretion. Sci Rep gels with clear background at nanogram sensi-
5:10912 tivity using Coomassie Brilliant Blue G-250
13. Wagner S, Königsmaier L, Lara-Tejero M, and R-250. Electrophoresis 9:255–262
Lefebre M, Marlovits TC, Galán JE (2010) 25. Shevchenko A, Wilm M, Vorm O, Mann M
Organization and coordinated assembly of the (1996) Mass spectrometric sequencing of pro-
type III secretion export apparatus. Proc Natl teins silver-stained polyacrylamide gels. Anal
Acad Sci U S A 107:17745–17750 Chem 68:850–858
14. Lara-Tejero M, Kato J, Wagner S, Liu X, Galán 26. Sessler N, Krug K, Nordheim A, Mordmüller
JE (2011) A sorting platform determines the B, Macek B (2012) Analysis of the Plasmodium
order of protein secretion in bacterial type III falciparum proteasome using blue native
systems. Science 331:1188–1191 PAGE and label-free quantitative mass spec-
15. Fischer M, Zilkenat S, Gerlach RG, Wagner S, trometry. Amino Acids 43:1119–1129
Renard BY (2014) Pre- and post-processing
 Chapter 27

In Situ Imaging of Bacterial Secretion Systems

by Electron Cryotomography
Gregor L. Weiss, João M. Medeiros, and Martin Pilhofer

Abstract
The unique property of electron cryotomography (ECT) is its capability to resolve the structure of mac-
romolecular machines in their cellular context. The integration of ECT data with high-resolution struc-
tures of purified subcomplexes and live-cell fluorescence light microscopy can generate pseudo-atomic
models that lead to a mechanistic understanding across size and time scales. Recent advances in electron
detection, sample thinning, data acquisition, and data processing have significantly enhanced the applica-
bility and performance of ECT. Here we describe a detailed workflow for an ECT experiment, including
cell culture, vitrification, data acquisition, data reconstruction, tomogram analysis, and subtomogram aver-
aging. This protocol provides an entry point to the technique for students and researchers and indicates
the many possible variations arising from specific target properties and the available instrumentation.

Key words Tilt series, Plunge freezing, Electron cryomicroscopy, Reconstruction, Cryogen,
Segmentation

1  Introduction

Bacterial cell–cell interactions are often mediated by the secretion

of effector proteins that act on target cells. Macromolecular
machines in the bacterial cell envelope are crucial for the transloca-
tion of effectors from the bacterial cytoplasm into the extracellular
space or directly into the target cell [1]. Much insight into the
mechanisms of these secretion systems are the result of high-­
resolution structure determination of purified subcomplexes and
live-cell fluorescence light microscopy. Structural studies of secre-
tion systems, however, are often challenged by the complexity of
the systems, the involvement of membrane proteins, and the
dependence on the removal of the system from its cellular context.
On the other hand, fluorescence light microscopy is limited to the
visualization of labeled components, by the dependence on func-
tional fluorescent tags, and by the achievable resolution. Electron
cryotomography (ECT) offers solutions to these problems. ECT

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_27, © Springer Science+Business Media LLC 2017

353
354 Gregor L. Weiss et al.

resolves unique structures in situ, in a lifelike state, in three dimen-

sions, and at a resolution of a few nanometers [2–6]. Flash freezing
a sample in noncrystalline ice avoids the introduction of artifacts
arising from conventional electron microscopy (EM) sample prep-
aration [7].
In a typical ECT experiment, a few microliters of a bacterial
culture are applied to an EM grid and excess liquid is removed.
The grid is plunged into a cryogen and transferred to a transmis-
sion electron microscope operating at cryogenic temperature.
A series of 2D projection images of a bacterial cell from different
angles is recorded and the data are reconstructed into a 3D image—
the tomogram. The resolution of an individual tomogram is around
2–5 nm. For instance, high-quality tomograms typically resolve the
double leaflet of bacterial cellular membranes or cytoskeletal struc-
tures in the crowded cytoplasm. Extracting, aligning, and averag-
ing of subvolumes (subtomogram averaging) allows for the
improvement of contrast and enables atomic resolution for suitable
targets [8]. The generated density maps can be integrated with
high-resolution structures of subcomplexes, resulting in pseudo-­
atomic models of the machines in their cellular context.
ECT has facilitated groundbreaking progress regarding the
understanding of diverse bacterial secretion systems [9–20], but it
faces a number of technical challenges. The incident electron beam
damages the sample, which limits the applicable total dose, result-
ing in data with low contrast. With increasing sample thickness,
noise is introduced by higher fractions of inelastic and multiple
scattering events. The physics of image formation requires data
acquisition at a defocus in order to be able to observe low-­
resolution features; however, this reduces high-resolution informa-
tion. Subcellular targets, including secretion systems, can only be
detected after data collection and reconstruction. Tilt series acqui-
sition is time consuming (20–75 min per series), and the limited
tilt range causes anisotropic resolution. Fortunately, many of these
issues have been addressed by recent methodological develop-
ments. One development is the introduction of direct electron
detectors, by which the field of electron cryomicroscopy entered a
new era [21]. These detectors offer much improved detective
quantum efficiency [22], and the fast readout allows for the correc-
tion of sample motion occurring during exposure [23]. Phase
plates were introduced to modulate the contrast transfer function
(CTF) in such a way that tilt series can be acquired in focus, signifi-
cantly enhancing contrast [24]. Cryo-light microscopy allows for
the detection of rare events prior to ECT data acquisition and for
the identification of subcellular structures [25, 26]. Cryo-focused
ion beam (FIB) milling is a potent approach to thin samples prior
to ECT [27]. It is likely that the aforementioned developments,
together with novel data collection schemes [28] and the p ­ ossibility
of subtomogram classification [29], will significantly enhance the
 Electron Cryo-Tomography 355

power and applicability of ECT. The integration with data from

cell and structural biology will resolve the mechanisms of secretion
machines across scales. This is particularly significant for secretion
systems that are dependent on cell–cell contact since their function
is likely only to be fully understood by including the secreting as
well as the receiving cell in the analysis.
It would require going far beyond the scope of this chapter to
detail all possible variations of the ECT workflow. The technology
is under active development, and the exact protocol is highly
dependent on the target and the available instrumentation. Here
we present one possible step-by-step workflow that can help stu-
dents and researchers to obtain a first insight into the technique.

2  Materials
2.1  Plunge Freezing 1. Sample support: EM grids with a holey carbon support such as
Quantifoils (Quantifoil Micro Tools, Germany) or C-Flats
(Protochips, USA). Here we use Quantifoil R2/2 copper grids
(200 mesh). For grid specifications see Note 1.
2. Cleaning and hydrophilization of grids: Glow discharge system
such as Emitech K100X (Quorum Technologies, UK) and a
glass slide as grid support.
3. Preparation of gold fiducial markers: Bovine serum albumin
(BSA), double-distilled (dd) water, 10 nm gold nanoparticles,
and a tabletop centrifuge.
4. Vitrification of cells on grids: Plunge-freezing device such as
EM GP (Leica Microsystems, Germany), Cryoplunge 3
(Gatan, USA), or a custom-built model. Here we use a Vitrobot
MK II (FEI, USA).
5. Blotting of excess liquid: Whatman filter paper (diameter: 47 mm,
grade #1) and a device for punching holes into the filter paper.
6. Cryogen for plunge freezing: Ethane/propane (37%/63%
vol/vol; ±2%) gas mixture.
7. Cooling agent: Liquid nitrogen (LN2) in a 4 L cryo-dewar.
8. Grid storage: Cryo-grid storage boxes and transfer release tool
(TGS Technologies) or custom-made systems. Long-term LN2
storage dewars and 50 mL Falcon tubes attached to strings.
9. Cryo-transfer dewars (350 mL) for grid box transfer.

2.2  Electron 1. 300 kV transmission electron cryomicroscope, equipped with

Cryotomography a field emission gun (FEG), imaging filter, and direct electron
detector. Here we use an FEI Polara G2 (FEI) with a
­postcolumn Gatan Imaging Filter (GIF) 2002 (Gatan) and a
K2 Summit detector (Gatan). For alternative instrumentation
see Note 2.
356 Gregor L. Weiss et al.

2. Data collection software for automated acquisition of tilt series

such as Latitude (Gatan), Leginon [30], SerialEM [31], or
Tomography 4.0 (FEI). Here we use UCSF Tomography [32].
3. Cooling agent: 20–40 L LN2 for a 2 day session.

2.3  Tomogram 1. Computation: Workstation equipped with a modern Intel or

Reconstruction, AMD processor and at least 8 GB memory. Current tomogra-
Analysis, phy software suites can take advantage of CUDA-capable
and Subtomogram graphics processing units for reconstruction, so an NVIDIA
Averaging card from the GeForce 700 family or newer is recommended.
Here we use an iMac 4 GHz Intel Core i7 with 16 GB mem-
ory size and an NVIDIA GeForce GTX 780 M (Apple, USA).
For the reconstruction of unbinned data or large subtomo-
gram averaging jobs, we use a Linux workstation with 2 × 10
core Intel Xeon 2.2 GHz processor, 256 GB memory, and
NVIDIA GTX 1080 8 GB graphics controller.
2. Drift correction software such as DigitalMicrograph (Gatan)
or MotionCorr [33]. Here we use Alignframes, which is part
of the IMOD software package [34, 35].
3. Reconstruction software: Several reconstruction algorithms
are available, such as supersampling SART [36], NUFFT [37],
Tomo3D [38], or Tomography 4.0 (FEI). Here we use the
IMOD software package. The package requires a Java runtime
environment and for Windows additionally the Unix toolkit
Cygwin.
4. Generating movies from image sequences: QuickTime 7 Pro
(Apple) or Fiji [39].
5. Subtomogram averaging software, such as Dynamo [40] or
Relion [29]. Here we use PEET [41], which is part of the
IMOD package and requires the MATLAB Compiler Runtime.

3  Methods

As a complementary source of information on the theoretical back-

ground, we recommend the online course “Getting Started in
Cryo-EM” (http://cryo-em-course.caltech.edu/).

3.1  Cultivation Culture the bacteria. Optimize the growth conditions (e.g.,
of Bacteria medium, temperature, gas atmosphere, shaking/static incubation,
liquid/solid medium) to obtain a high expression level for the
secretion system of interest (see Note 3) while at the same time
minimizing the cell diameter, resulting in higher-quality data (see
Note 4). Ideally, cells are grown in a few milliliters of liquid
medium. For a protocol on how to proceed with cells grown on
solid medium, see Note 5. For a description of how to grow
 Electron Cryo-Tomography 357

bacteria directly on EM grids, see Note 6. The expression level of

the secretion system of interest is ideally monitored just before
freezing, e.g., by fluorescence light microscopy.

3.2  Plunge Freezing To preserve bacterial cells in a frozen-hydrated near-native state,

of Bacterial Cells the sample is applied to an EM grid and plunge frozen in liquid
ethane/propane [42] using a Vitrobot (FEI) [43]. A video show-
ing the workflow has been published [44].
1. Prepare gold fiducial markers. Fiducials are added to the sam-
ple prior to freezing to allow for the alignment of individual tilt
images. Mix 400 μL 10 nm gold nanoparticles with 100 μL 5%
(w/v) BSA and vortex for 5 s. Centrifuge the mixture at
14,000 × g for 15 min, discard the supernatant, and wash the
pellet with 200 μL ddH2O with an additional centrifugation
step at 14,000 × g for 10 min. Pool the pellets of eight parallel
preparations and resuspend them in ddH2O (final volume
100  μL). Fiducial markers can be stored at 4 °C for up to
2 months.
2. Prepare sample support. Glow discharging is used for cleaning
the EM grids and rendering their surface hydrophilic. Measure
the distance between the specimen table and electrode inside
the glow-discharge device (Emitech K100X) and adjust it to
2 cm. Place the grids with the carbon side facing up onto a
glass slide and place the slide onto the specimen table inside
the glow discharger. Close the lid and set the following param-
eters: Negative discharge 15 mA for 60 s, initial pumping to
1 × 10−1 bar, bleed to 2 × 10−1 bar. For different samples and
grids, it might be necessary to adjust the table-electrode dis-
tance (typically 2–4 cm), current (typically 15–25 mA), and
discharge time (typically 15 s to 3 min). Treated grids should
be used for freezing within a few hours (otherwise the proce-
dure can be repeated).
3. Prepare the Vitrobot. Fill the Vitrobot humidifier with ddH2O
with a syringe. Mount the prepunched Whatman filter paper
carefully on the blotting pads in the Vitrobot chamber. For
bacterial cells, set the temperature to 22 °C and the humidity
to 100%. Set the following parameters: wait time 1 s (time
between applying your sample on the grid and blotting), blot-
ting time 1–10 s (time the filter paper is pressed onto the grid),
drain time 1 s (time from blotting to plunging), and blot offset
−2 to −3 mm. For more details on parameters see Note 7.
4. Prepare the cryogen. Cool the Vitrobot LN2 reservoir and the
central cryogen container with LN2. Keep filling the LN2 com-
partment and wait until the cryogen container is free of LN2.
Start filling the cryogen container with ethane/propane gas
mixture until it is completely filled with liquid cryogen. See
Note 8 for a comment on the choice of cryogen.
358 Gregor L. Weiss et al.

5. Mix the bacterial culture with gold fiducial markers (4:1 vol/
vol). Prepare this shortly before plunge freezing to avoid alter-
ations of the sample.
6. Pick up a glow-discharged grid with dry Vitrobot tweezers.
Mount the tweezers on the Vitrobot and lift the grid up into
the blotting chamber by pressing the foot pedal. Apply 3–4 μL
of the cell/gold suspension onto the carbon side of the grid
and continue by pressing the foot pedal. The grid is then auto-
matically blotted from both sides, removing excess liquid.
After the drain time, the grid is plunged into the liquid eth-
ane/propane mixture. Due to the high thermal conductivity of
the cryogen, freezing happens very rapidly without the forma-
tion of ice crystals, resulting in vitreous ice [42, 45]. See Note 9
for ways to increase the amount of cells on the grid.
7. From this point, the grids must be kept at liquid nitrogen tem-
perature to prevent thawing, ice contamination, and devitrifi-
cation. Carefully remove the grid from the cryogen and place
it into a labeled and precooled grid storage box.
8. In rare cases, the expression level or assembly of a secretion
system might be affected by the short incubation time on the
grid or blotting of excess liquid. To verify the assembly state of
a fluorescently tagged secretion system after plunge freezing,
the frozen grid can be analyzed in a cryogenic fluorescence
light microscope [25].
9. For storage, grid boxes are dropped into 50 mL Falcon tubes.
The tubes are closed with lids attached to strings and kept in
long-term liquid nitrogen storage dewars.

3.3  Loading Grids 1. Cool down the loading station chamber and cryopump. Place
into Electron the grid box and Polara cartridges into the chamber. Load a
Microscope C-clip ring into a C-clip tool and cool the tip using LN2. Using
cooled forceps, transfer a grid into a cartridge and clamp it
with the C-clip ring. Repeat for up to six grids/cartridges total.
2. Detach the cooled multispecimen holder (MSH) from the
Polara and attach it to the loading station. Vent the holder and
insert the cartridge rod into the chamber. Load the cartridges
into the parking positions of the MSH with the threads facing
inward. Retract the rod and pump the vacuum in the MSH
using the rough pump and cryopump. Start the turbomolecu-
lar pump of the Polara. Detach the MSH from the loading
station and attach it to the Polara. Click “Pump airlock” in the
“Setup” tab of the microscope software and open valves A and
B when prompted by the software.
3. Push the MSH to the intended cartridge position. The first
position refers to the lowest parking position in the cryo load-
ing station. Use the cooled insertion rod to pick up a cartridge
from the MSH. Retract the insertion rod and the MSH.
 Electron Cryo-Tomography 359

4. Press “Cartridge” in the “Setup” tab of the microscope soft-

ware and insert the cartridge into the beam path. Screw the
cartridge onto the stage by clockwise turns. Stop with the first
sign of resistance and retract the insertion rod.

3.4  Basic Microscope Microscope alignments change over time and must be adjusted
Alignments/ before data collection. Here we describe the procedure for a cooled
Preparations FEI Polara instrument equipped with a Gatan imaging filter and
K2 Summit direct electron detector.
1. Choose a magnification for data collection. The pixel size on
the specimen level should be at least two times smaller than the
intended resolution. Keep in mind that it might be necessary
to bin the data during processing, resulting in a larger final
pixel size. The following microscope preparations should be
made at the chosen magnification.
2. To center the condenser (C2) aperture, move the stage in X/Y
to an area of broken sample support (hole) and save this posi-
tion in the “Stage” tab of the microscope software. Drop down
the fluorescent screen, condense, and spread the beam repeat-
edly. Adjust the condenser aperture screws so that the beam
condenses and spreads centered without lateral movement.
3. To correct for an astigmatic beam, activate the condenser stig-
mators in the “Direct alignments” tab and adjust them using
the X/Y multifunction knobs.
4. Pivot points are aligned to ensure that the beam stays centered
on the camera when the focus is being changed. In “Direct
alignments,” select “Beam tilt ppx.” Condense the beam and
minimize movements by adjusting the X/Y multifunction
knobs. Repeat for “Beam tilt ppy,” click “done,” and spread
the beam. Lift the viewing screen.
5. To adjust the position of the grid so that the target does not
move laterally during tilting (eucentric height), move the stage
in X/Y to center a feature, e.g., an ice particle on carbon. Start
the “Wobbler” in the “Stage” tab and adjust the height of the
grid by pressing the “+/−“ Z-axis buttons to minimize move-
ments of the centered feature (use the K2 detector to view with
an exposure time of 0.1 s at data collection magnification).
6. Focus the feature using the focus knob.
7. Press the “Eucentric focus” button on the microscope control
panel
8. To center the beam along the optical axis of the objective lens,
select “Tomo rotation center” in “Direct alignments.”
Minimize movements of the image (viewed on K2 detector in
linear mode) using the X/Y multifunction buttons. Using the
focus step size knob, the amplitude of the objective lens cur-
rent oscillation can be adjusted, which results in different
intensities of image movement.
360 Gregor L. Weiss et al.

9. To center the objective aperture, drop the viewing screen,

switch to diffraction mode, and adjust the objective aperture
screws so that the bright spot is centered inside the illuminated
disk. Turn off diffraction mode.
10. To correct for objective lens astigmatism, lift the viewing
screen and start live-FFT in the DigitalMicrograph software
(DM). Activate the objective stigmators in “Direct align-
ments” and adjust them using the X/Y multifunction knobs,
so that the Thon rings on the live-FFT are circular.
11. Move the stage back to the saved “hole”-position (broken area
in carbon support) using the “Stage” tab.
12. To align the crossover, switch to an intermediate magnification
(e.g., 3000×). Select “Cross-over” in the “Tune” tab. Turn
multifunction knob X until the edge of the pumping aperture
is visible. Move multifunction knob X in the reverse direction
and count the turns until the opposite edge is visible. Center
by going back half the number of turns. Repeat procedure for
multifunction knob Y and press “Done.”
13. The Gatan imaging filter (GIF) is used to remove inelastically
scattered electrons from the beam by zero-loss filtering. GIF
calibration is performed on the Orius charge-coupled device
camera to avoid damage to the K2 chip (for newer models such
as the Quantum LS, calibration is performed using the K2).
Insert the Orius camera in DM, switch to data collection mag-
nification, set spot size to 3, start “Tune GIF” in DM, and
select “Full tune.”
14. To acquire gain and dark reference images, start “Prepare Gain
Reference” in DM. Follow the program’s instructions for
recording reference images for linear mode (at spot size 3) and
counting/superresolution mode (at spot size 8). Note that this
procedure should be repeated daily or if any fixed noise pattern
appears in the acquired images.

3.5  Automated The next step is the acquisition of tilt series (2D projection images
Sequential Tilt Series from different angles). The sequential recording of tilt series for a
Acquisition number of targets is performed automatically using a suitable soft-
ware package (see Subheading 2.2, step 2). Here we describe the
use of UCSF Tomography [32]. The program operates by record-
ing a low-magnification atlas of a grid region for the localization of
targets, intermediate-magnification images for the inspection of
potential targets, and final data collection on a selection of targets.
For this purpose, the program specifies five different modes that
correspond to different settings of the microscope and the camera:
Atlas, Search, Track, Focus, and Collect. Table 1 lists the main
function for each mode and a set of parameters that can serve as a
good starting point for data collection.
 Electron Cryo-Tomography 361

Table 1
UCSF tomography modes and example parameters

UCSF Slit Pixel

tomogr. Magnifi­ Defocus width Exposure Spot size Camera
mode Main function cation (μm) (eV) time (s) Binning size (nm) mode
Atlas Recording of a 4500× −50 80 2 1 8 67.5 Counted
montage image
of a defined grid
area
Search Recording of images 13,500× −50 40 2 1 8 15.0 Counted
of potential
targets for closer
inspection
Track Tuning eucentricity, 13,500× −20 40 1 1 8 15.0 Counted
targeting, and
tracking specimen
shift after first leg
Focus Automated focusing 42,000× 0 20 1 1 8 0.50 Counted
Collect Final data acquisition 42,000× −4 to 20 1–4 1 8 0.50 Counted
−10

1. Start the “TecnaiServer” on the microscope PC and UCSF

Tomography on the K2 computer.
2. Set the parameters from Table 1 for all modes. Select the
“Image” tab and the desired mode, click “Configure,” and
change parameters.
3. Set the magnification, beam intensity, and beam shift for each
mode. Select Atlas mode, set magnification, center the beam
on the detector, and spread it. The spot size should be chosen
so that the desired intensity is obtained by spreading the beam
just beyond the edges of the detector. Press “From Scope” to
save the beam settings for the Atlas mode. Repeat this proce-
dure for all other modes.
4. A series of calibrations is required. Move the stage to center an
ice particle on the carbon support. Go to the “Calibration” tab
in UCSF Tomography. Set the defocus values for all modes to
0. Bring the specimen to eucentric height (see Subheading 3.4,
step 5). Focus the specimen. Select Collect mode, click
“Configure,” and press “Read true focus.” Select the “Stage
shift” radio button and run this calibration for Atlas and Search
modes. Run “Image Shift” calibration for all modes. Run
“Focus” calibration for Track and Focus modes. Run
“Eucentric” calibration for Track mode. Run “Optical axis”
362 Gregor L. Weiss et al.

calibration for Focus mode. All calibrations should be repeated

in case the alignments of the microscope or the magnification
for any of the modes change.
5. Set the desired defocus values for all modes in the “Configure”
window. See Note 10 for comments on the defocus value for
final data acquisition.
6. To align the different modes to each other, center a feature (ice
particle or crack) that is visible in all modes. Switch to Collect
mode. Select the “Align modes” radio button and press
“Start.” Follow the instructions in the log window. Repeat the
procedure for Search mode.
7. The goal now is to record a map in Atlas mode covering around
4–16 squares of the grid (see Fig. 1a for an example). Screen
your grid at low magnification to identify a region with suit-
able ice thickness and target density. Move the stage to the
center of the identified grid area. Select the “Montage” tab in
UCSF Tomography. Enter the size of the atlas map in microns
(250 × 250 covers ~4 grid squares on a 200 mesh grid) and a
file name. Click “Build” and monitor the automated recording
of the atlas map. Note that the recording can be aborted at any
time in case a sufficient grid area has been imaged.
8. To identify targets on the atlas, load the atlas in the “Montage”
tab. Uncheck “Go to Target.” Double click an interesting tile
on the map to zoom in. Within this image, double click a spot
on the carbon support, which will be used for focusing and
finding eucentricity. With every additional double click, poten-
tial targets can be selected for closer inspection (collection of
Search images in the next step). Click “Zoom Out.” A red
circle is added on the atlas map to indicate that this area has
been visited. Repeat procedure for other tiles of interest.
9. To acquire intermediate-magnification search images of the
potential targets marked on the atlas map, switch to the
“Target” tab and select the corresponding atlas map file in the
“Pre-Rotation” field. Click “Acquire Targets” and monitor the
progress (see Fig. 1b for an example).
10. To inspect the potential targets, switch to the “Target Review”
tab. Target images are saved in a new file (named like the atlas
and appended with “tgt”). Load the file. Inspect the potential
targets and double click positions that will be used for final
data collection. Selected targets can be removed by right click-
ing. Note that on some images no targets can be selected since
they are dedicated to finding eucentricity and focusing.
11. Set tilt series parameters in the “Tomography” tab. Enter the
tilt range (−60° to +60°), tilt increment (+1°), and starting
angle (−20°). See Note 11 for comments on tilting schemes.
 Electron Cryo-Tomography 363

Fig. 1 Steps in ECT data acquisition and analysis. Examples are shown for (a) an atlas covering several
squares, (b) a search image of a bacterial cell lying across the carbon support (c) an individual tilt image of
arrays of metamorphosis-associated contractile structures, (d) a slice through a cryotomogram that was
reconstructed from the tilt series shown in c, (e) two views of a segmentation of the tomogram shown in d,
(f) a slice through an average generated from subvolumes of tomograms similar to d, (g) an isosurface of the
average shown in f. c–e were generated with data from [16]. Bars, 1000 nm in a and b, 100 nm in c–e,
10 nm in f and g

Insert the base name for tilt series into the “File” field and
locate the target file in the “Target” field. Press “More” to set
additional imaging parameters. Check “Dose Fract” to enable
dose fractionation and enter total exposure time and subframe
exposure time (see Note 12). Select “UShort” and “Cryo” for
“MRC Data Type.” Check “Align” for “ZLP Alignment” (to
align zero-loss peak before each tilt series) and “Close at End”
(to close column valves after data acquisition).
12. To set the electron dose, move to a hole in the carbon support
and switch to Collect mode. Acquire an image with DM and
read out the dose for the image (e−/pixel). By adjusting the
beam intensity, spot size, and exposure time, set the dose for
individual tilt images. See Note 12 for considerations on total
dose, dose rate, spot size, exposure time, and dose-­fractionation
parameters.
364 Gregor L. Weiss et al.

13. Start data collection by clicking “Start” in the “Tomography”

tab. The program will sequentially record tilt series of the
selected targets (see Fig. 1c for an example). Note that data col-
lection can be paused (“Pause”) or aborted (“Stop All”).
Pressing “Stop” will abort the current tilt series and the pro-
gram will move on to the next target in the list. Since batch
tomography can last many hours or even days, make sure that
the LN2 dewar of the microscope is always filled. Additional
grids can be kept at cryogenic temperature inside the MSH (see
Note 13 for long-term cooling of the MSH).
14. From this point, no physical presence at the microscope is
required. With remote-control software packages, such as
TeamViewer (TeamViewer GmbH, Germany) or VNC Viewer
(RealVNC Ltd., UK), the progress of tilt series acquisition can
be monitored and controlled remotely.

3.6  Data Processing 1. Motion correction. Imperfections in the microscope stage and
the incident electron beam result in sample movement during
exposure [23]. To correct for this motion, the projection
images for each tilt image are read out as subframes (dose frac-
tionation) and saved as image stacks. The subframes of a given
tilt image are then computationally aligned to each other, aver-
aged, and saved. Move the uncorrected tilt series and the cor-
responding subframes to an empty folder. Rename each of the
numbered subframe stacks from “(…).mrc001” to “(…)_001.
mrc,” “(…).mrc002” to “(…)_002.mrc,” and so on. Open a
terminal window within that folder and enter the command
“alignframes –stack <uncorrected tilt series.mrc> <sub-frames-
basename_*> <output file.mrc>.” The measured drift for each
frame set can be monitored in the terminal window.
2. The next step is the reconstruction of the tilt series into a 3D
image (tomogram). Move the motion-corrected tilt series to a
new directory. Open a terminal window, change to the preced-
ing directory, and start the IMOD program eTomo. The docu-
mentation for programs of the IMOD package can be accessed
via the Help menu. Locate the tilt series (stack) under the
“Build Tomogram” tab. See Note 14 for automated data
reconstruction.
3. Specify tilt series parameters. Enter the diameter of the gold
fiducials (10 nm), enter tilt series axis type (single axis), and
choose “cryosample.adoc” as the system template. Click “Scan
header” to read pixel size and image rotation from the tilt
series file. Click “View raw image stack” to inspect the selected
tilt series (see Subheading 3.7, step 1 on using 3dmod for data
visualization). Specify tilt images to be excluded from recon-
struction (e.g., images in which the grid bar blocks the beam
at high tilts) to the “Exclude views” field. Click “Create Com
Scripts.”
 Electron Cryo-Tomography 365

4. The procedure now follows the eTomo panels from top (“Pre-­
processing”) to bottom (“Clean Up”). Note that intermediate
files are saved and it is possible to go back and rerun certain
steps with changed parameters. Also note that moving the
mouse over an input field in IMOD will show more informa-
tion on the required input parameters.
5. “Pre-processing.” To remove pixels with aberrantly high inten-
sity. Set 12 for “Peak criterion” and 9 for “Difference crite-
rion.” Click “Create Fixed Stack” and inspect the processed tilt
series by clicking “View Fixed Stack.” Click “Use Fixed Stack”
and proceed by clicking “Done.”
6. “Coarse Alignment.” To align successive tilt images, click
“Calculate Cross-Correlation.” Bin the coarsely aligned image
stack by 2 and check “Reduce size with antialiasing filter”
(note that this binning will not affect the size of the final recon-
struction). Click “Generate Coarse Aligned Stack” and inspect
the aligned stack by clicking “View Aligned Stack in 3dmod.”
Use “Midas” to fix any misaligned image pairs. Proceed with
“Done.”
7. “Fiducial Model Gen.” The selection of gold markers and the
tracking of the selected markers throughout the tilt series can
be performed automatically using the program RAPTOR, or
manually. Here, we manually select the fiducial markers after
selecting “Make seed and track” and “Make seed model manu-
ally.” Open the aligned tilt series by clicking “Seed Fiducial
Model.” Select 10–20 gold fiducial markers distributed over
the entire 0° tilt image, and save the model before closing
3dmod. Select the “Track Beads” tab and track the fiducials
over the whole tilt series with “Track Seed Model.” Open the
fiducial model by clicking “Fix Fiducial Model” and modify
the model to minimize the number of gaps and to fix mis-
tracked fiducials. Save the model, click “Done,” and proceed.
8. “Fine Alignment.” Calculate the fine alignment with “Compute
Alignment.” Inspect by clicking “View/Edit Fiducial Model.”
Inspect fiducials by clicking “Go to Next Big Residual” and if
appropriate fix the model by clicking “Move Point by Residual.”
Repeat until there are no more residuals. Save model and
repeat “Compute alignment.” Proceed with “Done.”
9. “Tomogram Positioning.” Enter “1500” as the preliminary
tomogram Z-height in the field “Positioning tomogram thick-
ness” and click “Create Whole Tomogram.” Click “Create
Boundary Model” and flip the tomogram 90° using “Edit >
Image > Flip/Rotate.” Indicate the boundaries of the cell on a
central slice (“View axis position”) by drawing two horizontal
lines. Repeat this for one higher and one lower slice (“View
axis position”). Save the model by typing “s,” close 3dmod,
and press “Compute Z Shift & Pitch Angles.” Click “Create
Final Alignment” and proceed with “Done.”
366 Gregor L. Weiss et al.

10. “Final Aligned Stack.” Enter the binning factor in the “Aligned
image stack binning” field. We usually use a factor of 2, result-
ing in data with 1920 pixels in X and 1854 pixels in Y. Check
“Use linear interpolation” and “Reduce size with antialiasing
filter.” Click “Create Full Aligned Stack.” Other, optional pro-
cessing steps in this tab allow for the erasing of gold beads,
CTF correction, and 2D filtering.
11. “Tomogram Generation.” Choose between two different
tomogram reconstruction algorithms: weighted backprojec-
tion in Fourier space [46] or the Simultaneous Iterative
Reconstruction Technique (SIRT) [47]. Click “Generate
Tomogram” and proceed with “Done.”
12. “Post-processing.” Open the tomogram with “3dmod Full
Volume.” Select a slice with structures of interest. Use the con-
trast sliders to determine the maximum and minimum values
for “black” and “white,” needed to visualize the structure of
interest. Enter the numbers into the “Scale to match contrast”
fields. For “Reorientation,” choose either “Rotate around X”
or “Swap Y and Z dimensions.” Note that this option will
affect the handedness of the data. The correct option must be
determined as described previously [48]. Click “Trim Volume”
and “Done.”
13. “Clean Up.” Select all listed intermediate files and click “Delete
Selected” to save disk space. Finish by clicking “Done.” Note
that the “Com Scripts Interface” for a specific reconstruction
can be reopened using the “.edf” file.

3.7  Data 1. View the tomogram. Open the “.rec” file in 3dmod of the
Visualization IMOD package using the command “3dmod <input file.rec>”
(see Fig. 1d for an example). 3dmod offers different viewing
tools, including “ZAP,” “XYZ,” and “Slicer.” “Slicer” is par-
ticularly useful to detect macromolecular complexes (such as
secretion systems), based on the possibilities of rotating the
tomogram around all three axes (use sliders “X-, Y-,
Z-­rotation”), pan up and down slice by slice (use slider “View
axis position”), and average multiple slices to enhance contrast
(enter value in “Img”). To measure distances, select “model”
mode in the 3dmod information window, click the left mouse
button on point A, move the mouse to point B, and type “q.”
The A-B distance (in 3D) will be reported in the log of the
3dmod information window.
2. Generate 3D models by segmentation and visualize them.
Segmentation (visualization of selected pixels) can be per-
formed automatically, e.g., by applying a density threshold
(use 3dmod’s “Isosurface” in the “Image” drop-down menu)
or by algorithms that trace specific features such as membranes
or filaments (e.g., Shape, at bio3d.colorado.edu, or other
automated segmentation approaches [49–51]). The poor con-
 Electron Cryo-Tomography 367

trast in low-­dose, low-defocus cryotomograms, however, often

challenges available programs to deliver meaningful models.
Manual segmentation can be performed in such cases or for
structures for which no programs are available (see Fig. 1e for
an example). Amira (FEI) and 3dmod are frequently used for
manual segmentation.
Open the tomogram in 3dmod, select the “ZAP” viewer,
and open the “Drawing Tools” under the “Special” menu. To
segment membranes, use “sculpt” to outline the membrane in
a given slice (select “Edit > Object > Type… > Closed”).
Repeat segmentation for the same membranous structure in
every approximately tenth slice. Interpolate the contour for
the intermediate slices (“Special > Interpolator”). Inspect and
refine the model (can be viewed without tomogram slices in
“Image > Model View”). Use meshing (select “Model View”
window followed by “Edit > Objects… > Meshing”) to gener-
ate a surface based on the modeled contours. For filamentous
structures, change the object type to “open” and draw lines
with the middle mouse button. Apply “meshing” to represent
structures as rods or tubes. Save the model file as a “.mod” file.
3. Make a movie. 3D data (tomograms and models) can be effec-
tively presented as movies. Start 3dmod in a new folder (a
sequence of images will be saved inside this folder). Open a
tomogram. Remove red/yellow navigation markers by typing
“shift-T.” Select the “Movie/Montage” option under “File”
in the menu bar. Choose start and end frame and other options.
Make a test run by pressing the command key and middle
mouse button at the same time. Save an image sequence by
selecting an output format under “Snapshot” and start
sequence as described earlier. The “Movie/Montage…”
option for the Model viewer functions similarly; however, it
offers the possibility of introducing rotations and switching off
tomogram slices. Use QuickTime 7 Pro or Fiji to generate a
movie from the image sequence.

3.8  Subtomogram Contrast and resolution of cryotomographic data can be improved

Averaging by aligning and averaging repeating subvolumes. This is particularly
effective for secretion systems. Here we describe a protocol using
PEET [41, 52]. The workflow takes advantage of the cell-­envelope-­
spanning localization of the targets, which allows for a reduced
angular search space, resulting in reduced computational effort.
Note that PEET tutorials can be found at https://goo.gl/nsXEtn.
1. Mark subvolumes as model points and indicate their orienta-
tion. Open the tomogram in 3dmod and choose “Edit > Object
> Type” from the menu bar. Select object type “Open” and
add a “Symbol” to visually indicate modeled points. Select the
“Model” radio button in the small 3dmod information win-
368 Gregor L. Weiss et al.

dow. Using the “Slicer,” rotate the tomogram to the best pos-
sible view of a particular secretion system. Click with the
middle mouse button on the feature of interest to define the
center of a subvolume. Click a second point to indicate the
orientation of the subvolume (e.g., perpendicular to the cell
envelope or along a tube). The program draws a line between
both model points (green line and circles in Fig. 2.). Type “n”
(for new Contour) and repeat for all subvolumes in this tomo-
gram. Record the number of points within the object and save
the model as a “.mod” file. Repeat the procedure for addi-
tional tomograms and save the corresponding model files.
Note that all tomograms should have the same pixel size.
2. Compute an initial motive list. This list contains information
about the orientation of the modeled subvolumes. Run the
IMOD command “stalkInit <input file.mod>” in the folder
where the model is located.
3. Open PEET via the eTomo graphical user interface and select
a base name and a folder. Parameters from previous projects
can be imported.
4. Load the tomograms and corresponding models (“head.mod”
file, generated by the stalkInit command) into the respective
fields of the “Volume Table.”
5. As “Reference” for the first round of alignment, choose a ran-
dom subvolume.

Fig. 2 Modeling of subvolumes for subsequent subtomogram averaging. Shown

is a schematic of a cell envelope-spanning secretion system (OM outer mem-
brane, IM inner membrane). Points 1 and 2 and the connecting vector (green)
represent the model that is manually generated for each subvolume. Dashed
lines: axes assigned to each subvolume based on orientation of vector; arrows:
assignment of search angles in PEET
 Electron Cryo-Tomography 369

6. Enter the “Volume size” in voxels (volumetric pixels). Use the

distance measurement tool (see Subheading 3.7, step 1) in
3dmod to estimate a suitable box size.
7. In the Particle Y axis box, select “user supplied csv files.” This
option defines the Y-axis for all subvolumes as the vector that
was modeled for each subvolume, which allows for a smaller
angular search, saving computational resources. Make sure
that the “…_RotAxes.csv” file for each tomogram, generated
by the stalkInit command, is in the same folder as the corre-
sponding tomogram and that both share the same name
besides the “_RotAxes.csv” extension.
8. Click “User supplied csv files” in the “Initial Motive List” box
to load the “…_In it MOTL.csv” file generated by stalkInit
into the appropriate field in the “Volume Table.” This file
specifies rotations or translations required for an approximate
alignment of each particle to the reference.
9. Generate an “Iteration Table” in the “Run” tab. PEET aligns
individual subvolumes to a reference volume by rotational and
translational movements over several iterations (reducing
search space and step size for each iteration). A new reference
is generated from a subset of the aligned and averaged subvol-
umes at the end of each run for use in the next iteration. The
“Iteration Table” specifies the search and alignment parame-
ters (see Table 2 for values that are suitable for the approach

Table 2
PEET subtomogram averaging iteration table

Angular search range High-

frequency
Phi Theta Psi filter
Search Reference
Run # Max Step Max Step Max Step distance Cutoff Sigma threshold
1 60 20 7.5 2.5 7.5 2.5 15 0.2 0.01 2/3 of all particles
2 30 10 7.5 2.5 7.5 2.5 15 0.2 0.01 2/3 of all particles
3 15 5 7.5 2.5 7.5 2.5 15 0.2 0.01 2/3 of all particles
4 7.5 2.5 7.5 2.5 7.5 2.5 10 0.5 0.005 2/3 of all particles
5 3.75 1.25 3.75 1.25 3.75 1.25 10 0.5 0.005 2/3 of all particles
6 1.875 0.625 1.875 0.625 1.875 0.625 10 0.5 0.005 2/3 of all particles
7 1.0 0.3 1.0 0.3 1.0 0.3 5 0.5 0.005 2/3 of all particles
370 Gregor L. Weiss et al.

described here). Specify the maximum range of rotation

(“Max”) as well the “Step” for “Phi” (angle around particle
Y-axis), “Theta” (angle around particle Z-axis), and “Psi”
(angle around particle X-axis). Angles and axes are indicated in
Fig.  2. Enter the “Search Distance,” high frequency filter
“Cutoff” and “Sigma,” as well as “Reference Threshold.”
Move the mouse over an input field to get more information
about the parameters. Press “Insert” to enter another
iteration.
10. Calculate the total number of subvolumes and set the numbers
of particles to be averaged (“Start,” “Incr,” “End,” and
“Additional numbers”). For example, with 53 subvolumes you
can define 10 as the start, 15 as increment, 40 as end, and 53
as additional. This will result in averages with 10, 25, 40, and
53 subvolumes, respectively. The remaining parameters
(“Optional/Advanced Features”) can be left at their default
settings. Activate parallel processing under “Options >
Settings” as subtomogram averaging is computationally inten-
sive. Click “Run” to start the computation.
11. View averages by clicking “Open averages in 3dmod” and
select the “Slicer” window (see Fig. 1f for an example). Switch
between averages with different numbers of particles using the
arrowheads at the top of the window. For an intensity thresh-
old rendering view select “Image > Isosurface” (see Fig. 1g for
an example).

4  Notes

1. The choice of the hole size and the distance between holes in
the carbon film depends on the sample and the data collection
magnification. For Quantifoils, both parameters are specified
in the grid type, e.g., R2/1 designates a grid with 2 μm hole
diameter and 1 μm hole distance. Larger hole diameters and
smaller hole distances provide more area for imaging without a
carbon background; however, they result in higher fragility
(resulting in more breakage during transfers and more beam-­
induced sample movement during data collection). The mesh
number specifies the number of squares on the grid. A higher
mesh number results in smaller area per square, which in turn
provides higher stability.
2. Alternative 300 kV instruments are the Titan Krios (FEI),
Titan Halo (FEI), and JEOL3200 (JEOL, Japan). A 200 kV
instrument, such as a Tecnai F20 (FEI) with cryosample holder
(Gatan) or Talos Arctica (FEI), can be used for thin specimens
or for screening freezing conditions. The imaging filter is criti-
cal for improving the signal-to-noise ratio by removing inelas-
tically scattered electrons, in particular for thicker samples.
 Electron Cryo-Tomography 371

3. It is usually impossible to discern whether a given cell will

express a certain secretion system prior to collecting a tilt series
and reconstructing the tomogram. It is therefore crucial to
maximize both the percentage of cells expressing the secretion
system and the number of secretion systems per cell. Some
secretion systems are inducible by specific growth conditions
or genetic manipulations of regulatory genes [53–55].
4. Thicker samples produce noisier data based on inelastic and
multiple scattering events. Certain growth media (e.g., starva-
tion media) or genetic manipulations [56] can be used to
reduce the cell diameter. Gentle lysozyme treatment will also
result in a better signal-to-noise ratio as the cells lose some of
their cytoplasmic content [57].
5. If cells are cultured on a solid medium, collect some colonies
with an inoculation loop, resuspend cells in 200 μL liquid
medium, and immediately continue with plunge freezing.
6. If cells are directly grown on grids, gold mesh material is pre-
ferred over copper to avoid cytotoxic effects. Grids are steril-
ized under ultraviolet light in a sterile workbench for 15 min
and subsequently glow discharged (see Subheading 3.2,
step 2). Using sterile forceps, the grids are then placed on the
bottom of a bacterial liquid culture in a 12 well plate (Thermo
Fisher Scientific, USA). The plate is incubated, and the cell
density on the grid can be checked using a light microscope.
7. The critical parameters for the freezing process are blot time
and blot offset. Blot time implies the time (in seconds) the
blotting paper is pressed against the grid. Longer blotting
results in thinner ice, although excessive blotting can be detri-
mental. The blot time for bacterial cells usually ranges from
1 to 10 s. Blot offset is the vertical location of a grid before the
blot is applied. This will change the grid’s position on the
wedge, affecting the gradient in ice thickness across the grid.
A good starting point for bacterial cells is a blot time of 2 s and
an offset of −3 mm.
8. Pure ethane is also frequently used as a cryogen. The advan-
tage of ethane/propane, however, is that the mixture does not
solidify even in close thermal contact with LN2. This ensures
that the cryogen can be kept at low temperatures to achieve
optimal vitrification without the need for thawing solidified
cryogen [42].
9. The number of cells on the grid can be increased by repeated
cycles of applying a sample on the grid and blotting the grid
(change the number of blottings in the option panel of the
Vitrobot software). Alternatively, blotting from only one side
(opposite side of sample) can be very effective. This can be
achieved by blotting manually with a forceps-held filter paper
372 Gregor L. Weiss et al.

or by replacing one Whatman paper inside the Vitrobot

chamber with a teflon sheet (Miroslava Schaffer, personal
communication).
10. The defocus should be chosen as a compromise based on the
following considerations: For low defocus values, the CTF
oscillates slowly, resulting in good information transfer for
higher spatial frequencies and poor information transfer for
lower spatial frequencies. The goal is therefore to choose a
defocus value as close as possible to focus while still being able
to detect the secretion systems of interest in the individual
tomograms. In a typical ECT experiment with bacterial cells, a
defocus value between −4 and −10 μm is usually chosen.
11. The tilt increment needed to obtain a certain resolution
depends on the diameter of the sample and can be approxi-
mated by the Crowther criterion [58]. In practice, tilt incre-
ments of <1° are not applicable owing to dose limitation and
the resulting low contrast in individual tilt images. Bacterial
cells are typically imaged with an increment of 1°. For subto-
mogram averaging approaches, increments can be increased to
2° or even 4°. The tilt range is limited by the grid holder and
the extremes are typically chosen in the ranges of −60° to −70°
and +60° to +70°.
Low-tilt images provide higher-quality data compared to
high-tilt images owing to specimen thickness. It therefore
became popular to start the tilt series at −30° (instead of 0°),
tilting toward 60°, followed by a second leg from −30° to
−60°. This scheme results in reduced beam damage for the
most informative low-tilt projection images and allows for the
computational removal of high-tilt information from the final
reconstruction [28, 59, 60].
12. Choose data acquisition parameters based on the following
considerations: Choose the total electron dose between 60 and
180 e−/Å2. Higher doses result in more beam damage and can
be detrimental for resolving high-resolution features (which is
particularly important for subtomogram averaging approaches).
Choose the tilt increment and tilt range (see Note 11) and
calculate the dose per tilt image. Choose a dose rate <15 e−/
pix/s (<10 for high-resolution approaches) to avoid coinci-
dence loss during electron counting [34]. Choose a high spot
size to obtain a coherent beam (typically 8–11) that makes it
possible to illuminate the entire detector. Choose an exposure
time that results in the intended total dose (typically 1–5 s).
The fast readout of the K2 detector allows for dose fraction-
ation (reading out subframes) and correction for sample
motion during image acquisition (which particularly occurs at
high tilts). To ensure enough contrast to allow for the proper
alignment of the subframes, choose a sufficient subframe expo-
sure time (typically 0.2–0.5 s).
 Electron Cryo-Tomography 373

13. Because the standard multispecimen holder LN2 dewar lasts

for only ~4 h, we usually replace it with a 1.5 L coffee thermos
placed on a lifting platform, which will last for at least 14 h.
14. For screening experiments or during data collection, it can be
useful to run automated tomogram reconstruction using pro-
grams such as batchruntomo (part of IMOD), Raptor [61],
or Tomoauto [62].

Acknowledgments

We thank D. Böck, R. Kooger, and P. Szwedziak for comments on

the manuscript. G. L. Weiss was supported by a Boehringer
Ingelheim Fonds PhD Fellowship. The Pilhofer Lab is supported
by grants from ETH Zürich, the European Research Council, the
Swiss National Science Foundation, and the Helmut Horten
Foundation.

References

1. Costa TD, Felisberto-Rodrigues C, Meir A, requires a dynamic contractile phage tail-like

Prevost MS, Redzej A, Trocker M, Waksman G structure. Nature 483:182–186
(2015) Secretion systems in Gram-negative 10. Abrusci P, Vergara-Irigaray M, Johnson S,
bacteria: structural and mechanistic insights. Beeby MD, Hendrixson DR, Roversi P, Friede
Nat Rev Microbiol 13:343–359 ME, Deane JE, Jensen GJ, Tang CM, Lea SM
2. Gan L, Jensen GJ (2012) Electron tomogra- (2013) Architecture of the major component
phy of cells. Q Rev Biophys 45:27–56 of the type III secretion system export appara-
3. Harapin J, Eibauer M, Medalia O (2013) tus. Nat Struct Mol Biol 20:99–104
Structural analysis of supramolecular assem- 11. Kawamoto A, Morimoto YV, Miyata T,
blies by cryo-electron tomography. Structure Minamino T, Hughes KT, Kato T, Namba K
21:1522–1530 (2013) Common and distinct structural fea-
4. Briggs JAG (2013) Structural biology in situ— tures of Salmonella injectisome and flagellar
the potential of subtomogram averaging. Curr basal body. Sci Rep 3:3396
Opin Struct Biol 23:261–267 12. Kudryashev M, Stenta M, Schmelz S, Amstutz
5. Lučić V, Rigort A, Baumeister W (2013) Cryo-­ M, Wiesand U, Castaño-Diez D, Degiacomi
electron tomography: the challenge of doing MT, Münnich S, Bleck CKE, Kowal J, Diepold
structural biology in situ. J Cell Biol A, Heinz DW, Dal Peraro M, Cornelis GR,
202:407–419 Stahlberg H (2013) In situ structural analysis
6. Asano S, Engel BD, Baumeister W (2016) In of the Yersinia enterocolitica injectisome. elife
situ cryo-electron tomography: a post-­ 2:e00792
reductionist approach to structural biology. 1 3. Nans A, Saibil HR, Hayward RD (2014)
J Mol Biol 428:332–343 Pathogen-host reorganization during
7. Pilhofer M, Ladinsky MS, McDowall AW, Chlamydia invasion revealed by cryo-electron
Jensen GJ (2010) Bacterial TEM. Methods tomography. Cell Microbiol 16:1457–1472
Cell Biol 96:21–45 14. Pilhofer M, Aistleitner K, Ladinsky MS, König
8. Schur FKM, Obr M, Hagen WJH, Wan W, L, Horn M, Jensen GJ (2014) Architecture
Jakobi AJ, Kirkpatrick JM, Sachse C, Kräuslich and host interface of environmental chlamyd-
HG, Briggs JAG (2016) An atomic model of iae revealed by electron cryotomography.
HIV-1 capsid-SP1 reveals structures regulating Environ Microbiol 16:417–429
assembly and maturation. Science 353:506–508 15. Radics J, Königsmaier L, Marlovits TC (2014)
9. Basler M, Pilhofer M, Henderson GP, Jensen Structure of a pathogenic type 3 secretion sys-
GJ, Mekelanos JJ (2012) Type VI secretion tem in action. Nat Struct Mol Biol 21:82–87
374 Gregor L. Weiss et al.

16. Shikuma NJ, Pilhofer M, Weiss GL, Hadfield 28. Hagen WJH, Wan W, Briggs JAG (2016)
MG, Jensen GJ, Newman DK (2014) Marine Implementation of a cryo-electron tomogra-
tubeworm metamorphosis induced by arrays of phy tilt-scheme optimized for high resolution
bacterial phage tail-like structures. Science subtomogram averaging. J Struct Biol
343:529–533 29. Bharat TAM, Russo CJ, Löwe J, Passmore LA,
17. Hu B, Morado DR, Margolin W, Rohda JR, Scheres SHW (2015) Advances in single-­
Arizmendi O, Picking WL, Picking WD, Liu particle electron cryomicroscopy structure
J (2015) Visualization of the type III secretion determination applied to sub-tomogram aver-
sorting platform of Shigella flexneri. Proc Natl aging. Structure 23:1743–1753
Acad Sci U S A 112:1047–1052 30. Suloway C, Pulokas J, Fellmann D, Cheng A,
18. Kudryashev M, Diepold A, Amstutz M, Guerra F, Quispe J, Stagg S, Potter CS,
Armitage JP, Stahlberg H, Cornelis GR (2015) Carragher B (2005) Automated molecular
Yersinia enterocolitica type III secretion injec- microscopy: the new Leginon system. J Struct
tisomes form regularly spaced clusters, which Biol 151:41–60
incorporate new machines upon activation. 31. Mastronarde DN (2005) Automated electron
Mol Microbiol 95:875–884 microscope tomography using robust predic-
19. Nans A, Kudryashev M, Saibil HR, Hayward tion of specimen movements. J Struct Biol
RD (2015) Structure of a bacterial type III 152:36–51
secretion system in contact with a host mem- 32. Zheng SQ, Keszthelyi B, Branlund E, Lyle JM,
brane in situ. Nat Commun 6:10114 Braunfeld MB, Sedat JW, Agard DA (2007)
20. Chang YW, Rettberg LA, Treuner-Lange A, UCSF tomography: an integrated software
Iwasa J, Segaard-Anderson L, Jensen GJ suite for real-time electron microscopic tomo-
(2016) Architecture of the type IVa pilus graphic data collection, alignment, and recon-
machine. Science 351:1165–1172 struction. J Struct Biol 157:138–147
21. Kühlbrandt W (2014) The resolution revolu- 33. Li X, Mooney P, Zheng S, Booth CR, Braunfeld
tion. Science 343:1443–1444 MB, Gubbens S, Agard DA, Cheng Y (2013)
22. McMullan G, Faruqi AR, Clare D, Henderson Electron counting and beam-induced motion
R (2014) Comparison of optimal performance correction enable near-atomic-resolution
at 300 keV of three direct electron detectors single-­
particle cryo-EM. Nat Methods
for use in low dose electron microscopy. 10:584–590
Ultramicroscopy 147:156–163 34. Mastronarde DN (2008) Correction for non-­
23. Campbell MG, Cheng A, Brilot AF, Moeller A, perpendicularity of beam and tilt axis in tomo-
Lyumkis D, Veesler D, Pan J, Harrison SC, graphic reconstructions with the IMOD
Potter CS, Carragher B, Grigorieff N (2012) package. J Microsc 230:212–217
Movies of ice-embedded particles enhance 35. Kremer JR, Mastronarde DN, McIntosh JR
resolution in electron cryo-microscopy.
­ (1996) Computer visualization of three-­
Structure 20:1823–1828 dimensional image data using IMOD. J Struct
24. Danev R, Buijsse B, Khoshouei M, Plitzko JM, Biol 116:71–76
Baumeister W (2014) Volta potential phase 36. Kunz M, Frangakis AS (2014) Super-sampling
plate for in-focus phase contrast transmission SART with ordered subsets. J Struct Biol
electron microscopy. Proc Natl Acad Sci U S A 188:107–115
111:15635–15640 37. Chen Y, Förster F (2014) Iterative reconstruc-
25. Briegel A, Chen S, Koster AJ, Plitzko JM, tion of cryo-electron tomograms using non-
Schwartz CL, Jensen GJ (2010) Correlated uniform fast Fourier transforms. J Struct Biol
light and electron cryo-microscopy. Methods 185:309–316
Enzymol 481:317–341 38. Agulleiro JI, Fernandez JJ (2015) Tomo3D
26. Chang YW, Chen S, Tocheva EI, Treuner-­Lange 2.0—exploitation of advanced vector eXten-
A, Löbach S, Søgaard-Anderson L, Jensen GJ sions (AVX) for 3D reconstruction. J Struct
(2014) Correlated cryogenic photoactivated Biol 189:147–152
localization microscopy and cryo-­ electron 39. Schindelin J, Arganda-Carreras I, Frise E,
tomography. Nat Methods 11:737–739 Kaynig V, Longair M, Pietzsch T, Preibisch S,
27. Rigort A, Bäuerlein FJB, Villa E, Eibauer M, Rueden C, Saalfeld S, Schmid B, Tinevez JY,
Laugks T, Baumeister W, Plitzko JM (2012a) White DJ, Hartenstein V, Eliceiri K, Tomancak
Focused ion beam micromachining of eukary- P, Cardona A (2012) Fiji: an open-source plat-
otic cells for cryoelectron tomography. Proc form for biological-image analysis. Nat
Natl Acad Sci U S A 109:4449–4454 Methods 9:676–682
 Electron Cryo-Tomography 375

40. Castaño-Díez D, Kudryashev M, Arheit M, units in electron cryomicroscopy density maps.

Stahlberg H (2012) Dynamo: a flexible, user-­ J Struct Biol 156:432–441
friendly development tool for subtomogram 52. Heumann JM, Hoenger A, Mastronarde DN
averaging of cryo-EM data in high-­performance (2011) Clustering and variance maps for cryo-­
computing environments. J Struct Biol electron tomography using wedge-masked dif-
178:139–151 ferences. J Struct Biol 175:288–299
41. Nicastro D, Schwartz C, Pierson J, Gaudette R, 53. Galán JE, Curtiss R (1990) Expression of
Porter ME, McIntosh JR (2006) The molecu- Salmonella typhimurium genes required for
lar architecture of axonemes revealed by cryo- invasion is regulated by changes in DNA super-
electron tomography. Science 313:944–948 coiling. Infect Immun 58:1879–1885
42. Tivol WF, Briegel A, Jensen GJ (2008) An 54. Basler M, Ho BT, Mekalanos JJ (2013) Tit-­
improved cryogen for plunge freezing. Microsc for-­Tat: Type VI secretion system counterat-
Microanal 14:375–379 tack during bacterial cell-cell interactions. Cell
43. Iancu CV, Tivol WF, Schooler JB, Dias PD, 152:884–894
Henderson GP, Murphy GE, Wright ER, Li Z, 55. Eichelberg K, Galán JE (1999) Differential
Yu Z, Briegel A, Gan L, He Y, Jensen GJ (2006) regulation of Salmonella typhimurium type III
Electron cryotomography sample preparation secreted proteins by pathogenicity island 1
using the Vitrobot. Nat Protoc 1:2813–2819 (SPI-1)-encoded transcriptional activators
44. Chen S, McDowall A, Dobro MJ, Briegel A, InvF and hilA. Infect Immun 67:4099–4105
Ladinsky M, Shi J, Tocheva EI, Beeby M, 56. Farley MM, Hu B, Margolin W, Liu J (2016)
Pilhofer M, Ding HJ, Li Z, Gan L, Morris Minicells, back in fashion. J Bacteriol 198:
DM, Jensen GJ (2010) Electron cryotomogra- 1186–1195
phy of bacterial cells. J Vis Exp 39:e1943 57. Briegel A, Wong ML, Hodges HL, Oikonomou
45. Brüggeller P, Mayer E (1980) Complete vitrifi- CM, Piasta KN, Harris MJ, Fowler DJ,
cation in pure liquid water and dilute aqueous Thompson LK, Falke JJ, Kiessling LL, Jensen
solutions. Nature 288:569–571 GJ (2014) New insights into bacterial chemo-
46. Radermacher M (2007) Weighted back-­ receptor Array structure and assembly from
projection methods. In: Electron tomography. electron cryotomography. Biochemistry
Springer, New York 53:1575–1585
47. Wolf D, Lubk A, Lichte H (2014) Weighted 58. Crowther RA, DeRosier DJ, Klug A (1970)
simultaneous iterative reconstruction technique The reconstruction of a three-dimensional
for single-axis tomography. Ultramicroscopy structure from projections and its application
136:15–25 to electron microscopy. Proc R Soc A 317:
48. Briegel A, Pilhofer M, Mastronarde DN, 319–340
Jensen GJ (2013) The challenge of determin- 59. Wan W, Briggs JAG (2016) Cryo-electron
ing handedness in electron tomography and tomography and subtomogram averaging.
the use of DNA origami gold nanoparticle heli- Methods Enzymol 579:329–367
ces as molecular standards. J Struct Biol 60. Pfeffer S, Burbaum L, Unverdorben P, Pech
183:95–98 M, Chen Y, Zimmermann R, Beckman R,
49. Rigort A, Günther D, Hegerl R, Baum D, Förster F (2015) Structure of the native Sec61
Weber B, Prohaska S, Medalia O, Baumeister protein-conducting channel. Nat Commun
W, Hege HC (2012b) Automated segmenta- 6:8403
tion of electron tomograms for a quantitative 61. Amat F, Moussavi F, Comolli LR, Elidan G,
description of actin filament networks. J Struct Downing KH, Horowitz M (2008) Markov
Biol 177:135–144 random field based automatic image alignment
50. Volkmann N (2002) A novel three-dimensional for electron tomography. J Struct Biol
variant of the watershed transform for segmen- 161:260–275
tation of electron density maps. J Struct Biol 62. Morado DR, Hu B, Liu J (2016) Using
138:123–129 Tomoauto: a protocol for high-throughput
51. Baker ML, Yu Z, Chiu W, Bajaj C (2006) automated cryo-electron tomography. J Vis
Automated segmentation of molecular sub- Exp 107:e53608
 Chapter 28

Structural Analysis of Protein Complexes

by Cryo Electron Microscopy
Tiago R.D. Costa, Athanasios Ignatiou, and Elena V. Orlova

Abstract
Structural studies of biocomplexes using single-particle cryo-electron microscopy (cryo-EM) is now a
well-established technique in structural biology and has become competitive with X-ray crystallography.
The latest advances in EM enable us to determine structures of protein complexes at 3–5 Å resolution for
an extremely broad range of sizes from ~200 kDa up to hundreds of megadaltons (Bartesaghi et al.,
Science 348(6239):1147–1151, 2051; Bai et al., Nature 525(7568):212–217, 2015; Vinothkumar et al.,
Nature 515(7525):80–84, 2014; Grigorieff and Harrison, Curr Opin Struct Biol 21(2):265–273, 2011).
The majority of biocomplexes comprise a number of different components and are not amenable to crys-
tallisation. Secretion systems are typical examples of such multi-protein complexes, and structural studies
of them are extremely challenging. The only feasible approach to revealing their spatial organisation and
functional modification is cryo-EM. The development of systems for digital registration of images and
algorithms for the fast and efficient processing of recorded images and subsequent analysis facilitated the
determination of structures at near-atomic resolution. In this review we will describe sample preparation
for cryo-EM, how data are collected by new detectors, and the logistics of image analysis through the basic
steps required for reconstructions of both small and large biological complexes and their refinement to
nearly atomic resolution. The processing workflow is illustrated using examples of EM analysis of a Type
IV Secretion System.

Key words Cryo-electron microscopy, Sample preparation, Single particle analysis, Image processing,
Type IV secretion system

1  EM Advances in Studies of Macro-Complexes (Type IV Secretion Systems)

The complexity of experimental and computational procedures

used for studies of the structure–function relationships of biologi-
cal complexes is growing significantly. One has to use different
approaches to uncover conformational changes linked to the func-
tional activity of the complexes. X-ray, nuclear magnetic reso-
nance (NMR), and electron microscopy (EM), combined with
biochemical and biophysical methods, allow for a deeper under-
standing of the mechanisms which underlie such macromolecular
complex functions. This was clearly demonstrated by studies of a

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_28, © Springer Science+Business Media LLC 2017

377
378 Tiago R.D. Costa et al.

ribosome [1]. Recent advances in EM, such as the invention of

direct electron detection cameras, systems for automated data col-
lection, and the development of new powerful image processing
algorithms, have dramatically expanded the range of biological
macromolecules suitable for study by this technique. The main
advantages of EM are that it does not require crystallisation of
samples and it is able to work with biocomplexes within a large
range of sizes: from ~150 kDa to several hundreds of megadaltons
[2–5]. Additionally, image processing software packages have been
significantly improved to make it possible to analyse the quality of
images and distortions caused by the microscope that prevent
obtaining high-resolution structures. New approaches were devel-
oped to reveal more consistently the sample quality: to assess its
homogeneity, to separate different conformations within the sam-
ple, and evaluate distributions of particles between different con-
formations [6–8]. A number of different packages with sophisticated
image processing algorithms are used for the analysis of macromo-
lecular complexes with different symmetries or asymmetry. Another
positive aspect of modern achievements is that computing power is
steadily increasing, making it possible to analyse many hundreds of
thousands of particle images from heterogeneous samples.
Nonetheless, the basic workflow of sample imaging and image pro-
cessing remains the same (Fig. 1) [9].
This tremendous success in improving the resolution of EM
structures during the last decade would not have been possible
without cryo sample preparation and what we now call cryo-
­EM. This approach for sample preparation and the combination of
methods of X-ray crystallography with EM have made it possible to
achieve near-atomic-resolution details for some of the Gram-­
negative bacterial secretion systems. These structures have given us
unprecedented understanding of the mechanistic details of how
bacteria assemble these highly specialised nano-machines to secrete
proteins and DNA to the bacterial extracellular space and to the
eukaryotic or bacterial target cell. Among the Gram-negative bac-
terial secretion systems, Type IV Secretion Systems (T4SS) possess
the unique ability to secrete proteins, DNA, or protein–DNA com-
plexes in an adenosine triphosphate–dependent process. Given the
competency for T4SSs to secrete such a variety of substrates
involved in the pathogenesis and spreading of conjugative plasmids
encoding antibiotic resistance genes, this secretion system become
an important target for structural biology studies [10].
Among all Gram-negative T4SSs, those encoded by the pTi
plasmid of Agrobacterium tumefaciens, together with the conjuga-
tive pKM101 and pR388 plasmids from E. coli, are the best char-
acterized. This macromolecular structure is composed of 12
proteins: VirB1–VirB11 and VirD4 [11]. The first main advance in
the understanding of the general architecture of a T4SS took place
when the cryo-EM structure of the so-called core outer-membrane
 Cryo Electron Microscopy 379

Fig. 1 Workflow of EM structural analysis. In green is the experimental part of

structural analysis. The computational part is shown in light and dark blue; the
initial steps of processing are shown in light blue. They include image frame
alignment, CTF correction, normalisation, and filtering. The subsequent steps—
alignment, statistical analysis, determination of particle orientations, and initial
three-dimensional reconstruction (3D)—are shown in dark blue. The final step
(light purple) is the interpretation of the maps obtained

complex (OMC, encoded by the conjugative pKM101 plasmid)

was solved with a resolution of 15 Å. This 1.1 MDa structure,
which spans both the outer and inner membrane, is made of 14
copies of VirB7, VirB9, and VirB10 proteins (Fig. 2a) [12].
Further, the resolution of the same core OMC was improved to
12.4 Å, which provided further details on the structural organisa-
tion of the proteins that form this complex (Fig. 2b) [13]. Recently,
the almost complete full structure of the T4SS (VirB3–VirB10)
encoded by the conjugative R388 plasmid was solved by negative
staining (NS). This remarkable structure provided the first view of
both the outer and the bipartite inner-membrane complex (IMC)
and how these are linked by a structure called stalk (Fig. 2c) [14].
This review might not be complete from the point of view of a
specialist and it might not provide sufficient mathematical back-
ground for the reader. However, we will try to give a general
380 Tiago R.D. Costa et al.

Fig. 2 EM maps of different T4SS structures. (a) Cryo-EM structure of core outer-membrane complex at 15 Å
resolution. (b) Cryo-EM structure of the core outer-membrane complex at 12.4 Å resolution. (c) Structure of
almost complete the T4SS complex (in negative stain) at overall resolution of ~20 Å

overview of imaging using an electron microscope and the current

basic steps of structural analysis. This will include the outline of
sample cryo preparation, the effects of radiation damage, and
advances in the procedure of data collection. We will describe steps
considered to be pre-processing determination of orientations of
the particle images, and methods used to obtain structures and
how they can be evaluated. Since this is a rather short review, we
will not describe here how an image is obtained in an electron
microscope. This information can be found in other reviews and
books [8, 15]; readers interested in more details on the topics
described here may consult the references provided at the end of
the chapter.

2  Sample Preparation in Cryo-EM

Although EM provides much better resolution than light micros-

copy, it has the disadvantage that samples must be imaged in a
vacuum. This is due to the fact that images are created by a beam
of electrons in a transmission electron microscope. Without a vac-
uum, electrons become very quickly absorbed by air since they col-
lide with air molecules and lose their energy and direction of
scattering, therefore to obtain a high-quality image of a sample, it
is necessary to keep the electron path free of air molecules (under
a vacuum) to allow electrons to move directly to the sample. In
their native conditions biological objects (macromolecules and
cells) are immersed in water solutions. To be visualised in an elec-
tron microscope under vacuum the bio complexes have to be made
rigid and stable to avoid drying out or undergo structural changes 
during the exposure time at data collection.
 Cryo Electron Microscopy 381

Fig. 3 Cryo-EM sample preparation. (a) 3 mm copper mesh grid covered with a film of holey carbon. (b) Magnified
image of square patch showing microscopic holes in carbon. (c) Enlarged image of a single hole containing a
layer of vitrified ice with protein molecules. (d) Cross section of a hole with particles embedded in ice

Cryo-EM methods of sample preparation allow to preserve the

structural integrity of biocomplexes, keeping them in a nearly
native hydrated state in the vacuum system of the microscope. The
method proposed by Dubochet et al. [16, 17] is now a well-­
established, standard technique for freezing aqueous solutions of
samples on cryo-EM grids. The EM grid is a metallic round plate
(~3 mm in diameter and usually made of copper) with a fine mesh.
The size of the mesh is typically chosen depending on the experi-
ment, but the most commonly used type has 400 squares per inch
(Fig. 3). Depending on the sample, a continuous thin layer of car-
bon film or one perforated with irregular (lacey grids) or regular
holes should be put on the top of the metal grid. One can use
manufactured grids with regular holes in carbon films. The grids
should be chosen according to the size and shapes of holes and
distance between holes, which are most suitable for the given sam-
ple (e.g. Quantifoil grids, Quantifoil Micro Tools GmbH; C-flat
grids, Protochips, Inc.) with regularly arranged holes for auto-
mated and manual data collection (www.protochips.com; http://
www.agarscientific.com).
A drop of a sample (~3 μL) is applied to a glow-discharged (to
make the surface more hydrophilic) grid; the sample is kept for a
short time on the grid (0.5–2 min, depending on the sample) and
then the gird is maintained on a plunger. Excess sample is blotted
to make a thin layer of sample solution and then the grid ­immediately
plunged into liquid ethane (or propane) that has a temperature of
−182 °C. The ethane must be cooled prior sample freezing by
liquid nitrogen (Fig. 4). Plunge freezing in liquid ethane takes
place in ~10−5 s, trapping the biological molecules in their native,
hydrated state embedded in amorphous ice that is like solid water.
Cooling by plunging into liquid ethane is much faster than plung-
ing directly into liquid nitrogen because the liquid ethane is used
382 Tiago R.D. Costa et al.

Fig. 4 Sample vitrification. Left panel: grid with sample applied being held in twee-
zers; right panel : after blotting excess of a sample, grid is plunged into container
filled with liquid ethane. The top level of the container should be immersed in the
nitrogen atmosphere that has a temperature slightly above the temperature of
liquid nitrogen. The grid is then transferred into the grid holder. The transfer must
be done without taking the grid out of the nitrogen atmosphere

near its freezing point so it does not evaporate and produce an

insulating gas layer. This fast freezing prevents the formation of ice
crystals and keeps samples in a nearly native hydrated state [18, 19].
More detailed information on the vitrification of samples can be
found in papers by M. Samsó and R.A. Grassucci [20, 21].
The grids must be kept all the time at temperatures not higher
as –170 °C (in a storage, during transfer to the microscope using a
cryo-transfer holder, and at imaging in the electron microscope);
otherwise, ice will change its conformation and start to make crys-
tals that will destroy the sample and contaminate the grids. Another
important advantage of cryo-EM is that liquid nitrogen tempera-
tures reduce the radiation damage induced by electron beam when
it passes through samples [22, 23]. Nowadays automated and con-
trolled devices (Vitrobots) have been developed, thereby allowing
higher reproducibility in grid preparation [24, 25]. However, it is
recommended that the first estimation of the sample quality should
 Cryo Electron Microscopy 383

be done using the NS technique, which is fast, robust, and reliable

[26]. It allows a quick assessment of sample quality, concentration,
and suitability for the subsequent cryo preparations.

3  Image Acquisition by Digital Detectors

3.1  CCD Cameras EM, like all modern photographic systems, is related to the acqui-
sition of images and now uses digital cameras. All digital cameras
convert analogue optical signals into digital format so the steps in
developing and scanning film have become redundant. The first
digital cameras used charge-coupled device (CCD) sensors, which
had been invented in 1969 by W.S. Boyle and G.E. Smith at Bell
Telephone Laboratories (Nobel Prize, 2009) [27]. The concept
behind this device was based on the transformation of an analogue
signal, such as photon energy, into an electrical charge using spe-
cially designed photo sensors. The magnitude of the charge regis-
tered by the sensor is proportional to the energy of photons
absorbed by the sensor. CCD chips consist of an array of photosen-
sitive elements. In the readout mechanism, charges are successively
transferred to a reading register, amplified, and converted into a
digital signal. The number of reading registers determines the
speed of the CCD image recording. Since there are only a few of
them, the readout of these cameras is not very high.
However, in electron microscopes, the process of recording
electrons is more complicated than with photosensors, which are
not able to register electrons. Therefore, the sensors were modi-
fied in such a way that a mono- or polycrystalline scintillator that
converts electron energy into photons was placed on top of the
photosensor, and only then were photons converted into an elec-
trical signal (Fig. 5a) [28, 29]. Unfortunately, a CCD camera’s
sensitivity decreases with increased voltage of an electron micro-
scope, so thicker scintillator layers are needed to improve the elec-
tron detection efficiency. Thick layers of scintillators affect the
image quality, which is degraded because the higher-energy elec-
trons are scattered through several adjacent sensors, reducing the
image resolution. Nonetheless, these cameras provided experience
and understanding for the development of automated data collec-
tion in EM.

3.2  Direct Electron In the last decade, new digital detectors have been designed
Detectors enabling the detection of electrons without the intermediate step
of transforming electrons into photons and then into an electrical
signal (Fig. 5b). Direct detector devices (DDDs) use an array of
radiation-hardened active pixel sensors (a pixel circuit) which are
integrated into a silicon complementary metal-oxide semiconduc-
tor (CMOS) chip [30, 31]. In this case the electron energy is trans-
formed directly into an electrical signal. Another advance in this
384 Tiago R.D. Costa et al.

Fig. 5 Digital cameras. (a) In CCD-based cameras, the electrons hit a scintillator,
generating light, which is partially captured by fibre optics, and directed onto the
cooled CCD chip. (b) In direct detectors (two right panels ), active pixel sensors,
mostly based on CMOS technology, are capable of capturing and directly detect-
ing incident electrons

technology is that an amplifier is built into each pixel and allows

fast signal readout from each individual sensor (or pixel) nearly
simultaneously. This makes it possible to separate a single exposure
into a set of smaller subexposures. In cryo-EM, this provides a
valuable option for electron dose fractionation that is important in
studies of radiation-sensitive biological samples. The resultant
image subframes can then be used for specimen drift correction,
which is not possible with conventional CCD cameras.
The photon-electron  conversion step is removed by the
CMOS semiconductor technology, where the fibre optics became
unnecessary allowing to improve the signal-to-noise ratio (SNR) in
images registered by DDD compared to those from CCD. The
quality of DD detectors is best described in terms of the detective
quantum efficiency (DQE) [32, 33]. The DQE is a measure of the
efficiency of signal transfer by the camera and defined as a ratio of
the SNR in the output image registered by the camera sensors to
the SNR at the input image:

DQE = (SNRout ) / (SNRin )

2 2

The ratio depends on the spatial frequency (sizes of the details)
of the image. A perfect detector would not distort the input signal,
so in an ideal system the output should be the same as the input.
Therefore, the DQE of an ideal system would be equal to one for
all frequencies. In reality, cameras distort the fine details in images,
and this is reflected by a significant decline in the DQE at high
frequencies [34, 35].
Direct detectors make it possible to register electrons within a
high range of energy and are now used in 300 keV microscopes.
 Cryo Electron Microscopy 385

The high sensitivity of such a system has made it possible to reduce

the size of sensors, and additional improvements in software have
offered a new mode of image exposure, such as counting mode,
where the system records single electrons, as implemented in
Gatan’s K2 cameras [33, 34, 36, 37].

3.3  Micrograph Cryo-EM images using DDD cameras can record from 7 sub-
Subframe Alignment frames on Falcon (FEI) to 40–50 subframes using direct electron
(DE) or K2 (Gatan) cameras. Thus, the data recorded with DDD
cameras represent sets of image frames (movies) that can be
motion-corrected. Such high rates of image recording can reveal
distortions of images induced by drift of the grid (sample) within
the EM. Typically, frame alignment starts from frame N−1 that is
aligned to the last frame. Then these two images are summed, and
frame N−2 will be aligned to this sum. Then frames N−1 and N−2
will be summed, and frame N−3 is aligned to this new sum. The
process is repeated in the same way towards the first frame. There
are variations in algorithms when summation is done not for two
frames for the following alignment but for four or five frames or all
frames. Alignment is refined iteratively when on the next round of
alignment the total sum obtained during the previous round is
used as a reference. The entire procedure improves, firstly, SNR of
the reference and then the quality of the alignment. Nowadays a
number of software packages can be used for the frame alignment
[36, 38–41]. The image shown in Fig. 6a represents the sum of the
original frames without any correction. The trajectory of the image
shift during these several exposures indicates that the movement of
initial shifts of the sample is large initially but then slows down
(Fig. 6b). A power spectrum from the sum of the frames without
motion correction demonstrates that the Thon rings are not very
sharp: they fade fast owing to the small shifts in different directions
shown in Fig. 6b (red dots). When the movie frames are aligned
(the motion correction), the Thon rings become symmetrical,
going up to 3 Å (Fig. 6c). This indicates the presence of high-­
resolution details in the images. Summation of the motion-­
corrected subframes generates the final sharper image (Fig. 6d).

3.4  Radiation Images in an electron microscope are generated by the electron

Damage beam that illuminates the sample and then the image is formed by
electromagnetic lenses in the plane of the camera. While a short
wavelength of the electron beam improves dramatically the resolu-
tion of images of biological molecules, it was proved that biosam-
ples are very sensitive to the high-energy electron irradiation that
takes place during imaging. Changes in biological complexes
depend on the time of the overall exposure (cumulative) dose and
were estimated using spot fading diffraction experiments on two-­
dimensional (2D) crystals [42–44]. Therefore, a 3D structure
derived from experiments when a sample was overexposed can
386 Tiago R.D. Costa et al.

Fig. 6 Cryo-EM images with motion correction. (a) Representative cryo-EM image of vitrified T4SS particles. (b)
Trace of motion in X and Y directions of frames. (c) Left : power spectrum from the sum of raw movie frames
without motion correction. Right : power spectrum from the sum of movie frames after motion correction. (d) Sum
of movie frames that were shifted according to determined shifts shown in (b). Protein is black in these images

differ remarkably from the structure of the native molecule. High-­

energy electrons of the electron beam in EM may induce displace-
ments, bond breakage, and mass loss of low-atomic-­ number
elements such as carbon, nitrogen, and oxygen [45]. It has been
shown by crystallography that exposure of crystals to X-rays
induces decarboxylation of glutamate and aspartate residues, the
breakage of disulphide bonds, and the loss of hydroxyl groups
from tyrosine and the methylthio group of methionine [46].
Cryo-EM imaging carries the major benefit of reducing radia-
tion damage as samples are kept at cryogenic temperatures during
imaging. Vitrified samples preserve their native structure and are
 Cryo Electron Microscopy 387

imaged well at liquid nitrogen temperatures [22, 47]. Such low

temperatures increase tolerance to ionising radiation damage [44,
48] since the free radicals generated from inelastic scattering events
are unable to diffuse through the sample and cause secondary
damage [49]. In addition, the freezing also constrains the move-
ment and degrees of freedom of the atoms of a molecule after a
bond is broken, thereby limiting the structural rearrangement pro-
duced during irradiation [44]. As a result, keeping and imaging
samples at liquid nitrogen temperature improves radiation toler-
ance two- to sixfold over room temperature imaging [44, 48].
Another important approach has been used for a number of
years in EM which is the usage of the low-dose mode during data
collection. Low-dose imaging is based on reducing the amount of
time a sample is exposed to electrons by focusing on an adjacent
area that is sufficiently close to the area of interest but does not
overlap it. All modern electron microscopes that are used for bio-
logical studies come with pre-installed low-dose software allowing
for efficient exchange between imaging modes. The search mode is
a low-magnification overview image used to identify areas of inter-
est while the imaging (or photo) mode is used for actual data col-
lection at high magnification. The focus mode is typically set at a
higher magnification than imaging mode, but the beam is shifted
to an adjacent area. Such an interchange between these modes is
implemented in systems for automated data collection and allows
significant reduction of radiation damage.
The next and now very fast evolving method is the usage of
dose fractionation, which is provided by current direct detector
technology. DDDs have a very high speed of frame readouts.
Depending on the detector type (FEI, Gatan, or DE) and the
available software, it is possible to record from 7 to 60 subframes
per exposure. Typically, the images in the first two or three frames
demonstrate large sample shifts, while later on the movement slows
down. However, the last frames indicate often that the sample has
been damaged by the beam (Fig. 6b). Bartesaghi and co-authors
compared the density maps reconstructed from different fractions
of the total exposure (10, 20, or 30 e−/Å2). Analysis of these high-­
resolution cryo-EM structures show that densities for residues
with positively charged and neutral side chains are well resolved,
while the residues with negatively charged side chains, having
weaker densities, were less resolved [50]. The negatively charged
glutamate and aspartate show on average 30% less density than the
similarly sized neutral glutamine and asparagines [50], which is
consistent with observations in X-ray analysis [46]. Therefore, the
user can use all frames for the quality assessment of images and
samples and then only the first half or first two thirds of subframes
(depending on the type of the DDD used in experiments) for the
reconstruction of the native complex [42, 43, 48, 50, 51].
388 Tiago R.D. Costa et al.

4  Image Analysis of Micrographs

4.1  Contrast Transfer The aim of single-particle reconstruction is to obtain an accurate

Function representation of the 3D structure of a molecule using a set of 2D
projection image data. The macromolecular complexes are consid-
ered as thin objects so their images can be described as linear projec-
tions of the Coulomb potential of the molecular complex [44]. This
is a primary condition necessary for the subsequent reconstruction
procedure. However, images produced by electron microscopes do
not directly represent projections of the molecules under study.
Deviations from the real densities of projections are induced by
aberrations in the optical system of the microscope [52].
The function that describes the actual representation of every
single point in the registered image of a theoretically correct pro-
jection is called the contrast transfer function (CTF) of the micro-
scope [44, 52, 53]. The CTF is defined by the acceleration voltage
(electron wavelength), the type of electron source (beam coher-
ence), and aberrations of the objective lens (Cs, Cc, and astigma-
tism). The major factors affecting the CTF are the degree of
spherical aberration (Cs) of the objective lens and level of defocus
(Δf). As a result, the CTF modulates the amplitudes and phases of
the electron diffraction pattern formed in the back focal plane of
the objective lens. For any given defocus setting the features of a
specimen are modulated through positive and negative contrast.
The CTF limits the amount of information which can be obtained
from electron images. At zero crossings of the CTF no information
is transmitted, and specimen features corresponding to such spatial
frequencies will not be visible in the final image.
A transmission electron microscopy (TEM) image can be rep-
resented as a power spectrum (Fourier space), which demonstrates
the magnitude of the various frequency components contained
within the image (Figs. 7 and 8). The effect of the CTF on an
image is that the power spectrum looks like it is oscillating and
appears as concentric rings or Thon rings [54], which indicate the
location of the minima and maxima in frequency space. Dark
regions show the positions of all zero crossings of the CTF, and
bright regions correspond to areas where the CTF has either posi-
tive or negative contrast (Figs. 7 and 8).
The major effect on images of biological samples from the
spherical aberration of the objective lens is to cause phase changes,
and therefore the representation of densities in image is altered
significantly. On the other hand, biological samples viewed in ice
under close to focus conditions demonstrate very little amplitude
contrast since the difference in their densities and water density is
very small [8, 44]. Thus, the images are typically taken far from
focus (in underfocus mode) to increase the weight of low frequen-
cies and thereby improve the visibility of particles [8, 44]. Here it
 Cryo Electron Microscopy 389

Fig. 7 CTF with envelope function. Dotted blue line : amplitude of all frequencies
in perfect microscope; green line: effect of envelope function on CTF (red) result-
ing in suppression of high spatial frequencies

Fig. 8 Assessment of CTF parameters. (a) Comparison of theoretically calculated

CTF (left bottom quadrant ) with CTF seen in experimental spectrum. For an
accurate CTF determination the Thon rings from both image parts should match
accurately. (b) Identification of axes of astigmatism which are superimposed
over Thon rings of an actual observed power spectrum and compared with the
theoretical spectrum. The spectrum of a micrograph shown here indicates that
there is a small astigmatism, ~2%, and the axes of ellipse are slightly tilted,
shown in light blue

should be mentioned that low frequencies are responsible for the

overall shape and appearance of particles in images. However, high
defocusing induces changes in the distribution of density informa-
tion related to fine details that could be lost owing to the attenua-
tion of amplitudes at high frequencies. The level of defocus used
for imaging depends on the size of the biocomplex. The images of
small particles (~100–300 kDa) are taken with a large defocus,
sometimes up to 6–7 μm, while viruses with diameters of at least
50 nm can be imaged at 0.5–1.0 μm.
390 Tiago R.D. Costa et al.

The CTF for biological samples can be described by the

formula

Phase CTF = -2 sin éëp ( Df lq 2 - C s l 3q 4 / 2 ) ùû ,

Phase CTF Cs = spherical aberration constant; Δf = defocus; q =
spatial frequency; λ = electron wavelength. The spherical aberra-
tion coefficient and the electron wavelength are the only constants,
and these values remain fixed for each electron microscope [52].

4.2  Defocus To correct an image for CTF effects and obtain an image that cor-
Determination responds to the projection, it is necessary to determine its defocus
and Correction of CTF and to check it for astigmatism and drift. The nominal value of
defocus set on the microscope does not usually represent the actual
defocus obtained in the final digital image or micrograph. This
occurs because, although the acceleration voltage and spherical
aberration remain constant, the common deviations in sample
thickness and the position of the supporting film will cause local
variations in defocus. As a result, the CTF related to the defocus
should be determined for each image frame. Finding the exact level
of defocus and astigmatism in cryo-EM images is of absolutely cru-
cial importance when working on the production of a high-­
resolution structure.
CTF determination is performed by calculating the sum of
power spectra (or amplitudes) of small patches (256 × 256 or
slightly larger) from the sum of all subframes. This spectrum is cor-
related with a number of CTFs theoretically calculated in a range
of possible defocus values. A maximum correlation between the
observed and a theoretical CTF would indicate the actual defocus
of the values and will define frequencies where the phases must be
flipped (Fig. 8). Different options for (semi-) automated defocus
determination are available in a number software packages, such as
EMAN2, CTFIND, and IMAGIC5 [55–57].
Astigmatic images have power spectra which are not rotation-
ally symmetric, and this can complicate and reduce the accuracy of
CTF determination. Generally cryo-EM images which have greater
than 5% astigmatism are not used for further processing except in
special cases where strong astigmatism can be used to recover
information in areas where the CTF crosses zeroes. The level of
astigmatism can be calculated as follows:

(
Astigmatism = Defocus max - Defocus min / Defocus avg )
An EM projection image is only considered a faithful represen-
tation of the observed object of interest if it has been corrected for
the CTF modulation effects of the microscope. This can only be
done after CTF determination. Phase correction is carried out in
reciprocal space by multiplying the alternating rings of the CTF by
 Cryo Electron Microscopy 391

Fig. 9 CTF correction. (a) CTF oscillates changing contrast from negative to posi-
tive depending on frequencies. Information is lost only where CTF crosses zero
line. (b) Negative lobes of uncorrected CTF are flipped over to positive (correction
of CTF by phase flipping). The missing information can be recovered by collecting
images at different defocus levels which fill these zero regions with information

−1 at positions where the contrast transfer is negative and +1 where

the contrast transfer is positive. This has the effect of reversing or
“flipping” the negative lobes of the CTF into positive contrast,
thereby restoring the correct image phases (Fig. 9).
The missing information where the CTF crosses zeroes are
restored by combining images at different defocuses, so that where
some images lack spatial information at a particular frequency oth-
ers will provide the complementary missing information. High
spatial frequencies are suppressed by envelope decay, so amplitude
correction is also important for maximising high-resolution details.
This operation usually involves applying a Wiener filter [58] to
remove noise from the CTF prior to amplitude amplification.

4.3  Particle Selection The structural analysis process in EM begins with selecting images
of individual particles from micrographs. This involves recording
their unique locations (x,y) within the image field and saving these
coordinates in a data file that will be used in the next steps of pro-
cessing. This can be done interactively using packages like Xmipp
[59], EMAN2 [55], Ximdisp [60], RELION (semi-automated
selection of cryo-EM particles in RELION-2 [61], and others.
The simplest way to do this is to select a single particle image by
clicking on the image with a mouse. The coordinates of these
points will be stored and then used to extract individual particles
within a square box of designated dimension, for example, 500 ×
500 pixels. The cut-out area must be large enough to retain all the
image data around the object with as little background as possible.
Particles can also be selected automatically with particle identifica-
tion/selection programs, for example, Autopicker [62], BShow
[63], and FindEM [64]. These programs use the assessment of
local correlation to measure the degree of similarity between refer-
ence images and then a small area of the raw micrograph. Areas
which show maximum correlation to the references are boxed out.
392 Tiago R.D. Costa et al.

Fig. 10 Particle picking. (a) Cryo-EM micrograph of T4SS core outer-membrane

complex. (b) Particles outlined by yellow squares represent end views; side
views are outlined by squares in cyan

The poor contrast in ice images and the presence of artefacts which
could resemble target particles can generally increase the inaccu-
racy of automatic selection.
The images of selected particles must not overlap with other
particles and particles in images should not be distorted. In Fig. 10
we show an example of a micrograph of a vitrified sample of the
T4SS core OMC. Images were recorded on a 4096 × 4096 Gatan
CCD camera with a low electron dose on a Tecnai F20 FEG micro-
scope operating at a voltage of 200 kV, a magnification of 68,100,
and a defocus range of 1250–3500 nm.

4.4  Normalisation The normalisation of all images is an essential pre-processing step.

of Data The contrast and intensity can vary from image to image during
data acquisition, even when all the EM settings are the same. This
effect arises because of a number of factors, including differences in
the thickness of the carbon support film or ice, particle orienta-
tions, uneven staining, or the merging of images from different
data collection sessions. Normalisation standardises the densities of
images by setting the mean pixel grey value of each image to the
same level, commonly zero, and also rescaling the standard devia-
tion to an equal value for every particle image. Without normalisa-
tion the density variations, such as very bright or very dark regions
within an image, could bias the cross-correlation procedures which
are later used for alignment and calculation of particle classes.
Typically, normalisation is based on the following formula:

rinorm
,j = (( r
i, j ) )
- ravg / s old s new ,

σold and σnew are the standard deviations of the original and target
images respectively, and ρi,j is the density of a pixel in the image
array coordinates.
 Cryo Electron Microscopy 393

Fig. 11 Alignment and classification of T4SS core outer-membrane complex

images. Upper panel : representative images of core outer-membrane complex;
middle panel : class averages of images at nearly the same orientations; bottom
panel : corresponding reprojections from final 3D model

4.5  Alignment One of the most important steps in image processing is the reduc-
of Particle Images tion of noise and enhancement of the signal. Different factors,
such as insufficient coherence of the electron beam, quality of
amorphous ice (due to uneven distribution of salts and some other
effects in buffers), supporting films, and the noise of registering
cameras, contribute to decreasing the SNR in particle images.
Since these types of noise are not related to a signal from the sam-
ple, the averaging of particle images improves the SNR. However,
to retrieve reliable information using averaging, images must rep-
resent the same particles in the same orientations [44]. Therefore,
images of particles in the same orientations should be identified,
and before averaging, the images should be aligned rotationally
and translationally with respect to each other. Alignment compares
all images to a reference image and shifts them so that they are in
the same position as the reference. Usually, normalised particle
images should be centred or aligned to the reference that repre-
sents a typical view of the complex.
One of the possible options for starting analysis is to align all
images of a data set to the rotationally averaged total sum of all
images (with no alignment). The centred images are then sub-
jected to multivariate statistical analysis (MSA) (see following dis-
cussion) to obtain a number of averages of images which are
aligned only translationally and grouped according to common
features. The best characteristic views (averages of groups of images
with the lowest variations among them) are used for multi-­reference
alignment. The most reliable classes are centred and used as new
references for the next round of alignment (Fig. 11). The proce-
dure can be repeated several times alternating with MSA [65]. In
other cases one can use alignment in Fourier space [66] or so-­
called reference-free alignment implemented in EMAN2 and
SPIDER [55, 67].
394 Tiago R.D. Costa et al.

4.6  Statistical To improve the SNR, aligned images which have high similarity
Analysis between each other should be grouped together, and for this pur-
and Classification pose, statistical analysis and classification are used. Different
approaches have been developed to reduce a large number of vari-
ables to a limited number of important parameters [68].

4.6.1  Principal Principal component analysis (PCA) reduces the number of variables
Component Analysis to find the most significant variations in the measurements [44, 65].
The essence of the procedure is a transformation of a set of observa-
tions of possibly correlated variables (in our case images) into a set of
values of uncorrelated variables called principal components. In the
complete representation, a number of the principal components are
equal to the number of original variables. However, since images
contain a high level of noise, the number of the meaningful compo-
nents becomes much smaller. The principal components are described
by the eigenvectors of the data matrix. PCA is the simplest of the
true eigenvector-based multivariate analyses [65, 68].

4.6.2  Factor Analysis Factor analysis is designed to identify variations in a number of

original variables using predefined “significant” factors that are
often defined by a researcher [44]. That requires specific assump-
tions about the underlying features of the object under study, such
as average density or the perimeter and size of specific domains.

4.6.3  Maximum Maximum likelihood estimation (ML) is a method that assesses

Likelihood Estimation parameters that would correspond to a statistical model. When
applied to a data set (such as our image data set) and given a statisti-
cal model (the initial 3D model), ML provides estimates of how our
new reconstruction would correspond to the proposed model and
what sort of deviations could be observed. This concept can be
stated in different words: once a model is specified with parameters
(to a certain extent) and data have been collected (our EM images),
one is able to evaluate how well the model fits the observed data.
The quality of this correlation is assessed by finding parameter val-
ues of a model that best fit the data—a procedure called parameter
estimation. In the EM case, that would be a 3D model that corre-
sponds in the best way to the data set; otherwise, the model must be
modified. ML has many properties which should be taken into
account during estimation: sufficiency (complete information about
parameters reflecting features of interests), consistency (numbers of
images related to this or some other 3D model), efficiency (lowest
possible variance of parameter estimates), and perhaps other practi-
cal parameters [69, 70]. This method is successfully used in the
analysis of 3D reconstructions and implemented in RELION [71].

4.6.4  Classification Classification is done once the principal components or important

factors of the data are defined. Cluster analysis is a tool to identify
groups of similar objects. This kind of analysis is used for grouping
 Cryo Electron Microscopy 395

similar images (particles in the same orientations), and in EM, two

implementations are used. One is K-means (used in SPIDER,
EMAN, and XMIPP [55, 59, 67]), where the user defines a num-
ber of classes (K, typically not bigger than 10) that should be
obtained and the algorithm randomly assigns each image to one of
the classes [72]. These starting (random) points are called cen-
troids or seeds. The centroids should be placed as far away from
each other as possible. The next step is to take each point belong-
ing to a given data set and associate it to the nearest centroid.
Averages are calculated for each class, and the distance between
each image and the obtained averages are calculated (Fig. 12a).
A new class will be formed by the images that were closest to one
of the averages. Then the class averages are recalculated. This pro-
cess is done iteratively until the images stop moving between
classes. The K-means method is reasonably fast and works better in
low-­dimensional space since dimensionality increases time and a
local minima problem may occur.
Another method of classification is hierarchical ascendant clas-
sification (HAC) (implemented in IMAGIC, SPIDER, and
EMAN2). There are two main streams: agglomerative, which is a
“bottom-up” approach where each observation starts in its own
cluster and pairs of clusters are merged as one moves up the hierar-
chy, and divisive, a “top-down” approach where all observations
start in one cluster and splits are performed recursively as one
moves down the hierarchy. HAC is based on the Ward criterion
[73], which minimises the intra-class variance while maximising
the inter-class variance. In IMAGIC a version of the agglomerative
approach is used. This criterion is used in the pair-wise merging of
the classes that are supposed to be obtained to form a HAC tree
(Fig.  12b). The user chooses a number of classes that would be
needed for subsequent processing, and the HAC tree is cut at that
level. Sometimes it is hard to achieve the lowest intra-class variance
since this algorithm does not allow one to move images from one
class to another. However, this can be achieved by weighting
parameters during reclassification.

4.7  Determination To obtain the 3D structure of a biocomplex from EM images, the

of Particle Orientation orientation of each of the individual particle images must be deter-
mined. The location of an individual molecule can be identified by
the X, Y, and Z coordinates, and the shifts of different particles
with respect to each other can be described in the same way as
shifts in X, Y, and Z. The particles can also be rotated by α, β, and
γ angles, called Euler angles. This means that a molecule has six
degrees of freedom in space. In the microscope images correspond 
to the projections along the Z-axis of the translational system of
coordinates, so the shift in the Z direction is not significant (we
assume that the electron beam is parallel), but the shifts in the X
and Y directions should be determined. During translational
396 Tiago R.D. Costa et al.

Fig. 12 Principles of image classification. (a) K-means classification. Experimental

data represented by empty circles with their characteristic parameters v1 and
v2. The initial randomly selected seeds are shown in light blue and pink . Results
of classification steps are shown in circles coloured correspondingly. (b)
Hierarchical classification. The left panel shows position of points in plane and
their successive forming of classes; the right panel shows the classification tree

alignment, the centre of the molecule is set to X = 0 and Y = 0. To

calculate the 3D map from individual images or class sums, it is
necessary to determine the orientations of the characteristic views
(classes) relative to each other.

4.7.1  Random Conical The random conical tilt (RCT) technique is used to obtain an ini-
Tilt tial model of a new complex about which little is known. The
method is based on a typical property of samples having a prefer-
ential orientation on a grid when negatively stained. The RCT
 Cryo Electron Microscopy 397

approach is a reliable method for generating an unbiased initial 3D

model obtained by experimental measurements. Two electron
micrographs of the same part of the grid (at magnification ~30–40K)
are required; the first one is typically at a high tilt (45–60°, where
a goniometer is used), and the second image of the same area is
taken without tilt [74]. The tilt respectively to the Z axis is known
(the same as that used for the tilt of the first image), and rotation
around the Z axis is obtained from images of the same particles but
in the untilted micrograph. As soon the tilted images are centred,
angles are assigned to them (Euler angles). With the relative orien-
tation in space determined, a 3D reconstruction of the object can
be produced [44]. Although such a model could be far from pre-
fect, it will be a good start for subsequent refinement.

4.7.2  Projection Projection matching requires an initial model and is based on a

Matching simple principle of comparison of images with projections of the
model [44]. As a 3D template (an initial map), one can use the
low-resolution negative-stain EM 3D reconstruction, the low-pass
filtered X-ray model, or the EM map of a homolog. This template
is projected in all possible directions covering an entire Euler
sphere with a certain angular increment. Then the images or class
averages of the data set are compared with these references, and
the angles corresponding to the reference with the best cross-­
correlation will be assigned to the image [67]. Projection match-
ing helps to determine out-of-plane rotations of the object. During
the angular determination refinement, the angular increment
between projections is reduced, or additional projections with a
small increment can be calculated around the initial angle. This
method is easy to use; however, it is extremely time consuming due
to a long computation during which it is necessary to try all pos-
sible in-plane alignments and to compare each image to a set of
references. Nonetheless, multi-processor computers can be used to
speed up the process. Once the Euler angles are assigned to all
images or class averages, a new 3D reconstruction will be calcu-
lated and a new set of refined higher resolution model projections
computed for the next round of projection matching. Several dif-
ferent software programs, such as IMAGIC [57], EMAN2 [55],
and SPIDER [67], offer the option of projection matching.

4.7.3  Angular EM images represent projections of embedded molecules in ran-

Reconstitution dom orientations. If there is no initial model, the angular reconsti-
tution technique is a method of choice to determine the orientation
of images relative to the other images. The common line projection
theorem postulates that every pair of 2D projections of the same 3D
object has at least one mutual 1D line projection called a common
line projection [8, 75]. Thus, by matching 1D line projections for
different images, we can identify the relationship between the 2D
projections and determine the angles between the common lines
and, therefore, determine the relative Euler angles of the images.
398 Tiago R.D. Costa et al.

A set of 1D projections of an image is obtained when the image

is rotated 360° at 1° intervals. The set of 1D projections forms a
sinogram (because a trajectory of projections of one point at suc-
cessive rotations from 1° to 360° corresponds to a sine function).
During angular reconstitution [75] each 1D projection of the first
image is compared with each 1D projection calculated from the
second 2D image (Fig. 13). The line that has the highest correla-
tion (in theory it should be equal to 1) is considered as a common
line of these two projections. Sinograms of different images are
compared pair-wise line by line, checking the correlation to find
common 1D projections (the most similar 1D projections) between
two selected 2D images. At least three images are required to
determine an initial orientation with respect to the object.
Sinograms are generated for all classes, and the search for common
lines is performed for all images. The angles between common 1D
projections are used to assign an orientation to the class average.
Then other classes are added to the initial set of three [75, 76]. It
is also possible to determine orientations using common lines in
Fourier space [77, 78]. The central section theorem states that the
2D Fourier transform of a 2D projection represents a 2D central
section through the 3D Fourier transform of the 3D density. In
Fourier space, a common line corresponds to the cross section of
Fourier transforms of images. This means that two Fourier trans-
forms from two different 2D projections of the same 3D object
have one common central line [77, 78]. Here a comparison of the
radial lines of the Fourier transform of one image with all possible
radial lines of the Fourier transform of the other image is per-
formed. Again, as in real space, the angle between common lines of
the two images with respect to the third one gives the angle
between these two views. A combination of EM and angular recon-
stitution has become an important method for analysing 3D struc-
tures of non-crystallised molecules.

4.8  3D EM images are considered as 2D projections of a 3D object [44].

Reconstruction This is due to the large depth of focus in TEM images. The depth
of defocus is related to the acceleration voltage: the higher the
voltage, the greater the depth of focus, which can be up to 200 nm.
Therefore, the image should represent a projection (the total sum
of electron densities along the beam rays) in the image plane pro-
duced. However, for the images to be considered as real projec-
tions, they must be corrected for CTF effects (see earlier discussion).
Once the CTF correction is done and the Euler angles have been
assigned to each projection image or class average, the 3D electron
density for the particle can be determined.
Several approaches are used to calculate the 3D densities of the
molecules from their projections. Since the current trend is towards
complete automation of the image processing, two approaches are
commonly used in EM owing to the efficiency of their
 Cryo Electron Microscopy 399

Fig. 13 Sinograms and sinocorrelation functions for three projections. A sample object is a model of the T4SS
core outer-membrane complex, which has 14-fold symmetry. Images corresponding to projections are num-
bered 1–3. S1, S2, and S3 (sinograms) are sets of 1D projections of corresponding projections 1–3. CSC12 and
CSC32 correspond to cross-sinogram correlation (CSC) functions between projections 1 and 2 and projections
3 and 2, respectively. Each point of the CSC function represents the correlation coefficient of a pair of lines
from the two sinograms. There are 14 common lines between each pair of images since the object has 14-fold
symmetry. The highest correlations (rainbow circles ) point to the position of common lines. Some of them are
shown by dashed lines between projections 2 and 1, and the solid red line shows one common line between
projections 2 and 3 (the other are not shown). Each CSC function has all peaks doubled because projections
from 180° to 360° mirror those from 0° to 180°. The angular distance between the common lines (red solid
and dashed ) shows the angle α13 between projections 1 and 3
400 Tiago R.D. Costa et al.

Fig. 14 3D reconstruction in real space. Different class averages of the core

outer-membrane complex are shown and situated around the Euler sphere. Each
class average is back-projected along its assigned Euler angle in real space.
Densities of each  2D class average are stretched as rays through 3D space and
the crossover points of the intersecting rays will sum together defining the 3D
electron density of the entire object

implementations. In the first one, the reconstruction is calculated

in real space and based on filtered back-projection algorithms; in
the other one, reconstructions are performed using Fourier space
[79–81].
Three-dimensional reconstructionin real space. These methods cal-
culate the 3D distribution of densities in the space of objects. In
EM, the exact filtered back-projection method [82] is used more
often. In this method, each image of the data set (or classes) is
stretched along a direction defined by the found orientations of
images. Electron densities in three dimensions are obtained by the
summation of rays from stretched projections. The electron den-
sity generated by these summations produces densities for each
voxel. As more projections are included in the 3D reconstruction,
the voxels become better defined (Fig. 14). The angular distribu-
tion of different images should evenly cover the Euler sphere or
the asymmetric triangle (for particles with symmetry). This is
essential to achieve an even representation of all details in the struc-
ture, or at least the set of images chosen for the reconstruction
should have a distribution of angles that covers a great circle of the
Euler sphere [83]. A patchy distribution of angles leads to the
appearance of stripes in the 3D electron density map. To avoid an
additional low-frequency background induced by the projection-­
stretching procedure, images are filtered in advance (although in
some packages the filter is applied on the resulting 3D
 Cryo Electron Microscopy 401

reconstruction). A high-pass filtering of the input 2D projections

corrects the overweighting of the low-frequency components,
thereby restoring amplitude balance and, thus, minimisation of
blurring. The exact filter algorithm used in IMAGIC computes a
specific filter unique to each 2D projection [82, 84].
Three-dimensional reconstructionin Fourier space is based on a the-
orem which states that the Fourier transform of a 2D projection of
a 3D object constitutes a central section of the 3D Fourier trans-
form of the object [85]. This means that the Fourier transform of
projections from different angular views can be merged to fill up
the Fourier space with different 2D sections calculated from the
images (or classes). Recovery of the 3D structure of an object in
real space is done by reverse transformation of its 3D Fourier trans-
form (Fig. 15) [85–87].
The size of trustworthy resolved details can be assessed using a
formula derived by A. Cowther [86] assuming that projections are
evenly distributed:

R = D /N

where N is the number of views, D the particle diameter, and R the

target resolution. N can be significantly decreased for the same
resolution if the complex has a high order of point group
symmetry.
In Fourier space the large number of central sections used
causes overlapping of the central sections near and at the point of
origin. This leads to overweighting of the low-frequency compo-
nents in the Fourier transform and therefore to an effect similar to
a simple back-projection in real space, such as blurring of the
reconstruction. Thus, currently used algorithms based on Fourier
methods employ down-weighting of low frequencies or a high-­
pass map filter.

4.9  Structure All single-particle EM packages use nearly the same procedure for
Refinement the refinement of structures after the first 3D model is obtained
(Fig. 1). It is done by a realignment of single-particle images with
new references obtained from the new model. It is often combined
with the determination of angles: projection matching with a
smaller angular increment or a local refinement of the angles [88].
Reprojections of the new model can be used as a new anchor set in
angular reconstitution to refine the orientation of the classes. The
anchor set is a set of projections calculated from the first 3D model
which are used to determine orientations for new classes or
realigned images. The angular increment between projections used
as the anchor set is usually chosen so that 100–150 projections
(which is much less compared to projection matching) will be cal-
culated and used for the refinement of angles. New Euler angles
are found and assigned to the new class averages. Sorting classes
402 Tiago R.D. Costa et al.

Fig. 15 Reconstruction using Fourier transform (FT) of classes (images). (a) T4SS
core outer-membrane complex (14-fold symmetry) is observed in different ori-
entations on the supporting film. (b) Projections of these particles that are equiv-
alent to EM images in the direction of the electron beam. (c) FT from projections
shown in b. s : FT of side view; e : FT of end view. The corresponding images are
shown in b. (d) A pair of 2D transforms that share at least one common line in
reciprocal space. Common lines between the side view (s) projection and the end
view (e) projection are indicated by green lines ; purple line: common line
between side views (shown in s1 and s7). The angles between pairs of common
lines determine the relative Euler angle orientations. (e) An inverse Fourier trans-
formation of the combined 2D transforms generates an improved real-space
structure seen as electron density

according to the errors between classes and reprojections helps to

facilitate the refinement procedure. While all implementations
share the same principles of refinement, the details of the a­ lgorithms
and the degree to which the user can control the process vary sig-
nificantly [9].

4.10  Evaluation When the map of a new structure is obtained, the molecular mass
of Quality Structure and its oligomeric state should be estimated. The map should be
inspected at the 1σ threshold of the density level, which typically
corresponds to the molecular weight of the complex in the study.
If the complex consists of several interacting proteins, the map
 Cryo Electron Microscopy 403

should not have disconnected fragments of densities; they should

be continuous at densities well above the background noise. The
concept of resolution is based on an assessment of the minimal
distance between two points in an image at which they can still be
distinguished from one another. That criterion was formulated as
the Rayleigh criterion: when the centre of a peak of one point
image falls exactly on the first zero of the image of the second
point. In electron crystallography, the signal-related Fourier com-
ponents of the image are linked to the reflections on a regular lat-
tice, the reciprocal lattice, and a resolution is defined by frequencies
of the reflections that are above the background noise and there-
fore available for Fourier synthesis [89]. This crystallographic reso-
lution, Rc, and Raleigh’s point-to-point resolution distance, d, are
related by d = 0.61/Rc. How can this be done in an objective way
in single-particle analysis? For several decades researchers have
used a concept of Fourier shell correlation (FSC) (see following
discussion). However, in recent years, owing to the tremendous
achievements of single-particle cryo-EM, several modified
approaches have been proposed and methods of resolution assess-
ment have moved closer to criteria used in X-ray crystallography.

4.10.1  Fourier Shell In single-particle analysis there is no well-defined periodical pat-

Correlation and Gold tern in Fourier spectra. Common practice today is to look for data
Standard Approach consistency by splitting the data set randomly in half and compare
the two resulting averages (3D reconstructions). The resolution of
a 3D map (the size of reliable details) could be assessed by
FSC. Two Fourier transforms are calculated, and the correspond-
ing spherical shells are compared using normalised cross-­correlation
as a function of spatial frequency (R) (radius in the Fourier space).
The value of the cross-correlation is used to assess the frequencies
(or size of details) at which these two maps start to differ. If the
correlation falls below the 0.5 threshold, then the details are con-
sidered to be different. Currently, several criteria are used to deter-
mine the threshold used in FSC, and a threshold of 0.1432 has
become rather popular [90, 91]. It should be noted that FSC reso-
lution depends on how the data were split and which threshold of
FSC is used for the assessment. Scheres and Chen proposed a “gold
standard” method for the evaluation of structural quality [92].
According to this approach, the initial data set is divided into two
halves from the very beginning, and two models are refined inde-
pendently. As soon as the structures are obtained, the FSC curves
can be determined as usual. The FSC between two independent
reconstructions shows that when the gold-standard procedure is
used, the resolution of the final results depends on how well the
structures are aligned and what sort of mask was used to remove
surrounding noise (Fig. 16). This separation of the data set into
two equal subsets and the independent refinement helps to avoid
bias towards the same model used on initial steps for the alignment
and determination of angles.
404 Tiago R.D. Costa et al.

Fig. 16 Examples of Fourier shell correlation (FSC) and resolution assessments. (a) FSC curvature of two inde-
pendent 3D structures that have been well aligned and masked with a loose soft mask shown as a wide halo
around the particle images. A resolution assessed at the 0.5 level is 15.6 Å. (b) Here the second structure is
 Cryo Electron Microscopy 405

The disadvantage of the FSC is that for the resolution assess-

ment one needs to split the data into two halves, but this reduces
the resolution of the final 3D model since the entire image data set
was not used for the same 3D reconstruction. The other approach
to the evaluation of detail reliability reconstructed in 3D is the
randomisation of phases (or one can use both randomisation of
phases and amplitudes) above the frequencies critical for detail
assessment. The main concept behind the approach is to modify an
original data set of particle images in such a way that the ampli-
tudes and phases beyond a certain chosen frequency are substi-
tuted by random values. This randomisation of phases (and
sometimes amplitudes) at high frequencies is equivalent to replac-
ing high-frequency structural details by noise. The modified data
set is subsequently subjected to the same image processing proce-
dure used for the original experimental data. The FSC between
these two structures usually demonstrates a sharp drop at the same
frequency where substitution of phases and amplitudes was carried
out [93]. Any non-zero FSC values beyond the resolution where
the noise was introduced reflects a level of bias during image pro-
cessing. It is important to note that a data set with high-frequency
noise contains little information about the real structure compared
to the original data set, so the particle orientations may be less
accurately defined and may affect the value of FSC at frequencies
close to the threshold selected for the phase substitution. The
behaviour of the FSC curves at high frequencies may also be
affected by 3D masking. The FSC can be improved if the feature-
less regions are masked out. However, a tight mask with a very
sharp boundary can produce strange artefacts in FSC such as rising
up to the Nyquist frequencies, indicating that there are non-­reliable
details in the structure (Fig. 16).

4.10.2  Spectral The concept of resolution estimation using an assessment of the

Signal-to-Noise Ratio SNR in spectra of the reconstruction (SSNR) where an entire
data set was used has been suggested by Unser and collaborators
[94–97] and is similar to the Q-factor [98, 99]. The basis of the
method is to measure the consistency between the input data for
the r­ econstructed 3D map and a corresponding set of reprojec-
tions computed. The method estimates the relative energy

Fig. 16  (continued) rotated by 10° around the rotational axis. While the overall shape still coincides well
between both structures small details are not in the register.  It is reflected in the FSC curve which declines
much faster and the resolution at the 0.5 threshold indicates only 27 Å, corresponding to a size of major
domains in the structure. (c) The structures are not aligned. FSC falls down even earlier at lower frequencies
indicating only consistency in overall sizes. (d) FSC between two 3D structures when the same tight mask was
applied. The increase at high frequencies indicates the correlation between masks imposed on the structures,
and here the resolution is overestimated
406 Tiago R.D. Costa et al.

contribution of the reconstructed signal and noise components

by calculating two independent reconstructions [96]. SSNR
characterises the quality of the reconstruction as a function of the
radial frequency. The bottom line is that one will only trust those
signal frequency components whose energy is above what would
have been obtained had the algorithm been applied to noise only.

4.10.3  Local Estimation The third approach whose popularity has risen in recent years anal-
of Resolution yses the 3D density maps obtained from a whole data set by com-
puting the cross-correlation between neighbouring voxels in the
Fourier domain, the Fourier Neighbor Correlation (FNC). A 3D
mask is applied on a 3D structure in such a manner that values
outside the mask are altered for zeroes and pixels remain unchanged
only in part of the cube. Any density inside the structure mask
counts as signal plus noise, whereas any density outside this mask
counts as noise. This operation can be represented in Fourier space
as a convolution of the Fourier transform of the mask with the
Fourier transform of the noise. The convolution provides an assess-
ment of correlations between the Fourier terms. The method has
been implemented in a computer program called RMEASURE
[100]. It is used for 3D reconstructions only.
The ResMap algorithm [101] is based on initialising a
local sinusoid model at r = 2d, where d is the voxel spacing in ang-
stroms (Å). Likelihood ratio tests are conducted at all voxels in the
volume. At a fixed wavelength equal to d, the standard likelihood
ratio test can detect whether a local sinusoid is a meaningful part of
the model approximation. The test requires an estimate of the
noise variance, which can be evaluated from the region surround-
ing the structure. The smallest r at which the likelihood ratio test
passes at a given p-value defines the resolution. The p-value is the
measure of whether the outcome of the attempt is due to an actual
effect or a mere random chance. Voxels that pass the test are
assigned a resolution r, while those that fail the test will be exam-
ined at a larger r. The algorithm produces a local resolution map
with a number assigned to every voxel in the density map.

5  Interpretation and Fitting of Atomic Models

The final verification of the quality of the obtained EM map is

done by an analysis of fitting or “docking” of known atomic struc-
tures, homologous atomic models, or results obtained by de novo
tracing of the polypeptide chain into subunits or protein compo-
nent domains (Fig. 17). In recent decades the resolution (the
smallest reliable details in a structure) of most EM maps produced
by single-particle EM analysis was between 20 and 30 Å. At such a
low resolution large domains can be recognised according to their
overall shape. But this level of detail does not provide sufficient
information on the interaction between proteins and possible
 Cryo Electron Microscopy 407

Fig. 17 Fitting of atomic model (PDB: 3JQO) to cryo-EM map (EMD-2232) of core
outer-membrane complex. (a) Side view. (b) Top view. (c) Central cross section

active sites. The use of antibodies and different methods of label-

ling specific domains allow to localise domain positions and con-
struct reasonable pseudo-atomic models based on EM maps;
nonetheless, these results require substantial additional biochemi-
cal research to verify the interpretation. In subnanometer-­resolution
maps (6–9 Å range), densities corresponding to α-helices reveal
characteristic cylindrical densities with a twist that makes more
accurate fittings and determination of structural handiness. At the
4.5 Å level one can see a separation of strands in the β-layers. A
resolution of around 4 Å reveals densities corresponding to large
amino acid side chains [50, 102, 103], and at resolutions of around
3.7 Å one can do de novo tracing of a polypeptide chain using
methods developed in X-ray crystallography [104].
An efficient approach is when pseudo-atomic models obtained
by homology modelling are fitted into cryo-EM density maps to
build atomic models of individual proteins. The basic principle
behind the fitting procedure is the assessment of the correlation
between a density map and the model. The maximum cross-­
correlation value between the EM map and modelled densities of
the atomic structure indicates the best fit of the model. Such fitting
is performed for maps at a resolution greater than 3.7 Å. Depending
on the software used, a search can be carried out in reciprocal or real
space. The initial stage of the fitting procedure is carried out either
manually or automatically as a so-called rigid fit if a homologous
atomic model exists. If there is no such a model, it can be built up
with online homology servers like Phyre2 [105] or I-Tasser [106].
These models can be initially fitted into the EM density map using
Chimera [107]; and then locations of flexible domains can be iden-
tified in Coot [108] or in a more automated way with FlexEM
[109] or IMODFIT [110]. The flexible fitted structure can be opti-
mised and checked for clashes using PHENIX [111]. This last step
makes it possible to fix the positions of secondary elements simulta-
neously while conforming to geometrical restraints.
Recently, a new method, electron microscopy–iterative modu-
lar optimisation (EM–IMO) was published [112]. It endeavours to
build, modify, and refine the local structures of protein models
408 Tiago R.D. Costa et al.

using cryo-EM maps as constraints. A multi-parameter refinement

strategy which combines EM–IMO and molecular dynamics with
fine-tuning parameters allows one to build backbone models for
the different conformations of proteins at near-atomic resolution
in cryo-EM maps. The use of EM–IMO demonstrates that homol-
ogy modelling and a multi-parametric refinement protocol offer a
practical strategy for building atomic models based on medium- to
high-resolution cryo-EM density maps [113]. Recent develop-
ments in single-particle cryo-EM now allow structures to be
resolved at resolutions close to 3 Å. To facilitate the interpretation
of EM reconstructions, X-ray packages such as REFMAC and
PHENIX were modified for optimal fitting of atomic models to
EM maps because external structural information can enhance the
reliability of the derived atomic models, stabilise refinement, and
reduce overfitting [104].
An atomic model obtained as a result of flexible fitting should
be evaluated for its correctness and consistency with requirements
of interactions between atoms. A Ramachandran plot [114] is rou-
tinely used to visualise the distribution of dihedral torsion angles.
These angles of a polypeptide backbone are the most dominant
local structural factors which dictate protein folding. Geometrical
constraints and steric clashes between atoms of the main chain and
side chains of each residue generate sometimes incorrect orienta-
tions of neighbouring amino acids. The angles are grouped into
favoured and disallowed regions within the plot, which indicates
the overall structural quality. Also, for each type of secondary
structure, i.e. α-helix or β-sheet, there will be a characteristic range
of allowed torsion angles, and these are shown on the plot.

6  Conclusions

Although impressive improvements have been made recently in the

field of EM, many challenges remain in structural studies like large
multi-protein complexes with low or no symmetry. Large com-
plexes are typically flexible or can be unstable, so better approaches
to dealing with sample heterogeneity are needed, which means
more computer power will be required. Flexible, multi-domain
protein structures could be uncovered by an iterative approach
where their spatial organisation is determined by solving structures
of domains step by step. Here, the largest domain could be tackled
first, then “subtracted” computationally from the experimental
image to get at the next largest domain; this process could then be
repeated sequentially until the complete structure and overall
architecture have been determined. Cryo-EM and methods of
image analysis have become important and powerful tools in the
analysis of biocomplexes. Recent advances in EM have greatly con-
tributed to unravelling the so far elusive structural and mechanistic
 Cryo Electron Microscopy 409

details of Gram-negative bacterial secretion systems. Such a level of

detailed information has not only led to a better understanding of
the mechanistic traits underlying the process of substrate secretion
through the T4SS apparatus but also given us the unique opportu-
nity to visualise how this bacterial nano-machine is structurally
organised within both outer and inner membranes. This valuable
structural and mechanistic knowledge can now be used to design
and develop new antibacterial compounds which target critical
bacterial processes and could help us to curb the spread of patho-
genicity and antibiotic resistance.

Acknowledgments

The authors thank Dr. H. White for reading the manuscript and
useful discussions that led to improvements in the manuscript.
This work was funded by MRC Grant MR/K012401/1 to
E.V.O. The authors apologise for not covering all methods fully
owing to space constraints.

References
1. Steitz TA (2008) A structural understanding 9. Cheng Y, Grigorieff N, Penczek PA, Walz T
of the dynamic ribosome machine. Nat Rev (2015) A primer to single-particle ­ cryo-­
Mol Cell Biol 9(3):242–253 electron microscopy. Cell 161(3):438–449
2. Bartesaghi A, Merk A, Banerjee S, Matthies 10. Costa TR, Felisberto-Rodrigues C, Meir A,
D, Wu X, Milne JL, Subramaniam S (2015) Prevost MS, Redzej A, Trokter M, Waksman
2.2 A resolution cryo-EM structure of beta-­ G (2015) Secretion systems in Gram-negative
galactosidase in complex with a cell-permeant bacteria: structural and mechanistic insights.
inhibitor. Science 348(6239):1147–1151 Nat Rev Microbiol 13(6):343–359
3. Bai XC, Yan C, Yang G, Lu P, Ma D, Sun L, 11. Ilangovan A, Connery S, Waksman G (2015)
Zhou R, Scheres SH, Shi Y (2015) An atomic Structural biology of the Gram-negative bac-
structure of human gamma-secretase. Nature terial conjugation systems. Trends Microbiol
525(7568):212–217 23(5):301–310
4. Vinothkumar KR, Zhu J, Hirst J (2014) 12. Fronzes R, Schafer E, Wang L, Saibil HR,
Architecture of mammalian respiratory com- Orlova EV, Waksman G (2009) Structure of a
plex I. Nature 515(7525):80–84 type IV secretion system core complex.
5. Grigorieff N, Harrison SC (2011) Near-­ Science 323(5911):266–268
atomic resolution reconstructions of icosahe- 13. Rivera-Calzada A, Fronzes R, Savva CG,
dral viruses from electron cryo-microscopy. Chandran V, Lian PW, Laeremans T, Pardon
Curr Opin Struct Biol 21(2):265–273 E, Steyaert J, Remaut H, Waksman G, Orlova
6. Grant T, Grigorieff N (2015) Automatic esti- EV (2013) Structure of a bacterial type IV
mation and correction of anisotropic magnifi- secretion core complex at subnanometre reso-
cation distortion in electron microscopes. lution. EMBO J 32(8):1195–1204
J Struct Biol 192(2):204–208 14. Low HH, Gubellini F, Rivera-Calzada A,
7. Scheres SH (2010) Classification of structural Braun N, Connery S, Dujeancourt A, Lu F,
heterogeneity by maximum-likelihood meth- Redzej A, Fronzes R, Orlova EV, Waksman G
ods. Methods Enzymol 482:295–320 (2014) Structure of a type IV secretion system.
8. Orlova EV, Saibil HR (2010) Methods for Nature 508(7497):550–553
three-dimensional reconstruction of hetero- 15. Spence JCH (2003) High resolution micros-
geneous assemblies. Methods Enzymol copy, 3rd edn. Oxford University Press,
482:321–341 New York
410 Tiago R.D. Costa et al.

16. Dubochet J, Adrian M, Chang JJ, Homo JC, 31. Veesler D, Campbell MG, Cheng A, Fu CY,
Lepault J, McDowall AW, Schultz P (1988) Murez Z, Johnson JE, Potter CS, Carragher
Cryo-electron microscopy of vitrified speci- B (2013) Maximizing the potential of elec-
mens. Q Rev Biophys 21(2):129–228 tron cryomicroscopy data collected using
17. Jaffe JS, Glaeser RM (1987) Difference direct detectors. J Struct Biol
Fourier analysis of “surface features” of bacte- 184(2):193–202
riorhodopsin using glucose-embedded and 32. Cunningham IA (1999) Practical digital
frozen-hydrated purple membrane. imaging and PACS. Advanced Medical
Ultramicroscopy 23(1):17–28 Publishing for American Association of
18. Dubochet J, Lepault J, Freeman R, Berriman Physicists in Medicine, USA
A, Homo JC (1982) Electron microscopy of 33. McMullan G, Chen S, Henderson R, Faruqi
frozen water and aqueous solutions. J Microsc AR (2009) Detective quantum efficiency of
124(3):219–237 electron area detectors in electron micros-
19. Lepault J, Dubochet J (1986) Electron copy. Ultramicroscopy 109(9):1126–1143
microscopy of frozen hydrated specimens: 34. McMullan G, Faruqi AR, Clare D, Henderson
preparation and characteristics. Methods R (2014) Comparison of optimal perfor-
Enzymol 127:719–730 mance at 300keV of three direct electron
20. Cabra V, Samso M (2015) Do’s and don’ts of detectors for use in low dose electron micros-
cryo-electron microscopy: a primer on sample copy. Ultramicroscopy 147:156–163
preparation and high quality data collection 35. Ruskin RS, Yu Z, Grigorieff N (2013)
for macromolecular 3D reconstruction. J Vis Quantitative characterization of electron
Exp 95:52311 detectors for transmission electron micros-
21. Grassucci RA, Taylor DJ, Frank J (2007) copy. J Struct Biol 184(3):385–393
Preparation of macromolecular complexes for 36. Bammes BE, Rochat RH, Jakana J, Chen
cryo-electron microscopy. Nat Protoc DH, Chiu W (2012) Direct electron detec-
2(12):3239–3246 tion yields cryo-EM reconstructions at reso-
22. Adrian M, Dubochet J, Lepault J, McDowall lutions beyond 3/4 Nyquist frequency.
AW (1984) Cryo-electron microscopy of J Struct Biol 177(3):589–601
viruses. Nature 308(5954):32–36 37. Milazzo AC, Moldovan G, Lanman J, Jin L,
23. Baker LA, Rubinstein JL (2010) Radiation Bouwer JC, Klienfelder S, Peltier ST, Ellisman
damage in electron cryomicroscopy. Methods MH, Kirkland AI, Xuong NH (2010)
Enzymol 481:371–388 Characterization of a direct detection device
24. Tivol WF, Briegel A, Jensen GJ (2008) An imaging camera for transmission electron
improved cryogen for plunge freezing. Microsc microscopy. Ultramicroscopy 110(7):744–747
Microanal 14(5):375–379 38. Campbell MG, Cheng A, Brilot AF, Moeller
25. Vos MR, Bomans PH, Frederik PM, A, Lyumkis D, Veesler D, Pan J, Harrison SC,
Sommerdijk NA (2008) The development of Potter CS, Carragher B, Grigorieff N (2012)
a glove-box/Vitrobot combination: air-water Movies of ice-embedded particles enhance
interface events visualized by cryo-­ TEM. resolution in electron cryo-microscopy.
Ultramicroscopy 108(11):1478–1483 Structure 20(11):1823–1828
26. Jensen GJ (2010) Cryo-EM. Part A: sample 39. Li X, Mooney P, Zheng S, Booth CR, Braunfeld
preparation and data collection. Preface. MB, Gubbens S, Agard DA, Cheng Y (2013)
Methods Enzymol 481:xv–xvi Electron counting and beam-­induced motion
27. Boyle WS, Smith GE (1970) Charge coupled correction enable near-­ atomic-­
resolution sin-
semiconductor devices. J Bell Syst Tech gle-particle cryo-EM. Nat Methods
49(4):587–593 10(6):584–590
28. McMullan G, Cattermole DM, Chen S, 40. Abrishami V, Vargas J, Li X, Cheng Y,
Henderson R, Llopart X, Summerfield C, Marabini R, Sorzano CO, Carazo JM (2015)
Tlustos L, Faruqi AR (2007) Electron imag- Alignment of direct detection device micro-
ing with Medipix2 hybrid pixel detector. graphs using a robust optical flow approach.
Ultramicroscopy 107(4–5):401–413 J Struct Biol 189(3):163–176
29. Faruqi AR, Henderson R (2007) Electronic 41. Afanasyev P, Ravelli RB, Matadeen R, De
detectors for electron microscopy. Curr Opin Carlo S, van Duinen G, Alewijnse B, Peters PJ,
Struct Biol 17(5):549–555 Abrahams JP, Portugal RV, Schatz M, van
30. Ramachandra R, Bouwer JC, Mackey MR, Heel M (2015) A posteriori correction of cam-
Bushong E, Peltier ST, Xuong NH, Ellisman era characteristics from large image data sets.
MH (2014) Improving signal to noise in Sci Rep 5:10317
labeled biological specimens using energy-­ 42. Glaeser RM (1971) Limitations to significant
filtered TEM of sections with a drift correc- information in biological electron microscopy
tion strategy and a direct detection device. as a result of radiation damage. J Ultrastruct
Microsc Microanal 20(3):706–714 Res 36(3):466–482
 Cryo Electron Microscopy 411

43. Chiu W, Jeng TW (1982) Electron radiation 59. de la Rosa-Trevin JM, Oton J, Marabini R,
sensitivity of protein crystals. Ultramicroscopy Zaldivar A, Vargas J, Carazo JM, Sorzano CO
10(1–2):63–69 (2013) Xmipp 3.0: an improved software
44. Frank J (2006) Three dimensional electron suite for image processing in electron micros-
microscopy of macromolecular assemblies: copy. J Struct Biol 184(2):321–328
visualization of biological molecules in their 60. Smith JM (1999) Ximdisp–a visualization
native state, 2nd edn. Oxford University tool to aid structure determination from
Press, New York electron microscope images. J Struct Biol
45. Egerton RF, Li P, Malac M (2004) Radiation 125(2–3):223–228
damage in the TEM and SEM. Micron 61. Scheres SH (2015) Semi-automated selection
35(6):399–409 of cryo-EM particles in RELION-1.3.
46. Burmeister WP (2000) Structural changes in J Struct Biol 189(2):114–122
a cryo-cooled protein crystal owing to radia- 62. Langlois R, Pallesen J, Ash JT, Nam Ho D,
tion damage. Acta Crystallogr D Biol Rubinstein JL, Frank J (2014) Automated par-
Crystallogr 56(Pt 3):328–341 ticle picking for low-contrast macromolecules
47. Taylor KA, Glaeser RM (1976) Electron in cryo-electron microscopy. J Struct Biol
microscopy of frozen hydrated biological 186(1):1–7
specimens. J Ultrastruct Res 55(3):448–456 63. Heymann JB, Belnap DM (2007) Bsoft:
48. Chiu W (1986) Electron microscopy of fro- image processing and molecular modeling for
zen, hydrated biological specimens. Annu electron microscopy. J Struct Biol
Rev Biophys Biophys Chem 15:237–257 157(1):3–18
49. Knapek E, Dubochet J (1980) Beam damage 64. Roseman AM (2004) FindEM--a fast, effi-
to organic material is considerably reduced in cient program for automatic selection of par-
cryo-electron microscopy. J Mol Biol 141(2): ticles from electron micrographs. J Struct Biol
147–161 145(1–2):91–99
50. Bartesaghi A, Matthies D, Banerjee S, Merk 65. Van Heel M, Portugal RV, Schatz M (2009)
A, Subramaniam S (2014) Structure of beta-­ Multivariate statistical analysis in single parti-
galactosidase at 3.2-a resolution obtained by cle (Cryo) electron microscopy. In: Verkley A,
cryo-electron microscopy. Proc Natl Acad Sci Orlova E (eds) An electronic textbook: elec-
U S A 111(32):11709–11714 tron microscopy in life science. 3D-EM
51. Carlson DB, Evans JE (2012) Low-dose Network of Excellence, London
imaging techniques for transmission electron 66. Grigorieff N (2007) FREALIGN: high-­
microscopy. The transmission electron micro- resolution refinement of single particle struc-
scope. InTech, China tures. J Struct Biol 157(1):117–125
52. Erickson HP, Klug A (1971) Measurement and 67. Frank J, Radermacher M, Penczek P, Zhu J,
compensation of defocusing and aberrations by Li Y, Ladjadj M, Leith A (1996) SPIDER and
Fourier processing of electron micrographs. WEB: processing and visualization of images
Philos Trans R Soc B 261(837):105–118 in 3D electron microscopy and related fields.
53. Wade RH (1992) A brief look at imaging and J Struct Biol 116(1):190–199
contrast transfer. Ultramicroscopy 46:145–156 68. Bartholomew DJ, Steele F, Galbraith J,
54. Thon F (1966) Zur Defokussierungsabh ä Moustaki I (2008) Analysis of multivariate
ngigkeit des Phasenkontrastes bei der social science data. Statistics in the social and
­elektronenmikroskopischen Abbildung. behavioral sciences series, 2nd edn. Taylor &
Naturforschg 21a:476–478 Francis, USA
55. Tang G, Peng L, Baldwin PR, Mann DS, 69. Myung IJ (2003) Tutorial on maximum likeli-
Jiang W, Rees I, Ludtke SJ (2007) EMAN2: hood estimation. J Math Psyc 47(1):90–100
an extensible image processing suite for elec- 70. Sigworth FJ (1998) A maximum-likelihood
tron microscopy. J Struct Biol 157(1):38–46 approach to single-particle image refinement.
56. Rohou A, Grigorieff N (2015) CTFFIND4: J Struct Biol 122(3):328–339
fast and accurate defocus estimation from 71. Scheres SH (2012) A Bayesian view on cryo-
electron micrographs. J Struct Biol 192(2): EM structure determination. J Mol Biol
­
216–221 415(2):406–418
57. van Heel M, Harauz G, Orlova EV, Schmidt 72. Macqueen J (1967) Some methods for classi-
R, Schatz M (1996) A new generation of the fication and analysis of multivariate observa-
IMAGIC image processing system. J Struct tions. Proc Fifth Berkeley Symp Math Stat
Biol 116(1):17–24 Prob 1:281–297
58. Wiener N (1964) Extrapolation, interpola- 73. Ward JHJ (1963) Hierarchical grouping to
tion, and smoothing of stationary time series. optimize an objective function. J Am Stat
Wiley, New York Assoc 58(301):236–244
412 Tiago R.D. Costa et al.

74. Guan W, Lockwood A, Inkson BJ, Mobus G 89. Glaeser RM, Downing KH, DeRosier DJ,
(2011) A piezoelectric goniometer inside a Chiu W, Frank J (2007) Electron crystallog-
transmission electron microscope goniome- raphy of biological macromolecules. Oxford
ter. Microsc Microanal 17(5):827–833 University Press, New York
75. Van Heel M (1987) Angular reconstitution: a 90. van Heel M, Schatz M (2005) Fourier shell
posteriori assignment of projection directions correlation threshold criteria. J Struct Biol
for 3D reconstruction. Ultramicroscopy 151(3):250–262
21(2):111–123 91. Rosenthal PB, Henderson R (2003) Optimal
76. van Heel M, Orlova EV, Harauz G, Stark H, determination of particle orientation, abso-
Dube P, Zemlin F, Schatz M (1997) Angular lute hand, and contrast loss in single-particle
reconstitution in three-dimentional electron electron cryomicroscopy. J Mol Biol 333(4):
microscopy: historical and theoretical aspects. 721–745
Scanning Microsc 11:195–210 92. Scheres SH, Chen S (2012) Prevention of over-
77. Crowther RA (1971) Procedures for three-­ fitting in cryo-EM structure determination.
dimensional reconstruction of spherical Nat Methods 9(9):853–854
viruses by Fourier synthesis from electron 93. Chen S, McMullan G, Faruqi AR, Murshudov
micrographs. Philos Trans R Soc Lond Ser B GN, Short JM, Scheres SH, Henderson R
Biol Sci 261(837):221–230 (2013) High-resolution noise substitution to
78. Fuller SD (1987) The T=4 envelope of measure overfitting and validate resolution in
Sindbis virus is organized by interactions with 3D structure determination by single particle
a complementary T=3 capsid. Cell 48(6): electron cryomicroscopy. Ultramicroscopy
923–934 135:24–35
79. Herman GT (1980) Image reconstruction 94. Unser M, Trus BL, Steven AC (1987) A new
from projections: the fundamentals of com- resolution criterion based on spectral signal-­
puterized tomography. Academic, New York to-­noise ratios. Ultramicroscopy 23(1):39–51
80. Penczek PA (2010) Fundamentals of three-­ 95. Unser M, Trus BL, Frank J, Steven AC
dimensional reconstruction from projections, (1989) The spectral signal-to-noise ratio res-
vol 482. Methods in enzymology: Cryo-EM, olution criterion: computational efficiency
part B, 3-D reconstruction. Academic, Elsevier, and statistical precision. Ultramicroscopy
San Diego, CA 30(3):429–433
81. Orlova EV, Saibil HR (2011) Structural anal- 96. Unser M, Sorzano CO, Thevenaz P, Jonic S,
ysis of macromolecular assemblies by electron El-Bez C, De Carlo S, Conway JF, Trus BL
microscopy. Chem Rev 111(12):7710–7748 (2005) Spectral signal-to-noise ratio and res-
82. Harauz G, van Heel M (1986a) Exact filters olution assessment of 3D reconstructions.
for general geometry three-dimensional J Struct Biol 149(3):243–255
reconstruction. Optik 73:146–156 97. Penczek PA (2002) Three-dimensional spec-
83. Orlov SS (1976) Theory of three dimensional tral signal-to-noise ratio for a class of recon-
reconstruction—conditions of a complete set of struction algorithms. J Struct Biol 138(1–2):
projections. Sov Phys Crystallogr 20:312–314 34–46
84. Radermacher M (1988) Three-dimensional 98. Kessel M, Radermacher M, Frank J (1985)
reconstruction of single particles from ran- The structure of the stalk surface layer of a
dom and nonrandom tilt series. J Electron brine pond microorganism: correlation aver-
Microsc Tech 9(4):359–394 aging applied to a double layered lattice struc-
85. De Rosier DJ, Klug A (1968) Reconstruction ture. J Microsc 139(Pt 1):63–74
of three dimensional structures from electron 99. van Heel M, Hollenberg J (1980) The
micrographs. Nature 217(5124):130–134 stretching of distorted images of two-­
86. Crowther RA, DeRosier DJ, Klug A (1970) dimensional crystals. Electron microscopy at
The reconstruction of a three-dimensional molecular dimensions. Springer, Berlin
structure from projections and its application 100. Sousa D, Grigorieff N (2007) Ab initio reso-
to electron microscopy. Proc R Soc A 317 lution measurement for single particle struc-
(1530) tures. J Struct Biol 157(1):201–210
87. DeRosier DJ, Moore PB (1970) 101. Kucukelbir A, Sigworth FJ, Tagare HD

Reconstruction of three-dimensional images (2014) Quantifying the local resolution of
from electron micrographs of structures with cryo-EM density maps. Nat Methods
helical symmetry. J Mol Biol 52(2):355–369 11(1):63–65
88. Penczek PA (2008) Single particle recon- 102. Zhang R, Alushin GM, Brown A, Nogales E
struction. In: Shmueli U (ed) International (2015) Mechanistic origin of microtubule
tables for crystallography. Springer, New York, dynamic instability and its modulation by EB
pp 375–388 proteins. Cell 162(4):849–859
 Cryo Electron Microscopy 413

103. Clare DK, Orlova EV (2010) 4.6A cryo-EM refinement guided by cryo-EM density.
reconstruction of tobacco mosaic virus from Structure 16(2):295–307
images recorded at 300 keV on a 4k x 4k 110. Lopez-Blanco JR, Chacon P (2013) iMOD-
CCD camera. J Struct Biol 171(3): FIT: efficient and robust flexible fitting based
303–308 on vibrational analysis in internal coordinates.
104. Brown A, Long F, Nicholls RA, Toots J,
J Struct Biol 184(2):261–270
Emsley P, Murshudov G (2015) Tools for 111. Adams PD, Afonine PV, Bunkoczi G, Chen
macromolecular model building and refine- VB, Davis IW, Echols N, Headd JJ, Hung
ment into electron cryo-microscopy recon- LW, Kapral GJ, Grosse-Kunstleve RW, McCoy
structions. Acta Crystallogr D Biol Crystallogr AJ, Moriarty NW, Oeffner R, Read RJ,
71(Pt 1):136–153 Richardson DC, Richardson JS, Terwilliger
105. Kelley LA, Mezulis S, Yates CM, Wass MN, TC, Zwart PH (2010) PHENIX: a compre-
Sternberg MJ (2015) The Phyre2 web portal hensive python-based system for macromo-
for protein modeling, prediction and analysis. lecular structure solution. Acta Crystallogr D
Nat Protoc 10(6):845–858 Biol Crystallogr 66(Pt 2):213–221
106. Yang J, Yan R, Roy A, Xu D, Poisson J, Zhang 112. Zhu J, Cheng L, Fang Q, Zhou ZH, Honig B
Y (2015) The I-TASSER suite: protein struc- (2010) Building and refining protein models
ture and function prediction. Nat Methods within cryo-electron microscopy density maps
12(1):7–8 based on homology modeling and multiscale
107. Pettersen EF, Goddard TD, Huang CC,
structure refinement. J Mol Biol
Couch GS, Greenblatt DM, Meng EC, Ferrin 397(3):835–851
TE (2004) UCSF chimera—a visualization 113. Lindert S, Alexander N, Wotzel N, Karakas
system for exploratory research and analysis. M, Stewart PL, Meiler J (2012) EM-fold: de
J Comput Chem 25(13):1605–1612 novo atomic-detail protein structure determi-
108. Emsley P, Cowtan K (2004) Coot: model-­ nation from medium-resolution density maps.
building tools for molecular graphics. Acta Structure 20(3):464–478
Crystallogr D Biol Crystallogr 60(Pt 12 Pt 114.
Ramachandran GN, Ramakrishnan C,
1):2126–2132 Sasisekharan V (1963) Stereochemistry of
109. Topf M, Lasker K, Webb B, Wolfson H, Chiu polypeptide chain configurations. J Mol Biol
W, Sali A (2008) Protein structure fitting and 7:95–99
 Chapter 29

Bacterial Filamentous Appendages Investigated

by Solid-­State NMR Spectroscopy
Birgit Habenstein and Antoine Loquet

Abstract
The assembly of filamentous appendages at the surface of bacteria is essential in many infection mecha-
nisms. The extent of mechanical, dynamical, and functional properties of such appendages is very diverse,
ranging from a structural scaffold of the pathogen–host cell interaction to cell motility, surface adhesion,
or the export of virulence effectors. In particular, the architectures of several bacterial secretion systems
have revealed the presence of filamentous architectures, known as pili, fimbriae, andneedles. At the macro-
scopic level, filamentous bacterial appendages appear as thin extracellular filaments of several nanometers
in diameter and up to several microns in length. The structural characterization of these appendages at
atomic-scale resolution represents an extremely challenging task because of their inherent noncrystallinity
and very poor solubility. Here, we describe protocols based on recent advances in solid-state NMR spectros-
copy to investigate the secondary structure, subunit–subunit protein interactions, symmetry parameters,
and atomic architecture of bacterial filaments.

Key words Solid-state nuclear magnetic resonance, Structure determination, Pilus, Needle, Protein
assembly, Protein complex, Helical symmetry

1  Introduction

Filamentous appendages are present at the surface of numerous

bacteria to execute crucial functions ranging from DNA uptake
and toxin secretion to cell adhesion. Gram-negative bacteria and
their associated secretion systems display various extracellular
appendages, termed pili, fimbriae, and needles, which play essen-
tial roles during infection. For instance, the type III secretion sys-
tem contains an extracellular filament called a needle, made by the
helical assembly of multiple copies of a single protein subunit [1,
2] and acting as a conduit to export virulence factors from the
periplasm to the extracellular space [3]. The chaperone–usher
pathway forms an extracellular pilus rod, the type 1 pilus [4–6],
that mediates surface attachment [7]. Type IV secretion systems
can ­assemble at their surface filamentous assemblies called pili

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_29, © Springer Science+Business Media LLC 2017

415
416 Birgit Habenstein and Antoine Loquet

that promote conjugation, DNA uptake, or the export of effec-

tors [8]. Furthermore, type IV pili are used to adhere to surfaces
or to generate twitching motility [9]. Filamentous appendages are
often involved in cell or surface attachment, and other proteins, for
instance adhesin proteins, are commonly observed to be present
along the surface of filaments [10] or at the tips of filaments as with
type III secretion needles [11, 12] and chaperone–usher pili [13].
These filamentous appendages are formed by the noncovalent
assembly of several dozens to hundreds of copies of protein sub-
units, self-organized in a hierarchical and highly symmetrical
arrangement. The resulting supramolecular complex consists of a
thin and unbranched filament, usually observed with a diameter in
a range of 5–50 nm and up to several microns in length.
Mechanically resistant, they endure the molecular tumbling and
steric collisions that make up the extracellular environment. From
a structural point of view, bacterial filaments present various impor-
tant challenges to structural biologists with respect to solving their
architectures at an atomic-scale resolution. Indeed, their elongated
shape does not exhibit the long-range order required to crystallize,
limiting the use of X-ray crystallography in studying the native
assembled conformation. Moreover, the size of intact filaments
restricts their molecular tumbling in solution, hampering solution
nuclear magnetic resonance (NMR) spectroscopy. Several strate-
gies have been developed to circumvent these limitations, and inte-
grative approaches combining local structural information on the
monomeric subunits (from X-ray diffraction data or solution
NMR) and the molecular envelope determined by electron micros-
copy (EM) or high-order symmetry parameters obtained from dif-
fraction techniques have proven suitable for obtaining
near-atomic-resolution 3D models of filaments, as recently dem-
onstrated for the P pilus [6], the type II secretion pseudopilus
[14], or the type III secretion needle [2]. The fitting of mono-
meric subunits into cryo-EM data or the combination with X-ray
diffraction data nevertheless has several drawbacks: (1) the mono-
meric subunit structure might undergo conformational changes
between the soluble or crystalline state and the assembled state; (2)
some subunit proteins need to be truncated or mutated to obtain
a soluble or crystalline monomer and prevent its aggregation into
filaments, leading to a partial loss of structural information; and (3)
atomic-resolution reconstruction requires high-resolution EM
density maps, which are often not available for filamentous assem-
blies exhibiting structural heterogeneity at the macroscopic scale.
In the past two decades, magic-angle spinning (MAS) solid-­
state NMR spectroscopy (SSNMR) has emerged as a powerful
technique in structural biology to solve the 3D structure of com-
plex biomolecular systems [15–23]. The design of new SSNMR
probes and pulse sequences involves the setup of high field magnets
(>800 MHz up to 1 GHz), and, by adapting solution NMR-­based
strategic isotope-labeling schemes and developing SSNMR-based
 Solid-State NMR 417

hybrid approaches including complementary techniques (e.g., X-ray,

EM, solution NMR, electron paramagnetic resonance spectros-
copy (EPR), computational modeling), SSNMR has established a
cutting-edge approach to investigating the structures of biological
samples. One main attractive feature of SSNMR in structural biol-
ogy lies in its power to perform atomic-resolution characteriza-
tion of insoluble samples (e.g., precipitates, aggregates,
nanoparticles, fibers, filaments, capsids) lacking crystalline order
or homogeneity at the macroscopic level. Nevertheless, the tech-
nique is subject to several prerequisites: samples should have a
descent structural order at the local level, i.e., poor static disorder;
structural studies are limited to medium-size protein subunits
(<400 residues); and 13C- or 15N-labeled proteins need to be pro-
duced and the assemblies reconstituted in vitro. Bacterial filamen-
tous appendages are very often built by the noncovalent assembly
of a major protein subunit through high-order symmetries, in par-
ticular helical symmetries [24, 25], and usually present an impres-
sive structural homogeneity at the local level. They therefore
represent promising targets for high-resolution structure determi-
nation by SSNMR techniques. Several atomic 3D models of bac-
terial appendages have been solved by SSNMR-based approaches

Fig. 1 Overview of structural investigations of type III secretion needle and type 1
pilus by SSNMR-based approaches. (a) Schematic representation of type III injec-
tisome based on cryo-EM maps of Shigella flexneri needle (accession number
EMD-5352 [2]) and Salmonella typhimurium needle complex (accession number
EMD-1875 [70]). (b) SSNMR-based atomic model of S. typhimurium needle
filaments [1], side perspective and top view, PDB accession number 2LPZ. (c) Top
view of chaperone–usher pilus structure obtained by cryo-EM (accession number
EMD-3222 [6]). (d) SSNMR-based atomic resolution model (PDB accession
number 2N7H [5])
418 Birgit Habenstein and Antoine Loquet

(Fig. 1), including the type III secretion needles of S. typhimurium

[1] (Fig. 1a, b) and S. flexneri [26] and the type 1 pilus of E. coli
[5] (Fig. 1c, d). In this chapter, we describe a protocol based on
SSNMR spectroscopy to investigate the structure and assembly
architecture of bacterial filamentous appendages. The protocol
relies on the production and purification of isotopically 13C/15N
labeled proteins, their in vitro assembly into filamentous samples,
and their analysis by SSNMR. The protocol described here can be
adapted to numerous filamentous assemblies.
SSNMR studies are commonly based on 13C and 15N detection
because of the sensitivity and high spectral dispersion of these
nuclei. The monomeric protein subunit is produced in isotopically
labeled media to obtain 13C and 15N enrichment necessary for high
NMR sensitivity and then self-assembled into filaments. For each
filamentous sample, the protein subunits are expressed, purified,
and assembled differently to obtain a maximal quantity of pure
protein filaments, and the assembly conditions are optimized to
avoid structural polymorphism or heterogeneity. For a structural
characterization of bacterial filamentous appendages, it is impor-
tant to produce pure protein and to optimize the assembly condi-
tions to avoid structural polymorphism or heterogeneity. We here
globally describe the steps that are needed to obtain a protein fila-
ment sample for SSNMR analysis, even though the purification
techniques and assembly conditions are different for each protein
assembly. Figure 2 represents a general scheme for in vitro filament
reconstitution for SSNMR analysis.

Fig. 2 In vitro bacterial filament reconstitution for SSNMR analysis. (a) Expression of labeled protein subunits in
minimal medium containing labeled 13C, 15N sources. One colony of the transformed E. coli bacteria must be
selected from an agar plate to inoculate a preculture, which is further used for inoculating the main culture. (b) The
cells are destroyed and the protein subunit is harvested and purified on an adequate column system. (c) The pure
protein subunits are concentrated (or diluted) to a reasonable assembly concentration in the assembly buffer, which
should be meticulously tested and optimized at the optimal pH and salt concentration. The assembly conditions are
usually optimal under slow shaking. The subunits then adopt their native conformation by assembling (represented
here). (d) Filament formation should be visualized by transmission electron microscopy, as illustrated here for a
protein filament (subunit size of 14 kDa)
 Solid-State NMR 419

2  Materials

2.1  In Vitro Protein SSNMR structural investigations require the use of uniformly
Expression, (or selectively) 13C/15N samples. In this protocol, the sample pro-
Purification, and Self-­ duction is based on the in vitro polymerization of protein subunits
Assembly into filaments. Here, the purpose is to give an overall description
of Isotopically of the requirements and their particularities throughout the
13
C-/15N-Labeled SSNMR sample preparation process using Escherichia coli as a
Protein Subunits standard expression system (see Notes 1–3).
1. Expression vector encoding the protein construct with, for
example, a (His)7-tag. Expression vector, typically pET21,
containing the DNA fragment encoding the protein sequence
of interest; a (His)7-tag is often used for protein purification.
2. E.coli expression strain. A commonly used protein expression
strain for NMR studies is E. coli BL21 (DE3).
3. Isopropyl-thio-β-d-galactopyranoside (IPTG), stock solution
(1 M).
4. A culture medium. For producing unlabeled protein standard
Lysogeny Broth LB is used. For isotope-labeled proteins and
expression tests, M9 minimal medium is used, following the
composition (see Table 1).
5. In case an isotopically labeled sample is produced, different
labeled 13C sources will be needed (see Subheading 3.1.2).
To obtain a uniformly 13C/15N ([U-13C/15N])-labeled sam-
ple, the minimal medium is supplemented with uniformly
13
C-labeled glucose (glc) and 15N-labeled NH4Cl to corre-
spondingly labeled 13C and 15N atoms. As part of the struc-
tural studies it is necessary to selectively label the proteins by
supplementing the M9 medium instead of uniformly
13
C-labeled glc with selectively 13C-labeled glc or glycerol (gly)
[27–30]. The selectively 13C-labeled precursors are [1-13C]-glc
and [2-13C]-glc or [1,3-13C]-glyc and [2-13C]-glyc. The two
distinct selective labeling schemes, two for glc and two for glyc
labeling, are complementary in their resulting amino acid
labeling patterns, which arise from the metabolic pathways
[27–29].
6. Incubator/shaker.
7. Purification elements: usually Äkta (GE Healthcare) or Bio-­rad
chromatography system, purification chromatography column(s),
desalting column, centrifuges, centrifugal filter units.
8. pH meter.
9. Purification buffers.
420 Birgit Habenstein and Antoine Loquet

Table 1
M9 Minimal medium composition

Component M9 medium
NaCl 0.5 g/L
KH2PO4 3 g/L
Na2HPO4 6.7 g/L
MgSO4 1 mM
ZnCl2 10 μM
FeCl3 1 μM
CaCl2 100 μM

10. Buffer for protein assembly (see Note 4).

11. Ultracentrifuge.
12. 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS).

2.2  Solid-State NMR 1. Table centrifuge or ultracentrifuge and ultracentrifugal device

Spectroscopy [31, 32] for rotor filling procedure.
2. SSNMR rotors, with diameter ranging from 3.2 to 4 mm.
3. Spectrometer(s): a detailed structural analysis of macromolec-
ular protein assembly such as bacterial filaments should be
carried out with an NMR spectrometer at a magnetic field
≥500 MHz proton frequency (11.75 T) for sensitivity and
spectral resolution purposes. The spectrometer will be
equipped with MAS probes with double (1H/X) and triple
channels (1H/13C/15N) allowing the detection of 1H, 13C, and
15
N nuclei.

2.3  Solid-State NMR 1. NMR data processing software: NMRpipe [33].

Analysis 2. Graphical and analysis programs to visualize multidimensional
SSNMR spectra and perform resonance assignment.
3. CCPNMR [34], SPARKY (see Table 2).
4. Databases of 13C, 15N, and 1H chemical shifts: Biological
Magnetic Resonance Data Bank (BMRB) [35] (see Table 2) or
published data [36].
5. Computational routines to derive dihedral angles from SSNMR
chemical shifts:
TALOS+ [37], PREDITOR [38] (see Table 2).
6. Prediction of chemical shifts if monomer structure is
available:
SPARTA+ [39], SHIFTX2 [40], CamShift [41] (see Table 2).
 Solid-State NMR 421

Table 2
Websites of computational programs and databases

Name Website address

NMRpipe http://spin.niddk.nih.gov/NMRPipe/
CcpNmr analysis http://www.ccpn.ac.uk/software/analysis
BMRB http://bmrb.wisc.edu/
SPARKY https://www.cgl.ucsf.edu/home/sparky/
TALOS+ http://spin.niddk.nih.gov/bax/software/TALOS/
PREDITOR http://wishart.biology.ualberta.ca/shiftor/cgi-bin/preditor_current.py/
SPARTA+ http://spin.niddk.nih.gov/bax/software/SPARTA+/
SHIFTX2 http://www.shiftx2.ca/
CamShift http://www-vendruscolo.ch.cam.ac.uk/camshift/camshift.php
XPLOR-NIH http://nmr.cit.nih.gov/xplor-nih/
CNS http://cns-online.org/v1.3/
ARIA http://aria.pasteur.fr/
CS-ROSETTA http://spin.niddk.nih.gov/bax/software/CSROSETTA/
ROSETTA https://www.rosettacommons.org/
HADDOCK http://haddock.science.uu.nl/services/HADDOCK2.2/
CING https://code.google.com/archive/p/cing/
PSVS http://psvs-1_4-dev.nesg.org/
PROCHECK-NMR http://www.ebi.ac.uk/thornton-srv/software/PROCHECK/
Pymol https://www.pymol.org/
Swiss PDB Viewer http://spdbv.vital-it.ch/

2.4  Structure 1. Several computational programs can be employed in the

Modeling modeling process depending on the extent of available structural
information, in particular if SSNMR restraints are combined
with structural data obtained from other biophysical
techniques:
XPLOR-NIH [42], CNS [43], ARIA [44], CS-ROSETTA
[45], HADDOCK [46] (see Table 2).
2. Subsequent to structure modeling, structure validation programs
are used:
CING, PSVS [47], PROCHECK-NMR [48] (see Table 2).
3. Protein visualization software:
Pymol, Swiss PDB Viewer (see Table 2).
422 Birgit Habenstein and Antoine Loquet

3  Methods

3.1  Isotope-Labeled 1. Transform E.coli strain (e.g., BL21 DE3) with vector, includ-
In Vitro Protein ing protein-coding DNA, and plate it on LB agar plates con-
Expression, taining appropriate antibiotic.
Purification, 2. Inoculate a preculture of ≤10 mL LB medium with one iso-
and Self-Assembly lated culture picked from the agar plate (see Note 5). Incubate
3.1.1  Expression
at 37 °C while shaking (rotation values around 200 rpm) until
of Unlabeled Subunit
exponential growth phase is reached (0.6 < OD600 < 1).
Proteins 3. Inoculate main culture (LB) with preculture. To optimize pro-
tein expression, purification, and assembly conditions, as well as
to obtain a preliminary NMR signature (see Subheading 3.3), it
is appropriate to use unlabeled LB medium. With this low-cost
filament production, the conditions can be optimized and the
produced quantity estimated. Based on a 1D 13C spectrum
recorded at natural abundance (unlabeled material), a prelimi-
nary analysis of the structural complexity in the filament can be
obtained.
4. Induce protein expression with 0.75–1 mM IPTG at an opti-
cal density measured at a wavelength of 600 nm (OD600) of
0.8–1. Test the optimal expression time by sodium dodecyl
sulfate (SDS) gels, usually between 3 and 6 h at 37 °C, but
sometimes protein expression can be optimal at low tempera-
tures overnight. Harvest cells by centrifugation on a table cen-
trifuge (6000 rpm, 30 min, 4 °C) and store at −80 °C until
purification.

3.1.2  Expression 1. Transform E. coli strain (e.g., BL21 DE3) with vector, includ-
of Isotopically Labeled ing protein-coding DNA, and plate it on LB agar plates con-
Subunit Proteins taining appropriate antibiotic.
2. A preculture of ≤10 mL LB medium should be inoculated
with one isolated culture picked from the agar plate (see Note 5),
incubated at 37 °C while shaking (rotation values around
200 rpm) until the exponential growth phase is reached (0.6 <
OD600 < 1). Centrifuge down the preculture and remove
supernatant.
3. Inoculate main culture—M9 medium with unlabeled (step 4)
or labeled carbon and nitrogen sources (step 5)—with the
harvested cells (preculture of ≤10 mL LB per liter main
culture).
4. First perform a protein production in M9 medium supple-
mented with unlabeled glucose and NH4Cl to quantify the
production yield in M9 medium (see Note 6).
5. For the subsequent production of uniformly 13C/15N-labeled
([U-13C/15N]-labeled) protein in M9 medium (see Fig. 3a),
 Solid-State NMR 423

supplement the M9 medium with uniformly labeled 13C glucose

and 15NH4Cl. As mentioned in Subheadings 2.1 and 3.1.4,
selective labeling schemes can be achieved via different
13
C-labeled precursors: selectively [1-13C]-glc and [2-13C]-glc
or [1,3-13C]-glyc and [2-13C]-glyc (Fig. 3b–d). For the 15N
labeling, 15NH4Cl is always supplemented. The two distinct
selective labeling schemes for glc or glyc are complementary in
their resulting amino acid labeling pattern. After evaluation of
SSNMR data on uniformly labeled protein filament samples, a
selective labeling scheme can be chosen (see Note 7) [49].
6. Induce protein expression with 0.75–1 mM IPTG at OD600
0.5–0.8. Express protein for period determined in Subheading
3.1.1. Harvest cells by centrifugation on table centrifuge (6000
rpm, 30 min, 4 °C) and store at −80 °C until purification.

3.1.3  Purification The purification protocol should be set up using unlabeled pro-
and Polymerization teins to avoid unnecessary expenses.
of Subunit Proteins
1. Thaw cells on ice and resuspend in appropriate lysis buffer.
2. Lyse E. coli cells by appropriate lysis method (e.g., sonication,
mechanical disruption or chemical lysis).
3. Set up a purification protocol. If the protein is produced in
inclusion bodies, the purification generally needs to be carried
out under denaturing conditions (e.g., 6–8 M urea or guani-
dine). In this case, a refolding step needs to be scheduled,
which might be concurrent with the desalting step. If the pro-
tein stays folded in the cytoplasm, the desalting step might not
be required and a simple concentration after purification might
be sufficient.
4. Concentrate the soluble protein fraction to a final concentra-
tion favorable to correct assembly. Typical values for protein
concentration range between 0.1 and 1 mM.
5. Keep protein solution under smooth shaking conditions for at
least 1 week. The solution becomes turbid upon filament for-
mation. After 1 week, filament formation should be checked
by EM. If filaments have formed, it is recommended to keep
assembled fibrils at 4 °C and add an antibacterial agent such as
sodium azide (0.02% (w/v)) to avoid sample contamination.

3.1.4  Homogeneous 1. To distinguish between intra- and intermolecular restraints

Versus Asymmetric during the assignment process (see Subheading 3.5.3), several
Labeling Strategies strategies have been developed based on asymmetric labeling
for Intermolecular Restraint schemes. These approaches differ from homogeneous labeling
Detection strategies where all protein subunits are produced with the
same isotope labeled precursor.
2. (1:1) mixed labeled filaments correspond to the equimolar
mixture of two protein batches with different isotope-labeling
424 Birgit Habenstein and Antoine Loquet

Fig. 3 Strategies for isotopic labeling of protein subunits. Homogeneous labeling schemes, based on (a) uni-
formly 13C-labeled subunits (white ), (b) selectively [1-13C]-glucose-labeled subunits (green), and (c) selectively
[2-13C]-glucose-labeled subunits (pink ). [1-13C]- and [2-13C]-glucose labeling schemes are homogeneous in
the sense that all protein subunits are labeled with the same scheme, although the isotopically labeled carbon
positions are selective. (d) Labeling patterns for the amino acid valine for each of the labeling schemes.
Isotopically labeled atomic positions are underlined and colored according to the cases a–c

schemes prior to the assembly process. Two types of modus

operandi are conceivable, based on the detection of 13C-13C or
15
N-13C intermolecular subunit–subunit interactions. The dis-
crimination between intra- and intermolecular 13C-13C prox-
imities is based on the use of a mixture of (1:1) [1-13C]-glc/
[2-13C]-glc [30] or (1:1) [1,3-13C]-glyc/[2-13C]-glyc-labeled
filaments. For both labeling precursors, two complementary
schemes in terms of the resulting labeling pattern are available
([1-13C]-glc [27] and [2-13C]-glc [29] or [1,3-13C]-glyc and
[2-13C]-glyc [28, 50]).
3. To detect intermolecular 15N-13C proximities, a mixed labeled
filament sample containing (1:1) [U-15N]- and [U-13C]-
labeled subunits should be prepared [51]. The mixed labeled
samples are prepared by mixing purified differently labeled
subunits before filament assembly by gently vortexing.
4. A possibility arises aimed at highlighting intramolecular long-­
range contacts. To do so, a mixed labeled filament sample con-
taining 1:X (X = 3–5) [U-13C/15N]-labeled and unlabeled
subunits should be prepared, also termed a diluted sample.
The use of the mixed labeled and diluted samples for identify-
ing long-range distances is described in Subheading 3.5.3.

3.2  Solid-State 1. Centrifuge aggregated filament sample and remove and keep the
NMR Setup supernatant at 4 °C until SSNMR experiments are successful.
Wash sample one to five times with ultrapure water if a high con-
3.2.1  Rotor Packing
centration of salts was used in the assembly buffer (see Note 8).
Centrifugation is needed between the washing steps, and
 Solid-State NMR 425

resuspension of the filaments should be done by pipetting

without vortexing to avoid filament damage.
2. Centrifuge filament sample using an ultracentrifuge. The dura-
tion of this step depends on the consistency of the sample, with
samples having a high water content requiring longer centrifu-
gation (up to 24 h). Remove supernatant using a pipette.
3. Introduce pellet inside SSNMR rotor. Different sizes of MAS
rotors are illustrated in Fig. 4a. Depending on sample consis-
tency, rotor filling can be done using different devices. In case
of a solid sample (crystals, powder, lyophilized sample), a
regular spatula is used. For gel-like samples, as is the case for
bacterial filamentous samples, a pipet or a capillary pipet is
suitable.
4. Add DSS for temperature and chemical shift calibration dur-
ing SSNMR experiments (see Subheadings 3.2.3).
5. Close rotor with cap. If rotor is not fully filled, a top (or a top
and a bottom) spacer should be used to equilibrate the con-
tent and ensure stable spinning.

3.2.2  Magic-Angle Several preparatory and optimization steps are required to carry out
and Pulse Calibration MAS SSNMR spectroscopy on biological samples. SSNMR experi-
ments are designed using a so-called NMR pulse sequence, which
contains a single or a series of radiofrequency pulses and one or
more acquisition times to record the NMR signal. The procedures
and parameters described in what follows are typical in the opera-
tion of a Bruker spectrometer in MAS configuration. A probehead
for 3.2 mm MAS rotors of an 800 MHz spectrometer is shown
in Fig. 4b.
1. Adjust angle of rotor relative to magnetic field (called the magic
angle) at a value of ~54.7°. A sample of KBr in powder form is
used to record a 79Br detected 1D spectrum under a spinning
frequency of 5 kHz; a single scan is enough to observe the sig-
nal on most spectrometers. The spinning frequency gives rise to
so-called spinning sidebands at the distance of the spinning
­frequency (in this case 5 kHz) from the 79Br signal. Adjust the
angle by optimizing the linewidth of the spinning sidebands,
i.e., the sideband signal to the maximum intensity. The benefi-
cial effect of MAS is illustrated in Fig. 4c.
2. Calibrate the hard NMR pulses on standard reference samples,
usually Adamantane for calibrating the 1H pulse and a 13C/15N-­
labeled amino acid (e.g., 13C/15N-labeled histidine hydrochlo-
ride monohydrate in polycrystalline powder form) for 13C and
15
N pulses. 90° pulses are optimized by adjusting the pulse
length or the pulse power. For example, to obtain the 90°
pulse values, the 180° pulse can be found by optimizing the
signal decrease until it disappears.
426 Birgit Habenstein and Antoine Loquet

Fig. 4 Solid-state experimental setup. (a) SSNMR MAS rotors with 7, 4, 3.2, and
2.5 mm diameter. (b) Bruker triple resonance 1H/13C/15N MAS 3.2 mm probehead.
(c) Carbon-detected 1H-13C CP spectrum of a protein filament recorded under MAS
condition with 1H decoupling during acquisition (SPINAL64) [54] (black ), under
MAS condition without 1H decoupling (red ), and without MAS (blue )

3. The optimization of the different polarization transfer steps

involving through-space transfer, such as 1H-X Hartmann–
Hahn cross polarization (CP) [52] and 15N-13C-specific CP
(DCP) [53], 15N to 13Cα and 15N to 13CO, are realized on a
13
C/15N-labeled histidine (or a different amino acid) sample.
Proton decoupling during evolution and acquisition times is
performed with SPINAL64 [54] and optimized setting the
pulse frequency to 90 kHz (applied on 1H). The beneficial
effect on the spectral resolution and sensitivity of proton
decoupling is illustrated in Fig. 4c.

3.2.3  Sample Once the NMR setup on reference samples has been done, the
Temperature and Chemical sample of interest is introduced into the spectrometer to adjust the
Shift Calibration sample temperature and calibrate the chemical shift.
1. To obtain a precise estimation of the sample temperature dur-
ing an SSNMR experiment, it is recommended to use an inter-
nal calibration to avoid inaccurate measurements of the probe
thermocouple. On hydrated samples such as the filament sam-
ples described in this chapter, the sample temperature can be
calibrated using the difference between the DSS 1H resonance
(observed at 0 ppm) and the supernatant water resonance
 Solid-State NMR 427

[31]. Following the relationship δ(H2O) = 7.83 − T/96.9

ppm (with the temperature T in Kelvin), the temperature can
be measured with a precision of ±1–2 °C. The experimental
sample temperature is commonly set to 1–10 °C for all SSNMR
experiments.
2. Because DSS is an inert compound in most biological samples
(except at very low pH), its 1H frequency (representative of a
chemical shift value of 0 ppm) can be used directly to calibrate
the 1H frequency and indirectly to calibrate the 13C and 15N
chemical shifts following International Union of Pure and
Applied Chemistry (IUPAC) recommendations [55].

3.3  Preliminary A first global structural characterization of a sample of interest

Characterization of should be carried out on an unlabeled filament sample to avoid
an Unlabeled Sample unnecessary expenses arising from isotope-labeled precursors.
Several standard SSNMR experiments can be carried out on an
unlabeled sample. The aim is to obtain rapidly (1–2 days of NMR
spectrometer time) a rough evaluation of the structural order and
to determine the presence of mobile segments within the protein
assembly. Only for a more detailed analysis of mobile segments
(described in Subheading 3.3.3) is a [U-13C]- or [U-13C/15N]-
labeled sample necessary.

3.3.1  CP-Based
1.
Introduce the unlabeled filament sample into the
One-Dimensional spectrometer.
Experiment 2. Set the spinning frequency to 11 kHz and simultaneously cool
down the probehead to maintain a sample temperature in the
probehead of around 5–10 °C (see Subheading 3.2.3).
3. Setup a 1D 1H-13C CP experiment following the hard pulse
optimization on a reference sample (see Subheading 3.2.2). For
most natural abundance filament samples, the 13C signal is vis-
ible only after several hours of acquisition, making the optimi-
zation of pulse lengths, decoupling, and CP transfers extremely
time consuming and practically almost impossible. For this
experiment, the values should therefore be taken from the
values obtained in the reference sample optimization. Set the
number of scans to 20k (20,480), the CP contact time to 1 ms,
and the decoupling strength to 90 kHz. Different causes might
be at play if the spectrum shows no signal (see Note 9).
3.3.2  Interpretation The carbon-detected 1D 1H-13C CP spectrum is very informative
of a CP-Based One-­ since the linewidths are directly correlated to the degree of molec-
Dimensional Experiment: ular order and the resonance peak distribution indicates the sec-
Structural Elements ondary structure content. Figure 5 illustrates the information
and Homogeneity content of a 1D 1H-13C CP versus a 1D INEPT-based experiment
recorded on a bacterial filament.
428 Birgit Habenstein and Antoine Loquet

Fig. 5 SSNMR characterization of rigid and mobile protein segments in a protein filamentous assembly.
(a) Schematic representation of a bacterial filament, including rigid and mobile protein segments. (b) Cross
polarization (CP)-based experiments reveal residues contributing to the rigid core of the assembly. INEPT-
based experiments probe the presence of mobile segments, illustrated on a protein filament with a subunit
size of 14 kDa. (c) The chemical shift value of Cα, Cβ, and carbonyl C′ atoms is sensitive to the secondary
structure, and the chemical shift distribution is indicative of the secondary structure composition; illustrated
on a 1D CP spectrum of two protein filaments (both subunit sizes of 9 kDa) containing β-strand (blue) and
α-helical (red ) conformation
 Solid-State NMR 429

1. 13C linewidths experimentally observed in biological samples

under MAS and high field are typically between 0.2 and 5 ppm,
depending on the structural homogeneity of the protein subunit
conformation. Figure 5a shows a typical 13C-detected 1D CP
recorded for a filamentous protein assembly with linewidths
<70–80 Hz. Protein assemblies exhibiting a high structural het-
erogeneity (i.e., the protein subunits in the assembly adopt sev-
eral slightly different atomic structures or are not in the same
local environment) would lead to a 13C linewidth >~2 ppm.
2. If the examined sample exhibits linewidths >2 ppm in a spec-
tral region where nonoverlapped peaks can be examined, read-
justing/optimizing the purification or assembly conditions to
obtain a more ordered protein structure in the filaments could
be helpful. Nonetheless, the observation of large linewidths
for some distinguishable peaks might reveal a less homoge-
neously ordered protein segment in the assembly and would
not necessarily imply that all protein segments contributing to
the assembly have adopted a less-well-ordered structure. It is
therefore important to carefully judge all visible signals.
3. In insoluble and noncrystalline protein assemblies, the struc-
tural homogeneity can be intrinsically poor if the protein sub-
unit molecules have slightly different conformations within
the assembly. In this case, an atomic 3D structure determina-
tion is extremely challenging, and the identification of the resi-
dues participating in the rigid core, as well as the determination
of the secondary structure, should become the main objectives
of the SSNMR study.
4. The signal distribution and dispersion are representative of the
amino acid composition in the rigid structural segments and
of the secondary structure elements.
Each amino acid has its 13C SSNMR spectral fingerprint,
which arises from all carbon atoms present in the specific
amino acid. The signal distribution in a 1D CP 1H-13C spec-
trum already provides some indications about residues in a
rigid ­conformation; for example, resonance peaks at a chemi-
cal shift position of 75 ppm only arise from threonine Cβ reso-
nances (see Fig. 5b). Similar observations can also be carried
out for other carbon resonance signals.
5. As presented in Fig. 5c, the signal dispersion gives a first indi-
cation of the secondary structure elements contributing to the
rigid protein structure. Figure 5c compares two protein assem-
blies, in one case built of near-complete α-helical proteins
(red) and in the second case of near-complete β-strand sub-
units (blue). The position of C′, Cα, and Cβ carbon signals
depends on the secondary structure elements and are therefore
shifted toward specific spectral regions.
430 Birgit Habenstein and Antoine Loquet

3.3.3  INEPT-Based Although a CP-based experiment can reveal rigid residues, SSNMR
One-Dimensional also has the ability to probe residues with higher mobility. INEPT-­
Experiment: Mobility based experiments are used to perform a through-bond transfer
and probe protein segments with increased molecular mobility
(from picoseconds to nanoseconds). Such INEPT-based experi-
ments are useful for detecting the presence of mobile protein seg-
ments in comparison with the rigid core probed by CP-based
experiments; both approaches are complementary in the delinea-
tion of different protein segments or domains with different mobil-
ity (see Fig. 5). The following procedure is used to carry out the
INEPT-based approach.
1. Optimize a 1H-13C INEPT transfer using GARP [56] proton
decoupling with a radio frequency of 5 kHz. Set up a 1D spec-
trum with 4000 scans and a recycle delay of 1 s.
2. Analysis: The interpretation of a 1D 13C-detected INEPT
spectrum (e.g., Fig. 5b) depends mostly on the amount of vis-
ible signals. If a low amount of signals is present, the signal
assignment to the amino acid carbons is possible (residue-type
assignment). The shifts in an INEPT experiment are usually
very close to chemical shifts typical for random coil conforma-
tion and can be assigned using the published 13C random coil
chemical shifts (e.g., [36]). Always carefully check the reso-
nances of your buffer components that might remain after
washing. If signals do not correspond to protein signals and
cannot be identified, it is recommended to record a solution
NMR spectrum of the buffer only to clearly assign buffer
signals.
3. Analysis: Based on the 1H/13C/15N chemical shift database
(e.g., the useful paper by Jardetzky et al. [36]), residue spin
systems can mostly be assigned unambiguously. As described
earlier, the residues observed in INEPT-based experiments
often exhibit random coil chemical shifts. Therefore, a com-
parison of the chemical shifts assigned in the INEPT spectrum
to the random coil standard values allows for estimating the
degree of mobility of the residue present in this dynamic
regime. The amino acids observed in the INEPT spectrum
compared to the primary sequence of a protein might make it
possible to identify the mobile protein segment, sometimes
with high confidence if a single copy of one residue type is
present in the primary sequence.
4. For all other described analyses of the mobile elements a
[U-13C]- or [U-13C/15N]-labeled sample is necessary.
5. If multiple residues of the same residue type are present
showing slightly different chemical shifts or difficulties in
disambiguating signals, 2D 1H-13C INEPT and 1H-(13C)-13C
or (1H-)13C-13C INEPT-TOBSY [57, 58] can be recorded.
 Solid-State NMR 431

2D 1H-13C INEPT can connect the 13C signal visible in the 1D

13
C INEPT to its bonded 1H atom (1H random coil chemical
shift values can be used). An additional 13C-13C polarization
transfer with the TOBSY sequence then makes it possible to
connect the 1H to the neighboring carbon positions, establish-
ing the typical 1Hα-13Cα-13Cβ spin system aimed at the resi-
due-type assignment.
6. As reported by Baldus et al. [59], the peak intensity of each resi-
due type Cα-Cβ correlation in a through-bond (e.g., INEPT-­
TOBSY) or through-space (e.g., CP-PDSD) spectrum can be
used to compare the amino acid composition within the mobile
and the rigid protein segments, respectively.

3.4  Structural A major task consists in assigning experimentally observed reso-

Characterization nances to the 13C atoms where they originate. The optimal proce-
of Rigid Core dure is to first familiarize oneself with the amino acid composition,
of Filamentous the signal dispersion, and linewidth by identifying a maximum of
Assembly amino acid spin systems (see Subheadings 3.4.1 and 3.4.2 and as an
example Fig. 6). Once a maximum of spin systems has been identi-
fied, integrating information from selective labeling schemes if
available (Fig. 7), and the necessary spectra have been recorded to
perform the sequential assignment, the subsequent step consists in
assigning the amino acid spin systems to the primary sequence and
to identify missing spin systems using the sequential connections of
the N, C′, Cα, and Cβ resonances (see Subheading 3.4.3 and Fig. 8).
At this stage, create a project in your assignment program of choice

Fig. 6 Fingerprint SSNMR experiments for structural investigation of filamentous appendages. (a) 2D 13C-13C
PDSD spectrum recorded with a short mixing time to observe intraresidue correlations. (b) 2D 15N-13Cα spec-
trum to probe intraresidue backbone correlations. These experiments are illustrated on the S. typhimuriumtype
III secretion needle [1]
432 Birgit Habenstein and Antoine Loquet

Fig. 7 Comparison of theoretical 2D NCA spectra for each amino acid, for three different labeling schemes:
[U-13C]-glc (a), [1-13C]-glc (b), and [2-13C]-glc (c). (d) Spectral comparison of [U-13C] and [2-13C]-glc leads to
the identification of leucine correlations. (e) Spectral comparison of [U-13C] and [1-13C]-glc leads to the identi-
fication of several amino acid correlations

Fig. 8 Strategy for SSNMR resonance assignment. Intraresidue (a–d) and interresidue (e–h) correlations are
established through the connectivity between the carbon–carbon and the nitrogen–carbon atoms. (a) C-C PDSD
experiment for all intraresidue 13C-13C correlations (short mixing time). (b) N-C specific-CP for intraresidual 15N-13Cα
correlations. (c) N-(Cα)-Cβ (15N-13Cα specific-CP, DREAM) for intraresidual 15N-13Cα-13Cβ correlations. (d) N-(Cα)-C′
(15N-13Cα specific-CP, MIRROR) for intraresidual 15N-13Cα-13C’ correlations. (e) C-C PDSD experiment for predomi-
nantly intraresidue and sequential 13C-13C correlations (intermediate mixing time). (f) N-C′ (15N-13C′ specific-CP)
for sequential 15N-13C′ correlations. (g) Cα-(N)-C′ (13Cα-15 N specific-CP, 15N-13C’ specific-CP) for sequential
13
Cα-13C′ correlations. (h) N-(C′)-Cα (15N-13C’ specific-CP, MIRROR) for sequential 15N-13C’-13Cα correlations
 Solid-State NMR 433

(see item 1 in Subheading 2.3) and carefully set up all molecular

parameters of your system in the project.
Several system parameters obtained during the characterization
based on a [U-13C/15N]-labeled sample are important in terms of
choosing an adequate precursor for the selective labeling scheme.
As described in Step 3 in Subheading 3.2.1, glycerol-­based selec-
tive labeling should be chosen if the signal-to-noise ratio in the
spectra is a limiting factor. If the spectral overlap is the most limit-
ing factor, glucose-based selective labeling should be the correct
choice [49].

3.4.1  Identification 1. Set up a 2D 13C-13C proton-driven spin diffusion (PDSD) [60]

of Amino Acid Spin experiment using a mixing time of 50 ms on a [U-13C/15N]-
Systems labeled sample. This experiment is illustrated with a bacterial
secretion needle in Fig. 6a. The relatively short mixing time
allows for the observation of intraresidue correlations (Fig. 8a).
The 2D spectrum is processed using the program NMRpipe
(see item 3 in Subheading 2.1).
2. Compare the resulting 13C-13C spectrum with standard chemi-
cal shift values for amino acids based on the BMRB average
chemical shift statistics. Several amino acid spin systems, such
as isoleucine, threonine, serine, alanine, or proline, show a
typical resonance pattern that can be identified in a straight-
forward manner if the resolution is sufficiently high (narrow
signal linewidths and adequate molecular weight). Spectral
analysis should be carried out using one of the visualization
programs listed in item 2 of Subheading 2.3.
3. Set up a 2D 13C-13C DREAM experiment with a short mixing
time (typically less than 5 ms) [61]. The DREAM experiment
provides intraresidue correlations, such as the PDSD; how-
ever, positive and negative signals are observed owing to the
adiabatic double quantum polarization transfer. This experi-
ment has the advantage of high polarization transfer efficiency
and is therefore extremely useful in sensitivity problems. The
inversion of the signal intensity leads to the observation of
typical Cα-Cβ-Cγ correlations for which the Cγ signal is again
positive (as the diagonal is) and helps in identifying spin
systems.
4. If available, set up a 2D 13C-13C PDSD experiment using a mix-
ing time of 75 ms on a selectively 13C-labeled sample. The rela-
tively short mixing time allows for the observation of intraresidue
correlations, and the labeling scheme decreases the number of
signals, because several carbons are unlabeled in the selectively
13
C-labeled sample and no signal will be observed, and there-
fore disambiguates signal assignments. To assign the signals,
refer to the metabolic labeling pattern of the corresponding
scheme ([1-13C]-glc [27] and [2-13C]-glc [29] or [1,3-13C]-glyc
and [2-13C]-glyc [50]).
434 Birgit Habenstein and Antoine Loquet

5. The complementarity of the selective labeling schemes can be

used to rapidly identify specific amino acid types, as illustrated
in Fig. 7 in an N-Cα spectrum. For instance, leucine is the
only amino acid that is not mainly 13C labeled on its Cα posi-
tion in the [2-13C]glc labeling scheme. Comparison of N-Cα
spectra recorded for [U-13C]glc and [2-13C]glc allows for a
rapid identification of leucine residues (Fig. 7d). A consider-
able number of amino acid ambiguities can be lifted in this
manner as their N-Cα correlation will arise in either or both
[1-13C]glc and [1,3-13C]glyc or in the [2-13C]glc, [2-13C]glyc-
labeled sample. When performed on a 13C-13C correlation
spectrum, this type of spectrum comparison between the dif-
ferent labeling schemes is likewise useful for lifting assignment
ambiguities.

3.4.2  15N-13C 1. Set up a 15N-13C-specific CP with a mixing time ranging from

Intraresidue Linking 2 to 5 ms (to be optimized on a 1D spectrum). This experi-
ment is illustrated with  a bacterial secretion needle in Fig. 6b.
Although the experiment is based on through-space magneti-
zation transfer, short mixing times of 2–5 ms and relatively
low-­power pulses (e.g., 5–20 kHz on 13C and 15N) allow for
the observation of specific intraresidue 15N to 13Cα correla-
tions (Fig. 8b).
2. For amino acids with isolated Cα resonances, the Cα reso-
nance can be unambiguously linked to their nitrogen reso-
nance frequency.
3. The NCC-type experiment (e.g., N-(Cα)-Cβ, N-(Cα)-Cx,
N-(Cα)-C′) adds an additional dimension to identify the
amino acid spin systems (already described on a 13C-13C basis
in Subheading 3.4.1) (see Note 10).
4. For identifying N-Cα-Cβ correlations, the optimal transfer
efficiency is obtained by adding a DREAM 13Cα-13Cβ transfer
after the 15N-13C-specific CP, the resulting spectrum contain-
ing mostly positive 13Cα and negative 13Cβ signals [62]. The
13
C resonances can now be linked to their corresponding
intraresidual 15N signal, which will further be used for the
sequential assignment (Fig. 8c).
5. To identify the carbonyl resonances, an N-(Cα)-C′ experiment
(MIRROR transfer [63]) (Fig. 8d) is used to link Cα to their
attached carbonyl.

3.4.3  Sequential The assignment procedure described in what follows requires uni-
Assignment formly or uniformly and selectively 13C/15N-labeled samples for
steps 1 and 2 and a [U-13C/15N]-labeled sample for steps 3 and 4.
1. Set up a 2D 13C-13C PDSD experiment using a mixing time of
200 ms. The intermediate mixing time allows for the interresidue
 Solid-State NMR 435

correlations to build up. The comparison of an intermediate-­

mixing with a short-mixing PDSD makes it possible to iden-
tify spatial 13C-13C correlations that arise mostly from contacts
with neighboring residues, i.e., the signals visible only in the
long-­mixing PDSD arise from interresidue contacts (Fig. 8e).
Nevertheless, this information must be treated with caution
and needs to be used only in combination with exclusively
sequential spectra (see following discussion) since the signals in
a 13C-13C PDSD spectrum can arise from all spatially close car-
bons. For example, in an α-helix, the carbons of residue i + 3 are
very close to those of residue i and the contact signals between
the two residues can build up rapidly in a PDSD spectrum if the
helix is part of a very rigid segment in the protein.
2. If available, set up a 2D 13C-13C PDSD experiment using an
intermediate mixing time of 300 ms on a selectively 13C/15N-­
labeled sample. The intermediate mixing time allows for the
interresidue correlations to build up. As described in para-
graph 1 of this subheading, the comparison of an intermedi-
ate-mixing with a short-mixing PDSD makes it possible to
identify spatial 13C-13C correlations that arise from contacts
with carbons that are close in space, as are the neighboring
residues, i.e., the signals visible only in the intermediate-mix-
ing PDSD arise from interresidue contacts. Nevertheless, the
information needs to be treated with caution, as mentioned
earlier. To assign the signals, refer to the metabolic labeling
pattern of the corresponding scheme.
3. Set up the spectra linking the resonances between the residues
(see Note 10). Depending on the system complexity (residue
number, spectral dispersion, amino acid composition, line-
width, available spectrometer field strength), the spectra need
to be recorded in a 2D or 3D setup, and the number of neces-
sary spectra to perform the complete sequential assignment
will vary. The minimal amount of indispensable 2D spectra is
N-C′ (Fig. 8f), N-(Cα)-Cβ, N-(Cα)-C′, Cα-(N)-C′ (Fig. 8g),
N-(C′)-Cα (Fig. 8h) using the following magnetization trans-
fer schemes: 15N-13C and 13C-15N-specific CP; 13Cα-13Cβ
DREAM; 13C′-13Cα DARR/MIRROR or PDSD (less specific
toward 13Cα). These spectra can be set up with three acquisi-
tion times (N-Cα-Cββ, N-Cα-C′, Cα-N-C′, N-C′-Cα) if
required by the system. If necessary, supplemental spectra can
be recorded in a 2D or 3D setup. The most useful are the fol-
lowing: N-C′-(Cα)-Cβ or N-(C′)-Cα-Cβ; Cα-(N)-(C′)-Cα or
Cα-N-(C′)-Cα; N-(Cα)-Cx (DARR or PDSD 13Cα-13Cx
transfer); 3D CCC (with a short 1H-13Cα CP toward the Cα,
followed by a DREAM and a DARR or PDSD.
4. Choose an amino acid spin system with identified 15N, 13Cα,
and 13C′ resonances (residue i) and proceed to link those
436 Birgit Habenstein and Antoine Loquet

resonances to the neighboring residue. At each linking step

two resonance frequencies should be constant (i.e., identi-
fied). 15N, 13Cα of residue i can then be found in the Cα-N-C′
linked to the C′ of residue i − 1. In the N-C′-Cα, the N(i) and
C′(i − 1) are linked to the Cα of residue i − 1. The N-Cα-C′
then provides the N(i − 1) and the N-Cα-Cβ the Cβ (i − 1),
so that Cα and Cβ resonances of the spin system i − 1 are
identified and the side-chain resonances can be obtained from
the 2D 13C-13C or the 3D 13C-13C-13C. The supplemental
spectra mentioned earlier can help remove ambiguities and
solve difficult assignment stretches. For example, if the N, C′,
Cα resonances of two spin systems are identical, then using the
sequential connection that includes the Cβ resonance (present
in N-Cα-Cβ and N-(C′)-Cα-Cβ) could resolve the ambiguity.

3.4.4  Determination 1. The assigned 13Cα, 13Cβ resonances make it possible to iden-
of Secondary Structure tify the secondary chemical shift [64] ΔδCα-ΔδCβ, indicative
and Topology of Protein of the secondary structure.
Subunit 2. Calculate 13Cα(assigned)-13Cα(random coil) and 13Cβ(assigned)-
13
Cβ(random coil). Chemical shift values of amino acids in
random coil conformation can be obtained from [36].
3. Plot ΔδCα-ΔδCβ, negative values for >3 residues in a row are
indicative of β-strands and positive for α-helical conformation.
It appears that amino acids such as glycine and proline show
unusual chemical shift values, especially when they occur as
secondary structure breakers. Caution must be taken when
defining the limits of secondary structure elements from the
secondary chemical shifts.
4. Use TALOS+ [39] or PREDITOR [38] to predict the protein
dihedral angles from the assigned chemical shifts. The phi/psi
dihedral angle restraints will be used later as structural input
during the modeling process.

3.5  SSNMR Several types of structural restraints are used in the structural mod-
Structural Restraints eling of macromolecular protein filaments. The following list
(Subheadings 3.5.1–3.5.3) follows chronological order in a basic
SSNMR structure determination process of a macromolecular
assembly.

3.5.1  Dihedral Angle The dihedral angles can be derived from the SSNMR chemical shifts
Restraints based on the TALOS+ routine (step 2 in Subheading 3.4.4) and are
introduced as structural restraints in the protocol of the selected
modeling software to perform the structure determination.

3.5.2  Collection Following the nomenclature used in solution NMR, SSNMR dis-
of Distance Restraints tance restraints are classified in different categories depending
on the number of residues separating the residues involved in the
 Solid-State NMR 437

contact. With the spectral assignment, the sequential (residue i −

residue i ± 1) and medium-range (residue i − residue i ± 2, 3, or
4) contacts will be identified first because they can be derived
directly from the sequential assignment process. Long-range (resi-
due i − residue i > 4) contacts establish the 3D fold of the protein
subunit by defining the carbon network and thereby connecting
the different secondary structure elements (intramolecular long-
range), as well as by defining the subunit–subunit interfaces (inter-
molecular long-range). In this chapter, we focus on the collection
of 13C-13C and 15N-13C long-range distance restraints. While iden-
tifying a long-range contact, several considerations must be taken
into account (see Note 11).
1. Set up a 2D 13C-13C PDSD experiment using a mixing time of
400 ms on a [U-13C]- or [U-13C/15N]-labeled sample. The
long mixing time allows for long-range distance correlations
to build up. The comparison of this long-mixing (400 ms)
PDSD with the other intermediate-mixing (200 ms) PDSD
(see Subheading 3.4.3) makes it possible to identify additional
spatial 13C-13C correlations that might arise from contacts
with the neighboring residues or from residues that are far
away in the primary sequence. If the protein filament allows
for it, i.e., if the spectral resolution is high enough that the
signal can be unambiguously assigned, long-range contacts
might be encoded in the signals and be identified in a straight-
forward manner.
2. Set up a 2D 13C-13C PDSD experiment using a mixing time
of 800 ms on a selectively 13C-labeled sample. The labeling
scheme will result from the selectively labeled precursors
used for the protein production (see Subheading 3.1.2). In
selectively labeled samples, the buildup of long-range con-
tact signals is considerably enhanced when compared to
uniformly labeled protein because an SSNMR magnetiza-
tion transfer phenomenon called dipolar truncation is
reduced. In most cases, these spectra are the most promis-
ing in terms of identifying long-range 13C-13C contacts. The
long mixing time allows for long-range distance correla-
tions to build up. The comparison of this long-mixing (800
ms) PDSD with the intermediate-mixing (200 ms) PDSD
makes it possible to identify additional spatial 13C-13C cor-
relations that might arise from contacts with neighboring
residues or from residues that are far away in the primary
sequence. If the protein filament allows for it, i.e., if the
spectral resolution is high enough that the signal can be
unambiguously assigned, long-range contacts might be
encoded in the signals and be identified in a straightforward
manner.
438 Birgit Habenstein and Antoine Loquet

3.5.3  Intrasubunit 1. For the detection of intermolecular 15N-13C long-range con-

and Subunit–Subunit tacts, set up a 2D 15N-13C PAIN-CP [65] on a filament sample
Protein Interactions prepared with a mixture of [U-13C]- and [U-15 N]-labeled
subunits (see Subheading 3.1.4 for sample preparation). The
signals arise unambiguously from the intermolecular 15N-13C
contacts experiment (Fig. 9a).
2. To discriminate between intra- and intermolecular 13C-13C
long-range contacts, set up a 2D 13C-13C PDSD experiment
using a mixing time of 800 ms on a filament prepared with a
1:1 mixture of selectively 13C-labeled subunits (see Subheading
3.1.4 for sample preparation). The correlation peaks detected
between two carbons that are labeled in one labeling scheme
and unlabeled in the other must arise from an intermolecular
contact [30]. Therefore, a comparison of this spectrum with
the two spectra recorded on each labeling scheme ([1-13C]-glc

Fig. 9 Strategies for heterogeneous isotopically labeled filament samples. Heterogeneous labeling schemes
based on (a) an equimolar mixture of uniformly 13C and uniformly 15N labeled subunits (blue and white, respec-
tively), (b) an equimolar mixture of [1-13C]-glucose- and [2-13C]-glucose-labeled subunits (green and magenta,
respectively), and (c) a diluted mixture (1/4) of 13C-labeled subunits in an unlabeled subunit background (white
and yellow, respectively)
 Solid-State NMR 439

and [2-13C]-glc or [1,3-13C]-glyc and [2-13C]-glyc) makes it

possible to identify the intermolecular contacts (Fig. 9b).
3. Highlighting intramolecular long-range contacts. A mixed
labeled filament sample containing 1:X (with X = 3 to 5)
[U-13C/15N]-labeled and unlabeled subunits should be prepared,
also termed diluted sample. If long-range contact signals are
detected in a 13C-13C spectrum of a diluted sample, they arise
from intramolecular contacts. Signals present in the 13C-­13C long-
range correlation spectrum on a nondiluted [U-13C/15N]-labeled
and absent in the equivalent spectrum recorded on the diluted
sample can potentially arise from intermolecular long-range
contacts (Fig. 9c). This information should be treated with
caution since the overall decrease in signal intensity might also
cause the disappearance of intramolecular contact peaks.

3.6  SSNMR Structure Modeling the 3D structure aims at the generation of a PDB file
Calculation containing all atom coordinates for each protein subunit within the
filament assembly (Fig. 10). However, depending on the quality of
the SSNMR spectra, the amount of SSNMR structural restraints or
the access to complementary structural data from other techniques,
the modeling process can lead to 3D models with different degrees
of precision, ranging from pseudo-atomic resolution structures to
cartoonlike 3D models based on the molecular topology.
1. Prepare the SSNMR distance restraint lists (unambiguous and
ambiguous lists) encoding for intramolecular (i.e., intrasubunit)
distances and TALOS-based dihedral restraints.

Fig. 10 (a, b) T3SS needle filament and its basic building block, consisting of the smallest asymmetric unit contain-
ing all subunit–subunit interfaces. (c) Based on the detection of intermolecular SSNMR restraints, (d) distance
restraints at subunit–subunit interfaces can be derived. (e) SSNMR atomic structure of T3SS building block
440 Birgit Habenstein and Antoine Loquet

2. Generate subunit monomeric structures using standard CNS

or XPLOR-NIH routines based on simulated annealing
protocols with torsion angles as internal degrees of freedom.
At this stage, only unambiguous SSNMR restraints are used in
the calculation.
3. Carefully observe whether the structure converges to one
monomer fold. Select the ten lowest-energy monomers.
4. Assign the ambiguous restraints from the ambiguous restraint
list based on the preliminary experimentally determined 3D
fold. Rerun the structure calculation as described in step 1 by
adding the disambiguated restraints. This procedure can be car-
ried out manually, but it can also be automatized in programs
such as ARIA [66], UNIO [67], or HADDOCK [46].
5. Prepare the SSNMR distance restraint lists (unambiguous
and ambiguous lists) encoding for intermolecular (i.e. inter-
subunit) distances.
6. Choose the minimal number of subunit monomers that will
be calculated for the building block modeling. Typically, the
building block should be considered as the smallest multi-
meric protein unit encoding each existing subunit–subunit
interface of the assembly [49]. For example, for the T3SS
needle filament, composed of three (n = 3) different subunit–
subunit interfaces, the building block is an (N = n + 1 = 4)
tetramer.
7. According to the molecular topology encoded in the second-
ary chemical shifts and the monomeric 3D fold, structurally
ambiguous intermolecular distance restraints can already be
disambiguated.
8. Generate a homomultimeric subunit made of N subunits using
CNS or XPLOR-NIH routines. At this stage of the modeling,
the TALOS and intramolecular restraints are used as input in
addition to the intermolecular restraints. To guarantee the
symmetry between the different subunits, noncrystallographic
symmetries (NCS) [43] are used to reinforce the superimpos-
ability of the monomeric subunits. Please note that the quasi-­
symmetry of the local subunit structure is detected by the
presence of a single SSNMR resonance set.
9. Select the ten lowest-energy multimeric structures to perform
the quality assessment using PROCHECK-NMR [48] .

3.7  Integrative The structure determination of the filamentous protein assembly

Structural Analysis follows the previously described procedure. In this section, we
globally describe how and which data from other sources can be
integrated and used throughout the study. The integration of the
data can be performed at different stages of the structural study
(see Fig. 11).
 Solid-State NMR 441

Fig. 11 Integrative structure determination process based on SSNMR. The different tasks are numbered and
asterisks indicate the data that can be integrated. Green arrows and underlining highlight intermediate results,
if the necessary data are available. Step 1 assignment process; 2 assignment process of long-range contacts
in SSNMR data; 3 secondary structure determination; 4 structure modeling with integration of different struc-
tural data obtained from biophysical techniques. # The monomer structure can be obtained from x-ray crystal-
lography or solution NMR

3.7.1  SSNMR Analysis This subheading corresponds to steps 1 and 3 illustrated in Fig. 11.
Guided by Solution NMR
1. If the solution NMR 13C chemical shifts of a monomer or a
Chemical Shifts
truncated version of the monomer are available, predict a 2D
13
C-13C spectrum and superimpose it to the experimental
SSNMR 13C-13C spectrum recorded on the assembly, using,
for example, the CCPNMR analysis prediction tool. If a
monomer crystal structure is available, predict the 13C chemi-
cal shifts (seeprediction programs listed in Subheading 2.3) to
further predict a 2D SSNMR 13C-13C.
2. Plot the predicted 2D 13C-13C on the experimental SSNMR
spectrum and compare the two spectra. If an experimental
peak and a predicted peak correspond in an unoverlapped
region of the spectrum, you can use with caution the pre-
dicted peak as assignment to start a sequential assignment
stretch. Be aware that this assignment is only predicted and
very liable to be incorrect. Nevertheless, it can facilitate/
accelerate the sequential assignment process (described in
Subheading 3.4.3).
3. If the solution NMR chemical shifts of a monomeric subunit
are available, plot the chemical shift differences between
solution NMR and SSNMR data obtained on the protein
­
sequence. Where differences larger than the error bar are
observed, structural differences between the monomeric state
and the monomer in the assembly might occur or an interaction
442 Birgit Habenstein and Antoine Loquet

site between the monomers in the assembly might be located

in this segment.
4. The secondary structure of the subunit in the assembly can be
obtained following Subheading 3.4.4. If the monomeric protein
structure is available from X-ray crystallography or solution
NMR, the secondary structures of the monomeric subunit
and the subunit in the assembly can be compared and possible
structural rearrangements delineated.

3.7.2  SSNMR Restraint This subheading corresponds to step 2 illustrated in Fig. 11.

Assignment Guided The assignment of long-range SSNMR contacts in the 13C-13C
by a Solution NMR/Crystal spectra can be facilitated by different biophysical data, which can
Monomeric Structure or be used alone or in combination.
EM Data
1. If a monomeric subunit structure is available and the secondary
structure is revealed to be mostly conserved, the carbon–carbon
distances can be extracted using protein visualization software
(e.g., Pymol or SwissPDB viewer). Two different carbon–
carbon distance lists should be extracted: all distances below 4
Å or below 8 Å.
2. The distances should then be used to simulate a 13C-13C peak
list based on the assigned chemical shifts, which can be dis-
played on the 13C-13C spectra recorded for the detection of
long-range signals.
3. Intramolecular carbon–carbon distances below 4 Å should be
detected and below 8 Å could be visible in the spectra; 13C-13C
peaks can therefore be assigned based on the simulated peak
list. Nevertheless, the peaks should be assigned with caution
and the potential assignments discarded in regions where the
secondary structure is not conserved between the monomeric
and the assembled subunits. The remaining unassigned peaks
should arise from intermolecular carbon–carbon contacts.
4. EM provides data on the assembly dimensions and can thereby
exclude certain protein arrangements in the filament. On the
basis of scanning transmission electron microscopy (STEM),
mass-per-length data can be extracted; they give access to a
crucial filament parameter: the number of monomers per
length unit. First modeling based on the available data can
reduce the long-range intra- and intermolecular SSNMR con-
tact assignment ambiguities. If a cryo-EM map is available, a
first model based on the cryo-EM map will reduce significantly
the ambiguities arising during SSNMR long-range contact
assignment.

3.7.3  Integrative Several integrative modeling approaches have been developed to

Modeling Procedure efficiently deliver 3D models of filamentous assemblies based on
SSNMR and complementary data.
 Solid-State NMR 443

1. The protocol described in Subheading 3.6 can be used to gen-

erate an atomic resolution structure of the filament building
block. This building block structure can serve as a rigid body
core to introduce symmetry restraints (e.g., axial rise and heli-
cal angle per subunit) for energy minimization and refinement
to obtain a 3D model of the filament objet. Such a protocol,
already used without SSNMR data but starting from a mono-
meric crystal structure to determine the type II secretion pilus
structure, is currently being developed by Nilges et al. [14].
2. On the other hand, Baker et al. proposed a Rosetta protocol
[1, 26] to start from an ensemble of unfolded subunit polypep-
tide chains that are simultaneously minimized under all avail-
able experimental restraints (local SSNMR data such as
chemical shifts and distance restraints, EM density map, sym-
metry restraints). This protocol has been used to determine 3D
atomic models of the T3SS needles of Salmonella [1], Shigella
[26], and the M13 bacteriophage [68].
3. Also starting from extended subunit monomers, Habeck,
Lange et al. proposed a protocol [5] based on inferential
structure determination (ISD) [69] to integrate solution
NMR, SSNMR, and MPL data to simultaneously assign
SSNMR restraints from ambiguous data and model the E. coli
type 1 pilus structure in an iterative way. By disambiguating
the SSNMR signals with the structural model, which becomes
more precise at each step, the integration of ambiguous
SSNMR data is facilitated.

4  Notes

1. All glassware, tubes, pipet tips, stock solutions, media, and

buffer should be sterile (e.g., autoclave passage or filter
sterilization).
2. Throughout the entire expression and purification hand gloves
should be used.
3. If possible, expression and purification should be carried out
in a sterile environment to avoid contamination
4. To obtain optimal results during protein assembly, different
assembly conditions should be tested on unlabeled proteins
(salt concentrations, pH, protein concentration, seeding
effects). The resulting filament samples should be compared,
if possible recording a 1D SSNMR fingerprint spectrum
­(1H-­13C CP spectrum) and otherwise by visual comparison of
the filament pellet after centrifugation and EM.
5. Prewarmed medium (37 °C) can help the bacteria to take up
growth again.
444 Birgit Habenstein and Antoine Loquet

6. Between a protein production in LB and M9 medium the

quantity decreases usually by around 10–60%.
7. In the case of selectively labeled glc, only one-sixth of all carbon
atoms in the filament subunit will be 13C-labeled, i.e., visible in
SSNMR spectra. This gives rise to a maximum resolution but
simultaneously decreases the spectral sensitivity, one of the
most limiting factors in SSNMR analysis, by a factor of 6 and
an even higher factor with increasing dimensionality. In the
case of [1,3-13C]-glyc and [2-13C]-glyc, two-thirds and one-
third of all carbon atoms in the filament subunit will be
13
C-labeled, leading to a diminished sensitivity compared to
[U-13C/15N]-labeled protein assemblies but a significantly
higher sensitivity than for selectively glc-labeled proteins. For
a detailed comparison see [49].
8. High salt concentrations are a problem in NMR as they make
probe matching and tuning difficult, and they also increase
pulse durations.
9. Check any hardware or software issues such as cable connections.
The protein could still be in solution and salts could have
crystallized (improbable if washing has been performed).
Check the supernatant fraction of the centrifugation step after
filament recovery for protein content (UV spectroscopy or
SDS gel) or subsequently check a minimal amount of the fila-
ment sample for the protein of interest by SDS-PAGE (poly-
acrylamide gel electrophoresis). The pulse strengths could be
different from those required for experiments on this filament
sample. This situation can arise if the optimization was done in
a powder sample since NMR hard pulses in powder samples
require less power (or shorter duration) compared to those in
hydrated samples, this effect being reinforced for salty sam-
ples. Slightly higher values may be tried for the biological sam-
ples to compensate for these effects.
10. In a 2D setup of experiments containing a 13C-13C polarization
transfer, for example N-(Cα)-Cβ, the 13C frequency signals in
brackets (i.e., without acquisition before next transfer) are also
observed owing to the incomplete polarization transfer.
11. If ambiguity arises between a sequential/medium-range con-
tact and a long-range (residue i − residue i > 4) contact, the
signal should be assigned to the sequential or medium-range
contact as this usually corresponds to a shorter internuclear
distance. We recommend extensively assigning all possible
sequential and medium-range contacts prior to the long-range
assignment in the spectra designated for long-range distance
detection. The identification of a cross peak as an “unambiguous
long-range contact” requires that no other intraresidue,
sequential, or medium-range contact can explain the chemical
 Solid-State NMR 445

shift values with respect to the chosen chemical shift tolerance

window. Likewise, only a single long-range contact should
explain the observed cross peak. In case of several assignment
possibilities, the assignment should be considered spectrally
ambiguous. The NMR user defines the value of the chemical
shift tolerance window. It corresponds to the range in parts
per million for which assignment possibilities will be consid-
ered while assigning a cross peak. Typical values can be ±0.1–
0.25 ppm, depending on the experimental linewidth. The
detection of a long-range contact in homogeneously labeled
samples ([U-13C]glc; [1-13C]glc; [2-13C]glc; [1,3-13C]glyc;
[2-13C]-glyc) is always associated with an ambiguity arising
from the distinction between intramolecular and intermolecu-
lar interactions. From an NMR point of view, both possibili-
ties are not distinguishable and require the use of asymmetric
labeling strategies (see Subheading 3.1.4). In the present
chapter, distance restraints will be considered to encode for a
single distance range of 2–8.5 Å. This relatively large upper
distance (8.5 Å) is used to compensate multiple relayed
polarization transfers that can occur during a long recoupling
time, especially for 13C-13C mixing.

Acknowledgments

The authors thank their past and present colleagues, in particular Prof.
Adam Lange at the Leibniz-Institut für Molekulare Pharmakologie
for his guidance during the author postdoctoral periods and his main
intellectual contribution for the T3SS needle and type I pilus projects.
This work was further supported by the Fondation pour la Recherche
Médicale (FRM-AJE20140630090 to A.L.), the ANR (13-PDOC-
0017-01 to B.H. and ANR-­14-­CE09-0020-01 to A.L.), the FP7
program (FP7-PEOPLE-­2013-CIG to A.L.), the IdEx Bordeaux
University (Chaire d’Installation to B.H.)and the European
Research Council (ERC) under the European Union’s Horizon
2020 research and innovation program (ERC Starting Grant to
A.L., agreement 105945). Erick Dufourc is acknowledged for his
continuous support.

References
1. Loquet A, Sgourakis NG, Gupta R, Giller K, insights into its assembly and signaling mecha-
Riedel D, Goosmann C et al (2012) Atomic nisms. Proc Natl Acad Sci U S A 109:
model of the type III secretion system needle. 4461–4466
Nature 486:276–279 3. Blocker AJ, Deane JE, Veenendaal AK, Roversi P,
2. Fujii T, Cheung M, Blanco A, Kato T, Blocker Hodgkinson JL, Johnson S et al (2008) What’s
AJ, Namba K (2012) Structure of a type III the point of the type III secretion system needle?
secretion needle at 7-A resolution provides Proc Natl Acad Sci U S A 105:6507–6513
446 Birgit Habenstein and Antoine Loquet

4. Sauer FG, Futterer K, Pinkner JS, Dodson KW, 18. Tang M, Comellas G, Rienstra CM (2013)
Hultgren SJ, Waksman G (1999) Structural Advanced solid-state NMR approaches for
basis of chaperone function and pilus biogene- structure determination of membrane proteins
sis. Science 285:1058–1061 and amyloid fibrils. Acc Chem Res 46:
5. Habenstein B, Loquet A, Hwang S, Giller K, 2080–2088
Vasa SK, Becker S et al (2015) Hybrid struc- 19. Weingarth M, Baldus M (2013) Solid-state
ture of the type 1 pilus of Uropathogenic NMR-based approaches for supramolecular
Escherichia coli. Angew Chem Int Ed Engl structure elucidation. Acc Chem Res 46:
54:11691–11695 2037–2046
6. Hospenthal MK, Redzej A, Dodson K, Ukleja 20. Loquet A, Habenstein B, Lange A (2013)
M, Frenz B, Rodrigues C et al (2016) Structure Structural investigations of molecular machines
of a chaperone-usher pilus reveals the molecu- by solid-state NMR. Acc Chem Res 46:
lar basis of rod uncoiling. Cell 164:269–278 2070–2079
7. Geibel S, Waksman G (2014) The molecular 21. Yan S, Suiter CL, Hou G, Zhang H, Polenova T
dissection of the chaperone-usher pathway. (2013) Probing structure and dynamics of pro-
Biochim Biophys Acta 1843:1559–1567 tein assemblies by magic angle spinning NMR
8. Chandran Darbari V, Waksman G (2015) spectroscopy. Acc Chem Res 46:2047–2058
Structural biology of bacterial type IV secre-
22. Tycko R, Wickner RB (2013) Molecular
tion systems. Annu Rev Biochem 84:603–629 structures of amyloid and prion fibrils: con-
9. Melville S, Craig L (2013) Type IV pili in sensus versus controversy. Acc Chem Res
Gram-positive bacteria. Microbiol Mol Biol 46:1487–1496
Rev 77:323–341 23. Hong M, Zhang Y, Hu F (2012) Membrane
10. Stones DH, Krachler AM (2015) Fatal attrac- protein structure and dynamics from NMR
tion: how bacterial adhesins affect host signal- spectroscopy. Annu Rev Phys Chem 63:1–24
ing and what we can learn from them. Int 24. Egelman EH (2015) Three-dimensional
J Mol Sci 16:2626–2640 reconstruction of helical polymers. Arch
11. Cheung M, Shen DK, Makino F, Kato T, Biochem Biophys 581:54–58
Roehrich AD, Martinez-Argudo I et al (2015) 25. Egelman EH (2010) Reducing irreducible
Three-dimensional electron microscopy recon- complexity: divergence of quaternary structure
struction and cysteine-mediated crosslinking and function in macromolecular assemblies.
provide a model of the type III secretion sys- Curr Opin Cell Biol 22:68–74
tem needle tip complex. Mol Microbiol 26. Demers JP, Habenstein B, Loquet A, Kumar
95:31–50 Vasa S, Giller K, Becker S et al (2014) High-­
12. Rathinavelan T, Lara-Tejero M, Lefebre M, resolution structure of the Shigella type-III
Chatterjee S, McShan AC, Guo DC et al secretion needle by solid-state NMR and cryo-­
(2014) NMR model of PrgI-SipD interaction electron microscopy. Nat Commun 5:4976
and its implications in the needle-tip assembly 27. Hong M (1999) Determination of multiple
of the Salmonella type III secretion system. ***&phi;***-torsion angles in proteins by
J Mol Biol 426:2958–2969 selective and extensive (13)C labeling and two-­
13. Jones CH, Pinkner JS, Roth R, Heuser J, dimensional solid-state NMR. J Magn Reson
Nicholes AV, Abraham SN et al (1995) FimH 139:389–401
adhesin of type 1 pili is assembled into a fibril- 28. Castellani F, van Rossum B, Diehl A, Schubert
lar tip structure in the Enterobacteriaceae. Proc M, Rehbein K, Oschkinat H (2002) Structure
Natl Acad Sci U S A 92:2081–2085 of a protein determined by solid-state
14. Campos M, Nilges M, Cisneros DA, Francetic O ­magic-­angle-­spinning NMR spectroscopy.
(2010) Detailed structural and assembly model Nature 420:98–102
of the type II secretion pilus from sparse data. 29. Lundstrom P, Teilum K, Carstensen T,
Proc Natl Acad Sci U S A 107:13081–13086 Bezsonova I, Wiesner S, Hansen DF et al (2007)
15. Habenstein B, Loquet A. (2015). Solid-state Fractional 13C enrichment of isolated carbons
NMR: an emerging technique in structural using [1-13C]- or [2-13C]-glucose facilitates
biology of self-assemblies. Biophys Chem. the accurate measurement of dynamics at back-
16. Meier BH, Bockmann A (2015) The structure bone Calpha and side-chain methyl positions in
of fibrils from ‘misfolded’ proteins. Curr Opin proteins. J Biomol NMR 38:199–212
Struct Biol 30:43–49 30. Loquet A, Giller K, Becker S, Lange A (2010)
17. Miao Y, Cross TA (2013) Solid state NMR and Supramolecular interactions probed by 13C-­13C
protein-protein interactions in membranes. solid-state NMR spectroscopy. J Am Chem Soc
Curr Opin Struct Biol 23:919–928 132:15164–15166
 Solid-State NMR 447

31. Bockmann A, Gardiennet C, Verel R, Hunkeler in NMR structure calculation. Bioinformatics

A, Loquet A, Pintacuda G et al (2009) 23:381–382
Characterization of different water pools in 45. Shen Y, Vernon R, Baker D, Bax A (2009) De
solid-state NMR protein samples. J Biomol novo protein structure generation from incom-
NMR 45:319–327 plete chemical shift assignments. J Biomol
32. Bertini I, Engelke F, Luchinat C, Parigi G, NMR 43:63–78
Ravera E, Rosa C et al (2012) NMR properties 46. Dominguez C, Boelens R, Bonvin AM (2003)
of sedimented solutes. Phys Chem Chem Phys HADDOCK: a protein-protein docking
14:439–447 approach based on biochemical or biophysical
33. Delaglio F, Grzesiek S, Vuister GW, Zhu G, information. J Am Chem Soc 125:1731–1737
Pfeifer J, Bax A (1995) NMRPipe: a multidi- 47. Bhattacharya A, Tejero R, Montelione GT
mensional spectral processing system based on (2007) Evaluating protein structures deter-
UNIX pipes. J Biomol NMR 6:277–293 mined by structural genomics consortia.
34. Vranken WF, Boucher W, Stevens TJ, Fogh Proteins 66:778–795
RH, Pajon A, Llinas M et al (2005) The CCPN 48. Laskowski RA, Rullmannn JA, MacArthur
data model for NMR spectroscopy: develop- MW, Kaptein R, Thornton JM (1996) AQUA
ment of a software pipeline. Proteins and PROCHECK-NMR: programs for check-
59:687–696 ing the quality of protein structures solved by
35. Ulrich EL, Akutsu H, Doreleijers JF, Harano Y, NMR. J Biomol NMR 8:477–486
Ioannidis YE, Lin J et al (2008) BioMagResBank. 49. Loquet A, Habenstein B, Chevelkov V, Vasa
Nucleic Acids Res 36:D402–D408 SK, Giller K, Becker S et al (2013) Atomic
36. Wang Y, Jardetzky O (2002) Probability-based structure and handedness of the building block
protein secondary structure identification of a biological assembly. J Am Chem Soc
using combined NMR chemical-shift data. 135:19135–19138
Protein Sci 11:852–861 50. Higman VA, Flinders J, Hiller M, Jehle S,
37. Shen Y, Delaglio F, Cornilescu G, Bax A Markovic S, Fiedler S et al (2009) Assigning
(2009) TALOS+: a hybrid method for predict- large proteins in the solid state: a MAS NMR
ing protein backbone torsion angles from resonance assignment strategy using selectively
NMR chemical shifts. J Biomol NMR and extensively 13C-labelled proteins. J Biomol
44:213–223 NMR 44:245–260
38. Berjanskii MV, Neal S, Wishart DS (2006) 51. Etzkorn M, Bockmann A, Lange A, Baldus M
PREDITOR: a web server for predicting pro- (2004) Probing molecular interfaces using 2D
tein torsion angle restraints. Nucleic Acids Res magic-angle-spinning NMR on protein mix-
34:W63–W69 tures with different uniform labeling. J Am
39. Shen Y, Bax A (2010) SPARTA+: a modest Chem Soc 126:14746–14751
improvement in empirical NMR chemical shift 52. Hartmann SR, Hahn EL (1962) Nuclear dou-
prediction by means of an artificial neural net- ble resonance in rotating frame. Phys Rev
work. J Biomol NMR 48:13–22 128:2042
40. Han B, Liu Y, Ginzinger SW, Wishart DS 53. Baldus M, Petkova AT, Herzfeld J, Griffin RG
(2011) SHIFTX2: significantly improved pro- (1998) Cross polarization in the tilted frame:
tein chemical shift prediction. J Biomol NMR assignment and spectral simplification in hetero-
50:43–57 nuclear spin systems. Mol Phys 95:1197–1207
41. Kohlhoff KJ, Robustelli P, Cavalli A, Salvatella 54. Fung BM, Khitrin AK, Ermolaev K (2000) An
X, Vendruscolo M (2009) Fast and accurate improved broadband decoupling sequence for
predictions of protein NMR chemical shifts liquid crystals and solids. J Magn Reson
from interatomic distances. J Am Chem Soc 142:97–101
131:13894–13895 55. Harris RK, Becker ED, Cabral De Menezes
42. Schwieters CD, Kuszewski JJ, Tjandra N, SM, Granger P, Hoffman RE, Zilm KW et al
Clore GM (2003) The Xplor-NIH NMR (2008) Further conventions for NMR shield-
molecular structure determination package. ing and chemical shifts IUPAC recommenda-
J Magn Reson 160:65–73 tions 2008. Solid State Nucl Magn Reson
43. Brunger AT (2007) Version 1.2 of the crystal- 33:41–56
lography and NMR system. Nat Protoc 56. Shaka AF, Frenkeil T, Freeman R (1983) NMR
2:2728–2733 broadband decoupling with low radio-­
44. Rieping W, Habeck M, Bardiaux B, Bernard A, frequency power. J Magn Reson. 52:159–163
Malliavin TE, Nilges M (2007) ARIA2: auto- 57. Baldus M, Geurts DG, Hediger S, Meier BH
mated NOE assignment and data integration (1996) Efficient 15N-13C polarization transfer
448 Birgit Habenstein and Antoine Loquet

by adiabatic passage Hartmann-Hahn cross-­ 64. Luca S, Filippov DV, van Boom JH, Oschkinat
polarization. J Magn Reson Ser A 118: H, de Groot HJ, Baldus M (2001) Secondary
140–144 chemical shifts in immobilized peptides and
58. Andronesi OC, Becker S, Seidel K, Heise H, proteins: a qualitative basis for structure refine-
Young HS, Baldus M (2005) Determination of ment under magic angle spinning. J Biomol
membrane protein structure and dynamics by NMR 20:325–331
magic-angle-spinning solid-state NMR spec- 65. Lewandowski JR, De Paepe G, Griffin RG
troscopy. J Am Chem Soc 127:12965–12974 (2007) Proton assisted insensitive nuclei cross
59. Ader C, Frey S, Maas W, Schmidt HB, Gorlich polarization. J Am Chem Soc 129:728–729
D, Baldus M (2010) Amyloid-like interactions 66. Bardiaux B, Malliavin T, Nilges M (2012)
within nucleoporin FG hydrogels. Proc Natl ARIA for solution and solid-state
Acad Sci U S A 107:6281–6285 NMR. Methods Mol Biol 831:453–483
60. Szeverenyi NM, Sullivan MJ, Maciel GE 67. Guerry P, Herrmann T (2012) Comprehensive
(1982) Observation of spin exchange by two-­ automation for NMR structure determina-
dimensional Fourier-transform C-13 cross tion of proteins. Methods Mol Biol 831:
polarization-magic-angle spinning. J Magn 429–451
Reson 47:462–475 68. Morag O, Sgourakis NG, Baker D, Goldbourt
61. Verel R, Ernst M, Meier BH (2001) Adiabatic A (2015) The NMR-Rosetta capsid model of
dipolar recoupling in solid-state NMR: the M13 bacteriophage reveals a quadrupled
DREAM scheme. J Magn Reson 150:81–99 hydrophobic packing epitope. Proc Natl Acad
62. Westfeld T, Verel R, Ernst M, Bockmann A, Sci U S A 112:971–976
Meier BH (2012) Properties of the DREAM 69. Rieping W, Habeck M, Nilges M (2005)
scheme and its optimization for application to Inferential structure determination. Science
proteins. J Biomol NMR 53:103–112 309:303–306
63. Scholz I, Hodgkinson P, Meier BH, Ernst M 70. Schraidt O, Marlovits TC (2011) Three-­
(2009) Understanding two-pulse phase-­ dimensional model of Salmonella’s needle
modulated decoupling in solid-state NMR. J complex at subnanometer resolution. Science
Chem Phys 130:114510 331:1192–1195
 Chapter 30

Energy Requirements for Protein Secretion

via the Flagellar Type III Secretion System
Marc Erhardt

Abstract
Protein transport across the cytoplasmic membrane is coupled to energy derived from adenosine triphos-
phate hydrolysis or the protein motive force (pmf). A sophisticated, multi-component type III secretion
system exports substrate proteins of both the bacterial flagellum and virulence-associated injectisome sys-
tem of many Gram-negative pathogens. The type-III secretion system is primarily a pmf-driven protein
exporter. Here, I describe methods to investigate the export of substrate proteins into the culture super-
natant under conditions that manipulate the pmf.

Key words Type III secretion system, Bacterial flagellum, Protein export, Proton motive force, ∆pH
gradient, ∆Ψ gradient, Ionophore, Carbonyl cyanide m-chlorophenylhydrazone (CCCP), Valinomycin

1  Introduction

Bacterial protein transportation systems utilize energy derived

from the protein motive force (pmf) or adenosine triphosphate
(ATP) hydrolysis for the translocation of substrate proteins across
biological membranes [1].
Protein substrates of the bacterial flagellum and the evolution-
arily related virulence-associated injectisome or needle complex are
secreted by a homologous protein transportation system, termed
the type III secretion system (T3SS). Secretion of substrate proteins
via the flagellar-specific type III secretion system (f-T3SS) or viru-
lence-associated type III secretion system (v-T3SS) of the injecti-
some complex is essential for the assembly of the corresponding
nanomachine, as well as the secretion of effector proteins in the case
of the injectisome system (reviewed in detail in [2–4]). The T3SS is
intrinsically a pmf-driven protein exporter that exploits the activity
of an associated ATPase to facilitate substrate secretion [5–8].
The core T3SS consists of eight or nine proteins that directly
participate in the process of substrate protein translocation across
the inner membrane [2–4]. Five integral membrane proteins

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_30, © Springer Science+Business Media LLC 2017

449
450 Marc Erhardt

(f-T3SS: FlhA FlhB FliO FliP FliQ FliR; v-T3SS: SctV SctU SctR
SctS SctT) form the export gate and are involved in the primary
substrate recognition, substrate unfolding, energy transduction
and protein transport across the cytoplasmic membrane. An associ-
ated ATPase complex (f-T3SS: FliH FliI FliJ; injectisome: SctL
SctN SctO) has a facilitating role in substrate recognition, sub-
strate unfolding and energy transduction, but is not strictly required
for protein export [9].
A scaffold structure formed by an integral membrane ring
(MS-ring; flagellum: FliF; injectisome: SctD SctJ) is essential for
the assembly of a functional core export apparatus [10]. In addi-
tion, accessory proteins form a cytoplasmic ring that facilitates sub-
strate recognition and binding of ATPase complex components
(flagellum: FliG FliM FliN; injectisome: SctQ) [11–13].
The contribution of the pmf and ATP hydrolysis to the protein
export process via T3SS has been examined using T3SS-dependent
substrate protein secretion of mutant strains and after treatment
with compounds that modulate the pmf. Here, I describe a meth-
odology to analyse the protein export of an f-T3SS-specific sub-
strate into culture supernatant in the presence or absence of
compounds interfering with the pmf.

2  Materials

Standard chemicals are purchased in analytical quality from estab-

lished commercial suppliers. Prepare all solutions using ultrapure
water unless indicated otherwise.

2.1  FlgM-­ 1. Salmonella enterica serovar Typhimurium strains: TH3730

Secretion Assay PflhDC5451::Tn10dTc[del-25], TH10874 ∆flgM5628::FRT
∆araBAD923::flgM-FKF ParaBAD924 (see Note 1).
2. Lysogeny broth (LB): 10 g tryptone, 5 g yeast extract, 5 g
NaCl. Add 12 g agar for LB agar plates. Add water to a volume
of 1 L and autoclave.
3. Shaking incubator (see Note 2).
4. Spectrophotometer for OD600 determination.
5. Anhydrotetracycline: 0.2 mg/mL stock solution in 50% H2O
50% ethanol (see Note 3).
6. l-arabinose: 20% stock in H2O (see Note 4).
7. Tabletop centrifuge, refrigerated.

2.2  Protein 1. Spectrophotometer for OD600 determination.

Fractionation 2. Table-top centrifuge, refrigerated.
 Energy Requirements of Type-III Protein Secretion 451

2.2.1  Protein Extraction 1. Nitrocellulose filter, 0.45 μm pore size.

by Filtration Over
2. 2× sodium dodecyl sulfate (SDS) sample buffer: 100 mM Tris–
a Nitrocellulose Filter
HCl, pH 6.8, 4% SDS, 20% glycerol, 1% β-mercaptoethanol,
25 mM EDTA, 0.04% bromophenol blue (see Note 5).
3. Heating block for 1.6 and 2.0 mL centrifugation tubes.

2.2.2  Protein 1. Trichloroacetic acid (TCA). Store at 4 °C.

Precipitation Using 2. Acetone. Store at 4 °C.
Trichloroacetic Acid
3. Vortexer.
4. 2× SDS sample buffer: 100 mM Tris–HCl, pH 6.8, 4% SDS,
20% glycerol, 1% β-mercaptoethanol, 25 mM EDTA, 0.04%
bromophenol blue (see Note 5).
5. Heating block for 1.6 and 2.0 mL centrifugation tubes.

2.3  Immunoblotting 1. 4–20% precast gels. Store at 4 °C.

2. Mini-gel caster system and SDS-polyacrylamide gel electro-
phoresis (PAGE) apparatus.
3. SDS running buffer: 25 mM Tris–HCl, 192 mM glycine, 0.1%
SDS, pH 8.3.
4. Western blot transfer membranes: 0.2 μm pore size Hybond-P
polyvinylidene fluoride (PVDF) or 0.45 μm pore size nitrocel-
lulose (see Note 6).
5. Western blot transfer buffer: 25 mM Tris–HCl, 192 mM gly-
cine, 20% methanol, pH 8.3.
6. Trans-Blot apparatus for western blot transfer.
7. Serum-purified anti-FlgM rabbit polyclonal antibodies [14],
dilution 1:10,000.
8.
Horseradish-peroxidase-conjugated anti-rabbit polyclonal
antibodies, dilution 1:10,000–1:20,000.

2.4  Assays to Inhibit 1.

Ionophore carbonyl cyanide m-chlorophenylhydrazone
Proton Motive Force (CCCP): 20 mM stock solution: Dissolve 4.1 mg CCCP in 1
mL dimethyl sulfoxide (DMSO) (see Note 7).
2. Valinomycin: 20 mM stock solution: Dissolve 22.2 mg valino-
mycin in 1 mL distilled H2O.
3. Potassium chloride: 1 M stock solution: Dissolve 7.45 g KCl in
100 mL distilled H2O.
4. Potassium acetate: 1 M stock solution: Dissolve 9.81 g
CH3COOK in 100 mL distilled H2O (see Note 8).
452 Marc Erhardt

3  Methods

Perform experiments at room temperature unless indicated

otherwise.

3.1  FlgM-­ 1. Streak S. Typhimurium strains TH3730 (see Note 9) or

Secretion Assay TH10874 (see Note 10) for single colonies on fresh LB plates
(see Note 1). Incubate overnight at 37 °C.
2. Inoculate a single colony of S. Typhimurium strains TH3730
or TH10874 into 1 mL LB and incubate overnight at 37 °C in
a water bath incubator, shaking at 200 rpm.
3. Dilute overnight culture of S. Typhimurium strains TH3730
or TH10874 1:100 in 3 mL LB and incubate at 37 °C in a
water bath incubator, shaking at 200 rpm. Grow approximately
2 h until optical density (OD600) of 0.5. Induce flagellar gene
expression by addition of 100 ng/mL anhydrotetracycline and
resume incubation at 37 °C for 60 min. Grow TH10874 cul-
tures in LB supplemented with 0.2% l-arabinose to induce
expression of FlgM for 60 min at 37 °C.
4. Pellet cells by 5 min centrifugation at 10,000 × g and perform
permeabilization and washing steps as detailed in Subheading 3.4.
5. Resuspend in 3 mL LB medium containing appropriate dilu-
tions of pmf inhibitors and 100 ng/mL anhydrotetracycline or
0.2% l-arabinose for TH3730 or TH10874, respectively, and
resume incubation at 37 °C for 30 min in a water bath incuba-
tor, shaking at 200 rpm as outlined in Subheading 3.4.
6. Store bacterial culture on ice until further treatment.

3.2  Protein 1. Remove 0.5 mL of bacterial culture and record OD600 for nor-
Fractionation malization purposes (see step 4 in Subheading 3.2.1 and step
7 in Subheading 3.2.2).
2. Centrifuge 2 mL of bacterial culture at 10,000 × g for 5 min in
a tabletop centrifuge and transfer 1.8 mL of the supernatant to
a new centrifugation tube. Discard remaining supernatant and
store pellet on ice until further treatment (label ‘cellular
fraction’).
3. Centrifuge supernatant at 10,000 × g for 5 min in a tabletop
centrifuge and transfer 1.6 mL of the supernatant to a new cen-
trifugation tube (label ‘supernatant fraction’) (see Note 11).
4. Two alternative methods can be used for collecting proteins
from the culture supernatant: filtration on nitrocellulose filters
and TCA precipitation.
 Energy Requirements of Type-III Protein Secretion 453

3.2.1  Protein Extraction 1. Filter supernatant from step 3 in Subheading 3.2 through a
by Filtration over prewetted nitrocellulose filter with a pore size of 0.45 μm for
Nitrocellulose Filter protein binding.
(See Note 12) 2. Elute proteins by addition of 40 μL 2× SDS sample buffer and
heat treatment for 30 min at 65 °C.
3. Resuspend bacterial pellet from step 2, Subheading 3.2, in 50
μL 2× SDS sample buffer and heat for 10 min at 95 °C.
4. Adjust cellular and supernatant fractions to 20 OD600 equiva-
lents per microlitre by addition of 2× SDS sample buffer and
keep on ice or store at −20 °C until further use.

3.2.2  Protein 1. Add TCA to a final concentration of 10% to the supernatant

Precipitation Using from step 3, Subheading 3.2, and incubate on ice for 30 min.
Trichloroacetic Acid 2. Centrifuge at 20,000 × g for 30 min in a refrigerated tabletop
centrifuge at 4 °C and discard supernatant.
3. Resuspend pellet in 1 mL ice-cold acetone by vortexing.
4. Centrifuge at 20,000 × g for 30 min in a refrigerated tabletop
centrifuge at 4 °C and discard supernatant.
5. Dry pellet overnight or for 30 min in a laminar flow bench.
6. Resuspend supernatant pellet in 40 μL 2× SDS sample buffer
and heat for 10 min 95 °C.
7. Resuspend bacterial pellet from step 2, Subheading 3.2, in 50
μL 2× SDS sample buffer and heat for 10 min at 95 °C.
8. Adjust cellular and supernatant fractions to 20 OD600 equiva-
lents per microlitre by addition of 2× SDS sample buffer and
keep on ice or store at −20 °C until further use.

3.3  Immunoblotting 1. Load 200 OD600 equivalents of cellular and supernatant frac-
tions onto 4–20% precast gels and perform protein separation
by SDS-PAGE in standard Tris-glycine buffer.
2. Following separation, electrotransfer proteins to a 0.2 μm pore
size Hybond-P PVDF transfer membrane (see Note 6) or a
0.45 μm pore size nitrocellulose membrane using a Trans-Blot
transfer apparatus.
3. Perform immunodetection using appropriate concentrations
of primary and secondary antibodies. In case of TH3730 or
TH10874, secreted and cellular FlgM protein is detected
using serum-purified anti-FlgM rabbit polyclonal antibodies
(dilution 1:10,000, [14]) and horseradish-peroxidase-conju-
gated anti-­rabbit polyclonal antibodies (dilution 1:10,000,
BioRad).
454 Marc Erhardt

3.4  Assays to Inhibit 1. Grow bacterial cultures as described in Subheading 3.1, steps
Proton Motive Force 1–3.
3.4.1  Disruption 2. Pellet bacterial culture by 5 min centrifugation at 10,000 × g
of Proton Motive Force and resuspend in 3 mL LB medium containing 0–20 μM
Using Carbonyl Cyanide carbonyl cyanide m-chlorophenylhydrazone (see Notes 14
m-Chlorophenylhydrazone and 15).
(See Note 13) (See Fig. 1a) 3. Wash bacterial cells by centrifugation at 10,000 × g for 5 min.
Resuspend pellet in 3 mL LB medium containing 0–20 μM
carbonyl cyanide m-chlorophenylhydrazone and inducer as in
step 3, Subheading 3.1.
4. Resume incubation at 37 °C for 30 min in a water bath shaker
at 200 rpm and continue with step 6, Subheading 3.1.

Fig. 1 (a) Effect of inhibition of pmf on FlgM export. FlgM secretion was inhibited by
addition of 10 μM uncoupling agent carbonyl cyanide m-chlorophenyl hydrazone
(CCCP) and completely abolished by treatment with 20 μM CCCP in strain TH10874
(arabinose-inducible flgM ). Cytoplasmic FlgM levels remained constant. (b) Effect
of inhibition of ∆Ψ component of pmf on FlgM export. FlgM secretion was inhibited
by addition of valinomycin in the presence of K+. Cells were pretreated with 120
mM Tris–HCl to permeabilize the outer membrane to valinomycin where indicated.
(c) Effect of inhibition of ∆pH on FlgM export. Secretion of FlgM for cultures grown
in pH 5 was inhibited by addition of 34 mM potassium acetate. Adapted with per-
mission from Macmillan Publishers Ltd: Nature [6], copyright 2008
 Energy Requirements of Type-III Protein Secretion 455

3.4.2  Disruption of ΔΨ 1. Grow bacterial cultures as described in Subheading 3.1, steps
Component of Proton 1–3.
Motive Force by K+/ 2. Pellet bacterial culture by 5 min centrifugation at 10,000 × g
Valinomycin (See Note 16) and resuspend in 3 mL LB medium containing 120 mM Tris–
(See Fig. 1b) HCl, pH 7.3. Incubate for 2 min (see Note 17).
3. Pellet bacterial culture by 5 min centrifugation at 10,000 × g
and resuspend in 3 mL LB medium containing 120 mM Tris–
HCl, pH 7.3, and 0–40 μM valinomycin in the presence or
absence of 150 mM KCl (see Note 15).
4. Pellet bacterial culture by 5 min centrifugation at 10,000 × g,
discard supernatant and resuspend in 3 mL LB medium con-
taining 120 mM Tris–HCl, pH 7.3, and 0–40 μM valinomycin
in the presence or absence of 150 mM KCl.
5. Resume incubation at 37 °C for 30 min in a water bath shaker
at 200 rpm and continue with step 6, Subheading 3.1.

3.4.3  Disruption of ΔpH 1. Grow bacterial cultures as described in Subheading 3.1, steps
Component of Proton 1–3.
Motive Force by Potassium 2. Pellet bacterial culture by 5 min centrifugation at 10,000 × g
Acetate (See Note 18) (See and resuspend in 3 mL LB medium.
Fig. 1c)
3. Wash bacterial cells by centrifugation at 10,000 × g for 5 min
and discard supernatant.
4. Resuspend pellet in 3 mL LB medium at pH 7 or pH 5 in the
presence or absence of 34 mM potassium acetate, respectively.
Add inducer of flagellar genes transcription as required for
strains TH3730 and TH10874 (see Note 15).
5. Resume incubation at 37 °C for 30 min in a water bath shaker
at 200 rpm and continue with step 6, Subheading 3.1.

4  Notes

1. Alternatively, use a mutant strain where ATP synthesis is

uncoupled from the pmf: TH11802 ∆atpA::tetRA
∆flgM5628::FRT ∆araBAD923::flgM-FKF ParaBAD934,
which is deficient in a major subunit of the FOF1 ATP synthase.
The absence of the FOF1 ATP synthase results in a growth
defect, which can be partially rescued by growth in media con-
taining 0.2% glucose.
2. For best growth use a shaking water bath incubator.
3. Filter sterilize using 0.2 μm mixed cellulose ester or polyether-
sulfone filters.
4. Filter sterilize using 0.2 μm mixed cellulose ester or polyether-
sulfone filters.
456 Marc Erhardt

5. Add freshly prepared β-mercaptoethanol.

6. Transfer to a 0.2 μm pore size Hybond-P PVDF membrane, as
recommended for efficient electrotransfer of FlgM.
7. Dissolve freshly in DMSO.
8. Adjust pH to desired final pH by addition of HCl or NaOH.
9. This strain harbours the flagellar master regulatory operon
flhDC under control of an anhydrotetracycline-inducible pro-
moter and allows for constant expression of flagellar genes in
the presence of inducer.
10. This strain harbours flgM as a non-structural reporter substrate
of the flagellar T3SS under the control of an arabinose-­
inducible promoter.
11. This step minimizes potential contamination of the superna-
tant fraction by residual bacterial cells.
12. Protein binding to nitrocellulose filters is recommended for
efficient recovery of secreted FlgM from the culture
supernatant.
13. The pmf consists of two components, a proton concentration
gradient (ΔpH) and a charge difference between the periplas-
mic and cytoplasmic faces of the membrane (ΔΨ). The iono-
phore CCCP disrupts both the proton gradient ∆pH and the
membrane potential ∆Ψ by causing an influx of H+ into the
cytoplasm [6].
14. To control for DMSO-induced effects, add 0.5% DMSO to
samples not treated with CCCP.
15. Add inducer 100 ng/mL anhydrotetracycline or 0.2% l-­
arabinose for continuous expression of flagellar genes
(TH3730) or flgM (TH10874), respectively.
16. Valinomycin renders membranes permeable to potassium,
which dissipates the ΔΨ component of the pmf by balancing
the charge difference [6, 15].
17. Appropriate controls include samples not treated with 120
mM Tris–HCl, pH 7.3.
18. Weak acids such as acetate or benzoate cross the cytoplasmic
membrane in neutral form and release a proton in the cytoplasm.
The resulting decrease in the cytoplasmic pH essentially col-
lapses the proton gradient ∆pH at an external pH of 5 [6, 16].

Acknowledgments

This work was supported by the Helmholtz Association young

investigator grant VH-NG-932 and the People Programme (Marie
Curie Actions) of the European Union Seventh Framework
Programme (grant 334030).
 Energy Requirements of Type-III Protein Secretion 457

References

1. Wickner W, Schekman R (2005) Protein trans- tein secretion in Salmonella enterica. PLoS
location across biological membranes. Science Genet 10:e1004800
310:1452–1456 10. Morimoto YV, Ito M, Hiraoka KD, Che YS,
2. Erhardt M, Namba K, Hughes KT (2010) Bai F, Kami-Ike N, Namba K, Minamino T
Bacterial nanomachines: the flagellum and type (2014) Assembly and stoichiometry of FliF and
III injectisome. Cold Spring Harb Perspect FlhA in Salmonella flagellar basal body. Mol
Biol 2:a000299 Microbiol 91:1214–1226
3. Minamino T (2014) Protein export through 11. McMurry JL, Murphy JW, Gonzalez-Pedrajo
the bacterial flagellar type III export pathway. B (2006) The FliN-FliH interaction medi-
Biochim Biophys Acta 1843:1642–1648 ates localization of flagellar export ATPase
4. Diepold A, Wagner S (2014) Assembly of the FliI to the C ring complex. Biochemistry
bacterial type III secretion machinery. FEMS 45:11790–11798
Microbiol Rev 38:802–822 12. Erhardt M, Hughes KT (2010) C-ring require-
5. Wilharm G, Lehmann V, Krauss K, Lehnert ment in flagellar type III secretion is bypassed
B, Richter S, Ruckdeschel K, Heesemann by FlhDC upregulation. Mol Microbiol
J, Trulzsch K (2004) Yersinia enterocolitica 75:376–393
type III secretion depends on the proton 13. Diepold A, Kudryashev M, Delalez NJ, Berry
motive force but not on the flagellar motor RM, Armitage JP (2015) Composition, forma-
components MotA and MotB. Infect Immun tion, and regulation of the cytosolic c-ring, a
72:4004–4009 dynamic component of the type III secretion
6. Paul K, Erhardt M, Hirano T, Blair DF, injectisome. PLoS Biol 13:e1002039
Hughes KT (2008) Energy source of flagellar 14. Hughes KT, Gillen KL, Semon MJ, Karlinsey
type III secretion. Nature 451:489–492 JE (1993) Sensing structural intermediates in
7. Minamino T, Namba K (2008) Distinct roles bacterial flagellar assembly by export of a nega-
of the FliI ATPase and proton motive force tive regulator. Science 262:1277–1280
in bacterial flagellar protein export. Nature 15. Minamino T, Morimoto YV, Hara N, Namba
451:485–488 K (2011) An energy transduction mechanism
8. Lee PC, Zmina SE, Stopford CM, Toska J, used in bacterial flagellar type III protein
Rietsch A (2014) Control of type III secretion export. Nat Commun 2:475
activity and substrate specificity by the cyto- 16. Minamino T, Imae Y, Oosawa F, Kobayashi Y,
plasmic regulator PcrG. Proc Natl Acad Sci U Oosawa K (2003) Effect of intracellular pH on
S A 111:E2027–E2036 rotational speed of bacterial flagellar motors.
9. Erhardt M, Mertens ME, Fabiani FD, Hughes J Bacteriol 185:1190–1194
KT (2014) ATPase-independent type-III pro-
 Chapter 31

Identification of Effectors: Precipitation

of Supernatant Material
Nicolas Flaugnatti and Laure Journet

Abstract
Bacterial secretion systems allow the transport of proteins, called effectors, as well as external machine
components in the extracellular medium or directly into target cells. Comparison of the secretome, i.e. the
proteins released in the culture medium, of wild-type and mutant cells provides information on the secre-
tion profile. In addition, mass spectrometry analyses of the culture supernatant of bacteria grown in liquid
culture under secreting conditions allows the identification of secretion system substrates. Upon identifica-
tion of the substrates, the secretion profile serves as a tool to test the functionality of secretion systems.
Here we present a classical method used to concentrate the culture supernatant, based on trichloroacetic
acid precipitation.

Key words Supernatant, TCA precipitation, Secretome

1  Introduction

Bacterial secretion systems are macromolecular machines dedicated

to the transport of proteins across the cell envelope. These secretion
systems deliver effectors outside the cell, either in the medium
(T1SS, T2SS, T5SS, T9SS) or directly into target cells (T3SS, T4SS,
T6SS) [1]. Secretion of effector proteins into the milieu can be
observed in these systems, and the analysis of secretion supernatant
has been widely used either to identify new secreted effectors or to
probe the functionality of secretion systems. For contact-­dependent
systems such as the T3SS, in vitro secretion in the medium can be
observed under certain conditions (e.g. Ca2+ depletion, acidic pH)
[2, 3]. As it is not always possible to predict effectors by bioinfor-
matics approaches (see Chapter 2), analysis of the content of the
culture media, the so-called secretome, using global proteomic
approaches has been widely used to identify secretion system sub-
strates in T2SS [4–7], T6SS [8–10], T3SS [11] and T9SS [12].
Upon the identification of substrates, the secretion profile is
used to test the functionality of the secretion system using sodium

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_31, © Springer Science+Business Media LLC 2017

459
460 Nicolas Flaugnatti and Laure Journet

dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE)

of the supernatant fraction followed by coomassie blue staining or
immunostaining by western blot detection of specific effectors or
components of the machinery. In some secretion systems, such as
T3SS and T6SS, external structural components are released in the
milieu upon secretion and can also be used to test the proper
assembly of the system. For example, the Hcp release assay is widely
used to probe the functionality of T6SS (see also Chapter 32).
Such analyses of secretomes require concentrating the dilute
solutions that are the culture supernatant or the biological fluids.
This can be achieved using trichloroacetic acid (TCA) precipitation
and acetone-based protocols [13, 14]. Alternative protocols have
been proposed using acetone alone, methanol/chloroform [15], or
a combination of pyrogallol red, molybdate, and methanol [16].
Here we detail the most classical assay used to precipitate pro-
teins of bacterial culture supernatant based on TCA precipitation;
it is used widely in secretion system studies. First, cells and super-
natant are separated by centrifugation. Cell-free culture superna-
tant fraction samples are then obtained by further centrifugation
and filtration and subjected to TCA precipitation before analysis by
mass spectrometry or western blot.

2  Materials

1. Lysogeny broth (LB) or recommended medium to grow

strain of interest in secreting conditions.
2. TCA (CCl3COOH, MW: 163.39, TCA): 100% (w/v). Add
227 mL ultrapure water to previously unopened bottle con-
taining 500 g TCA (see Note 1). Wear personal protective
equipment and work under a fume hood.
3. Sodium deoxycholate (DOC): 16 mg/mL (optional, see Note 2).
Store at room temperature.
4. Acetone. Pre-chill before use.
5. Refrigerated centrifuge capable of 21,460 × g or tabletop
centrifuge (see Note 3).
6. 0.22-μm-pore-size syringe filters (see Note 4).
7. 2 mL syringe.
8. 3 M Tris–HCl, pH 8.8
9. SDS-PAGE loading buffer: 60 mM Tris–HCl, pH 6.8, 2% SDS,
10% glycerol, 5% β-mercaptoethanol, 0.01% bromophenol blue.
10. Boiling water bath or thermomixer.
11. Vortexer.
12. 2 mL microtubes (safe-lock) (see Note 3).
 TCA Precipitation 461

13. Fume hood and personal protective equipment for TCA

handling.
14. Spectrophotometer to measure absorbance at λ = 600 nm.
15. SDS-PAGE and protein transfer apparatus.

3  Methods

1. Grow a 10 mL bacterial strain culture in the appropriate

medium and conditions allowing secretion (see Notes 5 and 6).
Measure the optical density at λ = 600 nm (OD600).
2. Dispose the culture in 2 mL microtubes (see Note 7) and pel-
let cells by centrifugation at 6000 × g for 5 min.
3. Carefully remove 1.8 mL of supernatant and transfer it to a
new microtube and keep it on ice before performing step 5.
4. Carefully discard the remaining 200 μL of supernatant from the
cell pellet obtained in step 3. (Centrifuge again at 6000 × g for
5 min if the cells from the cell pellet started to resuspend.) Keep
this total cell fraction pellet on ice before resuspending the
pellet in an appropriate volume of SDS-PAGE loading buffer
(the equivalent of 0.2–0.5 OD600 units (ODU)/10 μL). Store
on ice (or at −20 °C).
5. Centrifuge the 1.8 mL supernatant fraction obtained in step 3
at 16,000 × g at 4 °C for 5 min. Carefully recover the super-
natant and transfer it to a new microtube. Avoid recovering
the remaining cells from the pellet, if any.
6. Filter-sterilize the supernatant using a 0.22 μm syringe filter
and transfer the filtered supernatant directly to a new micro-
tube. Check the volume (around 1.5 mL). This fraction con-
stitutes the cell-free fraction (see Note 8).
7. Add TCA to a final concentration of 20% (add 375 μL TCA to
1.5 mL filtered supernatant). Invert four times to mix, vortex
and keep on ice for 1 h to overnight.
8. Centrifuge at 21,000 × g for 30 min at 4 °C.
9. Discard the as much supernatant as possible (see Note 9).
10. Resuspend the pellet in 400–500 μL cold acetone. Vortex.
11. Centrifuge at 21,000 × g for 15 min at 4 °C. Discard the super-
natant with a pipet and further on a paper towel. Leave the tube
open at room temperature to dry the pellet (see Note 10).
12. Resuspend the pellet in appropriate buffer for further analysis
(such as mass spectrometry) or go to step 13 for SDS-PAGE
analysis.
13. Resuspend TCA-precipitated pellets of supernatant fractions
in an appropriate volume of SDS-PAGE loading buffer
462 Nicolas Flaugnatti and Laure Journet

(1 ODU/10 μL). If the TCA-precipitated sample turns yellow,

add 1 μL (or more) of Tris–HCl, pH 8.8.
14. Vortex. Heat the samples from step 4 (whole-cell fraction)
and step 13 (cell-free supernatant precipitated fraction) at
95 °C for 10 min (see Note 11).
15. Analyse whole-cell samples and cell-free supernatants by

SDS-­PAGE, followed by coomassie blue staining or immunob-
lot. If performing western blot, include a control for cell lysis,
using antibodies detecting an internal protein. Alternatively,
check the coomassie or silver staining profile.

4  Notes

1. For a safe and easy preparation, avoid weighting out the TCA
crystalline powder as it becomes easily syrupy upon contact
with humidity. The TCA solution must be kept in a dark glass
bottle. It is very corrosive and should be handled with care
with suitable protection. Do not use plastic containers.
2. DOC may be used as a carrier to assist protein precipitation.
If using DOC, add the DOC stock solution at the final con-
centration of 0.16 mg/mL to the cell-free fraction obtained in
step 6, vortex and leave on ice for 30 min; then proceed to
TCA precipitation as described in step 7. DOC should be
washed out with further acetone washing steps (repeat steps
10 and 11 three times). However, this could be a problem
with further mass spectrometry analysis.
3. Tabletop centrifuge at maximum speed may be sufficient;
however, we generally use a higher speed. TCA-resistant tubes,
such as Eppendorf tubes (check with your manufacturer for
tube compatibility), should be used.
4. In principle, any 0.22 μm filter may be used. However, we had
experience with a secreted protein that was retained on polyvi-
nylidene fluoride filters, so we moved to Polyether sulfone
(PES) filters. Be aware that the material of the filter may be of
importance.
5. A “non-secreting strain” should be used as a control, such as a
mutant in a core component, the ATPase energizing the assem-
bly of the secretion machinery or the substrate transport.
6. You must find conditions where secretion can be detected
in vitro. Because effectors can be secreted at low levels, high-­
sensitivity mass spectrometry methods may be required [10].
If the secretion system is not produced in laboratory conditions,
native endogenous promoter(s) may be swapped for an induc-
ible promoter (e.g. Ptac, Plac, PBAD) to artificially induce the
expression of the secretion system [17].
 TCA Precipitation 463

7. A 5–10 mL of culture is generally sufficient. We usually transfer

2 mL of supernatant in 2 mL microtubes, leading to the recov-
ery of 1.5 mL of cell-free supernatant. An equivalent of 1 OD600
unit will be loaded on the gel for supernatant fraction analysis.
To scale up experiments, you will have to use tubes with larger
volumes compatible with high-speed spin that are resistant to
TCA. Appropriate 50 mL tubes (polyether) may be used; check
first with your manufacturer for TCA compatibility.
8. At this stage, for bacteria producing high levels of vesicles (e.g.
for T9SS in Bacteroidetes), an additional ultracentrifugation
(30,000 × g for 4 h at 4 °C) will allow separation of vesicles
from vesicle-free supernatant [12].
9. Check the orientation of the microtube before the centrifuga-
tion step since the pellet is not always visible at this stage.
10. You may use a vacuum concentrator (SpeedVac or equivalent)
for 10 min to evaporate the acetone. However, pellets may be
more difficult to resuspend if too dry, and this step may
decrease recovery of the samples.
11. In some cases, we have observed that an additional freezing
at −20 °C in SDS-PAGE loading buffer helps resuspension of
TCA precipitates.

Acknowledgments

This work was supported by the Centre National de la Recherche

Scientifique, the Aix-Marseille Université and grants from the
Agence Nationale de la Recherche (ANR-14-CE14-0006-02 and
ANR-15-CE11-0019-01). The doctoral studies of N.F. are sup-
ported by the ANR-14-CE14-0006-02 grant.

References
1. Costa TR, Felisberto-Rodrigues C, Meir A, 4. Coulthurst SJ, Lilley KS, Hedley PE, Liu H,
Prevost MS, Redzej A, Trokter M, Waksman G Toth IK, Salmond GP (2008) DsbA plays a
(2015) Secretion systems in Gram-negative critical and multifaceted role in the production
bacteria: structural and mechanistic insights. of secreted virulence factors by the phytopatho-
Nat Rev Microbiol 13:343–359 gen Erwinia carotovora subsp. atroseptica.
2. Cornelis GR, Biot T, Lambert de Rouvroit C, J Biol Chem 283:23739–23753
Michiels T, Mulder B, Sluiters C, Sory MP, 5. Kazemi-Pour N, Condemine G, Hugouvieux-­
Van Bouchaute M, Vanooteghem JC (1989) Cotte-­Pattat N (2004) The secretome of the
The Yersinia yop regulon. Mol Microbiol 3: plant pathogenic bacterium Erwinia chrysan-
1455–1459 themi. Proteomics 4:3177–3186
3. Beuzon CR, Banks G, Deiwick J, Hensel M, 6. Sikora AE, Zielke RA, Lawrence DA, Andrews
Holden DW (1999) pH-dependent secretion PC, Sandkvist M (2011) Proteomic analysis of
of SseB, a product of the SPI-2 type III secre- the Vibrio cholerae type II secretome reveals
tion system of Salmonella typhimurium. Mol new proteins, including three related serine
Microbiol 33:806–816 proteases. J Biol Chem 286:16555–16566
464 Nicolas Flaugnatti and Laure Journet

7. Burtnick MN, Brett PJ, DeShazer D (2014) 12. Veith PD, Chen YY, Gorasia DG, Chen D,
Proteomic analysis of the Burkholderia pseudo- Glew MD, O’Brien-Simpson NM, Cecil JD,
mallei type II secretome reveals hydrolytic Holden JA, Reynolds EC (2014)
enzymes, novel proteins, and the deubiquitin- Porphyromonas gingivalis outer membrane
ase TssM. Infect Immun 82:3214–3226 vesicles exclusively contain outer membrane
8. Hood RD, Singh P, Hsu F, Güvener T, Carl and periplasmic proteins and carry a cargo
MA, Trinidad RR, Silverman JM, Ohlson BB, enriched with virulence factors. J Proteome
Hicks KG, Plemel RL, Li M, Schwarz S, Wang Res 13:2420–2432
WY, Merz AJ, Goodlett DR, Mougous JD 13. Hwang BJ, Chu G (1996) Trichloroacetic acid
(2010) A type VI secretion system of precipitation by ultracentrifugation to concen-
Pseudomonas aeruginosa targets a toxin to trate dilute protein in viscous solution.
bacteria. Cell Host Microbe 7:25–37 BioTechniques 20:982–984
9. Russell AB, Singh P, Brittnacher M, Bui NK, 14. Ozols J (1990) Amino acid analysis. Methods
Hood RD, Carl MA, Agnello DM, Schwarz S, Enzymol 182:587–601
Goodlett DR, Vollmer W, Mougous JD (2012) 15. Wessel D, Flügge UI (1984) A method for the
A widespread bacterial type VI secretion effec- quantitative recovery of protein in dilute solu-
tor superfamily identified using a heuristic tion in the presence of detergents and lipids.
approach. Cell Host Microbe 11:538–549 Anal Biochem 138:141–143
10. Fritsch MJ, Trunk K, Diniz JA, Guo M, Trost 16. Caldwell RB, Lattemann CT (2004) Simple
M, Coulthurst SJ (2013) Proteomic identifica- and reliable method to precipitate proteins
tion of novel secreted antibacterial toxins of the from bacterial culture supernatant. Appl
Serratia marcescens type VI secretion system. Environ Microbiol 70:610–612
Mol Cell Proteomics 12:2735–2749
17. Gueguen E, Cascales E (2013) Promoter
11. Deng W, de Hoog CL, Yu HB, Li Y, Croxen swapping unveils the role of the Citrobacter
MA, Thomas NA, Puente JL, Foster LJ, Finlay rodentium CTS1 type VI secretion system in
BB (2010) A comprehensive proteomic analysis interbacterial competition. Appl Environ
of the type III secretome of Citrobacter roden- Microbiol 79:32–38
tium. J Biol Chem 285:6790–6800
 Chapter 32

Screening for Secretion of the Type VI Secretion System

Protein Hcp by Enzyme-Linked Immunosorbent Assay
and Colony Blot
Brent S. Weber, Pek Man Ly, and Mario F. Feldman

Abstract
The bacterial type VI secretion system (T6SS) is a secretory apparatus encoded by many Gram-negative bac-
teria. The T6SS facilitates the secretion and injection of toxic effector proteins into host cells, providing a
competitive advantage to bacteria encoding this machinery. The activity of the T6SS can be monitored by
probing for the conserved tubule component Hcp, which is secreted to the supernatants by the T6SS. Detection
of Hcp in culture supernatants is indicative of an active T6SS, but this secretion system is often tightly regu-
lated or inactive under laboratory conditions and different bacterial strains display differing Hcp secretion
phenotypes. Herein, we describe an enzyme-linked immunosorbent assay (ELISA) and colony blot methods
to facilitate large-scale screening of isolates for Hcp secretion and, thus, T6SS activity.

Key words ELISA, Colony blot, Supernatant, Hcp, Effector

1  Introduction

Secretion systems encoded by Gram-negative bacteria secrete a

wide range of proteins and are often essential for virulence [1]. Of
the many secretory machines characterized, the type VI secretion
system (T6SS) has emerged as a potent mediator of antibacterial
and antieukaryotic activity [2, 3]. The T6SS is encoded by approx-
imately 13 conserved proteins, which assemble to form a secretory
apparatus capable of secreting effector substrates to adjacent bacte-
rial and eukaryotic cells [4]. The effector repertoire is highly vari-
able, both across different species and in individual isolates of a
single species [5]. Furthermore, the T6SS is often tightly regu-
lated, and myriad mechanisms exist among different bacteria to
control activation of T6SS [6]. However, activation of T6SS invari-
ably results in the secretion of Hcp, an essential structural compo-
nent of the secretory apparatus. Thus, Hcp secretion is a molecular
marker for T6SS, and detection of Hcp in culture supernatants is
indicative of an active T6SS [7]. Typically, Hcp secretion has been

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_32, © Springer Science+Business Media LLC 2017

465
466 Brent S. Weber et al.

detected by western blot or mass spectrometry. While some bacterial

strains encode a constitutively active T6SS, others maintain tight
regulation of this system using a diverse range of mechanisms.
Furthermore, different isolates of the same species often show dif-
ferent T6SS activity, for example between clinical and environmen-
tal isolates [8–10]. A single bacterial strain can also show variability
in T6SS activation [11, 12]. Screening large numbers of species,
strains, or colonies by western blot to assess T6SS activity is time
and resource consuming. Therefore, we have developed two meth-
ods to screen a large number of colonies, either from many differ-
ent isolates or from many colonies from the same isolate, for Hcp
secretion [12, 13]. These assays enable relatively rapid detection of
T6SS activity among hundreds or even thousands of isolates or
colonies and provide a time-saving alternative to more traditional
western blot or mass spectrometry methods. We describe the appli-
cation of an enzyme-linked immunosorbent assay (ELISA) and a
colony blot method to detect Hcp secretion by isolates and trans-
poson mutant libraries of Acinetobacter baumannii, which can be
adapted for use in other T6SS-encoding organisms. The ELISA
protocol is particularly suited to screening many different isolates
at once, while the colony blot method has obvious applications for
screening transposon mutant libraries. Both assays can be adapted
for the detection of any secreted protein in an organism of
interest.

2  Materials

2.1  ELISA 1. Lysogeny broth(LB) broth: Dissolve 10 g NaCl, 10 g ryptone,

5 g yeast extract in 1 L distilled water. Autoclave at 121 °C for
15 min to sterilize.
2. LB agar: Dissolve 15 g agar in 1 L LB broth. Autoclave at 121
°C for 15 min to sterilize.
3. Sterile toothpicks or other tool to inoculate colonies into
96-well plates.
4. Binding buffer: 100 mM sodium bicarbonate/carbonate, pH
9.6. Dissolve 3.03 g Na2CO3, 6.0 g NaHCO3 in 1 L distilled
water.
5. Phosphate buffered saline (PBS) wash buffer: Dissolve 8 g
NaCl, 0.2 g KCl, 1.44 g Na2HPO4, 0.24 g KH2PO4 in 1 L
distilled water. Prepare as a 10× solution and dilute prior to
use. Autoclave at 121 °C for 15 min to sterilize.
6. PBS-Tween 20 (PBST) wash buffer: Add 1 mL Tween 20 to
1 L PBS.
7. Blocking and antibody diluent solutions: Dissolve 5% skim
milk in PBS (initial blocking buffer) or 2.5% skim milk in PBST
(diluent for antibodies). Prepare fresh for each experiment.
 Screens for Protein Secretion 467

8. 96-well plates for growth cultures and high-binding ELISA.

9. Antibodies: Primary: Rabbit anti-Hcp (polyclonal, not com-
mercially available, developed against Hcp protein of interest)
(see Note 1).
10. Secondary: Goat anti-rabbit horseradish peroxidase (HRP)
conjugate.
11. TMB substrate and stop solution: 3,3′, 5,5″-tetramethyl-benzidine.
12. Standard laboratory equipment: Multichannel pipettes, centri-
fuges, incubators, microplate reader.

2.2  Colony Blot 1. LB agar: Dissolve 10 g NaCl, 10 g tryptone, 5 g yeast extract,

for Hcp Secretion 15 g agar in 1 L distilled water. Autoclave at 121 °C for 15 min
to sterilize.
2. Circular nitrocellulose blotting membranes: 0.45 μm, diameter:
82 mm. Autoclave in glass Petri dish at 121 °C for 15 min to
sterilize.
3. Tris-buffered saline (TBS) wash buffer: Dissolve 8.76 g NaCl,
1.21 g Tris in 1 L distilled water. Adjust to pH 8.0 with HCl.
4. TBST wash buffer: Add 1 mL Tween 20 to 1 L TBS.
5. Odyssey® blocking buffer TBS (LI-COR) or equivalent reagent.
6. Primary antibodies: Rabbit anti-Hcp (polyclonal, not commer-
cially available, developed against Hcp protein of interest)
(see Note 1), mouse anti-E. coli RNA polymerase β prime (or
any antibody raised against a cytoplasmic protein).
7. Goat anti-mouse and goat anti-rabbit secondary antibodies
coupled to infrared fluorescent dyes.
8. Fluorescence imaging system (LI-COR Odyssey or equivalent).
9. Standard laboratory equipment: Metal tweezers, incubator, 50
mL conical tubes, tumbling or rocking platform.

3  Methods

3.1  Hcp ELISA 1. The day prior to beginning the experiment, inoculate LB agar
plate(s) with the bacterial strain(s) of interest and incubate
overnight at 37 °C. This should be done in a manner that
ensures well-defined single colonies are present the next day.
2. Prepare 96-well plates for bacterial growth: Aseptically add
200 μL LB broth to the wells of a 96-well plate. Using sterile
toothpicks, inoculate individual colonies from plates grown
overnight into each well. If available, inoculate a positive and
negative control strain into the last two wells (e.g., strain that
secretes Hcp and an hcp mutant). Place 96-well plate in a
humidified chamber (see Note 2) at 37 °C with shaking (200
rpm) overnight (see Note 3).
468 Brent S. Weber et al.

3. The next day, retrieve 96-well plate and measure the OD600 in
a plate reader for a measure of bacterial cell growth. Then,
centrifuge the plate at 5000 × g for 10 min to pellet the bacte-
rial cells.
4. Meanwhile, transfer 25 μL blocking buffer to the wells of a
96-well high-binding ELISA plate.
5. Using a multichannel pipette, carefully transfer 75 μL of the
supernatants to the high-binding ELISA plate. Avoid pipetting
any cellular material (see Note 4). Be sure to keep the original
plate with the cell pellets as this will serve as the master stock
for any positive wells later on. The plate can be kept at 4 °C, or
alternatively add glycerol to the pellets and store at −80 °C for
the long term.
6. Incubate ELISA plate at room temperature for 1.5 h.
7. After incubation, remove solution from plate and wash three
times with PBS. Then completely fill wells with blocking buffer
(5% skim milk in PBS). Place on a rocking platform at room
temperature for 1 h.
8. Remove blocking solution, wash once with PBS.
9. Add 6 μL anti-Hcp antibody to 12 mL 2.5% skim milk in PBST
(1:2000 dilution) (see Note 5). Pipet 100 μL of this solution
into each well of the ELISA plate and place on room-­
temperature rocking platform for 1 h.
10. Remove solution and wash thoroughly three to five times with
PBST.
11. Add 2.4 μL goat anti-rabbit HRP conjugate to 12 mL 2.5%
skim milk in PBST (1:5000 dilution) and pipet 100 μL of this
solution into each well of the ELISA plate. Incubate for 1 h at
room temperature on rocking platform.
12. Remove solution and wash thoroughly three to five times with
PBST.
13. Add 100 μL TMB substrate to wells. Gently shake the plate or
place on rocking platform. Watch plate for appearance of blue
color, which may be very rapid or take several minutes(see
Note 6). At desired time, measure the A650nm of the plate (see
Note 7). This can be used in conjunction with the A600nm taken
earlier to compare wells relative to growth. Take a photograph
of the plate if desired (see Fig. 1).

3.2  Colony Blot All liquids should be discarded as per biohazard protocols. All steps
for Hcp SECRETION are performed at room temperature unless otherwise stated.
1. Inoculate LB agar plate(s) with bacteria of interest in a way
that allows defined single colonies to grow. For large-scale
screening of colonies, LB agar plates are inoculated by spread
 Screens for Protein Secretion 469

Fig. 1 (a) Example of using Hcp ELISA to screen multiple isolates of a given species at once. Here, 45 different
isolates of various Acinetobacter species were screened for Hcp secretion in duplicate. The results highlight
the variability among different species for T6SS activity within a given genus. (b) Screening a single strain, A.
baumannii ATCC 17978, for Hcp secretion. Single colonies of the strain were inoculated and used to perform
the Hcp ELISA. For this strain, a plasmid encodes the repressors for T6SS but can be lost upon culture without
selection [12]

plating of diluted bacteria cultures in such a way as to allow no

more than 200 colony-forming units per plate. Incubate at 37
°C overnight (see Note 8).
2. Working aseptically, transfer colonies onto a nitrocellulose
membrane by gently overlaying one sterile membrane using
tweezers on top of the colonies for about 30 s to 1 min, or
until entire membrane is moistened (see Note 9). Gently lift
membrane from plate and allow to dry, transfer side up, for
20 min.
3. Place the dried membrane to a 50 mL conical tube, transfer
side away from walls, and wash for 5 min (tumbling or rock-
ing) with 25 mL distilled water three times, or until colonies
are no longer attached to membrane. Next, wash the mem-
brane twice with 25 mL TBS.
4. In a new 50 mL conical tube, block the membrane with 25 mL
Odyssey blocking buffer TBS for 1 h at room temperature or
overnight at 4 °C.
5. In a new 50 mL conical tube, add 1.5 mL TBST and 1.5 mL
Odyssey blocking buffer TBS (1:1 ratio). Add 3.3 μL anti-­Hcp
antibody (1:9000 dilution) (see Note 5) and 1.2 μL anti-­RNA
polymerase antibody (1:2500 dilution) (see Note 10). Transfer
membrane to this tube, transfer side away from walls, and
incubate (tumbling or rocking) for 40 min.
6. Discard solution and wash membrane three times with 10 mL
TBST (10 min each).
470 Brent S. Weber et al.

Fig. 2 Use of colony blots to isolate active T6SS mutants from a T6SS-inactive strain. (a) Example of an Hcp
colony blot used to screen transposon mutants of A. baumannii strain 1225, which has an inactive T6SS and
does not secrete Hcp under normal laboratory conditions. Arrows indicate strong signals from single colonies
probed with the anti-Hcp antibody. (b) Hcp colony blot used on patch plated A. baumannii transposon mutants.
Colonies I and II show strong signals for anti-Hcp. (c) Western blot for Hcp expression in whole cells (WC) and
secretion in supernatant (SUP) of colonies I and II from figure (b). RNAP antibody used as lysis and loading
control. Wild-type A. baumannii 1225 expresses but does not secrete Hcp and is used as a negative control

7. In a new 50 mL conical tube, add 2.5 mL TBST and 2.5 mL

Odyssey blocking buffer TBS (1:1 ratio). Add 0.4 μL ­anti-­rabbit
and anti-mouse antibodies (1:12,500 dilution). Transfer mem-
brane to this tube, transfer side away from walls, and incubate
(tumbling or rocking) in the dark for 40 min (see Note 11).
8. Discard solution and wash membrane two times with 10 mL
TBST (10 min each) in the dark.
9. Wash membrane once with 10 mL TBS (10 min) in the dark.
10. Image membrane using LI-COR Odyssey CLx imaging system
(see Fig. 2).

4  Notes

1. For ideal results, an anti-Hcp antibody raised against purified

Hcp from the bacterial species of interest is preferred, whether
polyclonal or monoclonal. Hcp is generally well conserved
within a given genus of bacteria, and anti-Hcp antibodies have
been shown to cross react with Hcp proteins across different
 Screens for Protein Secretion 471

bacterial genera [14]. We have successfully employed our anti-­

Hcp antibody raised against A. baumannii to detect Hcp in
other Acinetobacter species, but the Hcp protein is highly con-
served. Validation of a given anti-Hcp antibody is an important
step before using it for the ELISA.
2. We typically use a large plastic container with a lid and place
damp paper towels in the bottom. This prevents excessive
evaporation of liquid from plate that would occur in the shak-
ing incubator.
3. Incubation time will vary depending on bacterial species under
study. Mid-log phase cultures may be desired rather than over-
night incubation, but we have found overnight (~16 h) gives
consistent results. Longer incubation times may result in exces-
sive lysis of bacterial cells, which may interfere with down-
stream analysis.
4. It is important not to transfer any cells into the ELISA plates
in order to only measure secreted Hcp. Extra centrifuging may
be required. Alternatively, 0.22 μm filters are available for
96-well plates; however, in our experience careful pipetting is
usually sufficient to avoid this problem.
5. Antibody titer must be empirically determined.
6. A solution of sulfuric acid can be used to stop the reaction, but
we typically do not include this step and simply read the absor-
bance of the plate at 650 nm. In our experience, the color
change that occurs after stopping the reaction can lead to a
signal that is too intense to be measured by our plate reader.
This is particularly problematic with strains that robustly
secrete Hcp.
7. Often bubbles will result during pipetting. These can affect
absorbance readings and should be removed (using a flame or
pipet tip) before measuring absorbance.
8. Dilution of bacterial cultures for spread plating must be opti-
mized and is species and strain specific. For best results, avoid
using inoculated plates over a day old.
9. Be careful not to press down onto agar surface. Avoid disrupt-
ing colonies as much as possible to limit cell lysis.
10. We use anti-RNA polymerase as a probe for cell lysis. Colony
blots can be performed by only probing for anti-Hcp; how-
ever, probing for anti-RNA polymerase eliminates single colo-
nies with strong anti-Hcp signals owing to lysis on the
membrane.
11. We use tin foil to wrap around conical tubes.
472 Brent S. Weber et al.

References

1. Costa TR, Felisberto-Rodrigues C, Meir A, natural transformation and contact-­dependent

Prevost MS, Redzej A, Trokter M, Waksman bacterial killing indicative of Type VI secre-
G (2015) Secretion systems in Gram-negative tion system activity. Appl Environ Microbiol
bacteria: structural and mechanistic insights. 82:2833–2842
Nat Rev Microbiol 13:343–359 9. Repizo GD, Gagne S, Foucault-Grunenwald
2. Pukatzki S, Ma AT, Sturtevant D, Krastins ML, Borges V, Charpentier X, Limansky AS,
B, Sarracino D, Nelson WC, Heidelberg JF, Gomes JP, Viale AM, Salcedo SP (2015)
Mekalanos JJ (2006) Identification of a con- Differential role of the T6SS in Acinetobacter
served bacterial protein secretion system in Vibrio baumannii virulence. PLoS One 10:e0138265
cholerae using the Dictyostelium host model sys- 10. Unterweger D, Kitaoka M, Miyata ST,
tem. Proc Natl Acad Sci U S A 103:1528–1533 Bachmann V, Brooks TM, Moloney J, Sosa
3. Mougous JD, Cuff ME, Raunser S, Shen A, O, Silva D, Duran-Gonzalez J, Provenzano D,
Zhou M, Gifford CA, Goodman AL, Joachimiak Pukatzki S (2012) Constitutive type VI secre-
G, Ordonez CL, Lory S, Walz T, Joachimiak tion system expression gives Vibrio cholerae
A, Mekalanos JJ (2006) A virulence locus of intra- and interspecific competitive advantages.
Pseudomonas aeruginosa encodes a protein PLoS One 7:e48320
secretion apparatus. Science 312:1526–1530 11. Tang L, Liang X, Moore R, Dong TG (2015)
4. Russell AB, Peterson SB, Mougous JD (2014) The icmF3 locus is involved in multiple adap-
Type VI secretion system effectors: poisons tation- and virulence-related characteristics in
with a purpose. Nat Rev Microbiol 12:137–148 Pseudomonas aeruginosa PAO1. Front Cell
5. Cianfanelli FR, Monlezun L, Coulthurst SJ Infect Microbiol 5:83
(2016) Aim, load, fire: the type VI secre- 12. Weber BS, Ly PM, Irwin JN, Pukatzki S,
tion system, a bacterial nanoweapon. Trends Feldman MF (2015) A multidrug resistance
Microbiol 24:51–62 plasmid contains the molecular switch for type
6. Silverman JM, Brunet YR, Cascales E, VI secretion in Acinetobacter baumannii. Proc
Mougous JD (2012) Structure and regula- Natl Acad Sci U S A 112:9442–9447
tion of the type VI secretion system. Annu Rev 13. Weber BS, Miyata ST, Iwashkiw JA, Mortensen
Microbiol 66:453–472 BL, Skaar EP, Pukatzki S, Feldman MF (2013)
7. Pukatzki S, McAuley SB, Miyata ST (2009) Genomic and functional analysis of the type VI
The type VI secretion system: translocation secretion system in Acinetobacter. PLoS One
of effectors and effector-domains. Curr Opin 8:e55142
Microbiol 12:11–17 14. Carruthers MD, Nicholson PA, Tracy EN,
8. Bernardy EE, Turnsek MA, Wilson SK, Tarr Munson RS Jr (2013) Acinetobacter bauman-
CL, Hammer BK (2016) Diversity of clinical nii utilizes a type VI secretion system for bacte-
and environmental isolates of Vibrio cholerae in rial competition. PLoS One 8:e59388
 Chapter 33

Effector Translocation: Cya Reporter Assay

Suma Chakravarthy, Bethany Huot, and Brian H. Kvitko

Abstract
An accurate and complete roster of the Type III effector (T3E) proteins translocated by the P. syringae
Type III secretion system (T3SS) into host cells is critical to understanding the pathogen’s interactions
with plants. The adenylate cyclase (Cya) reporter offers a highly sensitive and robust assay for monitoring
the translocation of T3Es. T3Es are fused to the calmodulin-dependent adenylate-cyclase domain of
CyaA. The T3E targets Cya for translocation through the T3SS into the host cell at which point it is acti-
vated by calmodulin and converts adenosine triphosphate into cyclic adenosine monophosphate (cAMP).
The T3SS translocation-dependent increase in cAMP concentration in plant cells is then measured with an
enzyme-linked immunosorbent assay kit. The Cya reporter can be used to determine whether a candidate
protein is translocated by T3SS or to measure relative levels of T3SS translocation in a semiquantitative
manner.

Key words Pseudomonas syringae, Type III secretion system, Type III translocation, Translocation
reporter, Adenylate cyclase, Calmodulin, cAMP, ELISA

1  Introduction

The plant pathogenic bacterium Pseudomonas syringae deploys a

Type III secretion system (T3SS) to translocate Type III effector
(T3E) proteins directly into plant host cells. The translocation of
T3Es is essential for P. syringae pathogenicity [1]. Although col-
lectively P. syringae strains infect a wide range of plant hosts, creat-
ing diverse symptoms, individual strains typically infect a limited
range of hosts. The host range of a given P. syringae strain is defined
largely by its repertoire of translocated T3Es [2]. T3Es act as viru-
lence factors, coordinating their actions to modify host cellular tar-
gets and creating a susceptible state amenable to bacterial
proliferation. However, individual T3Es may also be detected if
cognate plant resistance (R) protein immune receptors are present
in the host [3]. Detection of T3Es by R proteins results in a potent
effector-triggered immune response that blocks pathogen prolif-
eration. Collectively, the T3E repertoire of a given P. syringae strain

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_33, © Springer Science+Business Media LLC 2017

473
474 Suma Chakravarthy et al.

both delineates that strain’s capacity to create susceptibility in a

given host plant through the T3Es’ collective virulence functions,
as well as its capacity to induce a virulence through R protein–
mediated detection [4]. Therefore, having an accurate and com-
plete roster of a particular P. syringae strain’s repertoire of
translocated T3Es is critical to understanding its interactions with
plants.
The use of the Cya translocation reporter has been crucial in
confirming the T3E rosters of P. syringae strains [5, 6]. T3SS trans-
location reporter constructs generate a distinct output signal that
is only created when the candidate protein fused to the reporter is
delivered by T3SS competent bacteria into host cells [7–9].
Constructs used in the Cya translocation reporter assay have two
main components to accomplish this: a eukaryotic cell-specific
reporter and a T3SS-specific translocation signal. The reporter
used is the adenylate cyclase domain (Cya2-400) of the Bordetella
pertussis CyaA adenylate cyclase toxin. The specificity derives from
the enzyme’s calmodulin-dependent conversion of adenosine tri-
phosphate to cyclic adenosine monophosphate (cAMP), which
cannot occur in bacteria owing to a lack of calmodulin [8]. The
Cya2-400 domain alone is incapable of independent exit from bacte-
ria or entry in host cells as it lacks the proper translocation signal,
thereby preventing production of a signal not associated with the
delivery of T3Es [8]. The Cya2-400 reporter domain is fused to the
C-terminus of a candidate translocated effector of interest (EOI)
to preserve the N-terminal T3SS translocation signal. P. syringae
expressing the 3′ Cya2-400–T3E fusion protein is inoculated into
leaves of the test host at relatively high concentrations [6, 10].
Within a few hours following infiltration, P. syringae will deploy its
T3SS, initiating the translocation of T3Es into the plant cells. If
the expressed Cya2-400–T3E candidate protein fusion can be trans-
located via the T3SS, it will also be delivered into the host cells
along with the other effectors. Upon exposure to calmodulin
within the plant cytoplasm, Cya2-400 adenylate cyclase will be acti-
vated resulting in the accumulation of cAMP. The concentration of
cAMP in the leaf tissue can then be measured by an enzyme-linked
immunosorbent assay (ELISA) assay, which is standardized against
soluble protein. A mutant P. syringae strain with a defective T3SS
is used as a control to confirm that the cAMP accumulation is
T3SS-dependent. In addition to its use to verify candidate T3Es as
being translocated by P. syringae, the Cya reporter assay has also
been used to measure relative levels of T3SS translocation in a
semiquantitative manner since increased concentrations of cAMP
correlates with increased protein translocation [11, 12].
 Cya Reporter Assay 475

2  Materials

2.1  Construction 1. 1.5 mL microcentrifuge tubes.

of Plasmids 2. Tabletop microcentrifuge.
and Strains
3. Primers to amplify the EOI, P1 and P2, and universal primers
M13F (GTT TTC CCA GTC ACG AC) and M13R (CAG GAA
ACA GCT ATG AC).
4. Genomic DNA of P. syringae strain that encodes the EOI.
5. PrimeSTAR HS DNA polymerase (CloneTech) or equivalent.
6. Molecular-biology-grade water.
7. Agarose for gel electrophoresis.
8. TBE gel running buffer (10.8 g Tris, 5.5 g boric acid, 4 mL
0.5 M ethylenediaminetetraacetic acid (EDTA), fill to 1 L with
dH2O).
9. Ethidium bromide (10 mg/mL, final concentration 0.5 μg/mL).
10. DNA ladder.
11. DNA loading dye.
12. pCPP5371 (Phrp-GW-Cya destination vector) plasmid DNA
[13] (see Notes 1 and 2).
13. pENTR/SD/D-TOPO kit from Thermo Fisher Scientific or
equivalent.
14. P. syringae pv. tomato DC3000.
15. LR Clonase II from Thermo Fisher Scientific.
16. Competent E. coli cloning strain such as DH5α or TOP10.
17. Antibiotic stocks of kanamycin (50 mg/mL) and gentamicin
(10 mg/mL), dissolved in dH2O and filter sterilized.
18. KB (King’s medium B) liquid and agar solidified medium
(20 g Bacto peptone, 0.4 g MgSO4·7H2O, glycerol, 1.5 g
K2HPO4 (see Note 3); fill to 1 L with dH2O; add 18 g agar for
solidified media, autoclave).
19. LB (Luria-Bertani) liquid and agar solidified medium (10 g of
tryptone, 5 g yeast extract, 10 g NaCl, 1 mL 1 M NaOH, fill to
1 L with dH2O. Add 15 g agar for solidified media, autoclave).
20. Kit for plasmid minipreparation.
21. Primers to confirm final Cya reporter plasmid P3 (F primer, TGA
GCA TGC TAC CGA GTA ACG CAG CT) and P4 (R primer,
AGT GGT ACC GAT ATC GAA TTC TTA GCT GT).
22. 300 mM sucrose, filter sterilized.
23. 1 mm gap electroporation cuvettes.
476 Suma Chakravarthy et al.

24. Cell electroporator.

25. 14 mL disposable culture tubes.
26. 2× NEB OneTaq.

2.2  Plant Inoculation 1. Plants to inoculate (e.g., Nicotiana benthamiana, Arabidopsis,

tobacco) (see Note 4): Enough plants to conduct three infiltra-
tions per test strain.
2. P. syringae pv. tomato DC3000 pCPP5388 (pCPP5371::Phrp-­
avrPto-­Cya) secretion positive control.
3. P. syringae pv. tomato DC3000 T3SS (−) strain such as
CUCPB5113 (ΔhrcQ b-U::SpR) pCPP5388 (pCPP5371::Phrp-­
avrPto-­Cya) secretion negative control [14, 15].
4. Long wooden inoculation dowels.
5. 10 mM MgCl2 (see Note 5).
6. 1 mL needleless syringes.
7. Dissecting needle.
8. Spectrophotometer.
9. Kimwipe or paper towel.
10. Large soft-tip black Sharpie marker (see Note 6).
11. 2.2 mL round-bottom microcentrifuge tubes.
12. 4 mm diameter disposable biopsy punches or cork borers.
13. Liquid nitrogen and dewar.
14. 4.5 mm copper BBs.

2.3  cAMP ELISA 1. cAMP ELISA kit (Enzo).

and Bradford Assay 2. 12 × 75 mm Pyrex culture tubes or similar.
3. Multichannel or electronic repeater pipette.
4. Microplate reader that can read at 405 and 595 nm.
5. 0.1 M HCl (see Note 7).
6. Kimwipes.
7. Vacuum aspirator.
8. Orbital shaker.
9. 96-well plates.
10. dH2O.
11. Bradford reagent.
12. Bovine serum albumin (BSA) standard.
13. Ethanol vapor “bubble breaker.” A squirt bottle modified with
its feed tube cut to 5 cm. filled one-quarter with 96% ethanol.

2.4  Data Analysis Basic data analysis software package.

 Cya Reporter Assay 477

3  Methods

3.1  Construction 1. Design gene-specific primers to amplify the coding sequence

of a Cya Fusion of your EOI in order to clone into the Gateway entry vector
Reporter Construct pENTR/SD/D-TOPO. The forward primer, P1, must con-
with Effector tain a 5’CACC followed by the ATG start codon, and the
of Interest reverse primer, P2, must exclude the stop codon. This will
and Introduction ensure directional cloning into the entry vector.
into P. syringae 2. Use a high-fidelity DNA polymerase enzyme that generates
DC3000 (See Note 8) blunt-ended products to amplify your EOI from Pseudomonas
or the appropriate host. Set up the polymerase chain reaction
(PCR) as described in what follows.
10 μL 5× PrimeSTAR buffer with Mg++.
4 μL 2.5 mM dNTP mix.
1 μL 10 μM forward primer (P1).
1 μL 10 μM reverse primer (P2).
100 ng genomic DNA.
0.5 μL Taq DNA polymerase at 2.5 units/μL.
Sterile molecular-biology-grade water to a final volume of
50 μL. Add contents to a PCR tube and mix well with a pipette.
3. Run reaction in a thermocycler with the following conditions.
95 °C for 5 min.
98 °C for 10 s.
55 °C (or appropriate annealing temperature) for 15 s.
72 °C 1 min/kb.
Repeat cycle steps 2–4 for 30 cycles.
Final extension 72 °C for 5 min.
Hold at 12 °C.
4. Perform agarose gel electrophoresis using a 1% gel in TBE
running buffer, 0.5 μg/mL ethidium bromide, to confirm
product amplification. A strong, single band at the expected
size must be observed. Alternatively, gel-purify the PCR
desired product using the adequate kit.
5. Perform TOPO cloning reaction to introduce the PCR prod-
uct into pENTR/SD/D-TOPO.
4 μL fresh PCR product.
1 μL salt solution.
Sterile water to 5 μL.
1 μL TOPO vector.
Mix components in a tube and incubate at room temperature
for 5–30 min.
478 Suma Chakravarthy et al.

6. Transform 2 μL of the cloning reaction into chemically compe-

tent E. coli cells and select for transformants on LB with 50
μg/mL kanamycin. Screen colonies obtained on the selection
plate for presence of desired clone, using colony PCR with
gene-­specific primers.
7. For the colony PCR, prepare a PCR strip or plate by pipetting
9 μL/well of sterile water. Pick individual colonies with a tooth-
pick and mix it into the water, taking care to not pick too many
cells. The water should not be turbid to the eye. Boil the cells at
99 °C for 5 min in a thermocycler and cool to room temperature.
Give the tubes or plate a quick spin to bring down the contents
of the tube. Add 11 μL of enzyme mix per sample, prepared as
follows: 10 μL 2× NEB OneTaq, 0.5 μL 10 μM primer 1 (P1),
0.5 μL 10 μM of primer 2 (P2). Enzyme mix can be prepared as
a master mix based on the total number of samples. Run the
samples in a thermocycler with the following conditions:
94 °C for 30 s.
30 cycles of:
94 °C for 15–30 s.
45–68 °C for 15–30 s.
68 °C for 1 min/kb.
Final extension 68 °C for 5 min.
Hold at 12 °C.
8. Prepare purified plasmid DNA of selected clone by miniprep
and sequence confirm the EOI gene with M13F and M13R
primers (see Note 9).
9. Perform an LR (Clonase Gateway recombination reaction)
Gateway reaction to introduce the EOI into the pCPP5371
Cya destination vector. This will generate a fusion protein of
the EOI and the Cya at its C-terminus.
100–300 ng pENTR/SD/D-TOPO::EOI, entry clone.
300 ng pCPP5371, Phrp-GW-Cya destination vector.
4 μL 5× LR Clonase II reaction buffer.
TE buffer, pH 8.0 to final 16 μL.
Add all components to a microcentrifuge tube, mix well, and
incubate at room temperature for 1–2 h. Add 2 μL proteinase K to
each reaction and incubate for 10 min at 37 °C.
10. Transform 2 μL of the cloning reaction into chemically compe-
tent E. coli cells, and select for transformants on LB with
10 μg/mL of gentamicin.
11. Screen colonies obtained on the selection plate for the pres-
ence of the desired clone using colony PCR with primers P3
and P4. The expected product should be the size of the EOI
(bp) + 1.4 kb.
 Cya Reporter Assay 479

12. Prepare purified plasmid DNA by miniprep of the selected clone

and sequence the EOI with gene-specific primers P1 and P2.
13. Introduce Cya reporter plasmid pCPP5371::EOI into P. syrin-
gae DC3000 by electroporation. Grow DC3000 from frozen
stocks on KB agar medium at 30 °C for 1–2 days. From a
freshly grown plate, inoculate a 5 mL culture in liquid KB
medium and grow overnight at 30 °C.
14. Harvest cells by centrifugation at 3500 relative centrifugal
force (RCF) for 5 min at room temperature. Wash cells twice
with 5 mL 300 mM sucrose at room temperature. Resuspend
in a final volume of 100 μL 300 mM sucrose. Add 100–200 ng
pCPP5371::EOI and mix gently with a pipette.
15. Take a 1 mm electroporation cuvette, and add the mixture of
DC3000 cells and DNA. Electroporate the cells at room tem-
perature at 1.8 kV, 25 μF, 200 Ω. Electroporate an aliquot
with no plasmid DNA as a negative control.
16. Add 1 mL liquid KB, transfer the cells to a 14 mL culture tube,
and recover for 2 h at 30 °C with shaking at 250 rpm. Plate 50
and 150 μL each on separate KB agar plates with 10 μg/mL
gentamicin. Store the remainder of the electroporation mix at
4 °C in case it is required to replate higher cell volumes later.
17. Incubate plates for 3–4 days at 30 °C until well-spaced single
colonies are obtained.
18. Streak single colonies onto KB agar plates with 10 μg/mL
gentamicin and prepare frozen 15% glycerol stocks of the
DC3000 containing pCPP5371::EOI.

3.2  Plant Inoculation 1. Bring plants from the greenhouse or chamber 1 or 2 days prior
(See Notes 10 and 11) to plant inoculation and keep them on the laboratory bench.
We have used Nicotiana benthamiana, tobacco, and Arabidopsis
for inoculation of the Cya reporter strains. N. benthamiana
plants should be 4 to 6 weeks old, tobacco 5 to 8 weeks old,
and Arabidopsis 4 to 5 weeks old.
2. Start fresh plates for each DC3000 strain, pCPP5371::EOI,
pCPP5388 (positive control), and T3SS- pCPP5388 (negative
control) on KB agar with 10 μg/mL gentamicin and grow at
30 °C for 1–2 days.
3. From the primary plates use a wooden dowel to scoop up an
entire isolated colony, resuspend in 150–200 μL liquid KB,
and mix well by vortexing. Spot the entire volume onto a fresh
plate of KB agar with 10 μg/mL gentamicin. Swirl the plate
around to spread the bacteria on the entire surface of the plate.
Refresh all the strains similarly. Leave the plates open in the
laminar flow hood for about 5–10 min to dry and then incu-
bate at 30 °C overnight (up to 18 h).
480 Suma Chakravarthy et al.

4. Use an inoculating loop or pipette tip to harvest a pea-sized

scoop of bacterial cells from the plate and resuspend in 5 mL
10 mM MgCl2.
5. Use a spectrophotometer to measure the optical density at
600 nm (OD600) of the cell suspension. Prepare a 20-fold dilu-
tion of the resuspended cells by mixing 950 μL 10 mM MgCl2
with 50 μL culture. Measure the OD600 of the dilution and
calculate the OD of the culture by multiplying the observed
OD value by 20.
6. Adjust the resuspended cells to an OD600 of 0.05 with 10 mM
MgCl2 to a final volume of 10 mL for plant inoculation. Use
the following formula for the adjustment:
Volume  of  culture = (0.05/(observed OD600 × 20))∗10
Make up to 10 mL with 10 mM MgCl2.
In our spectrophotometer, this OD600 value corresponds to
5 × 107 CFU/mL. The exact correlation between OD600 and
CFU/mL will need to be optimized since different spectro-
photometers may vary. The important point to remember is
that plants should be inoculated with a culture adjusted to
5 × 107 CFU/mL.
7. At least three infiltrations should be done for each strain to
create biological replicates (EOI, positive and negative con-
trols). In N. benthamianaand tobacco, the different strains can
be inoculated on the same leaf as long as the inoculation zones
do not overlap. For N. benthamiana and tobacco, choose well-
expanded leaves, which are the fourth or fifth from the apex.
For Arabidopsis, choose leaves from the center triplet in the
whorl and inoculate two or three whole leaves per sample.
8. Using a 1 mL needleless syringe, inoculate plant leaves with
the different strains. Make sure to include the positive and
negative controls. A needle may be used to gently prick the leaf
and then inoculate the suspension. Pat dry the inoculation
zone with a Kimwipe or paper towel. For N. benthamiana and
tobacco, outline the inoculation zone with a wide-tip black
Sharpie to delineate the area. For Arabidopsis, mark the peti-
oles of inoculated leaves with the black Sharpie.
9. Leave the plants on the laboratory bench for 6 h (see Note 12).
10. Take 2.2 mL round-bottom microcentrifuge tubes and add
two copper BBs to each tube. Label the tubes with the name
of the different strains and the replicate numbers (e.g., EOI1,
EOI2, EOI3, and so on). Have these tubes and a dewar of liq-
uid nitrogen ready at the time of tissue harvesting.
11. After 6 h, use a biopsy punch or cork borer to excise ~1 cm2
leaf tissue from each infiltration zone. Eight leaf discs collected
with a 4 mm biopsy punch have a combined total area of
 Cya Reporter Assay 481

approximately 1 cm2 of leaf tissue. Working quickly, push a

long wooden dowel through the cork borer or biopsy punch to
eject the collected leaf discs into pre-labeled tubes. Close the
tubes and flash freeze in liquid nitrogen as soon as possible.

3.3  Direct Direct cAMP ELISA kit is used to quantitatively measure cAMP in
cAMP Assay the leaf samples. See the manufacturer’s instructions for additional
details on conducting the assay and data analysis.
1. Bring all cAMP ELISA buffers to room temperature before use
(see Note 13).
2. Vortex the frozen tubes rapidly so the copper BBs grind the
tissue into a fine powder. Proceed to perform the assay. Ground
tissue can also be frozen at −80 °C until use (see Note 14).
3. Add 300 μL 0.1 M HCl to the frozen tissue. Vortex vigorously
to mix. Tissue will turn brown in a few minutes. This is
normal.
4. Centrifuge at ≥12,000 relative centrifugal force (RCF) for
10 min. Resuspend the pellet by vortexing and physical agita-
tion and recentrifuge at ≥12,000 RCF for 10 min to create a
tight pellet. Transfer the supernatant to a fresh microcentri-
fuge tube, taking care to avoid any leaf debris (see Note 15).
5. Extracted supernatants of each sample need to be diluted so
that cAMP concentrations are within the range of the cAMP
standard curve. This must be empirically determined for each
EOI vector/host plant/bacterial concentration combination.
We have used 10-fold to 300-fold dilutions in different experi-
ments. More concentrated samples may be required to measure
lower levels of translocation. The 50-fold dilution described in
what follows has been used successfully in Arabidopsis.
6. Using a multichannel pipette and a 96-well plate, set up the
following dilution series:
Mix 50 μL extract supernatant with 200 μL 0.1 M HCl (1:5).
Mix 30 μL 1:5 diluted sample with 270 μL 0.1 M HCl (1:10)
(1:50 final).
7. Prepare cAMP standards. Diluted standards should be used
within 60 min of preparation. Glass or polypropylene, but not
polystyrene, tubes may be used with the standards.
8. Label five 12 × 75 mm Pyrex culture tubes #S1– #S5.
9. Pipet 900 μL of the 0.1 M HCl into tube #S1.
10. Pipet 750 μL of the 0.1 M HCl into tubes #S2– #S5.
11. Add 100 μL of the 2000 pmol/mL standard stock into tube
#S1. Vortex vigorously.
12. Add 250 μL of tube #S1 to tube #S2 and vortex vigorously.
482 Suma Chakravarthy et al.

13. Repeat step 12 for tubes #S2 to #S5 to prepare standard

curve. Standard concentrations for tubes #S1 to #S5 are as fol-
lows: 200 pmol/mL, 50 pmol/mL, 12.5 pmol/mL, 3.13
pmol/mL, and 0.78 pmol/mL.
14. Set up ELISA plate. Every well is set up in duplicate. Sixteen
wells will be used to generate assay standards (see Note 16).
The positive and negative control DC3000 strains will require
12 wells (2 strains with technical duplicates of 3 biological rep-
licates). Your EOI samples will require six wells apiece (techni-
cal duplicates of three biological replicates).
The following steps can be performed with a multichannel pipette.
15. Pipet 50 μL of neutralizing reagent into each well except the
blank wells.
16. Pipet 100 μL of the 0.1 M HCl into the B0 (0 pmol/mL stan-
dard) wells.
17. Add 150 μL of 0.1 M HCl to the nonspecific binding (NSB)
wells.
18. Pipet 100 μL of standards #S1 to #S5 to the bottom of the
appropriate wells.
19. Pipet 100 μL of the samples to the bottom of the appropriate
wells.
20. Pipet 50 μL of the blue conjugate into each well except the
blank wells.
21. Pipet 50 μL of the yellow antibody into each well except the
NSB and blank wells.
The blank wells should be empty, the NSB wells should be blue,
and all other wells should be green.
22. Seal the plate with sticker provided in kit and secure to an
orbital shaker with tape. Shake at approximately 100–200 rpm
for 2 h at room temperature.
23. During incubation, prepare the wash buffer and determine pro-
tein concentrations using a Bradford assay (see Subheading 3.4).
24. Wash buffer prep: Calculate the amount of wash buffer needed:
#wells*1.2 mL/well = total volume wash buffer. Prepare a
1:20 dilution of wash buffer concentrate in dH2O.
25. Wash the plate: Empty contents of the wells (into the sink) and
add 400 μL wash buffer to every well.
26. Repeat step 25 twice more for a total of three washes.
27. After the final wash, empty the wells and firmly tap plate on
Kimwipe to remove any remaining wash buffer.
28. Use a vacuum aspirator to remove the last traces of wash buffer.
29. Add 200 μL of the substrate solution to each well.
 Cya Reporter Assay 483

30. Incubate for 1 h at room temperature in the dark without

shaking. The wells with lower cAMP concentrations will
become more yellow.
31. Pipet 50 μL of the stop solution into each well.
32. Measure absorbance at 405 nm.

3.4  Bradford Assay All standards and samples should be run in duplicate. Sixteen wells
will be used for the standards. BSA concentration in the prediluted
standard kit: 125, 250, 500, 750, 1000, 1500, and 2000 μg/mL.
Use water for the 0 μg/mL “standard.”
These steps can be performed with a multichannel pipette.
1. In a 96-well plate add 99 μL water to each standard well and
80 μL water to each sample well.
2. Add 1 μL of the standards to the appropriate wells (1:100
dilution).
3. Add 20 μL of the 1:5 diluted samples to the appropriate wells
(1:25 final dilution).
4. Add 100 μL Bradford dye to each well, pipetting up and down
to mix.
5. Wells with higher protein concentrations should turn bluer.
6. Use a bubble breaker to remove any bubbles in the wells.
7. Incubate 10 min at room temperature.
8. Measure absorbance at 595 nm.

3.5  Data Analysis The principle behind the data analysis is as follows. (1) Generate
standard curves of pmol cAMP/mL and protein μg/mL concen-
tration. (2) Use the standard curves to calculate the pmol cAMP/
mL and protein μg/mL of your samples. (3) Divide the sample
cAMP value by the sample protein value to get pmol cAMP/μg
protein. This analysis can be conducted using Microsoft Excel.
Use the mean value of technical duplicate wells for all
calculations.
cAMP concentration calculation from 405 nm absorbance
values.
1. Subtract the blank well value from the NSB, B0, standard, and
sample values.
2. Subtract the blank-adjusted NSB value from the blank-adjusted
B0, standard, and sample values. These are the Net OD values.
3. Use the Net OD values to calculate the percentage of bound
cAMP for the standards. (S1/B0)*100, (S2/B0)*100….
4. Plot the percentage of bound (y) for the standards (linear scale)
vs. the pmol/mL values (x) of the standards (log scale). The
cAMP standard concentrations for S1–S5 are as follows:
484 Suma Chakravarthy et al.

200 pmol/mL, 50 pmol/mL, 12.5 pmol/mL, 3.125 pmol/mL,

and 0.781 pmol/mL.
5. Fit a logarithmic curve to the data points to determine
y = m*ln(x) + b.
6. Calculate the cAMP pmol/mL for each sample by solving for x.
x[sample cAMP pmol/mL] = e^((y[sample percent bound]
− b)/m).
7. Multiply the sample cAMP pmol/mL by the dilution factor (in
this protocol, 50) to get the cAMP pmol/mL of the undiluted
samples.
Protein concentration calculation from 595 nm absorbance values.
8. Plot the protein μg/mL (y) for the standards (linear scale) vs. the
A595 values (x) of the standards (linear scale). Protein standard
concentrations are 100-fold dilutions of the BSA standards P1–P5
= 1.25 μg/mL, 2.5 μg/mL, 5 μg/mL, 7.5 μg/mL, 10 μg/mL, 15 μg/
mL, and 20 μg/mL.
9. Fit a linear curve to the data points to determine y = m * x + b.
10. Calculate sample protein μg/mL by solving for y:
y[sample protein μg/mL] = (m*(x[A595])) + b.
11. Multiply the sample cAMP pmol/mL by the dilution factor of
25 to get the protein μg/mL of the undiluted sample.
Calculate sample pmol cAMP/μg protein
12. Divide the sample cAMP pmol/mL by the sample protein μg/
mL. The mL units cancel, resulting in pmol cAMP/μg protein
of the undiluted sample.
13. Determine the mean and standard deviation for the pmol
cAMP/μg protein values of the three biological repeats of each
sample. The T3SS- control strain should have a calculated
pmol cAMP/μg protein value of less than 1. The positive con-
trol strain will vary based on the host plant and experiment but
will likely be in the 10s to 100s range of pmol cAMP/μg pro-
tein (see Note 17).

4  Notes

1. pCPP5371 carries the ccdB toxin gene in the Gateway cassette

and must be maintained in a CcdB-resistant E. coli strain such
as DB3.1.
2. The hrp promoter is the avrPto promoter hrp box region.
3. For LM and KB liquid media do not add K2HPO4 prior to
autoclaving or it will precipitate. Mix up a 100× phosphate
stock, 75 g K2HPO4, in 500 mL dH2O, and filter sterilize. Add
to media after it cools.
 Cya Reporter Assay 485

4. Arabidopsis Col-0 is grown in an environmental growth cham-

ber at 23 °C constant temperature with a 14 hours of
light/10 hours of dark cycle and ~100 μmol light. Plants are
covered with humidity domes that are cracked open 2 cm and
benefit from growth with very limited airflow and 3–4 cm
between each plant. N. benthamiana and tobacco can be
chamber grown at 28 °C during the day and 23 °C at night on
a 12 L/12D cycle and ~200 μmol light or grown under green-
house conditions with 26 °C during the day and 22 °C at
night, 16 L/8D conditions. Nicotiana are fertilized every 2
weeks with 1 g/L of Peter’s 20:20:20 water-­soluble fertilizer.
5. Depending on how the Arabidopsis are raised, they may
wilt temporarily when infiltrated with 10 mM MgCl2. We have
used 0.25 mM MgCl2 regularly as a substitute, which does not
cause wilting.
6. Fine-point Sharpies should not be used as they will tear the
leaves.
7. Depending on your dilution scheme, the kit may not provide
sufficient 0.1 M HCl. You may mix your own to use in place of
what is provided in the kit.
8. The methods described here explain in detail the generation of
a reporter plasmid containing the EOI fused at the C-­terminus
with the Cya reporter gene. Expression of the EOI-Cya fusion
protein is driven by a hrp promoter, which is induced in plants.
9. Some long EOI genes may require additional sequencing to
obtain full coverage.
10. The effector AvrPto, encoded by P. syringae pv. tomato
DC3000, serves as a positive control for the assay because it
has been shown to be translocated to high levels, while a T3SS
mutant of DC3000, incapable of translocation, is used as the
negative control for the assay [10].
11. The assay is flexible and can be performed on different plants.
We have used Nicotiana benthamiana, Arabidopsis, and
tobacco plants for the assay. Other researchers have used
tomato plants using a similar protocol.
12. We sample at 6 h, as this time point precedes any dramatic
changes in bacterial population. Collection at later time points
may complicate analyses owing to an increased bacterial replica-
tion in some samples based on host compatibility or induction
of the hypersensitive response (HR) in incompatible hosts. We
would not recommend collecting samples from leaf tissue that
has undergone pathogen-induced cell death (disease or HR).
13. In multiplate versions of the kit, the blue conjugate should be
aliquoted.
14. The leaf discs frozen along with the copper BBs can be ground
to a powder by vigorous vortexing. It helps to first open the lid
486 Suma Chakravarthy et al.

Fig. 1 Nonlinear relationship between inoculum levels and cAMP accumulation in

Cya reporter-based T3SS translocation assays. Suspensions of Pto DC3000
transformed with avrPto1-cya-expressing plasmid pCPP5312 were infiltrated
into the panels of two tobacco leaves at concentrations ranging from around 2.6
× 108 to 2.7 × 106 CFU/mL. Two 1 cm diameter leaf discs were harvested 6 h
postinoculation per infiltration area to determine soluble pmol cAMP/μg protein.
The mean soluble protein concentrations of 88 processed discs per leaf were
used to calculate pmol cAMP/μg protein

of the microcentrifuge tube for a few seconds to release pres-

sure due to liquid nitrogen. Following this, vortex at full speed
for 15–20 s until the tissue is ground up. If large pieces of leaf
still remain, simply drop the tube back into liquid nitrogen and
proceed with other tubes, and complete the grinding after a
few minutes. This will prevent tissue thawing. Once the tissue
is ground up, add 0.1 M HCl and leave on the bench until all
the samples are processed. Proceed to the remaining steps once
all the samples have been resuspended in HCl.
15. HCl extracted supernatant can be refrozen for reanalysis,
although the measured cAMP values will decrease.
16. We do not run the Total Activity (TA) standard described in
the manufacturer’s protocol. The TA value is not used in later
calculations, and its omission allows all cAMP standards to be
run in two columns, simplifying plate setup design.
17. Special consideration should be taken when interpreting trans-
location values quantitatively. The pmol cAMP/μg protein val-
ues increase logarithmically with increasing bacterial
­concentrations. See Fig. 1, which is recalculated and plotted
from data collected and published in [16].
 Cya Reporter Assay 487

Acknowledgments

The authors would like to thank Dr. Lisa Schechter, Dr. Hai Li
Wei, Dr. Sebastien Cunnac, Dr. Alan Collmer, and Dr. Sheng Yang
He for significant contributions to the development and refine-
ment of the procedure described in this chapter. This work was
supported by National Science Foundation grant IOS-1025642
and the Gordon and Betty Moore Foundation (GBMF3037).

References
1. Cunnac S, Chakravarthy S, Kvitko BH, Russell 9. den Dulk-Ras A, Vergunst AC, Hooykaas PJ
AB, Martin GB, Collmer A (2011) Genetic dis- (2014) Cre reporter assay for translocation
assembly and combinatorial reassembly identify (CRAfT): a tool for the study of protein translo-
a minimal functional repertoire of type III cation into host cells. Methods Mol Biol 1197:
effectors in Pseudomonas syringae. Proc Natl 103–121
Acad Sci U S A 108:2975–2980 10. Schechter LM, Valenta JC, Schneider DJ,
2. Fouts DE, Badel JL, Ramos AR, Rapp RA, Collmer A, Sakk E (2012) Functional and
Collmer A (2003) A pseudomonas syringae pv. computational analysis of amino acid patterns
tomato DC3000 Hrp (Type III secretion) dele- predictive type III secretion system substrates
tion mutant expressing the Hrp system of bean in Pseudomonas syringae. PLoS One 7:e36038
pathogen P. syringae pv. syringae 61 retains nor- 11. Crabill E, Joe A, Block A, van Rooyen JM,
mal host specificity for tomato. Mol Plant- Alfano JR (2010) Plant immunity directly or
Microbe Interact 16:43–52 indirectly restricts the injection of type III
3. Alfano JR, Collmer A (2004) Type III secre- effectors by the Pseudomonas syringae type III
tion system effector proteins: double agents in secretion system. Plant Physiol 154:233–244
bacterial disease and plant defense. Annu Rev 12. Wei HL, Chakravarthy S, Worley JN, Collmer
Phytopathol 42:385–414 A (2013) Consequences of flagellin export
4. Wei CF, Kvitko BH, Shimizu R, Crabill E, through the type III secretion system of
Alfano JR, Lin NC, Martin GB, Huang HC, Pseudomonas syringae reveal a major difference
Collmer A (2007) A Pseudomonas syringae pv. in the innate immune systems of mammals and
tomato DC3000 mutant lacking the type III the model plant Nicotiana benthamiana. Cell
effector HopQ1-1 is able to cause disease in Microbiol 15:601–618
the model plant Nicotiana benthamiana. Plant 13. Oh HS, Kvitko BH, Morello JE, Collmer A
J 51:32–46 (2007) Pseudomonas syringae lytic transglycosyl-
5. Schechter LM, Vencato M, Jordan KL, ases coregulated with the type III secretion system
Schneider SE, Schneider DJ, Collmer A (2006) contribute to the translocation of effector pro-
Multiple approaches to a complete inventory teins into plant cells. J Bacteriol 189:8277–8289
of Pseudomonas syringae pv. tomato DC3000 14. Badel JL, Shimizu R, Oh HS, Collmer A
type III secretion system effector proteins. Mol (2006) A Pseudomonas syringae pv. Tomato
Plant-Microbe Interact 19:1180–1192 avrE1/hopM1 mutant is severely reduced in
6. Schechter LM, Roberts KA, Jamir Y, Alfano growth and lesion formation in tomato. Mol
JR, Collmer A (2004) Pseudomonas syringae Plant-Microbe Interact 19:99–111
type III secretion system targeting signals and 15. Lam HN, Chakravarthy S, Wei HL, BuiNguyen
novel effectors studied with a Cya translocation H, Stodghill PV, Collmer A, Swingle BM,
reporter. J Bacteriol 186:543–555 Cartinhour SW (2014) Global analysis of the
7. Garcia JT, Ferracci F, Jackson MW, Joseph SS, HrpL regulon in the plant pathogen
Pattis I, Plano LR, Fischer W, Plano GV (2006) Pseudomonas syringae pv. tomato DC3000
Measurement of effector protein injection by reveals new regulon members with diverse
type III and type IV secretion systems by using functions. PLoS One 9:e106115
a 13-residue phosphorylatable glycogen syn- 16. Kvitko BH, Ramos AR, Morello JE, Oh HS,
thase kinase tag. Infect Immun 74:5645–5657 Collmer A (2007) Identification of harpins in
8. Sory MP, Cornelis GR (1994) Translocation of Pseudomonas syringae pv. tomato DC3000,
a hybrid YopE-adenylate cyclase from Yersinia which are functionally similar to HrpK1 in pro-
enterocolitica into HeLa cells. Mol Microbiol moting translocation of type III secretion sys-
14:583–594 tem effectors. J Bacteriol 189:8059–8072
 Chapter 34

Monitoring Effector Translocation using the TEM-1

Beta-­Lactamase Reporter System
Julie Allombert, Anne Vianney, and Xavier Charpentier

Abstract
Among the bacterial secretion systems, the Type III, IV, and VI secretion systems enable bacteria to secrete
proteins directly into a target cell. This specific form of secretion, referred to as translocation, is essential
for a number of pathogens to alter or kill targeted cells. The translocated proteins, called effector proteins,
can directly interfere with the normal processes of the targeted cells, preventing elimination of pathogens
and promoting their multiplication. The function of effector proteins varies greatly depending on the
considered pathogen and the targeted cell. In addition, there is often no magic bullet, and the number of
effector proteins can range from a handful to hundreds, with, for instance, a substrate of over 300 effector
proteins of the Icm/Dot Type IV secretion system in the human pathogen Legionella pneumophila.
Identifying, detecting, and monitoring the translocation of each of the effector proteins represents an
active field of research and is key to understanding the bacterial molecular weaponry. Translational fusion
of an effector with a reporter protein of known activity remains the best method to monitor effector trans-
location. The development of a fluorescent substrate for the TEM-1 beta-lactamase has turned this
antibiotic-­resistant protein into a highly versatile reporter system for investigating protein transfer events
associated with microbial infection of host cells. Here we describe a simple protocol to assay the transloca-
tion of an effector protein by the Icm/Dot system of the human pathogen Legionella pneumophila.

Key words Effector protein, Type IV secretion system, β-lactamase fusion, CCF4, Fluorescence,
Legionella pneumophila

1  Introduction

Protein delivery from pathogen to host represents a major and uni-

fying theme in microbial pathogenesis. From the pathogen’s per-
spective, sending its own proteins to the target host cell represents
an efficient strategy for interfering with the host cellular functions,
preventing exposure to the host defense mechanism or even sub-
verting the cells for its benefit. There are, however, barriers pre-
venting the diffusion of proteins out of the pathogenic cell and
blocking the import of proteins to the host cell side. For instance,
in Gram-negative bacteria, a protein would have to get across three

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_34, © Springer Science+Business Media LLC 2017

489
490 Julie Allombert et al.

membranes; the inner and outer membranes of the bacteria and

the host cytoplasmic membrane.
Remarkably, bacteria have evolved several multisubunit molec-
ular machines to achieve the feat of translocating a protein from
the pathogen cytoplasm directly to the cytoplasm or the target cell.
The process is often referred to as translocation, and the trans-
located proteins, whose assumed function is to have an effect on
the cell functions, are called effectors. The multisubunit molecular
machines capable of translocating effectors include the Type III,
IV, and VI systems of Gram-negative bacteria and the Type VII
system of Mycobacteria [1]. The set of translocated effector mol-
ecules tends to be unique to each pathogen and reflects the unique
needs and specific niches of each bacterial species. A major chal-
lenge is to identify the substrates of these systems and track their
translocation in the host cell.
Several methods have been reported to specifically detect this
fraction of a bacterial protein that found its way to the host cell.
The first of them is the popular CyaA system involving transla-
tional fusions of effectors with the calmodulin-dependent catalytic
domain of the Bordetella pertusis toxin CyaA [2]. This enzyme
converts cellular adenosine triphosphate in cyclic adenosine mono-
phosphate (cAMP) in the presence of the eukaryotic protein
calmodulin. Levels of cAMP production can be subsequently
quantified. A less popular but clever method involves a transla-
tional fusion with the phosphorylable Elk peptide fused to the
nuclear localization signal (NLS) from the large T antigen of SV40
[3]. The NLS sequence directs the fusion protein to the cell nucleus
where the Elk tag is phosphorylated and can be detected with
phosphospecific Elk-peptide antibodies. Another method is based
on fractionation with digitonin that solubilise the eukaryotic
plasma membrane but not the prokaryotic membranes [4]. As for
the Cya and Elk-tag systems, these assays require disrupting the
eukaryotic cell and then performing an analysis. The beta-­lactamase
translocation assay was developed to overcome these limitations
and analyze translocation in living cells [5]. The beta-lactamase
translocation assay (see Fig. 1) takes advantage of the fluorescent
substrate CCF2-AM (or CCF4-AM) initially developed for the
detection of TEM-1 beta-lactamase activity within eukaryotic cells
[6]. The substrate consists of a coumarin and a fluorescein fluoro-
phore connected by a beta-lactam ring. Because of the fluoro-
phores’ spatial proximity, the fluorescence energy coming from the
excitation of the coumarin moiety is entirely transferred to the
fluorescein moiety, resulting in the emission of a green fluores-
cence. Enzymatic cleavage of the beta-lactam ring by the TEM-1
beta-lactamase frees the coumarin moiety, which, under excitation,
now emits a blue fluorescence. This shift in fluorescence can be
directly observed in the infected host cell with an epifluorescence
microscope or quantified with a spectrofluorometer. Translocation
in the eukaryotic host cytoplasm of an effector-TEM-1 fusion
 TEM-1 Beta-Lactamase Reporter System 491

Extracytosolic compartment
(media, vesicular compartement)

Bacterial cell

CCF2/AM
No fluorescence 'blaM-X TEM-X

520 nm 447 nm
FRET

409 nm 409 nm
Green Blue
fluorescence fluorescence

Intracellular compartment

Fig. 1 Schematic representation of TEM-1 reporter system to assess effector

protein translocation in live eukaryotic cells. Upon passive entry into eukaryotic
cell, the nonfluorescent esterified CCF2/AM (or CCF4/AM) substrate is rapidly
converted by cellular esterases into a charged and fluorescent CCF2. Excitation
of the coumarin moiety (circle) at 409 nm results in fluorescence energy transfer
(FRET) to the fluorescein moiety (hexagon), which emits a green fluorescence
signal at 520 nm. Injection of an effector fused to TEM-1 into a CCF2-loaded cell
induces catalytic cleavage of a CCF2 beta-lactam ring (square), disrupting
FRET. This produces an easily detectable and measurable change in fluores-
cence from green to blue emission

protein triggers a change in fluorescence of the host cell, which

makes the analysis of effector translocation rapid, easy, and reliable.
Because fewer than 100 molecules of TEM-1 can be readily
detected within a cell [6], the system was sensitive enough to
detect the translocation of a weakly produced fusion. Of note, the
TEM-1 enzyme is naturally secreted in the periplasmic space by the
Sec pathway. Thus, to use TEM-1 as a reporter of the ability of a
fused protein to drive it to another secretion system, it is necessary
to use a version deleted of its N-terminal secretion signal. Owing
to the properties of a secreted protein, TEM-1 can efficiently
unfold and refold and is highly permissive of protein fusion. This
likely makes it compatible with secretion by most secretion sys-
tems, and numerous studies have used it to demonstrate the trans-
location of effector proteins by the Type III, IV, and VI systems
(for a review, see [7]). The system has also been successfully used to
detect proteins secreted by the protozoan parasite Toxoplasma gon-
dii in its host [8]. More than just a convenient way to demonstrate
the translocation of an effector, the beta-lactamase translocation
reporter system can be used to monitor the kinetic parameters of
the translocation process [9, 10]. The assay can be miniaturized to
a 384-well format and is compatible with a high-throughput screening
to identify small molecules that could inhibit translocation and
492 Julie Allombert et al.

prevent infection [10, 11]. Translocated effector proteins can also

identify cells targeted by a pathogen in an infected host [12, 13].
Here we provide a protocol for testing the translocation of
effector proteins that are the substrate of the Icm/Dot Type IV
secretion system of the human pathogen Legionella pneumophila.
L. pneumophila infects alveolar macrophages in the human lungs,
and this cellular infection can be recapitulated in vitro using
monocyte-­ derived macrophages (THP-1, U937 cells). Upon
phagocytosis by macrophages, L. pneumophila delivers numerous
effector proteins in the host through its Icm/Dot Type IV secre-
tion system. Ectopically expressed as fusions with the mature form
of TEM-1 beta-lactamase, their translocation is detectable less than
an hour following infection. The beta-lactamase translocation assay
is particularly easy, straightforward, and quick. It requires only a
few pipetting steps and no sample processing. Typically, the results
of the assay are obtained about 3 h post infection. The protocol
provided here could easily be adapted to other pathogens, secre-
tion systems, and cellular infection models.

2  Materials

2.1  Bacterial Strains, 1. L. pneumophila strain (Paris, Lens, Philadelphia-1).

Beta-Lactamase 2. Plasmids for expression of beta-lactamase fusion proteins: The
Constructs, Host Cells plasmid pXDC61 and its derivative used in this protocol are
available from the nonprofit plasmid repository addgene (add-
gene.org, plasmids #21841, #21842, #21843, #21844) (see
Note 1 and Fig. 2).
3. U937 cell line ATCC number: CRL-1593.2™.

2.2  Legionella 1. ACES-buffered Yeast Extract (AYE) medium: For 1 L dissolve 12
pneumophila Media g yeast extract and 10 g N-(2-Acentamido)-2-aminoethanesulfonic
and Bacterial Growth acid (ACES), adjust pH to 6.9 with 1 M KOH. Add 10 mL cys-
teine 40 g/L and 10 mL iron pyrophosphate 30 g/L. Fill volume
to 1 L with distilled water and filter sterilize.
2. Charcoal yeast extract (CYE) plates: For 1 L dissolve 10 g yeast
extract and 10 g ACES, adjust pH to 6.9 with 1 M KOH, add
15 g agar and 2 g activated charcoal, and autoclave. Add 10
mL filter sterilized cysteine 40 g/L and 10 mL filter-sterilized
ferric nitrate 25 g/L. When appropriate add 5 μg/mL
chloramphenicol
­ and 1 mM isopropyl β-d-1-
thiogalactopyranoside (IPTG) (see Note 2).
3. Disposable 13 mL polypropylene snap-cap tubes, sterile.
4. 1.5 mL microcentrifuge tubes, sterile.
5. 30 °C incubator.
6. Orbital shaker, 30 °C.
7. Spectrophotometer and cuvettes.
 TEM-1 Beta-Lactamase Reporter System 493

Fig. 2 Map of plasmid pXDC61 to express beta-lactamase TEM-effector fusion

proteins. (a) The pXDC61 plasmid is a mobilizable (oriT) plasmid derived from the
broad host range plasmid RSF1010. The plasmid confers chloramphenicol resis-
tance. The 'blaM gene encodes the mature form of the TEM-1 beta-lactamase
deprived of its N-terminal secretion signal and is controlled by an IPTG-inducible
promoter. A polylinker (KpnI, SmaI, BamHI, XbaI) is placed at the 3′ end of the
'blaM gene. (b) Detail of polylinker for in-frame cloning of effector genes

2.3  Cell Culture 1. RPMI medium supplemented with 10% fetal bovine serum
and Differentiation (FBS) and glutamine (i.e., RPMI 1640 GlutaMAXTM, Gibco).
When appropriate add 5 μg/mL chloramphenicol and 1 mM
IPTG.
2. Phorbol 12-myristate 13-acetate (PMA): 0.1 M.
3. Culture flask, 25 cm2, sterile.
4. Disposable 15 mL polypropylene snap-cap tubes, sterile.
5. 96-well black polystyrene microplates with clear bottom, sterile.
6. CO2 incubator, 37 °C.
7. Malassez counting chamber.
494 Julie Allombert et al.

2.4  Translocation 1. LiveBLAzer-FRET B/G loading kit (Invitrogen). This kit

Assays includes the CCF4/AM substrate (see Note 3).
2. Probenecid stock solution: 0.1 M. Dissolve 1.25 g probenecid
(Sigma) in 22 mL 0.4 M NaOH by vigorous agitation. Add 22
mL 100 mM phosphate buffer, pH 8.0, and stir to dissolve the
precipitate that could form. Check pH and, if necessary, adjust
it to 8.0 with 1 M NaOH (if pH < 8) or HCl (if pH > 8).
Distribute in 1 mL aliquot and store at −20 °C.
3. RPMI medium.
4. Fluorescence microplate reader equipped with a dual mono-
chromator (e.g., Tecan Infinite M200) or with an excitation
filter at 405 nm and emission filters at 460 nm (blue fluores-
cence) and 530 nm (green fluorescence). Determine whether
the plate reader reads the plate from top or bottom.
5. Inverted fluorescence microscope equipped with a beta-­
lactamase filter set (Chroma Set # 41031; Excitation filter:
HQ405/20× (405 ± 10); dichroic mirror: 425 DCXR; emis-
sion filter: HQ435LP (435 long-pass)). Alternatively a
4′,6′-diamidino-2-phenylindole (DAPI) filter set (340 to
380 nm excitation and 425 nm long-pass emission) may be
used to observe the blue fluorescence, while the green fluores-
cence may be observed with a green fluorescent protein
(GFP)/fluorescein filter set.

3  Methods

3.1  Growth The infecting strain should be grown under conditions that have
of Infecting Legionella been previously determined to result in a successful infection. These
pneumophila Strains conditions may vary depending on the strain and species used but
should include chloramphenicol to maintain the plasmid and IPTG
to induce expression of the tested effector fusion proteins.
1. Streak L. pneumophila strains carrying pXDC61-derived plas-
mids from a frozen stock to CYE plates supplemented with
chloramphenicol and then incubated at 30 °C for 5 days.
2. With a sterile loop, scrape off a few colonies and transfer to a
1.5 mL microcentrifuge tube containing 1 mL sterile ultrapure
water. Resuspend bacteria by repeated pipetting.
3. Measure the optical density at 600 nm (OD600) of a tenfold
diluted bacterial suspension.
4. In a sterile 13 mL tube, inoculate 2 mL AYE supplemented
with chloramphenicol and IPTG with appropriate volume of
previous bacterial suspension to reach a starting OD600 of 0.3.
Incubate at 30 °C in orbital shaker for 3 days (see Note 4).
5. Validate the beta-lactamase fusions production by western blot
(see Note 5).
 TEM-1 Beta-Lactamase Reporter System 495

3.2  Maintenance U937 cells are monocytes that grow as a suspension and should be
and Differentiation maintained at a cell density of between 1.105 and 2.106 viable
of U937 Target Cells cells/mL.
1. Seed a 25 cm2 culture flask with U937 cells from a frozen stock
or from a previous culture flask in 10 mL RPMI medium sup-
plemented with glutamine and FBS. The culture flask is incu-
bated at 37 °C in a CO2 incubator for 5 days.
2. Determine the cellular concentration of the U937 cell culture
with a Malassez counting chamber.
3. Transfer 1.107 cells to a sterile 15 mL conical tube and centri-
fuge 5 min at 880 g.
4. Discard the supernatant and gently resuspend the pellet of cells
in 10 mL RPMI supplemented with glutamine and FBS (pre-
heated to 37°C) by slow repeated pipetting. This gives a cell
suspension of 1.106 cells/mL. Add 0.5 μL PMA.
5. Distribute 100 μL of cell suspension per well of a 96-well
microplate (105 cells/well). Leave three wells without U937
cells (medium alone). They will be used for the blank fluores-
cence measurement.
6. Incubate at 37 °C in CO2 incubator for 3 days. Following this
incubation, the previously spherical and nonadherent cells
should now be differentiated into macrophage-like cells that
adhere to the bottom of the well and display a spread-out
morphology.

3.3  Detection 1. Grow L. pneumophila strains as described in Subheading 3.1.

of Effector 2. Measure OD600 of liquid cultures of L. pneumophila strains.
Translocation Using Adjust to OD600 = 1 with sterile ultrapure water. This gives a
a Fluorescence Plate bacterial suspension at 109 bacteria/mL.
Reader 3. Add 200 μL of the resulting suspension to 800 μL RPMI sup-
plemented with glutamine, FBS, chloramphenicol, and
IPTG. Incubate these bacterial suspensions (2.108 bacteria/
mL) for 2 h at 37 °C in the CO2 incubator.
4. Add 10 μL of the bacterial suspension in wells of the 96-well
microplate containing differentiated U937 cells (Subheading
3.2). This gives a multiplicity of infection (ratio bacteria to dif-
ferentiated cell) of 20 (see Note 6). The bacterial suspension of
each tested L. pneumophila strain is added to three different wells.
Centrifuge the microplate for 10 min at 600 × g (see Note 7).
Incubate at 37 °C in a CO2 incubator for 1 h.
5. During the incubation time prepare the CCF4-AM loading
solution of the LiveBLAzer-FRET B/G loading kit.
Determined the number of tested wells, n. Mix n × 0.12 μL
CCF4-­AM 6X solution with n × 1.08 μL solution B. Vortex for
10 s. Add n × 15.8 μL solution C and n × 3 μL probenecid
496 Julie Allombert et al.

0.1 M (see  Note 8). Vortex for 10 s. This loading solution

should be kept away from strong light and is stable for 4 h at
room temperature.
6. Add 20 μL of the loading solution to each of the tested wells
of the 96-well microplate, including to the 3 wells that contain
media alone (see step 5 in Subheading 3.2). Incubate for 2 h in
the dark at room temperature.
7. If you have access to a fluorescence plate reader with bottom
read capabilities, go directly to step 9. If the microplate reader
is only equipped with a fluorescence top reading module, an
extra step is needed because the fluorescence signals are
quenched by the red solution C used in the loading CCF4
solution.
8. Following the 2 h CCF4 loading in the dark (step 6), discard
delicately the liquid contained in each well of the 96-well
microplate, including the 3 wells without cells. Replace with
50 μL RPMI at room temperature and without FBS (dispens-
ing medium with FBS tends to create bubbles that may inter-
fere with fluorescence measurements).
9. Put the plate (lid on) in a fluorescence plate reader to start the
measurements. Measure successively the blue fluorescence (Ex.
405 nm, Em. 460 nm) and green fluorescence (Ex. 405 nm,
Em. 530 nm). Both measurements should be performed on
blank wells (medium alone) in addition to the tested wells.
10. After collecting the raw data, perform a blank subtraction on
each fluorescence read. To evaluate the secretion efficiency of
the beta-lactamase fusions, divide the blank-subtracted blue flu-
orescence signal by the blank-subtracted green fluorescence sig-
nal (see Note 9). An example of expected results is shown in
Fig. 3.

3.4  Visualization Following fluorescence quantifications, the infected cells may also be
of Effector observed under a microscope to assess the percentage of translocation-­
Translocation Using positive (blue) and translocation-negative cells (green).
Fluorescence 1. Follow the protocol of Subheading 3.3 until step 8.
Microscope
2. Place the plate on an inverted microscope equipped with a 40×
or 60× objective.
3. Observe cells with the beta-lactamase filter set (Ex: 405 ± 10;
dichroic mirror: 425; Em: 435 long-pass). Using this filter set,
both green and blue cells can be visualized simultaneously.
4. Alternatively, blue cells can be visualized with a (DAPI) filter
set (340–380 nm excitation and 425 nm long-pass emission).
Green cells can be visualized with the filter set commonly used
for the visualization of GFP. Overexposure of the cells with this
filter set may bleach the fluorescein moiety of CCF2
 TEM-1 Beta-Lactamase Reporter System 497

Fig. 3 Typical results of a translocation assay. (a) Raw data, blank-subtracted fluorescence signals (relative
fluorescence unit, RFU), and fluorescence ratio (emission 460 nm/530 nm) obtained for secretion by wild-type
(WT) or ∆dotA L. pneumophila strain Lens of the known Dot/Icm effector LepA and the nonsecreted cytoplas-
mic protein FabI. Secretion assays were done in triplicate for each strain. Fluorescence measurements were
made with a Tecan M200 Infinite microplate reader equipped with a monochromator and a fluorescence top
reading module. The measurement program includes fluorescence readings with an excitation wavelength of
405 nm, emission wavelengths of 460 and 530 nm, and a gain set at 135. (b) Graphical representation of mean
emission ratio for LepA and FabI and corresponding standard deviations. Depending on the gain set for each
of the two fluorescence measurements, this ratio can change significantly. Data acquired under different gain
or in different plate reader can be normalized by setting the 460/530 ratio of the uninfected cells to 1. (c)
Fluorescence images of a typical translocation assay. Upon excitation at 405 nm, two images were captured
at 460 and 530 nm and merged

(or CCF4), which could result in the observation of blue fluo-

rescence even in the absence of effector translocation.
5. If the images of the blue and green cells are acquired sepa-
rately, the two images should be merged. Typical images are
shown in Fig. 3.

4  Notes

1. Beta-lactamase fusion plasmids are derived from the pXDC61

plasmid (Fig. 2) [14]. They are introduced into the L. pneu-
mophila by electroporation. These plasmids are constructed by
cloning the coding sequence of the candidate effector gene in
frame with the 'blaM gene encoding the mature form of the
TEM-1 beta-lactamase deprived of its N-terminal secretion
signal. The candidate gene is cloned downstream of the 'blaM
gene in order to leave intact the potential C-terminal secretion
signal of the candidate protein. The appropriate polylinker is
shown in Fig. 2b. If the nature of the secretion signal is
unknown, it is advisable to generate and test both fusion pro-
teins at the C- or the N-terminus of the TEM-1 beta-­lactamase.
498 Julie Allombert et al.

An NdeI site is available at the start codon of 'blaM. The

expression of these gene fusions is controlled by an IPTG-
inducible promoter.
2. Culture conditions must be optimized depending on the used
Legionella species and strains in order to obtain bacteria in a
virulent state (stationary phase). Here, we use CYE agar plates
and AYE liquid medium, but bacteria grown on buffered CYE
agar plates and LGM (Legionella growth medium) are equally
infective.
3. According to the supplier, “CCF2-AM and CCF4-AM differ
by two carbons in the bridge linking the coumarin moiety to
the lactam ring. Both are in the membrane-permeable, esteri-
fied forms, and can be used for assays in intact cells. CCF4-AM
has better solubility properties (soluble for >24 h) than
CCF2-AM and is thus best suited for screening applications. In
addition, CCF4-AM has slightly better FRET and thus slightly
lower background than CCF2-AM.” In our experience,
CCF2/AM and CCF4/AM perform equally well. We have not
found significant differences between the two compounds.
4. L. pneumophila grown on CYE agar plates may exhibit a het-
erogeneous population with a large part of filamentous bacte-
ria. Therefore, bacteria are grown in liquid cultures before host
­cell infection in order to work with a homogeneous and more
infective population.
5. It is advisable to assess the correct production and stability of
the beta-lactamase fusions. This can be done using conven-
tional western blot techniques, which will not be described
here. We recommend using the beta-lactamase monoclonal
antibody clone 8A5.A10, which is available from a variety of
suppliers.
6. The response of the beta-lactamase reporter should be deter-
mined as a function of multiplicity of infection (MOI). For L.
pneumophila and phagocytic cells, between an MOI of 1 and
10 the beta-lactamase system seems to behave linearly, and
above 25 bacteria/cell, the system appears to saturate [10].
Therefore, we use an MOI of 20 as a compromise between
sensitivity and linearity. Care should be taken to ensure that
the MOI is not too high, for instance, the TEM-FabI negative
control should not produce blue fluorescence. This situation
occurs in L. pneumophila when the MOI is over 50.
7. Motile L. pneumophila can make contact with host cells with-
out centrifugation. However, this experiment is based on the
fluorescence signals of a host cell monolayer at a specific time
point. Thus, the infection must be synchronized by centrifuga-
tion in order to work with a host cell monolayer that is homo-
geneously infected.
 TEM-1 Beta-Lactamase Reporter System 499

8. The addition of probenecid in the CCF4 loading solution is

required to inhibit organic anion transporter [15] and facili-
tates the loading of CCF4 by inhibiting efflux from cells.
9. Expected results are shown in Fig. 3. For Dot/Icm effector
proteins, you should see an increase of the blue fluorescence
and a decrease of the green fluorescence in comparison to the
nonsecreted protein (FabI) or in comparison with the TEM-­
effector fusion in the ∆dotA mutant impaired for its Icm/Dot
Type IV secretion system.

References
1. Costa TRD et al (2015) Secretion systems 8. Lodoen MB, Gerke C, Boothroyd JC (2010) A
in Gram-negative bacteria: structural and highly sensitive FRET-based approach reveals
mechanistic insights. Nat Rev Microbiol secretion of the actin-binding protein toxo-
13(6):343–359 filin during Toxoplasma gondii infection. Cell
2. Sory MP, Cornelis GR (1994) Translocation of Microbiol 12(1):55–66
a hybrid YopE-adenylate cyclase from Yersinia 9. Mills E, Baruch K, Charpentier X, Kobi S,
enterocolitica into HeLa cells. Mol Microbiol Rosenshine I (2008) Real-time analysis of
14(3):583–594 effector translocation by the type III secretion
3. Day JB, Ferracci F, Plano GV (2003) system of enteropathogenic Escherichia coli.
Translocation of YopE and YopN into eukary- Cell Host Microbe 3(2):104–113
otic cells by Yersinia pestis yopN, tyeA, sycN, 10. Charpentier X et al (2009) Chemical genetics
yscB and lcrG deletion mutants measured reveals bacterial and host cell functions critical
using a phosphorylatable peptide tag and for type IV effector translocation by Legionella
phosphospecific antibodies. Mol Microbiol pneumophila. PLoS Pathog 5(7):e1000501
47(3):807–823 11. Harmon DE, Davis AJ, Castillo C, Mecsas
4. Lee VT, Anderson DM, Schneewind O (1998) J (2010) Identification and characterization of
Targeting of Yersinia Yop proteins into the small-molecule inhibitors of Yop translocation
cytosol of HeLa cells: one-step translocation in Yersinia pseudotuberculosis. Antimicrob
of YopE across bacterial and eukaryotic mem- Agents Chemother 54(8):3241–3254
branes is dependent on SycE chaperone. Mol 12. Marketon MM, DePaolo RW, DeBord KL,
Microbiol 28(3):593–601 Jabri B, Schneewind O (2005) Plague bacteria
5. Charpentier X, Oswald E (2004) Identification target immune cells during infection. Science
of the secretion and translocation domain of 309(5741):1739–1741
the enteropathogenic and enterohemorrhagic 13. Geddes K, Cruz F, Heffron F (2007) Analysis
Escherichia coli effector Cif, using TEM-1 of cells targeted by Salmonella Type III secre-
beta-lactamase as a new fluorescence-based tion in vivo. PLoS Pathog 3(12):e196
reporter. J Bacteriol 186(16):5486–5495 14. de Felipe KS et al (2008) Legionella eukaryotic-­
6. Zlokarnik G et al (1998) Quantitation of tran- like type IV substrates interfere with organelle
scription and clonal selection of single living trafficking. PLoS Pathog 4(8):e1000117
cells with beta-lactamase as reporter. Science 15. Steinberg TH, Newman AS, Swanson JA,
279(5347):84–88 Silverstein SC (1987) Macrophages possess pro-
7. Pechous RD, Goldman WE (2015) benecid-inhibitable organic anion transporters
Illuminating targets of bacterial secretion. that remove fluorescent dyes from the cytoplas-
PLoS Pathog 11(8):e1004981 mic matrix. J Cell Biol 105(6 Pt 1):2695–2702
 Chapter 35

Effector Translocation Assay: Differential Solubilization

Irina S. Franco, Sara V. Pais, Nuno Charro, and Luís Jaime Mota

Abstract
The identification of effector proteins delivered into mammalian host cells by bacterial pathogens possessing
syringelike nanomachines is an important step toward understanding the mechanisms underlying the viru-
lence of these pathogens. In this chapter, we describe a method based on mammalian tissue culture infec-
tion models where incubation with a nonionic detergent (Triton X-100) enables solubilization of host cell
membranes but not of bacterial membranes. This allows the isolation of a Triton-soluble fraction lacking
bacteria but enriched in proteins present in the host cell cytoplasm and plasma membrane. Using appropriate
controls, this fraction can be probed by immunoblotting for the presence of bacterial effector proteins
delivered into host cells.

Key words Bacterial protein secretion system, Type III secretion, Effector, Translocation, Detergent
solubilization, SDS-PAGE, Immunoblotting

1  Introduction

Gram-negative bacteria possess different macromolecular structures,

known as type III, type IV, or type VI secretion systems, for the
delivery of effector proteins directly from the bacterial cytoplasm
into eukaryotic or prokaryotic host cells [1]. This protein delivery
or injection process is normally described as effector translocation.
Demonstrating that a particular bacterial effector protein is injected
into mammalian host cells during infection is not a trivial task, as
effectors are often delivered in minute amounts and can be short-
lived within the host cell. Various assays have been developed to
monitor effector translocation, for example, using Bordetella
pertussis calmodulin-dependent adenylate cyclase [2] or mature
TEM-1 β-lactamase [3] reporter assays (described in Chapters 33
and 34). Here we describe a method for assessing effector translo-
cation into mammalian cells by differential solubilization. It con-
sists in the infection of tissue culture cells by a bacterial pathogen,
followed by lysis of the infected mammalian cells using a detergent

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_35, © Springer Science+Business Media LLC 2017

501
502 Irina S. Franco et al.

that does not affect the integrity of bacterial membranes. The nonionic
detergents Triton X-100 [4] and digitonin [5] have been widely
used for this purpose, based on the inability of Triton X-100 to
solubilize outer membranes of Gram-negative bacteria (although it
can solubilize inner membranes) [6–8], and on the specificity of
digitonin for cholesterol-rich membranes [9]. Subsequent high-
speed centrifugation allows the separation of detergent-soluble
(supernatant) and insoluble components (pellet) of the lysate,
where the supernatant comprises cytoplasmic and plasma mem-
brane components (including delivered effector proteins) and the
pellet retains unbroken bacteria and nuclei that remained intact.
Analysis of these fractions by immunoblotting makes it possible to
confirm the presence of an effector protein of interest in the super-
natant fraction, which is taken as evidence of effector translocation.
A critical control involves probing for a bacterial protein that is not
delivered into host cells. This ensures that during experimental
manipulation there was no contamination of the supernatant frac-
tion with cytosolic bacterial proteins.
Differential solubilization can be applied to monitor effector
translocation by different Gram-negative bacteria and types of host
cells. Furthermore, if effector-specific antibodies are available, dif-
ferential solubilization can be used to monitor the translocation of
endogenously expressed and nonmodified effector proteins. This is
in contrast to other effector translocation assays that require the
modification of the gene encoding the effector to produce a protein
with an epitope tag or fused to a reporter protein, usually expressed
from a plasmid and often from an exogenous promoter. The dif-
ferential solubilization procedure using Triton X-100 is illustrated
by the two protocols detailed in what follows, consisting in the
monitoring of type III secretion (T3SS)-mediated translocation (1)
of the Yersinia enterocolitica effector YopE, expressed from its own
promoter using a nonmodified wild-type strain to infect RAW
264.7 murine macrophage-like cells (see Fig. 1), and (2) of the
Salmonella enterica serovar Typhimurium (S. Typhimurium)
effector SteA with a C-terminal double hemagglutinin epitope tag
(SteA-2HA), expressed from its own promoter but encoded in an
exogenous low-copy plasmid, during infection of HeLa cells
(see Fig. 2). While YopE is translocated by extracellular Yersinia
into host cells [2], SteA can also be translocated into the host cell
cytoplasm from intracellular Salmonella residing within a mem-
brane-bound vacuole [10], which further illustrates the versatility
of the procedure.
 Assessing Effector Translocation by Differential Solubilization 503

Fig. 1 Effector translocation by Yersinia enterocolitica during infection of RAW

264.7 macrophage-like cells. RAW 264.7 cells were infected by wild-type (wt) or
T3SS-defective (yscN mutant) Y. enterocolitica bacterial strains. Triton-soluble and
Triton-insoluble fractions from uninfected cells (UI) and from cells infected by wt or
yscN mutant (ΔYscN) bacteria were prepared as described in Subheadings 2.1
and 3.1. The sample fractions were analyzed by SDS-PAGE and immunoblotting as
described in Subheadings 2.2, 2.3, and 3.3. YopE is a Y. enterocolitica effector pro-
tein; SycO is a Y. enterocolitica T3S chaperone [15], used to control for possible
significant bacterial cross-contamination of the Triton-soluble fraction (and as a
loading control of the Triton-insoluble fraction); tubulin is a host cell protein used as
a loading control of the Triton-soluble fraction

Fig. 2 Effector translocation by Salmonella enterica serovar Typhimurium

(S. Typhimurium) during infection of HeLa cells. HeLa cells were infected by
S. Typhimurium steA mutant bearing a plasmid encoding C-terminal 2 × HA
epitope-­tagged wild-type SteA (SteAWT-2HA) or mutant SteA with lysine residue 36
replaced by alanine (SteAK36A-2HA). Triton-soluble and Triton-insoluble fractions
from cells infected by the two strains were prepared as described in Subheadings
2.1 and 3.2 (see Note 29). The samples were analyzed by SDS-­PAGE and immu-
noblotting as described in Subheadings 2.2, 2.3, and 3.3. SteA is a Salmonella effec-
tor protein [10, 14], DnaK is a bacterial molecular chaperone used to control for
possible significant bacterial cross-contamination of the Triton-soluble fraction
(and as a loading control of the Triton-insoluble fraction), and tubulin is a host cell
protein used as a loading control of the Triton-soluble fraction. Note the detection
of residual levels of tubulin in the Triton-insoluble fraction
504 Irina S. Franco et al.

2  Materials (See Note 1)

2.1  Cell Culture, 1. Cell lines: HeLa (clone HtTA-1) and RAW 264.7 cells
Infection, (European Collection of Authenticated Cell Cultures,
and Preparation of Cell ECACC).
Extracts 2. Bacterial strains and plasmids: Y. enterocolitica E40 (pYV40)
(wild-type) and Y. enterocolitica E40 (pMSL41) (yscNΔ169–177;
deficient in YscN ATPase that is essential for the activity of the
Yersinia T3SS) [11], S. Typhimurium steA mutant (an isogenic
derivative of S. Typhimurium strain NCTC 12023 [identical to
ATCC 14208s]) [12], carrying low-copy pWSK129-derived
plasmids (six to eight copies per cell) [13] expressing C-terminal
2 × HA epitope-tagged wild-type SteA (SteAWT-2HA) or mutant
SteA with lysine residue 36 replaced by alanine (SteAK36A-2HA)
under the control of the steA promoter [10, 14].
3. Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented
with 10% (v/v) heat-inactivated fetal bovine serum (FBS)
(DMEM + FBS), stored at 4 °C. Commercial 500 mL bottles
of heat-inactivated FBS are stored at −20 °C. The FBS is
thawed by incubation at 4 °C for 48 h, followed by prepara-
tion of aliquots in 50 mL tubes stored at −20 °C. To prepare
DMEM + FBS, the aliquots are thawed in a 37 °C water bath
and added to a commercial 500 mL bottle of DMEM. Do not
add antibiotics to cell culture medium.
4. Earle’s Buffered Salt Solution pH 7.4 (EBSS). Store at room
temperature.
5. TrypLE™ Express (Thermo Fisher Scientific). Store at room
temperature.
6. Phosphate-buffered saline (PBS 1×): 137 mM NaCl, 2.7 mM
KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, pH 7.4. Store at
room temperature. Prepared by diluting a stock of commercial
PBS 10× in double-distilled water (ddH2O), followed by ster-
ilization by autoclaving.
7. Nalidixic acid 3.5 mg/mL: Dissolve appropriate amount in
0.1 M NaOH and filter (0.22 μm) sterilize. Store at −20 °C.
Keep working aliquots at 4 °C.
8. Kanamycin 50 mg/mL: Dissolve appropriate amount in
ddH2O and filter (0.22 μm) sterilize. Store at −20 °C. Keep
working aliquots at 4 °C.
9. Lysogeny broth (LB) medium: Dissolve appropriate amount
of LB powder in ddH2O and sterilize by autoclaving. Store at
room temperature. Freshly supplemented with kanamycin
(to 50 μg/mL) to grow S. Typhimurium strains.
10. LB agar: Dissolve appropriate amount of LB powder in

ddH2O, add agar to 1.6% (w/v), and sterilize by autoclaving
 Assessing Effector Translocation by Differential Solubilization 505

(store at room temperature). Allow to cool to 55 °C, and add

adequate amounts of nalidixic acid (to 35 μg/mL) or kanamy-
cin (to 50 μg/mL) to grow Y. enterocolitica or S. Typhimurium
strains, respectively. The plates can be stored at 4 °C for up to
2 months.

11. Brain heart infusion (BHI) medium: Dissolve appropriate
amount of BHI powder in ddH2O and sterilize by autoclaving.
Store at room temperature. Freshly supplemented with nali-
dixic acid (to 35 μg/mL) to grow Y. enterocolitica strains.
12. Triton X-100, stock solution at 10% (v/v) in PBS 1× (stored
at 4 °C): Incubate Triton X-100 at 37 °C for 30 min, within
the biological safety cabinet measure an adequate volume of
Triton X-100, and add it to appropriate volume of sterile PBS
1× (e.g., 5 mL Triton X-100 to 45 mL PBS 1× in a 50 mL
tube), mix well, and incubate 30 min at 37 °C.
13. Gentamicin 10 mg/mL. Store at 4 °C.
14. Protease inhibitor cocktail. Store at −20 °C.
15. CO2 incubator, microbiology incubators, class II biological
safety cabinet, water bath, shaking water bath with adjustable
temperature, mini centrifuges.

2.2  Sodium Dodecyl 1. 1.5 M Tris–HCl, pH 8.8: Dissolve an appropriate amount of

Sulfate Polyacrylamide Tris base in ddH2O, adjust pH to 8.8 with HCl, adjust to
Gel Electrophoresis desired volume with ddH2O, and sterilize by autoclaving.
Store at room temperature.
2. 1.0 M Tris–HCl, pH 6.8: Dissolve an appropriate amount of
Tris base in ddH2O, adjust pH to 8.8 with HCl, adjust to
desired volume with ddH2O, and sterilize by autoclaving.
Store at room temperature.
3. Acrylamide/bis-acrylamide (37.5:1 solution). Store at 4 °C.
4. Sodium dodecyl sulfate (SDS) 20% (w/v): Dissolve an appro-
priate amount of SDS in ddH2O. It is not required to sterilize
the solution. Store at room temperature.
5. Ammonium persulfate (APS) 10% (w/v). Store at 4 °C
(see Note 2).
6. N, N, N, N′-tetramethyl-ethylenediamine (TEMED). Store
at 4 °C.
7. 12% SDS polyacrylamide gel electrophoresis (PAGE) gels (for
two mini gels): Prepare resolving gel: 6.5 mL H2O, 4.5 mL
acrylamide/bis-acrylamide (37.5:1 solution), 3.8 mL 1.5 M
Tris–HCl, pH 8.8, 75 μL 20% (w/v) SDS, 150 μL 10% (w/v)
APS, 6 μL TEMED (see Note 3). After polymerization, prepare
stacking gel: 7.34 mL H2O, 1.25 mL acrylamide/bis-­acrylamide
(37.5:1 solution), 1.25 mL 1 M Tris–HCl, pH 6.8, 50 μL
506 Irina S. Franco et al.

SDS 20% (w/v), 100 μL APS 10% (w/v), 10 μL TEMED

(see Note 4).
8. Protein molecular weight marker. Store at −20 °C (see Note 5).
9. Tris-glycine buffer: 0.025 M Tris, 192 mM glycine, 0.1%
(w/v) SDS. Prepare a 10× Tris-glycine stock solution without
SDS (0.25 M Tris and 1.92 M glycine, using adequate amount
of Tris base, glycine, and ddH2O). Store at room temperature.
This stock solution is used to prepare the Tris-glycine running
buffer, using adequate amounts of ddH2O and 20% (w/v)
SDS. Store at room temperature.
10. SDS-PAGE loading buffer 5×: 0.25 M Tris–HCl, pH 6.8, 10%
SDS (w/v), 50% (v/v) glycerol, 0.5 M β-mercaptoethanol,
0.5% (w/v) bromophenol blue. Store at −20 °C.
11. SDS-PAGE mini-gel caster and migration apparatus.

2.3  Immunoblotting 1. Transfer buffer: 0.025 M Tris, 192 mM glycine, and 20%
(v/v) methanol. Store at room temperature.
2. PBS 10×: 1.37 M NaCl, 0.027 M KCl, 0.1 M Na2HPO4,
0.02 M KH2PO4. Weigh adequate amounts of each of the
reagents, dissolve in ddH2O, adjust to final volume, and steril-
ize by autoclaving. Store at room temperature.
3. Washing solution (PBST): PBS 1× containing 0.2% (v/v)
Tween 20. Store at room temperature.
4. Blocking solution: PBST containing 4% (w/v) skim milk
powder: Dissolve the appropriate amount of skim milk pow-
der in PBST (see Note 6). Prepare fresh and store at 4 °C for
up to 2 days.
5. Stripping buffer: 25 mM glycine, pH 2, 1% (w/v) SDS:
Dissolve an adequate amount of glycine in ddH2O, adjust pH
to 2 with HCl, add 20% (w/v) SDS to a final concentration of
1% (w/v), and adjust to desired volume using ddH2O. Store
at room temperature.
6. Nitrocellulose membranes, 0.2 μm pore size (see Note 7).
7. Ponceau solution: 0.1% (w/v) Ponceau solution in 0.5% (v/v)
acetic acid: Dissolve Ponceau S in H2O and glacial acetic acid.
8. Whatman paper.
9. Autoradiography films.
10. Primary antibodies (all stored at −20 °C): Mouse monoclonal
anti-DnaK (clone 8E2/2; Millipore; used at 1:5000); rat
monoclonal anti-HA (clone 3F10; Roche; used at 1:1000),
mouse monoclonal anti-TEM-1 (QED Bioscience; used at
1:500); mouse monoclonal anti-α-tubulin (clone B-5-1-2;
Sigma-Aldrich; used at 1:1000); rabbit polyclonal anti-SycO
([15]; used at 1:500); rabbit polyclonal anti-YopE ([16];
used at 1:1000).
 Assessing Effector Translocation by Differential Solubilization 507

11. Secondary antibodies: Mouse and rabbit horseradish peroxidase

(HRP)-conjugated secondary antibodies (used at 1:10,000).
Store working aliquots at 4 °C and stocks at −20 °C.
12. Immunodetection kit such as Western Lightning Plus-ECL
(Perkin Elmer) or similar reagent.
13. Gel transfer apparatus.
14. Gel imaging apparatus.

3  Methods

3.1  Infection of RAW 1. RAW 264.7 cells are maintained in DMEM + FBS (with no
264.7 Cells by Y. antibiotics) at 37 °C in a humidified atmosphere with 5% (v/v)
enterocolitica CO2. The cells are used for up to 15–20 passages. The cells are
and Preparation routinely tested for mycoplasma contamination, using
of Triton-Soluble Venor®GeM Advance (Minerva Biolabs GmbH).
and Triton-Insoluble 2. The day before the infection prepare RAW 264.7 cells and
Fractions grow Y. enterocolitica strains: (1) seed RAW 264.7 cells at a
density of 1 × 106 cells per well in six-well tissue culture plates;
and (2) grow Y. enterocolitica in 5 mL BHI, overnight at 26 °C,
with continuous shaking (130 rpm).
3. Dilute the bacterial cultures grown overnight to an optical
density at 600 nm (OD600) of 0.2 in fresh BHI and resume
growth at 26 °C with continuous shaking (130 rpm) for 2 h
(see Note 8).
4. To induce expression of the Yersinia T3SS genes, quickly shift
the bacterial cultures to a shaking water bath (130 rpm) at
37 °C and incubate for an additional 30 min (see Note 9).
5. Centrifuge 1.5 mL of the bacterial culture (17,000 × g, 1 min;
see  Note 10) and resuspend the bacterial pellet in 1 mL
DMEM + FBS and measure the OD600.
6. Calculate the volume of the bacterial suspension that needs to
be added to the RAW 264.7 cells to have a multiplicity of
infection (MOI) of 50, i.e., 5 × 107 bacteria per well
(see Note 11).
7. Add the calculated volume to the seeded RAW 264.7 cells.
Carefully swirl the plates to obtain an even infection.
8. Incubate the infected cells for 3 h at 37 °C in a humidified
atmosphere of 5% (v/v) CO2.
9. After 3 h of infection, replace the medium of the infected cells
by DMEM + FBS (previously warmed at 37 °C) containing
50 μg/mL gentamicin to kill extracellular bacteria (50 μg/mL)
and incubate for an additional 2 h at 37 °C in a humidified
atmosphere of 5% (v/v) CO2.
508 Irina S. Franco et al.

10. From this point, all manipulation should be done on ice and
using ice-cold solutions.
11. Wash infected cells twice with ice-cold 1× PBS.
12. Add 250 μL 1× PBS containing 0.1% (v/v) Triton X-100 and
a protease inhibitor cocktail (see Notes 12 and 13).
13. Incubate cells for 10 min on ice.
14. To remove cells from the wells, pipet up and down several
times (about 15–20 times) and transfer cells to a 1.5 mL tube.
15. Centrifuge samples at 17,000 × g for 15 min (see Note 10) at
4 °C. Remove the top 200 μL of supernatant and repeat this
centrifugation step (see Note 14). Recover the top 100 μL of
this second centrifugation step and add 25 μL 5× SDS-PAGE
loading buffer (this is the Triton-soluble fraction).
16. Remove all supernatant from the pellet of the first centrifuga-
tion and resuspend it in 200 μL 1× SDS-PAGE loading buffer
(this is the Triton-insoluble fraction).
17. Incubate samples for 10 min at 95–100 °C.
18. Use immediately 30 μL of the Triton-soluble fraction and
20 μL of the Triton-insoluble fraction for immunoblotting
(see subsequent discussion) or keep samples at −20 °C or
−80 °C until use.

3.2  Infection 1. HeLa cells are maintained in DMEM + FBS (with no antibiotics)
of HeLa Cells at 37 °C in a humidified atmosphere with 5% (v/v) CO2. The
by S. Typhimurium cells are used for up to 15–20 passages. The cells are routinely
and Preparation tested for mycoplasma contamination using Venor®GeM
of Triton-Soluble Advance (Minerva Biolabs GmbH).
and Triton-Insoluble 2. The day before the infection prepare HeLa cells and grow
Fractions S. Typhimurium strains: (1) seed HeLa cells at a density of
2.5 × 105 cells per well in six-well tissue culture plates; and (2)
grow S. Typhimurium in 5 mL LB overnight at 37 °C, with
continuous shaking (130 rpm) (see Note 15).
3. Dilute 1:33 the bacterial cultures grown overnight in 5 mL
fresh LB medium and grow the bacterial culture for 3 h 30 min
at 37 °C with continuous shaking (130 rpm) (see Note 16).
4. 5–10 min before the bacterial incubation has ended (step 3),
wash once the seeded HeLa cells with previously warmed
EBSS and incubate for 15–20 min at 37 °C in a humidified
atmosphere of 5% (v/v) CO2.
5. Measure OD600 of the bacterial culture.
6. Dilute the bacterial culture in 5 mL EBSS (previously warmed at
37 °C) to have a MOI of 100 (i.e., to 1.25 × 106 bacteria/mL)
when adding 2 mL of this suspension to the seeded HeLa cells
(see Note 17).
 Assessing Effector Translocation by Differential Solubilization 509

7. Remove EBSS and add 2 mL bacterial suspension to monolayer

of HeLa cells. This corresponds to the beginning of the infec-
tion (time zero).
8. Incubate the infected cells for 15 min at 37 °C in a humidified
atmosphere of 5% (v/v) CO2.
9. Wash the infected cells three times with DMEM + FBS (previ-
ously warmed at 37 °C) containing 100 μg/mL gentamicin
(added fresh from the gentamicin 10 mg/mL stock solution
just before the washing steps).
10. Incubate the infected cells in DMEM + FBS containing 100
μg/mL gentamicin for 1 h at 37 °C in a humidified atmosphere
of 5% (v/v) CO2.
11. Replace the medium of the infected cells by DMEM + FBS
(previously warmed at 37 °C) containing 16 μg/mL genta-
micin (added fresh from the gentamicin 10 mg/mL stock
solution just before the washing steps).
12. Incubate the infected cells for a total of 14 h of infection,
using as reference the time zero of infection (see step 7 of
Subheading 3.2) (see Note 18).
13. Wash HeLa cells with 1× PBS.
14. Add 250 μL TrypLE Express to the HeLa cell monolayer and
incubate the cells for 5 min at 37 °C.
15. Add 1 mL DMEM + FBS and collect the cells into a 1.5 mL
tube by extensively pipetting up and down (15–20 times).
16. Centrifuge at 17,000 × g for 1 min (see Note 10), discard
supernatant, and wash cells in 1 mL ice-cold 1× PBS.
17. Repeat centrifugation and washing step (step 16) (see Note 19).
18. From this point, all manipulation should be done on ice and
using ice-cold solutions.
19. Resuspend the pellet in 100 μL ice-cold 1× PBS containing
0.1% (v/v) Triton X-100 and a protease inhibitor cocktail.
20. Incubate for 10 min on ice with occasional homogenization.
21. Centrifuge cell lysates at 18,620 × g for 15 min at 4 °C

(see Note 10) to separate the Triton-soluble from Triton-­insoluble
fraction as described in steps 15–17 of Subheading 3.1.
22. Centrifuge samples at 18,620 × g for 15 min (see Note 10) at
4 °C. Remove the top 80 μL of supernatant and repeat this
centrifugation step (see Note 14). Recover the top 40 μL of
this second centrifugation step and add 10 μL 5× SDS-PAGE
loading buffer (this is the Triton-soluble fraction).
23. Remove all supernatant from the pellet of the first centrifuga-
tion and resuspend it in 100 μL 1× SDS-PAGE loading buffer
(this is the Triton-insoluble fraction).
510 Irina S. Franco et al.

24. Incubate samples for 10 min at 95–100 °C.

25. Use immediately 30 μL of the Triton-soluble fraction and 20 μL
of the Triton-insoluble fraction for immunoblotting (see
following discussion) or keep samples at −20 °C or −80 °C
until use.

3.3  SDS-PAGE 1. Load samples on separate wells of a 12% SDS-PAGE (see Notes 3,

and Immunoblotting 20, and 21).
2. Run the SDS-PAGE for 70 min at 150 V (see Note 22).
3. Process the SDS-PAGE for transfer into nitrocellulose mem-
branes (see Note 23).
4. After transfer, evaluate the efficiency of protein transfer by
staining the membrane(s) with a 0.1% (w/v) Ponceau solu-
tion: Sink the membrane in a few milliliters of the Ponceau
solution and incubate with gentle shaking for 1–5 min, destain
with distilled H2O until protein bands are visible, and use a
pen or a pencil to label the bands of the molecular weight
marks and the position of lanes. If appropriate, cut the blot-
ting membrane in strips to be detected against specific primary
antibodies.
5. Using a flat-bottom incubation vessel (e.g., a Petri dish),
incubate the membrane(s) in blocking solution for at least 1 h
at room temperature, with gentle rocking (see Note 24).
6. Dilute primary antibody in the blocking solution and incubate
for at least 1 h at room temperature, with gentle rocking
(see Note 25).
7. Remove the primary antibody solution and store it at −20 °C
(see Note 26).
8. Add an excess volume of PBST and rinse the membrane(s) by
gentle swirling of the immunoblotting incubation vessel.
Discard the PBST solution.
9. Add an excess volume of PBST and incubate the membrane(s)
for 10 min at room temperature, with gentle rocking. Repeat
twice (see Note 27).
10. Discard the PBST solution and incubate the membrane(s) for
1 h with appropriate HRP-conjugated secondary antibodies
diluted in blocking solution.
11. Discard the secondary antibody solution and wash the membrane
as indicated earlier in steps 8 and 9.
12. Perform immunoblot detection using ECL detection system
and acquire the final image using an imaging system or by expo-
sure to ECL autoradiography films followed by processing in a
dark room using photography developer and fixer solutions
(see Note 28).
 Assessing Effector Translocation by Differential Solubilization 511

13. If a membrane must be reprobed with other primary antibodies,

wash the membrane in PBST (as described earlier in steps 8
and 9) and incubate it in an excess volume of stripping solu-
tion for 20 min at room temperature, with gentle rocking.
14. Wash with PBST (as described earlier in steps 8 and 9).
15. Proceed with the immunoblotting procedure, restarting from
the earlier step 5.

4  Notes

1. Prepare all solutions using ddH2O, unless otherwise indicated,

and analytical-grade reagents. Prepare all reagents at room
temperature and store them at the indicated temperatures.
Follow regulations and guidelines for the manipulation and
disposal of chemicals, mammalian cell cultures, and biosafety
level class II organisms. All solutions and materials used for the
manipulation of mammalian cell cultures must be sterile and
manipulated only within a biological safety cabinet, and those
used for bacterial cultures must also be sterile and manipulated
by aseptic techniques.
2. The recommended procedure in classical molecular biology
laboratory textbooks (e.g., Sambrook et al., Molecular
Cloning: A Laboratory Manual) is that the 10% (w/v) APS
solution should be prepared fresh. We normally prepare a 10
mL stock solution of 10% (w/v) APS that we store at 4 °C and
use within several weeks with no noticeable effect in the per-
formance of the SDS-PAGE.
3. The description is for 12% SDS-PAGE, but the concentration
of the resolving gel can be adjusted according to the molecular
mass of the proteins being analyzed by recalculating the vol-
umes of acrylamide/bis-acrylamide and ddH2O.
4. To facilitate the visualization of the wells, we normally add
~50 μL of a 2% (w/v) solution of Orange G (Sigma-Aldrich)
to the stacking gel.
5. It is convenient to follow running of the SDS-PAGE and to
label the bands in the nitrocellulose membrane with colored
protein markers (see step 4 in Subheading 3.3), but other
types of markers can be used.
6. Other blocking agents (e.g., bovine serum albumin (BSA)
or fish gelatin) can be used, but skim milk powder works
very well.
7. Polyvinylidene difluoride (PVDF) membranes can also be used,
but they need to be soaked in methanol for 15–30 s prior to use
in immunoblotting.
512 Irina S. Franco et al.

8. The bacterial growth and infection conditions described are

for Y. enterocolitica and need to be adapted to each particular
bacterial species.
9. It is critical that the shift be done quickly and that a water bath
be used to incubate the bacterial cultures at 37 °C.
10. Basically top speed in a microcentrifuge for 1 min; this is the g
force at maximum rotations per minute in the microcentrifuge
we normally use for this.
11. To accurately calculate the MOI for infection, it is necessary to
establish the relation between the colony forming units
(CFU)/mL of a culture grown in liquid medium and its cor-
responding OD600. For Y. enterocolitica cultures we consider
that an OD600 of 1 corresponds to 5 × 108 CFU/mL.
12. Optimal conditions to detect effector translocation for each
particular experiment might have to be determined empirically,
such as by the duration of the infection or the solubilization
agent used. In our hands, 0.1% (w/v) Triton-X100 works well
for monitoring effector translocation by Y. enterocolitica or
S. Typhimurium into tissue culture cells. Other commonly
used detergents in this type of assay include 0.2% (w/v)
saponin [17] and 0.02% (w/v) digitonin [5, 18].
13. The infected cells can also be recovered as described for the
S. Typhimurium infection of HeLa cells (steps 13–20 in
Subheading 3.2).
14. This is a critical step, and disturbing the pellet when recover-
ing the supernatant must be avoided. The second centrifuga-
tion step described is designed to circumvent this problem.
15. The bacterial growth and infection conditions described are
for S. Typhimurium and need to be adapted to each particular
bacterial species.
16. These are incubation conditions that induce expression of the
genes encoding the Salmonella pathogenicity island-1-­encoded
T3SS (SPI-1 T3SS) whose effectors promote invasion of HeLa
cells. If infecting macrophages, the overnight culture of
S.  Typhimurium can be opsonized (or not) and used directly
in macrophage infection [19].
17. See Note 11. For S. Typhimurium cultures we consider that an
OD600 of 1 corresponds to 1 × 109 CFU/mL. Make sure to
mix extremely well (vortex and invert the tube 10–15 times)
the bacterial suspension.
18. These infection conditions are to monitor the translocation of
S. Typhimurium effector proteins by the SPI-2 T3SS, as SteA
in this protocol [10, 14], which is induced within the vacuole
where Salmonella resides within host cells. Translocation of
SPI-2 effector proteins normally can be detected 6–8 h after
 Assessing Effector Translocation by Differential Solubilization 513

bacterial inoculation of the tissue culture cells. Incubating the

HeLa cells with S. Typhimurium for 14 h is convenient because
the infection can be done late in the afternoon and the samples
collected early in the morning.
19. This procedure makes it possible to obtain a pellet of infected
cells and to concentrate the protein extract by adjusting the
volumes of 1× PBS containing 0.1% (w/v) Triton X-100 used
to lyse the infected cells.
20. Several controls should be used to rule out possible cross-­
contamination of the obtained fractions. To rule out the possi-
bility of bacterial lysis and consequent release of bacterial
components into the Triton-soluble fraction, the blots should
be probed with an antibody against a bacterial nontranslocated
protein (see Figs. 1 and 2). Additionally, the presence of a host
cell cytosolic protein in the Triton-soluble fraction (e.g., tubulin)
can be confirmed (see Figs. 1 and 2). Demonstration of translo-
cation of an effector by a particular secretion system is normally
done by using a secretion-defective strain (see Fig. 1).
21. If the proteins to be detected have significantly different

molecular weights, the blotting membrane can be cut in strips
to incubate each one separately with the appropriate primary
antibody. If the bands of the proteins are expected to be too
proximal, the membrane must be stripped and reprobed.
22. The running conditions indicated can be adjusted accordingly,
depending on the molecular mass of the proteins that need to
be analyzed, and the running time of the SDS-PAGE can be
controlled visually based on the migration of the bromophe-
nol blue dye or of the prestained protein markers.
23. Semi-dry or wet electroblotting apparatus can be used. Wet
electroblotting transfer is known to facilitate transfer of pro-
teins with a molecular mass >100 kDa.
24. The blocking step can also be done overnight at 4 °C or after
blocking for 1 h at room temperature; the membranes can be
kept at 4 °C overnight or for 2 or 3 days. Use an excess vol-
ume of blocking to ensure that the membranes are fully cov-
ered at all times.
25. The incubation with the primary antibody can also be done
overnight at 4 °C. Exact conditions should be optimized for
each specific antibody. The volume of the antibody solution used
must ensure that the membranes are fully covered at all times,
but the volume should be minimized because antibodies are
usually expensive or scarce. For a small membrane, 5 mL of
antibody solution in a 90 mm Petri dish is usually enough.
26. The antibodies diluted in blocking solution can normally be
reused at least five or six times. If a decrease in performance is
514 Irina S. Franco et al.

noticed, the procedure can be repeated with a fresh dilution of

the antibody.
27. If appropriate, the membranes can be left for longer in PBST
(at least 1–2 h).
28. Other detection reagents or image acquisition systems can be
used.
29. The use of an uninfected control is recommended (see Fig. 1)
but can be dispensable if the primary antibody used is known
not to originate a background signal in immunoblotting (as
was the case with the anti-HA antibody used in the experiment
illustrated in Fig. 2).

Acknowledgments

This work was supported by the Unidade de Ciências Biomoleculares

Aplicadas—UCIBIO, which is financed by national funds from
Fundação para a Ciência e a Tecnologia (FCT) (UID/Multi/
04378/2013) and cofinanced by the ERDF under the PT2020
Partnership Agreement (POCI-01-0145-­FEDER-007728) and by
FCT research grants PTDC/BIA-­MIC/2821/2012 and PTDC/
BIA-MIC/116780/2010. Irina Franco is recipient of apostdoc-
toral fellowship (SFRH/BPD/102378/2014) from FCT and Sara
V. Pais holds a fellowship (PD/BD/52210/2013) within the scope
of the PhD program Molecular Biosciences (PD/00133/2012)
funded by FCT.

References
1. Costa TR, Felisberto-Rodrigues C, Meir A, 5. Lee VT, Anderson DM, Schneewind O (1998)
Prevost MS, Redzej A, Trokter M, Waksman G Targeting of Yersinia Yop proteins into the
(2015) Secretion systems in Gram-negative cytosol of HeLa cells: one-step translocation of
bacteria: structural and mechanistic insights. YopE across bacterial and eukaryotic mem-
Nat Rev Microbiol 13:343–359 branes is dependent on SycE chaperone. Mol
2. Sory MP, Cornelis GR (1994) Translocation of Microbiol 28:593–601
a hybrid YopE-adenylate cyclase from Yersinia 6. Schnaitman CA (1971) Effect of ethylenedi-
enterocolitica into HeLa cells. Mol Microbiol aminetetraacetic acid, Triton X-100, and lyso-
14:583–594 zyme on the morphology and chemical
3. Charpentier X, Oswald E (2004) Identification composition of isolate cell walls of Escherichia
of the secretion and translocation domain of coli. J Bacteriol 108:553–563
the enteropathogenic and enterohemorrhagic 7. Schnaitman CA (1971) Solubilization of the
Escherichia coli effector Cif, using TEM-1 beta-­ cytoplasmic membrane of Escherichia coli by
lactamase as a new fluorescence-based reporter. Triton X-100. J Bacteriol 108:545–552
J Bacteriol 186:5486–5495 8. Birdsell DC, Cota-Robles EH (1968) Lysis of
4. Collazo CM, Galan JE (1997) The invasion-­ spheroplasts of Escherichia coli by a non-ionic
associated type III system of Salmonella detergent. Biochem Biophys Res Commun
typhimurium directs the translocation of Sip 31:438–446
proteins into the host cell. Mol Microbiol 9. Esparis-Ogando A, Zurzolo C, Rodriguez-­
24:747–756 Boulan E (1994) Permeabilization of MDCK
 Assessing Effector Translocation by Differential Solubilization 515

cells with cholesterol binding agents: dependence subcellular targeting within host cells. Cell
on substratum and confluency. Am J Phys Microbiol 18:949–969
267:C166–C176 15. Letzelter M, Sorg I, Mota LJ, Meyer S, Stalder
10. Domingues L, Holden DW, Mota LJ (2014) The J, Feldman M, Kuhn M, Callebaut I, Cornelis
Salmonella effector SteA contributes to the control GR (2006) The discovery of SycO highlights a
of membrane dynamics of Salmonella-containing new function for type III secretion effector
vacuoles. Infect Immun 82:2923–2934 chaperones. EMBO J 25:3223–3233
11. Sory MP, Boland A, Lambermont I, Cornelis 16. Diepold A, Amstutz M, Abel S, Sorg I, Jenal
GR (1995) Identification of the YopE and U, Cornelis GR (2010) Deciphering the
YopH domains required for secretion and assembly of the Yersinia type III secretion
internalization into the cytosol of macro- injectisome. EMBO J 29:1928–1940
phages, using the cyaA gene fusion approach. 17. VanRheenen SM, Luo ZQ, O’Connor T,
Proc Natl Acad Sci U S A 92:11998–12002 Isberg RR (2006) Members of a Legionella
12. Figueira R, Watson KG, Holden DW, Helaine pneumophila family of proteins with ExoU
S (2013) Identification of Salmonella pathoge- (phospholipase A) active sites are translocated
nicity island-2 type III secretion system effec- to target cells. Infect Immun 74:3597–3606
tors involved in intramacrophage replication of 18. Denecker G, Totemeyer S, Mota LJ,
S. enterica serovar Typhimurium: implications Troisfontaines P, Lambermont I, Youta C,
for rational vaccine design. MBio 4:e00065 Stainier I, Ackermann M, Cornelis GR (2002)
13. Wang RF, Kushner SR (1991) Construction of Effect of low- and high-virulence Yersinia
versatile low-copy-number vectors for cloning, enterocolitica strains on the inflammatory
sequencing and gene expression in Escherichia response of human umbilical vein endothelial
coli. Gene 100:195–199 cells. Infect Immun 70:3510–3520
14. Domingues L, Ismail A, Charro N, Rodriguez-­ 19. Drecktrah D, Knodler LA, Ireland R, Steele-­
Escudero I, Holden DW, Molina M, Cid VJ, Mortimer O (2006) The mechanism of
Mota LJ (2016) The Salmonella effector SteA Salmonella entry determines the vacuolar envi-
binds phosphatidylinositol 4-phosphate for ronment and intracellular gene expression.
Traffic 7:39–51
 Chapter 36

Quantitative Determination of Anti-bacterial Activity

During Bacterial Co-culture
Juliana Alcoforado Diniz, Birte Hollmann, and Sarah J. Coulthurst

Abstract
Anti-bacterial activity assays are an important tool in the assessment of the ability of one bacterium to kill
or inhibit the growth of another, for example during the study of the Type VI secretion system (T6SS) and
the anti-bacterial toxins it secretes. The method we describe here can detect the ability of a bacterial strain
to kill or inhibit other bacterial cells in a contact-dependent manner when co-cultured on an agar surface.
It is particularly useful since it enumerates the recovery of viable target cells and thus enables quantification
of the anti-bacterial activity. We provide a detailed description of how to measure the T6SS-dependent
anti-bacterial activity of a bacterium such as Serratia marcescens against a competitor prokaryotic organ-
ism, Escherichia coli, and also describe possible variations in the method to allow adaptation to other
attacker and target organisms.

Key words Gram-negative bacteria, Protein secretion system, Type VI secretion system, Co-culture
assay, Anti-bacterial activity, Bacterial competitive fitness, Toxin/immunity pair

1  Introduction

To gain a fitness advantage in mixed microbial communities, many

bacteria have developed the ability to kill competitor prokaryotic
cells. Protein secretion systems are an important weapon in this
war, particularly the Type VI secretion system (T6SS), which can
be utilised to kill both closely and distantly related competitors
efficiently [1]. This versatile nanomachinery [2], which in some
cases can also be used against eukaryotic targets, is widespread in
Gram-negative bacteria. In the last few years, work by different
groups has resulted in the identification of many new anti-bacterial
toxins, also called effectors, delivered directly into target bacterial
cells by the T6SS. These include a variety of enzymes able to
­disrupt the bacterial cell wall, cell membrane and nucleic acid.
Together with these toxins, the secreting organism possesses
­cognate immunity proteins that are responsible for specific neu-
tralisation of the cognate effector to provide self-resistance [1, 3, 4].

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9_36, © Springer Science+Business Media LLC 2017

517
518 Juliana Alcoforado Diniz et al.

Fig. 1 Schematic overview of anti-bacterial activity assay. Following overnight growth on a solid agar surface,
the target strain (Ta) and different attacker strains (wild type, WT; T6SS inactive mutant, ΔT6; a mutant lacking
a toxin of interest, ΔTX; and a no-attacker, medium-only control, LB) are mixed in a ratio of 5 attackers to 1
target and spotted onto a solid agar surface. After a defined incubation period at the required temperature, the
co-culture spots (LB + target, green; wild type + target, blue; ΔT6SS + target, red; and Δtoxin + target, grey)
are scraped off, the cells resuspended and serial dilutions prepared. For an initial trial, these dilutions from
neat (N) to 10−6 (−6) are spotted on an agar plate supplemented with antibiotic selective for growth of the
target only. After incubation, estimation of target recovery from the trial plate is used to determine the dilution
which will provide a few tens of single colonies per plate in the actual experiment. For the actual experiment,
an appropriate volume of the correct dilution of the co-culture is spread on a selective plate with a glass
spreader and the colonies are counted following overnight incubation; replicate experiments provide fully
quantitative data. See text for full details

An important tool in identifying new toxin/immunity pairs and

also monitoring the level and impact of the T6SS-dependent
­anti-­bacterial activity of a particular strain is a co-culture-based
anti-­bacterial activity assay, such as that described here. This assay
is performed on the solid surface of an agar plate, since T6SS-­
mediated anti-bacterial activity requires intimate cell–cell contact
to permit the puncturing device of the machinery [5] to physically
interact with the target cell [6, 7].
In brief, the accessible method described here involves co-­
culture of an attacker and a target strain of bacteria on the surface
of an agar plate for a defined time, followed by the use of antibiotic
selection to kill the attacker strain and allow the recovery and enu-
meration of viable target cells. An overall schematic depiction of this
method is given in Fig. 1. To determine the number of surviving
 Anti-bacterial Activity Assay 519

target cells following exposure to the T6SS-wielding attacker, a

standard serial dilution-based viable count of target cells is per-
formed. This method is quantitative and reliable, with an extended
dynamic range (target cell recovery from 101 to >1010 colony-­
forming units (cfu) can be quantified owing to the serial dilution
approach). The alternatives to this assay use colorimetric or fluo-
rescent reporters to distinguish target cells within a co-culture [7–9].
These approaches have the advantage of convenience but have dis-
advantages in that they reduce the dynamic range of the output or
have potential issues in separating or distinguishing attacker and
target cells during quantification.
In Serratia marcescens, the technique described here has been
successfully implemented to demonstrate the existence and impact
of T6SS-mediated anti-bacterial activity and to identify new T6SS-­
dependent anti-bacterial toxins [10–13]. This technique, or minor
variations of it, has also been used to measure T6SS-mediated anti-­
bacterial activity in other organisms, including Vibrio cholerae,
Pseudomonas aeruginosa and Agrobacterium tumefaciens [14–16].
Overall, this assay can be an important tool in demonstrating
T6SS-dependent activity against competitor organisms, character-
ising the functionality of mutants in the T6SS and confirming the
identification of new toxin/immunity pairs. It could also be applied
to other inter-bacterial competitive strategies beyond the T6SS.
Here we describe in detail how to perform an anti-bacterial activity
assay of S. marcescens (attacker) against Escherichia coli (target) as
an example, along with ways to adapt the assay to other systems of
interest.

2  Materials

1. Liquid LB medium: 10 g tryptone, 10 g NaCl, 5 g yeast

extract, 1000 mL deionised water. Mix, adjust pH to 7.5 and
autoclave at 121 °C for 20 min.
2. LB agar: 10 g tryptone, 10 g NaCl, 5 g yeast extract, 1000 mL
deionised water, 12 g Select agar. Mix, adjust pH to 7.5 and
autoclave at 121 °C for 20 min.
3. Antibiotic: Dissolve 10 mg streptomycin sulfate in 1 mL deion-
ised water and pass through a syringe filter with pore size
0.2 μm to sterilise. Aliquot and store at −20 °C.
4. LB agar plates: Following autoclaving, bring the molten agar to
55 °C. Under sterile conditions, dispense 20 mL into each 90 mm
single-vented plastic Petri dish, then allow to cool and set.
5. LB agar plates plus antibiotic: Prepare LB agar plates as just
described, except this time add streptomycin to a final concen-
tration of 100 μg/mL (1/100 dilution) while the molten agar
is at 55 °C and mix well before pouring the plates.
520 Juliana Alcoforado Diniz et al.

6. Sterile disposable inoculation loops 10 μL.

7. Bent glass rod and ethanol for spreading cell suspensions on
agar plates.
8. Optional: Tally counter or pen-style colony counter for count-
ing colonies.
9. 30 and 37 °C static laboratory incubator.
10. Laminar flow cabinet (if unavailable, alternative plate-drying
methods can be used).

3  Methods

Carry out all procedures at room temperature and under sterile

conditions.
1. Streak out the required target and attacker strains from freezer
stocks and grow to single colonies according to their normal
requirements. In this example, the target strain is a streptomycin-­
resistant strain of E. coli K12, such as strain MC4100 [17]
(see Note 1). The attacker strains are selected according to the
experiment; they should include, at a minimum, the wild type
(e.g. S. marcescens Db10) and a T6SS inactive mutant strain
(lacking one of the core components), plus as many other
mutants as required.
2. Using a sterile inoculating loop, inoculate a single colony from
each of the strains onto an LB agar plate as a ‘patch’ (Fig. 1)
and incubate overnight at the optimal growth temperature for
that strain. Up to five strains can be patched together on the
same plate, taking care to keep them apart.
3. Dry the LB plates to be used for the co-culture spots the next
day; for this, keep them open for 2 h in a laminar flow cabinet
and then store at room temperature overnight.
4. Using a sterile disposable inoculating loop, scrape off a stripe
of cells from each of the overnight patches described in step 2
and resuspend in 0.5 mL of sterile LB in a sterile 1.5 mL
microcentrifuge tube (agitate the loop, remove and then vor-
tex the tube).
5. Measure the optical density (OD) of the resuspended cells for
each of the target and attacker strains and normalise to an
OD600 of 0.5 in a final volume of 100 μL of sterile LB medium
(e.g. if OD600 = 2.5, then add 20 μL of the culture to 80 μL of
medium).
6. Mix together normalised attacker and target cells at a ratio of
5 attacker: 1 target (e.g., 50 μL attacker +10 μL target)
(see Note 2). Do this for each attacker and also include 5 LB:
1 target as a no-attacker control.
 Anti-bacterial Activity Assay 521

7. Spot 25 μL of each mixture onto an LB agar plate (all the spots
from one replicate, i.e. one co-culture spot for each attacker,
go on the same plate). When the co-culture is being performed
at 37 °C (see Note 3), pre-warm this plate.
8. Wait 5 min to allow the spots to dry and immediately put plates
in the incubator at 37 °C for 4 h (see Note 4).
9. Following the incubation period, scrape off each spot using a
sterile disposable loop and resuspend all the cells on the loop
in 1 mL sterile LB; mix thoroughly on a vortex for approxi-
mately 30 s or until the pellet is completely resuspended.
10. Prepare serial tenfold dilutions of each resuspension from neat
to 10−6 using sterile LB (e.g. 90 μL LB plus 10 μL preceding
dilution). Make sure to change pipette tips and to vortex 5 s
between each dilution step. Then put diluted samples onto anti-
biotic-containing selective medium to enumerate the surviving
target cells, in one of two formats described in steps 11 and 12.
11. Trial: When testing a particular attacker/target combination
for the first time, it is advisable to perform a trial experiment to
determine the correct dilution to spread on a selective plate to
yield a few tens of single colonies. For this, perform a full serial
tenfold dilution of the resuspended cells, from neat to 10−6,
and spot 5 μL of each dilution onto an LB + streptomycin
plate, as shown in Fig. 1. Incubate at 37 °C (or the target
organism’s preferred growth temperature) overnight or until
single colonies have grown.
12. Proper experiment: Prepare a suitable dilution of the resus-
pended cells for each attacker/target pair (based on the trial)
and spread 50 μL or 100 μL onto LB + streptomycin plates.
We recommend using a bent glass rod, sterilised by dipping in
ethanol and removal using a flame, to spread the cell suspen-
sion evenly over the agar surface. Incubate at 37 °C (or the
target organism’s preferred growth temperature) overnight or
until single colonies have grown. It is important to obtain
­well-spaced single colonies; if this does not happen, adjust the
volume or dilution spread on the plate.
13. Count the colonies on the enumeration plates, using a tally or
pen-style colony counter if preferred. Calculate the number of
viable target cells recovered, expressed as cfu per co-­culture
spot. See also Notes 5 and 6.
Example (Fig. 2):
Target with LB only: 30 colonies in 50 μL of 10−6 dilu-
tion → 30 × (1000/50) × 106 = 6 × 108 cells/spot.
Target with wild-type S. marcescens: 11 colonies in 100 μL of
10−2 dilution → 11 × (1000/100) × 102 = 1.1 × 104 cells/spot.
Target with T6SS mutant: 45 colonies in 100 μL of 10−6 dilu-
tion → 45 × (1000/100) × 106 = 4.5 × 108 cells/spot.
522 Juliana Alcoforado Diniz et al.

Fig. 2 Graphical representation of data generated by anti-bacterial activity assay.

Top : Table showing number of colony-forming units per co-culture spot resulting
from a typical assay (from the example in the main text). Here just the first repli-
cate (R1) is represented so the mean corresponds to a single replicate and there
is no standard deviation (SEM); however, in the proper experiment at least four
replicates should be performed and the mean ± SEM presented. Bottom : Graph
generated using Microsoft Excel with data provided in table. The y-axis shows
the number of recovered target cells, presented as the number of colony-forming
units per co-culture spot, and the x-axis shows the attacker strains: no-attacker
control (LB), wild type (WT) and an inactive T6SS mutant (ΔT6SS)

4  Notes

1. Streptomycin-resistant strains: Target selection does not have to

utilise streptomycin; however, an antibiotic to which the
attacker is fully sensitive and the target is fully resistant is
required. This can be achieved using an intrinsic resistance of
the target strain or, alternatively, a chromosomally encoded,
stable resistance determinant can be introduced using standard
genetic methods.
2. Co-culture ratios: The initial ratio of attacker:target in the
­co-­culture spot can be varied. In our experience, 5:1 or 1:1
normally gives the best outcome. However, especially if the
 Anti-bacterial Activity Assay 523

two organisms are mismatched in terms of growth rate or the

killing effect is very small, more extreme ratios can work
better.
3. Incubation temperatures: The temperature at which co-culture
spots are incubated can be varied and optimised to suit the
attacker/target combination. The attacker strain S. marcescens
can grow at 37 °C or 30 °C; in this case, the choice is based on
the target strain.
4. Incubation times: The incubation time for the co-culture of
attacker and target can also be varied.
5. Number of replicates: To obtain quantitative data, at least four
replicates are required. For convenience, two replicates per
day, obtained an hour apart, is ideal. Start the replicates from
fresh patches of cells.
6. If desired, the recovery of the attacker following co-culture can
also be determined simultaneously, for example if the final
attacker:target ratio is of interest. In this case, the cells recov-
ered from the co-culture are enumerated in parallel on plates
containing a second antibiotic, one to which the target is sensi-
tive and the attacker is resistant.

Acknowledgements

This work was supported by Coordenação de Aperfeiçoamento de

Pessoal de Nível Superior (CAPES, Ph.D. studentship to JAD) and
the Wellcome Trust (Senior Fellowship to SJC).

References
1. Alcoforado Diniz J, Liu YC, Coulthurst SJ type VI secretion system spike. Nature 500:
(2015) Molecular weaponry: diverse effectors 350–353
delivered by the Type VI secretion system. Cell 6. Russell AB, Hood RD, Bui NK et al (2011)
Microbiol 17:1742–1751 Type VI secretion delivers bacteriolytic effec-
2. Cianfanelli FR, Monlezun L, Coulthurst SJ tors to target cells. Nature 475:343–347
(2016) Aim, load, fire: the Type VI secretion 7. Schwarz S, West TE, Boyer F et al (2010)
system, a bacterial nanoweapon. Trends Micro­ Burkholderia Type VI secretion systems have
biol 24:51–62 distinct roles in eukaryotic and bacterial cell
3. Durand E, Cambillau C, Cascales E et al interactions. PLoS Pathog 6:e1001068
(2014) VgrG, Tae, Tle, and beyond: the versa- 8. Gueguen E, Cascales E (2013) Promoter
tile arsenal of Type VI secretion effectors. swapping unveils the role of the Citrobacter
­
Trends Microbiol 22:498–507 rodentium CTS1 type VI secretion system in
4. Russell AB, Peterson SB, Mougous JD (2014) interbacterial competition. Appl Environ Micro­
Type VI secretion system effectors: poisons with biol 79:32–38
a purpose. Nat Rev Microbiol 12:137–148 9. Hachani A, Lossi NS, Filloux A (2013) A visual
5. Shneider MM, Buth SA, Ho BT et al (2013) assay to monitor T6SS-mediated bacterial
PAAR-repeat proteins sharpen and diversify the competition. J Vis Exp 20:50103
524 Juliana Alcoforado Diniz et al.

10. Alcoforado Diniz J, Coulthurst SJ (2015) 14. Hood RD, Singh P, Hsu F et al (2010) A type
Intraspecies competition in Serratia marcescens VI secretion system of Pseudomonas aeruginosa
is mediated by Type VI-secreted Rhs effectors targets a toxin to bacteria. Cell Host Microbe
and a conserved effector-associated accessory 7:25–37
protein. J Bacteriol 197:2350–2360 15. Ma LS, Hachani A, Lin JS et al (2014)
11. English G, Trunk K, Rao VA et al (2012) New Agrobacterium tumefaciens deploys a super-
secreted toxins and immunity proteins encoded family of type VI secretion DNase effectors
within the Type VI secretion system gene clus- as weapons for interbacterial competition in
ter of Serratia marcescens. Mol Microbiol 86: planta. Cell Host Microbe 16:94–104
921–936 16. Macintyre DL, Miyata ST, Kitaoka M et al
12. Fritsch MJ, Trunk K, Diniz JA et al (2013) (2010) The Vibrio cholerae type VI secretion
Proteomic identification of novel secreted anti- system displays antimicrobial properties. Proc
bacterial toxins of the Serratia marcescens Type Natl Acad Sci U S A 107:19520–19524
VI secretion system. Mol Cell Proteomics 17. Casadaban MJ, Cohen SN (1979) Lactose
12:2735–2749 genes fused to exogenous promoters in
13. Murdoch SL, Trunk K, English G et al (2011) one step using a Mu-lac bacteriophage:
The opportunistic pathogen Serratia marces- in vivo probe for transcriptional control
cens utilizes Type VI secretion to target bacte- sequences. Proc Natl Acad Sci U S A 76:
rial competitors. J Bacteriol 193:6057–6069 4530–4533
 Index

A Cell envelope�������������������������������������277, 301, 353, 368, 459

Cell lysate�����������������89, 93, 94, 117, 212, 248–251, 253, 254
Adenylate cyclase (Cya)����������������������������159–161, 169, 170, Cell surface exposure����������������������������������������������������87–94
201, 202, 486, 490, 501 Cell wall��������������������������������������������� 1, 30, 49, 62, 151, 152,
Affinity chromatography������������������������������������������ 222, 300 326, 327, 344, 345, 517
Affinitypurification. See Affinity chromatography Cell-envelope�������������������������������������������������������������������367
Agarose pad�������������������������������������������������������������� 292, 297 cI repressor��������������������������������������������������������������� 200, 201
Agar pad���������������������������������������������������������������������������293 Co-culture����������������������������������������������������������������518–523
Agrobacterium tumefaciens�������������������������������3, 66, 179, 213, Co-immunoprecipitation�������������������177, 213, 218, 248, 258
214, 378, 519 Colony blot���������������������������������������������������������������466–471
Alkaline phosphatase������������������������������������������������ 130, 174 Confocal microscope����������������������������������������������������67, 69
Amber suppressor tRNA������������������������������������������ 234, 237 Conformational changes������������������������������������ 98, 109, 110,
Analyte��������������������������������������258–261, 263–267, 270–274 277–286, 377, 416
Anti-bacterial activity�����������������������������������������������517–523 Consensus sequence�����������������������������������������������������������31
Arabidopsis������������������������������������������������ 476, 479–481, 485 Coomassie staining�����������������������������������317, 322, 325, 333,
Assembly pathways����������������������������������������� 3, 82, 289–297 339, 344, 460, 462
Co-sedimentation��������������������������������������������� 144, 145, 148
B
Cross-linking��������������������������������������������� 79, 143–146, 148,
Bacterialcompetition. See Anti-bacterial activity 213–217, 223, 249
Bacterial two-hybrid������������������159–162, 172, 174, 247, 258 Cryogen���������������������������������������������354, 355, 357, 358, 371
Bait����������������������������������� 166, 167, 178, 179, 182–183, 190, Cryogenic�������������������������������������������������������������������������364
212, 222, 226, 248, 249, 251–255 Cya reporter������������������������������������������������������ 475, 479, 486
Bayesian network����������������������������������������������������������������47 Cysteine������������������������������������������ 38, 65, 68–69, 75, 88, 98,
BCIP. See X-Pho 107–110, 112–126, 235, 237, 239, 243, 303
Beta-barrelprotein. See Outer membrane protein (OMP) Cytology-based two-hybrid (C2H)���������� 189–193, 195, 196
BIAcore. See Surface plasmon resonance (SPR) Cytoplasmic membrane (or inner membrane)�����������������38, 44,
Bimolecular Fluorescence Complementation 76, 79, 109, 115, 117, 124, 277–279, 285, 450, 456, 490
(BiFC)����������������������������������������������������������189–196
Biogenesis. See Assembly pathway D
Bioinformatics����������������2, 23, 24, 34, 36, 40, 44, 51, 66, 111 DDM. See n-dodecyl- β -D-maltopyranoside)
Biotinylation����������������������������������������89, 115, 117–119, 125 Decyl maltose neopentyl glycol (DM-NPG)������������������302,
Bitopic��������������������������������������������������������������������������97, 98 305, 307
Blue native polyacrylamide gel electrophoresis (Blue native Detergent������������������������������� 63, 89, 94, 102, 107, 120, 121,
PAGE)�������������������������������������������������� 233, 321–350 126, 215–217, 307, 312, 322, 345, 347, 501, 512
Bordetella pertussis������������������������������������� 159, 201, 474, 501 Dialysis������������������������������������������������������������� 313, 315–318
Bpa. See p-benzoyl-l-phenylalanine (Bpa) Differential solubilization�������������������������������������������������514
Bradford assay�����������������������������������261, 347, 476, 482, 483 Digitonin�������������������������������������������302, 307, 490, 502, 512
Dose fractionation������������������������������������������������������������387
C
Dual-color immunoblotting������������������������������������� 340, 341
Calmodulin�������������������������������� 222, 227, 229, 474, 490, 501
Calmodulin binding peptide (CBP)���������������� 222, 224, 225, E
229–231, 300 Effector���������������������������40, 41, 59, 226, 247, 257, 353, 449,
cAMP signaling�������������������������������������������������������� 159, 162 459–463, 465, 473–486, 490–497, 499, 501–514, 517
Carboxypeptidase Y����������������������������������������������������99–102 Effector translocation����������������������������������������������� 486, 514

Laure Journet and Eric Cascales (eds.), Bacterial Protein Secretion Systems: Methods and Protocols, Methods in Molecular
Biology, vol. 1615, DOI 10.1007/978-1-4939-7033-9, © Springer Science+Business Media LLC 2017

525
 Bacterial Protein Secretion Systems: Methods and Protocols
526  Index
  
Electron microscopy (EM) Immunoprecipitation (IP)������������������������115, 117, 120, 125,
cryo-electron microscopy������������������������������������ 381, 409 126, 162, 169, 211, 213, 233, 236, 238, 240–242, 324,
electron cryo-tomography�����������������������������������353–373 330, 334–335, 347
EM grid������������������������������������������������������������������� 354, 355, Infection����������������������������������� 179, 201, 305, 415, 492, 494,
357, 381 495, 498, 501–505, 507–510, 512
Enzyme-linked immunosorbent assay (ELISA)����������������90, Insertion����������������32, 66, 106, 107, 109–111, 126, 134–136,
466–469, 471, 474, 476, 481, 482 151, 171, 172, 200, 207, 235, 242, 305, 306, 358, 359
Escherichia coli (E. coli)���������������������������������59, 60, 62, 76, Interactome����������������������������������������������������������������������221
77, 79, 81, 82, 89, 102, 106, 110, 119, 125, 130, 131, Ion electrochemical gradient���������������������������� 277, 279, 282
133–137, 139, 140, 146, 152, 156, 160, 163, 165,
167, 169–171, 174, 179, 180, 190–192, 202, 203, L
205, 206, 222, 234–237, 244, 253, 260, 269, 280, Lambda Red based recombination�����������������������������������222
283, 296, 300, 301, 303–305, 334, 346, 378, 418, Legionella pneumophila�����������������������492, 494–495, 497, 498
419, 422, 423, 443, 467, 475, 478, 519, 520 Lipobox������������������������������������������������������ 66, 68–69, 71, 75
Lipoprotein labeling�����������������������������������������������������������77
F
Lipoproteins�����������������������������������38, 65–72, 75–78, 87, 89,
Fluorescence microscopy�����������������67–69, 72, 292, 293, 297 91, 235, 279, 302, 306
Fluorescent reporter����������������������������������������� 191, 192, 519 Lysozyme������������������������������� 60–63, 67–71, 82, 84, 99, 100,
Fractionation�������������������� 59–64, 83–85, 323–324, 329–333, 102, 153, 156, 214, 215, 217, 280, 281, 285, 300, 302,
346, 347, 363, 364, 372, 384, 387, 450–453, 490 304, 322, 326, 329, 344, 371
Fractionator���������������������������������������������������������� 84, 85, 332 Lytic transglycosylases (LTGs)����������������� 151, 152, 155, 156
Freeze and thaw����������������������������������������������������������������327
M
G
MacConkey/maltose medium�����������163, 166, 169, 173, 208
β-galactosidase����������������������������������130, 131, 133, 135, 137, Machine learning���������������������������������������29, 31–34, 40, 41,
139, 140, 162, 163, 167–170, 190, 200, 206–208, 44–46, 48, 50, 51, 110
281–283, 286 MacSyFinder�������������������������������������������4, 5, 7, 9–10, 18, 19
GAL4 DNA binding domain (DNA-BD)�������������� 178, 180 Maleimide����������������������������������������88, 91, 92, 94, 113–115,
Gal4 transcriptional activation domain����������������������������178 120–121, 124–126
GALLEX������������������������������������������������� 200–204, 206–208 Mass spectrometry����������������������������152, 177, 212, 221, 222,
Genetic assay������������������������������������������������������������ 159, 162 230, 322, 335, 340–344, 349, 350, 460–462, 466
Globomycin������������������������������������������������������������ 76, 77, 79 MCherry. See Fluorescent reporters
Membrane fraction�������������������������������������60, 61, 63, 81, 82,
H 125, 138, 323–324, 329–334
HeLa cells������������������������������������������������502, 503, 508–510, Membrane preparations��������������������� 82, 322–323, 327–329
512, 513 Membrane protein complexes��������������������������������������44, 51, 62,
Helix–helix interactions�������������������������������������������� 200, 202 81–86, 97–102, 105–126, 129–141, 199, 235, 301, 321,
Hidden Markov model (HMM)����������������������� 4–12, 16, 31, 322, 324, 326–329, 334–335
32, 36, 38, 39, 44–46, 111 Muramic acid assay�������������������������������������������������� 145, 147
Host cell����������������������������������� 116, 125, 226, 247, 248, 473,
N
474, 489, 490, 492, 498, 501–503, 512
Hybrid protein������������������������������������������131, 136, 159–164, Naïve Bayes classifier����������������������������������������������������32, 48
166–171, 226, 228 n-dodecyl- β -D-maltopyranoside (DDM)���������������������302,
307, 322, 327, 328, 334, 341, 345
I n-dodecyl-β-D-maltoside. See n-dodecyl- β
Identification -D-maltopyranoside (DDM)
of components���������������������������������������������� 2, 3, 5, 7, 17 Negative staining (NS)������������������������������������� 307, 317, 379
of effectors������������������������������������������������������������������463 N-hydroxysuccinimide (NHS)-based reagent������������ 88, 259
of the secretion systems������������������������������������ 1–19, 459 Nicotiana benthamiana������������������������������ 476, 479, 480, 485
Image processing������������������������������������������51, 69, 291, 294, Ni-NTA agarose beads����������������������������� 248–251, 253, 254
378, 393, 398, 405
O
Immunoblotting. See Western blot
Immunodetection���������������������������������69, 90, 100–102, 130, One-dimensional BN PAGE�������������������������������������������322
162, 203, 204, 252, 349, 453 One-hybrid�����������������������������������������������������������������������200
 Bacterial Protein Secretion Systems: Methods and Protocols
527
Index      

Orientation����������������������������� 94, 97, 98, 105, 107–112, 115, Pseudomonas aeruginosa���������������������������������� 66–69, 71, 267,
119, 123, 124, 129, 267, 306, 367, 368, 395–398, 401 311–313, 317, 318, 519
Ortho-nitrophenyl-β-D-galactoside (ONPG)����������������131, Pull-down������������������������������������������������� 169, 177, 248–254
133, 137, 139, 163, 168, 204, 207 Pulse-chase��������������������������������109, 234, 235, 237–241, 243
Osmotic shock��������������������������������������������������������������60–63 Purified peptidoglycan���������������������� 144–147, 149, 152–156
Outer membrane protein (OMP)���������������������47, 82, 85, 87,
97, 235, 342 R
Receiver operating characteristic (ROC) curve������������35, 36
P
Reconstruction������������������������� 296, 354, 356, 364–366, 372,
Palmitate labeling���������������������������������������������������������76, 77 373, 379, 387, 388, 394, 397–403, 405, 408, 416
p-benzoyl-l-phenylalanine (Bpa)������������� 234–236, 239, 242 Red-Gal (6-chloro-3-indolyl-β-D-galactoside)��������������131,
Peptidoglycan�����������������������������������59, 62, 79, 84, 143–149, 133, 139
151–156, 278, 285 Regular expression��������������������������������������������������������������39
Peptidoglycan-binding domain����������������������������������������145 Remazol blue���������������������������������������������������� 152, 153, 155
Performance measures��������������������������������������������������������35 Reporter gene�����������������������������������130, 132, 134, 135, 170,
Peripheral proteins��������������������������������������������������������63, 87 178, 200–202, 208, 485
Pho-lac dual reporter system (or pho-lac reporter Restriction site/ligation free cloning methods���������� 136, 300
fusions)������������������������������������������� 98, 132, 135, 140
Pilus��������������97, 109, 131, 143, 312–318, 415–418, 443, 445 S
Plant inoculation���������������������������������������������� 476, 479–481 Saccharomyces cerevisiae���������������������������������������������� 179, 180
Plunge freezing����������������������������������355, 357–358, 371, 381 Salmonella enterica serovar Typhimurium���������� 450, 502, 503
p-nitrophenyl phosphate (pNPP)������������� 131, 133, 138, 139 SDS-boiling������������������������������������������������������������� 144, 146
Polytopic������������������������ 97, 98, 102, 108–112, 119, 131, 139 Secretomes������������������������������������������������������������������������460
Positive-inside rule�������������������������������������������������������������44 Sec system��������������������������������������������������������������������������38
Prediction���������������������������� 52, 111, 112, 129, 134, 135, 441 Segmentation��������������������������������������������������� 363, 366, 367
Prey��������������������������� 178, 179, 182–183, 190, 248, 250–255 Sequence logo��������������������������������������������������� 31, 32, 37, 38
Pronase����������������������������������������������������� 144, 147, 153, 154 Sequence similarity search���������������������������2, 5, 8, 15, 16, 51
ProtA�������������������������������������������������222, 224–227, 229, 231 Serratia marcescens��������������������������������������������� 519, 521, 523
Protease���������������� 84, 88–90, 94, 98–102, 181, 186, 214, 215, Serva Blue G�������������������������������������322, 324, 325, 339, 344
217, 222, 224–227, 229–231, 236, 249, 250, 279, 300, Shearing�������������������������������������������������������������������311–319
302, 304, 322, 326, 327, 329, 344, 505, 508, 509 Signal peptide (or Signal sequence)����������������� 23, 31, 36–39,
Protease accessibility (or Proteinase accessibility)���������98–102, 45, 66, 75, 76, 91, 152, 300
281–283 Silver staining����������������������������222, 322, 325, 339–340, 344
Protease inhibitor��������������������������� 63, 90, 99, 102, 181, 182, Single-particle analysis�����������������������������������������������������403
214, 215, 217, 230, 249, 302, 304, 322, 326, 327, 329, Site-directed mutagenesis������������ 71, 119, 136, 191, 192, 235
344, 505, 508 Site-directed photocrosslinking�������������������������������� 234, 235
Protein A/G-agarose affinity resin��������������������������� 120, 122 Sodium carbonate������������������������������������������������� 60, 62, 325
Protein A/G Sepharose����������������������������������������������������117 Sodium dodecyl sulphate polyacrylamide gel electrophoresis
Proteinase K��������������89–91, 93, 99–102, 165, 166, 278, 279, (SDS-PAGE)���������������������������������61, 62, 77, 79, 82,
281–283, 285, 286, 478 85, 100–102, 109, 115, 116, 120, 122, 146, 148, 152,
Protein complex�������������������91, 190, 199, 212–214, 216, 221, 203–206, 211, 214, 216, 222, 226, 228, 230, 231, 236,
222, 231, 248, 299–301, 321, 322, 326–329, 334–336, 241, 243, 248, 250, 252, 261, 266, 313, 314, 316–318,
338, 340, 341, 344, 345, 377–409 322, 324–325, 328, 330, 334, 336–340, 348, 451,
Protein G-Sepharose����������������������������������������������������76, 78 453, 460, 461, 463, 503, 506, 508–511, 513
Protein-lipid interaction���������������������������������������������������199 Solid-state NMR������������������������������������������������������415–445
Protein overproduction�����������������������������������������������������299 Solubilisation�������������������������������������������� 60–63, 76, 78, 126
Protein-peptidoglycan interaction������������������������������������149 Sorting signals���������������������������������������23, 24, 29, 30, 52, 69
Protein-protein interaction�������������������������������� 81, 159–174, Spheroplast�����������������������������������������59–63, 67–72, 99–102,
177–186, 199, 211–218, 221–231, 247–255, 277, 279 278–283, 285, 286
Protein sorting��������������������������������������������������������������23–52 Structure determination���������������������������������� 106, 353, 417,
Proteolysis��������������� 63, 89–93, 100–102, 277, 278, 280, 286 429, 436, 440, 441, 443
Proton motive force Subcellular location (SCL) (or Subcellular
∆pH���������������������������������������������������������������������������454 localization)���������������������������������23, 29, 48, 131, 135
∆Ψ������������������������������������������������������������������������������� 454 Substituted cysteine accessibility method
Protonophore���������������������������������������������������� 279, 282, 286 (SCAM)��������������������������������������������������������� 98, 126
 Bacterial Protein Secretion Systems: Methods and Protocols
528  Index
  
Sucrose gradient (or Sucrose density Trypsin����������������������������������������������������������� 63, 89, 90, 102
gradient)����������� 69, 82, 84, 85, 323–324, 329–334, 344 Twin-arginine protein translocation�����������������������������������39
Superfolder green fluorescentprotein. See Fluorescent Two-dimensional BN/SDS PAGE������������������ 322, 336, 340
reporters Two-hybrid system������������������������������������������� 177, 178, 180
Supernatant������������������������������������������������������� 60–63, 70, 78, Type III secretion (T3SS)������������������������2–4, 6, 8, 15, 17, 41,
84, 100, 101, 122, 148, 156, 182–184, 205, 206, 212, 42, 50, 66, 98, 143, 151, 179, 200, 213, 226, 289, 295,
215, 216, 218, 240, 250, 253, 261, 281–283, 286, 304, 301, 307, 321, 322, 336, 340–342, 345, 377–380, 386,
316, 327–329, 333–335, 357, 422, 424, 426, 444, 450, 392, 393, 399, 402, 409, 415–418, 431, 439, 440, 443,
452, 453, 455, 456, 463, 465, 468, 470, 481, 486, 495, 449–456, 459, 473, 474, 476, 492, 499, 502, 503
502, 508, 509, 512 Type IV secretion (T4SS)������������������������������� 6, 98, 109, 143
Support vector machine������������������� 24, 31, 33, 40, 44, 47, 49 Type VI secretion system (T6SS)���������������������� 6, 17–19, 40,
Surface exposed lipoproteins����������������������������������������87, 89 66, 68–69, 98, 152, 179, 199, 213, 226, 269, 270, 289,
Surface plasmon resonance (SPR)����������������������������257–274 301–303, 460, 465–471, 501, 517–522

T U
Tandem Affinity Purification (TAP)�������� 222–226, 228–231 Ultracentrifugation���������������������������������61, 81, 85, 146, 148,
TEM-1 beta-lactamase reporter��������������� 490, 492, 493, 497 154–156, 304, 328–330, 332, 346
Tilt series�����������������������������������354, 356, 360–365, 371, 372 Urea�������������������������������������������������������������� 60, 63, 125, 423
Tobacco Etch Virus (TEV)�������222, 224–227, 229, 231, 300
Topology W
lipoprotein topology�����������������������������������������������������91 Western blot��������������������������������� 61, 62, 112, 116, 117, 120,
membrane/Inner membrane protein 122, 125, 162, 168, 169, 181, 183–184, 186, 205, 206,
topology�������������������������������������������������� 97, 101, 102 212–214, 216, 218, 226, 228, 248, 253, 279, 341, 451,
transmembrane β-barrel/OMPs topology��������������44, 47 460, 462, 466, 470, 494, 498
T7 overexpression�������������������������������������������������������������300 Whole-cell dot blot������������������������������������������������������92, 93
TOXCAT�����������������������������������������������������������������200–206
Toxin/immunity pairs����������������������������������������������� 518, 519 X
Transglycosylase. See Lytic transglycosylase
X-Gal�����������������������������������������������163, 164, 166, 167, 169,
Transient protein interaction (or transient
173, 204, 206–208
interaction)��������������������������������������������������� 177, 244
X-Pho (5-bromo-4-chloro-3-indolyl-phosphate or
Transmembranedomain. See Transmembrane segment
BCIP)�������������131, 133, 135, 137, 138, 140, 163, 204
Transmembranehelix. See Transmembrane segment
Transmembraneprotein. See Membrane protein Y
Trans-membrane segment���������������������������� 36, 98, 129, 303
Tris-Tricine gel electrophoresis Yeast transformation������������������������������������������������� 181, 182
(Tricine-SDS-PAGE)�������������� 77–79, 313, 316–318 Yeast two-hybrid (Y2H)���������������������������159, 162, 178, 179,
Triton X-10060, 62, 76, 79, 99–102, 121, 126, 213–217, 307, 182, 186, 213, 247
345, 502, 505, 508, 509, 513 Yersinia enterocolitica������������������������� 502–505, 507–508, 512