Introduction to Cell Mechanics and Mechanobiology

Introduction to Introduction to
Cell Mechanics and Cell Mechanics and
Mechanobiology Mechanobiology
Christopher R. Jacobs, Hayden Huang, Ronald Y. Kwon

Introduction to Cell Mechanics and Mechanobiology teaches a quantitative

understanding of the way cells detect, modify, and respond to the physical
aspects of the cell environment. Coverage includes the mechanics of single
molecule polymers, polymer networks, two-dimensional membranes, whole-
cell mechanics, and mechanobiology, as well as primer chapters on solid,
fluid, and statistical mechanics.

This textbook is designed for advanced undergraduate and graduate

students taking a course in the mechanics of the cell.  Researchers in

Jacobs Huang Kwon
biomedical engineering, bioengineering, and mechanical engineering
will also find it a useful reference.  Advanced analyses and mathematical
derivations are presented as optional boxes along with solved examples in
the main text; each chapter ends with problems and annotated references.

www.garlandscience.com Christopher R. Jacobs

ISBN 978-0-8153-4425-4
Hayden Huang
9 780815 344254
Ronald Y. Kwon
Introduction to Cell
Mechanics and
Mechanobiology
This page intentionally left blank
to match pagination of print book
Introduction to Cell
Mechanics and
Mechanobiology
Christopher R. Jacobs
Hayden Huang
Ronald Y. Kwon
Garland Science
Vice President: Denise Schanck
Editor: Summers Scholl
Senior Editorial Assistant: Allie Bochicchio
Production Editor and Layout: Natasha Wolfe
Illustrator: Laurel Muller, Cohographics
Cover design: Andrew Magee
Copyeditor: Christopher Purdon
Proofreader: Mary Curioli
Indexer: Indexing Specialists (UK) Ltd

©2013 by Garland Science, Taylor & Francis Group, LLC

This book contains information obtained from authentic and highly regarded sources. Every effort has been made to trace copyright holders
and to obtain their permission for the use of copyright material. Reprinted material is quoted with permission, and sources are indicated.
A wide variety of references are listed. Reasonable efforts have been made to publish reliable data and information, but the author and the
publisher cannot assume responsibility for the validity of all materials or for the consequences of their use. All rights reserved. No part of this
publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means—graphic, electronic, or mechanical,
including photocopying, recording, taping, or information storage and retrieval systems—without permission of the copyright holder.

ISBN 978-0-8153-4425-4

Library of Congress Cataloging-in-Publication Data

Jacobs, C. R. (Christopher R.)

Introduction to cell mechanics and mechanobiology/Christopher R. Jacobs, Hayden Huang, Ronald Y. Kwon.
p. cm.

  Includes bibliographical references.

  Summary: “Introduction to Cell Mechanics and Mechanobiology teaches advanced undergraduate students a quantitative understanding
of the way cells detect, modify, and respond to the physical properties within the cell environment. Coverage includes the mechanics of single
molecule polymers, polymer networks, two-dimensional membranes, whole-cell mechanics, and mechanobiology, as well as primer
chapters on solid, fluid, and statistical mechanics”–Provided by publisher.

ISBN 978-0-8153-4425-4 (pbk.)

I.  Huang, Hayden.  II.  Kwon, Ronald Y.  III.  Title.
[DNLM: 1. Cell Physiological Phenomena. 2.  Biomechanics. QU 375]
572–dc23
2012018504

Published by Garland Science, Taylor & Francis Group, LLC, an informa business,
711 Third Avenue, New York, NY, 10017, USA, and 2 Park Square, Milton Park, Abingdon, OX14 4RN, UK.

Printed in the United States of America

15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

Visit our website at http://www.garlandscience.com

Preface

In recent years, mechanical signals have become widely recognized as being criti-
cal to the proper functioning of numerous biological processes. This has led to the
emergence of a new field called cellular mechanobiology, which merges cell biol-
ogy with various disciplines of mechanics (including solid, fluid, statistical, com-
putational, and experimental mechanics). Cellular mechanobiology seeks to
uncover the principles by which the sensation or generation of mechanical force
alters cell function. Introduction to Cell Mechanics and Mechanobiology presents
students from a wide variety of backgrounds with the physical and mechanical
principles underpinning cell and tissue behavior.
This textbook arose from a cell mechanics course at Stanford University first
offered by two of us in 2005. Over several iterations, we taught from a set of course
notes and chapter excerpts—having found no textbook to cover the necessary
breadth of topics. Our colleagues had similar experiences teaching with the same
adhoc approach, which convinced us of the need for a comprehensive instruc-
tional tool in this area. Another reason we felt compelled to write this text is that
cell mechanics provides an excellent substrate to introduce many types of
mechanics (solid, fluid, statistical, experimental, and even computational). These
topics are traditionally covered in separate courses with applications largely
focused on engineering structures. As authors, we have varied backgrounds, but
share a common fondness for the insights mechanical engineering brings to cell
biology.
Introduction to Cell Mechanics and Mechanobiology is intended for advanced
undergraduates and early graduate students in biological engineering and bio-
medical engineering, including those not necessarily in a biomechanics track. We
do not assume an extensive knowledge in any area of biology or mechanics. We do
assume that students have a mathematics background common to all areas of
engineering and quantitative science, meaning exposure to calculus, ordinary dif-
ferential equations, and linear algebra.
The field of cell mechanics encompasses advanced concepts, such as large defor-
mation mechanics and nonlinear mechanics. We do not expect our audience to
have a strong background in the advanced mathematics of continuum mechan-
ics. Our intent is to avoid graduate-level mathematics wherever possible. In our
approach, the treatment of tensor mathematics—central to large deformation
mechanics (common in cell mechanics)—poses unique difficulties. To show sim-
plified mathematical derivations measuring mechanical parameters in the context
of living cells, we present tensors “by analogy” as matrices, rather than introduc-
ing them in a fully rigorous fashion. For example, we skip index notation entirely.
Admittedly, this approach may be less satisfying to mechanicians, which we also
consider ourselves. However, we hope that the advantages of this approach will
outweigh our oversimplifications.
The book is grouped into two parts: (I) Principles and (II) Practices. We have writ-
ten the chapters to allow instructors flexibility in presentation, depending on the
level of students and the length of the course. After introducing cell mechanics as
vi PREFACE

a framework in Chapter 1, we provide a review of cell biology in Chapter 2. The

next four chapters establish the necessary concepts in mechanics with enough
depth that the student attains a basic competency and appreciation for each topic.
Chapter 3 covers solid mechanics—including rigid and deformable bodies as well
as a short overview of large-deformation mechanics. Fluid mechanics (Chapter 4)
is important for cell mechanics not only in cytoplasmic flow, but also as a physical
signal that regulates cell mechanobiological behavior. Chapter 5 dives into statis-
tical mechanics, with descriptions of energy, entropy, and random walks, com-
mon themes for understanding the aggregate behavior of systems composed of
many objects. In Chapter 6, we describe experimental methods, an area that is
always changing, but is essential in demonstrating how theory may be reconciled
with actual experiments. These fundamentals in Part I are followed by cell
mechanics proper in Part II. Chapters 7–9 begin with a discussion of an aspect of
cell biology followed by analysis of the mechanics. We undertake polymer
mechanics in Chapter 7 from a continuum and a statistical viewpoint and exam-
ine situations in which both need to be considered simultaneously. These tools
are applied to individual cytoskeletal polymers as well as to other polymers such
as DNA. Polymer networks are presented in Chapter 8, with a focus on the role of
the cytoskeleton in regulating physical properties, such as red blood cell shape
and limitations on cell protrusion lengths. Chapter 9 examines the bilayer mem-
brane, from both the perspective of matter floating around within it (diffusion) as
well as a mechanical perspective of bending and stretching. The last two chapters
address mechanobiology. Chapter 10 is focused on cellular force generation and
the related processes of adhesion and migration. Chapter 11 discusses the process
of mechanosensing or mechanotransduction and intracellular signaling. These
last chapters do not have as much rigorous mechanical engineering mathematics,
but are an integral part of cell biomechanics.
Given the varied backgrounds of our students and the interdisciplinary nature of
the subject, we have attempted to provide some guidance on the treatment of
variables and units. At the start of the book, we present a master list of all the vari-
ables used in the text that specifies exactly what each variable is used for in a par-
ticular chapter. We have retained the “contextual” usage in each chapter, accepted
within each field, to prepare students for reading the literature. Three types of
boxes supplement the main text: “Advanced Material” challenges readers to think
critically and problem-solve; interesting and noteworthy asides are denoted as
“Nota Bene”; “Examples” provide in-depth solved calculations and explanations.
Each chapter concludes with a set of Key Concepts, Problems that can be used as
homework sets, and Annotated References that guide students for further study.

Online Resources
Accessible from www.garlandscience.com/cell-mechanics, Student and Instructor
Resource websites provide learning and teaching tools created for Introduction to
Cell Mechanics and Mechanobiology. The Student Resources site is open to every-
one, and users have the option to register in order to use book-marking and note-
taking tools. The Instructor’s Resource site requires registration; access is available
to instructors who have assigned the book to their course. To access the Instructor’s
Resource site, please contact your local sales representative or email science@gar-
land.com. Below is an overview of the resources available for this book. Resources
may be browsed by individual chapters and there is a search engine. You can also
access the resources available for other Garland Science titles.
For students:
• Computer simulation modules in two formats: ready-to-run simulations that simu-
late the mechanical behavior of cells and tutorial MATLAB modules on simulation
of cell behavior with the finite element method.
• Color versions of several figures are available, indicated by the figure legend in the
text.
 PREFACE vii

• A handful of animations and videos dynamically illustrate important concepts

from the book.
• Solutions to selected end-of-chapter problems are available to students.
For instructors:
• In addition to color versions of several figures, all of the images from the book are
available in two convenient formats: Microsoft PowerPoint• and JPEG. They have
been optimized for display on a computer. Figures are searchable by figure number,
figure name, or by keywords used in the figure legend from the book.
• The animations and videos that are available to students are also available on the
Instructor’s Resource website in two formats. The WMV-formatted movies are cre-
ated for instructors who wish to use the movies in PowerPoint presentations on
computers running Windows•; the QuickTime•-formatted movies are for use in
PowerPoint for Apple computers or Keynote• presentations. The movies can easily
be downloaded to your personal computer using the “download” button on the
movie preview page.
• Solutions to selected end-of-chapter problems are available to qualified adopters.

The origin of the book is rooted in teaching from sections of outstanding books
by David H. Boal, Jonathon Howard, and Howard C. Berg. We thank Roger
Kamm, Vijay Pande, and Andrew Spakowitz, who taught some of us at various
times and have unselfishly shared course materials and handouts and, in the
case of Dr. Kamm, unpublished drafts of his own textbook. With their permis-
sion, we have incorporated their approach to some topics in Chapters 4, 5, 7, 8,
and 9 and adapted several problems into sections of our book. We are grateful
for their amazing willingness to share their intellectual product in the name of
improving the educational experience of students around the world. We also
thank reviewers Roland R. Kaunas and Peter J. Butler, who shared notes from
their own courses in cell mechanics. We are profoundly appreciative of the tire-
less work of those who have preceded us, without whom we never could have
completed this task. We thank the additional reviewers of the book, Dan Fletcher,
Christian Franck, Wonmuk Hwang, Paul Janmey, Yuan Lin, Lidan You, and
Diane Wagner, for their valuable insight and critiques of our drafts. We are also
grateful to Summers Scholl and the editorial and production teams at Garland
who took a chance on three textbook neophytes and guided us unerringly
through uncharted waters. Finally we are each deeply indebted to our families,
including Roberta, Jolene, VH, YYH, LHH, Joyce, Melody, Tae, and Cynthia.
Without your support, patience, and understanding—as this project took us
away from you on so many nights and weekends—we never could have contem-
plated this undertaking, much less completed it.

Christopher R. Jacobs
Hayden Huang
Ronald R. Kwon
This page intentionally left blank
to match pagination of print book
Detailed Contents

Prefacev CHAPTER 2: Fundamentals in Cell

Biology 19
PART I: PRINCIPLES 1 2.1 ​Fundamentals in cell and molecular
biology19
CHAPTER 1: Cell Mechanics as a Framework 3
Proteins are polymers of amino acids 20
1.1 Cell mechanics and human disease 4
DNA and RNA are polymers of
Specialized cells in the ear allow you to hear 5
nucleic acids 22
Hemodynamic forces regulate endothelial
Polysaccharides are polymers of sugars 24
cells 6
Fatty acids store energy but also form
To keep bone healthy, bone cells need
structures 24
mechanical stimulation 6
Correspondence between DNA-to-RNA-
The cells that line your lungs sense stretch 7
to-protein is the central dogma of
Pathogens can alter cell mechanical properties 7 modern cell biology 25
Other pathogens can use cell mechanical Phenotype is the manifestation
structures to their advantage 7 of genotype 27
Cancer cells need to crawl to be metastatic 8 Transcriptional regulation is one way
Viruses transfer their cargo into cells they that phenotype differs from genotype 28
infect 8 Cell organelles perform a variety
1.2 The cell is an applied mechanics grand of functions 29
challenge8 2.2 Receptors are cells’ primary chemical 
Computer simulation of cell mechanics sensors30
requires state-of-the-art approaches 9 Cells communicate by biochemical signals 30
1.3 Model problem: micropipette aspiration 9 Signaling between cells can occur
What is a typical experimental setup for through many different mechanisms 31
micropipette aspiration? 9 Intracellular signaling occurs via small
The liquid-drop model is a simple model molecules known as second messengers 32
that can explain some aspiration results 11 Large molecule signaling cascades
The Law of Laplace can be applied to a have the potential for more specificity 34
spherical cell 12 Receptors use several mechanisms to
Micropipette aspiration experiments can initiate signaling 35
be analyzed with the Law of Laplace 12 2.3 Experimental biology 36
How do we measure surface tension Optical techniques can display cells clearly 37
and areal expansion modulus? 13
Fluorescence visualizes cells with lower
Why do cells “rush in”? 15 background 38
Cells can behave as elastic solids Fluorophores can highlight structures 39
or liquid drops 16
Fluorophores can probe function 40
Key Concepts 16
Atomic force microscopy can elucidate
Problems 17 the mechanical behavior of cells 40
Annotated References 17 Gel electrophoresis can separate molecules 41
x DETAILED CONTENTS

Visualizing gel-separated products employs Torsion of a solid cylinder can be

a variety of methods 42 modeled as a torsion of a series of
PCR amplifies specific DNA regions shells of increasing radius 61
exponentially 43 Kinematics, equilibrium, and constitutive
2.4 ​Experimental design in biology 46 equations are the foundation of solid
mechanics 62
Reductionist experiments are powerful
but limited 46 Kinematics in a beam are the strain–
displacement relationship 62
Modern genetics has advanced our
ability to study in situ 48 Equilibrium in a beam is the stress–
moment relationship 64
Bioinformatics allows us to use vast
amounts of genomic data 49 The constitutive equation is the
stress–strain relationship 65
Systems biology is integration rather
than reduction 49 The second moment of inertia is a
measure of bending resistance 65
Biomechanics and mechanobiology are
integrative 49 The cantilevered beam can be solved
from the general beam equations 66
Key Concepts 50
Buckling loads can be determined from
Problems 50
the beam equations 67
Annotated References 52
Transverse strains occur with axial loading 68
The general continuum equations can be
CHAPTER 3: Solid Mechanics Primer 53
developed from our simple examples 68
3.1 Rigid-body mechanics and free-body
Equilibrium implies conditions on stress 69
diagrams53
Kinematics relate strain to displacement 71
What is a “rigid” body? 53
The constitutive equation or stress–strain
One of the most powerful, but underused,
relation characterizes the material behavior 73
tools is a free-body diagram 53
Vector notation is a compact way to express
Identifying the forces is the first step in
equations in continuum mechanics 74
drawing a free-body diagram 54
Stress and strain can be expressed as matrices 76
Influences are identified by applying the
equations of motion 54 In the principal directions shear stress is zero 76
Free-body diagrams can be drawn for parts 3.3 Large deformation mechanics 78
of objects 55 The deformation gradient tensor
3.2 Mechanics of deformable bodies 55 describes large deformations 78
Rigid-body mechanics is not very Stretch is another geometrical measure
useful for analyzing deformable bodies 55 of deformation 79
Mechanical stress is analogous to pressure 56 Large deformation strain can be defined
in terms of the deformation gradient 80
Normal stress is perpendicular to the
area of interest 56 The deformation gradient can be
decomposed into rotation and
Strain represents the normalized
stretch components 82
change in length of an object to load 57
3.4 Structural elements are defined by
The stress–strain plot for a material
their shape and loading mode 83
reveals information about its stiffness 57
Key Concepts 84
Stress and pressure are not the same
thing, because stress has directionality 58 Problems 84
Shear stress describes stress when forces Annotated References 87
and areas are perpendicular to each other 59
Shear strain measures deformation Chapter 4: Fluid Mechanics Primer 89
resulting from shear stress 59 4.1 Fluid statics 89
Torsion in the thin-walled cylinder can Hydrostatic pressure results from
be modeled with shear stress relations 60 gravitational forces 89
 DETAILED CONTENTS xi

Hydrostatic pressure is isotropic 91 5.1 Internal energy 120

Resultant forces arising from hydrostatic Potential energy can be used to make
pressure can be calculated through predictions of mechanical behavior 120
integration 92 Strain energy is potential energy stored
4.2 ​Newtonian fluids 92 in elastic deformations 122
Fluids obey mass conservation 93 Equilibrium in continuum mechanics is
Fluid flows can be laminar or turbulent 94 a problem of strain energy minimization 123
Many laminar flows can be solved analytically 95 Changes in mechanical state alter
internal energy 123
Many biological fluids can exhibit
non-Newtonian behavior 97 5.2 Entropy 124
4.3 ​The Navier–Stokes equations 98 Entropy is directly defined within
statistical mechanics 124
Derivation of the Navier–Stokes equations
begins with Newton’s second law 99 Microstates, macrostates, and density
of states can be exemplified in a
Constitutive relations and the continuity
three-coin system 124
equation are necessary to make Navier’s
equations solvable 102 Microstates, macrostates, and density
of states provide insight to macroscopic
Navier–Stokes equations: putting it all
system behavior 127
together 103
Ensembles are collections of microstates
4.4 ​Rheological analysis 103
sharing a common property 127
The mechanical behavior of viscoelastic
Entropy is related to the number of microstates
materials can be decomposed into
associated with a given macrostate 127
elastic and viscous components 104
5.3 Free energy 128
Complex moduli can be defined for
viscoelastic materials 106 Equilibrium behavior for thermodynamic
systems can be obtained via free energy
Power laws can be used to model frequency-
minimization 128
dependent changes in storage and
loss moduli 108 Temperature-dependence of end-to-end
length in polymers arises out of
4.5 Dimensional analysis 110
competition between energy and entropy 129
Dimensional analysis requires the
5.4 Microcanonical ensemble 131
determination of base parameters 110
The hairpinned polymer as a non-
The Buckingham Pi Theorem gives the
interacting two-level system 132
number of dimensionless parameters
that can be formed from base parameters 111 A microcanonical ensemble can be
used to determine
Dimensionless parameters can be found
constant energy microstates 132
through solving a system of equations 111
Entropy can be calculated via
Similitude is a practical use of
combinatorial enumeration of the
dimensional analysis 113
density of states 133
Dimensional parameters can be
Entropy is maximal when half the
used to check analytical expressions 114
sites contain hairpins 133
Key Concepts 115
S(W) can be used to predict equilibrium
Problems 116 behavior 133
Annotated References 117 The number of hairpins at equilibrium is
dependent on temperature 134
Chapter 5: Statistical Mechanics Primer 119 Equilibrium obtained via the
Statistical mechanics relies on the use microcanonical ensemble is identical to
of probabilistic distributions 119 that obtained via free energy minimization 135
Statistical mechanics can be used to 5.5 Canonical ensemble 136
investigate the influence of random Canonical ensemble starting from the
molecular forces on mechanical behavior 119 microcanonical ensemble 136
xii DETAILED CONTENTS

Probability distribution from the canonical 6.2 Measurement of forces produced by cells 160
ensemble gives Boltzmann’s law 138 Traction force microscopy measures the forces
The free energy at equilibrium can be exerted by a cell on its underlying surface 160
found using the partition function 139 Cross-correlation can be used for
The internal energy at equilibrium can be particle tracking 160
determined using the partition function 141 Determining the forces that produced a
Using the canonical approach may be displacement is an inverse problem 163
preferable for analyzing thermodynamic Microfabricated micropillar arrays can
systems 142 be used to measure traction forces directly 165
5.6 Random walks 143 Surface modification can help determine
A simple random walk can be how a cell interacts with its surroundings 166
demonstrated using soccer 143 6.3 Applying forces to cells 167
The diffusion equation can be derived Flow chambers are used for studying
from the random walk 145 cellular responses to fluid shear stress 167
Key Concepts 147 The transition between laminar and
Problems 148 turbulent flow is governed by the
Reynolds number 168
Annotated References 149
Parallel plate flow devices can be designed
for low Reynolds number shear flow 168
Chapter 6: Cell Mechanics in the Fully developed flow occurs past the
Laboratory 151 entrance length 169
6.1 Probing the mechanical behavior of cells Cone-and-plate flow can be used to
through cellular micromanipulation 151 study responses to shear 170
Known forces can be applied to cells through Diverse device designs can be used to
the use of cell-bound beads and an study responses to fluid flow 171
electromagnet 152 Flexible substrates are used for
The dependence of force on distance subjecting cells to strain 172
from the magnet tip can be calibrated Confined uniaxial stretching can lead
through Stokes’ law 152 to multiaxial cellular deformations 172
Magnetic twisting and multiple-pole Cylindrically symmetric deformations
micromanipulators can apply stresses generate uniform biaxial stretch 172
to many cells simultaneously 153 6.4 Analysis of deformation 173
Optical traps generate forces on particles Viscoelastic behavior in micromanipulation
through transfer of light momentum 153 experiments can be parameterized
Ray tracing elucidates the origin of through spring–dashpot models 173
restoring forces in optical tweezers 154 Combinations of springs and dashpots can
What are the magnitudes of forces in be used to model viscoelastic behavior 174
an optical trap? 155 Microscopy techniques can be adapted
How does optical trapping compare to visualize cells subject to mechanical 
with magnetic micromanipulation? 156 loading 177
Atomic force microscopy involves the Cellular deformations can be inferred
direct probing of objects with a small from image sequences through
cantilever157 image correlation–based approaches 178
Cantilever deflection is detected using Intracellular strains can be computed
a reflected laser beam 157 from displacement fields 179
Scanning and tapping modes can be 6.5 Blinding and controls 181
used to obtain cellular topography 158 Key Concepts 182
A Hertz model can be used to estimate  Problems 183
mechanical properties 158 Annotated References 184
 DETAILED CONTENTS xiii

PART II: Practices 187 Force is the gradient of free energy in

thermodynamic systems 208
CHAPTER 7: Mechanics of Cellular
The behavior of polymers tends
Polymers 189 toward that of an ideal chain in
7.1 Biopolymer structure 189 the limit of long contour length 209
Microfilaments are polymers composed 7.5 Freely jointed chain (FJC) 210
of actin monomers 189 The FJC model places a limit on
F-actin polymerization is influenced by polymer extension 210
the molecular characteristics of G-actin 189 The force–displacement relation for the
Microtubules are polymers composed FJC can be found by the canonical
of tubulin dimers 191 ensemble 211
MT polymerization is affected by Differences between the ideal chain
polarity and GTP/GDP binding 191 and the FJC emerge at large forces 213
Intermediate filaments are polymers 7.6 Worm-like chain (WLC) 214
with a diverse range in composition 192 The WLC incorporates energetic
Intermediate filaments possess a coiled-coil effects of bending 214
structure 192 The force–displacement relation for
Intermediate filaments have diverse the WLC can be found by the
functions in cells 192 canonical ensemble 216
7.2 Polymerization kinetics 194 Differences in the WLC and FJC emerge
Actin and MT polymerization can when they are fitted to experimental data
be modeled as a bimolecular reaction 195 for DNA 217
The critical concentration is the only Persistence length is related to Kuhn
concentration at which the polymer length 218
does not change length 195 Key Concepts 219
Polarity leads to different kinetics on Problems 220
each end 196 Annotated References 221
Polymerization kinetics are affected by
ATP/ADP in actin and GTP/GDP
binding in tubulin 197 CHAPTER 8: Polymer Networks and
Subunit polarity and ATP hydrolysis lead to the Cytoskeleton 223
polymer treadmilling 197 8.1 Polymer networks 223
7.3 Persistence length 198 Polymer networks have many degrees
Persistence length gives a measure of freedom 223
of flexibility in a thermally fluctuating Effective continuums can be used to
polymer 198 model polymer networks 223
Persistence length is related to 8.2 Scaling approaches 225
flexural rigidity for an elastic beam 200
Cellular solids theory implies
Polymers can be classified as stiff, flexible, scaling relationships between
or semi-flexible by the persistence length 202 effective mechanical properties and
7.4 Ideal chain 203 network volume fraction 225
The ideal chain is a polymer model Bending-dominated deformation results
for flexible polymers 203 in a nonlinear scaling of the elastic
The probability for the chain to have different modulus with volume fraction 225
end-to-end lengths can be determined Deformation dominated by axial strain
from the random walk 204 results in a linear scaling of the
The free energy of the ideal chain can be elastic modulus with volume fraction 227
computed from its probability The stiffness of tensegrity structures
distribution function 207 scales linearly with member prestress 228
xiv DETAILED CONTENTS

8.3 Affine networks 229 The fluid mosaic model of the cell membrane
Affine deformations assume the describes its physical properties 252
filaments deform as if they are 9.2 Phospholipid self-assembly 252
embedded in a continuum 229 Critical micelle concentration depends
Flexible polymer networks can on amphiphile molecular structure 253
be modeled using rubber elasticity 230 Aggregate shape can be understood
Anisotropic affine networks can be from packing constraints 254
modeled using strain energy approaches 233 9.3 Membrane barrier function 255
Elastic moduli can be computed The diffusion equations relate
from strain energy density 233 concentration to flux per unit area 256
Elastic moduli of affine anisotropic networks Fick’s second law shows how spatial
can be calculated from appropriate concentration changes as a function of time 257
strain energy density and angular
9.4 Membrane mechanics I: In-plane shear
distribution functions 235
and tension 259
8.4 Biomechanical function and
Thin structures such as membranes
cytoskeletal structure 236
can be treated as plates or shells 260
Filopodia are cross-linked bundles
Kinematic assumptions help describe
of actin filaments involved in cell motility 236
deformations 260
Actin filaments within filopodia can be
A constitutive model describes
modeled as elastic beams undergoing
material behavior 262
buckling 236
The equilibrium condition simplifies
The membrane imparts force on the
for in-plane tension and shear 262
ends of filopodia 238
Equilibrium simplifies in the case of
The maximum filopodium length before
shear alone 265
buckling in the absence of cross-linking is
shorter than what is observed in vivo 238 Equilibrium simplifies in the case of
equibiaxial tension 266
Cross-linking extends the maximum
length before buckling 238 Areal strain can be a measure of
biaxial deformation 267
Is the structure of the red blood cell’s
cytoskeleton functionally advantageous? 239 9.5 Membrane mechanics II: Bending 267
Thin structures can be analyzed using the In bending the kinematics are
two-dimensional shear modulus and governed by membrane rotation 268
the areal strain energy density 240 Linear elastic behavior is assumed
Sixfold connectivity facilitates resistance for the constitutive model 269
to shear 242 Equilibrium places conditions on
Fourfold connectivity does not resultant forces and moments 269
sustain shear as well as sixfold 245 Which dominates, tension or bending? 272
Key Concepts 246 9.6 Measurement of bending rigidity 272
Problems 246 Membranes undergo thermal
Annotated References 247 undulations similar to polymers 272
Membranes straighten out with tension 273
Key Concepts 275
Chapter 9: Mechanics of the Cell
Problems 275
Membrane 249
Annotated References 277
9.1 Membrane biology 249
Water is a polar molecule 249
Cellular membranes form by interacting CHAPTER 10: Adhesion, Migration, and
with water 250 Contraction 279
The saturation of the lipid tails determines 10.1 Adhesion 279
some properties of the membrane 251 Cells can form adhesions with the substrate 279
The cell membrane distinguishes inside Fluid shear can be used to measure
and outside 251 adhesion strength indirectly 281
 DETAILED CONTENTS xv

Detachment forces can be measured Key Concepts 308

through direct cellular manipulation 281 Problems 308
The surface tension/liquid-drop model Annotated References 310
can be used to describe simple
adhesion 282
Adhesive peeling can be modeled Chapter 11: Cellular
using continuum mechanics 284 Mechanotransduction 311
Adhesion energy density can be 11.1 Mechanical signals 311
obtained through consideration Vascular endothelium experiences
of strain energy 286 blood-flow-mediated shear stress 312
Targeting of white blood cells during Lumen-lining epithelial cells are
inflammation involves the formation subjected to fluid flow 313
of transient and stable intercellular
Fluid flow occurs in musculoskeletal tissues 313
adhesions 287
Fluid flow during embryonic development
Kinetics of receptor–ligand binding can
regulates the establishment of left–right
be described with the law of mass action 288
asymmetry 315
The Bell model describes the effect
Strain and matrix deformation function
of force on dissociation rate 290
as regulatory signals 316
Shear enhances neutrophil adhesion—
Smooth muscle cells and cardiac
up to a point 291
myocytes are subjected to strain in
10.2 Migration 292 the cardiovascular system 316
Cell migration can be studied in vitro Cellular strain in the musculoskeletal
and in vivo 292 system is dependent on tissue stiffness 317
Cell locomotion occurs in distinct steps 292 The lung and bladder are hollow elastic
Protrusion is driven by actin organs that are regulated by stretch 317
polymerization 293 Cells can respond to hydrostatic pressure 317
Actin polymerization at the leading edge:
11.2 Mechanosensing organelles and structures 318
involvement of Brownian motion? 294
Stereocilia are the mechanosensors of the ear 319
Cell motion can be directed by external cues 295
Specialized structures are used in touch
Cell migration can be characterized
sensation 320
by speed and persistence time 296
Primary cilia are nearly ubiquitous, but
Directional bias during cell migration
functionally mysterious 320
can be obtained from cell trajectories 298
Cellular adhesions can sense as well as
10.3 Contraction 298
transmit force 321
Muscle cells are specialized cells for
The cytoskeleton can sense mechanical loads 322
contractile force generation 299
Mechanosensing can involve the
Studying cardiac function gave early
glycoproteins covering the cell 323
insight into muscle function 299
The skeletal muscle system generates The cell membrane is ideally suited
skeletal forces for ambulation and mobility 300 to sense mechanical loads 324
The Hill equation describes the relationship Lipid rafts affect the behavior of proteins
between muscle force and velocity 300 within the membrane 325
Non-muscle cells can generate 11.3 Initiation of intracellular signaling 326
contractile forces within stress fibers 301 Ion channels can be mechanosensitive 326
Stress fiber pre-strain can be measured Hydrophobic mismatches allow the
from buckling behavior 302 mechanical gating of membrane channels 327
Myosin cross-bridges generate sliding Mechanical forces can expose
forces within actin bundles 303 cryptic binding sites 328
Myosin molecules work together to Bell’s equation describes protein
produce sliding 304 unfolding kinetics 329
The power-stroke model is a mechanical Molecular conformation changes
model of actomyosin interactions 305 can be detected fluorescently 329
xvi DETAILED CONTENTS

11.4 Alteration of cellular function 330 Mechanical stimulation can induce

Intracellular calcium increases in extracellular matrix remodeling 333
response to mechanical stress 330 Cell viability and apoptosis are altered
Nitric oxide, inositol triphosphate, and by different processes 334
cyclic AMP, like Ca2+, are second Key Concepts 334
messenger molecules implicated in Problems 334
mechanosensation 331
Annotated References 335
Mitogen-activated protein kinase activity
is altered after exposure to mechanical
stimulation 332 Abbreviations337
Mechanically stimulated cells exhibit List of variables and units 338
prostanglandin release 332
Index343
Mechanical forces can induce
morphological changes in cells 332
PART I: PRINCIPLES
This page intentionally left blank
to match pagination of print book
CHAPTER 1
Cell Mechanics as
a Framework

B iological cells are the smallest and most basic units of life. The field of cell
biology, which seeks to elucidate cell function through better understanding
of physiological processes, cellular structure, and the interaction of cells with the
extracellular environment, has become the primary basic science for better
understanding of human disease in biomedical research. Until recently, the
study of basic problems in cell biology has been performed almost exclusively
within the context of biochemistry and through the use of molecular and genetic
approaches. Pathological processes may be considered disruptions in biochemi-
cal signaling events. The regulation of cell function by extracellular signals may
be understood from the point of view of binding of a molecule to a receptor on
the cell surface. Basic cellular processes such as cell division are considered in
terms of the biochemical events driving them. This emphasis on biochemistry
and structural biology in cell biology research is reflected in typical curricula and
core texts traditionally used for cell biology courses.
Recently, there has been a shift in paradigm in the understanding of cell func-
tion and disease primarily within the analytical context of biochemistry. In
particular, it has become well established that critical insights into diverse cel-
lular processes and pathologies can be gained by understanding the role of
mechanical force. A rapidly growing body of science indicates that mechanical
phenomena are critical to the proper functioning of several basic cell processes
and that mechanical loads can serve as extracellular signals that regulate cell
function. Further, disruptions in mechanical sensing and/or function have
been implicated in several diseases considered major health risks, such as
osteoporosis, atherosclerosis, and cancer. This has led to the emergence of a
new discipline that merges mechanics and cell biology: cellular mechanobiol-
ogy. This term refers to any aspect of cell biology in which mechanical force is
generated, imparted, or sensed, leading to alterations in cellular function. The
study of cellular mechanobiology bridges cell biology and biochemistry with
various disciplines of mechanics, including solid, fluid, statistical, experimen-
tal, and computational mechanics.
The primary goal of this introductory chapter is to motivate the study of cell
mechanics and cellular mechanobiology by: (1) demonstrating its role in basic
cellular and pathological processes; and (2) showing how cell mechanics pro-
vides an ideal framework for introducing a broad mechanics curriculum in an
integrated manner. We first present cell mechanics in the context of human dis-
ease by providing a survey of physiological and pathological processes that are
mediated by cell mechanics and can be better understood through mechanical
analyses. Next, we propose cell mechanics as an ideal substrate for introducing
principles of solid, fluid, statistical, experimental, and even computational
mechanics, and put forth the argument that cell mechanics may be the grand
challenge of applied mechanics for the twenty-first century. Finally, we present
a simple model problem: micropipette aspiration, in which a cell is partly
“sucked” into a narrow tube by a vacuum. This example will help you develop
a feeling for how cell mechanics is studied and demonstrate how a relatively
4 CHAPTER 1: Cell Mechanics as a Framework

simple approach can give important insight into cell mechanical behavior (and
how this behavior can dictate cellular function).

1.1 CELL MECHANICS AND HUMAN DISEASE

Most of our understanding of biomedicine, both in terms of health and disease,
is biological or biochemical in nature. There are some exceptions, of course, such
as the component of mechanics at the tissue or whole organism level when we
think about fracture of a bone, soft tissue trauma, or surgical repair. Further,
when you think of your senses that involve mechanics, such as hearing and
touch, the fact that mechanically specialized cells are involved is unsurprising.
By contrast, we do not typically think about mechanics of cells in relation to can-
cer, malaria, or viral infections—but they are related. What may be even more of
a surprise is that many of the causes of human suffering involve cell mechanics
to some degree or other.
For instance, the health of several tissues, particularly tissues of the skeleton
(bone and cartilage) and of the cardiovascular system (the heart and arteries),
is heavily dependent on mechanical loading, which in turn comes from physi-
cal activity and the environment (gravity). To be clear, we are not simply saying
that these physiological systems have a mechanical function (bones support
the body and the heart pumps blood)—which they do. We are also saying that
these systems actively change and respond to changes in mechanical forces at
the cellular level—bones will reinforce certain regions and actively degrade
others. In this chapter we hope to convince the reader that mechanics is in fact
involved in virtually every aspect of life, although its influence may be subtle or
indirect.
Understanding human health and disease often requires an understanding of
biomechanics and mechanobiology at the cellular level, for example:
• When bone cells do not experience proper mechanical stimulation, bone forma-
tion ceases and bone resorption is initiated. So, in prolonged space travel, where
gravity is virtually nonexistent, astronauts face major bone loss, even with rigorous
exercise regimens.
• In coronary artery disease, changes in the temporal and spatial patterns of fluid
shear stress on endothelial cells are linked to the formation of atherosclerotic
plaques.
• The pathogenesis of osteoarthritis occurs due to changes in physical loading that
lead to altered mechanical signals experienced by chondrocytes.
• Lung alveolar epithelial cells and airway smooth muscle cells are regulated by
cyclic mechanical stretch during breathing, and hypersensitization due to airborne
pathogens that can lead to sustained hypercontractility, which in turn can cause
asthmatic attacks.
• Infection can be initiated from mechanical disruption of the cell membrane by
viruses delivering foreign genetic material. This is a serious problem—if we could
deliver genes as easily as viruses, we could potentially cure many genetic diseases
by having cells express the corrected version (of the mutated gene). But the cell
membrane is actually an excellent mechanical barrier.
• Metastatic cancer cells must be able to migrate through tissue and attach at distant
sites to spread. Why certain cancers appear to metastasize preferentially to particu-
lar locations is still a mystery.
• Mechanical stimuli regulate fibroblast behavior during wound healing. Further,
there is a difference between “normal” wound healing, where the wound is grown
over, and the development of scar tissue.
• Physical forces are also known to be a critical factor in the regulation of the tissue-
specific differentiation of adult and embryonic stem cells. For example, it is thought
that the beating of some mammalian embryonic hearts is more for shaping the
 CELL MECHANICS AND HUMAN DISEASE 5

heart muscle rather than for functional pumping, given that the heart does not
need to pump blood in any serious manner in utero.
• Post-birth, brain development and angiogenesis all centrally involve cells’ ability to
interact with their dynamic mechanical environment.
• Cardiovascular diseases such as hypertension and heart failure often result from
long-term mechanical influences. Indeed, cardiac hypertrophy is one of the most
common responses to changes in forces. The distinction between healthy hypertro-
phy (resulting from exercise) versus pathological hypertrophy (resulting from poor
health) is still not well understood.
• The fundamental cellular processes of membrane trafficking, endocytosis and exo-
cytosis (the ways in which a cell engulfs or expels substances, respectively), micro-
tubule assembly and disassembly, actin polymerization and depolymerization,
dynamics of cell–matrix and cell–cell adhesions, chromosome segregation, kine-
tochore dynamics (such as DNA motion during cell division), cytoplasmic protein
and vesicle sorting and transport, cell motility, apoptosis (“programmed cell
death”), invasion (motion of a cell to where it is not usually located), and prolifera-
tion and differentiation (specialization of a cell to a phenotype with a particular
function) are all regulated, at least in part, by mechanical forces.

In the sections that follow, we examine a few of these examples in more detail.

Specialized cells in the ear allow you to hear

At its most basic level, hearing is a process of transduction (transduction being the
conversion of a signal from one type to another). A physical signal in the form of
sound (pressure) waves is converted into electrical impulses along a nerve.
Mechanotransduction (transduction in which the incoming signal is mechani-
cally based) occurs in the ear via a specialized cell called the inner ear hair cell.
This cell has small hairs called cilia (singular: cilium) extending from the apical
(top) surface of the cell into the lumen of the cochlea. Sound in the form of pres-
sure waves caused by vibrations of the inner ear bones travels through the fluid in
the cochlea.
Investigators have recently deduced the remarkable mechanism of transduction
in the hair cell. Filaments (fibers) of the cytoskeletal protein actin were identified
linking the tip of one cilium to the side of an adjacent cilium (Figure 1.1). The
actin filaments are anchored to proteins that span the cell membrane and form
small holes or pores known as channels. These channels are normally closed, but,
when open, permit the passage of small ions (in the case of hair cells, calcium

(A) (B) (C) pressure

wave
ion channel
Ca2+

Figure 1.1 Hearing occurs via mechanotransduction by the inner ear hair cell. (A) The bundle of cilia extending from the apical
surface of the cell is deflected by pressure waves in the cochlear lumen. (B) Tiny actin bundles called tip links are stretched as the cilia
deflect due to the pressure wave. (C) The tip links are attached to calcium channels or pores that open and allow calcium into the cell
where it eventually leads to a nerve impulse. (A, Courtesy of Dr. David Furness; B, from, Jacobs RA, Hudspeth AJ (1990) Symp. Quant. BioI.
55, 547–561. With permission from Cold Spring Harbor Press.)
6 CHAPTER 1: Cell Mechanics as a Framework

ions) along their concentration gradient. In the resting state, the cell keeps its
internal calcium level extremely low (<1 mM) relative to the calcium concentra-
tion outside the cell. When sound is transmitted to the inner ear, the vibrations
cause the cilia to deflect, which in turn stretches the actin filaments. This stretch-
ing creates tension that is transmitted to the channels, causing them to open. So,
when the channel is opened, calcium flows down its concentration gradient, and
the intracellular calcium concentration increases. The kinetics of signaling pro-
teins inside the cell are altered by this change in concentration, and a cascade of
biochemical events is initiated that eventually leads to a depolarization of the cell
and a nerve impulse.
As you might imagine, mechanics is very important in this process. The cilia
need to have the right mechanical characteristics to stand upright, but remain
flexible enough that they can be deflected by sound waves. The actin tip links
need to be strong enough to open the channel and to have the appropriate poly-
mer mechanics behavior so that they are stretched by cilium deflection, but are
not affected by thermal noise (recall that these are very small objects, so the
soup of molecules floating around will periodically collide with them, and can
generate some forces that need to be ignored). In this text, our goal is to build a
foundation and present a framework so that you can consider these questions
effectively.

Hemodynamic forces regulate endothelial cells

Blood vessels are not passive piping for the blood. They are very responsive and
are constantly changing their radius (via vascular tone or under the influence of
vasodilators and vasoconstrictors) and leakiness. The cells lining these vessels are
called endothelial cells (or collectively, the endothelium). Endothelial cells are
very responsive to mechanical forces generated by the circulatory system, includ-
ing the shear from flow, stretch from the distension of the (larger) vessels, and
transmural pressure differences (pressure differences between the inside of the
vessel and outside). The response of the endothelial cells is varied—they can
change shape to align their long axis in the direction of flow, alter their internal
structure (the cytoskeleton and adhesive plaques), and release a variety of signal-
ing molecules. These actions help maintain blood flow and homeostasis (mainte-
nance of physiological conditions at some baseline), and there is strong evidence
that pathophysiological changes (such as atherosclerosis) occur in regions where
mechanical signaling is disrupted.

To keep bone healthy, bone cells need mechanical stimulation

Physical loading is critical for skeletal health. Indeed, one of the most important
factors in keeping bone healthy is for it to receive normal mechanical stimulation.
When bone is not loaded it is said to be in a state of partial disuse, perhaps owing
to a sedentary lifestyle, or complete disuse, which might occur because of bed rest
or during long-duration spaceflight. In these extreme latter cases bone loss has
been documented to occur at rates as high as 1–2% of total bone mass per month.
Bone loss puts people at increased risk of fracture, even when trauma is absent or
relatively low. These fragility or osteoporotic fractures can be devastating both to
individuals and as a public health issue, costing billions of dollars annually. In
fact, one-half of all women and one-quarter of all men older than 50 today will
experience an osteoporotic fracture in their lifetime. Hip fractures are the most
devastating result of low bone mass, and for most patients the first step in a down-
ward spiral of lost ambulation, lost independence, institutionalization, and
secondary medical morbidity and mortality. Shockingly, within 1 year of a hip
fracture, 50% of patients will be unable to walk unaided, 25% will be institutional-
ized, and 20% will have died.
 CELL MECHANICS AND HUMAN DISEASE 7

The good news is that physical loading on your bone from staying active will pro-
tect you from losing bone, although some activities appear to be better than oth-
ers. Ballet is better than swimming, presumably because of the impact loading
involved. In fact, it has been shown that high-level athletes can actually build
bone specifically in regions of the skeleton that experience higher loading during
their sport. Despite its critical importance for human health and its status as a
compelling scientific question, the mechanism that allows bone cells (osteocytes
and osteoblasts primarily) to sense and respond to loading by coordinating the
cellular response remains basically unknown. It has been suggested that the sens-
ing mechanism might involve the cytoskeleton, focal adhesions, adherens junc-
tions, membrane channels, and even the biophysical behavior of the membrane
itself. Indeed there is evidence for each of these and many others, so it seems
likely that several cellular sensors exist, perhaps forming a redundant system.

The cells that line your lungs sense stretch

During respiration, the lung is exposed to constant oscillatory stresses arising
from expansion and contraction of the basement membrane. These mechanical
signals are postulated to play an important role in maintaining normal lung func-
tion and morphology. Stretch regulates pulmonary epithelial cell growth and
cytoskeletal remodeling, as well as secretion of signaling molecules and phospho-
lipids. These mechanical loads may be increased, for example, when a patient is
subjected to mechanical ventilation. The physiological consequences of altered
cellular function in response to such perturbations in mechanical loading are not
yet fully understood.

Pathogens can alter cell mechanical properties

Malaria provides an interesting example of subtle mechanical alterations at the
cellular level. Malaria is a mosquito-transmitted parasite that infects red blood
cells (RBCs). Because the parasite resides in the RBCs during a large part of its
lifetime, it is generally protected from the immune system. Because infected RBCs
can be destroyed by the spleen, the parasite causes the infected RBC to increase
its stickiness by inducing the expression of adhesive surface proteins on the RBC
membrane. This allows the RBC to stick to the vessel walls and avoid being filtered
in the spleen. Because there are many variations of this class of malaria surface
proteins, the immune system is slow to adapt and remove these infected RBCs. As
you can imagine, there has to be some deftness in the change in adhesion so that
the cells will tend to stick a bit more, but not so much more that all the blood
clumps together. Indeed, one effect of having stickier RBCs is that occasionally
there will be an accumulation of RBCs in smaller blood vessels, resulting in a
hemorrhage.

Other pathogens can use cell mechanical structures

to their advantage
Bacteria of the genus Listeria act similarly by hiding within cells to evade the
immune system. To invade other cells, the bacteria take over part of the cell’s actin
machinery (part of the cell cytoskeleton). Actin is polymerized to form fibers within
the cell to provide structure and anchorage. The bacteria “sit” on the tip of the
growing actin polymer and wait for a polymer to grow long enough for the bacteria
to be pushed out of the cell and into an adjacent cell. Once the host is infected, the
bacteria can spread throughout the host’s body without ever exposing themselves
to the immune system. For this mechanism to work, the bacteria must be able to
achieve sufficient force to break through two cell membranes. The bacteria have to
be able to “know” where to sit on the actin filament to be ­propelled—this is a source
of active investigation for use in generating molecular machines.
8 CHAPTER 1: Cell Mechanics as a Framework

Cancer cells need to crawl to be metastatic

Cancer metastasis is the process whereby an individual cancer cell(s) detaches
from the main tumor, enters the bloodstream, reattaches at some new location,
exits the blood vessel, and starts growing in its new location. Metastasis causes
most cancer deaths, but many aspects of this process have yet to be fully under-
stood. Cell migration is a critical component that is mediated by mechanical pro-
cesses such as adhesion and intracellular force generation. Changes to aspects of
these processes (such as the migratory speed of the cells) are generally tied to the
long-term prognosis of the cancer, but the ways in which this occurs are not well
understood. Adhesion may not only be important for allowing cancer cells to
migrate but also for them to home in on a particular location. Not all tumor cells
metastasize the same way; certain tumors will preferentially metastasize to spe-
cific regions or tissues. Whether this is due to selective adhesion at the preferred
sites or diminished survival at other sites is not clear.
Solid tumors, as a whole, also exhibit altered physiology. Not only do the cells
within tumors exhibit increased (“out-of-control”) division rates, but tumors can
redirect blood flow to allow themselves to grow faster. Further, many primary
solid tumors tend to be stiffer than the surrounding tissues, even though they gen-
erally originate from the same tissue mass. Whether this increased stiffness alters
cancer cell function through a mechanosensing function is not known, but it does
have a practical use—many superficial (i.e., close to the skin) tumors can be
detected by performing a self-examination, by feeling for a “lump” or “bump” that
is somewhat harder compared to the surrounding tissue.

Viruses transfer their cargo into cells they infect

When a cell is invaded by a virus, the viral cargo of genetic material must be intro-
Nota Bene
duced into the cell. There are two mechanisms by which this can occur: endocy-
Membrane fusion is a chemical tosis or membrane fusion. In the case of the former, the binding of proteins (called
process by which a pore is ligands) on the surface of the virus to proteins (called receptors) on the surface of
introduced into the cell membrane at the cell initiates a process called receptor-mediated endocytosis. In this process,
the point where the membranes of
the virus and the target cell are fused
the virus is enveloped by the cell, allowing it to deliver its genetic cargo and repli-
together. The mechanism underlying cate. This process relies on a coordinated sequence of mechanical events, includ-
this is not understood well and may ing adhesion, membrane pinching, and generation of cytoskeletal force that may
not be as mechanically dependent as provide potential targets for therapeutics aimed at inhibiting viral invasion. In
endocytosis. But a certain degree of addition, given the highly efficient means by which viruses invade cells, there is
adhesion and membrane bending will interest in understanding these mechanical processes for purposes of biomim-
invariably have to occur.
icry, such as virus-based methodologies for nanoparticle delivery into cells.

1.2 THE CELL IS AN APPLIED MECHANICS

GRAND CHALLENGE
In the twentieth century, the state of the art for applied mechanics was structural
analysis on the large scale. Amazing achievements were realized in construction,
such as high-rise buildings and beautiful bridges. Architecture was allowed to
move beyond bulky stone and brick to elegant steel and glass. Transportation was
revolutionized with cars and trains and modern aircraft representing outstanding
examples of highly efficient structures that could only be created once their
mechanical behavior had been analyzed in detail—engines, streamlining, brakes,
lift, power, heat, etc., all having to be characterized and then applied together.
Mechanics also played a major role in allowing people to reach the moon and
explore the planets. Mechanics was and is key to military advances (missile tech-
nology, armor, advanced planes and drones, robotics, etc.). However, the theories
and analysis required to design and build these impressive structures are, to a
great extent, mature. For instance, much of car body design is based on computa-
tional analysis and not on the development of new laws or principles. Although
 MODEL PROBLEM: MICROPIPETTE ASPIRATION 9

applied mechanics experienced great growth in the past, more recently it has
undergone some degree of contraction.
Comprehensive mechanical analysis of a cell is extremely complex. There are
one-dimensional linear elements in the cytoskeleton and two-dimensional
curved shells in the cell membrane. There are also three-dimensional solids and
enormous potential for pressure effects and fluid–solid interaction. Indeed, the
overall cell is part–solid, part–liquid, something we call viscoelastic. Not only that,
but the properties of the cellular “substance” change depending on the frequency
with which forces are applied. On top of this, cellular structures are so small that
thermal and entropic effects can play an important role in their mechanics, often
requiring the analytical framework of statistical mechanics to understand their
behavior. Thus, in terms of difficult challenges in applied mechanics with poten-
tial for critical new advances and fundamental insight, it is hard to imagine a more
compelling problem than the cell. The overall picture we wish to present to you is
that much can be explained using basic mechanics, but there is still much to be
done using only slightly more advanced mechanical analysis.

Computer simulation of cell mechanics requires state-of-the-

art approaches
Just as cell mechanics is a compelling challenge in applied mechanics, it is also a
difficult, but rewarding, challenge in computational mechanics. For example,
multi-scale modeling involves coupling a large-scale simulation with another
simulation representing microscopic behavior. For cells one might simulate the
behavior of individual actin and tubulin polymers and couple that to models of
cytoskeletal networks or even the whole cell. The full mechanical behavior of the
cell is a synthesis of solid, fluid, and statistical mechanics such that there is an
opportunity for multi-physics formulations. There is also potential for fluid-
structure problems, contact, even nonlinear material models. There is hardly an
area within advanced computational mechanics that does not have application
within cell mechanics.

1.3 MODEL PROBLEM: MICROPIPETTE

ASPIRATION
We conclude this chapter with a simple model problem to give you a first glimpse
into approaches for investigating cell mechanics and what sort of understanding
we can gain from these analyses. Micropipette aspiration was one of the first
methods used to examine cellular behavior and is responsible for some remarka-
bly important and surprising insights into cellular behavior. Micropipette aspira-
tion involves relatively simple instrumentation, and the experimental analysis
can be very straightforward. The introduction of micropipette aspiration here is
meant to foreshadow the level of abstraction and rigor that will follow throughout
the text.

What is a typical experimental setup for

micropipette aspiration?
Some of the earliest mechanical measurements of cell membranes were made
using micropipette aspiration experiments. These measurements were based
partly on the concept that cells were pouches with fluid interiors (use of RBCs
eliminated the problem of the nucleus, because RBCs have none). The versatility
of micropipette aspiration and ease of interpretation of experimental results con-
tinue to make this an important experimental approach for studying the mechan-
ics of cells (and not just RBCs). As we will see, these experiments not only allow
one to make measurements of cell membrane mechanical properties, but they
also provide insight into the mechanical behavior of whole cells.
10 CHAPTER 1: Cell Mechanics as a Framework

Figure 1.2 Red blood cell being

drawn into a micropipette. (Courtesy
of Richard Waugh, University of
Rochester.)

A micropipette is a rigid tube (usually glass) that tapers to a diameter of several

micrometers at the tip (near the tip, the diameter is constant). It is hollow all the
way through its length and a suction (negative) pressure is applied to the inte-
rior (the lumen). If the end is brought in proximity to a cell while suction is
applied, a seal will form, and the cell will be drawn into the micropipette, form-
ing a protrusion (Figures 1.2 and 1.3). The negative pressure can be applied in
a variety of ways. One way is to apply the suction by mouth—this method actu-
Nota Bene ally provides a lot of control, and researchers commonly do this when forming
the seal.
Micropipette aspiration is harder
in winter. Early investigators Another common way is to connect the micropipette to tubing that runs to a
conducting micropipette experiments water-filled reservoir with controllable height. In this case, decreasing the height
found that the drift was much worse
of the fluid surface in the reservoir relative to the height of the fluid surface in the
in the winter. Why? The evaporation
rate at the top of the fluid reservoir dish in which the cells are cultured creates a suction pressure within the micro-
was faster because the air tends to pipette. In theory, the minimum suction pressure that can be applied is deter-
be drier in winter and the evaporation mined by the minimum change in height of the fluid reservoir that can be
rate was therefore higher. achieved (typically on the order of ~0.01 Pa). In practice, the resolution is worse
Contemporary engineering tools have (usually on the order of ~1 Pa), owing to drift caused by water evaporating from
made this reservoir-associated the reservoir. Typically, the maximum pressure that can be applied is on the
problem obsolete.
order of atmospheric pressure, resulting in a wide range of forces, from ~10 pN to
~100 nN.
Once the cell is drawn into the micropipette, the morphology of the cell relative to
the pipette can be divided into three regimes, as in Figure 1.4. The first regime is
when the length of the protrusion of the cell into the pipette Lpro is less than the
radius of the pipette Rpip, or Lpro/Rpip < 1. The second regime is when the protru-
sion length is equal to the pipette radius, or Lpro/Rpip = 1, and the protrusion is
hemispherical. The third regime is when Lpro/Rpip > 1, and the protrusion is cylin-
drical with a hemispherical cap. The radius of the hemispherical cap is Rpip, since
the radius of the protrusion cannot change once the hemispherical cap is formed.

Figure 1.3 A micropipette aspiration 3 protrusion forms

experiment. The micropipette tip is
placed in the proximity to the cell, and
suction pressure is applied. A seal forms
between the cell and the micropipette,
forming a cell protrusion into the 1 suction pressure
micropipette.

2 cell forms seal 

 MODEL PROBLEM: MICROPIPETTE ASPIRATION 11

Figure 1.4 Three regimes of cell

aspiration into a micropipette.
Rpip (Upper) The length of the protrusion of
Lpro/Rpip < 1 the cell into the pipette is less than the
radius of the pipette, or Lpro/Rpip < 1.
(Middle) Lpro/Rpip = 1. In this case, the
Lpro protrusion is hemispherical. (Lower) Lpro/
Rpip > 1. The protrusion is cylindrical with
a hemispherical cap of radius Rpip.

Rpip
Lpro/Rpip = 1

Lpro

Rpip
Lpro/Rpip > 1

Lpro

At this point one should be able to see that geometrically the radius of the protru-
sion of the cell in the first regime is larger than Rpip.

The liquid-drop model is a simple model that can explain

some aspiration results
When researchers performed early micropipette aspiration experiments on
cells such as neutrophils (a type of white blood cell), they noticed that after the
micropipette pressure exceeded a certain threshold, the cells would continu-
ously deform into the micropipette (in other words, the cells would rapidly
“rush into” the pipette). This observation led to the development of the liquid-
drop model. In this model, the cell interior is assumed to be a homogeneous
Newtonian viscous fluid, and the surrounding membrane is assumed to be a
thin layer under a constant surface tension, and without any bending resist-
ance. Further, it is assumed that there is no friction between the cell and the
interior walls of the pipette.
Surface tension has units of force per unit length, and can be thought of as the
tensile (stretching) force per unit area (stress) within the membrane, integrated
through the depth of the membrane. For example, if a membrane of thickness d is
subject to a tensile stress σ that is constant through its depth, then we can model
the membrane as having a surface tension of n =  σd. If the membrane is very thin
compared to the radius of the cell, we can ignore the membrane thickness for
analysis and rely exclusively on the surface tension n (Figure 1.5).
Why is this model called the liquid-drop model? A drop of cohesive liquid (such
as water) suspended in another less cohesive fluid (such as air) has a thin layer of
water molecules at the surface of the drop, and, because of the imbalance of
intermolecular forces at the surface, packs together and causes surface tension.
In brief, every molecule exerts, on average, an attractive force on every other
molecule. So, a molecule deep within the drop is pulled in every direction about
equally and has no net average force on it. However, a molecule at the surface is
pulled with a net force into the “bulk” of the drop, resulting in a spherical shape
for the drop and some resistance to deformation at the surface, which we call
12 CHAPTER 1: Cell Mechanics as a Framework

Figure 1.5 A membrane of thickness d

d subject to a tensile stress σ that
is constant through its depth can
be modeled as infinitely thin with a
surface tension of n = σd. n = σd
σ

surface tension. This surface tension is what allows some types of insects to walk
on the surface of water.

The Law of Laplace can be applied to a spherical cell

By modeling the cell as a liquid drop, we can analyze micropipette aspiration
experiments using the Law of Laplace, which relates the difference in pressure
between the inside and outside of a thin-walled pressure vessel with the surface
tension within the vessel wall. It can be derived from a simple analysis using a free
body diagram, a topic discussed in detail in Chapter 3. Consider a spherical thin-
walled vessel with radius R, a pressure of Pi inside the vessel, and a pressure of Po
outside the vessel (Figure 1.6).
If we cut the sphere in half, then there are two equal and opposite resultant
forces acting on the cut plane. The first is due to pressure, and it is calculated as
Fp = (Pi - Po)πR2. The second resultant force is due to surface tension on the
wall. If the surface tension is given by n, then the resultant force due to surface
Nota Bene tension Ft is s Ft = n2πR (the surface tension multiplied by the length of the edge
exerting such tension, which would be the circumference of the circle). Setting
History of Young and Laplace. The
Fp = Ft, we arrive at the Law of Laplace,
Law of Laplace is also known as the
Young–Laplace equation in honor of
2n
Thomas Young, who made initial Pi − Po = . (1.1)
qualitative observations of the R
curvature of liquid menisci in 1804,
and Pierre-Simon Laplace, who later Note that we set the forces equal because we assume that the droplet is not accel-
introduced the mathematical erating. In this case, Newton’s second law implies that the forces due to pressure
formalism. It is also sometimes called
the Young–Laplace–Gauss equation.
and surface tension must sum to zero. Because they act in opposite directions, the
forces are equal.

Micropipette aspiration experiments can be analyzed with

the Law of Laplace
We can use the Law of Laplace to analyze micropipette aspiration experiments by
relating suction pressure to the morphology of the cell as it enters the pipette.
Consider the configuration in Figure 1.7, where Patm is the pressure of the envi-
ronment, Pcell is the pressure within the cell, Ppip is the pressure within the pipette,
Rcell is the radius of the cell outside the pipette, Rpip is the radius of the pipette,

Figure 1.6 A spherical vessel with

an inner pressure exceeding the
outer pressure balanced by a
membrane tension n (left) and the
resulting free-body diagram (right).
Pi n

Po
 MODEL PROBLEM: MICROPIPETTE ASPIRATION 13

Figure 1.7 Schematic depicting

quantities used in analyzing
Rcell a micropipette aspiration
Rpro Rpip
Patm Ppip experiment.

Pcell
Lpro

Rpro is the radius of the protrusion, Lpro is the length of the protrusion into the
micropipette and Rpip is the radius of the micropipette.
For the portion of the cell that is not in the micropipette, from Equation 1.1 we
know that
2n
Pcell − Patm = , (1.2)
Rcell

where n is the surface tension of the cell. For the protrusion, or the portion of the
cell in the pipette,
2n
Pcell − Ppip = . (1.3)
Rpro

Combining equations 1.2 and 1.3, we obtain

 1 1 
Patm − Ppip = ∆P = 2n  − , (1.4)
 Rpro Rcell 

which relates the difference between the pressure in the surroundings and in the
pipette with the radius of the cell inside and outside of the pipette for a cell with a
given surface tension. Note that we assume the surface tension is constant through-
out the cell, even in the “crease” region where the cell contacts the micropipette,
and in the protrusion area and the cell membrane outside the micropipette.

How do we measure surface tension and areal

expansion modulus?
One can now easily measure surface tension from Equation 1.4. Once the cell is
drawn into the micropipette such that the protrusion is hemispherical (Lpro = Rpip),
then the radius of the protrusion also equals the radius of the pipette, Rpro = Rpip.
Here,

 1 1 
∆P = 2n  − . (1.5)
 pip
R Rcell 

The pressure ΔP is controlled by the user, and Rpip is also known. The cell radius
Rcell can be measured optically under a microscope, allowing one to calculate the
surface tension n. Evans and Yeung performed this experiment using different
micropipette diameters and found a surface tension of (~35 pN/μm) in neutro-
phils. This tension was found to be independent of the pipette diameter, which
supports the validity of the technique. We will see in the next section that surface
tension does not stay exactly the same if the cell is further deformed.
For a liquid drop, the surface tension will remain constant as it is aspirated into a
micropipette. In reality, cells do not behave like a perfect liquid drop. This is
because the cell membrane area increases as it is aspirated, resulting in a slight
increase in surface tension. The increase in tension per unit areal strain is given
14 CHAPTER 1: Cell Mechanics as a Framework

by what is called the areal expansion modulus. Needham and Hochmuth quanti-
fied the areal expansion modulus in neutrophils by aspirating them through a
tapered pipette (Figure 1.8). Applying progressively higher pressures, caused the
cell to advance further into the taper, increasing its surface area while main­
taining constant volume (maintaining constant volume is a condition called
incompressi­bility). The radii at either end of the cell, Ra and Rb, the total volume
V, and the apparent surface area A were measured from the geometry. By appar-
ent we mean that the surface area of the cell is approximated to be smooth, and
we ignore small folds and undulations. The surface tension was calculated using
the Law of Laplace as

 1 1
∆P = 2n  − . (1.6)
 Ra Rb 

The “original” radius Ro of the cell was calculated from the volume (assuming the
volume remained constant) as V = 4 3 πRo3, allowing the “original” or undeformed
apparent surface area to be calculated as Ao = 4πRo2 . The areal strain ( A − Ao )/Ao
and surface tension could therefore be found for the same cell as the pressure was
increased, and the cell advanced through the taper. The surface tension was plot-
ted as function of areal strain, and the data was fitted to a line. The areal expansion
modulus was found from the slope of the line and was calculated to be 39 pN/μm.
Extrapolating the fit line to zero areal strain resulted in a resting surface tension of
24 pN/μm in the undeformed state.

Figure 1.8 Cell being aspirated (A)

within a tapered pipette. The radius
of the pipette opening is 4 μm. (A) A cell
was aspirated into the tapered pipette
and allowed to recover to its resting
spherical shape. A positive pipette
pressure was then applied, and the cell
was driven down the pipette. Final resting
configuration after (B) ΔP = 2.5 Pa, (C)
ΔP = 5.0 Pa, and (D) ΔP = 7.5 Pa. (Adapted (B)
from, Needham D & Hochmuth RM (1992)
Biophys J. 61, 1664–1670.)

(C)

(D)
 MODEL PROBLEM: MICROPIPETTE ASPIRATION 15

Figure 1.9 Electron micrograph of

a neutrophil. Ruffles in the membrane
can be clearly seen. (Adapted from,
Needham D & Hochmuth RM (1992)
Biophys J. 61, 1664–1670.)

Why is this tension in the undeformed state important in neutrophils? Remember

that neutrophils circulate in the blood and therefore need to squeeze through
small capillaries with diameters smaller than the cells themselves, similar to RBCs.
As neutrophils squeeze through capillaries, the shape of the cells transform from a
sphere into a “sausage” (i.e., a cylinder with hemispherical caps at both ends).
Because the cells contain mostly fluid, they are usually incompressible and so
must maintain constant volume during this shape change. The surface area of the
“sausage” is greater than a sphere of the same volume, and the increase in surface
area grows larger as the radius of the “sausage” decreases. However, we will see
later that biomembranes are quite inextensible. So how do neutrophils undergo
this increase in surface area when squeezing through small blood vessels?
As can be seen in Figure 1.9, neutrophils contain many microscopic folds in their
membrane. What this means is that their “apparent” surface area is much less
than the actual surface area of the membrane if one were to take into account all
the folds and ruffles. The folds allow neutrophils to substantially increase their
apparent surface area without actually increasing the surface area of the mem-
brane, as long as the folds are not completely smoothed out. The tension within
the cortex of these cells has a crucial role: it allows the cells to have folds and ruf-
fles in the membrane while maintaining their spherical shape.

Why do cells “rush in”?

Remember that the liquid-drop model was developed in large part in response to
the observation that some cells would “rush in” after applying any pressure greater
than the critical pressure at which Lpro = Rpip. Why do liquid drops do this?
Remember that Equation 1.4 is a relation that must be satisfied for equilibrium.
Suppose we apply the critical pressure such that Lpro = Rpip, and then we increase
ΔP. Let us examine what happens to the terms on the right-hand side of Equation
1.4. We already learned that n is constant for liquid drops, and approximately con-
stant for neutrophils as they are aspirated. The radius of the protrusion, Rpro, will
also remain constant, because the radius of the hemispherical cap will be equal to
Rpip for any Lpro > Rpip. Rcell cannot increase, meaning that the volume of the cell
remains essentially constant over the time of the experiment. We therefore have
increased the left-hand side of Equation 1.4 with no way to increase the right-
hand side. The result is that equilibrium cannot be satisfied. This produces an
instability, and the result is the cell will rush into the pipette.
16 CHAPTER 1: Cell Mechanics as a Framework

Figure 1.10 ΔP as a function of cannot distinguish between

Lpro = Rpip for a cell that behaves elastic solid and liquid drop
like an elastic solid (dotted line)
and a cell that behaves like a liquid
drop (solid line). When Lpro = Rpip = 1,
an instability occurs for the cell that
behaves like a liquid drop, and the
cell rushes into the micropipette. By elastic solid: no instability
contrast, a cell that behaves like an

∆P
elastic solid will not have this instability.
liquid drop: instability

1
Lpro /Rpip

Cells can behave as elastic solids or liquid drops

Micropipette aspiration is an extremely versatile technique for measuring the
properties of membranes. It can apply a wide range of forces, and the experiments
are conducive to mechanical analysis. In addition to making mechanical measure-
ments of membranes, micropipette aspiration experiments play perhaps an even
more important role in understanding cell mechanics, in that they easily allow
insight into the fundamental mechanical behavior of different cell types. For exam-
ple, when researchers performed experiments on endothelial cells or chondro-
cytes, they found that they would not rush into the pipette, even after Lpro = Rpip.
Why? Put simply, these cells do not behave like a liquid drop. Subsequent experi-
ments and analyses showed that their mechanical behavior is much more like that
of an elastic solid, so it will not have this instability. Therefore, by observing whether
a particular cell does or does not rush into the pipette after Lpro = Rpip, one can eas-
ily distinguish whether its mechanical behavior is more like a liquid drop or an
elastic solid. Note that this simple method for classification is only possible if the
critical pressure is exceeded. If the experiment is terminated before Lpro = Rpip,
then one cannot (as easily) distinguish between elastic solid and liquid-drop
behavior, as can be seen in Figure 1.10.

Key Concepts

• The study of cellular mechanobiology bridges • Micropipette aspiration is an early and straightforward
cell biology and biochemistry with various approach for investigating cell mechanics. By modeling
disciplines of mechanics, including solid, fluid, the cell as a liquid drop, we can analyze micropipette
statistical, experimental, and computational aspiration experiments using the Law of Laplace.
mechanics. • Liquid-drop cells exhibit an instability when the radius
• A wide variety of devastating human diseases such as of the aspirated protrusion is equal to the radius of
osteoporosis, heart disease, and even cancer involve the pipette. At this point the cell cannot resist
cell mechanics in a fundamental way. increased pressure and rushes into the pipette.
• Cell mechanics is an excellent substrate for • By observing the movement of cells within the
introducing students to a wide variety of cutting-edge micropipette at the instability point, one can
approaches in mechanics. distinguish two types of cellular behavior cells: those
• Understanding the mechanical behavior of cells whose behavior is dominated by membrane tension
presents a grand challenge in theoretical, (similar to a liquid drop), and those that act as a
computational, and experimental mechanics. continuum solid.
 Annotated References 17

Problems

1. As we have seen, when a liquid drop–type cell such as known as alveoli. There are around 150 million alveoli
a neutrophil is subjected to micropipette aspiration, in your lungs. During respiration the alveoli are filled
it becomes unstable. This occurs when the radius of and emptied through the action of the diaphragm and
the protrusion is equal to the radius of the pipette. intercostal muscles in the chest wall. During inhalation
For a cell whose behavior is better approximated by a the pressure outside the alveoli can drop by up to 200
continuum such as a chondrocyte, do you think there is Pa. The layer of cells that line the alveoli is constantly
a maximum pressure that can be exerted on the cell? hydrated. So, we can model them as a bubble of air
What configuration would the cell be in at this point? surrounded by water, assuming that the surface tension
of water is 70 dyn/cm. With this information, what is
2. In our analysis we ignored the friction between the radius of the alveoli? In actuality, the radius of an
the inside of the pipette and the cell wall. Is this a alveoli is about 0.2 mm. They can be this small because
reasonable assumption? Why? How might friction affect the epithelial cells secrete a protein called surfactant
the results of an aspiration experiment if it were large? that reduces the water surface tension. If fact, there is
a developmental condition known as infant respiratory
3. The classic micropipette aspiration experiment can
distress syndrome (IRDS) that occurs when insufficient
be modified to address the instability of a cell with
surfactant protein is produced. What must the surface
liquid-drop behavior so that membrane behavior can be
tension be to keep the alveoli inflated?
measured. This is done with a tapered pipette (Figure
1.8). With this approach the pipette radius changes 5. Consider two adjacent alveoli and the end of a
along its length. Derive the relationship between bronchus such that air can pass easily between them
membrane tension and the pressure in the micropipette as well as to the outside. Assume that they have the
assuming that the two sides of the cell have radii of Ra same radius and that they are in a state of equilibrium
and Rb and pipette pressures are Pa and Pb. with respect to pressure and surface tension. What
would happen if a small volume of air moved from one
4. There are other structures similar to cells that can
alveolus to the other? What must be true of surfactant
be treated with the liquid-drop model. The bronchial
as a result?
passages of the lungs terminate in small spherical sacs

Annotated References

Evans E & Yeung A (1989) Apparent viscosity and cortical tension of documenting the prevalence of, as well as the social and economic
blood granulocytes determined by micropipet aspiration. Biophys. J. costs associated with, musculoskeletal disease.
56, 151–160. An early report on micropipette aspiration can be used
Janmey PA & Miller RT (2011) Mechanisms of mechanical signal-
to determine membrane tension.
ing in development and disease. J. Cell Sci. 124, 9–18. This article
Hochmuth RM (2000) Micropipette aspiration of living cells. J. Biome- discusses what is known about the ability of cells to sense local
chanics 33, 15–22. An excellent review of the use of micropipette stiffness and includes a discussion of diseases and developmental
aspiration to characterize the fundamental behavior of cells and processes.
membranes.
Krahl D, Michaelis U, Peiper HG et  al. (1994) Stimulation of bone
Huang H, Kamm RD & Lee RT (2004) Cell mechanics and mecha- growth through sports. Am. J. Sports Med. 22, 751–157. Early evi-
notransduction: pathways, probes, and physiology. Am. J. Physiol. Cell dence that loading of bone through physical activities leads to bone
Physiol. 287, C1–11. An overview of mechanosignaling with some formation.
discussion of techniques as well as an overview of cell-force interac-
Malone AM, Anderson CT, Tummala P et  al. (2007) Primary cilia
tions.
mediate mechanosensing in bone cells by a calcium-independ-
Jacobs CR, Temiyasathit S & Castillo AB (2010) Osteocyte mecha- ent mechanism. Proc. Natl. Acad. Sci. USA 104, 13325–13330.
nobiology and pericellular mechanics. Annu. Rev. Biomed. Eng. 12, Early evidence that primary cilia act as mechanosensors in bone
369–400. An extensive review of how mechanics and biology inter- cells.
act at the level of the cell to regulate the skeleton in health and
Needham D & Hochmuth RM (1992) A sensitive measure of sur-
disease. Includes extensive supplemental online material.
face stress in the resting neutrophil. Biophys. J. 61, 1664–1670. An
Jacobs J (2008) The Burden of Musculoskeletal Disease in the United early demonstration of how micropipette aspiration can be used to
States. Rosemont, IL. This monograph is an extensive collection determine membrane properties.
This page intentionally left blank
to match pagination of print book
CHAPTER 2
Fundamentals in
Cell Biology

W hen thinking about cell mechanics, it is easy to get caught up in models

and mathematics, and thinking about cells as somewhat exotic material
that can be subjected to the same sort of testing as is performed on inert metals
and plastics. However, an equally important part of cell mechanics is the study
of how mechanics interacts with the biological behavior of cells. The former is
termed “biomechanics”; the latter is increasingly referred to as “mechanobiol-
ogy,” as mentioned in Chapter 1. The former is meant to emphasize biomechan-
ics as the subdiscipline of mechanics that considers the mechanical properties
of biological structures and mechanobiology as the subdisipline of biology that
is focused on how mechanics regulates biological processes, or how biological
processes generate and regulate physical forces. Oftentimes this distinction is
artificial and may not be useful, as some of the most fascinating problems can-
not be understood without considering biology and mechanics with equal
depth and rigor.
Just as our treatment of cell biomechanics is built on a firm understanding of the
fundamental principles of mechanics, to appreciate mechanobiology fully we
must take a step back and form an understanding of the fundamentals of biology.
This must include not only the underpinnings of cell and molecular biology, but
also some of the modern experimental techniques that allowed these insights to
be made. Our goal is not only to provide you with enough biology background to
understand cell mechanics, but to allow you to read and understand the basics of
modern biology scientific publications.
Modern biology is undergoing a revolution of understanding that began in 1953
with Watson and Crick’s determination of the double-helix structure of DNA. This
has led to the sequencing of the human genome, modern genetics, and, in an
amazingly short period of time in terms of the history of science, to dramatic
advances in biology. This explosion of knowledge, known as the molecular revo-
lution, has given rise to the field of molecular biology sometimes referred to as
modern biology. J. Craig Venter, one of the pioneering founders of the high-
throughput approaches to the study of genes and their regulation (who also par-
ticipated via Celera in the Human Genome Project), declared this to be the
opening of the “Century of Biology.” Many have speculated that the discoveries
currently being made will lead to the next wave of social change following the
Industrial Revolution and the Information Revolution.

2.1 ​FUNDAMENTALS In CELL AND MOLECULAR

BIOLOGY
Formal writing in biology is heavy on terminology, in part due to the need to be
precise in descriptions while being succinct. Let us say you develop a hairline
fracture on one of your ribs. If the location of the fracture is closer to the sternum,
we call it a medial fracture (closer to the “middle” of the body). If it is closer to your
sides/arms/shoulders, we call it a lateral fracture. The terms medial and lateral
20 CHAPTER 2: Fundamentals in Cell Biology

Table 2.1 The monomer subunits and resulting polymers that make up the
biochemical constituents of the cell.

Monomer Polymer

nucleic acid RNA, DNA genes

amino acid peptide, protein gene products

fatty acid lipid not coded by genes

sugar polysacharide, carbohydrate not coded by genes

(along with other terms) can be used to quickly describe locations without having
to reference specific sites, much like north and south can be used if you are not
familiar with local landmarks.
In a similar way, there are a few terms that recur in biomedical writing with which
you should be familiar. We may not use them all in this book, but the diligent stu-
dent will want to get to know these terms. Some common ones are:
• Cell culture: the process of extracting live cells from biological tissue, and the
maintenance and growth of those cells, typically in vitro, to study the physiological
behavior of the cells under controlled conditions.
• In vitro: in a laboratory dish; not inside an organism; literally, “in glass”.
• Ex vivo: sometimes interchanged with in vitro, but occasionally referring to entire
tissues that are cultured in a laboratory dish.
• In vivo: inside a living organism, typically but not always at the natural location.
• In situ: inside a living organism, in the natural location. Growing an ear in a mouse’s
back would be an in vivo but not an in situ experiment.
• Amino acids: the fundamental building blocks (monomers) of proteins.
• DNA: deoxyribonucleic acid, the “hard copy” of genes and their regulatory
components.
• RNA: ribonucleic acid, the “working copy” of genes that assembles proteins.
• Gene: sequence of DNA that encodes a protein.
• Promoter: sequence of DNA that regulates when a particular gene is expressed.
Typically, but not always, near the gene being regulated.
• Probe: A fragment of DNA or RNA that is complementary to a specific target
sequence. The probe is able to bind to the target in a process called hybridization.
Many cellular components are polymers, assembled from subunits called mono-
Nota Bene mers (Table 2.1). The major exception to this is the most abundant constituent of
Inorganic carbon compounds. cells, water, which makes up roughly 70% of a cell’s mass. Inorganic ions are
There are several carbon compounds another exception and are critical to cell metabolism and signaling (“organic”
that are inorganic. Diamond and compounds have carbon, “inorganic” compounds do not). These components are
graphite are obvious examples. vital for the non-inert response of cells to mechanical forces. In some cases, bio-
However, there are no hard-and-fast logical polymers form the physical structure of the cell, and the biomechanical
rules to make the distinction. characterization of the cell relies on the properties of these polymers. In other
cases, the molecules are responsible for signaling. Such molecules can be impor-
tant for signaling in response to mechanical stresses (mechanosignaling), which
we discuss further in Chapter 11.

Proteins are polymers of amino acids

In general, proteins consist of many (> 50) amino acids. Short chains of amino
acids or fragments of proteins are called peptides. Each amino acid shares a com-
mon backbone structure of an amino group (H2N), a carbon atom, and a carboxyl
group (COOH) (Figure 2.1). The central carbon, known as the a-carbon, attaches
to a side chain. The diversity of charges, sizes, and interactions of these side chains
 FUNDAMENTALS In CELL AND MOLECULAR BIOLOGY 21

H O H O Figure 2.1 ​Chemical structure of a

polypeptide. Note the repeating
H2N C C OH H 2N C C OH backbone structure and the residues
that distinguish one amino acid from
another.
R R
+ H2O

H O H H O
amino end carboxyl end
N-terminus H2N C C N C C OH C-terminus

R R

distinguishes one amino acid from another. The properties of these side chains
additionally determine the hydrophilic/hydrophobic nature of the local peptide
sequence, which is useful for determining regions of proteins that span mem-
branes and the orientation of the protein. Remarkably, there are only 20 common
amino acids (although there are several more in nature, plus some synthetic ones
that have been created in laboratories). The sequence, or order, in which these
amino acids are assembled gives rise to the remarkable diversity of structure and
function found in proteins. The amino acid sequence is formally referred to as the
primary (1°) structure of the protein.
Amino acids are asymmetric because one side has an amino group and the other
has a carboxyl group. This directionality is preserved in peptides and proteins,
with one end terminated with the amino group, the N-terminal, and the other ter-
minated with the carboxyl group, the C-terminal. When amino acids are assem-
bled, one oxygen from the carboxyl group and two hydrogens (one from each
group) form a molecule of water. This process of assembly is called dehydration
synthesis or condensation. When these bonds are broken, a molecule of water is
utilized in a reaction called hydrolysis.
Once a protein is assembled, it can fold based on the distribution of charges and
steric interactions, the latter being space available for movement based on the
size of the side groups. Some common local folding patterns include a-helices
(single spring-like structures) and b-sheets (mostly flat loops) (Figure 2.2). These
patterns are typically referred to as the secondary (2°) structure, and are mainly a
result of hydrogen bonding. The global folding pattern of the entire protein is
referred to as the tertiary (3°) structure. How proteins fold is important for their
function, because the local shape/charge distribution determines which other
molecules the proteins can interact with. The prediction of folding patterns based
on amino acid sequence is an open problem currently, with computational
­models only able to handle short peptide fragments. Finally, the association of
multiple proteins is termed the quaternary (4°) structure. Molecular dynamics
computer simulations are capable of generating information regarding second-
ary and tertiary structures, but are known for being computationally expensive.
Simulation of quaternary structures, which would be a boon for pharmaceutical
design, but are also extremely computationally expensive.
The mechanics involved in folding or unfolding of proteins is a subject of intense
investigation, because such unfolding may occur in your cells. Not only does the
resistance to unfolding contribute to the “stiffness” of the molecule but there are
signaling pathways that may be activated by protein unfolding, as discussed in
later chapters. Knowing the charges and sterics of proteins and how a protein
folds is not only important in biology, but also in mechanobiology.
22 CHAPTER 2: Fundamentals in Cell Biology

Figure 2.2 Schematic of the two (A) (B)

most common protein secondary
folding structures, (A) an α-helix
and (B) a β-sheet. (Adapted from,
Alberts B, Johnson A, Lewis J et al. (2008)
Molecular Biology of the Cell, 5th ed.
Garland Science.)

carbon

nitrogen

Example 2.1: What determines how hard it is to fold a protein?

Consider the effects of side chains on protein folding and a­ queous environment, is the protein harder or easier to
unfolding. Start with a long protein that is entirely unfold?
uncharged with polar side chains. It takes some force to
coil it up, because of the bending of the bonds between In the first case, with all same charges throughout the
­adjacent amino acids. protein, the protein becomes harder to fold compared
with the uncharged state. This is because of like charges
If you somehow charge the protein with the same charge repelling, so that to fold the protein, not only do you have
throughout the amino acids, is it easier or more difficult to work against the bonds between the amino acids
to fold? bending, but also in bringing charges closer.
Next, take the original uncharged, polar protein, but in In the second case, the protein becomes harder to unfold
its folded configuration. compared with the original protein with polar side chains,
If you introduce hydrophobic side chains such that because hydrophobic groups tend to cluster together in an
large sections of these side chains are clustered together aqueous environment. As a result, it will take more work to
in the folded protein, and keep the protein in an separate the groups from each other to extend the protein.

DNA and RNA are polymers of nucleic acids

Nucleotides are the monomers that make up the genetic polymers DNA (deoxyri-
bonucleic acid) and RNA (ribonucleic acid). As one can deduce from the names,
in RNA the sugar backbone is ribose, and in DNA it is deoxyribose. In both cases,
there is a side group known as a base for its ability to bind hydrogen in aqueous
conditions. For both RNA and DNA there are four types of bases: cytosine (C),
guanine (G), adenine (A), and thymine (T) in DNA or uracil (U) in RNA. Bases are
 FUNDAMENTALS In CELL AND MOLECULAR BIOLOGY 23

building blocks of DNA DNA strand Figure 2.3 ​The various molecular
units that combine to form the
phosphate double-stranded DNA polymer are
sugar shown at different scales, from the
5′ 3′
components making up a
+ G
nucleotide to the assembled double
G G C A T
sugar base helix. (Adapted from, Alberts B, Johnson
phosphate nucleotide A, Lewis J et al. (2008) Molecular Biology
of the Cell, 5th ed. Garland Science.)
double-stranded DNA DNA double helix
3′ 3′
5′ 5′

C G G C

A T T A

T A A T

T A A

G C sugar–phosphate G C
backbone
C G C G

C G C G

A T A

G C C G

T A A T

5′ 5′
3′ 3′

hydrogen-bonded
base pairs

able to form hydrogen bonds with each other, but only in a complementary fash-
ion: C with G, and A with T (DNA) or U (RNA). Long nucleic acids can bind to each
other in the famous double-helix arrangement, but only when they have comple-
mentary sequences (Figure 2.3). Although nucleotides can have many roles in the
cell, they are primarily responsible for the storage of genetic information through
the sequence in which the bases are assembled and replicated d ­ uring cell division
(Figure 2.4).

template S strand Figure 2.4 The ability to replicate is

5′ 3′ one key property of DNA, allowing
it to pass on genetic sequence
information. The parental S strand
S strand 3′ 5′ and its complement S’ strand are
5′ 3′ new S′ strand separated and new (daughter) S
and S’ strands are synthesized.
(Adapted from, Alberts B, Johnson A,
3′ 5′ new S strand Lewis J et al. (2008) Molecular Biology of
S′ strand 5′ 3′ the Cell, 5th ed. Garland Science.)
parental DNA double helix
3′ 5′
template S′ strand
daughter DNA double helices
24 CHAPTER 2: Fundamentals in Cell Biology

The forces required to unfold nucleic acid polymers are of interest for several
reasons. DNA is usually tightly coiled around proteins called histones to con-
serve space. Further, because DNA is maintained in a double helix, there is
some twisting to the molecule as it unwinds to allow access for DNA replication
or transcription (more on transcription shortly). Finally, DNA activation some-
times depends on binding or association of one part of the gene to another, and
the two parts that require interaction may be separated by a long distance. The
distance between these two parts is guided, at least in part, by how flexible the
DNA is. Understanding the mechanism by which DNA (and RNA) works
requires some comprehension of the mechanical behavior underlying these
molecules.

Advanced Material: DNA and RNA are directional

Each of the carbons in the sugar backbone are labeled 5′  carbon is known as the 5′ end. Genetic sequence is
numerically 1′ through 5′. Adjacent rings are bound to usually given from the 5′ end to the 3′ end in a series of
each other through a phosphate group. This phosphate single-letter codes corresponding to the bases. Many
covalently links the 5′ carbon of one monomer to the 3′ processes involving nucleic acids are also directional, as
carbon of the next. Similar to proteins, this gives a direc- seen with the polymerase chain reaction later in this
tionality to nucleotides and their chains, nucleic acids. chapter.
The end of the polymer where the phosphate binds the

Polysaccharides are polymers of sugars

Another class of cellular polymers is the complex sugars. Familiar to us as an
(sweet) energy source, sugars also have important structural roles—the most
abundant organic molecule on the planet is cellulose from the plant cell wall.
The “crunchiness” of vegetables and the strength of wood are derived from cel-
lulose. Also, many proteins are decorated with sugars in their functional form.
A single sugar molecule is known as a monosaccharide. More complex oligo-
saccharides can be formed as disaccharides, trisaccharides, etc. Large sugar
polymers are known as polysaccharides. All sugars and the complex polymers
made from them are known collectively as carbohydrates. Similar to proteins,
bonds between monosaccharides are formed through condensation and bro-
ken through hydrolysis.

Fatty acids store energy but also form structures

The final major class of molecules introduced here is the fatty acids. A fatty acid is
made up of a long hydrocarbon-rich chain terminated with an acidic carboxyl
group. Again, we primarily think of fats or lipids as a more concentrated source of
energy than sugars. When their hydrocarbon tails are broken down, they produce
many times more energy than carbohydrates on a per-weight basis. Fatty acids
also perform critical biochemical and structural functions in the cell. Although
they do not form linear polymers such as ­nucleotides, amino acids, or monosac-
charides, they are able to aggregate to form globular and sheet structures, such as
the cell membrane. These structures not only help to segregate the interior of the
cell from the exterior, but also help to maintain the cell’s shape. Because fats and
water do not mix well, deforming a cell requires work to overcome the tendency of
fats to aggregate. Fatty acids are partly responsible for the mechanical resistance
of a cell to external forces.
 FUNDAMENTALS In CELL AND MOLECULAR BIOLOGY 25

Correspondence between DNA-to-RNA-to-protein is the central

dogma of modern cell biology
Our interest in the biological aspects of cells extends beyond passive structural
components. As we mentioned in the introduction to this chapter, we also wish
to consider the molecular response of cells to mechanical forces. But what
responses can there be, other than molecules being deformed? Different sig­
naling pathways are engaged by different cells in response to different types of
mechanical stimuli. A cell will change what proteins it makes and what part of
the cell they are transported to in response to mechanical stimuli. But to under- Nota Bene
stand this, one must first understand how a cell synthesizes proteins and how
that process is regulated. We now briefly review the central dogma of molecular Combinatorics of nucleic acids
biology, which is the link between the genetics of a cell to eventual protein and amino acids. Since there are
four base pairs, and 20 amino acids,
creation. a triplet codon is the minimum
Once the structure of DNA was discovered in the 1950s, it quickly became clear necessary to have at least one
unique sequence per amino acid.
that the sequence of nucleotides in DNA holds information that is passed from
However, a triplet codon has 64
cell to cell. There are two key properties of DNA that allow this to happen. The first possible combinations, whereas only
is replication, in which the entire sequence of DNA is copied resulting in two sep- 20 are needed for each amino acid.
arate strands with identical sequences. The original DNA double-strand is pulled Most amino acids, therefore, are
apart, and the complementary base sequences are assembled using the separated represented by several codon triplets.
original strands as templates. This results in two identical copies of the DNA Additionally, there are three triplets
polymer. for a “stop” codon that does not code
for a particular amino acid. This
The second critical property of DNA is that its sequence holds the instructions codon designates the end of the
for assembly of amino acids in a particular sequence to form specific proteins. protein. The “start” codon that
The region of DNA that contains the sequence information for a particular order designates the start of the protein
corresponds to the amino acid
is defined as a gene (Figure 2.5). Fundamentally there are two steps in the crea- methionine (Met). Note that Met can
tion of a protein from a gene. The first is to create a polymer of complementary appear in the middle of a peptide or
RNA from the gene in a process known as transcription. This RNA leaves the protein, so not all instances of Met
nucleus carrying its sequence or “message”. This type of RNA is known as mes- (ATG) represent the start of a new
senger RNA or simply mRNA. The mRNA then docks with a ribosome that is gene.
responsible for assembling the peptide sequence specified by the mRNA in a The redundancy is not evenly
spread among the amino acids (see
process called translation (Figure 2.6). The base pairs of the mRNA strand are
Figure 2.8). There is a loose pattern in
grouped into triplets, each of which is known as a codon that corresponds to a which an amino acid appears to be
particular amino acid in the forming peptide (Table 2.2). A useful memory tool coded by sequences that differ only
to distinguish transcription from translation is that transcription uses the same in the last base pair. Current thinking
“alphabet”—the sequence of bases in the DNA and RNA nucleic acids—but is that the redundancy takes
translation involves converting from sequences of bases to sequences of amino advantage of common errors to build
acids, or a different alphabet. This flow of information from DNA to RNA to pro- in some sort of protection against
mutations so that key amino acids
tein was termed the central dogma of molecular biology by Francis Crick, co-
are not as affected by such errors.
discoverer of the structure of DNA, in 1958.
One might wonder why a cell requires RNA. Why not just simply build a protein
straight from the DNA template? It turns out that only about 5–10% of the

gene A gene B gene C Figure 2.5 ​Genes are units of

sequence information that encode
for specific proteins. This
DNA correspondence between genes and
double helix proteins is known as the central dogma
of molecular biology. (Adapted from,
gene expression Alberts B, Johnson A, Lewis J et al. (2008)
Molecular Biology of the Cell, 5th ed.
Garland Science.)

protein A protein B protein C

26 CHAPTER 2: Fundamentals in Cell Biology

Table 2.2 ​The 20 common amino acids and their codon sequences. Three codons
DNA replication code for the “stop” sequence. Note that there is significant redundancy.
DNA repair DNA
5′ 3′ Ala-Alanine A GCA GCC GCG GCU
3′ 5′
mRNA synthesis Arg-Arginine R AGA AGG CGA CGC CGG CGU
(transcription)
Asp-Aspartic acid D GAC GAU
mRNA
Asn-Asparagine N AAC AAU
5′ 3′

protein synthesis Cys-Cysteine C UGC UGU

(translation)
Glu-Glutamic acid E GAA GAG
protein
H2N COOH Gln-Glutamine Q CAA CAG
amino acids Gly-Glycine G GGA GGC GGG GGU
Figure 2.6 ​Transcription is the His-Histidine H CAC CAU
production of mRNA from DNA;
translation is the production of a Ile-Isoleucine I AUA AUC AUU
protein from mRNA. (Adapted from,
Alberts B, Johnson A, Lewis J et al. (2008) Leu-Leucine L UUA UUG CUA CUC CUG CUU
Molecular Biology of the Cell, 5th ed.
Garland Science.) Lys-Lysine K AAA AAG

Met-Methionine M AUG

Phe-Phenylalanine F UUC UUU

Pro-Proline P CCA CCC CCG CCU

Ser-Serine S AGC AGU UCA UCC UCG UCU

Thr-Threonine T ACA ACC ACG ACU

Trp-Tryptophan W UGG

Tyr-Tyrosine Y UAC UAU

Val-Valine V GUA GUC GUG GUU

Stop UAA UAG UGA

entire genomic DNA strand encodes genes. These encoding regions are called
exons (“e” for encoding, Figure 2.7). Noncoding regions are called introns;
their function is not precisely known, but some intron regions are known to be
important for ­alternative splicing. Other intron regions are thought to be
important for DNA folding. mRNA is a compact subset of the genome that con-
tains the gene of interest, but also can be manipulated without disturbing the
original template stored in the DNA. Further, multiple mRNAs can be gener-
ated from a single DNA template, allowing for a relative amplification of pro-
tein expression at the mRNA level than if DNA were used for protein s­ ynthesis
directly.

Figure 2.7 ​Exons (blue), the units of

sequence that encode amino acid introns
sequences, are typically a small 1 5 10 14 22 25 26
fraction of the entire gene. (Adapted
from, Alberts B, Johnson A, Lewis J et al.
(2008) Molecular Biology of the Cell, 5th
ed. Garland Science.) exons
 FUNDAMENTALS In CELL AND MOLECULAR BIOLOGY 27

Example 2.2: Transcription and translation

The general idea underlying protein synthesis, following Note that the 3′ and 5′ are reversed. This is important,
the central dogma of cell biology, starts from a hard- because if the directionality were not reversed, then the
coded gene and ends in a protein (Figure 2.8). Suppose strand would not hybridize to the original sequence, and
you have a DNA sequence known to contain a gene, would not be a complement.
without introns, as follows:
We start from the 5′ end in the original sequence and
5′-ATACCCTATGAACAGATGAACC-3′ look for a start codon, and then separate into triplets:
What is the sequence of the complementary strand and 5′-ATACCCT ATG AAC AGA TGA ACC-3′
peptide generated from this sequence?
So we get Met-Asn-Arg. That is it, since the TGA is a stop
The complementary strand would be:
codon. Of course, most proteins are far longer than this
3′-TATGGGATACTTGTCTACTTGG-5′. three-amino-acid example peptide.

Figure 2.8 ​Summary of the process

cytoplasm of protein synthesis from DNA to
fully functional protein. (Adapted
nucleus from, Alberts B, Johnson A, Lewis J et al.
(2008) Molecular Biology of the Cell, 5th
introns exons
ed. Garland Science.)
DNA

transcription unit
transcription
“primary RNA transcript”

5′ capping
RNA splicing
RNA cap 3′ polyadenylation

mRNA AAAA
export

mRNA AAAA
translation 
protein

Phenotype is the manifestation of genotype

The collection of genes represented in a cell is known as its genome. With few
exceptions, all of the cells in a particular organism contain the same DNA
sequence and therefore the same genes. Almost all of the cells in a person’s body
contain the same DNA sequence and are said to be of the same genotype (notable
exceptions include the red blood cells, which lack nuclei and genomic DNA, and
reproductive cells). Of course, your genetic makeup likely differs from others as
you probably have different genotypes. What, then accounts for the difference
between the behavior of one of your liver cells from one of your bone cells if they
are genotypically identical? The difference in cellular behavior is a function of
which genes are being transcribed or expressed at any particular time. The word
28 CHAPTER 2: Fundamentals in Cell Biology

phenotype was coined in 1911 by Wilhelm Johannsen to describe the behavior or

character of a cell or organism in contrast with its genotype, which is inherited.
During development or certain repair processes, cells change their phenotype
and become more specialized in a process known as differentiation.

Advanced Material: Phenotype can be elusive

By contrast with genotype, a cell’s phenotype can be It is difficult to determine when a cell has changed
much more difficult to define and is often only the result its  behavior sufficiently to have differentiated into
of long-term consensus. This is confounded by the fact another cell type. These changes must, in some sense, be
that the behavior of a cell changes considerably over irreversible and permanent. Yet, in some situations cells
time as a result of changes in the extracellular environ- can lose the phenotypic markers of differentiation in a
ment. In some ways, phenotypic change and differentia- process known as dedifferentiation. Organisms that can
tion is meant to imply more enduring changes to a cell’s regenerate lost limbs are thought to be able to induce a
behavior than simple responses to external stimuli. dedifferentiation process. In some conditions osteocytes
When these stimuli persist for a sufficient time, they can can be induced to make bone. Does this mean that they
lead to changes in phenotype. For example, in bone, the have dedifferentiated into osteoblasts again? Cells often
cell responsible for forming new bone is called the oste- lose differentiation markers of their original cell type.
oblast. An osteoblast may or may not form bone in Have they dedifferentiated? These  questions, although
response to biochemical and physical signals, though it they appear simple on the surface, can be scientifically
remains an osteoblast. Occasionally, an osteoblast ambiguous and are often based on functional defini-
becomes embedded in the same bone that it is forming, tions. These issues can become quite contentious and
and becomes more quiescent than an osteoblast. This vary substantially from field to field.
cell is said to have differentiated into a different cell type
called an osteocyte.

Nota Bene
Progenitor cells have the potential to differentiate into multiple different cell
types, thereby acquiring different phenotypes. A cell that has the potential to dif-
RNA undergoes maturation ferentiate into multiple cell types is known as multipotent or pluripotent.
modifications. The newly formed Pluripotent cells can differentiate into cells or all three germ layers (endoderm,
mRNA transcript often undergoes
post-transcriptional modification
mesoderm, and ectoderm), while multipotent cells can differentiate into multi-
before translation into protein. For ple, but more limited, lineages. Cells that can differentiate into any cell type in an
instance, the 3′ end of the mRNA is organism are referred to as totipotent. Progenitor cells that can differentiate into
usually supplemented with a more specialized cell types, but can also undergo cell division without differenti-
polyadenosine tail (poly(A) tail), which ating, are known as stem cells. Both embryonic stem cells (currently only obtain-
allows protein complexes to bind to able from a blastocyst) and adult stem cells (obtainable from a variety of tissues,
the mRNA transcript. These protein
including bone marrow, umbilical cord blood, fat, and muscle) are being studied
complexes can splice out the introns
of the gene to leave behind only the for their potential use in regenerative medicine and tissue engineering. One key
exons for translation. It is also challenge in working with stem cells is that once the desired cell is generated, it
possible to splice out some exons must be conditioned to work with existing tissues. For example, if a partial tissue
that the introns flank, thereby replacement is required, the cells must integrate with the host cells and exhibit
creating several protein variants from the correct strength and durability to continue functioning. Many of these charac-
the same DNA source gene. This teristics depend on cell mechanical properties, as well as mechanobiological
modification is referred to as
alternative splicing.
adaptation, and understanding and manipulating these characteristics remains
a fundamental challenge in tissue engineering.

Transcriptional regulation is one way that phenotype differs

from genotype
One of the ways in which the cell regulates its behavior is by altering which genes
are transcribed and how often. RNA polymerase is the enzyme responsible for
transcribing DNA into RNA. Its activity is regulated by controlling its proclivity
to bind to a particular region of DNA, known as a promoter, usually proximal to
 FUNDAMENTALS In CELL AND MOLECULAR BIOLOGY 29

the coding sequence (the region that encodes the amino acid sequence) for a par-
ticular gene, and begin transcription. The cell is not constrained to up- or down-
regulate (increase or decrease) transcription of all genes equally. To regulate
function and behavior, the cell needs to control transcription of specific genes
without affecting transcription of other genes. DNA-binding proteins known as
transcription factors recognize specific DNA sequences and regulate the binding
of RNA polymerase. Given the complexity of cellular function, there is a dizzying
array of transcription factors. Indeed, most genes require a complex of transcrip-
tion factors to form a protein cluster for transcripts to be produced. Finally, tran-
scription factors are themselves coded from genes, so their availability is also
subject to transcriptional regulation.

Cell organelles perform a variety of functions

The interior of a cell is not a simple amorphous soup of materials. It is organized into
a wide variety of small subcellular units with particular functions, some of which
we have already discussed. They are called organelles by way of analogy to the organs
of a larger organism, a term coined in 1884 by Karl Möbius (not of the Möbius strip).
In our discussion of protein synthesis, we have already introduced a few critical orga-
nelles—the nucleus, the ribosome, the Golgi apparatus, and the endoplasmic reticu-
lum. Table 2.3 lists major organelles and their functions for eukaryotic cells.
Nota Bene
The “central dogma” is important for cell mechanics and mechanobiology for sev-
eral reasons. First, the proteins that exist to support cells structurally do not usually Often proteins need some final
tweaks. To become fully functional,
pre-exist within the cell. They have to be made by the cell, typically in an ongoing
many proteins need to be further
process involving rather intricate pathways. The cell produces proteins from amino modified after translation. This is
acids using instructions from genes. So, unlike the framework of a building, the known as post-translational
skeleton of the cell is constantly shifting and rebuilding, and anything that disrupts modification. Modifications may
this process can result in significant changes to cell properties. include formation of disulfide bonds
within a protein (linking together two
Second, for mechanobiology in particular, cells respond to several signals, parts of the protein normally far
including physical forces, by altering their expression profile. The phenotype of apart) and attachment of non-amino-
the cell depends in part on mechanical influences acting upon the cell. How the acid functional groups such as
phenotype adjusts is a matter of much interest—cells will change morphology, carbohydrates, lipids, phosphates,
increase levels of some proteins, decrease others, and alter their overall behavior etc. During post-translational
modification, the protein is often
(motion, activity, proliferation, etc.). To study these responses at the molecular folded to acquire its secondary and
level, we need to know what to examine. Now that we know the dogma we can tertiary structure, typically in the Golgi
apparatus and endoplasmic reticulum
(ER). X-ray crystallography is used to
Table 2.3 ​Some of the eukaryotic cellular organelles and their functions. These determine the structure of a protein
functions are described to provide an idea of their general function, but in some cases experimentally.
they may be involved in other tasks.

Organelle Primary function(s)

nucleus DNA repository, site of transcription

ribosome responsible for mRNA-to-protein translation

mitochondrion responsible for energy production by conversion of glucose

metabolites into ATP; has its own maternal DNA

endoplasmic reticulum site of translation and some post-translational modification

and folding of proteins

golgi apparatus post-translational modification and sorting of proteins;

important for secreting proteins

lysosome degradation of proteins and carbohydrates; it is important

that the degradative enzymes responsible remain
sequestered from the cytoplasm

vesicle intracellular transport and trafficking

30 CHAPTER 2: Fundamentals in Cell Biology

Nota Bene consider several strategies. Examination of mRNA levels would be an indication
of the activation of certain genes. One might complement mRNA profiles with
Bacteria do not have nuclei. A
protein measurements—and typically protein measurements are made after
bacterium is a prokaryotic cell—a cell
without a nucleus. Eukaryotic cells mRNA is known to change. This timing is selected because proteins come from
have a nucleus plus many of the mRNA, so one would normally expect the changes in protein concentration to
other organelles. Prokaryotic cells are lag behind the changes in mRNA, which we know only because of the central
nearly always unicellular and do not dogma.
form complex organisms. Originally
they were thought to have no Finally, there is the challenge of determining the exact mechanism by which cells
intracellular organelles, but is now sense and respond to mechanical forces. Although it is established that, stretch-
known that they do have some (such ing a cell will lead to up-regulation of mRNA and protein from certain genes, less
as ribosomes). Their DNA is often is known about how the up-regulation occurs. There are many hypotheses,
organized into a single closed loop including proteins and nucleic acids being directly unfolded by the forces, the
known as a plasmid.
cell membrane itself being flexed, and others. The central dogma provides a
much better idea of what to look for if we know roughly the pathway such up-
regulation takes.

2.2 ​RECEPTORS ARE CELLS’ PRIMARY

extracellular CHEMICAL SENSORS
space
Receptors are proteins that bind to specific target molecules or ligands and initi-
ate a cellular response, typically in the form of a biochemical intracellular signal-
ing cascade. Often receptors are suspended in the cell membrane, but they can
also be found in the cytoplasm or the nuclear membrane. Membrane receptors
cytosol can be identified by one or more hydrophobic membrane-spanning domains
plasma
(Figure 2.9). The specific structure of a receptor can be determined by X-ray crys-
membrane
tallography or by computer simulation predictions based on the protein’s
Figure 2.9 ​A receptor typically sequence. Of course there is an enormous diversity of receptors and correspond-
contains several membrane- ing ligands. A cell can change its sensitivity to a given ligand by up-regulating or
spanning domains, an extracellular down-regulating the number of receptors on its surface. Often this regulation
component that binds to the target occurs in response to activation of the receptor itself, thereby creating a form of
or ligand, and an intracellular feedback mechanism. When a cell senses a small number of a particular ligand, it
component that initiates
intracellular signaling or other
commonly makes more receptors for that ligand. This is a positive feedback loop.
forms 25 of cellular regulation. It allows the cell to be highly sensitive to a huge number of extracellular signals
(Adapted from, Alberts B, Johnson A, and to do so very efficiently.
Lewis J et al. (2008) Molecular Biology of
the Cell, 5th ed. Garland Science.)
By being adaptive to its environment, the cell need not maintain large numbers of
receptors for every conceivable signal, but at the same time, it does not sacrifice
sensitivity. The converse is also true. When a particular ligand is abundant, the
cell may decrease its sensitivity to that signal selectively. Receptors may become
internalized upon activation and no longer be available for ligand binding.
Receptors can also become desensitized or less likely to bind their ligand if they
are repeatedly activated, providing a mechanism of cellular accommodation. For
a receptor to function properly it must undergo a change upon binding its ligand
that allows this information to be propagated. This occurs through the process of
biochemical signaling, which is a broader subject not limited to receptors.

Cells communicate by biochemical signals

Cellular signaling concerns the flow of information across time and space. It is
fundamental to how cells regulate their behavior and interact with their environ-
ment. Temporal flow simply refers to chains of events in signaling cascades or
pathways, with prior events being upstream and subsequent events being down-
stream. Spatial information flow can take the form of molecules diffusing, being
transported, crossing membranes, or being sequestered. Information flow in the
world of the cell occurs by molecular information carriers moving from place to
place. We refer to these as biochemical signals to reflect their role as c­ arriers of
information. It is also possible for information to be transmitted by physical
 RECEPTORS ARE CELLS’ PRIMARY CHEMICAL SENSORS 31

means, such as electric fields and physical forces. These physical signals can play
a fundamental role in many aspects of cell biology and have only recently begun
to be investigated. Physical signaling and the emerging field of cellular mechano-
biology will be discussed in some detail in Chapter 11. We begin here by first con-
sidering traditional biochemical signaling.

Signaling between cells can occur through many different

mechanisms
Extracellular biochemical signals are one of the most important ways that cells
sense and respond to changes in their environment. They are critical to coordinat-
ing the behavior of individual cells so they can assemble into tissues, organs, and
ultimately entire organisms. Hormones are substances that can have a dramatic
effect on cell behavior, often in very low concentrations. They are distinguished
from other biochemical signals in that they must be secreted by a specific organ
and circulate throughout the organism. Examples include adrenaline, parathyroid
Nota Bene
hormone, insulin, melatonin, and sex hormones such as estrogen and testoster-
one. Cytokines are distinguished from hormones in that they do not originate from Neurotransmitters are a special
a particular organ. Rather, cytokines mediate cell-to-cell communication and tend class of extracellular cell-to-cell
to affect a nearby collection of cells, although some can have systematic effects. signal that medicate communication
between neurons and adjacent
Cytokines are not only important in the functioning of the adult organism, they are cells. They typically act over the
central mediators of cell-to-cell regulation of development. Extracellular signals exceedingly short distance of a
are further classified in terms of the spatial extent of their effects. Substances that neural synapse. Steroids are another
have effects throughout an organism are known as endocrine signals, substances particular type of extracellular signal
that are responsible for cells regulating the behavior of nearby cells are known as that owning to their hydrophobic
paracrine signals, and cells regulating themselves are mediated by autocrine sig- nature, cross the cell membrane to
bind a cytoplasmic or nuclear
nals (Figure 2.10). Of course, it is nearly impossible for a cell to selectively signal to
receptor.
itself, so nearly all autocrine signals are also paracrine signals.
Signal transduction is the process of one type of chemical (or physical) signal
being changed or transduced into another. The most common type of transduc-
tion is when an extracellular biochemical signal arrives at the cell membrane and
is transduced into an intracellular signal (Figure 2.11). The vast majority of extra-
cellular signals are transduced by membrane receptors. The intracellular domain
of a receptor typically undergoes a conformational change as a result of ligand
binding. This might enhance enzymatic activity within the structure of the recep-
tor or expose a previously hidden binding site with which another protein can
interact. Signaling molecules will eventually become activated and leave the
membrane to propagate the signal intracellularly and initiate a complex set of bio-
chemical reactions sometimes termed a signaling cascade.

(A) contact-dependent (B) paracrine (C) endocrine

signaling cell target cell signaling endocrine cell receptor target
cell cell

target hormone
cells
membrane-
bound signal local
molecule mediator
bloodstream target
cell
Figure 2.10 Signaling between cells can occur through many different mechanisms. (A) Cell–cell contact may or may not be
required. (B) Autocrine and paracrine signaling occurs through the release of cell-secreted soluble chemicals. Paracrine signaling is
depicted here, in which the target and signaling cells are different. (C) Endocrine signaling involves a soluble chemical produced in the cells
of specialized organs and released into circulation. (Adapted from, Alberts B, Johnson A, Lewis J et al. (2008) Molecular Biology of the Cell,
5th ed. Garland Science.)
32 CHAPTER 2: Fundamentals in Cell Biology

Figure 2.11 ​When an extracellular extracellular signal molecule

signal (or ligand) is bound by a
receptor, the intracellular domain of
receptor protein
the receptor changes conformation
in such a way as to initiate an
intracellular signaling cascade. This
plasma membrane of
can eventually produce changes in cell target cell
metabolism, gene expression, or cell
morphology or motility. (Adapted from,
Alberts B, Johnson A, Lewis J et al. (2008)
Molecular Biology of the Cell, 5th ed.
Garland Science.)
intracellular signaling proteins

effector proteins
metabolic gene regulatory cytoskeletal
enzyme protein protein

altered cell
altered altered gene
shape or
metabolism expression
movement

Intracellular signaling occurs via small molecules known

as second messengers
Second messengers are rapidly generated, quickly diffuse though the cell to
activate their targets, and then return to a low baseline concentration. The
potency of a second messenger to deliver information is directly related to its
ability to become rapidly elevated when activated and to remain at a low level
otherwise. By analogy, this is a second messenger’s signal-to-noise ratio. The
level to which it can be raised is its “signal” and the background concentration
its “noise”. In addition, a molecule that is held at a very low concentration will
diffuse more rapidly when elevated than one that is not. Second messenger sys-
tems are therefore always associated with active cellular processes that quickly
return them to low concentrations after activation by removing them from the
intracellular space, sequestering them, or degrading them. The target of a sec-
ond messenger can be activation of an ion channel, activating another signaling
protein such as protein kinases (discussed below), or affecting the kinetics of
another reaction.
Second messengers fall into one of three categories. (1) Molecules that do not
freely cross lipid membranes, such as intracellular calcium (Ca2+), and cyclic
nucleotides, such as cyclic adenosine monophosphate (cAMP) and cyclic guano-
sine monophosphate (cGMP). (2) Hydrophobic molecules that are typically
­associated with the cell membrane and form as the metabolic products of phos-
pholipids; owing to their insolubility they tend to have proteins in the membrane
as their targets. Diacylglycerol (DAG) and inositol triphosphaste (IP3) are the clas-
sic examples. (3) Some dissolved gases such as nitric oxide (NO) and carbon mon-
oxide (CO) can be potent second messengers, owing to their ability to quickly
diffuse though the cytoplasm and across lipid membranes.
 RECEPTORS ARE CELLS’ PRIMARY CHEMICAL SENSORS 33

channel pump Figure 2.12 Channels and pumps

are examples of proteins or protein
complexes that regulate the
movement of second messengers
across the plasma membrane and
intracellular membranes that
distinguish various intracellular
organelles. Channels regulate transport
closed open along a concentration gradient and
pumps against it. In the case of pumps,
conversion of adenosine triphosphate
(ATP) to adenosine diphosphate (ADP) is
required to provide a source of energy.

Intracellular calcium signaling is one of the most studied and understood of the
second messenger systems. Calcium is maintained at a very low level in the
cytoplasm. This is done either by pumping it out of the cell or by sequestering it
in the endoplasmic reticulum (sarcoplasmic reticulum in muscle cells), and to a
lesser extent mitochondria. Transient increases in intracellular calcium can
occur through several molecular mechanisms. Calcium channels in the cell
membrane can allow calcium to flow from outside the cell into the cytoplasm in
response to an extracellular signal. There are also channels to allow the seques-
tered intracellular stored calcium to be released (Figure 2.12). Calcium signal-
ing can be readily studied, by using depletion (removing calcium) or via the use
of fluorescent reporters (Figure 2.13).

Advanced Material: Calcium channels

Calcium channels fall into two categories, the IP3 recep- tion. It is typically finished in a matter of seconds, after
tor and the ryanodine receptor. IP3, as we have dis- which the calcium is again pumped out and seques-
cussed, is a hydrophobic second messenger released tered in preparation for the next transient. Calcium is
into the cytoplasm from the membrane. The ryanodine known to regulate the activity of several downstream
receptor is sensitive to calcium concentration itself. proteins. One of the most well-understood of these pro-
This forms a highly sensitive positive feedback loop teins is calmodulin. Calmodulin, in turn, regulates the
such that a small elevation in intracellular calcium activities of other enzymatic proteins known as, not
induces a much larger and more rapid release from surprisingly, calmodulin-dependent protein kinases or
intracellular stores. This increase in calcium concentra- CaM kinases.
tion is transient and sometimes called a wave or oscilla-

Figure 2.13 Cells filled with the dye

(A) (B) Fura-2, whose fluorescent emission
changes as a function of calcium
concentration. The cells in (A) are in a
baseline state and the cells in (B) have
been stimulated with a calcium agonist.
The ease with which calcium signaling
can be studied optically has contributed
to its being one of the better-understood
second messenger systems. See www.
garlandscience.com for a color version
of this figure.
34 CHAPTER 2: Fundamentals in Cell Biology

Large molecule signaling cascades have the potential for

more specificity

Nota Bene
Signaling cascades involving large molecules and proteins tend to be more
slower acting than second messenger systems, but have the potential for more
What is ryanodine? The ryanodine specific activity. Some complex but necessary nomenclature is required to
receptor gets its name because it can understand the dynamics of these signaling cascades. The activity of many sign-
be modulated by ryanodine, to which
ryanodine receptors show very high
aling and effector proteins is altered when they become phosphorylated, or
affinity. However, ryanodine is a experience the addition of a phosphate (PO4) group. Typically, the phosphate is
poisonous alkaloid only appearing in donated by a high-energy donor such as ATP. This is a reversible reaction that
plants, and it has no role in cell induces a conformation change, allowing a protein to switch from an inactive to
physiology. The name was given an active state. Likewise, some proteins switch from an active to an inactive
because ryanodine was used in the state when they are phosphorylated. Enzymes that phosphorylate other pro-
first purification of the receptor. It is
teins are known as kinases and those that dephosphorylate other proteins are
an accident of history.
called phosphatases (Figure 2.14). In general phosphorylation only occurs on
three specific amino acid residues within a protein: tyrosine, serine, or threo-
nine. Whereas a kinase phosphorylates another protein, a “tyrosine kinase” is
an enzyme that is restricted to adding a phosphate only to the tyrosine residue
of another protein. The diversity and complexity of protein kinase signaling is
immense. More than 500 different kinases have been identified in humans. The
target of a kinase might itself be a kinase that becomes activated upon phospho-
rylation. These are known as kinase kinases. There are also kinase kinase kinases,
but at this point it becomes so cumbersome that they begin to be referred to as
3-kinases, 4-kinases, etc.
Nota Bene As protein–protein signaling cascades progress, relatively small events, such as
Why are intracellular messengers
binding of a few receptors, can lead to larger and larger biochemical responses.
“second”? The term second This can occur through positive feedback loops at the receptor level or from cata-
messenger system stems from the lytic reactions within the cascade itself, in which a single protein can catalyze a
concept that the extracellular signal downstream reaction repeatedly, amplifying the signal. This intracellular amplifi-
represents the first “message.” The cation can be quite profound and sometimes results in cells being exquisitely sen-
cell then converts this first message sitive to small extracellular signals. Protein–protein signaling pathways are critical
to a second message by because they allow for the intricate signaling cascades that the cell needs to
transduction.
orchestrate its varied and complex functions (Figure 2.15).

Figure 2.14 ​Sequential protein plasma membrane

phosphorylations form signaling
cascades or networks that are
generally slower than second MAP kinase kinase kinase cytosol
messenger signaling, but are GTP
potentially more specific. The ATP
active Ras P P
ultimate target may be either regulation ADP
of protein activity or altered gene
protein
expression. Phosphate can be provided MAP kinase kinase
by guanosine triphosphate (GTP) or ATP.
MAP kinase is mitogen activated protein ATP
kinase. (Adapted from, Alberts B, P P ADP
Johnson A, Lewis J et al. (2008) Molecular
Biology of the Cell, 5th ed. Garland MAP kinase
Science.)
ATP
ADP

P protein P protein P gene P gene

X Y regulatory regulatory
protein A protein B

changes in protein activity changes in gene expression

 RECEPTORS ARE CELLS’ PRIMARY CHEMICAL SENSORS 35

signal molecule

P P
GPCR
RTK
P P
P P

G protein G protein phospholipase C Grb2 PI 3-kinase

Ras-GEF (Sos)
adenylyl cyclase IP3 diacylglycerol PI(3,4,5)P3
Ras
Ca2+
MAP kinase kinase kinase
cyclic AMP PDK1
calmodulin MAP kinase kinase

PKA CaM-kinase PKC MAP kinase Akt kinase

gene regulatory proteins many target proteins

Figure 2.15 Intracellular signaling networks can be exceedingly complex. They can involve negative and positive feedback,
cross-talk and even redundancies. Although the details of the pathways in this schematic are beyond the scope of this course, it illustrates
most of the major signal systems at work in typical cellular regulation and allows cellular responses that are both sensitive and specific
(Adapted from, Alberts B, Johnson A, Lewis J et al. (2008) Molecular Biology of the Cell, 5th ed. Garland Science.)

Advanced Material: Proteins may need to form complexes for signaling

Adaptor proteins do not have any enzymatic activ- when bound to each other. Alternatively, protein com-
ity themselves, but are required to bind to other proteins plexes may be subsequently transported to other regions
to allow them to become active. For instance, they can of the cell or their formation may be a signal for degrada-
bind to particular recognition sites on a protein and induce tion of the protein complex. Adaptor proteins add another
a conformational change that exposes a particular residue layer of regulation allowing additional specificity in cell
for phosphorylation. This may be necessary for the forma- signaling.
tion of complexes of multiple proteins that only function

Receptors use several mechanisms to initiate signaling

As we have seen, receptors, as mediators of transduction, form an interface
between the extracellular signals a cell receives and the complex signaling path-
ways the cell activates in coordinating its response. It might not be surprising that
there are many molecular similarities between signaling pathways and the intra-
cellular domains of receptors that initiate signaling pathways. Receptors fall into
36 CHAPTER 2: Fundamentals in Cell Biology

several classes depending on how they acheive this. One such class is the receptor
tyrosine kinases. These are transmembrane proteins with an extracellular domain
separated from an intracellular domain by one or more membrane-spanning
domains. The extracellular domain is responsible for recognizing and specifically
binding to the ligand target of the receptor. Ligand binding leads to a conforma-
tional change in the intracellular domain. Receptor tyrosine kinases, like many
receptors, are actually dimers of two proteins. These dimers are only stable in the
presence of the ligand. When a stable dimer forms, tyrosines in the cytoplasmic
domains of the receptor are autophosphorylated and are thereby activated to ini-
tiate the intracellular signaling cascade. At this point, the specific signaling path-
way that is activated will depend on the particular receptor.
Another class of membrane receptors comprises the G-protein-coupled receptors
(GCPRs). They contain seven membrane-spanning domains and are each linked
to a G-protein, as the name suggests. Before ligand binding, the G-protein coupled
to the receptor is in its deactivated guanosine diphosphate (GDP)-bound form.
Upon ligand binding, the receptor acts as a guanine nucleotide exchange factor
(GEF) and allows a guanosine triphosphate (GTP) to substitute for the GDP,
thereby activating the G-protein. The activated G-protein dissociates from the
receptor and is free to diffuse and activate several downstream targets. A hydro-
phobic domain in the G-protein tends to keep it membrane-bound. Not surpris-
ingly then, the targets of the liberated G-protein are other membrane-bound
proteins. GCPRs also activate second messenger signaling systems. These include
production of IP3 and DAG through the action of the G-protein on phosphodiester-
ases and phospholipases and production of cAMP and cGMP by activation of
membrane-bound adenylyl cyclases. GPCRs can even potentiate intracellular cal-
cium increases by opening membrane calcium channels.
Integrins are a particular type of receptor that also act as mechanical linkers of the
cell to its extracellular environment (although they are not the only receptors that
do this, integrins are among the best-studied). Integrin ligands are extracellular
matrix proteins such as collagen, fibronectin, and laminin. In addition to providing
a mechanical connection between the cell and its environment, when integrins
bind they also activate intracellular signaling through integrin-associated kinases.
This signaling results in increased integrin production, recruitment and clustering
of integrins, and the eventual assembly of compact attachments known as focal
adhesions. In many cell types integrin is also an important cell survival signal as
well as a regulator of proliferation and differentiation. As we shall see, it also
appears to be important in helping the cell sense mechanical stress upon itself.
This is known as outside-in integrin signaling, in contrast with inside-out signaling.
In the case of inside-out signaling, cytoplasmic signals regulate the ability of an
integrin to bind its ligand by exposing or hiding the ligand-binding domain.

2.3 EXPERIMENTAL BIOLOGY

Progress in cell biology has been facilitated by the development of increasingly
­sophisticated experimental techniques. The inner workings of the cell may be
understood through many different approaches, each with its complement of pos-
sibly unfamiliar terms. Here we review the basic terminology so that you can read
the cell mechanics research literature with confidence.
In experimental biology, microscopy is used because cells and subcellular com-
ponents are typically too small to be seen unassisted. Many cell mechanical stud-
ies also require microscopy, which may involve the terms below:
• Lens: an optical component that is used to change the convergence of a light beam.
A convex or converging lens focuses a beam of light, whereas a concave or diverg-
ing lens spreads out the beam of light.
• Resolution: the distance between two point sources of light whereby the two sources
can be distinguished (optical microscope resolutions are typically at 0.1–1 μm).
 EXPERIMENTAL BIOLOGY 37

• Wavelength: the peak-to-peak distance between adjacent crests in the electromag-

netic wave representing light (optical microscope wavelengths typically range from
400 to 900 nm).
• Index of refraction: the change in the speed of light as the light beam encounters
different media (for example air to water) (typical indices in microscopy range from
1 to 1.5).
• Noise: unwanted signal, regardless of whether the signal constitutes actual data.
An incredible amount of information can be gained just by looking at cells. The
compound microscope was originally developed by Galileo in 1609 and consisted
of one convex–concave lens. Using these early simple instruments, Robert Hooke
(the same person who described the deformation of elastic bodies in Hooke’s law),
along with Antoine van Leeuwenhoek, described microscopic living structures in
cork bark and coined the word cell. Since the structures that make up a cell are
almost all transparent, simple light microscopy is limited as an investigative tool in
cell biology, yet a great number of cell properties relevant to cell mechanics can be
obtained through observation. Cell morphology—whether the cell is round,
spread, whether it has many protrusions like a starfish—is one example of a major
factor in cell mechanics. Cells that are spread out will tend to exert higher forces on
the surfaces to which they are attached. How the cell is shaped can reveal informa-
tion about the cell’s tendency to move or respond to forces— some cells will actu-
ally align in the direction of, or perpendicular to, applied forces.
Other factors of interest include both active processes, such as cell migration, and
structural components, such as the density of the cell’s support skeleton (the
cytoskeleton). Some of these can be visualized with minor adjustments to optical
tools; others require more sophisticated techniques.

Optical techniques can display cells clearly

Many microscopy techniques are  used to increase contrast. Cells are generally
difficult to image with just plain illumination, owing to their thinness. There are
several different ways to generate contrast (Figure 2.16), each with different char-
acteristics and uses. Phase–contrast microscopy (for which Frits Zernike, a physi-
cist, was awarded the Nobel Prize in 1953) solves the problem of cells being too
thin to directly observe via white light illumination (also called brightfield) by
adding contrast to the optical image based on small phase shifts introduced by the
specimen. Light from the source is passed through a ring-shaped mask and a lens
or condenser before ­passing though the specimen, magnifying objective, and eye-
piece. In this way, there are two sets of focal planes. The one associated with the
objective lens creates an image of the specimen on the observer’s retina as it
would in a standard microscope. In addition, the condenser creates an image of
the mask inside the microscope itself, typically within the multiple lenses of the

(A) (B) (C)

Figure 2.16 ​Differet types of contrast illumination available for imaging cells. (A) Brightfield, (B) phase contrast, (C) differential
interference contrast (Courtesy of An Nguyen).
38 CHAPTER 2: Fundamentals in Cell Biology

retina objective (Figure 2.17). The objective of a phase-contrast microscope has a sec-

ond mask inserted in the objective that is the inverse of the mask in the condenser.
eye When the microscope is properly aligned, this image comes into focus exactly on
the mask in the objective. Generally, the objective mask is a ring, and the con-
denser mask is an annular slit. In the absence of a specimen, all light in a properly
aligned phase-contrast microscope is blocked by one of the two masks. However,
when an object is introduced it causes the light to bend slightly through diffrac-
eyepiece
tion and refraction, allowing those photons to pass through the objective. The
(ocular)
resulting image has a greatly increased contrast when compared to brightfield
microscopy. Objects in a phase-contrast image appear bright or dark depending
not on their optical density, but rather on the phase shifts they introduce. Typically
a structure will appear dark and be surrounded by a bright halo of light. Phase-
contrast is so useful in visualizing cells that it is virtually ubiquitous in cell culture
laboratories. Differential interference microscopy (DIC) further enhances con-
trast using polarized light to produce brighter diffraction halo around each object.

objective Unfortunately, basic optical techniques are subject to numerous limitations,

including the inability to distinguish easily between different components within
the cell. Some structures, such as microtubules, are not visible at all.
specimen
condenser Fluorescence visualizes cells with lower background
Some molecules, when illuminated by specific wavelengths of light, exhibit a par-
light source ticular type of excitation whereby electrons absorb the incident photon and jump
to a higher energy level. Some energy is lost to heat as a result of the jump “over-
Figure 2.17 ​Optical schematic of a
typical light microscope. In a
shooting” the stable higher energy level (like throwing a ball higher than a roof in
phase-contrast approach masks are order to land it on the roof ). After a short time, the electron will decay back to its
inserted in the condenser and objective. original orbital and baseline energy level. This decay is accompanied by the
(Adapted from, Alberts B, Johnson A, release of a new photon, at a longer wavelength (lower energy). Such molecules
Lewis J et al. (2008) Molecular Biology of are called fluorescent and can enhance the use of microscopy to examine cellular
the Cell, 5th ed. Garland Science.) structures and processes.
Fluorescence microscopy is based on the excitation of molecules within the cell
with fluorescent properties. Some endogenous molecules can fluoresce (called
autofluorescence), but those are few and limited in scope. In general, flourescent
molecules are introduced into the cell externally. These molecules are called fluo-
rophores or fluorescent dyes. As was described in the previous paragraph, fluores-
cence occurs when a molecule absorbs a photon at one ­frequency (called the
excitation) and emits a photon at a lower frequency (called the emission). This
can be used to obtain very high-contrast images by illuminating the specimen at
one wavelength and observing it at a different wavelength.
A typical epifluorescence microscope actually uses the same objective lens for
illumination and observation (the prefix epi- means “beside”, referring to the illu-
mination and observation taking place on the same side of the specimen). A high-
intensity excitation source (a mercury lamp, LEDs, or lasers) is filtered to produce
a narrow band of wavelengths centered at or near the maximum absorbance
wavelength of the dye being used. The light is then reflected by a dichroic mirror
onto the specimen. The emitted fluorescent light, now at a longer wavelength,
travels back though the objective, this time passing through the dichroic mirror.

Advanced Material: Dichroic mirrors change reflectance with wavelength

A dichroic mirror is the heart of an epifluorescence the illumination light and the emitted light from the
microscope. It is reflective below a critical wavelength specimen to travel through the same objective, but for
and transparent to light above that wavelength. It allows the emitted light to be separated and observed.
 EXPERIMENTAL BIOLOGY 39

Figure 2.18 ​A light-path schematic of

a typical epifluorescence
microscope. The darker blue light path
is the excitation and would be at a
shorter wavelength. Fluorophores in the
eyepiece specimen absorb photons at the
excitation wavelength and emit them at
the emission wavelength. (Adapted from,
emission filter Alberts B, Johnson A, Lewis J et al. (2008)
Molecular Biology of the Cell, 5th ed.
Garland Science.)

light
source beam-splitting
(or dichroic) mirror

excitation filter objective lens

object

An emission filter helps improve the image by removing light at wavelengths dif- Nota Bene
ferent from a band centered around the dye’s emission wavelength being used
Cytoskeletal and nuclear
(Figure 2.18). In this way, high-quality fluorescence images can have a virtually
fluorescence imaging. A common
black background with a very high ­signal-to-noise ratio. cell structure of interest is the
cytoskeleton. Actin filaments can be
Fluorophores can highlight structures visualized using antibodies against
F-actin via immunofluorescence, or
Fluorescence microscopy can be used to examine specific structures within with various compounds such as
phalloidin, (which is typically attached
cells. Certain compounds can be attached to molecules that only bind to par- to, or conjugated to, a fluorescent
ticular proteins within the cell to determine their distribution. A broad way of molecule). Phalloidin is generally
doing this is using a technique called immunofluorescence. In immunofluores- specific for actin (phalloidin staining
cence, an antibody is generated that is specific for the protein of interest. This is very fast compared with immuno­
antibody (called the primary antibody) can be linked to a fluorescent molecule. fluorescence, hence it is widely used
The cell is incubated with this antibody and then washed to remove unbound despite requiring the cells to be fixed
antibodies. Upon examination using fluorescence microscopy, only the pro- as with immuno­fluorescence).
Unfortunately, the microtubules and
teins that were tagged with the fluorescent antibody would appear, providing intermediate filaments do not
information about the location of the protein, and to some extent its quantity. currently have widely available
Alternatively, a nonfluorescent primary antibody can be followed with a fluo- fluorescence agents and must be
rescent secondary antibody, which binds the primary antibody. This can also visualized using immunofluorescence
or transfection. The nucleus can be
visualized using several dyes; DAPI
(4′,6-diamidino-2-phenylindole) and
propidium iodide are good for
imaging fixed cells, Hoechst is a
secondary antibodies
primary antibody cell-permeant fluorescent dye.
marker Nuclear labels are generally toxic to
cells; these labels should not be used
immobilized for long-term (over days) cell tracking.
Other structures (mitochondria) can
antigen
be highlighted using various
compounds; an exhaustive listing
would be impractical here.
Figure 2.19 ​The use of a secondary antibody means that fluorophores or other
means of visualization only need to be linked to one antibody. Additionally, there is
potential for signal amplification, since multiple secondaries can bind to a single primary.
(Adapted from, Alberts B, Johnson A, Lewis J et al. (2008) Molecular Biology of the Cell, 5th ed.
Garland Science.)
40 CHAPTER 2: Fundamentals in Cell Biology

amplify the signal, as many secondary antibodies can attach to each primary
antibody (Figure 2.19).

Fluorophores can probe function

Fluorescence microscopy is not limited to telling us what a cell looks like; it can
also provide information about a cell’s biological activity. This is because the fluo-
rophors introduced into the cell can be designed to interact with the cell in spe-
cific ways. Green fluorescent protein (GFP), was originally isolated from a species
of jellyfish (Aequorea victoria). Many fluorescent compounds are synthetically
manufactured, which makes it difficult to introduce them into living cells. When
the gene sequence for GFP was identified, it became possible to introduce this
gene into nonfluorescent cells and make them fluoresce by hijacking the cell’s
own internal machinery. This in turn allowed living cells to be easily observed
with fluorescence microscopy (Figure 2.20). Even more powerful artificial gene
constructs can be created that allow fluorescent versions of normal proteins to be
created by the cell. These can be used to determine their cellular location and traf-
ficking. Constructs can also be created with a target gene’s promoter placed
upstream of GFP to determine gene transcription in real time. Variations of the
original GFP have been created by mutating the protein to change its fluorescent
properties. Pioneered by Roger Tsien, cyan (CFP), yellow (YFP), and red (dsRED)
Figure 2.20 ​Epifluorescence image are commonly available and their numbers continue to grow (more recently, with
obtained by inserting the GFP gene the “fruit”-based labels, such as mCherry, mBanana, etc.).
driven by a promoter that is only
turned on in a specific set of In addition to providing enhanced contrast, fluorescent dyes may change proper-
neurons in a fly. Thus, the cells that ties in response to their environment. Aequorin’s (also originally isolated from
appear bright are expressing the jellyfish) fluorescence increases as a function of calcium concentration. Calcium,
promoter of interest. In this way, GFP as we have seen, is an important cellular second messenger and knowing its con-
can be used to interrogate function in centration in living cells in real time can give important insight into cell metabo-
living cells and tissues. See www.
lism and signaling. A similar calcium reporter dye, Fura-2, also developed by
garlandscience.com for a color version
of this figure. (Adapted from, Alberts B, Roger Tsien, relies on a wavelength shift in the absorbance of the molecule when
Johnson A, Lewis J et al. (2008) Molecular
Biology of the Cell, 5th ed. Garland
Science.)

Example 2.3: Uses of fluorescence

List some possible reasons why one would use fluores- nates multiple components, only the fluorescent tag is
cence microscopy to study cell- or subcellular-level biol- visible under fluorescence microscopy. Fluorescence
ogy. Then describe some disadvantages of using imaging can be calibrated to yield quantitative results
fluorescence. using specialty dyes and techniques.
Fluorescence is desirable because labeling the structure Fluorescence labeling can be time-consuming, expen-
of interest with a fluorescent tag allows you to then image sive, and difficult to achieve. Currently, only a limited
the structure with minimal interference from other parts number of fluorescent labels may be used with living
of the cell. Unlike phase-contrast imaging, which illumi- cells.

it is bound to calcium. Fluorescent reporters for other ions, pH, and even voltage
have been developed.

Atomic force microscopy can elucidate the mechanical

behavior of cells
Originally a derivative of scanning tunneling microscopy (STM), atomic force
microscopy (AFM) replaces the tunneling current with physical contact between
the probe tip and the specimen. An image is formed either by dragging the tip
across the sample or repeatedly pushing on the sample and withdrawing it. AFM
 EXPERIMENTAL BIOLOGY 41

may also expose cells to well-defined mechanical perturbations in studies of their

mechanical behavior and response to mechanical stimulation. We discuss AFM in
more detail in Chapter 6, since it is one elegant way to measure the stiffness of cells.

Gel electrophoresis can separate molecules

Suppose you stretch a group of cells and want to see if production of a particular
protein that affects cell migration is altered in response. One way to do this is to
stretch the cells, lyse them, and then measure the amount of RNA or protein of inter-
est. Another way is to manufacture a short gene that links the promoter of the pro-
tein to a fluorescent molecule, and insert the gene (chemically, physically, or virally)
into the cell. When the promoter is activated, the fluorescent molecule is generated
and you can obtain a visual readout. Any one of these approaches, however, requires
you to be able to isolate, purify, or quantify the molecules of interest.
Gel electrophoresis is used routinely to separate heterogeneous mixtures of DNA,
RNA, and proteins from one another according to their molecular weight. A sam-
ple is forced through a polymer hydrogel and molecules that move quickly through
the gel (typically smaller molecules) are able to travel further in a given amount of
time than molecules that move more slowly (typically larger molecules). The mol-
ecules are driven by an externally applied electric field that produces a force on
them proportional to their charge (Figure 2.21). Nucleotides (RNA and DNA)
inherently contain a negative charge on their sugar backbone (from phosphates)
and therefore travel toward the positive electrode. By contrast, proteins may con-
tain a positive, negative, or no net charge. As a result, proteins are usually placed
in a powerful detergent such as sodium dodecyl phosphate (SDS) that covers the
proteins with negative charge and allows them to migrate through the gel. The
SDS treatment also causes the proteins to unfold or denature when bonds that
contribute to their tertiary and secondary structures are lost because of shielding
of the protein’s internal charge structure by the coating detergent.
Denaturing of proteins before gel electrophoresis has another important ben-
efit. The speed at which a protein migrates through a gel is a function of how
well it can slip through the polymer network of the gel. The drag that is exerted
on the protein as it migrates is, in turn, related to the cross-sectional area that
the protein presents to the network. Although cross-sectional area generally
increases larger proteins, there is not a strict relationship between size and friction

sample loaded onto gel Figure 2.21 A common sodium

by pipette dodecyl sulphate–polyacrylamide
gel electrophoresis (SDS-PAGE)
cathode vertical format system. (Modified
plastic casing from, Alberts B, Johnson A, Lewis J et al.
(2008) Molecular Biology of the Cell, 5th
ed. Garland Science.)

buffer

gel + anode

buffer
42 CHAPTER 2: Fundamentals in Cell Biology

Nota Bene during migration since large proteins can fold into a tightly packed small struc-
ture. Conversely, smaller proteins that are not well packed may actually migrate
Antibodies can be used for
more slowly. In denaturing conditions, proteins acquire an unfolded linear
protein detection and
recognition. Antibodies bind to structure and migrate inversely with their molecular weight.This means gels for
specific protein fragments known as proteins in denaturing conditions (such as SDS) separate samples according to
antigens that contain a sequence of their molecular weight. It should be noted that similar restrictions apply for
interest (the epitope). A polyclonal nucleotide gels—if a DNA is in a circular plasmid configuration, it will appear
antibody is isolated from the blood of smaller than a linear strand of the same number of base pairs (actually, you will
an animal that has been injected with tend to get multiple bands). Therefore, DNA is generally linearized before, it is
a target protein. Polyclonal antibodies
run through a gel.
will bind to several epitopes on the
target protein. A monoclonal antibody The vast majority of gels use one of two polymers to form the hydrogel network.
is produced in cell culture by High-density gels of polymerized acrylamide are used to separate proteins and
immortalizing a single immune cell
from the exposed animal. Monoclonal
smaller nucleic acid polymers. There are many easy-to-use PAGE (polyacrylamide
antibodies recognize a single specific gel electrophoresis) systems and they generally hold the gels between two glass or
epitope. Polyclonal antibodies can plastic plates that force the electric current to pass directly through the gel.
amplify better (more antibodies stuck Typically, these systems hold the gel vertically with the sample moving from the
to the same target) but variability top positive electrode toward the bottom negative electrode. Higher–molecular-
each time you make more. weight nucleotides (over a few hundred base pairs) are separated in lower density
Monoclonals are more restricted in agarose gels. Agarose systems keep the gel horizontal because polymerized aga-
targeting, but the repeatability is
much higher. rose is not strong enough to support its own weight. To keep the gel hydrated, the
gel is submerged and a larger current passes through both the gel and the bathing
fluid. Although automated systems that perform gel electrophoresis in tiny capil-
lary tubes can greatly increase throughput, running agarose and SDS–PAGE gels
remains a standard procedure in most molecular biology labs.

Visualizing gel-separated products employs a variety

of methods
Most gel-running systems allow for several samples to undergo electrophoresis
simultaneously in adjacent lanes of the gel. One lane is typically used to a
molecular-weight standard that is a mixture of proteins or nucleic acids that
have very distinct, known molecular weights. This standard separates into
clearly identifiable bands that are then used to convert the distance migrated
into the molecular weights of the samples. After a sample has been separated, a
particular point in the gel now corresponds to a particular molecular weight. In
this way gel electrophoresis can be used to isolate a fraction of a sample of a
given molecular weight by physically cutting out a particular band and depo-
lymerizing the surrounding hydrogel. Gels of this variety are known as prepara-
tory gels. The purpose of running the gel can be to determine the molecular
weight of an unknown sample. In these so-called analytical gels, the next step in
the process is to visualize the sample. In some cases this can be accomplished
by saturating the gel with a stain that provides color or fluorescence to all of the
proteins or nucleotides in the lane, but more specific target detection is often
needed. This can be accomplished with reagents that bind specifically to the
target of interest (typically antibodies for proteins and probes for nucleotides).
These can then be visualized by radioactively labeling the probe or antibody and
detecting them via autoradiography, attaching fluorescent reporter molecules,
or making them emit light via chemiluminescence. In the latter case, the light is
detected by exposure to highly sensitive X-ray film or by digital cameras.
The binding of antibodies or probes to their targets can involve reactions that
require hours to occur. The hydrogels used in electrophoresis are generally not
structurally sound enough to stand up to this treatment. Therefore a second blot-
ting step is required, in which the sample moves perpendicular to the plane of the
gel and is immobilized on a strong membrane of nylon or nitrocellulose. The
migration of the sample during blotting can also be accomplished by e­ lectrophoresis
or by driving a large volume of water through the gel (Figure 2.22). The blotting of
DNA is referred to as Southern blotting after its inventor, Edwin Southern. In a pun
 EXPERIMENTAL BIOLOGY 43

(A) (B) (C)

unlabeled nucleotide stack of paper towels remove nitrocellulose
nitrocellulose
or protein paper with tightly bound
elec paper
trop nucleic acids
h ores
is gel

labeled nucleotide gel

or protein of known
sizes serving as sponge
probe or antibody
size markers alkali solution
hybridized to separated DNA
sample separated according separated sample blotted onto
to size by gel electrophoresis nitrocellulose paper by suction
of buffer through gel and paper
(D) sealed
plastic
bag

labeled
probe
in buffer

(E)

positions
labeled
of
labeled bands
markers

Figure 2.22 ​Example of the steps involved to detect DNA by its radioactivity. To visualize proteins in a gel, you must first transfer
or blot them onto a membrane. Although this can be done electrophoretically, it can also be accomplished simply by soaking water with
paper towels. The membrane can then be incubated with tagged antibodies or nucleotide probes, and visualized with color or fluorescent
dyes so that the presence of a desired protein or nucleic acid fragment can be detected. (Adapted from, Alberts B, Johnson A, Lewis J et al.
(2008) Molecular Biology of the Cell, 5th ed. Garland Science.)

on the accepted name for the Southern blot, when the same approach was used
with RNA in 1977 at Stanford, it was named the northern blot. The good-natured
name-play continued with the development of the western blot for proteins. The
tradition continues with, for instance, blotting of DNA-binding proteins being
called a southwestern blot. Although there is no eastern blot, a method of blotting
lipids developed in Japan was termed far-eastern blotting.

PCR amplifies specific DNA regions exponentially

The development of the polymerase chain reaction (PCR) by Kary Mullis in 1983
has reshaped modern biology. The dramatic increase in detection power that PCR
44 CHAPTER 2: Fundamentals in Cell Biology

Advanced Material: Antibody uses in biology

Why are antibodies useful in experimental biology? stick to a wide variety of epitopes while not recognizing
Antibodies are proteins that have a “flexible” binding native proteins (so that the antibodies stick to unwanted
site. These proteins are typically drawn in the shape of a stuff but not to parts of your own body). In some cases,
“Y”. The binding site is usually targeted to a desired pep- this is not desirable; if you receive an organ transplant
tide sequence called an epitope, which is usually part of you do not want your immune system to attack it, but
a foreign organism (say, a surface protein on some virus). since it is  technically a foreign body, you will generate
The idea is that when a foreign organism invades your antibodies against it, so immunosuppresants will gener-
body, your immune system generates antibodies that are ally be ­necessary). In other cases, there are mistakes—in
capable of binding key surface proteins on the foreign autoimmune syndromes, your immune system gener-
organism, thereby rendering the proteins inactive. Since ates antibodies against your own tissues, which can
there are multitudes of foreign organisms, the immune cause all sorts of complications (lupus is a classic exam-
system must be capable of generating antibodies that ple of an autoimmune dysfunction).

provides has led to it be a virtually ubiquitous tool in the modern biology lab. In
recognition of the incredible importance of PCR, Mullis received one of very few
Nobel prizes to be awarded for a technique rather than a scientific discovery. The
key to PCR is to use a natural cellular enzyme, DNA polymerase, to replicate DNA
in vitro rather than in the cell nucleus. As each new DNA strand is synthesized, it
becomes in turn a template for subsequent synthesis in a chain reaction that pro-
duces an exponential increase in DNA and gives PCR its remarkable power to
detect even a single molecule of DNA in the initial sample. At its core, PCR is made
possible by two key components: the requirement of primers for DNA polymerase
binding and the identification of thermally stable DNA polymerases.
PCR would be much less useful if it amplified all DNA equally, but in fact it can
selectively amplify a target sequence of DNA (corresponding to a particular
gene, say) and leave non-target DNA alone. This is done by exploiting a critical
property of DNA polymerase. Specifically, DNA polymerase only adds nucleo-
tides where a portion of single-stranded DNA already has some double-stranded
DNA in place. During replication double-stranded DNA is separated into two
single-strands. DNA polymerase converts each single-strand back to a double-
strand by adding complementary bases to the 3′ end of the newly synthesized
strand in a process known as extension. The fragment of complementary DNA
that binds to the DNA being replicated and allows DNA polymerase to get
started is known as a primer. During cell division these primers are made by
specialized enzymes and allow the entire DNA strand to be replicated. By con-
trast, in PCR, primers are designed to flank a specific region of DNA and amplify
only that DNA fragment. Since DNA polymerase only proceeds from the 3′ end
of the primer and adds new base pairs in the 5′ direction, the flanking primers
are directional. That is, a forward primer is designed to be complementary to a
sequence of DNA that is unique and located upstream (in the 5′ direction) of the
target. It will produce a long strand of complementary DNA starting at the for-
ward primer and extending along the DNA template. A reverse primer is designed
to bind uniquely to the complementary DNA at a point downstream of the tar-
get. The end result is that with each cycle of extension the number of fragments
produced that are flanked by the primers doubles. In short order, the exponen-
tially growing fragments vastly outnumber any other DNA sequence in the reac-
tion (Figure 2.23).
In the 1970s scientists were already using DNA polymerase to replicate DNA.
However, since DNA polymerase can only add bases to single-stranded DNA, they
had to first separate the two strands by heating to the melting or denaturing tem-
perature (typically 95°C) and allowing it to cool to around 65°C, the temperature
 EXPERIMENTAL BIOLOGY 45

(A)
5′

+DNA polymerase DNA

heat to +dATP
hybridization synthesis
separate +dGTP
of primers from
double- strands +dCTP primers
stranded +dTTP
DNA 5′

Step 1 Step 2 Step 3

First Cycle

(B) separate the DNA

separate the DNA DNA strands and synthesis
DNA strands and synthesis anneal primers
separate the DNA anneal primers
DNA strands and synthesis
anneal primers

etc.
DNA oligonucleotide
primers

region of
double-stranded
chromosomal
DNA to be
amplified

First Cycle Second Cycle Third Cycle

(producing two double-stranded (producing four double- (producing eight double-
DNA molecules) stranded DNA molecules) stranded DNA molecules)

Figure 2.23 ​Schematic of DNA amplification in a typical polymerase chain reaction (PCR). The fragment flanked by the forward
and reverse primers grows exponentially with the number of amplification cycles, while the other fragments grow linearly. Therefore the
concentration of the fragment flanked by the primers quickly overwhelms all of the other species. (Adapted from, Alberts B, Johnson A,
Lewis J et al. (2008) Molecular Biology of the Cell, 5th ed. Garland Science.)

at which extension can occur (the annealing temperature). The problem was that
each time the sample was raised to the melting temperature, all of the DNA poly-
merase was denatured and needed to be re-added manually. This was very cum-
bersome and greatly limited yield. The answer came from the realization that
thermophilic bacteria thrive in high-temperature environments. Therefore, they
must have a form of DNA polymerase that can withstand high temperatures. One
of the most common forms of DNA polymerase used today is Taq polymerase
which was originally isolated from Thermus aquatiqus, a bacterium that naturally
occurs in hot springs and hydrothermal vents. With this component in place, it
was possible for modern PCR to be developed. In principle, it is simply a matter of
mixing a sample, DNA polymerase, forward and reverse primers, and an abun-
dance of DNA monomer in a tube and alternating quickly between high and low
temperatures. This is typically done in a solid-state device developed just for this
purpose known as a thermal cycler. The reaction will continue and the target is
exponentially amplified until one of the reagents is exhausted.
46 CHAPTER 2: Fundamentals in Cell Biology

2.4 ​EXPERIMENTAL DESIGN IN BIOLOGY

One striking difference between mechanics and biology is the vast diversity of
biological systems. In the case of the cell, seemingly unending systems of regula-
tion and control can be particularly intimidating to an engineer. The number of
genes in a human (~40,000) means that there are about the same number of pro-
teins that can potentially interact within a cell (more, when you consider alterna-
tive splicing and post-translational modifications). Although not all proteins are
expressed in a particular cell at one time, the number of interactions is clearly too
large for pen-and-paper analysis. Further, unlike most macroscopic mechanical
systems, few components can be directly probed or visualized at the molecular
level. As a result, most studies in cell biomechanics and mechanotransduction
involve probing a limited aspect of overall cell behavior. Because of this, ad hoc
snippets of behavior and knowledge predominate—a scientific paper entitled “X
controls Y in a Z-dependent manner” suggests that the regulation of Y by X can be
influenced by blocking or overexpressing Z. This situation is further exacerbated
by the fact that currently so much of the cell’s function and machinery remains
virtually unknown. Of course, this is what makes the study of the cell and its
diverse workings so exciting to many of us. Nonetheless, there are often funda-
mental differences in how experiments are conducted in biology versus mechan-
ics, and fundamental differences in how the results can be interpreted. In cell
mechanics it is not unusual for two different experiments to provide disparate
answers to the same question. Consider the estimated value for cell stiffness
across multiple techniques ranges from 100 Pa to 1 MPa. The fact that different
methods were used on different cell types under presumably different cell culture
conditions makes interpretation of these results all the more difficult.
In experimental mechanics the goal is generally to recreate the real-world situa-
tion with as much detail and fidelity as possible. If a material is to be character-
ized, one attempts to recreate its physical, chemical, and mechanical environment
to the greatest extent possible. The environment of the cell is elaborate, making
this task challenging indeed. When engineers first work with cells they can be
tempted into visualizing working within a three-dimensional, multi-phase, physi-
cally and chemically representative model of a given cellular environment, with
an abundance of variables to control. One should realize that many of the critical
insights in cell biology have come from taking a more reductionist experimental
approach, and engineers, although appropriately skeptical of how these findings
might translate to more complex situations (such as an intact organism), should
become familiar with them.

Reductionist experiments are powerful but limited

Reductionism refers to the process of understanding the very basic building
blocks of a system, and then predicting macroscopic phenomena as a conglomer-
ate of their behavior. Reductionist experimental design involves stripping the sys-
tem of as much complexity as possible, while preserving the key behaviors under
study. If one wants to study DNA replication, it is often sufficient to reproduce the
biochemical reactions in a test-tube without having to reproduce the environ-
ment of an intact cell. Indeed, current cell culture techniques are based on simpli-
fying the cell’s external environment. Tissue matrix and substrates are replaced by
plastic or glass, sometimes coated with adhesion molecules. The cell culture
medium is often designed to encourage cell proliferation and activity, sometimes
at the expense of realism and consistency. One frequently used supplement to
encourage proliferation is bovine serum, which is used not only for culturing cow
cells but cells from mice, rats, humans, etc.
These simple systems can yield important insights. Much of modern medicine
and biology is based on discoveries made in the cell culture lab. But by its very
nature, a simple system designed to produce straightforward results in an
 EXPERIMENTAL DESIGN IN BIOLOGY 47

e­ conomical manner is scientifically limited. These systems do not address the

question of whether the finding will translate when complexity is reintroduced.
Verifying that a discovery made in a simple system holds true in a cell or organism
is just as important as making the initial insight. Indeed, it has been reported that
treatments and drugs that work in laboratory mice do not usually translate to
humans—suggesting that creating a representative biological environment is no
guarantee that the results will be any more useful than in simplified systems. We
therefore consider simplified systems to be models of something much more
complicated. One might say that the model system is useful in showing that some-
thing can occur, but it must be followed by a determination of whether it indeed
occurs in reality. There is always the danger that a cell can be made to do some-
thing in the laboratory that it never does as part of a functional organism. There is
also the alternative, that something can be done in an organism that cannot be
reproduced in culture (for example, a hepatitis virus may freely infect and repli-
cate in an organism, but this is actually difficult to achieve in cell culture). This
latter issue is rarely considered, because random experimentation on whole
organisms can be difficult, expensive, and can carry a higher ethical burden.
Designing appropriate model systems and knowing their limitations is critical to
successful science and is often as much the product of intuition and personal
experience as rational logic and fact.
Another advantage to studying simple systems is that they more easily allow
investigators to understand mechanisms. This concept is often proffered as a con-
trast to so-called descriptive experiments. This is a subtle argument because all
science ultimately involves describing something or other. However, there is a
bias, sometimes unspoken, that it is better to understand how something ­happens
rather than merely to describe the fact that something happens. Showing that a
particular mechanical stimulation may encourage stem cells to differentiate into
a particular cell type might be descriptive, whereas determining the receptor that
mediates this would provide a groundwork of the mechanism. One advantage to
understanding the mechanism behind a particular behavior or response is that
this knowledge takes us one step closer to being able to manipulate the response
to a desired therapeutic end.
Investigations of proposed or hypothetical mechanisms generally involve an
intervention or series of interventions applied to a system that models the desired
behavior. Suppose we can show that, a given mechanical stimulation leads to cel-
lular differentiation. We hypothesize that a particular receptor is responsible.
Repeating the experiment in cells that have had this receptor removed and observ-
ing a failure to differentiate with the same stimulation would help support this
hypothesis (hypotheses can be supported, but not proven). Another approach
might be to use a biochemical blocker to suppress a particular pathway or target
molecule. Currently, blockers exist with varying degrees of specificity for kinases
and phosphatases, intracellular and extracellular receptors, ion channels, and
some other assorted signaling molecules. “Blocking antibodies” bind to specific
target proteins, but also inhibit their function—usually by interfering with the
binding site for the ligand. A recently developed intervention is small inhibitory
RNA (siRNA). It was discovered that as part of the cell’s natural defense against
certain viral attacks, the cell responds to the presence of short fragments of
­double-stranded RNA by producing RNA that is complementary to the introduced
sequence. These fragments bind to the target mRNA and mark it for degradation.
Production of a specific protein can be interrupted (or knocked down) at the post-
transcriptional level by using knowledge of the gene sequence.
If a response to a stimulus is lost when a particular protein or signaling event is
blocked, it is strong evidence that the blocked activity is involved in the pathway
between the stimulus and the response. Often the activity is termed necessary for
the response. An additional level of support can be shown if the activity can be
supplied through exogenous means and the response observed even in the
absence of the original stimulation. In this case, the activity is said to be sufficient
48 CHAPTER 2: Fundamentals in Cell Biology

for the response. When a step in a pathway is both necessary and sufficient, it is
strong evidence indeed. While this is the current strategy behind biological and
biomedical engineering approaches to mechanotransduction, it should be
acknowledged that this strategy also has weaknesses. For instance, when a new
blocker is discovered, it is impractical to test the blocker against every possible
combination of molecules. As a result, only a few controls are used to assess the
efficacy and specificity of the blocker. Blocking is also imperfect—siRNAs do not
typically completely knock down the end protein expression, rather they suppress
expression to varying degrees, depending on the cell type, design of the fragment,
and method of introducing the siRNA to the cell.
Because of the nature of signaling, it is not entirely accurate to discuss elements in
isolation. If stem cell differentiation by mechanical stimuli can be blocked by
removing compound A, and restored by overexpressing compound A in the
absence of stimuli, we might conclude that compound A, not mechanical stimuli,
is the necessary (and sufficient) ingredient for differentiation (and if we show that
mechanical stimuli up-regulate compound A, we might have the beginning of a
signaling pathway). Such a linear approach is useful in the short term, but may be
an oversimplification. For instance, there is no distinction between a major player
(engine’s spark plug) and a necessary component that is not a major player (igni-
tion key). For engineering purposes, the former is far more useful than the latter,
but often there is no way to tell which of these two is being investigated without far
more comprehensive experiments. It is possible to have two or more necessary
and sufficient pathway components, that do not have any overlap, but describe
the same process. These processes are then typically discovered to be earlier
(upstream) or later in the pathway (downstream).

Modern genetics has advanced our ability to study in situ

Reductionist approaches are powerful and particularly useful in generating
potential molecular targets for further investigation. However, these models are
limited and there is always a concern that what is observed in the laboratory may
not occur in living organisms. Modern genetics has led to a wide variety of
approaches that can help answer this question. For instance, transgenic animals
(typically mice) can be created with a particular gene removed (knocked out) from
their genomic DNA. The case for the function of that particular protein is greatly
strengthened if the expected loss of function is seen in the phenotype of the trans-
genic animal.
Often knockout animals have such functional impairments that they are not via-
ble. The limitations of global deletion approaches can be overcome by using a
site-specific recombinase (SSR). Rather than deleting a gene from all of the cells in
an organism, SSR technologies allow deletion in only certain cells or tissues. The
most popular SSR system is the Cre-Lox strategy, which relies on the activity of
Cyclic Recombinase (Cre). Isolated from a class of viruses that infect bacteria, Cre
recognizes a targeted gene or gene fragment and deletes it. By creating an artifi-
cial genetic construct in which Cre is placed downstream of a particular pro-
moter and delivering the construct to an animal, expression of Cre (and hence
gene deletion) is limited to only those cells in which the promoter is activated.
Promoters are available that restrict Cre to specific tissues and cells. Inducible
promoters are even available that can activate Cre expression at the time of the
investigator’s choosing by administration of an otherwise innocuous drug. By
studying the loss of function in an animal with a missing protein, the function of
that protein can sometimes be inferred. Perhaps the most convincing support for
a proposed function is a rescue experiment in which reintroducing the deleted
gene restores the lost function (the analog of necessary and sufficient experi-
ments in organisms instead of in cell culture). These tools and others have
sparked a revolution in genetics and allow translation and validation of insights
made in simple systems.
 EXPERIMENTAL DESIGN IN BIOLOGY 49

Bioinformatics allows us to use vast amounts of genomic data

A dramatic advance in the capacity of computer technology is one of the defining
characteristics of the last decade. The Information Revolution has led to advances
in the computing power available to scientists and their ability to compute and to
transmit vast amounts of information. In biology, the sequencing of the human
genome has led to vast quantities of data as a result of the large-scale, high-
throughput biotechnologies. Sequencing the genome of entire organisms, once
thought impractical, is now relatively routine (the current price for sequencing an
entire individual human is $200,000, down from the millions the genome project
cost. The exome, the part of the genome formed by exons, can be sequenced for
less than that $1,000.). The complete genetic sequences of hundreds of organisms
are now available to anyone online.
Microarray technology allows investigators to determine which genes and mRNA
sequences a biological sample (cells, tissues, or even an organism) is expressing at
any given time. Modern mass spectrometers allow the identification of hundreds
to thousands of the proteins produced by a cell over time. This deluge of informa-
tion and the computer power to analyze it has led to the formation of a new field,
bioinformatics.
In terms of genetic information or genomics, the ability to search a genome data-
base for a specific sequence has greatly simplified primer/probe design. Pieces of
sequence can be added or deleted, allowing the creation of new gain- or loss-of-
functional mutations to dissect how specific proteins work. Searching for sequence
homology allows proteins to be identified that might have similar structure and
function. Functional annotation allows rapid identification of what role a particu-
lar gene might play in organisms or regulatory and signaling pathways. A similar
advance in high-throughput protein sequencing is the foundation of proteomics,
a field that is only now beginning to yield benefits. As a field, bioinformatics is
extremely new. No longer merely a tool supporting traditional approaches, it has
grown into a paradigm in which unique hypotheses can be advanced and tested.
Although it is not possible to predict the future, large-scale, information-based
approaches are likely to continue to produce important insights and tools.

Systems biology is integration rather than reduction

Bioinformatics has led to the development of another new field in modern biology.
Systems biology espouses a philosophy of integration rather than reduction. At the
beginning of the twenty-first century, the systems approach began incorporating
the interaction of a biological network in all of its inherent complexity, rather than
reducing it to its simplest isolated components. It is expected that biological net-
works, because of their highly interconnected nature, can be simulated, and emer-
gent behaviors can be identified. Such properties would be indicative of the
network itself and could not be understood in terms of the behavior of any single
component. The databases and information manipulation tools from bioinformat-
ics are significant enabling technologies for systems biology. Systems analysis has
been applied to gene expression patterns from DNA microarrays, protein expres-
sion as determined by mass spectrometry, analysis of all of the small molecule
metabolites in a cell or tissue, and simulation of the kinetics of kinase/phosphatase
signaling cascades and regulatory networks. Traditional reductionist tools have
also been scaled up to the systems approach with systematic exposure of cells to
libraries of chemical blockers and arrays of siRNAs. Although examples of emer-
gent behavior are still few and therapeutics nonexistent, it is clear that much of cell
biology can only be understood in terms of complex network behavior.

Biomechanics and mechanobiology are integrative

Finally, it is important to acknowledge the differences in approaches between
biologists and engineers. Traditionally, the former tend to use detection methods
50 CHAPTER 2: Fundamentals in Cell Biology

that are empirical, coupled with statistical testing. Mechanistic approaches are
important, in terms of elucidating regulatory pathways. Repetition and rigorous
protocols are highly valued because of the degree of variability found among sam-
ples. By contrast, engineering approaches tend to be mathematical. They value
long-established physical principles that can be applied analytically or simulated.
To study biomechanics/mechanobiology, an appreciation for both approaches is
desirable. Not all aspects of mechanobiology are easily modeled and not all
aspects of biomechanics can be experimentally measured, but one key to
approaching this sort of research is identification of where the two areas are close
and attempting to bridge the gap between them.

Key Concepts

• Mechanobiology deals with the biological response of • Second messengers transfer information within the
cells to mechanical stimuli. cell quickly, but are broad in effect. Protein signaling
• Structurally, cell components are influenced by cascades are more specific, but act slower.
polymers. Proteins are amino acid polymers. DNA and • Microscopy is used to examine cells visually. Many
RNA are nucleic acid polymers. Polysaccharides are fluorescence techniques exist to help probe cell
sugar polymers. Lipids are fatty acid polymers. behavior.
• The central dogma of cell biology states that proteins • Gels electrophoresis is used to separate molecules
are translated from mRNA, which in turn is for analysis of various biological polymers.
transcribed from DNA. • Biological experiments tend to be reductionist,
• Cell organelles play various roles in cell physiology, mechanistic, and hypothesis-driven. Engineering
but several components are also involved in making approaches are more quantitative but complement
or modifying proteins. biological approaches.
• Receptors are proteins that span the cell membrane
and receive external signals and engage signaling
pathways inside the cell.

Problems

1. Describe a scenario whereby stretching a cell will lead two proteins, X and Y, are engaged at a particular time.
to changes in transcription of some gene, with calcium How would you experimentally determine this, given
signaling as an intermediary. Your response does not that you do not have access to a microscope sensitive
have to be based on a real pathway, but should involve enough to see the proteins themselves?
the material discussed in this chapter.
5. Here is an abstract from a 1997 article in Science (275,
2. You are examining a cell that was stained for a 1308–1311): “The small guanosine triphosphatase
component of the cytoskeleton called actin. You see a (GTPase) Rho is implicated in the formation of stress
single fiber at a location but suspect there may be two fibers and focal adhesions in fibroblasts stimulated
fibers instead. The objective you are using is a 40×, NA by extracellular signals such as lysophosphatidic acid
1.3 oil immersion. What is the closest distance between (LPA). Rho-kinase is activated by Rho and may mediate
two actin fibers at which you can still distinguish them? some biological effects of Rho. Microinjection of the
catalytic domain of Rho-kinase into serum-starved
3. A particular gene is known to be activated upon
Swiss 3T3 cells induced the formation of stress fibers
mechanical stimulation. When you assay the mRNA
and focal adhesions, whereas microinjection of the
concentration 1 hour after applying stimuli, you do
inactive catalytic domain, the Rho-binding domain, or
not see any changes. List possible reasons for this
the pleckstrin-homology domain inhibited the LPA-
observation.
induced formation of stress fibers and focal adhesions.
4. Some signaling pathways will involve multiply-linked Thus, Rho-kinase appears to mediate signals from Rho
proteins that either engage or disengage in response to and to induce the formation of stress fibers and focal
some stimuli. You are interested in examining whether adhesions.”
 Problems 51

To help you better understand the abstract, please sufficient, or whether one cannot tell, the down-
note the following: Swiss 3T3 cells are a type of regulation of gene given the findings below. Note that we
fibroblast from mice; microinjection is using a special do not know where in this pathway N acts, but under
microsyringe to inject selected chemicals into cells normal conditions mechanical stimuli will also increase
directly, rather like a shot from a conventional syringe. N phosphorylation.
Do not worry about the pleckstrin-homology domain
mechanical stimuli → calcium increase → gene 1
(which means the local protein sequence is nearly
up-regulation → gene 2 down-regulation
identical to that of a protein called pleckstrin). Do not
worry about what LPA actually does, except for the (a) U nder mechanical stimulation N phosphorylation
function outlined in the abstract. was inhibited by a chemical. Gene 1 failed to up-
(a) B
 riefly explain the hypothesis of the authors and regulate, but gene 2 was down-regulated.
how they tested the hypothesis. (b) Without mechanical stimuli, phoshorylated N was
(b) If you had a technique for measuring how stiff a induced by injecting phosphorylated N directly into
cell is, and you compared LPA-treated cells versus cells. Gene 2 was down-regulated.
untreated cells, which would I expect to be stiffer? (c) Under mechanical stimuli, calcium changes were
(c) Why did the authors serum-starve the cells? suppressed by a chemical. N exhibited increased
(d) If you had unlimited resources, what other controls phosphorylation, but gene 2 was not down-
would you add to what the authors did to further regulated.
test their hypothesis? 10. Discuss the following statement. “To produce one
6. If you wanted to examine the expression of a particular molecule of each possible kind of polypeptide chain,
protein, there are several molecular readouts you can 300 amino acids in length, would require more
use. Suppose you predict that in response to stretching atoms than exist in the universe.” Given the size of
a cell for an hour, the protein SSP (stretch-sensitive the universe, do you suppose this statement could
protein) will be more abundant in the cell. possibly be correct? Since counting atoms is a tricky
business, consider the problem from the standpoint
(a) H
 ow can you experimentally validate your
of mass. The mass of the observable universe is
prediction?
about 1080 g, give or take an order of magnitude or
(b) S
 uppose you did the experiment from (a) and
so. Assuming that the average mass of an amino acid
found that SSP increases. What would you think
is 110 Da, what would be the mass of one molecule
if a colleague told you that she also stretched the
of each possible kind of polypeptide chain 300 amino
cell but the SSP mRNA levels decreased? Does it
acids in length? Is this greater than the mass of the
matter when she assayed the cells?
universe?
7. A gene mutates so that the generated protein has
a single amino acid substitution somewhere in the 11. It is often said that protein complexes are made from
middle whereby an Asp is replaced with a Glu. (You may subunits (that is, individually synthesized proteins)
consult sources to determine what these side chains rather than as one long protein because the former is
actually are.) more likely to give a correct final structure.
(a) W
 hat difference in protein structure, if any, do you (a) A
 ssuming that the protein synthesis machinery
expect to result from this substitution? That is, do incorporates one incorrect amino acid for
you expect the protein to fold exactly the same each 10,000 it inserts, calculate the fraction of
way? Why or why not? bacterial ribosomes that would be assembled
(b) N
 ext you grab the ends of the protein and try to correctly if the proteins were synthesized as
pull it apart. You find that the protein is “stiffer” one large protein versus built from individual
(i.e., harder to extend) with the Glu mutation. Could proteins. For the sake of calculation, assume
this affect the signaling capacity of the protein? If that the ribosome is composed of 50 proteins,
so, how? each 200 amino acids in length, and that the
subunits—correct and incorrect—are assembled
8. We are interested in the interactions of two proteins, with equal likelihood into the complete ribosome.
JP (junctional protein) and JPAP (JP-associating [The probability that a polypeptide will be made
protein). We know these proteins are linked at the cell correctly, Pc, equals the fraction correct for each
boundaries between adjacent, adhering cells. We also operation, fc, raised to a power equal to the
know that JPAP will dissociate (i.e., separate from) JP number of operations, n, Pc = (fc)n. For an error
when the cell is mechanically stimulated. We do not rate of 1/10,000, fc = 0.9999.]
know whether JP leaves the cell boundaries when (b) Is the assumption that correct and incorrect
JPAP dissociates from it. Describe an experiment using subunit assembly is equally likely true? Why or why
fluorescent-conjugated JP and JPAP to determine not? How would a change in that assumption affect
whether JP leaves the cell boundaries in response to the calculation in part (a)?
mechanical stimuli.
12. If a sample of human DNA contains 20% cysteine (C) on
9. For the following experimental results involving protein a molar basis, what are the molar percentages of A, G,
N, determine whether N phosphorylation is necessary, and T?
52 CHAPTER 2: Fundamentals in Cell Biology

13. All small intracellular mediators (second messengers) 17. A typical mammalian cell is about 1000 mm3 in volume.
are water soluble and diffuse freely through the cytosol The concentration of protein within a typical cell is
(indicate true or false, and explain why). 200 mg/ml. Using western blotting, you can detect
10 ng of a specific protein from loading 100 mg of total
14. Cells communicate in ways that resemble human protein. The specific protein you are interested in has a
communication. Decide which of the following forms molecular weight of 100 kDa.
of human communication are analogous to autocrine,
(a) H ow many cells do you need to harvest to be able
paracrine, endocrine, and synaptic signaling by cells and
to load 100 mg of total protein?
briefly say why.
(b) How many copies of a given protein per cell are
(a) A telephone conversation required to detect the western band?
(b) Talking to people at a cocktail party
(c) A radio announcement 18. After treating cells with a chemical mutagen, you isolate
(d) Talking to yourself two mutants. One carries alanine and the other carries
methionine at a site in the protein that normally carries
15. Why do signaling responses that involve changes in valine. After treating these two mutants again with the
proteins already present in the cell occur in milliseconds mutagen, you isolate mutants from each that now carry
to seconds, whereas responses that require changes in threonine at the site of the original valine. Assuming
gene expression require minutes to hours? that all mutations involved single nucleotide changes,
deduce the codons that are used for valine, methionine,
16. You want to amplify DNA between the two stretches of
threonine, and alanine at the affected site. Would you
sequence in the figure. Of the listed primers, choose the
expect to be able to find valine-to-threonine mutations in
pair that will allow you to amplify the DNA by PCR.
one step?
DNA to be amplified

5′-GACCTGTGGAAGC CATACGGGATTGA-3′
3′-CTGGACACCTTCG GTATGCCCTAACT-5′

Primers

(1) 5′-GACCTGTCCAAGC-3′ (5) 5′-CATACGGGATTGA-3′

(2) 5′-CTGGACACCTTCG-3′ (6) 5′-GTATGCCCTAACT-3′
(3) 5′-CGAAGGTGTCCAG-3′ (7) 5′-TGTTAGGGCATAC-3′
(4) 5′-GCTTCCACAGGTC-3′ (8) 5′-TCAATCCCGTATG-3′

Annotated References

Alberts B, Johnson A, Lewis J et al. (2008) Molecular Biology of the 335–350. This is the original article that describes the PCR reaction,
Cell, 5th ed. Garland Science. A comprehensive text on molecular although there are previous publications reporting its use. The tech-
biology that covers in great detail the process of transcription and nique was rapidly adopted.
translation, and the structure and function of cells. This text is a
Periasamy A (2001) Methods in Cellular Imaging. Oxford University
useful reference not only for Chapter 2, but for many parts of the
Press. A good, broad introductory reference to imaging techniques
rest of the book.
used in cell biology.
Lodish H, Berk A, Kaiser CA et al. (2007) Molecular Cell Biology, 6th
Watson JD (2001) The Double Helix: A Personal Account of the Dis-
ed. W.H. Freeman. Another good textbook on molecular biology.
covery of the Structure of DNA. Touchstone. An autobiographical
Actually, there are several texts and you may choose the one that is
explanation of the discovery of the double-helical structure of DNA
most appropriate for your level of interest.
by one of the Nobel Prize–winning team.
Mullis KB & Faloona FA (1987) Specific synthesis of DNA in vitro
via a polymerase-catalyzed chain reaction. Methods Enzymol. 155,
CHAPTER 3

Solid Mechanics Primer

I n the field of cell mechanics, there is more than one way to understand the
mechanical behavior of cells. Our review of basic solid mechanics begins with
rigid-body mechanics and goes on to consider small deformations in solid bodies.
We discuss simple loading configurations, including axial, torsion, and bending
configurations. From there we discuss some aspects of large deformation mechan-
ics, although a complete treatment is beyond the scope of this text.

3.1 RIGID-BODY MECHANICS AND FREE-BODY

DIAGRAMS
What is a “rigid” body?
In this section, we confine ourselves to purely mechanical changes in state to a
given body. We exclude such things as chemical, radioactive, or other changes to
overall mass. Mechanical changes can be described as a combination of transla-
tion, rotation, and deformation to the body. A surprising amount can be learned
by ignoring the deformation and only considering translation and rotation. By
considering only the balance of forces on a body, we are able to determine its
acceleration and vice versa.
First, we consider all bodies to be rigid—that is, undeformable. This assumption is
only good for examining force balances and approximating distributions; under
close examination it will lead to ludicrous conditions. We will see later that stress
is the force applied to a given surface divided by the surface area. If a perfectly
rigid sphere hits a perfectly rigid flat wall, the stress at the point of impact is infin-
ity because there is zero area in the single point of contact. For now, we ignore
these issues and focus on the broader problem of determining force balance.

One of the most powerful, but underused, tools

is a free-body diagram
Free-body diagrams are representations of an object, a part of an object, or a col-
lection of objects for which all external forces are described. They are derived
from the fact that Newton’s laws of motion apply not only to complete structures,
but also to all components or sub-structures we care to define. The fundamental
process in a free-body diagram analysis is to (1) identify a component (or compo-
nents) of interest by surrounding it with an imaginary boundary, (2) imagine
removing this portion of a larger structure or system, and (3) identify all forces
across our imaginary boundary that must be acting to replace the external sur-
roundings. The forces that act across our imaginary boundary may be known
fixed forces or they may be unknown internal forces. Therefore, the key is know-
ing where to place the boundary so that the desired forces can be obtained. As we
will see, in deformational mechanics the fundamental equations of equilibrium
in solid mechanics, or the Navier–Stokes ­equation in fluid mechanics, are the
54 CHAPTER 3: Solid Mechanics Primer

Figure 3.1 A crawling cell can pull direction of travel

itself forward. The cell is crawling from
left to right. It has produced an
extension called a pseudopod and is pseudopod
pulling itself forward with it.

forward attachment point

result of free-body diagram analysis. For now, let us ignore deformation and
focus on problems in which the forces are simple and discrete.

Identifying the forces is the first step in drawing

a free-body diagram
It is important to identify all of the forces applied to the isolated component.
Explicitly drawing the imaginary boundary needed to isolate the structure of
interest from the larger system will ensure that no mistakes are made. Consider a
cell that is slowly crawling away from a wall that has left a cellular extension
behind, pulling the cell back toward the wall (Figure 3.1). Assume that the motion
is slow enough that it can be ignored (velocity = constant = 0).
The free-body diagram of the cell can be constructed by first drawing the imagi-
nary boundary that would isolate the structure of interest (the cell). Then we
redraw the cell in isolation, but representing the forces acting across the boundary
that would need to be applied to keep the cell in place (Figure 3.2).
A frequent error in drawing the free-body diagram of the cell is to locate the force
at the extension tip pointing toward the cell—it is wrongly imagined that the ten-
sion in the extension is pulling inward toward the cell. If you remember that the
forces in a free-body diagram are those that act across the boundary needed to
replace the external environment, you will never go wrong. The force being drawn
is the force of the wall on the cell, which is the reaction force in the opposite direc-
tion and of equal magnitude. Because the cell has a constant velocity (of nearly
zero), there must be a force that counterbalances it (Newton’s first or second law,
net change in acceleration is zero), and this is at the base of the cell. The arrows
down and up represent the weight of the cell and the reaction force from the
ground. These are both external forces, as gravitation is the force exerted by the
planet on the cell, and the reaction force is the force exerted by the substrate on
the cell.

Influences are identified by applying the equations of motion

Once the free-body diagram is complete, one may then apply Newton’s second
law (ΣF = 0, or ΣF = ma if there is acceleration) and moment balance [ΣM = 0, or
ΣM = Iα where I is the (mass, or first) moment of inertia and α is the angular

Figure 3.2 The imaginary boundary (A) (B)

around the cell (A) and the resulting
free-body diagram (B). The arrow
inside the cell represents its weight.
 MECHANICS OF DEFORMABLE BODIES 55

acceleration if there is rotation]. For cases in which accelerations are permitted,

we may use the inertial vector and the moment of inertial torque on the free-
body diagram to have all influences displayed. On the scale of a typical cell, how-
ever, inertial forces are generally small and therefore can generally be ignored.

Free-body diagrams can be drawn for parts of objects

In the example above, the forces between the cell and substrate are easy to iden-
tify because they are clearly separate things. However, free-body diagrams can
also be very useful for determining forces through something, even when there is
no obvious separation or boundary. Remember, we can draw our imaginary
boundary anywhere we like, including right through a solid structure, to identify
the forces acting internally. Consider the mechanics of a polymer network, which
is considered in more detail later (Figure 3.3).
Say you wish to know the forces in the individual polymers as a function of the Figure 3.3 Schematic of a polymer
overall force on the network. We can again do this by constructing a free-body network. Individual polymers are
diagram. Start by drawing our imaginary boundary (Figure 3.4). represented as springs connected by
freely rotating hinges or pins. We would
We selected this boundary because it takes advantage of the symmetry of the like to determine the internal forces on
problem. In addition, it cuts through the center polymer and we would like to the polymers as a function of the total
determine the force going through that element of the network. Next, we isolate force on the network.
the portion of the structure within the boundary and replace the boundary
with the forces that must act across it to maintain equilibrium (Figure 3.5).
The force balance and moment balance can now be applied. However, there are
some important things to note. First, we could have used the left part of the network
to draw the free-body diagram. This would have produced the same results. Second,
we have drawn all of the horizontal forces pointing to the left. A quick inspection of
the network reveals that the center polymer is under compression. Therefore, it
might have made more sense to draw this arrow pointing to the right. Actually, it
does not make a difference—drawn as it is, the analysis produces a negative value
for this force, indicating that the force is indeed applied from left to right.
For this free body, we now apply the force and moment balance. Notice that we
know the downward force on the lower pin. It is simply half the applied force (by
symmetry we know that the other half of the applied force is supported by the left-
hand side of the network). We can also see that force in the middle polymer is twice
the horizontal component of force in the obliquely oriented polymers. However,
we cannot determine the magnitude of these forces. We say that it is statically inde- Figure 3.4 Imaginary boundary drawn
terminate. Nevertheless, we can create another free-body diagram—there is noth- through the polymer network.
ing to prevent more than one free-body solution or analysis for a given structure.
Specifically, it is informative to isolate each of the pins (Figure 3.6), which tells us
that the vertical and horizontal forces at each pin must sum to zero. With this infor-
mation, and some simple geometry, we can solve the problem relatively easily.

3.2 MECHANICS OF DEFORMABLE BODIES

Rigid-body mechanics is not very useful for analyzing
deformable bodies
Although we can learn a great deal about the mechanical behavior of structures by
considering their elements to be rigid, this approach is limited. It cannot tell us
about the distribution of forces within a structure. To do this we need to consider
deformation. When we first consider that objects that appear solid and hard can
actually deform, it can be difficult to understand intuitively. However, in reality, Figure 3.5 Component of the
even the most rigid material undergoes deformations when exposed to loading. network isolated from Figure 3.4.
There are forces across the boundary at
In the mechanics of rigid bodies, we were concerned with characterizing the the upper and lower pins as well as the
forces and displacements of a structure. Because we idealized the behavior of the middle polymer.
56 CHAPTER 3: Solid Mechanics Primer

Figure 3.6 Free-body diagram for (A) (B) (C)

the lower pin (A), middle pin (B),
and upper pin (C). Notice that we
have reversed the direction of the force
in the middle polymer from Figure 3.5 to
better reflect its compressive nature.

F individual components as rigid, we did not need to be concerned with the

mechanical properties of the materials. When we now relax this assumption and
consider material deformations, we take an important step forward in developing
A1 our mechanical understanding. However, we will find that the tools of forces and
deformations alone are insufficient to describe material deformation. This is
because two distinct things contribute to the mechanical behavior of a structure:
the properties of the material from which it is formed and its shape. Consider a
cantilevered beam such as a swimming pool diving board. The stiffness at the end
of the board can be increased by using a stiffer material or by making a thicker
board. Therefore, forces and deformations need to be scaled in some way to the
shape of structure to which they are applied. Now, our first order of business is to
define our scaled forces (stress) and our scaled displacement (strain).

F2 = PA 2 Mechanical stress is analogous to pressure

Figure 3.7 A container filled with a When it is initially encountered, the concept of stress can be quite baffling. It is
fluid has a piston applied to its not necessarily intuitive to imagine forces “flowing” and being distributed
upper surface of area A1. The piston throughout a structure. Conceptually, stress is the abstract idea of force being
force F is supported by the substance, distributed over an area. We already have an intuitive understanding of distrib-
meaning that the contents are placed
under pressure P = F/A1. A small region
uted forces when we consider pressure. Imagine a simple vessel filled with a
of the bottom wall of the container with fluid, such as a gas or liquid. A force F is then applied to a piston that pressurizes
area A2 < A1 will experience a smaller the contents (Figure 3.7).
force F2 = PA2 = F(A2/A1) acting on it.
We can determine the pressure of the contents as the force F divided by the area of
the piston A1, P = F/A1. Furthermore, if we imagine a small cutout or window
in the bottom of the vessel, a force would be exerted on the area by the pressurized
fluid. The distributed pressure across the face of the piston adds up to produce a
net force on the cutout. And we know that this force is proportional to the pressure
and to the area of the piston, F2 = PA2. When the pressure is larger, the force is
larger, and when the area is larger, the force is also larger. This is all very straight-
forward and intuitive.
What is critical to realize is that we can imagine a distributed force due to pressure
anywhere within the fluid. For example, we can imagine a plane within the fluid of
area A3. The force across this area is F3 = PA3 and is exerted by the fluid on itself,
regardless of the orientation of the plane. So, the idea of pressure is not limited
only to those locations where there are pistons and cutouts. The pressure is exerted
everywhere within and is defined as the force per unit area exerted on any small
imaginary cut-plane regardless of position or orientation, so long as the plane is
wholly immersed in the fluid.

Normal stress is perpendicular to the area of interest

Let us try to apply our analogy with pressure to a solid structure and see how far it
can take us. Imagine a simple cylindrical column supporting a load at its top. In
 MECHANICS OF DEFORMABLE BODIES 57

the interior of the column the force resulting from the load is evenly distributed F
through the cross section, assuming the column is uniform. If we move far away
from the end, the force becomes evenly distributed throughout the cross section.
As with pressure, we can imagine a cut-plane through the column. Consider a cut- σ = F/A
plane that is perpendicular to the long axis of the column (Figure 3.8). Because
the force across this area is distributed evenly, we can apply an analogy to pres-
sure and define stress, σ, to be the force per unit area

F
σ = . (3.1)
A
Figure 3.8 Column loaded with a
This stress is considered a normal stress, because the force is acting perpendicu- distributed force F at both ends
larly across the area. Historically, tensile stresses, as illustrated in Figure 3.8, are experiences internal normal stress
assumed to be positive by convention, and compressive stresses are negative. F/A at any cut-plane oriented
perpendicular to the long axis of
Unfortunately for the analogy with pressure, the opposite sign convention is used
the cylinder.
for pressure as stress.
Nota Bene
Strain represents the normalized change in length Tensile stresses are positive. The
of an object to load convention that tensile stresses are
positive comes about because much
Just as we defined stress to be a force scaled by the area it is acting across, we can of the material-testing literature is
also define strain as a change in the length of an object, scaled by its original focused on tensile testing.
length. Consider again the column under tension from Figure 3.8. Let the initial
length of the column be denoted as L. When the load is applied, the column Nota Bene
extends by a small amount (ΔL) and the total length of the column is now L + ΔL
(Figure 3.9). Here, the strain is defined as Distinguishing material and
structural behaviors. The column is
a good example of why scaling can
∆L be so critical. Imagine that the column
ε= . (3.2)
L was twice as long initially: the amount
by which it extends under the same
load would be twice as large.
A strain associated with normal stress (that is, along the same axis as the applied However, the stress in the material
force) is called normal strain. Because strain is defined as the ratio of two would not have changed. This larger
lengths, it is inherently dimensionless. It is sometimes expressed as either per- displacement at the tip would be due
cent strain (ΔL/L × 100) or microstrain (ΔL/L × 1,000,000). Also note that the entirely to the shape of the column, in
same sign convention for stress holds for strain (tensile positive, compressive this case its length. This is an example
of the influence of geometry on
negative).
mechanical behavior and is
sometimes called the structural
behavior to distinguish it from the
The stress–strain plot for a material reveals information influence of the material properties.
about its stiffness
We have now defined stress and strain in our simple uniaxial column example.
Before we go on to more complicated examples, let us see how they relate to each ΔL
other to help us understand the behavior of a material. Our column is loaded with
axial tension and we record the deformation of the tip as a function of the load we
apply. Assume that the column is uniform in its material properties (homogene-
ous), that it behaves as an elastic material, and that its material behavior is the ΔL
ε=—
L
same in all directions (isotropic). Although the two quantities we directly measure L
are the tip displacement and the force, we calculate the stress (σ = F/A) and strain
(ε = ΔL/L) and plot them versus each other (Figure 3.10).
For a broad range of sizes and shapes of columns, this plot will exhibit the same
overall shape. For many materials, this plot is a simple line. Such materials are
known as linear (or linear elastic) materials. The slope of this line is known as the Figure 3.9 Same column from Figure
3.8 with the same load (load not
Young’s modulus of the material. Typically this proportionality constant is given
shown) will lengthen along the long
by E in the linear relationship σ = Eε, known as Hooke’s law (physics students axis of the cylinder. The ratio of the
may recognize its similarity to the one-dimensional Hooke’s law F = kx from change in length (Δ L) to the original
modeling a spring). As the load increases, a stress is eventually reached, known length (L) of the column is the strain.
58 CHAPTER 3: Solid Mechanics Primer

Nota Bene (A) (B)

Who was Hooke? The relationship proportional limit
between stress and strain for linear
elastic materials, σ = Eε, is known as yield point ultimate stress
Hooke’s law after Robert Hooke, the
seventeenth-century physicist. failure
Actually, Hooke studied linear springs
in terms of force and displacement
(F = kX). Hooke never would have
used the equation that bears his
name. Nonetheless, his contribution σ
to understanding elastic behavior E (Young’s modulus)
was very significant. Coincidentally,
for the subject of cell mechanics, he
is also known as the father of
microscopy and coined the term cell.
ε

Nota Bene Figure 3.10 For the loading configuration (A), a stress–strain relationship can be
obtained (B). For small strains, the relationship can be a straight line; this is the linear region
Who was Young? Young’s modulus with the slope representing the stiffness, or elastic modulus. As the load increases, the
is named after the eighteenth- material will pass through the proportional limit (extent of linear region), yield point (where
century British physician Thomas deformation is not recovered upon unloading), ultimate stress (maximum stress), and failure.
Young. He conducted research in
light and vision, circulation,
mechanics, and even helped
translate some of the Rosetta Stone. as the proportional limit where it no longer behaves linearly. The stress at failure
He extended the work of Hooke, who is the ultimate stress or strength of the material. Independence from geometry is
characterized elastic behavior in a major benefit of using stress and strain rather than force and displacement.
terms of force and displacement
(F = kX) into normalized stress and
strain, whereby, the proportionality Stress and pressure are not the same thing, because stress
constant k—known as the stiffness has directionality
and a function of both material and
geometry—becomes the Young’s One key distinction between stress and pressure has to do with orientation. As we
modulus, E, and is a true material mentioned previously, the pressure at any point in a fluid does not depend on the
property. It is also sometimes orientation. For a given point inside a fluid, the vertical and horizontal pressures
referred to as the elastic modulus or
at that location are always the same. By contrast, stress depends strongly on orien-
modulus of elasticity. However, it is
only one of several possible elastic tation. In our axially loaded column example, the stress in the vertical direction is
moduli, such as the bulk modulus or F/A, yet the stress in the horizontal direction is zero, because there is no load in
shear modulus.

Example 3.1: Cytoskeletal proteins in extension

At this point in our discussion we can already create a sim- a structure to give us some intuition of the mechanical
ple model for cytoskeletal proteins. We can treat them as behavior of the cell’s components. This stiffness would be
simple rods and we can ask how stiff they are. We are influenced by both the material behavior and geometry of
interested not in the stiffness as a material property per se; the polymer. EA is the relevant measure of structural stiff-
rather, what is relevant is the axial stiffness of a polymer as ness (per unit length) in this case Table 3.1.

Table 3.1 Cytoskeletal axial stiffnesses

R A E EA

Microtubule 12.5 nm 492 nm2 1.9 × 109 N/m2 0.934 × 10−6 N

Intermediate filament 5.0 nm 79 nm2 2 × 109 N/m2 0.16 × 10−6 N

Actin filament 3.5 nm 38 nm2 1.9 × 109 N/m2 0.23 × 10−9 N

R, radius; A, cross-sectional area; E, Young’s modulus.

 MECHANICS OF DEFORMABLE BODIES 59

the horizontal direction. The reason behind this difference is the ability of the
molecules in a fluid to flow under load and reorient the internal force distribution.
Molecules in a solid, on the other hand, are more strongly bound to their neigh-
bors and are not free to redistribute under load. In mathematical terms, pressure
is a scalar quantity, depending (as do forces, displacements, stresses, and strains)
on the location within the material where the quantity is measured. Because
forces and displacements are vectors (these quantities are described by a magni-
tude and a direction), directionality is built into them. Stress and strain, which F
depend on the direction of forces and displacements, therefore exhibit direction-
ality-like force and displacement. However, stress and strain are even more com- τ = F/A
plicated, because they depend not only on the direction of forces and
displacements, but also on the direction of the area (or displacement) to which
they apply. To specify a component of stress, we must know the direction of the
forces relative to the orientation of the surface (the surface direction or orienta-
tion is given by the vector normal to its local area). Stress and strain are mathe-
matical quantities known as tensors.
Figure 3.11 Shear stress occurs
when the loading force is
Shear stress describes stress when forces and areas are perpendicular to the area across
perpendicular to each other which the stress acts.

Let us return to the column-loading example. Initially, the forces in the cut-plane
were perpendicular to the plane, resulting in normal stresses. What sort of
stress results if we load the column perpendicular to the long axis (that is, parallel
to the plane)? Consider again the imaginary cut-plane some distance away from
the region of load application. As before, the force is distributed evenly through
the cross section, but now the small distributed forces are lined up parallel to the
plane rather than perpendicular to it (Figure 3.11). We term this type of stress
shear stress,

F
τ = . (3.3)
A

In general, the cut-plane through the column can be in any orientation we desire.
If we select a plane that is oriented obliquely to the axis of the column, the distrib- Figure 3.12 A general cut-plane
uted forces across the surface will be a mixture of parallel and perpendicularly through a column with a normal
oriented forces. Therefore, in general stress is a mixture of normal stress and shear load will experience both shear and
normal stresses (since the vector
stress (Figure 3.12).
components can be decomposed
into components parallel and
perpendicular to the surface). This is
Shear strain measures deformation resulting from true even if the load is a mixture of
shear stress shearing and normal forces.

In addition to shear stress, we need to define the complementary deformation

measure, shear strain. In the case of normal strain, we divided the vertical dis-
placement at the tip (representing the length change) by the column length. We δ
can similarly define the shear strain to be the transverse displacement divided by
the original vertical length, or γ =  δ/L. Because this is a result of shear stress, the
direction of displacement is perpendicular to the direction of the original length
(Figure 3.13). Like normal strain, shear strain is a dimensionless quantity and is
typically expressed in percent strain or microstrain.
L θ
To illustrate shear strain, consider a rectangular body. Assume the block is fixed at
the bottom and exposed to a shear force at the top. Finally, assume that points along
the top surface of the block displace by an amount δ as a result of the application of
the shear force, sometimes referred to as “pure shear.” The shear strain is the ratio of Figure 3.13 Shear strain is
δ to L or, γ = δ/L. The shear strain can be related to the amount by which the right measured by dividing the lateral
angle at the base of the block is reduced, which we denote θ, with δ/L = tan θ. If displacement (δ) by the original
δ  L, tan θ ≈ θ when θ is measured in radians. As with Hooke’s law, we can then normal length (L).
60 CHAPTER 3: Solid Mechanics Primer

define the linear relationship between shear stress and shear strain with the shear
modulus, G,
τ = Gγ = Gθ (3.4)

for small θ.

Torsion in the thin-walled cylinder can be modeled

with shear stress relations
Let us work through an example using shear stresses and strains. We are given a
thin-walled cylinder that is clamped at one end and a moment is applied at the
other end (Figure 3.14).
Let M be the torsional moment applied to the end of the shell of radius R. In
response to this torque, the shell will twist along its length. The overall deformation
is related to the internal strain, and the relationship between the displacements
and strain is known as the kinematics of the problem. We can measure the amount
of this twist by the angle of twist at the end, θ. For the same strain in the material, a
longer shell will exhibit a larger angle of twist compared with that of a shorter shell.
So we relate the twist angle to the angle by which a line drawn on the surface of the
shell parallel to the long axis of the shell rotates (which is simply the shear strain,
γ). We start by considering how much a point on the surface displaces, denoted δ,

δ = θR, (3.5)

as a function of the displacement along the circumference of the shell. In terms of

shear strain,
δ θR
γ= = , (3.6)
L L
where L is the length of the rod.
Next, we relate the applied moment to the angle of twist. Similar to the axial load-
ing case, the resultant load (or moment) is the net resultant of the stress acting on
the end of the shell. The moment M =  τAR, where R is the radius of the shell (we
ignore the thickness as being small compared with the radius R) and τ is the shear
stress on the end surface of the shell which has area A. The area A = 2πRt, where t
is the rod thickness (again, t  R). So,

M = τ 2πR 2t , (3.7)

which can be rearranged to obtain the shear stress

M
τ = . (3.8)
2πR 2t
Then, by using our material model, τ = Gγ from Equations 3.4 and 3.6
γL ML
θ= = . (3.9)
R 2GπR 3t

Figure 3.14 Torsion of a thin-walled t

cylindrical shell results in a twisting
of the shell. We can model this by
cutting open the shell and treating this γ δ θ M
object as a plane rectangle.

L
 MECHANICS OF DEFORMABLE BODIES 61

Torsion of a solid cylinder can be modeled as a torsion of a

series of shells of increasing radius
Now consider a solid cylindrical rod. We need to think about how the mechanics
changes at different points throughout the thickness rod. Indeed, we can solve the
solid rod problem by dividing it up into several thin rods and superimposing those
solutions (Figure 3.15).
Starting with the kinematics, let us determine how the strain changes through the
thickness. As we determined for the thin-walled shell, at a given angle of twist,
shear strain (and shear stress) varies linearly with radius R (Equations 3.5 and
3.6), where we now replace the constant shell radius R with the variable radius R,
and turn the shear into functions of R:
θ
γ (R) = R. (3.10)
L
For a small element of thickness, t = dR  R, the shear stress in that thin-walled
tube is
GθR
τ ( R) = Gγ = . (3.11)
L
Multiplying by the cross-sectional area of the shell (along which the applied
torque) gives the contribution to the net moment of the shell,
2GθπR 3dR
dM = τ 2πR 2dR = . (3.12)
L
To find the total moment for the entire rod, we integrate the contribution for each
shell,
R R
2GθπR 3 GθπR 4
∫
M = dM =
∫ L
dR =
2L
. (3.13)
0 0
Equation 3.13 can also be expressed in terms of the polar moment of inertia, J,
R

J =
∫ R dA = ∫ 2πR dR,
2 3
(3.14)
A 0
which is a measure of a given cross-sectional geometry to resist torsion. This
allows us to express Equation 3.13 in a more compact form,
GJθ
M= . (3.15)
L

Notice that this analysis only works for cylindrical-shaped rods. If a rod is not a cyl-
inder, if it has a square or rectangular cross section, our kinematic assumption
breaks down. Plane sections will not deform by simply sliding over one another, but
they will also become deformed (no longer remain planar). This deformation along
the axis of the rod is known as warping and occurs even for small deformations.

Figure 3.15 A solid rod can be

modeled as an integral over thin
+ + shells of varying radii. Each shell has
a radius of r and a thickness of dr.

+ ...
62 CHAPTER 3: Solid Mechanics Primer

Example 3.2: Cytoskeletal proteins in torsion

We wish to determine the rough structural stiffness of an order of magnitude by simply taking E for G. We also
typical cytoskeletal proteins in torsion. Although it is make the assumption that the polymers are solid cylin-
not possible to compare axial structural stiffness with ders. This is not the case for microtubules but it is still a
torsional stiffness directly, it is still helpful to make a reasonable order-of-magnitude approximation. We leave
general order-of-magnitude comparison to develop it to you to determine the size of this error. From this
our intuition at this level. The parameter that captures simple calculation, we can expect that the forces due
both the material behavior and the structural behavior to  axial deformation are much more important than
is GJ (Table 3.2). The effective shear modulus is not those due to torsion because of their very low relative
readily available experimentally, but we can get within magnitude.

Table 3.2 Cytoskeletal torsional stiffnesses

R J G GJ

Microtubule 12.5 nm 38,400 nm4 1.9 × 109 N/m2 73 × 10−24 Nm2

Intermediate filament 5.0 nm 980 nm4 2 × 109 N/m2 1.8 × 10−24 Nm2

Actin filament 3.5 nm 230 nm4 1.9 × 109 N/m2 0.45 × 10−24 Nm2

Kinematics, equilibrium, and constitutive equations are the

foundation of solid mechanics
At this stage in our review of mechanics, let us work through a more challenging
problem. Beam bending is illustrative, because we can solve the problem with
only the background and definitions we have developed thus far. Yet it also
requires the three fundamental relationships of solid mechanics: kinematics,
equilibrium, and constitutive behavior. Although we have not stated it explicitly,
we have been using these equations in the simple examples so far. Indeed, all
problems in solid mechanics require these three components, and, with the addi-
tion of boundary conditions, each must be included to find a solution. It is helpful
to consider each of these in turn in this simple situation before we develop them
Nota Bene in the general continuum mechanics formulation.
Cytoskeletal polymers often
experience small strains. Beams
Kinematics in a beam are the strain–displacement relationship
following these simplifications
conform to Euler–Bernoulli beam
In every mechanics problem there needs to be a relationship between displace-
theory. It is particularly useful for
slender structures in which shear ment and strain. In our simple axial and torsion examples, this was straightfor-
deformation is quite small compared ward. In the case of a beam, it is a bit more complex. Consider a beam that is
with displacements due to flexion. loaded by a bending moment such that it takes on a curved shape (Figure 3.16).
The long slender structural polymers
of the cells satisfy these assumptions We make three simplifying assumptions about the deformation that character-
quite well. izes the kinematics of the problem. First, we assume that any y–z-plane in the
undeformed beam will remain a plane in the deformed situation. In other words,

Figure 3.16 Schematic of a beam y

exposed to bending moments.

x
z
 MECHANICS OF DEFORMABLE BODIES 63

y–z-planes rotate, but they do not deform to lose their planar shape. This assump-
tion is sometimes called the “plane sections remain plane” assumption. Next, we
assume that the plane sections only rotate and do not slide relative to each other. S
In other words, we are going to assume that we can ignore the effects of shear.
Finally, we assume that as the beam deforms, the y–z-planes rotate in such a way
as to remain perpendicular to any imaginary line in the beam that was originally
oriented in the x-direction. These three assumptions serve to define fully the
deformation of every point in the beam. Furthermore, the deformed state is fully R
defined by knowing the displacement of the central axis of the beam only. The
beam becomes a simple one-dimensional problem of determining the displace-
ment in the y-direction, which we denote w(x). Next we need to determine the S – ∆S
strain in the beam as a function of the displacement w(x). ∆R
S
To accomplish this, we need to determine the amount of stretch or strain that lines
initially oriented parallel to the x-axis experience as the beam deforms. Consider a S + ∆S
very small portion of the deformed beam (Figure 3.17). Further, consider three
lines on the beam; the top surface, the bottom surface, and the central axis of the Figure 3.17 Small section of the
beam S long, ± ΔR thick.
beam. If the section is small enough, the deformed shape of each of these lines can
be approximated as sections of a circle (this is known as the osculating circle and
can be uniquely defined at every point in a curve as the best-fitting circle at that
point). Let the radius of the deformed central axis be R, the top surface R − ΔR, and
the bottom surface R + ΔR. Next, assume that the initial length of a line along the
central axis in our section of beam is S. In the deformed configuration, the central
axis is neither lengthened nor shortened. In other words, the strain at the central
axis is zero and is therefore often referred to as the neutral axis. Because the ratio of
circumference to radius is constant for similar circular sections,

S + ∆S S
= . (3.16)
R + ∆R R

Equation 3.16 can be rearranged to show

∆S ∆R
= . (3.17)
S R

Notice that the quantity ΔS/S is simply the change in length of one of our imaginary
lines normalized by its initial length. That is exactly our definition of strain. Also
notice that ΔR is the distance to the imaginary line in the y-direction measured
from the neutral axis. If we position the origin of our coordinate system so that it
lies on the neutral axis, we have a simple expression for the strain in the beam,
y
ε( y ) = . (3.18)
R

So we have used the term “central axis” without really defining it. This was an
intentional oversight to simplify the explanation. It would have been more appro-
priate to refer to the neutral axis. For cross sections that are symmetric about the
central axis, the two coincide, but this need not be the case in general.
Now, notice that the radius, R, is not a constant; it is a function of the displace-
ment w(x). To formulate equations that we can solve, we must make this depend-
ency explicit. To do this, notice that the quantity 1/R is also the local curvature of
the beam typically denoted as κ. In calculus, we learned that the local slope of a
curve is given by its first derivative, but the local curvature is given by the second
derivative. Therefore strain is the product of the second derivative of the displace-
ment, and the distance from the neutral axis becomes

y d 2w
ε (y) = = yκ = y 2 . (3.19)
R dx
64 CHAPTER 3: Solid Mechanics Primer

Equilibrium in a beam is the stress–moment relationship

The next step in deriving the beam-bending equations is to make use of equilib-
rium to relate the moment at a section within the beam to the stress at each point
within the beam. Let us think intuitively about the distributed forces or stresses
within our beam (Figure 3.18). At the inner side, the beam will be compressed to
a shorter total length, whereas at the outer side, the beam will be stretched. The
stress in a beam varies linearly through the cross section. At the inner side, the
stress is negative or compressive while at the outer side, it is positive or tensile. At
the neutral axis, the stress is zero. The stress resultant of this linearly varying stress
is the bending moment M.
Consider a plane that cuts through an arbitrary point in the beam at a location x.
For a small strip across the face in the y–z-plane of thickness dy and width h, the
contribution to the total moment by the stress exerted on the strip is simply the
total force times the moment arm. The moment arm can be calculated from any
arbitrary point; however, in this example, the most logical place is the origin of
the coordinate system that we located at the neutral axis. Also, note that we must
define the sign of the moment, which is arbitrary. You can define a positive
moment to be one that causes the beam to bend with its inner surface facing up
(as we have done here). It is equally acceptable to define it in the other sense, but
just make sure you keep track of which way you have defined it, or you will end
up with a minus sign that will not go away. It is a good idea to make a note of the
convention you are using on any problem you work out. In our case remember
that the stress is positive (tensile) on the lower surface and negative (compres-
sive) on the upper surface. We now know the moment associated with our strip is
dM = −yσ(y)hdy. The total moment is obtained by integrating the contributions
of all the strips that make up the beam or

t /2 t /2

M=
∫ dM = ∫ yσ ( y)hdy.
− t /2 − t /2
(3.20)

Figure 3.18 The resultant moment M+ convention

on the end of the beam is the
integral of the stresses through the load–moment relation
Consider a small
thickness of the beam.
y y strip dy thick.
3dy
x z t

h
 MECHANICS OF DEFORMABLE BODIES 65

The constitutive equation is the stress–strain relationship

The last part of our development is the constitutive equation. This is the equation
that relates stress to strain. It is the fundamental description of how a material
behaves. For this example, assume a simple linear elastic material behavior.
Hooke’s law gives us

σ ( y ) = Eε ( y ). (3.21)

Remarkably, with the kinematics, equilibrium, and constitutive equations, we
have completed the theoretical development for this problem. All that remains is
to combine what we know. Substituting the stress–strain relationship (Equation
3.21) into the equilibrium equation (Equation 3.20) yields

t /2

M=E
∫ yε( y)hdy.
− t /2
(3.22)

Next, using the strain–displacement (Equation 3.19), we have

t /2
d 2w
M=E
∫
− t /2
y2
dx 2
hdy . (3.23)

Because w does not depend on y, it can come out of the integral. The end result is
a governing relationship between the moment and the curvature,

t /2
d 2w d 2w
M=E
∫
− t /2
y 2hdy
dx 2
= EI
dx 2
, (3.24)

where
t /2

I=
∫ y hdy.
− t /2
2 (3.25)

The second moment of inertia is a measure of

bending resistance
Note that the integral term I in Equation 3.25 is a constant function of the cross-
sectional geometry only. It is known as the second moment of area, the second
moment of inertia, or area moment of inertia. It is a measure of a cross section’s
geometric contribution to a beam’s resistance to bending. Also notice that unlike
for the polar moment of inertia, J, for the second moment of inertia we must spec-
ify the direction in which the moment is taken. For this reason a subscript is some-
times added,

t /2 t /2

Iy =
∫ y hdy
− t /2
2
and I x =
∫ x hdx .
− t /2
2 (3.26)

The term that relates curvature to moment, EI, is a measure of the combined
effects of material stiffness and geometry to resist bending. To reflect this, it is
known as the flexural rigidity.
66 CHAPTER 3: Solid Mechanics Primer

Example 3.3: Cytoskeletal proteins in bending

Simple bending occurs frequently in the world of the cell. polymers are relatively flexible in bending and torsion
One of the key elements that gives the cell its structure is and relatively stiff axially. In a way, they are similar to
the cytoskeleton. There are three components that make strings or ropes, hard to stretch, but easy to bend and
up the cytoskeleton: microtubules, intermediate filaments, twist. Also notice that of the three, microtubules are the
and the actin cytoskeleton. We can estimate the flexural most able to resist bending (and torsion). Actin is the most
rigidity of a polymer of each of these components from its string-like, and intermediate filaments are in between. As
Young’s modulus and radius (Table 3.3). Note that, again, seen in future analyses, very flexible polymers, like actin,
we have neglected to account for the hollow shape of the can have their axial behavior affected by their flexibility
microtubule. due to the thermal behavior of the molecules surrounding
them.
As with torsion, notice the very low relative magnitude of
bending forces compared with axial forces. Cytoskeletal

Table 3.3 Cytoskeletal bending stiffnesses

R I E EI

Microtubule 12.5 nm 19,175 nm4 1.9 × 109 N/m2 364 × 10−25 Nm2

Intermediate filament 5.0 nm 491 nm4 2 × 109 N/m2 10 × 10−25 Nm2

Actin filament 3.5 nm 118 nm4 1.9 × 109 N/m2 2 × 10−25 Nm2

The cantilevered beam can be solved from the general

beam equations
Now that we have all the tools in hand, let us solve an example beam-bending
problem. One classic problem is the cantilevered beam. In this problem, the
beam is clamped at one end and has a single force applied to the other end
(Figure 3.19).
As we have derived, the governing equation for the transverse displacement of the
beam is

d 2w M ( x )
= . (3.27)
dx 2 EI
Notice that M is a function of x and E and I are constants. Given a beam of length
L, the moment produced by the force on the end is simply the force multiplied by
the moment arm, M(x) = (x – L)F. That takes care of the right side, but what about
the boundary conditions on the left side? Clearly, the displacement must be zero,
w(0) = 0. But it is also clamped, so the slope must also be zero, dw(0)/dx = 0. We
need to solve

d 2w ( x − L)F dw(0)
= w(0)= = 0. (3.28)
dx 2 EI dx

Figure 3.19 The cantilevered beam F

is clamped at one end and a
vertical force on the other.
+M

M = (X – L)F
 MECHANICS OF DEFORMABLE BODIES 67

Integrating (Equation 3.28) twice yields

F  x3 x2 
w= − L + C1 x + C 2  . (3.29)
EI  6 2 

Using our boundary conditions:

w(0) = 0 ⇒ C 2 = 0

dw(0) (3.30)
= 0 ⇒ C1 = 0.
dx
Therefore, the final solution is

F  x3 x2 
w= − L . (3.31)
EI  6 2 

Buckling loads can be determined from the beam equations

Another important problem in cytoskeletal mechanics that we can solve with
the beam equations is that of buckling. Consider a beam that is loaded axially
(Figure 3.20).
This is different from the axially loaded rod, because we are considering the
deformation of the beam in the y-direction (not the x-direction) as a function of
the axial loading at the tip. Consider a small element of the deformed beam
(Figure 3.21).
Notice from this free-body diagram that we have replaced the portion of the
beam to the right. To compensate for what we have removed, we need to add an
equivalent force and moment, Fs and Ms. Equilibrium requires that the force on
the section, Fs, is simply equal to the force on the tip F. Because this force is not
applied along the same line as F, it produces a moment, Fsy, that must be coun-
tered in order for the free body to be in equilibrium. This moment acts across the
cut-plane and must be Ms = −Fy. Our beam equation is now

d 2w M ( x ) Fw
= =− . (3.32)
dx 2 EI EI
This is a little more challenging to solve than Equation 3.28. We need a function
that returns itself when differentiated twice. This suggests

w( x ) = C1 sin(kx ) + C 2 cos(kx ). (3.33)

To satisfy the boundary condition, y(0) = 0, C2 = 0, and

d 2w Fw
=− = −C1k 2 sin(kx ). (3.34)
dx 2 EI

L Figure 3.20 Beam exposed to an

+M axial load.

w(0) = w(L) = 0
68 CHAPTER 3: Solid Mechanics Primer

y Substituting into Equation 3.32 implies that

Fs Ms
x  F
C1  k 2 −  sin(kx ) = 0.
w
 EI  (3.35)
Figure 3.21 Free-body diagram for a One solution to Equation 3.35 is k = 0 or y(x) = 0. This trivial solution corresponds
small section of the beam shown in to the beam remaining in a straight configuration. We will come back to this pos-
Figure 3.20. sibility in a moment. Alternatively, if we assume that k is not zero, we can see that k
must be

F
k=± .
EI (3.36)

To satisfy our second boundary condition, y(L) = 0,

 F  F
C1 sin  L  = 0 or L = nπ . (3.37)
 EI  EI

Each of the solutions corresponding to different values of the integer, n, is a differ-
n=1 ent mode of bending (Figure 3.22).
Notice that none of the solutions to Equation 3.36 place a constraint on C1. The
n=2 etc. bending mode that corresponds to the lowest possible force is the n = 1 mode. The
force that corresponds to this mode is
Figure 3.22 Deformation patterns
corresponding to the first two π 2 EI
bending modes of the beam. Fb = , (3.38)
L2

which is known as the Euler buckling load. If the applied force is < Fb, the beam
remains straight and y(x) = 0 is the solution. It can also be shown that this solution
is stable. In other words, if the beam is deflected slightly away from y(x) = 0, it will
return to the y(x) = 0 configuration. On the other hand, if F > Fb, the y(x) = 0 solu-
tion is not stable, and the beam will buckle, taking on a deformed configuration.
This simple analysis to determine the buckling load can be used to find the EI of a
biopolymer by measuring the longest length it is able to attain.

Transverse strains occur with axial loading

When we conducted our analysis of axial deformation, we neglected one important
thing. When we stretch a rod axially, not all the deformation is in the axial direction.
There is also a reduction in the width of the rod in the transverse direction. This effect
is quantified by Poisson’s ratio (Siméon Poisson, 1781–1840), the ratio of the axial
and transverse strains, and is typically denoted by ν which is defined as ν =  εt/εa.

The general continuum equations can be developed

from our simple examples
Now that we have worked a few problems and developed a bit of intuition for
mechanics, we are ready to state the general form of these equations. Of course,
we are making several simplifying assumptions. Specifically, the type of con-
tinuum mechanics we are going to apply is useful for linear elastic materials
undergoing small deformations. This is a bit limiting, because the mechanical
behavior of the cell is far from a linear–elastic–infinitesimal structure. Also,
what is being presented here is not meant to be a complete treatment of this
topic. Indeed, elastic small-deformation theory is typically taught as a year-
long subject in mechanical engineering departments. The linear elastic infini-
tesimal theory of plates and shells is typically another entire year-long subject.
 MECHANICS OF DEFORMABLE BODIES 69

We aim to provide you some familiarity with these tools, but the presentation is
far from complete.
As we described before, there are three parts to a problem in continuum mechan- Sx y Sx
ics: kinematics, equilibrium, and constitutive behavior. So, let us take each one
in turn.

Sx x
Equilibrium implies conditions on stress
S xz
In our discussion of stress and equilibrium, let us first introduce a notation system
for stress. In our examples, we defined stress conceptually to be the distributed y
force per unit area through a given test plane. In the example, the location of the
test plane was obvious. Now we want to create a general definition and notation x
system associated with our x-, y-, z-coordinate system. To specify the orientation
of the test plane, we will make use of the normal vector to the plane. z

Next consider the resultant force denoted by Sx (Figure 3.23). The x refers to Figure 3.23 Forces on an imaginary
the cut-plane we selected, which has a surface normal aligned in the x-direc- cut-plane in the y–z-plane.
tion. Like any vector, Sx can be broken into its three components, Sx x , Sx y , Sx z .
So, for this face, we have three potential stresses, one associated with each of
the components of Sx. To express the components of stress, we need a double
subscript notation system with the first referring to the direction of the cut-
plane normal, and the second referring to the direction of the internal force
acting through that plane. Now we have everything we need to define the stress
components in this coordinate system. For a cut-plane oriented perpendicular
to the x-direction,

Sx x (3.39)
σ xx = lim .
A→0 A

Similarly, for the other components for the x-cut-plane

Sx y Sx (3.40)
τ xy = lim and τ xz = lim z .
A→0 A A→0 A

And for the y-cut-plane

Sy x Sy Sy (3.41)
τ yx = lim , σ yy = lim y and τ yz = lim z
A→0 A A → 0 A A → 0 A

and for the z-cut-plane

Sz x Sz Sz (3.42)
τ zx = lim , τ zy = lim y and σ zz = lim z .
A→0 A A→0 A A→0 A

Now that we have a consistent notation system, let us see what equilibrium can
tell us about stress. Remember that the equilibrium condition is a condition on
forces. Specifically, for a nonaccelerating body, the forces must sum to zero.
Imagine a small piece of material within a general solid body, and consider the
forces on the infinitesimal element as shown (Figure 3.24). We assume that the
dimensions of the element are dx, dy, and dz respectively. Each face has one nor-
mal force and two shear forces. However, because the element is vanishingly
small, we can approximate the force on the faces far from the origin as a function
of the force on the closer face and its derivative. Taking the first term in a Taylor
series,
dSx x ( x )
Sx x ( x + dx ) = Sx x ( x ) + dx . (3.43)
dx
70 CHAPTER 3: Solid Mechanics Primer

Figure 3.24 All the forces on a small dSzz

element of volume oriented with Szz + — dz
the coordinate axes for dSzx dz
Szx + — dz
convenience. dz
dSzy
Szy + — dz
dz
dSyy
Sy y + — dy dSyx
dy Sy x + — dy
dSyz dy
Sy z + — dy
dy
dSxy
Sx y + — dx
dx
Sx z Szy
dSx x
Sx x + — dx
Sxx dx
Szx dSxz
Sxz + — dx
dx
Sx y Szz

y S yz
S yx

x
S yy
z

Now we simply sum the forces in the x-, y-, and z-directions in turn:

dSx x   dSy x  
∑f x

= 0 ⇒  − Sx x + Sx x +
 dx
dx  +  −Sy x + Sy x +
  dy
dy  +  −Sz x + Sz x +
 
dSz x 
dz
dz 


dSx y   dSy y  
∑f y

= 0 ⇒  − Sx y + Sx y +
 dx
dx  +  −Sy y + Sy y +
  dy
dy  +  −Sz y + Sz y +
 
dSz y 
dz
dz 


dSx z   dSy z  
∑f z

= 0 ⇒  − Sx z + Sx z +
 dx
dx  +  −Sy z + Sy z +
  dy
dy  +  −Sz z + Sz z +
 
dSz z 
dz
dz  .

(3.44)

Notice that the first two terms in each quantity in parentheses cancel. Next, we
can divide each row by the infinitesimal volume (dx dy dz) and simplify

 dSx x   dSy x   dSz x 

     
dx   dy   dz 
+ + =0
dydz dxdz dxdy

 dSx y   dSy y   dSz y 

     
dx   dy   dz  (3.45)
+ + =0
dydz dxdz dxdy

 dSx z   dSy z   dSz z 

     
dx   dy   dz 
+ + = 0.
dydz dxdz dxdy

Now, notice that in each term we have a differential area. In each case, this differen-
tial area does not depend on the derivative in the numerator. Therefore, we can write
 MECHANICS OF DEFORMABLE BODIES 71

d  Sx x  d  Sy x  d  Sz x 
+  + =0
dx  dydz  dy  dxdz  dz  dydz 

d  S x y  d  Sy y  d  Sz y 
+  + =0 (3.46)
dx  dydz  dy  dxdz  dz  dydz 

d  Sx z  d  Sy z  d  Sz z 
+  + = 0.
dx  dydz  dy  dxdz  dz  dydz 

The differential area for each term in parentheses is the area normal to the respec-
tive force vector component. Therefore, each of these terms is simply our defini-
tion of stress. The equilibrium equations then take the following remarkably
simple form:
dσ xx dσ yx dσ zx
+ + =0
dx dy dz
dσ xy dσ yy dσ zy
+ + =0 (3.47)
dx dy dz

dσ xz dσ yz dσ zz
+ + = 0.
dx dy dz

Example 3.4: Symmetry of stress

Our free-body analysis of the resultant forces on an there is no additional information in the σyx, σ zx, and
infinitesimal element can tell us one other important σzy terms they are typically replaced by σxy, σxz, and
fact about stress. Notice that the equilibrium equation is σyz, respectively.
the result of requiring the forces to sum to zero. What
about the moments? Remember that in a body that is not
accelerating, the moments about any arbitrary axis must
also sum to zero. In our example, calculate the moments Szz
about an axis passing through the center of the element
in the x-direction (Figure 3.25).
Summing moments implies Szx

∑M x = 0 ⇒ 2 Sy z + 2 Sz y = 0 Sx z

Sx x Sx x
or Sy z = Sz y . In terms of stress
Sx z
∫∫ σ zy dxdy =
∫∫ σ yz dxdy or σ zy = σ yz .
Szx
Likewise for the y- and z-axes, we obtain z

σ xy = σ yx and σ zx = σ xz . Szz
x
This important property of stress is that it must be Figure 3.25 The forces on a small two-dimensional
symmetric for any nonaccelerating body. Because element.

Kinematics relate strain to displacement

What is strain? The equations that relate strain and displacement are the kine-
matic equations and serve as the formal definition of strain. Unlike in our simple
examples above, these are general equations that characterize the deformation of
72 CHAPTER 3: Solid Mechanics Primer

a physical body. We begin our discussion by assuming we have specified a deform-

able body in a coordinate system specified x, y, and z. There is a deformation at
each point in the body given as u, v, w, the displacements in the x-, y-, and z-direc-
tions respectively.
normal strain: Let us start by defining normal strains. Imagine a general body undergoing a small
y deformation. We can define a test line within the body in the undeformed condi-
tion and ask what happens to it during the deformation (Figure 3.26). We are
going to consider every possible deformation and orientation of the test line in
turn, but, for now, we assume that the test line is oriented along the x-direction.
A B The ends of the test line are denoted by A and B, and are able to displace
u independently.
∆x
x Now we need to determine how much the test line is elongated. In general, this
will be a function of u, v, and w. However, if the deformations are “small” the
Figure 3.26 An imaginary line in the extension of the test line is dominated by u. The strain quantity we are defining
x-direction embedded in the here is the so-called infinitesimal or small deformation strain. Technically, it is
material being analyzed.
known as the Cauchy strain. The extension of the line is given by uA – uB and the
average strain in the test line is simply the change in length over the original length
(uA – uB)/(xA – xB) = Δu/Δx. The strain at any given point can now be defined as
the average strain in the test line as the length of the line shrinks to zero:
∆u
ε = Lim , (3.48)
∆x →0 ∆x
which is the definition of the derivative. To specify that we are referring to the dis-
placement in the x-direction of a test line originally oriented along the x-axis,
strain components are typically denoted with two subscripts,
du
ε xx = . (3.49)
dx
Similarly in the y- and z-directions,

dv
ε yy = ,
dy
(3.50)
dw
ε zz = .
dz
Now we consider the shear strain. Perhaps the most logical thing to do would be
to define them as similar to normal strains. We could define
dv
ε xy = . (3.51)
dx
However, there is a problem with this approach. Strain defined in this way is not
symmetric, because in general,
dv du
≠ . (3.52)
dx dy

It will simplify things greatly to define strain as symmetric, and this can be easily
achieved by taking our definition of shear strain to be

1  dv du 
ε xy = + , (3.53)
2  dx dy 

which preserves the symmetry condition, because εxy =  εyx, εxz =  εzx, and εyz =  εzy.
To clearly distinguish normal and shear strains, the symbol γ is sometimes used
for the components of the shear strain. The notation γ refers to the engineering
 MECHANICS OF DEFORMABLE BODIES 73

strain defined in our pure shear and torsion examples. It differs from the contin-
uum strain ε by a factor of two, that is, γ xy = 2ε xy , γ xz = 2ε xz , and γ yz = 2ε yz .

The constitutive equation or stress–strain relation

characterizes the material behavior
How are strain and stress related? As we have noted, the equations that relate
stress and strain are the constitutive equations. These are the equations that cap-
ture the behavior of the material. They will change depending on the material
being considered. Earlier in this chapter, we introduced Hooke’s law in the one-
dimensional case, σ = Eε. Let us see if we can generalize Hooke’s law to describe
the material behavior of a three-dimensional, isotropic, linearly xy elastic solid. In
fact, we have already described three parts of Hooke’s law in simple example
cases. We described the uniaxial behavior σ = Eε, the transverse contraction due
to Poisson’s ratio εp = −νεa, and the shear behavior τ = Gγ. Now, let us see if we
can figure out the general case. If we are given a set of six stresses (three normal
stresses and three shear stresses) can we figure out the set of six strains? In other
words, what are the 36 (6  ×  6) coefficients that must multiply the stresses to get
the strains. Note, it is easier to determine the coefficients that multiply the stresses
to get the strains, because the situations in which the material’s properties were
defined were given with several components of stress being zero. Because we are
describing a linear material, we can simply apply each of these behaviors in our
new notation system and add them up. We know from our example and definition
of Young’s modulus, when we hold all other stresses to be zero and apply a stress
in a normal direction, the coefficient multiplying the strain in the same direction
is 1/E. This gives us three of our coefficients.
Now, what can Poisson’s ratio tell us? If the stress is applied in the x-direction, it
tells us that the normal strain in the y- and z-directions are –ν times the strain in
the x-direction. However, we already know that strain is 1/E times the applied
stress. This gives us six more coefficients. For shear strain, we know that for an
applied pure shear stress, the normal strains and the shear strains in the other
directions are all zero. This gives us 24 more coefficients that must be zero. Finally,
we know from the relationship between shear strain and shear stress (1/G) the
final three coefficients we need. We can write the general form of the equations:

1
ε xx = σ xx − νσ yy − νσ zz  , γ xy = (1/G )τ xy
E
1
ε yy =  −νσ xx + σ yy − νσ zz  , γ xz = (1/G )τ xz (3.54)
E
1
ε zz =  −νσ xx − νσ yy + σ zz  , γ yz = (1/G )τ yz .
E
Notice that there are three material constants (E, ν, G) in Equation 3.54, though
only two of these are independent constants. The shear modulus can be expressed
in terms of Young’s modulus and Poisson’s ratio (the proof of this is left to you),
G = E/2(1 +  ν).
Therefore,

1 2(1 + ν )
ε xx = σ xx − νσ yy − νσ zz  , τ xy = γ xy
E E
1 2(1 + ν )
ε yy =  −νσ xx + σ yy − νσ zz  , τ xz = γ xz (3.55)
E E
1 2(1 + ν )
ε zz =  −νσ xx − νσ yy + σ zz  , τ yz = γ yz .
E E
74 CHAPTER 3: Solid Mechanics Primer

These equations can be inverted to yield the more traditional form of stress as a
function of strain:

E E
σ xx = (1 − ν )ε xx + νε yy + νε zz  τ xy = γ xy
( )(
1 + ν 1 − 2ν ) 2(1 + ν)

E E
σ yy = νε xx + (1 − ν )ε yy + νε zz  τ xz = γ xz (3.56)
(1 + ν )(1 − 2ν )  2(1 + ν )

E E
σ zz = νε xx + νε yy + (1 − ν )ε zz  τ yz = γ yz .

(1 + ν )(1 − 2ν ) 2(1 + ν )

Vector notation is a compact way to express equations

in continuum mechanics
Writing out in detail and manipulating the continuum equations of solid mechan-
ics can become quite cumbersome, therefore many compact forms of notation
have been introduced. One very powerful and extensively used system is known
as Voigt notation or vector notation. In this approach, the components of stress
and strain are organized into vectors,

σ xx   ε xx 
   
σ yy  ε yy 
σ zz   ε zz 
σ =   and ε =   . (3.57)
 τ xy  γ xy 
 τ xz  γ xz 
   
 τ yz  γ yz 

Using this notation the stress–strain relationship Equation (3.55) is simply

σ xx 
 
σ yy 
σ 
 zz  E
 =
 τ xy (1 + ν ) (1 − 2ν )
 τ xz 
 
 τ yz 

 
(1 − ν ) ν ν 0 0 0 
 ν (1 − ν ) ν 0 0 0   ε xx 
  
 ν ν (1 − ν ) 0 0 0  ε yy 
   ε zz 
 0 0 0
(1 − 2ν ) 0
 
0    or σ = Cε
 2  γ xy 
  γ 
 0 0 0 0
(1 − 2ν ) 0  
xz

 2  γ yz
   
 0 0 0 0 0
(1 − 2ν ) 
 2 
(3.58)
 MECHANICS OF DEFORMABLE BODIES 75

and

 
 1 −ν −ν
0 0 0 
E E E 
   −ν 1 −ν  
 ε xx   0 0 0  σ xx 
ε yy   E E E  σ 
   −ν −ν 1   yy 
 ε zz   0 0 0  σ zz  or ε = Dσ
 = E E E  
γ xy   2(1 + ν )   τ xy 
γ   0 0 0 0 0  τ 
E
 xz    
xz

γ yz   2(1 + ν )  τ yz 
0 0 0 0 0  
 E 
 
 0 2(1 + ν ) 
0 0 0 0
 E 
(3.59)

Nota Bene: Lamé constants allow a compact form of Hooke’s law

Equation 3.58 can be expressed in a more compact form.

    
σ xx   λ + 2µ λ λ 0 0 0  ε xx 
σ yy   λ λ + 2µ λ 0 0 0  ε yy 
    
σ zz   λ λ λ + 2µ 0 0 0  ε zz 
 =   
 τ xy   0 0 0 2µ 0 0  ε xy 
    
 τ xz   0 0 0 0 2µ 0  ε xz 

 τ yz   0 0 0 0 0 2 µ  ε yz  ,

where

Eν
λ =
(1 + ν ) (1 − 2ν )
and

E
µ = ,
2(1 + ν )

note that μ and λ are known as the Lamé constants.

Advanced Material: Coordinate rotations

It is important to note that the vector notation we use is a another specified by another set of vectors x′, y′, and z′ (x,
notational system only. We arrange the components of y, z, and x′, y′, and z′ are known as basis vectors). Also,
stress and strain in a vector to make them easy to manipu- define the angle between any of these vectors to be θxx′,
late. However, stress and strain are not mathematically θxy′, etc. A rotation matrix Q can be defined such that
vectors. As you remember, a vector is a quantity that has
both magnitude and direction, like force, deformation, or  
cos (θ xx ′ ) cos (θ xy ′ ) cos (θ xz ′ )
velocity. An implication of this is that a vector maintains
its magnitude and direction from one coordinate system Q = cos (θ yx ′ ) cos (θ yy ′ ) cos (θ yz ′ ) .
 
cos (θ zy ′ )
to another. This implies that the components of a vector  cos (θ zx ′ ) cos (θ zz ′ ) 
must behave in a very specific way in terms of how they  
relate to one another when expressed in different coordi-
nate systems. Assume that we have two (orthogonal) coor- Then any vector can be expressed in the new coordinate
dinate systems, one specified by the vectors x, y, and z and system by multiplying the vector expressed in the old
76 CHAPTER 3: Solid Mechanics Primer

coordinate system by Q. If P is a vector in the x,y,z-sys- In this text we will not be using tensors or tensor mathe-
tem then P′ in the x′,y′,z′-system is given by P′ = QP. matics. However, some of the quantities are in fact ten-
Obeying this coordinate transformation rule is the for- sors, like stress and strain. In Section 3.3 we define some
mal definition of a vector. However, our stress and strain new quantities and refer to them by their names from
“vectors” do not obey this rule. In fact, stress and strain continuum mechanics, so that you can follow up with
are more general mathematical quantities known as ten- additional reading. However, we avoid the use of any
sors. The appropriate transformation rule for tensors of tensor mathematics, and for our purposes they can be
this type requires the terms to be written in matrix nota- manipulated like matrices.
tion. For example,

σ σ xy σ xz 
 xx 
σ = σ yx σ yy σ yz  and σ′ = Q Tσ Q.
 
σ zx σ zy σ zz 


Stress and strain can be expressed as matrices

Using vector notation for stress and strain makes it easier to express the three-
dimensional Hooke’s law as a matrix equation. Notice that each component of
stress and strain has two subscripts referring to coordinate directions. Therefore,
it might make more sense to depict stress and strain as matrices,

   
σ xx σ xy σ xz  ε xx ε xy ε xz 
σ = σ σ yy σ yz  and ε = ε yx ε yy ε yz 
. (3.60)
yx
   
σ zx σ zy σ zz   ε zx ε zy ε zz 

Known as matrix notation, this approach is useful for mathematical manipula-
tions of stress and strain and proving new relationships. Bear in mind that the
same amount of information is contained in each notation (here, there are nine
components each for stress and strain, but three of them are redundant because
of symmetry). To illustrate one advantage of using this notation, we next derive
what are called the principal stresses and strains.

In the principal directions shear stress is zero

Recall from linear algebra that a matrix that is symmetric and positive definite has
an associated eigenvector problem. Specifically, there is a vector that, when mul-
tiplied by the matrix in question, yields the same vector multiplied by a scalar,

Av = ϕ v , (3.61)

where v is the eigenvector, and φ is the eigenvalue. In general, for a 3 × 3 matrix,
there are three eigenvectors/values. We also know from linear algebra that the
principal stresses are solutions of the so-called characteristic equation

 A11 − ϕ A12 A13 

 
| A − ϕ I |= det  A21 A22 − ϕ A23  , (3.62)
 A A32 A33 − ϕ 
 31

Nota Bene
where I is the identity matrix (one of the diagonal zeros elsewhere). This is actu-
Where does the word ally a third-order polynomial equation. The three eigenvalues of stress are
“eigenvalue” come from? Eigen is known as principal stresses, and we will denote them φ1, φ2, and φ3. Likewise,
a German word that means “own,” as the principal strains are denoted, ε1, ε2, and ε3. For isotropic materials, the
in “inherent to.”
eigenvectors for the stress and strain are aligned with each other and are known
 MECHANICS OF DEFORMABLE BODIES 77

as the principal directions v1, v2, and v3. We call these principal stresses and
strains because they are invariant with respect to the coordinate system. That is,
regardless of how you arrange the coordinate axes, the principal stresses and
strains for a given deformation will always be the same. Similarly, the principal
directions will not change. If you were to grab a square-shaped object at oppo-
site corners and pull, the material will deform into a kite-shape. One of the prin-
cipal directions is along the long diagonal of the kite. This remains true regardless
of which coordinate system is used.
The principal directions possess some particularly useful properties. They are
mutually perpendicular or orthogonal to each other. This means that they can
serve as a coordinate system, and it is possible to rotate our coordinate system
from our original x-, y-, and z-axes into v1, v2, and v3. When we do this, the stress
and strain matrices also change to match the new coordinate system, with the
result that the off-diagonal components of the stress matrix vanish, and the diago-
nal components are the principal stresses. In other words, using the principal
directions as our coordinate system, σ becomes

σ 1 0 0
 
σ =0 σ2 0 . (3.63)
0 0 σ 3 
 

This remarkable property means that no matter how complex the state of stress,
there is always a coordinate system in which all the shear stresses vanish and only
normal stresses remain. Indeed, specifying the principal stresses and principal
directions is sufficient to fully define the state of stress. The same holds true for
strain.

Example 3.5: Principal strains

Suppose you are analyzing the deformation of a flexible is the characteristic equation relating our strain ε, the
membrane on which you plan to seed cells to study principal strains εa, and the principal directions va.
their reaction to substrate stretch. You make three Noticing that we can rearrange the equation,
marks that have initial coordinates (0, 0), (1, 0), and
(0,  1). After you apply the deformation, they move to (ε va − Iε ava ) = 0.  (3.66)
(0,  0), (1.015, 0.005), and (0.005, 1.015), respectively.
What would the strains be? What are the principal We can then write,
strains and principal directions?
(ε − Iε a )va = 0.  (3.67)
We start by determining our strains. We have placed our
marks purposely to make this very easy. We have one test This must be true for non-zero va (our eigenvectors) and
line ranging from (0, 0) to (1, 0) and the other from (0, 0) therefore the quantity in parentheses must be non-
to (0, 1). Both of them are originally of length one. invertible. That means the determinant of (ε – Iεa) = 0, or
Consider the normal strains first. Our original line in the
x-direction extends by 1.5%, so εxx = 0.015. It is also dis-
placed by 0.5% in the y-direction, so εxy =  0.005. Likewise, 0.015 − ε a 0.005 = 0. (3.68)
εyy = 0.015 and εyx = 0.005. 0.005 0.015 − ε a

 0.005  Some algebra leads to
ε = 0.015 . (3.64)
0.005 0.015  0.000225 − 0.03ε a + ε a2 − 0.000025
 
(3.69)
To find the principal strains, we need to find the eigen- = ε a2 − 0.03ε a + 0.0002 = 0. 
values of ε.
The two roots of this equation are ε1 = 0.02 and ε2 = 0.01.
ε v a = ε av a  (3.65) Physically this means that the two principal strains are
78 CHAPTER 3: Solid Mechanics Primer

2% and 1%. What directions are these strains in? To In this new coordinate system
answer this we need the principal directions. We simply
substitute back into the characteristic equation 0.02 0.0 
ε= . (3.72)
ε v a = ε av a  0.0 0.01
  
for each εa in turn. We find that the eigenvectors (1, 1) An illustration of the principal strains can be obtained
(corresponding to ε1) and (1, –1) (corresponding to ε2). with the strain ellipse (Figure 3.27). It also shows that
Note that when you are solving for the eigenvectors, any the maximum and minimum normal strains occur in the
eigenvector can be multiplied by a scalar, and the result- principal directions.
ing vector is also an eigenvector. So, you will either need
to constrain the length or to set a component to an arbi- y normal strain
trary value (we use ones for clarity). (Alternatively, you
can construct eigenvectors of unit length.)
principal direction
Finally, notice that the eigenvectors are oriented 45
degrees from our original coordinate system. We can
define a new coordinate system rotated by 45 degrees
counterclockwise such that
x
1 1 1  1 
x ′=   , y ′=  . (3.70)
2 1 2 −1
 

The rotation matrix (described in Advanced Material,

Coordinate rotations p. 73) between the two coordinate Figure 3.27 The strain ellipse is a depiction of the
systems would be normal strain as a function of direction. In this example,
the larger of the two principal strains is in the (1,1) direction
    and the smaller is in the (1, −1).
cos  −π   −3π  
cos   2
−
2
  4   4    2 2 
Q= =  . (3.71)
 cos  π   −π  
cos   2 2 
  4  4    2 2 
  

3.3 LARGE DEFORMATION MECHANICS

In mechanics, if the deformations we are considering are too large to be approxi-
mated as infinitesimal, they fall in the realm of large deformation or finite defor-
mation mechanics. Cellular deformations can be quite large (exceeding 5–10%).
We are going to discuss how such deformation can be ­quantified ­(kinematics),
forgoing discussion of the other two components of mechanics—equilibrium
(stress) and constitutive models.

The deformation gradient tensor describes large

deformations
Consider a general object in its initial undeformed state (Figure 3.28). Every point
Nota Bene
in the object can be described by a vector A. The object undergoes a deformation
A note about notation. Typically, such that it takes on a new configuration. In the new configuration it is now
the undeformed vector is denoted by described by a vector a.
(capital) X and the deformed vector
by (lowercase) x. We have changed This is a general description and can describe any deformation as long as certain
notation here to distinguish “x” from simple requirements are met, such as no holes or cracks and the deformation
the x-coordinate. being smooth. Consider a small line in the undeformed configuration, dA, which
 LARGE DEFORMATION MECHANICS 79

y Figure 3.28 A general object can be

undeformed defined in its undeformed
configuration by a set of vectors A
and in its deformed configuration
by a set of vectors a.

deformed
A

becomes a small line, da, in the deformed configuration. We can relate these two
small elements by approximating the deformation with a Taylor series expansion
and neglecting higher-order terms with the matrix equation:

da ≈ FdA , (3.73)

where F is known as the deformation gradient and is given by

 ∂a Nota Bene
∂a x ∂a x 
 x  Vernacular of large deformation
 ∂Ax ∂Ay ∂Az  mechanics. Much of the notation
 
∂a  ∂ a y ∂a y ∂a y  (3.74)
and terminology of large deformation
F= = . mechanics was introduced by Clifford
∂A  ∂Ax ∂Ay ∂Az  Truesdell and Walter Noll (who had
 
 ∂a z ∂a z ∂a z  been Truesdell’s graduate student) in
the classic 1965 text The Non-Linear
 ∂Ax ∂Ay ∂Az 
  Field Theories of Mechanics. More
recently, Noll has suggested that the
term deformation gradient, which
F is an incredibly useful quantity and forms the basis of almost all of large defor- they introduced, may be misleading
mation kinematics. Here we develop a few particularly relevant applications. and should be replaced with
transplacement gradient, which was
adopted by Truesdell and Noll in
Stretch is another geometrical measure of deformation subsequent work. However,
deformation gradient is now so
Although strain is a natural way to quantify deformation, it is far from the only deeply embedded in the literature
one. Indeed several other deformation measures exist depending on the applica- that it seems unlikely to replaced.
tion. Another popular measure of deformation is stretch. Conceptually, stretch is
simple. If something is pulled twice as long, we say it has a stretch of two. If it does
not change, it has a stretch of one. Stretch is quantified by the stretch ratio, λ, the
ratio of the deformed length divided by the original length, in contrast to strain,
which is the change in length over original length. In our infinitesimal deforma-
tion notation

∆L L + ∆L
ε= ,λ = . (3.75)
L L

In our large deformation notation

|da|
λ= . (3.76)
|dA |

80 CHAPTER 3: Solid Mechanics Primer

Unlike strain, which can be normal or shear, the stretch ratio is always a measure
of the normal deformation of a prescribed differential line element. The stretch
ratio is related to the extension ratio, given by λ − 1, which is similar, but not identi-
cal, to our concept of strain. In particular, stretch is measured along a line ­segment
of interest and follows the line segment as it rotates and translates. In Example 3.5,
the test line that originally goes from (0, 0) to (1, 0) is deformed to a new line seg-
ment from (0, 0) to (1.015, 0.005). We decomposed the strain of this segment into
the εxx and εxy components. For stretch, we do not perform this decomposition.
Instead, we work with the entire vector and create F to describe how the line seg-
ment changes. For this particular deformation, the stretch ends up being just
above 1.015, so there is only a small difference between stretch and strain meas-
ures. However, for larger deformations, the two measures can be divergent. If
instead of (1.015, 0.005) the test line ends up at (1.5, 0.5) then εxx is 0.5 and εxy = 0.5,
but stretch is 1.58 and the extension ratio is 0.58. So although many of the math-
ematical manipulations will be similar (such as determining principal directions),
these quantities are distinct.
Stretch ratios can be defined for any particular line segment. One might choose
dA or da to be aligned with one of the coordinate axes. One particularly useful
approach is to use line segments that are aligned with the principal directions of
stretch. Neglecting rigid-body motions, the deformation in the principal direc-
tions is entirely normal (that is, there is no shear). The stretches are known as the
principal stretches (λ1, λ2, λ3). If the coordinate system is aligned with the princi-
pal directions, F takes a particular form with the principal stretches on the diago-
nals and zeros elsewhere,

λ 0 0
∂a  
1

F= = 0 λ2 0 . (3.77)
∂A  
 0 0 λ3 


The principal stretches can capture the full deformation, and formulations of
mechanical behavior in terms of principal stretches are particularly common in
large deformation constitutive modeling because of their simplicity.

Large deformation strain can be defined in terms of the

deformation gradient
The concept of strain is a measure of how much distortion an object experiences.
In large deformation mechanics, there are actually several quantities that are
acceptable measures of strain. However, we will only define one, the Green–
Lagrange strain, because it can be thought of as the small deformation strain with
an additional term. In terms of F, The Green–Lagrange strain, E, is defined as

E=
2
(
1 T
F F−I . ) (3.78)
Nota Bene

Right Cauchy–Green deformation The term FTF is a special quantity denoted as C = FTF.
tensor. In continuum mechanics, C
is called the right Cauchy–Green It is special because it is related to local stretch. In terms of our small test lines, it
deformation tensor. It is called “right” can be shown that da2 = (dA)C(dA). It also has interesting properties: its trace
not because it is correct, but because (sum of the diagonals) is equal to the sum of the squares of the principal stretches,
it is related to the deformation matrix and its determinant is equal to the product of the squares of the principal stretches.
of a polar decomposition when the On an intuitive level, the Green–Lagrange strain is a measure of how much the
deformation matrix is on the right of
local stretches, C, differ from one. Written out in detail in terms of deformation
the rotation matrix.
the Green–Lagrange strain is
 LARGE DEFORMATION MECHANICS 81

du 1  du 2 2
dw  
2
+    +   + 
dv
E xx =
dx 2   dx   dx   dx  

1   du   dw  
2 2 2
dv  dv 
E yy = +   +   +  
dy 2   dy   dy   dy  
dw 1  du 2 2
dw   .
2
+    +   + 
dv
Ezz = 
dz 2   dz   dz   dz   (3.79)

1  du dv  1  du du dv dv dw dw 
E xy = E yx + +  + +
2  dy dx  2  dx dy dx dy dx dy 
1  du dw  1  du du dv dv dw dw 
E xz = Ezx  +  +  + + 
2  dz dx  2  dx dz dx dz dx dz 
1  dv dw  1  du du dv dv dw dw 
E yz = E yx + +  + + .
2  dz dy  2  dy dz dy dz dy dz 

Notice that each component of strain contains a first term corresponding to small
deformation strains and an additional higher-order term that captures nonlinear
effects of large deformation. With linear homogeneous deformations, the Green–
Lagrange strain and the small strain measures are the same, because these higher-
order terms vanish.

Example 3.6: Large deformation principal strains

Consider our stretching example from Example 3.5, but We also know that da = FdA, so
this time  we apply much larger stretches. Again, our
three markers have initial coordinates (0, 0), (1, 0), and 2.5  F F12  1  0.5  F F12  0 
(0, 1). After the deformation is applied, the markers are  =  11   ,  =  11  
0.5  F21 F22  0  2.5  F21 F22  1 
at (0, 0), (2.5, 0.5), and (0.5, 2.5), respectively. Determine      
the principal strains and principal directions of the (3.81)
deformation. 2.5 0.5  .
⇒F= 
This is similar to our previous principal strain example, 0.5 2.5 
  
except that the deformations are not very small. Our
deformation may be represented by Figure 3.29 (note
that in the previous example, the changes would be so
3
small as to be invisible if drawn). The black lines repre-
sent the undeformed state and the blue lines represent
the deformed state. You can see that our test lines are
quite a bit different, so we need to use the finite deforma- 2
tion approach.
Let us build our F first. From the data, we have the posi- 1
tions of two lines in both the undeformed and deformed
configurations,
0
1  0  2.5  0.5  0 1 2 3
dA1 =   , dA2 =   , da1 =   , da2 =   (3.80)
0 1 0.5 2.5 Figure 3.29 Two lines in the original configuration
         (black) and deformed configuration (blue).
82 CHAPTER 3: Solid Mechanics Primer

With F in hand, we can compute the Green–Lagrange strains is aligned 45 degrees to the original coordinate
strain, E, system. In that direction the strain is 400% and in the
perpendicular direction it would be 150%. If we were to
2.75 1.25  rotate our coordinate system 45 degrees counterclock-
E=
2
(
1 T
F F−I = )
1.25
.
2.75 
(3.82) wise, the new E would be
 
 4.0 0 .
To find the principal strains, we need to find the eigen- E=  (3.83)
values of E. Using the same approach as in Example 3.5,  0 1.5 
 
we obtain principal strains of 1.5 and 4.0 and principal
directions of (1, –1) and (1, 1). The largest of the two Notice that the shear strains are zero.

The deformation gradient can be decomposed into rotation

and stretch components
F can be expressed by polar decomposition as

F = RU, (3.84)
Nota Bene where R is what is an orthonormal matrix. That is, it has special properties such
F is also called the deformation that it is orthogonal, meaning that all of its rows or columns are pairwise orthogo-
gradient tensor. It can be nal (dot products are one if self-multiplied, zero otherwise) and it is normal, or its
manipulated to some extent like a determinant is 1. The properties of R are exactly those required for it to be a rota-
matrix, and as far as we cover in this tion matrix like Q discussed in the Advanced Materials on coordinate rotation
textbook, matrix manipulations are (Section 3.2). R represents a rigid-body rotation of the object. This part of F does
enough to get the results we need.
However, like stress and strain, F
not lead to distortions or strain, but is an integral part of measurements made if
depends on the coordinate system, the principal stretch directions are not known ahead of time. The remaining term,
because it describes a physical U, captures the distortion of the object.
process. Certain rules apply in
manipulating F, including the rotation
U sits on the right side of the decomposition of F. You can alternatively decom-
rules in the Advanced Materials on pose F into VR, where the stretch component V sits on the left side. U and V are
coordinate rotation (p. 75). not the same, because the order of matrix multiplication is important (it is not
commutative), although the rotations are the same.
In Figure 3.30, we see that a square can be rotated by 45 degrees counterclock-
wise first, then stretched to expand one side to twice the original length.
Alternatively, we can stretch the square to twice the original length, and then
rotate it by 45 degrees counterclockwise. The rotations are identical. The material
stretch is the same, but the stretch tensors are different, because in the rotate-first
case, the square is being stretched along a 45-degree line, whereas the stretch-first
case stretches the square along the horizontal axis.
Note that if we include translation in this analysis, we will have a general mathe-
matical transformation for any mechanical deformation—a general deformation
can be expressed as a single translation, a single rotation, and a single distortion.

Figure 3.30 Deformation of an

object can be decomposed into a R deformed
stretch and rotation, or a rotation
and stretch. Whereas rotation is the
same in both cases, the stretch is not. U
F

undeformed V

R
 STRUCTURAL ELEMENTS ARE DEFINED BY THEIR SHAPE AND LOADING MODE 83

Example 3.7: Rotation can be extracted from F

In our previous examples, we mostly focused on defor- We obtain cos θ = 1/2 and θ is either 60 or –60 degrees
mations without really understanding what physical (±π/3 radians). The sin term is negative, meaning that
process was being represented by F. Now that we know θ is –60 degrees. Note that cos θiz = 0 for i = x, x′, y or y′
that F can be decomposed, let us see how R and U result because the z-axis is not moved in this two-­dimensional
in a given deformation. Consider a square that gets problem. Also note that cos θxy ′ = sin θ because the x- and
deformed as in Figure 3.31, where the black lines are the y-axes are 90 degrees apart and the only rotation is about
undeformed configuration and the blue lines are the the z-axis.
deformed configuration.
If we were to use the previous approach of taking eigen-
Say we measure the deformation such that A deforms to values, we would obtain the principal stretches as 1 and 2.
a as We also would obtain the principal stretch directions,
based on U, as 0 and 90 degrees. But it is important to note
1  0   1/2    that these principal stretch directions are not in the coor-
   3 .
A1 =   , A2 =   , a1 =   , a 2 =   (3.85) dinate system we started out with—these are for U, where
0
 
1
   − 3 /2   1  the rotation is already removed. Example 3.6 gives us the
 principal directions in the coordinate system we started
out with, because there’s no rotation in that deformation.
From this we can construct F as in Example 3.6,
So, how much is the rotation? Because we derived R, we
  know that the object being deformed is rotated clockwise
1/ 2 3 .
F= (3.86) by 60 degrees. So now we know how the deformation
− 3 / 2 1 
   occurs; we stretch the square so it doubles in length in
one direction and remains unstretched in the other, and
In some cases, we can extract rotation by finding U first. then rotate it by 60 degrees clockwise. The principal
To obtain U, we note that C = FTF is also U2, because RT directions are along –60 degrees and 30 degrees (0 and 90
is the inverse of R (this characteristic of R is true for rota- degrees, rotated clockwise by 60 degrees).
tion matrices in general; you may show it for yourself
using the Q in Advanced Material, Coordinate rotations
p. 75). If we’re lucky, we can then obtain U. Therefore,
1.5
1 0 ,
F TF =   (3.87)
0 4 1.0
  

from which U is easily derived (noting that stretches can- 0.5

not be negative):

1 0 0.0
U= . (3.88) 0.5 1.0 1.5 2.0 2.5
0 2 
  –0.5
This means that we can now solve for R, given that
F = RU,
–1.0
  cosθ − sinθ  1 0
1/ 2 3
F= =    . (3.89)
− 3 / 2 1   sinθ cosθ  0 2 Figure 3.31 Two boxes in the original configuration
    
 (black) and deformed configuration (blue).

3.4 STRUCTURAL ELEMENTS ARE DEFINED BY

THEIR SHAPE AND LOADING MODE
These equations of solid continuum mechanics are rather general and, as you might
agree, somewhat complicated. However, many structural elements are much more
84 CHAPTER 3: Solid Mechanics Primer

simple. They have been simplified through assumptions of the dimensionality of

the element and the dimensionality of the loading that it is exposed to. In our review
of solid mechanics, we have already employed several examples that we used to
develop our intuition (the axial rod, torsional rod, and beam). We have also seen
how these simplifying assumptions have led to specialized governing equations
between force and displacement. These tend to be much simpler to solve than the
general continuum equations. Here is a list of structural elements, each with its set of
assumptions and the order of the resulting governing equations:
Truss: One-dimensional straight element exposed to axial loading. Axial defor-
mation only. Second order.
Beam: One-dimensional straight element exposed to transverse loading. Bending
deformation only. Fourth order.
Wall: Two-dimensional flat structure exposed to in-plane loading and experienc-
ing in-plane deformation. Second order.
Plate: Two-dimensional flat structure exposed to transverse loading. Bending
deformation only. Fourth order.
Membrane: Two-dimensional curved structure exposed to in-plane loading.
Deformation is only considered within the plane of the membrane. Second order.
Shell: Two-dimensional curved structure exposed to transverse loading. Bending
deformation only. Fourth order.

Key Concepts

• The forces and moments acting on a body that is not • Cytoskeletal proteins are relatively stiff in axial
accelerating must sum to zero. The forces and tension, but flexible in torsion and bending.
moments acting across any closed boundary must also • A column will collapse when it is exposed to
sum to zero. This gives rise to the free-body diagram. axial compression that exceeds the Euler
• For deformable bodies, stress is a normalized buckling load.
measure of force analogous to pressure. • In large deformation mechanics, the strains are not
• Strain is a normalized deformation. assumed to be infinitesimal.
• Stress and strain can be normal or shear. • The deformation gradient is the fundamental large
• Continuum mechanics gives rise to three fundamental deformation kinematic measure, and captures a
relationships: kinematics, constitutive behavior, and body’s rotation and deformation.
equilibrium. • The Green–Lagrange strain is a common large
• Euler–Bernoulli beam theory describes small deformation strain measure, although others exist
transverse deformations of slender bodies and as well.
accounts only for lengthwise stresses.

Problems

1. Let V1 be the volume of a hollow cylinder of outer 2. A cylinder is a composite of two materials, with shear
radius Ro and inner radius Ri (do not use the thin- moduli G1 and G2. The cylinder is composed of material
shell approximation) consisting of material with shear with modulus G1 from 0 ≤ R ≤ Ri and G2 from Ri ≤ R ≤ Ro.
modulus G. This cylinder is fixed at one end; a moment Derive the net effective shear modulus G for the
of M is applied to the other end, resulting in an angular composite material for torsion.
twist of θ. Let V2 be the volume of a solid cylinder of
radius ro consisting of material with shear modulus 3. If an object of arbitrary shape is completely submerged
G. Under the same loading conditions as the hollow in water without touching any other surface (such as
cylinder, the angular twist is θ/2. Derive an expression the bottom floor), is there any pressure on the surface
for ri in terms of the other parameters provided. of the object? Is there net pressure on the object (that
 Problems 85

is, if you integrate the pressure over the entire surface E = 10 GPa (about the stiffness of wood), how much
of the object, is there a net force in any direction)? strain do you expect in the column when the table
Justify your response. is assembled under the minimal stress condition
derived in (a)? Neglect the weight of the column
4. You are given two materials, one with elastic modulus itself.
E1 called Mat A and the other with elastic modulus
E2 = 10E1 called Mat B. The materials are shaped into (c) The strain in part (b) is quite small, so you may
cubes of 1 m3 and loaded with 1 kN of force, evenly neglect it for this part. Someone sitting at the
across the top surface (and supported by an infinitely table is tilting their chair back and resting on the
rigid bottom surface). Estimate the strain (in terms of E1) edge of the tabletop. You model this as a block
for the following: of mass 70 kg with an arrangement as shown in
the illustration below. Assume that there is no
(a) The cube is made entirely of Mat A. friction between the block and tabletop. If the
shear modulus of the column is 5 GPa, find the
(b) The cube is made entirely of Mat B.
total displacement of the tabletop in the lateral
(c) The cube is made of 50% Mat A and 50% (horizontal) direction. Neglect the deformation of
Mat B, with the different materials layered the tabletop. Does the orientation of the column
horizontally (interface is parallel to the load). matter?

(d) The cube is made of 50% Mat A and 50% Mat

B, with the different materials layered vertically
(interface is perpendicular to the load).

(e) The cube is made up of 50% Mat A and 50% kg

Mat B, with the different materials arranged in a
“patchwork” pattern, composed of 1 cm cubes 1m
placed in alternating patterns (like a chessboard,
but in both directions).

5. A toy balances on the edge of a table or on a small 0.5 m 1m

platform (see below). These toys typically look
like they are about to fall over, but because of the 7. Starting over with setup similar to Problem 6, you
presence of a counterweight, they stay balanced. instead wish to maximize the resistance of the bar
Even if they are nudged, they just rock back and forth. to being twisted, as this bar will be used to support
Draw a free-body diagram for such a system at rest. a table, and it is quite irritating using a table whose
All relevant information should be provided in your top twists back and forth. Here, you wish to use
response, including location of the forces and relative a cylinder to support the table. This cylinder is
magnitudes. symmetrically centered with the table and is 1 m
tall (ignore any changes in height during torque
application).

You are using plastic (G = 100 MPa) to build the

column. The design constraint is that you want a
maximum angular twist of 0.01 radians under a torque
of 100 Nm.

(a) You can select from a set of plastic tubes that

have a wall thickness of 1 cm but with any outer
6. You want to build a table with a single support column radius you want. Find the smallest outer radius,
underneath it. Assume that the tabletop is symmetrically to the nearest centimeter, that satisfies the design
balanced and centered on the column. The design constraint. Hint 1: use Matlab/Excel/calculator to
constraints are that the height of the table fixed, and the solve the inequality. Hint 2: do not round down if
column requires a rectangular cross section of constant the rounded solution does not satisfy the design
perimeter, for which you can alter the aspect ratio (the requirement.
ratio of the lengths of adjacent sides).
(b) You now can select from a set of solid plastic
(a) If the sides of the rectangle are a and b, determine cylinders of the same material, with any radius you
the relationship between a and b such that the want. Again, find the smallest radius (to the nearest
stress (magnitude) in the column is minimized. centimeter) that satisfies the design constraint, and
Briefly justify your answer. calculate the ratio of areas of the suitable solid to
the hollow cylinder you found in (a). What does the
(b) Given that a + b = 20 cm, and the tabletop has ratio tell you about where the torsion load is being
a mass of 10 kg and the column material has most resisted?
86 CHAPTER 3: Solid Mechanics Primer

8. Assuming a model of a cell as a thin-walled sphere 17. Solve the beam equation for a beam that is free to
enclosing a pressurized fluid, derive the relationship rotate at the supports and is loaded with a single point
between the stress in the membrane and the cytoplasm load at mid-span.
pressure.
18. We solved for the buckling load assuming that the base
9. In calculating GJ for cytoskeletal proteins (Table 3.2), of the beam is free to rotate. What is the buckling load
we ignored the fact that microtubules are hollow. for the clamped configuration? This can be done with
Recalculate G and GJ for a hollow microtubule, symmetry arguments. Do not solve the beam equation.
assuming that the inner radius is 11 nm. How large
was our error? Was our order-of-magnitude estimate 19. Solve the beam equation for a beam supporting a
acceptable? distributed pressure of Q N/m that is free to rotate at
the supports.
10. The other error we made in Table 3.2 was that we
could estimate the shear modulus G with the Young’s 20. Often we are interested in the energy stored in a
modulus E. Assume an effective Poisson’s ratio of 0.3. structure because of its deformation or strain
How large was the error? Was our order-of-magnitude energy. One approach is to integrate the strain
estimate acceptable? What is the result if you assume a energy density (strain energy per unit volume) over
Poisson’s ratio of 0.0 or 0.5? the structure,

11. Determine the second moment of inertia I for a 1

dw = σε
cylindrical cross section of a beam of diameter D. You 2
may wish to use cylindrical coordinates.
where the product of stress and strain is taken
12. Determine the second moment of inertia I for a hollow component-by-component. Taking this approach, show
cylinder, diameter D, and lumen Do. You may wish to that for a beam
use cylindrical coordinates.
2 2
13. Why is it advantageous for a microtubule to be 1 1  d2w  1  y
E ( yκ ) = E  y
2
dw = = E  .
hollow? Using your result from the preceding question, 2 2  dw 2  2  R
determine the mass ratio and flexural rigidity ratio for
a microtubule with inner and outer radii 11.5 nm and
21. Provide a one-sentence description for each of the
14 nm respectively, compared with a solid microtubule
three basic relationships in continuum mechanics:
with the same outer radius. What is the most efficient
kinematics, constitutive behaviour, and equilibrium.
use of proteins to gain bending rigidity (attaining the
most rigidity with the least mass): one solid microtubule 22. Using the result from Problem 20, show that the strain
or several hollow ones? energy of a beam that is bent in a hairpin (180 degrees)
14. Show that J = Ix + Iy. Eπ Iy
is .
2R
15. Glass micropipettes in bending are often used as cell 23. Assume a state of stress σx = σ, σy = −σ, and σz = 0.
and molecular force transducers. The tip defection Show that this is a state of pure shear. Use this
acts as a linear spring, obeying Hooke’s law. Determine information to derive the shear modulus in terms of the
the spring constant of the tip for a glass rod of radius Young’s modulus and Poisson’s ratio.
0.25 μm, length 100 μm, and Young’s modulus 70 GPa.
The spring constant can be decreased further by 24. Consider a membrane that is in a bi-axial state of stress
increasing the length or decreasing the radius. What is such that σxx + σyy = σ and σzz = 0. We can define the
the effect on the spring constant of doubling the length? bi-axial strain to be εb = εxx + εyy. Determine the bi-axial
What is the effect on the spring constant of halving the modulus Eb = σ/εb in terms of E and ν. Also, determine
radius? how εb is related to the change in area over the original
area for a differential portion of the membrane.
16. Imagine that a cantilevered beam is loaded until the
point of failure by a single transverse load at the tip. 25. Consider Examples 3.5 and 3.6. You will notice that the
Assume that the beam is made of a material that has a displacements were selected in a specific way—the
tensile failure stress that is lower than its compressive large displacement example is 100 times the small
failure stress. Where in the beam will failure initiate? If displacement example. However, E is not simply 100
the failure stress of the material is σf, what is the failure times ε. Why is that? What happens if you neglect the
load at the tip? second-order terms in E?
 Annotated References 87

Annotated References

Fung YC (1977) A First Course in Continuum Mechanics. Prentice- Truesdell C & Noll W (1919) The Non-Linear Field Theories of Mechan-
Hall. An outstanding introduction to infinitesimal fluid and solid con- ics. Springer. A seminal book that first presented a standard unified
tinuum mechanics that includes a very lucid introduction to tensor set of concepts and notations for large deformation mechanics.
analysis. Currently out of print, but the third edition incorporates corrections
from the late Professor Truesdell’s personal copy. The history and
Malvern LE (1969) Introduction to the Mechanics of a Continuous
rationale for updating the terminology is outlined on Noll’s (cur-
Medium. Prentice-Hall. A comprehensive and classic text on large
rently at Carnegie Mellon University) website (http://www.math.
deformation mechanics. Mathematically intense and rigorous.
cmu.edu/~ wn0g/noll).
Timoshenko SP & Goodier JN (1934) Theory of Elasticity. McGraw-
Hill. A classic text in small deformation elasticity. Includes easy-to-
follow treatments of beams, plates, torsion, etc.
This page intentionally left blank
to match pagination of print book
CHAPTER 4

Fluid Mechanics Primer

T he mechanics of fluids can play a critical role in maintaining normal cellular

physiology and in mediating pathological processes. Many physiological
processes depend on the presence of extracellular fluid flow to transport nutri-
ents and waste products to and from various locations. In addition, mechanical
loads imparted by the fluid in the form of pressure and fluid shear stress can
function as potent regulatory signals to cells. Because cells are largely composed
of fluid, it comes as no surprise that within cells, fluid mechanics can influence a
variety of processes, such as those associated with cell motion or intracellular
transport. Given that fluid mechanics is essential to understanding many aspects
of cellular mechanobiology, in this chapter, we give a brief introduction to fluid
mechanics principles. As in Chapters 2 and 3, this primer cannot cover the field
in its entirety. Instead, our goal is to provide a basic understanding of fluid
mechanics principles that will allow readers to have a better grasp of topics asso-
ciated with fluid statics and dynamics in subsequent chapters. It will facilitate the
understanding of the process of mechanotransduction in cells regulated by fluid-
derived forces (covered in Section 11.1), as well as some experimental methods
in cell mechanics (covered in Section 6.3). In this chapter, we cover the basic
fluid statics and dynamics, the Navier–Stokes equations, the distinction between
Newtonian and non-Newtonian fluids, and rheological analysis. We conclude
the chapter with a treatment of dimensional analysis.

4.1 ​FLUID STATICS

We begin our discussion of fluid mechanics with a basic introduction to fluid
statics. Fluid statics, which is sometimes referred to as hydrostatics, is a subdis-
cipline within fluid mechanics that deals with the mechanics of fluids at rest. In
being at rest, the fluid is assumed to take the shape of the container that it is in
(this ability is what is often used to defined a substance as being a fluid rather
than a solid). In this section, we will see that although the fluid is at rest, it exerts
forces on surfaces it contacts. We will also see that these forces, which are
derived from the gravitational forces acting on the fluid, can be understood from
the fluid pressure.

Hydrostatic pressure results from gravitational forces

To introduce the concept of hydrostatic pressure, consider a cylindrical glass con-
tainer filled with water. Because the water has mass, it exerts a force on the bottom
of the glass from gravity. Assume that the container has cross-sectional area A, the
height of the water within the container is h, and the density of water is constant
and equal to ρ (Figure 4.1). The total gravitational force exerted on the bottom of
the container is

F = ρgAh , (4.1)

90 ChApTER 4: Fluid Mechanics Primer

Figure 4.1 ​A glass of water

experiences hydrostatic pressure
on its bottom from the weight of
the water. The height of the water
column within the glass is h, and the
cross-sectional area of the glass is A.
The density of water is ρ.

Nota Bene
height = h
Incompressible fluids. In deriving density = ρ
Equation 4.1, we assumed that the
density of the water was constant, in
other words the density of the fluid is
not altered even when subjected to
large pressures. Such fluids are said to area = A
be incompressible. Most fluids
(including water) are generally
pressure = p
assumed to be incompressible under
most relevant circumstances. However,
there are cases in which pressures are
extremely large and compressibility
may need to be accounted for, even where g is the acceleration due to gravity. In this case, F acts in the direction nor-
for fluids that are normally considered mal to the bottom of the container and in a downward direction. As we discussed
incompressible. For example, die in Section 3.2, pressure is the force per unit area in the direction normal to the
casting involves the injection of molten surface on which it is acting. Given this definition, we can express Equation 4.1 in
metal into a die at extremely high terms of pressure,
pressures. This pressure causes the
liquid metal to compress, and this
compressibility must be taken into p = F /A = ρgh. (4.2)

account for in the design of the die.
Such a pressure is referred to as hydrostatic pressure, because it is the pressure
exerted by a fluid at equilibrium, resulting here from gravitational forces.
Equation 4.2 states that the hydrostatic pressure acting on the bottom of the
container is proportional to the height of the fluid. More generally, the hydro-
static pressure experienced by a cell or organism in contact with a fluid is
dependent on the height of the fluid above it. Cells and organisms may be
exposed to hydrostatic pressures that vary quite a bit. For instance, aquatic
organisms may experience different hydrostatic pressures as they swim at dif-
ferent depths. If one were to hang upside down by one’s feet (Figure 4.2), vascu-
lar cells in the feet and skull would experience a decrease and increase in
hydrostatic pressure, respectively.

Figure 4.2 ​Schematic of a man low hydrostatic pressure

hanging upside down during
inversion therapy, which has been
used in the past by patients
seeking relief from back pain. Upon
hanging upside down, vascular cells in
the feet and skull experience a reversal
in hydrostatic pressure due to the
change in fluid height above the cells.

high hydrostatic pressure

 FLUID STATICS 91

Figure 4.3 ​Schematic depicting an

tubing apparatus for micropipette
aspiration. The negative pressure at
the micropipette tip only depends on the
micropipette height difference between the surface
of the fluid in the reservoir and the tip of
the micropipette, and is independent of
the specific path the tube traverses.

h
cells

reservoir

Hydrostatic pressure is isotropic

In the last section, we found that the water in a cylindrical container will exert
Nota Bene
pressure on the bottom of the container. But what about the sides? Consider an
infinitesimally small cube of fluid within the container. Because the fluid is at Absolute pressure versus gauge
rest, it follows from equilibrium that the pressure on all sides of the cube must pressure. In many cases, pressures
are measured relative to a
be equal. Therefore, the pressure within a fluid at rest is isotropic. If we assign a convenient baseline that is taken to
coordinate system to our container where the z-axis is perpendicular to the be zero. Often this baseline is the
container bottom, and the x–y plane is parallel to it, and we were to traverse to atmospheric pressure (the pressure
any point within a given x–y plane, the pressure would be identical, no matter due to the weight of the air in the
the path taken or distance traveled. An example of this principle can be atmosphere). These types of pressure
observed in a typical experimental setup for micropipette aspiration, which we are called gauge, or gage, pressures.
Blood pressure is a gauge pressure.
introduced in Section 1.3. In such a setup, one seeks to generate a negative
In contrast, absolute pressure is
pressure at the tip of a micropipette by coupling a micropipette and a reservoir relative to a vacuum.
with fluid-filled tubing (Figure 4.3). In this instance, the pressure only depends
on the height difference between the surface of the fluid in the reservoir and
the tip of the micropipette, and is independent of the path the tube traverses in
the x- and y-directions.
The isotropic nature of hydrostatic pressure also suggests that, within our fluid-
filled container, the water will exert force against the sides of the container in
addition to exerting force against the bottom. Consider a small fluid element in
contact with the side of the container (Figure 4.4). The fluid element experi-
ences a force from the glass that is equal in magnitude to the force exerted by the
surrounding water. If the glass were suddenly removed, the fluid element would
experience an imbalance in force, and would flow, collapsing into a puddle.

force from glass

Figure 4.4 ​The fluid element adjacent to the container side is subjected to forces
from the glass and the surrounding fluid. In this case, the glass exerts a force that
resists the net force imparted on the element by the surrounding forces which keeps it from
accelerating. If the glass were suddenly removed, the fluid would flow, and the column of
fluid would collapse into a puddle.
92 ChApTER 4: Fluid Mechanics Primer

Resultant forces arising from hydrostatic pressure can be

calculated through integration
In the last section, we saw that hydrostatic pressure is isotropic, and that our fluid-
filled container experiences forces from the fluid on its bottom and sides. Because
pressure is depth-dependent, the pressure exerted by the fluid on the container
sides will vary spatially: low near the top of the water surface and high near the
bottom of the container. We might ask, how do we calculate the resultant forces
acting on the sides of the container if the pressures are depth-dependent?
Integration is required, as is shown in the following example.

Example 4.1: Force of fluids being held by a wall

Consider the case in which we wish to construct an z
open-top, rectangular-shaped culture vessel, or biore-
actor, that will be filled completely to the top with cul-
ture media. The walls of the vessel are oriented vertically g
and have height h and width w. Calculate the force
exerted by the media on the walls of the bioreactor.
The pressure acting on a small fluid element is p = ρgz,
where z is the vertical distance between the surface of
the water and the fluid element, assuming the pressure
at the surface is zero (gauge pressure) indicating we are d
using gauge pressure. For an infinitesimally thin strip H
along the wall of the vessel of height dz and width w
located at depth z (Figure 4.5), the resultant force is w
dz
dF = ρgzdA,
dA = wdz
where Figure 4.5 ​Schematic depicting a single wall of a fluid-
filled cell culture vessel.
dA = wdz.

The total force can be found as

h
ρgwh 2
∫
F = dF =
∫
z =0
ρgzwdz =
2
.

4.2 ​NEWTONIAN FLUIDS

We now turn our attention from fluid statics to fluid dynamics. We previously
stated that a fluid is classified as a substance that adapts to the shape of its
container. A more precise definition can be stated as follows: a fluid is a sub-
stance that responds to shear stress with continual deformation. The specific
way in which deformation occurs under shear can vary between different
types of fluids. A Newtonian fluid is a substance that responds to shear stress
by deforming at a shear rate proportional to the applied shear stress. To illus-
trate, consider a thin layer of fluid between two plates exposed to a constant
shear stress (Figure 4.6). Such a shear stress can be imposed by sandwiching
fluid between two large parallel plates, and moving the top plate at a constant
velocity V0, while the bottom plate is kept stationary. This generates a flow
that is assumed to be steady, in other words the velocity field is not changing
 NEWTONIAN FLUIDS 93

Figure 4.6 ​A fluid is sandwiched

between two plates. The bottom
plate is stationary and the top plate is
V0 moving to the right (in the positive
x-direction) with a velocity V0. A
Newtonian fluid develops a linear flow
h y profile and shear stress that is
proportional to the shear rate. In this
x case, the shear rate is V0/h.
fluid between
plates
stationary

with time. Between the plates, the fluid velocity varies linearly in the y-direc-
tion, with zero velocity at y = 0, and a velocity of V0 at y = h. This characteriza-
tion of the fluid velocity is called a flow profile. In the case of a linear flow
profile, a constant shear stress is applied to the upper plate to maintain this
profile. For a Newtonian fluid, this shear stress is
∂u
τ =µ , (4.3)
∂y

where τ is the shear stress, u is the fluid velocity and y is a direction perpendicular
to the direction of u. The derivative of u with respect to y is called the shear rate or
velocity gradient. The constant μ represents the dynamic viscosity, and has units
kg/ms. We will see later in Section 4.3 that Equation 4.3 comes from a more gen-
eral set of constitutive relations for Newtonian fluids. Nota Bene

Given Equation 4.3, we can determine the magnitude of the shear stress in our No-slip condition. In our example
of the linear flow profile, the fluid
chamber as contacting the plates themselves has
the same velocity as the plates. This
V0
τ =µ . (4.4) condition is called the no-slip
h condition, which stipulates that fluid
immediately adjacent to a solid
Upon inspection of Equation 4.4, we see that how fast the fluid deforms in surface cannot slip along that
response to a given shear stress is determined by the fluid viscosity. A highly vis- surface, and will travel with the same
velocity as the surface itself. An
cous fluid such as honey will undergo a much slower velocity for the same-sized everyday example of the no-slip
gap, h, in response to a given shear stress, compared with a fluid with a lower condition is that dust on fan blades
viscosity, such as water. Alternatively, it takes a lot more shear stress to move the tends to stay in place no matter how
upper plate with a desired velocity, V0, for a fluid like honey. fast the fan spins.

Fluids obey mass conservation

Fluids, just like solids, obey mass conservation. For incompressible fluids, mass
conservation states that the volumetric flow rate into a fixed volume must equal
the volumetric flow rate out, or

Vin Ain = Vout Aout , (4.5)

where Vin is the average velocity of all the fluid entering, Ain is the cross-­
sectional area normal to the flow entry, and Vout and Aout are similarly defined Nota Bene
for fluid exiting (Figure 4.7). The product (V)(A) is referred to as the volumetric
Streams and rivers. An everyday
flow rate. The reason why volumetric flow rate must be conserved is that for example of conservation of mass is
incompressible fluids, where the density is assumed constant, volume is the water speed of streams and
directly proportional to mass. Conservation of mass is one of the most useful rivers. In deep regions of a channel
concepts in fluid mechanics. It explains why a pipe, a narrowing of the diame- (which have a larger cross-sectional
ter in a pipe leads to increased velocity of the fluid flowing within. As we will area), the water tends to flow slowly,
see, this relation will be critical in allowing us to solve the Navier–Stokes equa- while shallow regions tend to be
moving quickly.
tions in Section 4.3.
94 ChApTER 4: Fluid Mechanics Primer

Figure 4.7 ​In some fixed volume,

the rate that fluid mass enters the Vin
volume must be matched by the
fluid mass leaving the volume. Vout
Under incompressible conditions, this
can be expressed mathematically as the
velocity times the area must be
constant. In this figure, fluid leaving the
volume must be moving faster than the
fluid entering the volume, because the Aout
exit area, Aout is smaller than the inlet
area Ain.
Ain

Fluid flows can be laminar or turbulent

Before we present some analytical solutions, it is necessary to discuss the con-
cepts of laminar and turbulent flow. In general, turbulent flow can be roughly
Nota Bene
characterized as chaotic and irregular and involving some degree of stochasticity
Convection versus diffusion. A and/or randomness. Laminar flow is any flow in which turbulence is not exhib-
fluid that is in motion is able to carry ited. You may already have some intuition for the differences between such flows
things with it as a consequence of
based on everyday occurrences. Pouring a thick fluid, such as honey or certain
that motion by convection. Typically
this would be either a dissolved oils, into a dish generally results in a flow that is laminar. Opening the faucet
solute or heat. Convection is an slightly, so that the water is clear, will also generally result in laminar flow (assum-
important mechanism of mixing in ing you do not have an aerator). Opening the faucet more will result in a chaotic
turbulent flow. In contrast to flow, with the water churning and mixing—a characteristic of turbulent flow
convection, diffusion is the motion of (Figure 4.8).
particles in a fluid that occurs
independent of the bulk velocity. More formally, laminar flows are flows without any internal convective mixing, in
which fluid elements travel in well-defined “lines.” These lines can be visualized
by injecting a small amount of dye into the flow. The flow can be injected with dye
at many spots simultaneously to visualize the flow field. For laminar flows, the
lines of dye will remain cohesive. Because there is no active mixing, such flow
visualization generally will yield many parallel strands, from which one can imag-
ine layers of, without crossing from one layer to the next. This pattern of layers is
the source of the name “laminar.”
By measuring certain aspects of the flow and the geometry through which it is
occurring, we can distinguish between laminar and turbulent flow with a dimen-
sionless number called the Reynolds number. Abbreviated Re,
ρVL
Re = (4.6)
µ

Figure 4.8 ​Laminar versus turbulent dye injection point

flows. Laminar flow tends to be steady
and smooth, with unbroken streamlines
(which may be curved or straight and
may change with time). Turbulent flows
will exhibit mixing and disrupted laminar flow
streamlines and are time-dependent.

turbulent flow
 NEWTONIAN FLUIDS 95

where ρ and μ are the fluid density and dynamic viscosity, respectively, V is what Nota Bene
is called a characteristic velocity, and L is some length scale. For flow in tubes and
Transition to turbulence. Though
pipes, L is generally the diameter, and V is generally the average velocity (although
Re = 1 is often considered a transition
the radius and peak velocity may also be used). point at which flows begin to exhibit
The Reynolds number measures the relative magnitudes of inertial versus viscous turbulence, for steady pipe flows, Re
can rise as high as 2000 or so before
forces in a given flow and is widely considered one of the most important quanti-
such a transition occurs. The reason
ties in fluid mechanics. In general, a Reynolds number greater than one indicates for this is that in pipe flows, the fluid
that inertial forces, in other words how much momentum the fluid has, dominates is highly constrained, moving in a
the flow. Such inertial forces tend to drive mixing, and high Reynolds numbers are single dimension. In this instance, the
associated with turbulent flow. Reynolds numbers of less than 1 indicate that vis- changes in inertial terms tend to be
cous forces dominate. Note that at Re = 1, we have a balance between viscous and very small and mixing is more a
inertial terms, and this is sometimes referred to as the transition region, where the function of surface roughness.
flow can start developing instabilities, but is not yet fully turbulent.

Many laminar flows can be solved analytically

The ability to distinguish between laminar and turbulent is of great importance Nota Bene
when attempting to solve for a flow profile. Turbulent flows can generally not be Laminar flow at the cellular level.
solved analytically; however, many laminar flows can. In this section, we solve for Most flows at the cellular level are
pressure-driven flow between two parallel plates using a simple free-body analy- laminar by virtue of the small length
scales and low velocities prevalent in
sis of a differential fluid element. Consider a left-to-right, pressure-driven flow
the cellular environment. For
between parallel plates. We assume the fluid is Newtonian and incompressible instance, a typical capillary has a
and that the flow is steady and laminar. We also assume that the flow is fully devel- diameter of 10 μm and a blood
oped, in other words the flow is far from any inlet or outlet, so that the flow profile velocity on the order in 0.1 mm/s. If
is not changing in the flow direction. Finally, we assume that the flow is exclu- we assume the blood has a density
sively in the x-direction, with zero velocity in the y- and z-directions. Now, con- of approximately 1000 kg/m3 and a
sider a small fluid element of volume (dxdydz) within the flow and construct a viscosity of 0.001 kg/ms, plugging
these numbers into Equation 4.6, we
free-body diagram for the fluid element as in Figure 4.9.
obtain a value of Re = 0.001. This
In the x-direction, the fluid element is subjected to shear forces acting on the top indicates that viscous forces
and bottom surfaces, as well as forces due to pressure exerted by the surrounding dominate the flow, suggesting that
turbulence is unlikely to occur.
fluid. Summing these forces, we obtain the expression

 ∂P   ∂τ 
Pdydz −  P + dx  dydz +  τ dy dxdz − τ dxdz = 0 (4.7)
 ∂x   ∂y 

Figure 4.9 F ​ orce balance in

pressure-driven flow. A pressure-
driven flow can be solved by considering
a small infinitesimal fluid element (a
differential element) within the fluid.
h There are shear stresses acting on the
top and bottom surfaces and pressures
y acting on the left and right surfaces. The
remaining two surfaces experience no
x shear stress, and the pressures on them
are equal and opposite.

∂τ
τ + — dy
∂y
dz
∂p
p + — dx
p dy ∂x

dx τ
96 ChApTER 4: Fluid Mechanics Primer

which can be simplified to

∂P ∂τ
− dxdydz + dydxdz = 0. (4.8)
∂x ∂y

Dividing Equation 4.8 by the volume of the fluid element (dxdydz) yields
∂P ∂τ
− + = 0. (4.9)
∂x ∂y

Next, we examine the spatial dependence of the pressure and shear gradient in
the above expression. Consider the dependence of the shear gradient, ∂τ/∂y, on
x, y, and z. First, because the flow is fully developed and not changing in the
x-direction, shear stress is also not changing in the x-direction. Because we
assume no flow in the z-direction, this implies that the shear gradient can only be
a function of y,
∂τ
= f ( y ). (4.10)
∂y

Now consider the pressure gradient. Because the flow velocity is assumed to be
zero in the y- and z-directions, this implies that pressure is not a function of y and
z, and the pressure gradient can only be a function of x,
∂P
= g ( x ). (4.11)
∂x

However, Equation 4.9 states that the difference of Equations 4.10 and 4.11 must
equal zero,
− f ( y ) + g ( x ) = 0. (4.12)

The condition in Equation 4.12 can only be met if f ( y) and g(x) are both equal to
some constant, or

f ( y ) = g ( x ) = constant. (4.13)

To determine this constant, one can solve for ∂τ/∂y in Equation 4.9 and inte-
grate with respect to y to obtain
∂P
y + C1 = τ . (4.14)
∂x

Substituting in Equation 4.3 for the shear stress in the right-hand side of
Equation 4.14, we can express this relation as a function of u as
∂P ∂u
y + C1 = µ , (4.15)
∂x ∂y

where u is the velocity in the x-direction. Dividing Equation 4.15 by viscosity and
integrating with respect to y once more, we obtain the velocity profile
1 ∂P 2
y + C1 y + C 2 = u( y ), (4.16)
2 µ ∂x

where C1 and C2 are unknown constants.

To solve for these constants, we use the boundary conditions. In particular,
we rely on the no-slip characteristic of fluids—that the fluid velocity at a solid
 NEWTONIAN FLUIDS 97

boundary is exactly equal to the velocity of the solid boundary. The top and bot- Nota Bene
tom plates are stationary and u( y = 0) = 0, and u( y = h) = 0. This gives us the
Bernoulli’s equation. Consider a
final flow profile,
steady and laminar flow in which we
1 ∂P 2 inject dye, leaving a line of dye that is
( y − hy ) = u( y ). (4.17) tangent everywhere to the velocity of
2 µ ∂x the particle. Formally, we call this line
a streamline. As a particle moves
Equation 4.17 describes a parabolic velocity profile, which is commonly encoun- along a streamline, it will speed up or
tered in a variety of situations involving fluid flow. We will encounter this profile slow down due to the different forces
acting on it, such as those due to
again in our discussion of flow chambers in Section 6.3. In addition, it is straight- changes in fluid pressure and from
forward to show—using a similar approach—that within a pipe of circular cross gravity. If the flow is inviscid, in other
section, a parabolic profile is also obtained. words the fluid is assumed to have
zero viscosity, then we can relate
these parameters using a relationship
Many biological fluids can exhibit non-Newtonian behavior called Bernoulli’s equation,
In the last section we found that a Newtonian fluid will exhibit a parabolic veloc- 1 2
ity profile when subjected to pressure-driven flow between parallel plates. This P+ ρV + ρg z = constant
2
solution depended on the incorporation of the constitutive equation for
Newtonian fluids, Equation 4.3, which related shear stress and shear rate (or In the above expression, P is the local
pressure, ρ is the fluid density, V is
velocity gradient) through a constant, viscosity. Not all fluids follow the linear
the local fluid velocity, g is the
relation between shear rate (or velocity gradient) and shear stress that character- gravitational constant, and z is the
ize Newtonian fluids. The fluid here is considered non-Newtonian. For many bio- local elevation. We can see that along
logical fluid flows, viscosity is not constant, but a function of other parameters a streamline, for a given height, as
such as shear rate. pressure increases, velocity
decreases. A great way to experience
One non-Newtonian fluid is the Bingham plastic or Bingham fluid, which has this phenomenon is to hold two
been used to model the flow of biological fluids such as blood and mucus that pieces of paper close together and
exhibit solidlike behavior at low stresses but fluidlike behavior at large stresses. gently blow between them. They will
For this fluid there is a critical value of shear stress, τ0, below which the shear rate tend to move together due to the
is zero, such that increased velocity of air, which
results in a decrease in pressure
∂u between the two pieces of paper.
= 0 for τ < τ 0
∂y
(4.18)
∂u τ − τ 0
= for τ ≥ τ 0 .
∂y µ

If the shear stress is below τ0, the fluid does not deform or flow. To demonstrate
this, consider our parallel plate example in the previous section, except we substi-
tute the Newtonian fluid with a Bingham fluid. Assume that the peak level of shear
stress computed using Equations 4.13 and 4.17 is less than τ0. Here, ∂u/∂y = 0 and
u must be constant. However, we know that the velocity is zero at the plate sur-
faces by the no-slip condition and u = 0 everywhere. If this level of shear is greater
than or equal to τ0, then the fluid can flow. A classic example of the application of
the Bingham fluid model is to model paint sticking to a wall—if the shear forces
arising from gravity are not sufficiently large, the fluid stays motionless until it
dries out.
Another non-Newtonian fluid is a power-law fluid, which can be modeled using
the relationship
α
 ∂u 
τ = β  , (4.19)
 ∂y 

where β is a constant viscous factor (not viscosity) and α is a constant scaling
exponent. Equation 4.19 can be recast as

∂u
τ = µ eff , (4.20)
∂y
98 ChApTER 4: Fluid Mechanics Primer

Figure 4.10 Flow profiles for (A) (A)

Newtonian, (B) Bingham plastic,
and (C) shear-thinning fluids. The
Newtonian fluid exhibits a parabolic Newtonian
profile. Both the Bingham plastic and the
shear-thinning fluid have blunted
profiles, reminiscent of having cells
concentrated in the middle, and are thus
used to model blood flow in certain
circumstances. The Bingham plastic has
a sharply flat profile in the middle, where (B)
shear is lower than the critical shear.

Bingham plastic

(C)

power law
shear thinning

where
α −1
 ∂u 
µ eff = β   . (4.21)
 ∂y 

In this formulation, the effective viscosity of the fluid is shear rate–dependent, and
that dependency changes whether the scaling exponent is greater or less than
one. When α is greater than one, the fluid undergoes shear-thickening, in other
words the effective viscosity increases as the shear rate increases. There are not
too many everyday examples of shear-thickening solutions, but a watery mixture
of cornstarch can act as one—so much so that even though you can stir a vat of it,
you can also walk across its surface if you step quickly enough.
If the exponent is less than one, the fluid undergoes shear-thinning, where the vis-
cosity decreases as the shear rate increases. An example of a shear-thinning fluid is
ketchup, which shows its non-Newtonian behavior when one is trying to get it out of
a glass bottle. Initially, getting the ketchup to flow is difficult, but once it starts flow-
ing, it tends to flow easily. So, stirring or shaking the bottle of ketchup will shear the
fluid and temporarily decrease its viscosity and allow it to flow faster. Blood is a well-
known biological fluid that tends to exhibit shear-thinning behavior. One reason
that blood exhibits shear-thinning is that with flow the red blood cells tend to
become oriented with one another, thereby reducing viscocity. When the scaling
exponent is equal to one, we recover Newtonian fluid behavior (Figure 4.10).

4.3 ​THe NaVieR–StoKes EQuations

In Section 4.2, we derived an expression for flow with simple geometry assuming
the flow was laminar, steady, and fully developed. Such conditions may not be
applicable in vivo, where typical flows may be unsteady, spatially heterogeneous,
turbulent, and occurring over complicated geometries. We now derive a set of
equations called the Navier–Stokes equations that provide a mathematical char-
acterization of general fluid flows.
The Navier–Stokes equations are a general set of equations that describe the
motion of fluids. They are incredibly important and are used describe a variety of
 THe NaVieR–StoKes EQuations 99

phenomena such as air flow around an airplane wing, water flow in pipes, and
currents in the ocean. With regard to cellular mechanobiology, they have been
critical in shedding light on the types of mechanical forces that cells are exposed
to in vivo. For example, they have been used to predict fluid shear stresses and
pressure profiles for cells subjected to blood or interstitial fluid flow. In addition,
they are useful for modeling flow profiles for cells subjected to fluid flow in vitro,
such as within flow chambers and bioreactors.

Derivation of the Navier–Stokes equations begins with

Newton’s second law
In this section, we will derive the Navier–Stokes equations for incompressible,
time-dependent flow. We begin our derivation with an application of Newton’s
second law of motion. This is analogous to our application of the equilibrium con-
dition in Section 3.2 (recall that the key components of continuum mechanics are
equilibrium, constitutive, and kinematic relationships); however, this case is
broader because we do not assume accelerations are zero. Our strategy is to con-
sider an infinitesimally small fluid element (similar to our approach in deriving
our definitions of stress), determine the external forces acting on it, and then use
these forces to apply Newton’s second law of motion.
Consider a flowing fluid defined by the velocity vector u(x, y, z, t), where x, y, z are
spatial coordinates, and t is time. The velocity vector has three components, u, v,
and w, which give the velocities in the x-, y-, and z-directions respectively

 u( x , y , z , t ) 
u( x , y , z , t ) =  v( x , y , z , t )  . (4.22)
w( x , y , z , t )

We now consider a small rectangular-shaped fluid element moving with velocity
u(x, y, z, t). The fluid element is assumed to be aligned with the coordinate axes,
centered at point (x, y, z), and assumed to have dimensions Δx, Δy, and Δz, in the
x-, y-, and z-directions, respectively. The external forces are assumed to arise from
one of two sources; stresses imparted by the surrounding fluid acting on the faces
of the control volume, and body forces (such as that due to gravity).
First, consider the forces that act on the faces of the control volume. For simplicity,
we will begin with the forces in the x-direction. Because the fluid element has six
faces, there are six forces we must consider, one on each face: Fx(x − dx/2, y, z),
Fx(x + dx/2, y, z), Fx(x, y − dy/2, z), Fx(x, y + dy/2, z), Fx(x, y, z − dz/2), and Fx(x, y,
z + dz/2) (Figure 4.11). These forces can be computed as the product of the stress
acting on the face and the area of the face over which it is acting. In the x-direction,

 ∆x   ∆x 
Fx  x − , y , z  = σ xx  x − , y , z  ∆y ∆z
 2   2 

 ∆x   ∆x 
Fx  x + , y , z  = σ xx  x + , y , z  ∆y ∆z
 2   2 

 ∆y   ∆y 
Fx  x , y − , z  = σ xy  x , y − , z  ∆x ∆z
 2   2 

 ∆y   ∆y  (4.23)
Fx  x , y + , z  = σ xy  x , y + , z  ∆x ∆z
 2   2 

 ∆z   ∆z 
Fx  x , y , z −  = σ xz  x , y , z −  ∆x ∆y
 2  2 

 ∆z   ∆z 
Fx  x , y , z +  = σ xz  x , y , z +  ∆x ∆y .
 2   2 

100 ChApTER 4: Fluid Mechanics Primer

As in Section 3.2, we express each of the forces as a Taylor series expansion

about the center of the fluid element (0, 0, 0), and ignore second-order and
higher terms

 ∆x   ∂σ xx ∆x 
Fx  x − , y , z  ≈  σ xx ( x , y , z ) − ∆y ∆z
 2   ∂y 2 

 ∆x   ∂σ xx ∆x 
Fx  x + , y , z  ≈  σ xx ( x , y , z ) + ∆y ∆z
 2   ∂y 2 

 ∆y   ∂σ xy ∆y 
Fx  x , y − , z  ≈  σ xy ( x , y , z ) − ∆x ∆z
 2   ∂y 2 
(4.24)
 ∆y   ∂σ xy ∆y 
Fx  x , y + , z ≈  σ xy ( x , y , z ) + ∆x ∆z
 2   ∂y 2 

 ∆z   ∂σ xz ∆z 
Fx  x , y , z −  ≈ σ xz ( x , y , z ) − ∆x ∆y
 2   ∂y 2 

 ∆z   ∂σ xz ∆z 
Fx  x , y , z +  ≈  σ xz ( x , y , z ) + ∆x ∆y .
 2  ∂y 2 

If we let Fxext be the sum of the forces in the x-direction,

 ∆x   ∆x   ∆y 
Fxext = − Fx  x − , y , z  + Fx  x + , y , z  − Fx  x , y − , z
 2   2   2 
(4.25)
 ∆y   ∆z   ∆z 
+ Fx  x , y + , z  − Fx  x , y , z −  + Fx  x , y , z + 
 2   2 2 

Figure 4.11 Schematic depicting dz

external forces acting on faces of Fx ⎛⎝x, y, z + — ⎠⎞
2
the fluid element that is immersed
in the fluid and moving at the same
local velocity. In other words, the
boundaries follow the motion of the
fluid. The surface forces arise from dx
Fx ⎛⎝x – —, y, z⎠⎞
stresses imparted on the control volume 2
by the surrounding fluid.
dy
Fx ⎛⎝x, y – —, z⎠⎞
2

dz dx
dx
Fx ⎛x + —, y, z⎞ Fx ⎛x, y + —, z⎞
⎝ 2 ⎠
⎝ 2 ⎠

dx

dy
z dz
Fx ⎛x, y, z – —⎞
⎝ 2⎠
y
x
 THe NaVieR–StoKes EQuations 101

and following the substitution of the expressions in Equation 4.24 into 4.25,

 ∂σ ∂σ yx ∂σ zx 
Fxext =  xx + + ∆x ∆y ∆z. (4.26)
 ∂ x ∂y ∂z 

The volume is also subject to a body force. If fx is the body force in the x-direction
per unit mass, and ρ is the mass density of the fluid, then the total body force is the
product of fx and the mass of the fluid in the control volume,

Fxbody = f xρ∆x ∆y ∆z. (4.27)

Now that we have expressions for the external forces acting on the control volume,
we are able to invoke Newton’s second law (F = ma). Our fluid element is moving
in the fluid with velocity u(x, y, z, t). Because u is dependent on both time and
space, the velocity of the fluid element can be changing due to two distinct phe-
nomena; changes in the flow velocity with time, or changes in the flow field with
space. Consider the case in which the flow is spatially uniform (u is identical at all
points in space), but increasing with time. The fluid element is accelerating solely
due to the increase in velocity with time. Now consider the case in which u does
not change with time (it is steady), but the flow field is spatially heterogeneous. The
fluid element will undergo changes in velocity due to spatial alterations in the flow
profile. As an example, refer to Figure 4.7—there, the flow can be steady (time-
invariant), but as you move from the entrance to the exit, the fluid must accelerate
to satisfy mass conservation. This spatial change in velocity is sometimes referred
to as convective acceleration.
To account for the temporal and spatial dependence of u in calculating the accel-
eration of the fluid element, we use the chain rule. In particular, if a = {ax, ay, az}T
is the acceleration of the fluid element, then ax can be calculated as
du( x , y , z , t ) ∂u ∂u ∂x ∂u ∂y ∂u ∂z
ax = = + + + . (4.28)
dt ∂t ∂x ∂t ∂y ∂t ∂z ∂t

The partial derivatives ∂x/∂t, ∂y/∂t, and ∂z/∂t give the instantaneous change in
position of the fluid element with respect to time in the x-, y-, and z-directions,
respectively. Because the fluid element is moving with velocity vector u, then
these expressions are equivalent to the velocity of the fluid,
∂x ∂y ∂z
u= , v= , and w = , (4.29)
∂t ∂t ∂t

and Equation 4.28 can be rewritten as

∂u ∂u ∂u ∂u
ax = +u +v +w . (4.30)
∂t ∂x ∂y ∂z

With the external forces and fluid element acceleration in hand, we now apply
Newton’s second law. The mass of the element is
m = ρ∆x ∆y ∆z. (4.31)

Setting the sum of forces in Equations 4.26 and 4.27 equal to the product of the
mass in Equation 4.31 and acceleration in Equation 4.30 we find

 ∂σ xx ∂σ yx ∂σ zx 
 ∂x + ∂y + ∂z  ∆x ∆y ∆z + f xρ∆x ∆y ∆z
(4.32)
 ∂v ∂v ∂v ∂v 
= ρ∆x ∆y ∆z  +u +v +w ,
 ∂t ∂x ∂y ∂z 

102 ChApTER 4: Fluid Mechanics Primer

which simplifies to

 ∂σ xx ∂σ yx ∂σ zx   ∂v ∂v ∂v ∂v 
 ∂x + ∂y + ∂z  + ρf x = ρ  ∂t + u ∂x + v ∂y + w ∂z  (4.33)

when we divide out the volume of the control volume, ΔxΔyΔz. Repeating these
steps for the forces and accelerations in the y- and z-directions,

 ∂σ xy ∂σ yy ∂σ zy   ∂v ∂v ∂v ∂v 
 ∂x + ∂y + ∂z  + ρf y = ρ  ∂t + u ∂x + v ∂y + w ∂z  (4.34)

 ∂σ xz ∂σ yz ∂σ zz   ∂w ∂w ∂w ∂w 
 ∂x + ∂y + ∂z  + ρf z = ρ  ∂t + u ∂x + v ∂y + w ∂z  . (4.35)

Equations 4.33, 4.34, and 4.35 are called Navier’s equations, named after Claude-
Louis Navier.

Constitutive relations and the continuity equation are

necessary to make Navier’s equations solvable
Upon inspection of Navier’s equations, we see that there are more unknowns (six
independent stress components and three velocity components) than equations
(three), which means they cannot be solved unless further relations are specified.
To make these equations solvable, George Gabriel Stokes proposed a set of consti-
tutive relations that related stress to fluid velocity, viscosity, and pressure. These
are a more general form of the constitutive relation we introduced in Section 4.2
for a Newtonian fluid. Specifically,

 ∂u ∂v 
σ xy = σ yx = µ  + (4.36)
 ∂y ∂x 

 ∂v ∂w 
σ yz = σ zy = µ  + (4.37)
 ∂z ∂y 

 ∂u ∂w 
σ xz = σ zx = µ  +
 ∂z ∂x 
(4.38)

2  ∂u ∂v ∂w  ∂u
σ xx = − P − µ + + + 2µ (4.39)
3  ∂x ∂y ∂z  ∂x

2  ∂u ∂v ∂w  ∂v
σ yy = − P − µ + +  + 2µ (4.40)
3  ∂x ∂y ∂z  ∂y

2  ∂u ∂v ∂w  ∂w
σ zz = − P − µ + +  + 2µ . (4.41)
3  ∂x ∂y ∂z  ∂z

The above six equations, together with Navier’s equations, give nine equations in
total. They also introduce another unknown, pressure, making 10 unknowns in
total. We still require one more equation to make the system solvable. Because we
 RHEOLOGICAL ANALYSIS 103

have used the equilibrium and constitutive relations, you may not be surprised Nota Bene
that what is missing is a statement of kinematics. Specifically, this is known as the
continuity equation and is a mathematical statement of the conservation of mass. Continuity equation. A simple
though somewhat unrigorous
As we ­discussed previously, for incompressible fluids, the flow rate into a fixed
analysis can be used to demonstrate
volume must be equal to the flow rate out of it, the relation between Equations 4.43
and 4.44. Consider a small, imaginary
Vin Ain = Vout Aout . (4.42) cubical volume immersed in a flow
field. The volume is aligned with the
If Ain = Aout = A, then Equation 4.42 can be rewritten as coordinate system and with
dimensions Δx, Δy, and Δz in the x-,
(∆V ) A = 0, (4.43) y-, and z-directions. The volume is
assumed to be fixed in space, with
fluid entering it at some velocity, and
where ΔV = Vout − Vin. A differential form of Equation 4.43 is leaving it at a different velocity. Let
Δu, Δv, and Δw be the changes in fluid
∂u ∂v ∂w velocity in the x-, y-, and z-directions
+ + = 0. (4.44) that occur within the volume. From
∂x ∂y ∂z Equation 4.42, we know that

∆u∆y ∆z + ∆v ∆x ∆z + ∆w ∆x ∆y = 0.
Equation 4.44 is called the differential form of the continuity equation for incom-
pressible fluids. (4.45)
Dividing Equation 4.45 by the volume
of ΔxΔyΔz, and letting the volume go
Navier–Stokes equations: putting it all together to zero, we obtain Equation 4.44.
With Navier’s equations (three total), the constitutive relations proposed by Stokes
(six total), and the continuity equation, we have 10 equations and 10 unknowns
(u, v, w, P, and the six independent components of stress). Although this is a solv-
able system of equations, it can be greatly simplified by substituting in the rela-
tions for stress in Equations 4.36–4.41 into the stress derivatives in Equations
4.33–4.35. Taking the term in parentheses in the left-hand side of Equation 4.33,
we can substitute in Equations 4.36, 4.38, 4.39, and after some manipulation and
invoking Equation 4.44, we arrive at the following expression

∂P  ∂ 2u ∂ 2u ∂ 2u   ∂u ∂u ∂u ∂u 
− + µ  2 + 2 + 2  + Pf x = P  +u +v + w . (4.46)
∂x  ∂ x ∂ y ∂ z   ∂ t ∂ x ∂ y ∂z 

It can be similarly be shown that

∂P  ∂ 2v ∂ 2v ∂ 2v   ∂v ∂v ∂v ∂v 
− + µ  2 + 2 + 2  + Pf y = P  +u +v +w  (4.47)
∂y  ∂x ∂y ∂z   ∂t ∂x ∂y ∂z 

∂P  ∂ 2w ∂ 2w ∂ 2w   ∂w ∂w ∂w ∂w 
− + µ  2 + 2 + 2  + Pf z = P  +u +v +w . (4.48)
∂z  ∂x ∂y ∂z   ∂t ∂x ∂y ∂z 

Equations 4.47 and 4.48 are the Navier–Stokes equations. Together with the conti-
nuity equation, they form a set of four equations to solve for the four unknowns
u, v, w, and P.

4.4 ​RHEOLOGICAL ANALYSIS

So far, we have discussed the mechanics of two fundamentally different materials
elastic solids (Section 3.2) and viscous materials (Section 4.3). Although we have
discussed these two types of material in isolation, many materials cannot be
characterized as being, purely elastic or viscous, as they may exhibit both solid-
or fluidlike behavior, depending on the circumstances. For instance, like a solid,
a dab of toothpaste can hold its shape under its own weight. However, we can still
squeeze it out of the tube because, like a fluid, toothpaste has little capacity to
104 ChApTER 4: Fluid Mechanics Primer

Example 4.2: Fluid flow within parallel plates

In Section 4.2, we derived the flow profile for steady, ∂P ∂ 2u
incompressible, fully–developed laminar flow between = µ 2.
∂x ∂y
two infinite parallel plates, driven by pressure. We now
seek to perform the same calculation using the Navier– To solve for u( y), we integrate u with respect to y twice
Stokes equations. Consider plates separated by height and obtain
h, with one plate at y = 0, and another at y = h. Calculate
u( y) assuming a pressure gradient in the x-direction of 1 ∂P y 2
u( y ) = + Ay + B.
∂P/∂x, with no body forces present. µ ∂x 2
To calculate u, we begin with Equation 4.46. We can
We can solve for A and B by requiring that u = 0 at y = 0
simplify this expression in several ways. First, we can
and y = h, the no-slip condition. Doing so, we arrive at
set ∂u/∂t = 0 because the flow is steady. Second, there
our final solution, which is identical to that achieved via
should be no flow in the y- and z-directions, thus we
differential analysis
can set v = 0 and w = 0. Third, because the flow is fully
developed, u does not depend on x, and we can set any 1 ∂P 2
x derivatives equal to zero. Finally, because we assume u( y ) = ( y − hy ).
2 µ ∂x
no body forces, then f x = 0, and Equation 4.46 simpli-
fies to

sustain shear or recover from deformations. Within the body, tissues and organs,
as well as the cells residing within them, may be composed of both fluid and
­solidlike materials. The cytoplasm of a given cell may contain a solidlike cytoskel-
etal network immersed in an aqueous environment in which numerous proteins
are densely dispersed. In this case, one can see why cells would be expected to
exhibit both solid and fluidlike mechanical behavior.
Rheology is a scientific discipline that can be broadly characterized as the study of
materials that have some capacity to flow, but which cannot be adequately
described using classical fluid mechanics. Rheology is considered a distinct
branch of continuum mechanics that bridges solid and fluid mechanics. The need
for such a discipline can be better understood by revisiting our examples of non-
Newtonian fluids; power-law fluids and Bingham plastics. Although both models
are able to capture some aspects of nonlinear fluid behaviors, they do not contain
any fundamental solidlike behaviors. No matter how you adjust the exponent in a
power-law fluid, it will always continuously deform under shear. A Bingham plas-
tic is able to resist shear stress if it is below a critical level, but here the material is
completely rigid and does not exhibit elastic behavior. Rheological methods allow
us to better understand materials that exhibit solid and fluidlike behavior, such as
cells (Figure 4.12). A subset of rheology is the study of viscoelasticity, in which
one seeks to decompose mechanical behavior into purely elastic and purely vis-
cous components. We will discuss some basic rheological approaches for investi-
gating viscoelastic substances.

The mechanical behavior of viscoelastic materials can be

decomposed into elastic and viscous components
Because a viscoelastic substance is one that exhibits both elastic and viscous
mechanical behavior, studies of such substances often rely on oscillatory stimuli.
Recall from Section 3.2, we introduced the notion of a linearly elastic material.
Specifically, when subjected to stress, such a material would undergo strain in a
manner proportional to stress, and would recover completely upon removal of the
load. So, consider a material that is exposed to an oscillatory stress of the form

σ = σ 0 cos(ωt ), (4.49)
 RHEOLOGICAL ANALYSIS 105

Figure 4.12 Silly putty is a

substance that exhibits both elastic
and fluid responses. It can clearly
hold its shape, as seen on the left where
it is molded into a rectangular solid.
However, within 30 min it has flowed
under gravity, forming a “puddle” on the
table, a result of its fluidlike behavior.
Characterizing it as a non-Newtonian
fluid is somewhat limiting. If one were to
mold this putty into a sphere, one could
bounce it like a ball, which requires
elastic behavior.

where σ0 is the magnitude of the stress, and ω is the frequency of oscillatory load-
ing (in radians per unit time). If the material is linearly elastic, it would deform
proportionally to the stress as

ε = A cos(ω t ), (4.50)

where A is a constant. Unlike an elastic material, the stress in a purely viscous
material is not dependent on strain, but on strain rate. This is similar in concept
to a Newtonian fluid, in which shear stress is proportional to shear rate. For the
oscillatory stress described in Equation 4.49, a purely viscous material would
deform such that the time derivative of strain is proportional to stress,

ε = B sin(ω t ). (4.51)

Notice that the strain profile of a viscous material given by Equation 4.51 can be
rewritten as

ε = B cos(ω t − π /2), (4.52)

indicating that the strain in Equation 4.52 is exactly π/2 radians, or 90 degrees, out
of phase with the stress. Here the strain is said to be completely out of phase with
the stress. Materials that exhibit some combination of elastic and viscous behav-
ior will exhibit a phase shift that is between 0 and π/2 radians. In particular, if

ε = ε 0 cos(ω t − δ ) (4.53)

is the strain of the material, then the phase shift is δ radians. The parameter δ is
called the phase lag (Figure 4.13). The phase lag is useful because it gives a ­single

Figure 4.13 If the stimulus applied

to a material is a pure sine wave,
then the response of a viscoelastic
material, in blue, will typically
exhibit a phase lag, δ, which can be
response

used as a measure of how solid- or

fluidlike the material is.

δ
106 ChApTER 4: Fluid Mechanics Primer

value that characterizes the degree to which the mechanical behavior is elastic
relative to that which is viscous. If the phase lag is close to zero, then the material
is behaving primarily like an elastic material. If the phase lag is close to π/2, then
the material is behaving primarily like a viscous material.
Next, consider the case in which we subject a material to oscillatory loading and
obtain the strain profile given in Equation 4.53. We can decompose this profile
into purely in-phase and out-of-phase components. First, we rewrite Equation
4.53 as a sum of cosine and sine functions using the identity

cos(u + v ) = cos u cosv − sin u sin v. (4.54)

Equation 4.54 can then be rewritten as

ε = ε 0′ cos(ω t ) − ε 0″ sin(ω t ) , (4.55)

where

ε 0′ ε 0 cos(δ ), (4.56)
and

ε 0″ = −ε 0 sin(δ ). (4.57)

Equations 4.55, 4.56, and 4.57 demonstrate that the mechanical response given by
Equation 4.53 can be decomposed into two components, one that is in phase (the
first term, with cos(ωt)) with the driving stress, and one that is out of phase (the
second term, with sin(ωt)). The relative magnitudes of the in-phase and out-of-
phase terms are related to the magnitude of the phase lag.

Complex moduli can be defined for viscoelastic materials

Equation 4.55 indicates that, given a strain profile for an oscillatory loaded mate-
rial, we can decompose this strain into in-phase and out-of-phase components
and identify the degree to which the material is deforming like an elastic versus a
viscous material. However, such a decomposition would not necessarily allow us
to predict how the material would deform if subjected to a different oscillatory
stress profile, such as with a different frequency or magnitude. To make such pre-
dictions, we require some information regarding the material properties. In
Section 3.2, we described the concept of an elastic modulus, which was a material
property that related stress to strain for a linearly elastic material. We introduced
a similar relationship earlier in this chapter for fluids, using viscosity. We now
seek to define an analogous “modulus” for viscoelastic materials. This task is not
as simple as it was for the pure elastic or pure fluid cases. In particular, for a line-
arly elastic material, the ratio between stress and strain is fixed, and we could sim-
ply define the elastic modulus as the ratio of the two quantities. For our viscoelastic
material, the ratio between stress and strain changes with time, and we are unable
to define a modulus in the same way.
A solution to this problem is to use complex numbers and define quantities called
the complex stress, complex strain, and complex modulus. The reason why com-
plex numbers provide a nice solution to our dilemma is that functions involving
sines and cosines can be dealt with in an elegant fashion through the use of Euler’s
formula. In particular, let ℘ be a complex function of time,

℘ = cos(ω t ) + i sin(ω t ). (4.58)

In the above expression, the real part of ℘ is

Re{℘} = cos(ω t ), (4.59)

 RHEOLOGICAL ANALYSIS 107

the imaginary part is

Im{℘} = sin(ω t ), (4.60)

and i is the imaginary unit satisfying i 2 = −1. We can rewrite ℘ using Euler’s for-
mula, which states that complex functions like Equation 4.58 can be expressed as
an exponential imaginary function, Nota Bene

e ix = cos( x ) + i sin( x ). (4.61) Euler’s Formula. Leonhard Euler

published this formula in the
mid-1700s. Euler’s formula was
Using Euler’s formula, ℘ can be written as
described by Richard Feynman as
“one of the most remarkable, almost
℘ = cos(ω t ) + i sin(ω t ) = e iωt . (4.62) astounding, formulas of all

mathematics.”
Given the relation in Equation 4.62, we can now define a complex stress and
­complex strain. Let the complex stress be defined as

σ * = σ 0 cos(ω t ) + iσ 0 sin(ω t ) = σ 0e iω t . (4.63)

We have intentionally defined the complex stress such that the real part of σ* is the
stress σ applied to the material as in Equation 4.59,

Re {σ *} = σ 0 cos(ω t ). (4.64)

Similarly, the complex strain is defined as

ε * = ε 0 cos(ω t − δ ) + iε 0 sin(ω t − δ) = ε 0e i (ω t −δ ) , (4.65)

where again, we have intentionally defined the complex strain such that the real
part of e* is the observed strain in Equation 4.54,

Re {ε *} = ε 0 cos(ω t − δ ). (4.66)

With the complex stress and complex strain defined, we can now define a complex
modulus as the ratio of these two quantities,
σ*
E* = . (4.67)
ε*

Substituting expressions for the complex stress and strain in Equations 4.63 and
4.65 into Equation 4.67, and following some simplification, we arrive at a compact
expression for the complex modulus,

σ 0e iω t σ 0e iω t σ
E* = i (ω t −δ )
= = 0 e iδ . (4.68)
ε 0e ε 0e iω t e −iδ ε0

Because E * is a complex number, we can define E′ as the real part of E * and E″ as
the imaginary part, in other words

E * = E ′ + iE ″ (4.69)

with

σ0
E′ = cos(δ ) (4.70)
ε0
108 ChApTER 4: Fluid Mechanics Primer

and

σ0
E″ = sin(δ ). (4.71)
ε0

E ′ is known as the elastic modulus or storage modulus. The storage modulus E ′ is
associated with the in-phase component of resistance to stress. This becomes
apparent by substituting δ = 0 into Equations 4.69, 4.70, and 4.71. Here, the com-
plex modulus is equivalent to the storage modulus, E * = E ′. Because in-phase
deformation is a characteristic of elastic materials, the storage modulus can be
considered a measure of the elastic behavior of the material. E ″ is referred to as
the damping or loss modulus, and is associated with the out-of-phase resistance
to stress. For instance, for δ = π/2, E * = iE ″. In this case, the magnitude of the
complex modulus is equal to the loss modulus. Because this phase lag is associ-
ated with viscous materials, the loss modulus can be considered a measure of the
viscous behavior of the material.
Nota Bene Before concluding this section, it is worthwhile to note that in our development,
Historical roots of complex we designated the complex modulus as E *, suggesting a modulus analogous to
modulus. The complex modulus was Young’s modulus. There is a corresponding complex shear modulus G *, which is
introduced by the German physicist actually more common in the literature because dynamic rheological measure-
Carl Gauss in the early 1800s—the ments are often made under shear. For the rest of the chapter, we will use the com-
same Gauss for whom the SI unit of
plex shear modulus G * and associated complex shear stress τ * and complex shear
magnetic strength was named.
strain γ * in our developments.

Power laws can be used to model frequency-dependent

changes in storage and loss moduli
In the last section, we saw that for a viscoelastic material subject to oscillatory
loading, the storage modulus and loss modulus capture the in-phase versus out-
of-phase resistance to stress, respectively. For many viscoelastic materials, these
moduli depend on the frequency of loading. Furthermore, the frequency depend-
ence for each modulus may be different, such that the relative amount of in-phase
versus out-of-phase deformation changes at different frequencies. Such frequency
dependence is common in many biological materials, including cells. A variety of
models have been proposed to capture this frequency dependence. One such
approach involves modeling the frequency dependence of the storage and loss
moduli as power laws. For instance, the following relation has been demonstrated
to accurately describe the stiffening of cells subjected to oscillatory loading over
several orders of magnitudes

Example 4.3: Calculating the behavior of semi-fluid substances

Let us say you stimulate a cell with a pure sine wave via and G ″ = G0 sin(δ). Thus, the complex shear modulus has
shear, and the shear strain response is shifted exactly 45 equal contributions from both the storage modulus (G ′)
degrees (π/4 radians). Does it have exactly equal contri- and the loss modulus (G ″), but these moduli are individu-
butions from solid and fluid behavior? If so, does this ally more than half the value of G0, which would be the
mean that the storage loss moduli is half what the elastic elastic modulus were the cell to be a pure solid. The rea-
modulus would be if the cell were a pure solid with the son for this is that G * has magnitude of G0, so that the indi-
same response amplitude, but no phase shift? vidual components must be 2 2 times G0. That is, a
substance that is half solid and half-fluid has more than
Our input is τ = τ0sin(ωt), and the output is γ = γ0
half the elastic modulus of a pure solid responding the
sin(ωt − π/4). Our δ is therefore π/4, and we can write the
same way.
complex shear modulus G * = G ′ + iG ″ with G ′ = G0cos(δ)
 RHEOLOGICAL ANALYSIS 109

Advanced Material: Complex viscosity

Just as we defined a complex modulus, we can also Comparing the expressions for Equations 4.75 and 4.76, it
define a complex viscosity. More specifically, the com- becomes apparent that Equation 4.75 can be rewritten as
plex viscosity can be defined as the ratio of the complex
shear stress and complex shear strain rate as G * G + iG ′ iG ′ G ″
µ* = = = + . (4.77)
iω iω ω ω 
τ*
µ* = , (4.72) Letting µ* = µ′ + iµ″ , then
γ*

G″
µ′ = (4.78)
where ω 

τ * = τ 0e iω t  (4.73) and
G′
and µ″ = , (4.79)
ω 
dγ ∗ d where μ′ is the dynamic viscosity, and μ″ is the “out-of-
γ∗ = = γ 0e i (ω t −δ ) = γ 0iωe i (ω t −δ ) . (4.74)
dt dt  phase” viscosity that yields the change of the storage
modulus with respect to frequency. Upon inspection of
Equations 4.78, and 4.79, one might see that the complex
Substituting Equations 4.73 and 4.74 into 4.72, we viscosity is closely related to the complex shear modu-
obtain lus. In fact, μ′ and μ″ are the loss modulus and storage
modulus normalized by frequency, respectively. This
τ 0e iω t τ fact allows one to use these parameters to better under-
µ* = = 0 e iδ . (4.75)
γ 0iωe i (ω t −δ ) γ 0iω stand how the storage and loss moduli change with fre-
 quency. For instance, assume μ′ and μ″ are relatively
constant over some frequency range, and μ′ > μ″.
Similar to Equation 4.68, the complex shear modulus Increasing the loading frequency within this range
can be computed as would result in greater increases in G ″ relative to that in
G ′, and the material would be expected to exhibit more
τ 0 iδ
G* = e . (4.76) viscous behavior and less elastic behavior with increas-
γ0  ing frequency.

α
ω  πα 
G * (ω ) = G0   (1 + iξ )Γ(1 − α )cos  + i µω (4.80)
Φ   2 

where

 πα 
ξ = tan  (4.81)
 2 

and

Γ(n) = (n − 1)!, (4.82)

Go is a parameter that gives the frequency-independent component of the elastic

response, Φ is a normalization factor, ξ is a structural damping coefficient, μ is a
viscous coefficient, and α is the scaling exponent. The real part of G * is the stor-
age modulus G ′, and the imaginary part is the loss modulus G ″. The coefficients
μ and ξ represent distinct viscous components; the term iμω is meant to ensure
110 ChApTER 4: Fluid Mechanics Primer

10 5 that regardless of the scaling exponent α, high-frequency stimuli will result in

viscous effects dominating the behavior. The ξ term is much larger than the μ
10 4 G′ term for most relevant frequency ranges, and in most cases is the primary con-
(Pa)

tributor to the loss modulus. The scaling exponent determines not only how G ′
10 3 and G ″ change with frequency but the relative magnitudes of each modulus. If
G″ we assume that μ is small, we can see that as α approaches zero, ξ approaches
10 2 zero as well, and G * approaches G0, which gives the elastic component of the
10 –2 10 0 10 2 response. However, as α approaches one, ξ increases without bound, indicating
f (Hz) that the imaginary term will dominate and the material will exhibit strongly vis-
Figure 4.14 Storage and loss moduli cous effects.
as a function of frequency in cells Experiments in a variety of cell types and using various techniques for mechanical
subjected to oscillatory mechanical
loading have demonstrated that cells exhibit a scaling exponent of around 0.2–0.3
loading. Because the results are plotted
on a log–log scale, the slope of G′ gives (Figure 4.14). This power-law dependence is typical of a class of materials called
the scaling exponent α, which is soft glassy materials, which includes materials such as emulsions, slurries, and
approximately 0.2 in this case. (Adapted pastes. Soft glassy materials are characterized by their possession of some degree
from, Fabry et al., [2011] Phys. Rev. Lett. of disorder in which the discrete elements of which they are composed are entan-
87, 148102.) gled or aggregated via weak interactions. Therefore cells have often been described
in the literature as soft glassy materials.

Nota Bene 4.5 ​DIMENSIONAL ANALYSIS

Order-of-magnitude analysis.
Another analytical approach that can Within fluid mechanics, most flows cannot be determined analytically—that is,
reveal quantitative information in the the flow profile cannot be derived from first principles alone, and one must turn
absence of an exact solution is to experiments to investigate the flow of interest. In such situations, proper exper-
order-of-magnitude analysis. With this imental design may be difficult if one does not have a priori knowledge of relevant
strategy, instead of trying to find parameters that affect a parameter of interest. Dimensional analysis is a mathe-
precise functional relationships, we
estimate how parameters might be
matical technique used to collapse a set of experimental parameters into a
related within an order of magnitude reduced set of dimensionless quantities that influence a parameter of interest. In
or two. As an example, say you are dimensional analysis, the goal is not to obtain an exact formula for the parameter
designing a skyscraper and need an in question. Though dimensional analysis does not provide an exact analytical
estimate for the predicted weight of solution, it provides useful quantitative relationships that often allow one to draw
the building. Assume that the building valuable conclusions.
will be approximately 100 stories high
and made of steel. To within one
order of magnitude, one could Dimensional analysis requires the determination of base
estimate the square footage for each parameters
floor as 100 m × 100 m, and the
height of a single story as 1 m, which Consider the case where we wish to perform an experiment to determine the fluid
results in a volume of 100,000 m3.
in which drag forces on a swimming bacterium. We are interested in dimension-
One could also estimate the average
density of the building as 1 ton/m3,
less quantities that may affect the fluid drag force, as this will aid in the experi-
based on the fact that a cubic meter mental design.
of water weighs 1000 kg, which is
To begin, we must identify all the potential factors affecting our parameter of
approximately 1 ton, and although
steel is denser than water, much of interest, drag force, and specify their units. To establish this list of base parame-
the inside of the building is air. An ters, we require some level of intuition for the problem. Because we are interested
order-of-magnitude estimate for the in the fluid drag force on a swimming bacterium, one might make an educated
weight is 100,000 tons. This turns out guess that this force would depend on parameters such as the density of the fluid
to be a good estimate, as the Empire (units of kg/m3), the viscosity of the fluid (units of kg/ms), the velocity of the bac-
State Building has 102 floors and terium (units of m/s), and some characteristic length scale of the bacterium (units
weighs 365,000 tons; the Sears Tower
is 108 floors and weighs 223,000 tons.
of m). Including drag force (units of kg ⋅ m/s2), we have five parameters that we
  Similar calculations can be done consider to be directly relevant to the modeling of the problem, the base param-
for estimating the number of eters (Figure 4.15). We can express the fact that we expect drag force to be depend-
receptors in a cell for signaling, the ent on the selected parameters as
speed of molecular signals within
cells, and some properties of cells f ( F , L ,µ ,ρ,v ) = 0, (4.83)

that are otherwise difficult to
measure.
where f represents one or more as yet unknown function(s).
 DIMENSIONAL ANALYSIS 111

Figure 4.15 ​A bacterium swimming

μ, ρ in a fluid at velocity V (moving to
the left) in a fluid of density ρ and
viscosity μ with length L
experiences some drag slowing it
down. We wish to characterize the drag
V force, but because the geometry is odd,
it is difficult to obtain an exact analytic
solution. However, we can obtain some
basic relationships using dimensional
analysis.

The Buckingham Pi Theorem gives the number of dimensionless Nota Bene

parameters that can be formed from base parameters Roots of the Buckingham Pi
Theorem. The Buckingham Pi
With our list of base parameters established, we may now use them to form Theorem is a formalization of a
dimensionless parameters. The number of dimensionless parameters that we can nondimensionalization method
form from our base parameters is given by the Buckingham Pi Theorem, which introduced by Rayleigh. Edgar
states that the number of dimensionless parameters is equal to the number of Buckingham proposed the theorem
base parameters minus the number of independent physical units (dimensions). in 1914 and he termed each of the
In our example, we have five base parameters and three independent units nondimensional parameters Pi (π1, π2,
. . ., πp), showing that if j is the
(length, speed, and mass), leaving two dimensionless parameters to be found.
number of base parameters and k is
Bear in mind that an independent unit in this case does not have to be a funda- the number of dimensions
mental unit such as (m), (s), or (kg). They can be derived units such as (m/s), so (independent units), p = j − k. The
long as they are independent. Using velocity (m/s) and length (m) would be a theorem also shows that the general
valid alternative to using length (m) and time (s). One would not use velocity, length, form for each nondimensional
and time, as they are not independent. Similarly, you would not use area (m2) and parameter is the product of base
length (m) since one is a power of another and therefore not independent. parameters with each raised to an
unknown integer power. The use of
Pi denoted that the nondimensional
Dimensionless parameters can be found through solving parameters were products of base
a system of equations parameters.
  It should be noted that determining
At this point in the analysis, we have identified our base parameters and dimen- the number of Pi groups is a bit more
involved than we are describing here,
sions (units). The Buckingham Pi Theorem tells us how many dimensionless but the method used here is
parameters we are looking for and the form they will take. The next step is actually applicable to enough situations as a
to identify the dimensionless parameters. Although a hard and fast procedure is very good rule of thumb.
elusive, there is a strategic outline that is effective. First, identify j base parameters
(generally independent) that span the dimensions. In other words, these base
parameters should involve all of the physical units and will be termed repeating
parameters. The procedure is simply to combine the remaining parameters with
the repeating parameters one at a time.
In our bacterium example, we have identified five base parameters F, L, ρ, μ, and
v and three units, so we expect to find two dimensionless parameters. Because
j = 3, we need to identify three repeating parameters that should generally be sim-
ple and independent. Let us start with length and velocity. They are simple and
span two of our units (m and s). For our third repeating parameter we need to add
kg. Density or viscosity would both be acceptable, and in fact, both produce the
same dimensionless groups. Here we arbitrarily select viscosity, μ.
Now we need to combine each of our remaining parameters with the repeating
parameters one at a time. Starting with force, we want to develop a product of the
dependent parameters F, L, μ, and v such that the end expression has no units.
That is, for the product F aLbμcv d, we wish to find the values for the exponents a, b,
c, and d such that the product is unitless, or
112 ChApTER 4: Fluid Mechanics Primer

F a Lb µ cv d = 1. (4.84)

The “1” on the right side of Equation 4.84 does not refer to its numerical value, but
to the fact that the product has no units. To do this, we decompose each of these
parameters in Equation 4.84 into their basic units and then pick the exponents so
that they all cancel out. If we replace the parameters in Equation 4.84 with their
basic units, then
a c d
 kg ⋅ m  b  kg   m 
 2  (m )     = 1. (4.85)
s m ⋅ s  s 

For this relation to hold, each unit must have a final exponent of zero. If we con-
sider the exponents for length (m), we find that the following relation must be true:

a + b − c + d = 0. (4.86)

Similarly, for mass

a + c = 0, (4.87)

and for time

−2a − b − d = 0. (4.88)

Equations 4.86, 4.87, and 4.88 form a system of simultaneous algebraic equations.
The system can be simplified to yield

a = −b

a = −c (4.89)

a = −d.

This system has three equations and four unknowns, so it is indeterminate. We

can arbitrarily set a, the exponent for force, equal to 1 (a general rule of thumb is
to choose small whole numbers for one exponent). In this case, the exponents for
the rest of the parameters are equal to −1. We have arrived at our first dimension-
less term,
F
F 1µ −1L−1v −1 = .
Lµv (4.90)

Next, we examine density, ρ, our second remaining parameter. We need to com-

bine it with L, μ, and v, and do a similar analysis, in which

e g h
 kg  f  kg   m 
 3  (m )     = 1. (4.91)
m m ⋅ s  s 

Following the same steps as before, we obtain our second dimensionless term,

ρLv
ρ1µ −1L1v1 = . (4.92)
µ

The second term is the Reynolds number, which we previously discussed.

Equation 4.84 can be recast as a function of dimensionless quantities as

 F ρLv 
f = 0, (4.93)
 Lµv ′ µ 

 DIMENSIONAL ANALYSIS 113

which is a simpler functional form because it has fewer terms compared to the
original expression.

Similitude is a practical use of dimensional analysis

Now that we have derived our dimensional parameters, we can use them in a
variety of ways to facilitate experimental design or interpretation of results.
Dimensional analysis is useful from a modeling perspective, because it informs
us not only about the number of parameters we need to pay attention to (the
dimensionless numbers), but also the relationship between the variables that
appear in those parameters. This allows us greater flexibility in creating experi-
mental simulations.
Suppose we wish to measure drag force on our bacterium by building a large-
scale model and submerging it in a defined flow. We do this because it is easier to
handle larger objects (bacteria being submicrometer in size) and because it is dif-
ficult to measure such small forces (both in terms of the magnitude of the forces
and in terms of how we can attach instruments to a bacterium). We seek to meas-
ure the drag force on a large-scale model and then scale the observed forces to
what we would expect at normal scale. This is known as a similitude experiment.
For the results to scale accurately, a strict requirement is that dimensionless
parameters be held constant.

Example 4.4: Deformation of a cell under shear

Assume that a cell is well anchored to its substrate and to b = − c − 2d = −1 + 2 = 1 using the second equation.
that fluid shear is deforming the cell. It can be treated as Finally, using the first equation, a + 1 − 1 + 1 = 0 means
an elastic solid for this problem. Determine the Pi groups a = −1. So our last Pi group is Vμ/hG. Our expression is
for estimating the angle of deformation of the cell, and then
estimate the shear strain.
f (γ , Re , A/h 2 ,V µ/hG ) = 0.
Base parameters are fluid properties—density ρ and vis-
cosity μ, the cell height h, the fluid velocity V, the cell
Note that instead, we can select another set of repeating
area exposed to shear A, the cell shear modulus G, and
parameters. We can select, for example, shear modulus,
the shear strain, which is what we want to find.
height, and density—remember, we need kg, m, and s all
Density is kg/m3, viscosity is kg/ms, height is m, area is represented. If we go through the algebra with these new
m2, velocity is m/s, and the shear modulus is N/m2 = kg/ independent parameters, we get the following Pi groups
ms2. As our repeating parameters, we pick height, vis- in our functional form
cosity, and velocity. The remaining parameters are den-
sity, area, shear strain, and shear modulus. Right away, F (γ ,V ρ/G , A/h 2 , µ/hρG ).
we recognize that the Reynolds number is one parame-
ter. Area is normalized easily by A/h2. The units of shear Here, the expression is much messier, and the functional
strain and radians are length/length or dimensionless. form looks different. Could we have made a mistake? The
However, the shear modulus takes some work. Let (ma) answer is that despite their different appearances, we
(mb/sb)(kgc/mcsc)(kgd/mds2d) = 1. The algebraic equa- can convert from one to the other. The shear strain, γ,
tions lead to and the A/h2 term remains the same. But what happens
when we multiply the second and last Pi groups in the
a+b−c −d = 0 bottom expression? We get Vμ/hG, which is just the last
Pi group of the first expression. Similarly, if we divide the
− b − c − 2d = 0 second Pi group by the last Pi group in the bottom
expression, we get ρhV/μ, which is Re, the second Pi
c + d = 0.
group in the first expression. So, they are equivalent, but
Let us pick c = 1. Note that we are starting, not with a, one was easier than the other. And it was not obvious in
but with c, so that it is easier to solve for the rest of the the beginning which parameters to select. Thus, there
variables. From c = 1, we get d = − 1. Then that leads can be an element of art involved.
114 ChApTER 4: Fluid Mechanics Primer

Suppose we wish to build a model of the bacterium 1000 times its normal size,
expose the model bacterium to fluid flow, and measure the drag force exerted
on the model. Recall that our second dimensionless term is ρLv/μ, which is the
Reynolds number. If we simply increase the length of the bacterium by a factor
of 1000, the Reynolds number will also increase by a factor of 1000. To scale the
experiment correctly and keep the Reynolds number constant, one could use a
fluid that has the same density but with 1000 times the viscosity and run the
bacterium model at the same velocity as the actual bacterium. We can use our
first dimensionless term to see how this would affect our measured force. In
particular, the first dimensionless parameter was F/Lμv, indicating that if L and
μ are scaled up by 1000, the drag force will be scaled up by a factor of
1000 × 1000 = 1 × 106. In this case, the measured drag must be scaled down by
the same amount—a million, not a thousand—to obtain the actual full-scale
drag acting on the bacterium.

Dimensional parameters can be used to check analytical

expressions
Another use of dimensionless quantities is that they can be used to check derived
analytical expressions for inconsistencies. Consider the case in which we use sim-
ple physical arguments to derive a scaling relationship for the drag force imparted
on a fixed object of cross-sectional area L2 subjected to a flow of velocity V. If we
assume viscous forces are small, we know from Bernoulli’s equation that for a
streamline which passes near the front of the object,
P ~ ρV 2 , (4.94)

where P is pressure, and we have assumed that the change in height is negligible.
The resultant force from this pressure would be expected to scale as the product of
pressure and area,
F ~ PL2 . (4.95)

Combining Equations 4.94 and 4.95, we obtain the scaling relationship
F ~ ρV 2 L2 . (4.96)

We now wish to check whether this relation is correct from a functional perspec-
tive by expressing the solution in terms of the dimensionless parameters we previ-
ously found. To do this, we rewrite Equation 4.96 as
 ρVL 
F ~ (µVL ) , (4.97)
Nota Bene  µ 

Scaling. Like dimensional analysis,
the goal in scaling analysis is to obtain where we have separated the V 2 term and inserted viscosity. Finally, dividing by
functional relationships that allow one the first term on the right-hand side we obtain
to see roughly how some dependent
parameter varies with other
independent parameters. In this case,
 F   ρVL  (4.98)
one may use a somewhat unrigorous
 µVL  ~  µ  ,
analytical approach (as we did in

deriving Equation 4.96). The goal is
not an exact formula, but rather a which implies that the relation in Equation 4.96 can be expressed as a function of
relation that describes how one our dimensionless parameters. Although this does not necessarily mean that
parameter changes with the rest. Equation 4.96 is correct, it gives confidence that the relation is accurate, since as
Here, scaling problems can be Equation 4.98 is consistent with what we expect (it turns out that Equation 4.98 is
considered a combination of indeed correct, as it gives the same scaling found in the so-called drag equation,
dimensional analysis and order-of- which gives the drag force experienced by objects caused by movement through
magnitude estimation. In scaling, as in
dimensional analysis, constants such
fluid). If the relation could not be expressed as a function of our dimensionless
as π or ½ are commonly dropped. quantities, that would imply there was an error in either the derivation of Equation
4.96 or our dimensionless parameters.
 Key Concepts 115

Example 4.5: Determining the Reynolds number using a scaling relationship from the
Navier–Stokes Equations
For a one-dimensional flow, we assume that the flows in form a ratio of inertial to viscous terms, we make the
y- and z-directions are negligible, and so are changes in quotient of the left side versus the right side
those directions. The Navier–Stokes equations in x sim-
plifies to (ρU 2 /L)/(µU /L2 ) = ρUL/µ,

ρ(∂u∂t + u∂u∂x ) = ρg − ∂P ∂x + µ∂ 2u∂x 2 . which is the Reynolds number. You can make similar
ratios using the time-varying term to get the Womersley
We are interested in the viscous versus inertial terms, so parameter, and other parameters based on gravity, pres-
we ignore the time-varying term (the first term), and the sure, etc. This also explains why in pipe flows, the transi-
gravitation and pressure terms. This simplifies to tional Reynolds number can be above 1; in fully
developed, steady, one-dimensional pipe flows, the left-
ρu∂u∂x = µ∂ 2u∂x 2 . hand side is actually zero, because the term ∂u∂x is zero
and there are technically no inertial terms at all. At the
Let us use U as a characteristic velocity and L as a char- microscopic scale, however, there are imperfections in
acteristic length scale. Then, ∂u∂x can be estimated as the pipe surface, which leads to non-zero v and w, as
U/L, and, ∂2u∂x2 can be estimated as (U/L)/L = U/L2. well as changing u. Those perturbations can (with
The inertial terms (associated with momentum) are on increased velocity) induce turbulence, but usually at Re
the left, which simplifies to rU 2/L, and the viscous terms much greater than 1, more typically around 1000–2000.
are on the right, which simplifies to μU/L2. If we want to

Key Concepts

• A fluid differs from a solid in that it takes on the • The Navier–Stokes equations results from combining
shape of the container in which it is placed. equilibrium (conservation of momentum), constitutive
• Hydrostatics is the study of pressure in a stationary (Newtonian behavior), and kinematic (compatibility
fluid, and exerts pressure equally in all directions. assumption) relationships. They describe the behavior
of a very wide class of problems in fluid mechanics.
• Newtonian fluids are those in which the velocity
gradient is proportional to shear stress. The • For non-Newtonian fluids there is a nonlinear
proportionality constant is called the dynamic relationship between shear stress and velocity
viscosity. gradient. Power law fluids and Bingham plastics are
• Laminar flow involves ordered or aligned streamlines. two examples.
In contrast, turbulent flows involve tortuous • Rheology is the study of materials with some capacity
streamlines and mixing. The dimensionless quantity to flow. Viscoelastic materials exhibit both solid- and
Reynolds number (Re) is a measure of velocity and fluidlike behavior. The complex modulus can be used
governs the transition from laminar to turbulent flow. to describe viscoelastic behavior. It may depend
• Generally, Reynolds numbers much less than one are nonlinearly on frequency, and power-law relationships
dominated by viscosity and are laminar. Reynolds can be effective descriptions of complicated
numbers much greater than one are dominated by viscoelasticity.
inertia and are turbulent. The specific Reynolds • Dimensional analysis, scaling, and estimation are
number at which the transition occurs depends on methods that can provide insight into functional
the specific geometry. At the cellular scale, flow is relationships when a specific equation is not
typically laminar. available. Often these relationships are sufficient to
solve important problems.
• The velocity profile of laminar Newtonian flow for
simple geometries can sometimes be solved in • Dimensional analysis exploits the requirement for units
closed form. to be consistent in order to arrive at potential functional
relationships and novel dimensionless numbers.
• If viscosity is small enough to be neglected, the flow is
termed inviscid and is governed by Bernoulli’s equation.
116 ChApTER 4: Fluid Mechanics Primer

Problems

1. Determine the flow profile in a pipe with circular cross canceled out). So in the example, we can eliminate the
section, with inner radius r0, using differential analysis, time component of velocity and acceleration, showing
assuming pressure-driven flow with the same flow them to be redundant with respect to time. If we
conditions as we used for the parallel-plate problem. eliminate acceleration, we get the rank = 2, and time
Determine the ratio of the peak velocity to the average and velocity are left.
velocity. Is this ratio the same or different from that of (a) Based on this, go back to the fluid drag problem
the parallel-plate solution? presented in the text and determine the unit matrix
for that scenario, and show that the rank is 2. What
2. Show that for a Newtonian fluid, the complex shear
other base parameter groups could we use? (Hint: in
modulus and complex viscosity simplify to pure fluid
general, the dependent variable, force in this case,
values under an oscillatory input.
is typically not used in the base parameter groups,
3. Plot the flow profiles of a power-law fluid where β = 1 to avoid having it show up in multiple places in
and α = 0.5 or 2. You may select the plate gap and peak the function, thereby making it hard to isolate. It is
velocity for convenience, but make sure you specify partly for this reason some people prefer leaving the
them in your response. dependent variable outside the function.)
(b) Using a second base parameter group, derive the
4. Using the Navier–Stokes equations, derive the shear dimensionless form of the function. Based on this
stress acting on the bottom plate of a parallel-plate new set, if you decrease the length by a factor of
setup, with gap h, if there is both a pressure gradient 1000, keep the viscosity the same, keep the velocity
(dp/dx = C) and an upper plate moving with velocity V0, the same, increase the density by 1000, and multiply
with the bottom plate stationary. The Newtonian fluid the end force by 1000, can you still scale the
between the plates has density ρ and viscosity μ. experiment correctly? Does that make sense? Why?
5. If a cell can be treated as a fluid-filled bag with the fluid 8. Suppose you are considering a spring-dashpot system
having viscosity about 10 times that of water, determine as a model to describe some process. The spring
whether the flow within a cell during micropipette behavior can be described by Hooke’s law, F = kx, where
aspiration is laminar, turbulent, or transitional. You may x is a displacement, F is the force, and k is the spring
assume the pressure gradient is applied so that the cell constant. The Newtonian dashpot behavior is described
takes about 1 min to enter the pipette completely, at by F = μv, where v is the velocity, and μ is the viscous
roughly uniform velocity. coefficient (not viscosity). You wish to determine
the characteristic frequency of the system. Using
6. Determine a relationship for the radii of blood vessels dimensional analysis, find the dimensionless parameters
that branch out from a parent vessel of radius R to two governing this frequency.
progeny branches of radii R1 and R2 (not necessarily
equal) such that the shear stress resulting from laminar 9. Dimensional analysis is frequently used in fluid
blood flow on the inside of these vessels remains simulation. Let us say you are given a smooth sphere of
constant. Ignore fluid effects at the bifurcation itself. You radius r, suspended in a fluid of density ρ and viscosity
may assume that blood is Newtonian and incompressible. μ with a spring of constant k, under gravity g. You
stretch the spring out and release it to let the system
7. In the text, we discussed the number of base oscillate. Use dimensional analysis to determine the
parameters and units. A unit is independent if it cannot relationship of the “half-life” time of damping to the
be expressed in terms of the other units. To derive the other parameters. This half-life damping time refers to
number of independent units, one can create a matrix the time it takes the oscillations to reach one-half of the
based on the units at the top of each column and the initial amplitude when the system is initialized.
parameters beside each row. Example:
10. How much power do you give off in body heat? That
is, your presence in a room will warm the room from
time 0 1
1 your body heat. If I were to replace you with a lamp of
velocity −1 some wattage to achieve the same warming rate, what
 
acceleration  1 −2  wattage bulb would I need?

11. Each cell in your body is very roughly the same density
Where the first column is meters and the second as water (in reality, a bit higher, but ignore that for this
column is seconds. So time has no meters units question). If half of a typical body weight is in matrix
and a single time component (t = s1). The velocity (bones, cartilage, etc.), how many cells does a typical
has a meters component and a 1/s component adult have?
(v = m/s = m1s−1), and similarly for acceleration. The
rank of the matrix tells you the number of independent 12. How much force could your biceps muscle generate
units, and if you perform Gaussian elimination, you can if it were directly connected to a weight? Note that in
derive terms that can serve as the parameters on which the body, the bicep is actually levered to your forearm
you will base the dimensional analysis (these are not bones with the elbow as a fulcrum.
 Annotated RefeRences 117

Annotated References

Fabry B, Maksym GN, Butler JP et al. (2001) Scaling the microrheol- Pritchard PJ (2011) Introduction to Fluid Mechanics. John
ogy of living cells. Phys. Rev. Lett. 87, 1481–2. This article describes Wiley. This textbook on fluid mechanics covers in much more
how cells may be characterized as soft, glassy materials by analysis mathematical detail topics including hydrostatics, differential
of the frequency response of beads attached the cells. analysis, mass conservation, dimensional analysis, and the
Navier–Stokes equations.
Kamm R (2001) Molecular, Cellular, and Tissue Biomechanics. Lec-
ture notes from course number 20.310, Massachusetts Institute Stamenovic D, Suki B, Fabry B, et al. (2004) Rheology of airway
of Technology. This course introduced the concept of scaling smooth muscle cells is associated with cytoskeletal contractile
analysis and estimation methods applied to cell mechanics. It stress. J. Appl. Physiol. 96, 1600–1605. This article uses the power-
was the inspiration for several of the authors to undertake law presented in Equation 4.17 and shows an application of rheo-
this text, which has been heavily influenced by it as a result. logical analysis to studying cell response.
Kollmannsberger P & Fabry B (2011) Linear and nonlinear rheology Vogel S (1996) Life in Moving Fluids. Princeton University Press. This
of living cells. Annu. Rev. Mater. Res. 41, 75–97. A review of the rheo- book has a minimally mathematical description of biological fluids,
logical findings as applied to cells. This article contains many refer- with more focus as the organisms scale. It covers many key con-
ences to trace how this sort of modeling developed. cepts about biology and fluids in an accessible way.
This page intentionally left blank
to match pagination of print book
CHAPTER 5
Statistical Mechanics
Primer

T he structural components of cells can often be considered as collections of

many smaller, individual pieces. As we discussed in Chapter 2, polymers are
large molecules consisting of individual monomers or groups of individual mon­
omers joined together. It can be beneficial to determine how the properties of the
smaller pieces affect the collective behavior of the whole. We might want to
determine how much a given polymer curves (an  experimentally observable,
“macroscopic” property) as a function of some “microscopic” property, such as
the number, size, charge, of the monomers. Alternatively, we may want to deter­
mine the force required to straighten the polymer, or to extend it to a certain
length, as a function of these same microscopic properties. These questions can
be addressed using the analytical framework of statistical mechanics. In statisti­
cal mechanics, our goal is to relate the behavior of a system’s macroscopic behav­
ior (characterized by “macrostates”) to what we know about its microscopic
properties and behavior (characterized by “microstates”).

Statistical mechanics relies on the use of

probabilistic distributions
Statistical mechanics is so named because it relies on the use of probabilistic distri­
butions to form these relationships. The use of probability distributions allows for
the analysis of systems with large numbers of degrees of freedom by assuming their
collective behavior can be described by an appropriate statistical distribution.
Consider the case in which we come across an imaginary (and very large) pool
table with 1000 balls colliding. We wish to calculate the total kinetic energy of the
balls at some moment in time. In classical mechanics, we would track each ball’s
velocity, calculate each ball’s kinetic energy, and sum these values to get the total
kinetic energy. In statistical mechanics, our approach is to assume some probabil­
ity distribution for the velocities (10% of the balls have a velocity of 0–1 m/s, 15% of
the balls have a velocity of 1–2 m/s, etc.), and use this distribution to calculate an
expected value for the total kinetic energy.

Statistical mechanics can be used to investigate the influence

of random molecular forces on mechanical behavior
The ability to formulate relationships between microscopic and macroscopic
behavior is not the only beneficial feature of statistical mechanics. As we will
see, statistical mechanics is particularly useful for analyzing the behavior of very
soft structures, such as biopolymers and membranes under the influence of
random molecular forces. The notion of random molecular forces influencing
mechanical behavior can perhaps best be conveyed by considering a phenom­
enon called Brownian motion. Suppose we were to observe a very small particle
(such as a pollen grain, which was used by Einstein in his seminal studies of this
phenomenon; see Section 5.6) suspended in water at room temperature. If we
were to observe its movements under a microscope, we would see that the par­
ticle is not stationary, but experiences small random fluctuations in its position.
120 CHApTER 5: Statistical Mechanics Primer

Figure 5.1 Schematic depicting

Brownian motion of a particle in
fluid and thermal fluctuations of a
polymer. In both cases, the behavior or
configurations are caused by random
forces due to the surrounding
molecules.

particle: Brownian motion polymer: thermal fluctuations

This phenomenon is termed Brownian motion and is caused by collisions

between the water molecules and the particle, with small, instantaneous imbal­
ances in the forces causing small fluctuations in position. Because the kinetic
energy of the water molecules is associated with temperature, it comes as no
surprise that the fluctuations are dependent on temperature. If the temperature
is increased, the motion of the particle will also be increased.
These same random molecular forces that give rise to Brownian motion also exert
an influence on the mechanical behavior of soft structures within cells such as
actin microfilaments (as well as other biopolymers) which have been observed to
undergo a phenomenon called thermal fluctuations. These fluctuations are mani­
fested as polymer movements in the form of small wiggles or undulations. The
tendency of the polymer to wiggle is due to the same molecular forces that cause
Brownian motion. Just as increasing the temperature leads to greater Brownian
Nota Bene motion–induced displacements, it also leads to a greater degree of thermal fluc­
tuations. If we were to observe an actin polymer suspended in solution, we would
Degree of freedom. A degree of
freedom refers to the number of see that as temperature is increased, the polymer would tend to exhibit more
independent “ways” something can curvy or wiggly configurations (Figure 5.1). Conversely, if we were to lower the
maneuver. If the object in question temperature, we would observe configurations that tended to be straighter. We
can be treated as a small point, then will see that statistical mechanics is able to account for the influence of random
typically the degrees of freedom refer molecular forces in this and other phenomena by accounting for the influence of
to the number of dimensions entropy on equilibrium behavior.
available for motion. A bead on an
abacus is provided one degree of We now present some of the basic analytical tools of statistical mechanics. We will
motion, along the abacus rod. A be covering fundamental concepts and relations in statistical mechanics, such as
single helium atom in the air would internal energy, entropy, free energy, the Boltzmann distribution, and the parti­
have three degrees of motion.
However, if the object in question has
tion function. We’ll also be discussing random walks, a large class of mathemati­
geometry, then the degrees of cal problems often used in statistical mechanics analyses of membranes and
freedom must also account for polymers. By covering these topics, our goal is not only to give insight into the
rotations around different axes. A mathematical and/or physical origins of some of the polymer models in Chapters
book, for example, has six degrees of 7 and 8, but also to provide you with a basic understanding of statistical mechan­
freedom because, in addition to being ics sufficient for more advanced topics in polymer physics in the future. Because
able to translate along the three statistical mechanics considers the thermodynamic energy of an object, we will
spatial axes; it can also rotate about
each axis.
begin this chapter with a discussion of a particular form of energy called internal
energy.

5.1 INTERNAL ENERGY

Potential energy can be used to make predictions of
mechanical behavior
As you may know, there are many forms of energy: kinetic, potential, thermal,
electromagnetic, etc. These different forms of energy are unified by one common
 INTERNAL ENERGY 121

principle: within a closed system, any form of energy can be transformed into
another form, but the total energy of the system remains constant. In thermody­
namics, we are concerned with several forms of energy, one of the most important
being internal energy. For now, we will simply consider it as a form of energy that
is the sum of multiple types of energy, one of the primary types being potential
energy.
In mechanics, potential energy is defined as the capacity to do work. It is a par­
ticularly important form of energy in analyzing mechanical systems because of
the principle of minimum total potential energy. This principle states that when a
structure is subject to mechanical loading, it shall deform in such a way as to min­
imize the total potential energy in reaching equilibrium. The implications of this
principle are that it does not matter if a structure is relatively simple (such as a
single polymer) or relatively complex (such as an entangled network of one mil­
lion polymers), potential energy can take the form of a single scalar quantity that
represents the mechanical state of the structure. In addition, by finding the con­
figuration that minimizes the potential energy, we can determine the equilibrium
configuration of the structure under a given mechanical load.
Let us consider an example using springs to demonstrate how potential energy
can be used to determine the equilibrium state of a mechanical system. Con­
sider a Hookean spring. The force required to separate the ends of the spring by a
distance x is given by F = k1(x − x1), with a spring constant k1 and equilibrium
length x1. If we have a second Hookean spring, we know that F = k2(x − x2). In
this  case, the potential energies of the two springs are W1 = ½k1(x − x1)2 and
W2 = 0.5k2(x − x2)2. If we link the ends of both springs together so that their lengths
are the same, and pull them by their ends so they begin to stretch, the total poten­
tial energy is

W = W1 + W2 = ½k1( x − x1 )2 + ½k2 ( x − x 2 )2 , (5.1)

because their lengths are the same (Figure 5.2). As the length x is changed, the
total potential energy of our spring system also changes. We can see that although
this system contains multiple bodies (that is, two springs), the potential energy
gives a single (scalar) representation of the mechanical state of this system. This
quantity has a useful physical meaning, as it is equal to the work this system is
capable of performing.
Now, suppose we allow the system to equilibrate without applying any external
forces—we just let the system relax. We know that each spring has its own equilib­
rium length, however, now that their ends are joined, the two-spring system has
its own equilibrium length. As we mentioned earlier, the principle of minimum k1
total potential energy states that we can find the equilibrium state by finding the spring 1:
configuration that minimizes the total potential energy. For our spring system,
this means finding the length x at which Equation 5.1 is minimized. We can find k2
this length by taking the derivative of W with respect to x, and setting this equation spring 2:
to zero:
k1
dW/dx = k1( x − x1 ) + k2 ( x − x 2 ) = 0. (5.2)
two-spring
Solving for x, we get system:
k2
x = (k1 x1 + k2 x 2 )/(k1 + k2 ), (5.3)
x
which is the equilibrium length of our system.
Figure 5.2 Schematic of a two-
Note that finding the equilibrium length via the principle of minimum total poten­ spring system. The equilibrium length
tial energy gives the same answer as that obtained by performing a force balance. of the system, x, can be found by
We know that in equilibrium, without any external forces, there should be no net minimizing the potential energy.
122 CHApTER 5: Statistical Mechanics Primer

internal forces. Therefore any compression from one spring should balance the
extension from the other. Here, the equilibrium length x can be computed as

F = k1( x − x1 ) = −k2 ( x − x 2 ), (5.4)

where we use a negative sign on the second spring to indicate that it is mechani­
cally opposing the first spring (the second spring is compressing if the first spring
is extending and vice versa). As expected, Equation 5.4 is identical to Equation 5.3.
The reason performing a force balance is equivalent to potential energy minimi­
zation is that fundamentally, forces are the gradient of potential energy. Forces are
balanced (net forces are zero) when there is no gradient of potential energy, which
occurs at maxima or minima. Energy maxima are technically equilibrium points
but are unstable, so they are omitted from the discussion (the classic example is
an inverted pendulum). While performing a force balance and potential energy
minimization produce equivalent results, in analyzing complex mechanical sys­
tems, the latter approach is often preferred due to mathematical simplicity.
Potential energy has the added advantage that it allows us to marry the analytical
machinery of continuum mechanics with that of statistical mechanics, as we will
see in the next section.

Strain energy is potential energy stored in elastic deformations

When a body is subjected to a mechanical force, the body deforms. These defor­
mations may either be elastic (in other words, they self-reverse upon removal of
the force) or plastic (permanent deformations that do not reverse upon removal of
the force). When a structure is subjected to elastic deformations, potential energy
is stored in the structure; this form of potential energy is called strain energy. Note
that all strain energy is a form of potential energy, but not vice versa. For example,
raising an object a certain height increases its potential energy via opposition of
the gravitational forces, but does not increase its strain energy, because the object
is not deformed.
In Chapter 3, we learned about the concepts of stress and strain. Conveniently,
strain energy can be described in terms of these quantities. To demonstrate this,
we first derive the strain energy for a rod axially loaded at the tip. In Chapter 3, we
found that the force was proportional to the displacement,

F = EA∆L/L. (5.5)
Nota Bene The potential energy input into the rod to achieve this displacement is
Microscopic strain energy.
Microscopically, by strain energy we W = ½F ∆L. (5.6)
mean that when we displace the tip
of the rod a bit, all of the molecules In the rod example we considered the material response from the force and dis­
or atoms making up the rod are placement at the tip to be distributed throughout the material in the rod, in the
separated or deformed a bit more form of stress and strain. The potential energy input into the rod by the force at the
from their equilibrium distances and tip is analogously distributed through the material in the rod. For each small vol­
shapes, resulting in spring-like ume in the rod, a small element of potential energy is stored that we denote as dW.
potential energy storage throughout
the rod.
Thus, the total potential energy stored in the rod is the product of the volume of
the rod and this small element of energy

W = LAdW . (5.7)

If we equate the potential energy input at the tip with the internal potential energy
stored, we get,

dW = ½F /A ∆L/L = ½σ ε. (5.8)

The internal potential energy is strain energy, because it is the energy stored in the
form of strain (similar to the way the potential energy of a spring depends on its
stretch). The internal potential energy per unit volume is the strain energy density.
 INTERNAL ENERGY 123

We showed for an axially loaded rod, the strain energy density can be found as
one-half of the product of stress and strain. In general, it can be shown that the
strain energy density is one half of the component-wise product of stress and
strain
dW = ½(σ 11ε11 + σ 12ε12 + σ 13ε13 + σ 21ε 21 + σ 22ε 22
(5.9)
+ σ 23ε 23 + σ 31ε 31 + σ 32ε 32 + σ 33ε 33 ).

Equilibrium in continuum mechanics is a problem of strain

energy minimization
In the development above, we assumed that the energy input into the rod by the
force at the tip was equal to the stored internal strain energy. This assumption is
actually a statement of the conservation of energy principal of classical mechan­
ics. Outside of nuclear processes, energy can neither be created nor destroyed.
We have also seen that the three fundamental relations that make up a pro­blem
in continuum mechanics are kinematics, constitutive, and equilibrium.
However, there is an even more fundamental problem statement in continuum
mechanics that arises from the principle of minimum total potential energy.
Specifically, the problem can be given as finding the internal deformation state
that both minimizes the internal strain energy and satisfies the boundary condi­
tions (just as the two-spring problem could be given as finding the length that
minimized the potential energy). Indeed, it can be shown that the equilibrium
equation follows from assuming the minimization of strain energy and that the
two are equivalent.
Nota Bene
Changes in mechanical state alter internal energy Stress can be defined through
the strain energy. One of the most
With our discussion of potential energy and strain energy in hand, we now intro­ useful properties of strain energy is
duce the concept of internal energy. In thermodynamics, internal energy is the that it can be used to define stress. In
total energy contained within the system and is defined as the capacity for a ther­ our initial discussion of stress, we
defined it in an intuitive fashion as a
modynamic system to do work plus release heat. The two major components of distributed or normalized force.
internal energy are potential energy and kinetic energy. However, this is a somewhat
We have already seen that elastic deformations are associated with changes in dissatisfying approach in that it is
conceptually far from a rigorous
potential energy. These changes in potential energy arise out of configuration- thermodynamic quantity. On the
dependent changes in potential energy between interacting atoms or molecules. other hand, the strain energy is the
The potential energy due to interactions between two molecules can vary greatly, increase in potential energy stored
depending on the distance between them. There may be van der Waal forces that locally in a material due to its
result in long-range attractive forces, if they are dipoles. The potential energy due deformation (or strain) and thus is a
to these van der Waals forces would decrease as the distance between the mole­ thermodynamic quantity with a clear
definition. Critically, stress is simply
cules decreased. However, if the two molecules came too close together, the
the first derivative of the strain
potential energy would quickly rise, owing to steric interactions between atoms of energy density with respect to strain,
neighboring molecules (such as the energetically unfavorable interaction between
the electron clouds of atoms). If the atoms of each molecule were charged, there σ = dW/dε.
would also be a Coulomb potential that would contribute to the energy. If the Often, this is a very useful definition
atoms within each molecule both had an equal charge of the same sign, the poten­ of stress. Indeed the constitutive
tial energy would increase as they came closer together. behavior of many complex materials
is defined not in terms of a
In contrast to potential energy, which changes with alterations in mechanical relationship between stress and
configuration, changes in kinetic energy arise out of alterations in the velocity of strain, but in terms of their strain
the system’s particles. There is an association between kinetic energy and tem­ energy functions.
perature, because the temperature of an object is related to the speed of its funda­
mental particles. If we were to raise the temperature of an ideal gas, the average
velocity of the gas particles would increase.
In our calculations, we will generally be interested in obtaining the equilibrium
mechanical configuration for some system of interest. We will assume a reference
state with zero internal energy, with changes in configuration to this reference state
124 CHApTER 5: Statistical Mechanics Primer

resulting in increases or decreases in internal energy. This will allow us to ignore con­
tributions to the internal energy that (1) do not change (or which change relatively
little) compared with the reference state, or (2) are largely decoupled from changes
in configuration. To illustrate this, in many of our calculations, we will assume that
changes in internal energy arise solely out of changes in potential energy, and ignore
changes in internal energy arising from alterations in kinetic energy.

5.2 ENTROPY
Entropy is directly defined within statistical mechanics
Our focus now turns to another important thermodynamic quantity, entropy.
You may have heard the vague explanation that entropy is related to disorder.
Nota Bene Many people have heard the “messy room” analogy: a messy room is more dis­
ordered than a clean room, and so has higher entropy. One reason why the con­
Entropy and the first law of cept of entropy can be difficult to grasp is that no direct relation for entropy
thermodynamics. There are
exists within thermodynamics. Instead of a direct definition, an incremental
classically three laws of
thermodynamics: definition relates the change in entropy ΔS to a change in heat Δq of a system. In
particular, for a constant temperature (isothermal) and reversible process (one
(1) Conservation of energy. that can be reversed without changing the system or its surroundings), the
(2) The entropy of an isolated system
change in entropy is defined as
(one that does not exchange
energy or mass with the outside)
∆S = ∆q/T , (5.10)
never decreases.
(3) As the temperature of a system
reaches absolute zero, its entropy where T is the absolute temperature and Δq is the amount of heat absorbed. We
reaches a minimum. see that this is an indirect, incremental definition of entropy, as it only relates the
change in entropy to thermodynamic quantities, rather than directly defining
The incremental definition of entropy
is a consequence of the first law.
entropy itself.
Specifically, the change in internal In contrast, within statistical mechanics, entropy is directly defined. This defini­
energy, ΔW is equal to the change in tion of entropy was developed by Ludwig Boltzmann and was one his most impor­
“entropic” energy SΔT minus the
mechanical work done by the
tant contributions. In particular, the Boltzmann definition of entropy is
system.
S = kB ln Ω. (5.11)

Nota Bene Here, kB is the Boltzmann constant and is equal to 1.38 × 10−23 J/K. Ω is defined as
the density of states and is equal to the number of microstates for a given mac­
Ludwig Boltzmann’s tombstone. rostate. Each of these terms—density of states, microstates, and macrostates—will
Ludwig Boltzmann (1844–1906) was
be discussed in detail next.
an Austrian physicist who made
seminal contributions to thermody­
namics and statistical mechanics. He
was a full professor by age 25 and an Microstates, macrostates, and density of states can be
advocate of the atomic theory of exemplified in a three-coin system
matter when it was unpopular. His
tombstone is inscribed with his Before formally defining microstates, macrostates, and density of states, we present
famous equation (Figure 5.3). an example to gain some intuition for these quantities. Consider a set of three coins
placed inside a container (such as a coffee can): a nickel, a dime, and a quarter. We
designate the nickel as coin 1, the dime as coin 2, and the quarter as coin 3. When
we shake the container, this flips all three coins simultaneously. Each of the coins
can land heads (denoted as h) or tails (denoted as t). Whether each coin lands heads
or tail is random (mathematically, this is a random variable).
When we shake the container, we consider each microstate to be a possible out­
come of a shaking event specified by whether each coin comes up heads or tails.
If we were to shake the can, and the nickel comes up heads, the dime comes up
tails, and the quarter comes up heads, this microstate, which we denote as m, can
be written as
m = hth
 ENTROPY 125

Figure 5.3 Ludwig Boltzmann’s

tombstone in Vienna, Austria.
(Courtesy of Thomas D. Schneider).

where the ith letter designates the outcome (that is, h or t) of coin number i. We
say that this system’s current microscopic state, or microstate, is “hth”
Given that there are three coins and two possible outcomes per coin, there are
eight possible microstates. These individual microstates, denoted by mx, are:

m1 = hhh

m2 = hht

m3 = thh

m4 = hth

m5 = htt

m6 = tht
m7 = tth

m8 = ttt.

We now turn our attention to macrostates. Imagine a case where in which we are
not able to observe whether each coin lands heads or tail, but only some property
that is dependent on the number of coins that lands heads. For instance, imagine
that we place a small elf into the coffee container with the coins and cover the top
of the coffee can with a lid (Figure 5.4). We tell the elf inside to count the number
of coins that lands heads after each shake of the container, and yell out that num­
ber. We shake the container, and after the dime and the nickel (but not the quar­
ter) lands heads, we hear a diminutive voice yell out, “two!” We shake the container
again, and after all the coins land heads, we hear “three!”
We consider this number, the total number of heads, to be representative of the
current system’s macroscopic state, or macrostate. It is considered a macroscopic
value because we cannot observe the exact configuration of the microscopic con­
stituents that led to this macrostate (in other words, whether each coin landed
heads or tails). Rather, we are only able to observe some macroscopic property. If
126 CHApTER 5: Statistical Mechanics Primer

Figure 5.4 Example of a

macroscopic property. Three coins
and an elf are placed inside a coffee
can. The can is covered and shaken. The
elf yells out the number of coins that
landed heads. The number of heads is
considered a macroscopic value,
because we cannot observe the exact
configuration of the microscopic
constituents associated with this
macrostate (in other words, which coins
came up heads or tails). Two!

we define W(mx) to be the number of heads in microstate mx, then for each of the
eight microstates, the corresponding system macrostate is

m1 = hhh: W (m1 ) = 3

m2 = hht: W (m2 ) = 2

m3 = thh: W (m3 ) = 2

m4 = hth: W (m4 ) = 2

m5 = htt: W (m5 ) = 1

m6 = tht: W (m6 ) = 1

m7 = tth: W (m7 ) = 1

m8 = ttt: W (m8 ) = 0.

Notice that W can range from 0 to 3, and that the number of microstates associ­
ated with each macrostate is different. In particular, the macrostate correspond­
ing to W = 3 has only one microstate associated with it (hhh), whereas the
macrostate corresponding to W = 2 has three microstates associated with it (hht,
thh, and hth). The specific microstates associated with each macrostate are:

W = 3 : hhh

W = 2 : hht , thh, hth

W = 1: htt,tht,tth

W = 0 : ttt.

We can now introduce the density of states (sometimes called the multiplicity of
states), Ω(W), as the total number of microstates associated with each macrostate
W. In our three-coin system, we know that for W = 3 (three heads), there is only one
such microstate (hhh). In this case, Ω(W = 3) = 1. For W = 2, there are three such
microstates: hht, thh, and hth. Therefore, Ω(W = 2) = 3. Similarly, Ω(W = 1) = 3 and
Ω(W = 0) = 1.
 ENTROPY 127

Microstates, macrostates, and density of states provide insight

to macroscopic system behavior
Now let us introduce more formal definitions for microstates, macrostates, and
density of states. Remember, our goal is to take knowledge of the microscopic
behavior and properties of some system (for example, a structure or body of inter­
est) and gain insights into its macroscopic behavior. We do this by considering the
microscopic behavior in a statistical, rather than a deterministic, fashion. In other
words, we want to describe the system’s average large-scale behavior and not its
specific microscopic behavior over time. With this in mind, we provide the follow­
ing definitions:
• Microstate. Short for microscopic state, it is the state of the system with regard to
the specific (detailed) configuration of its microscopic components.
• Macrostate. Short for macroscopic state, it is the state of the system with regard to
some macroscopic property. There may be multiple microstates associated with a
given macrostate.
• Macroscopic property. A scalar property that describes the thermodynamic state
of a multi-bodied system. These properties are considered macroscopic in that they
are typically observable thermodynamic variables, such as pressure, temperature,
or volume. Another example might include end-to-end length for a polymer.
• Density of states. A function of some macrostate characteristic, it describes the
number of microstates associated with each macrostate with regard to some mac­
roscopic property of interest.

Ensembles are collections of microstates sharing

a common property
Before concluding this section, it is worthwhile to introduce the concept of ensem­
bles. An ensemble refers to a collection of microstates that share a common prop­
erty. In our three-coin system, one ensemble we could define is the collection of
all microstates with at least one coin that is heads. This would consist of all possi­
ble microstates (hhh, hht, thh, hth, htt, tht, and tth) except for one (ttt). Another
ensemble we could define is the collection of all microstates in which the nickel
(coin 1) is heads (hhh, hht, hth, and htt). Although these may not seem like terri­
bly useful ensembles, within statistical mechanics there are three important
ensembles. The first, known as the microcanonical ensemble, consists of all micro­
states that share the same internal energy. The microcanonical ensemble is used Nota Bene
in analyses of systems that are energetically isolated. The second, the canonical
ensemble, consists of all microstates associated with a constant temperature sys­ What is meant by canonical? The
term canonical is derived from the
tem. This ensemble is used in analyses of systems that are allowed to exchange Greek word kanon or “rule”. In
energy with a large thermal reservoir. The last, the grand canonical ensemble, is mathematics and statistical
used in analyses of systems in which exchange of both energy and mass (or parti­ mechanics it refers to a standard
cles) can occur. We will use both the microcanonical ensemble and canonical presentation or analysis approach of
ensemble, but the grand canonical ensemble is beyond the scope of this text. a simple system. However, the
canonical forms can often provide
great insights to the extent that they
Entropy is related to the number of microstates associated are similar to several more complex
with a given macrostate problems.

With our definitions of microstate, macrostate, and density of states in hand, we

now revisit the Boltzmann definition of entropy, Equation 5.11. We saw that
entropy is equal to the product of the Boltzmann constant and the log of the den­
sity of states. Because the density of states is a function of some macroscopic char­
acteristic (in particular, it gives the number of microstates associated with each
value of some macroscopic property), we can use this definition to calculate how
entropy varies with a macroscopic property, so long as we can enumerate the den­
sity of states. Upon inspection of Equation 5.11, it is easy to see that for a given
macroscopic property W, S(W) is large when Ω(W) is large, and conversely, S(W)
128 CHApTER 5: Statistical Mechanics Primer

is small when Ω(W) is small. Note that the minimum number of microstates asso­
ciated with a particular macrostate is one. Therefore, the lowest possible value of
Ω is 1; in this case, S = kBln(1) = 0.

Example 5.1: Entropy calculation

Consider the density of states from our three-coin We know that S = kBlnΩ, thus
system:
S (W = 0) = kBln(1) = 0
Ω(W = 0) = 1
S (W = 1) = k B ln(2) = 0.7k
Ω(W = 1) = 2
S (W = 2) = k B ln(2) = 0.7k
Ω(W = 2) = 2
S (W = 3) = kBln(1) = 0
Ω(W = 3) = 1
In this case, entropy is highest when W is equal to 1 or 2
Calculate the entropy for each value of W. For which (it is equally high in both cases), and lowest when W is
value(s) of W is entropy the highest? The lowest? equal to 0 or 3 (it is equally low in both cases).

5.3 FREE ENERGY

Equilibrium behavior for thermodynamic systems can be
obtained via free energy minimization
We now introduce the concept of free energy, and explore ways in which this ther­
modynamic potential can be used to predict equilibrium behavior of thermody­
namic systems. In Section 5.1, we learned about the principle of minimum total
potential energy, which states that when a structure is subject to mechanical load­
ing, it shall deform in such a way as to minimize the total potential energy. Within
thermodynamics, there is an analogous principle called the principle of minimum
free energy. The free energy Ψ (also known as the Helmholtz free energy or the
energy available to do work at a constant temperature and volume) is the sum of
internal energy W and the entropy S, where the latter is weighted by the negative
absolute temperature T:

Ψ = W − TS. (5.12)

The principle of minimum free energy states that a closed system (namely, a sys­
tem which can exchange energy in the form of heat or work, but not matter, with
its surroundings) at constant temperature will spontaneously adapt itself to lower
its free energy. A thermodynamic equilibrium is reached when free energy is min­
imized. Just as potential energy minimization can be used to find the equilibrium
state of mechanical systems, free energy minimization can be used to find the
equilibrium state of thermodynamic systems.
Upon inspection of Equation 5.12, it becomes evident that for a given body, the
free energy can be minimized in two distinct ways: through decreasing the inter­
nal energy W, or increasing the entropy S. The particular influence of energy ver­
sus entropy in the process of free energy minimization is determined by the
relative magnitudes of W and S, as well as the temperature T. For example, at zero
temperature, the entropy term vanishes, and the free energy is equal to the inter­
nal energy. Conversely, at very high temperatures, the TS term will dominate the
free energy. When the entropic contribution to the free energy TS is comparable
to the energetic contribution W (when W = TS.), a transition occurs between
energy-dominated and entropy-dominated behavior.
One question that may arise is if we have two principles of energy minimization,
one based on potential energy and one based on free energy, how do we know
 FREE ENERGY 129

which one is appropriate for a given system? Also, would these two principles give
­contradictory behaviors? For everyday (nonmicroscopic) structures such as a Nota Bene
steel beam in a bridge, we know that the principle of minimum potential energy
dictates its deformation under loading. However, we could also consider the beam Helmholtz and Gibbs free energy
are used for similar purposes but
to be a thermodynamic body subject to the laws of thermodynamics, which state
in different contexts. Free energy
that its deformation should be dictated by the principle of minimum free energy. denoted by Ψ here is known as the
So which one is correct? Helmholtz free energy. As the
difference between energy and
The answer is they are both correct. It is important to point out that the principle
entropy, it is often thought of as a
of potential energy minimization is not contradicted by the principle of free measure of the work that a system
energy minimization; rather, it is encompassed in it. In particular, energetic effects could do at a constant temperature
for everyday structures are so dominant that they mask any entropic influences on and volume. However, there are other
their behavior. Mathematically, this can be expressed as W  TS; the free energy “thermodynamic potentials” that are
is equivalent to the internal energy. Remember from Section 5.1 that the primary useful in various contexts. For
components of internal energy are potential energy and kinetic energy; however, instance, the enthalpy is defined as
the heat transferred to an isolated
in many cases we can assume that for the microscopic structural components of system at constant pressure and is
cells such as polymers and membranes, kinetic energy is generally not dependent defined as W + pV. The Gibbs free
on their mechanical configurations. The same is true for everyday structures as energy is often used in chemistry. It is
well. Therefore, changes in internal energy can be attributed solely to changes in the maximum work that can be
potential energy. In this case, for bodies in which W  TS, free energy minimiza­ obtained from a closed system at
tion is equivalent to potential energy minimization. For brevity, we will refer to constant volume, W − TS + pV.
systems in which W  TS as mechanical systems; systems in which W ~ TS or
W  TS will be referred to as thermodynamic systems.
Nota Bene
Temperature-dependence of end-to-end length in polymers Zero temperature limit. We could
arises out of competition between energy and entropy consider our treatment of solid
mechanics in Chapter 3 to be the
We now explain the concept of the competing influences between energy and zero temperature limit. This does not
entropy as a system spontaneously minimizes its free energy. Let us revisit our literally mean that the cell is frozen to
absolute zero. Rather, it is a common
example of a polymer subject to thermal fluctuations. If we were to observe a sus­ way of expressing that for the
pended actin polymer over time at room temperature, it would not be completely particular process we are interested
rigid, rather, it would continuously wiggle. Recall that the tendency of the polymer in, the entropic contribution TS is
to wiggle is due to random molecular forces similar to those that cause Brownian negligible as compared with the
motion. In the same way that increasing the temperature increases the small fluc­ energetic contribution W.
tuations in position of a suspended particle during Brownian motion, increasing
the temperature increases the tendency of the polymer to take on an increasingly
wiggly configuration (Figure 5.5).
To develop some intuition for the competition between energy and entropy in this
temperature-dependent phenomenon, we present an example of a model poly­
mer subject to thermal fluctuations. Note that the formal mathematical frame­
work for this introductory example will be given in subsequent sections. We first
need to define two lengths (Figure 5.6). The contour length, L, is the length of the
polymer if it were completely straight. The end-to-end length, R, which, as the
name implies, is the length of a straight line between one end of the polymer and
the other, and depends on the actual configuration of the polymer. If the polymer
is straight, R = L; in all other cases, R < L. Therefore, R is always less than or equal
to L but greater than or equal to 0. If we were to observe a thermally fluctuating

Figure 5.5 At low temperature, a

suspended actin polymer tends to
take on configurations that are
relatively straight. As the temperature
is increased, it tends to take on
configurations that are more wiggly.

low temperature high temperature

130 CHApTER 5: Statistical Mechanics Primer

actin polymer and quantify its end-to-end length as a function of temperature, we

b would see that on average, its end-to-end length decreases with increasing tem­
perature, indicating that as temperature is increased, the configuration of the
polymer tends to go from a straight configuration to a more wiggly one.
Now, consider a model polymer consisting of n rigid links of size b. The links are
connected to each other by freely rotating hinges. We constrain the potential ori­
R
entations of the links to simplify our analysis. In particular, we confine the orienta­
tions to be within a two-dimensional plane, and further, the links can be oriented
only vertically or horizontally (so at each joint the relative angle between two
adjacent segments is either 180°, 90°, or 0°; note that in the case of the former, the
polymer is overlapping onto itself, which we allow for simplicity). Note that
because there are n links of size b, then L = nb.
Remember that entropy is defined as the product of the Boltzmann constant and
Figure 5.6 Model polymer. The links
the log of the density of states Ω, and that Ω is defined as the number of micro­
of size b are connected to each other by states for a given macroscopic quantity. In our example, the microstates are the
freely rotating hinges. The contour different possible configurations of the polymer, and the macroscopic quantity of
length L is the length of the polymer if it interest is the end-to-end length R. Thus, finding Ω boils down to counting the
were completely straight. In this case, if number of ways in which our polymer can have a particular end-to-end length.
there are n links, L = nb. The second Computing the density of states for every possible value of end-to-end length
length is the end-to-end length R, which
(Ω(R) for 0 ≤ R ≤ L) would then allow us to determine the entropy associated with
is the length from one end of the
polymer to the other, and depends on each value of R, S(R).
the polymer configuration. While the mathematics for enumerating Ω(R) will be presented in later sections, a
simple thought experiment will allow us to obtain a qualitative sense for how the
density of states and entropy change with different values of R. Suppose the poly­
mer is completely straightened out. The polymer will have an end-to-end length
equal to the contour length, or R = L. Because we are only interested in the
dependence of Ω on R, we do not distinguish between rotated states (to illustrate,
a polymer straightened in the horizontal direction is considered to be equivalent
to a polymer straightened in the vertical direction) or reflected states. In this case,
there is only one polymer conformation in which this can occur, and so
Ω(R = L) = 1. Now suppose that the polymer is not completely straightened out, or
R < L. There are multiple polymer conformations that may have the same end-
to-end length (Figure 5.7) (in fact, mathematically, it can be shown that Ω is a

Figure 5.7 The model polymer in R=L

fully extended and less extended
configurations. If R = L, there is only
one possible configuration. In contrast, if
R < L, there are multiple possible
configurations. R<L

R<L
 MiCroCanoniCal EnsemBle 131

R=L Figure 5.8 Schematic of a polymer

in which the links are connected by
rotational springs. No energy is stored
in the springs (swirls) when the polymer
is straight (R = L). However, work is
R<L required to bend the polymer (R < L), and
this work is stored in the springs as
strain energy.

maximum when R = 0), therefore,

Ω( R < L) > Ω( R = L). (5.13)

The entropy is proportional to ln Ω, so the higher the value of Ω, the higher the
entropy. Therefore, polymer macrostates with end-to-end lengths of R < L are
associated with higher values of entropy than the polymer macrostate R = L.
Recall that equilibrium is achieved when the free energy is minimized, and that
increasing the entropy decreases the free energy. In our example, the smaller the
end-to-end length, the higher the entropy, and we therefore say that entropically,
a coiled configuration is more favorable than a straight configuration.
Now let us look at the effect of internal energy on the end-to-end length. Consider
the same polymer model as before, but this time imagine that instead of the links
being connected by freely rotating hinges, we now straighten out the polymer and
attach rotational springs between each link such that the springs are at their equi­
librium position when adjacent links are straight (180° between the links) (Figure
5.8). The addition of these springs is a theoretical construct that represents an
increase in internal energy if the polymer is bent. In this case, bending the poly­
mer requires work, which becomes stored in the springs as strain energy. Thus,
the polymer macrostate associated with R = L is associated with zero internal
energy (W = 0), while all other macrostates are associated with non-zero internal
energy (W > 0). Because the free energy is decreased when the internal energy is
reduced, we say that energetically, a straight configuration is favorable.
Now we know that entropically, a bent or coiled configuration is favorable in our
model polymer, while energetically, a straight configuration is favorable. So the
question becomes, what configuration will it tend to adopt? The answer is it depends
on the temperature. If we look at the expression for free energy, we can see that the
influence of energy or entropy on the free energy is determined by the temperature.
At lower temperatures, internal energy dominates the free energy. The polymer will
favor macrostates that are energetically favorable and will tend to be straighter. At
higher temperatures, entropy dominates the free energy; the polymer will favor
macrostates that are entropically favorable and will tend to be more wiggly.
In the next several sections, we will quantify this behavior by formulating a rela­
tionship that relates the equilibrium configuration of the polymer with tempera­
ture and its microscopic properties, such as the number of links and the energy of
bending. We will perform the same calculation using two distinct analytical
approaches: the microcanonical ensemble and the canonical ensemble.

5.4 MiCroCanoniCal EnsemBle

In the previous section, we used a model polymer system to gain a qualitative
understanding of the competition between energy and entropy in minimizing the
system’s free energy. The next section of this chapter focuses on developing an
132 CHApTER 5: Statistical Mechanics Primer

analytical framework to quantify this behavior. Indeed, determining macroscopic

behavior from microscopic properties is a general goal of statistical mechanics. As
we will see, there are several different ways within statistical mechanics to do this.
In the next two sections, we present two separate analytical approaches for per­
forming the same calculation: the microcanonical ensemble and the canonical
ensemble. In analyzing our model polymer, both approaches ultimately give the
same answer; however, we will see that their mathematical complexity can be
vastly different.

The hairpinned polymer as a non-interacting two-level system

Before we begin with our calculations using the microcanonical ensemble, we
first establish our model system. In this section, we will analyze a system of N par­
ticles called a non-interacting two-level system. It is called a two-level system
because each particle can be in only one of two energy levels. The system is called
non-interacting because each of the particles is considered independent. The
energy level that each particle takes on is unaffected by the energy levels of the
other particles. This type of system can be used to model a variety of biological
and physical phenomena. In this example, we use it to idealize the behavior of a
slightly modified version of the “hinged” polymer discussed earlier.
Consider a polymer consisting of N + 1 monomers and N linkages. The monomers
are rigid links, with torsional springs between adjacent links. For the sake of sim­
plicity, assume that the polymer is constrained to move in only one dimension.
Instead of sites of 90° bending between adjacent links, we can think of the poly­
mer as being able to undergo a 180° “hairpin” turn between adjacent links.
Although this does not seem realistic, there are certain cases in which this can
occur (Figure 5.9).
Because there are N linkages, there are a total of N possible hairpin sites. For
simplicity, we shall assume that the energetic cost of a hairpin bend is  (assume
 is positive). This energetic cost is the work that is necessary to twist the torsional
springs 180°, and is stored in the springs as strain energy. If Nh is the number of
sites that contain a hairpin, then the internal energy of a polymer with Nh hair­
pins is

W = N h. (5.14)

A microcanonical ensemble can be used to determine

constant energy microstates
With our model established, we now attempt to find the equilibrium configuration
using an analytical tool called the microcanonical ensemble. Recall that the micro­
canonical ensemble refers to a group of microstates, all of which are associated

Figure 5.9 Schematic and (A)

micrograph of hairpin polymers.
180° “hairpins”
(A) Depiction of the model polymer
undergoing 180° “hairpins.” Between
each pair of subunits, there can either
be a hairpin, or not. The polymer is
constrained to one dimension. (B)
Sequential images of a hairpin within an
(B)
actin filament that is straightening with
time. (B, from, Dogic Z, Zhang J, Lau AWC
et al. (2004) Phys Rev Letter 92. With
permission from the American Physical
Society.)
 MiCroCanoniCal EnsemBle 133

with the same internal energy W. To apply the microcanonical ensemble to our
model polymer, we consider it to be thermally and mechanically isolated from the Nota Bene
surroundings. We consider a constant energy ensemble of microstates, with each
microstate within the ensemble having the same energy W. The ensemble can be Are all microstates within an
ensemble equally likely to occur?
considered to be the collection of all microstates associated with the macrostate
The assertion that all microstates
having energy W. Within such an ensemble, each microstate is equally likely to within an ensemble are equally likely
occur. Therefore, if there are Ω microstates at a particular energy W, each micro­ to occur is actually an assumption. It
state within the ensemble occurs with the probability of p = 1/Ω. Our goal is to is sometime called the ergodic
calculate S(W), or the entropy for this ensemble of microstates having energy W. hypothesis or the fundamental
postulate of statistical mechanics
or the equal a priori probability
Entropy can be calculated via combinatorial enumeration postulate. At its core, we assume that
of the density of states we are considering the system for a
long enough period of time that all
Because entropy depends on the density of states, we must first calculate Ω(W), or possible microstates in the ensemble
the density of states Ω for each ensemble with energy W. We can find this using occur for equally long periods.
combinatorics. If we have N possible hairpin sites, the total number of configura­
tions can be found using the bionomial coefficient. The binomial coefficient is
Nota Bene
 m m!
 n  = n !(m − n )! , (5.15) Simplification of large factorials.
Factorials often appear in statistical
mechanics because calculating
which gives the number of ways in which a group of n items can be chosen from a microstates involves combinatorics.
set of m. In our problem, we can use the binomial coefficient to find the total num­ Factorials of large numbers are
ber of microstates in which Nh hairpins occur, given N total sites difficult to calculate. However, an
approximation can often be
sufficient. For instance, notice that:
N N!
Ω( N h ) =   = .
 N h  N h !( N − N h )! n n
(5.16) ln n! = ∑i =1
ln i ≈
∫ ln x dx ,
Now, we know that the entropy S = kB lnΩ, or 1

where we have used the trapezoid

S ( N h ) = kB (ln N ! − ln N h ! − ln( N − N h )!). rule to approximate the sum.
(5.17)
Integration by parts yields
But from Stirling’s approximation, assuming N is very large, ln n! ≈ n ln n − n + 1.
S ( N h ) = kB ( N ln N − N h ln N h − ( N − N h )ln( N − N h )). Stirling’s approximation for large
(5.18)
factorials can be obtained by taking
We know that W = Nhe, so we can write the entropy in terms of the energy by sub­ the exponential of the above
stituting the expression Nh = W/e equation.
If n is extremely large, n lnn  n, and
 W W  W  W  so
S (W ) = kB  N ln N − ln −  N −  ln  N −   .
         (5.19) ln n! ≈ n ln n.

Notice that for relations derived using

Entropy is maximal when half the sites contain hairpins Stirling’s approximation, significant
errors can occur if n is not large
Let us analyze our expression for S(W), which is plotted in Figure 5.10. We see enough. For example, if n = 10,
ln(n!) = 15.1 and n ln(n) = 23, a 52%
that the expression for entropy is equal to zero at its bounds, that is at W = 0 and
error.
W = Ne. These points correspond to the cases in which none or all of the sites con­
tain a hairpin, respectively. In both cases, the corresponding value for the density
of states is one, resulting in zero entropy. The entropy is maximal between these
two points, at W = 0.5Ne. This value of internal energy corresponds to the case
when half the sites contain a hairpin. This indicates the number of possible micro­
states is maximized when half the sites contain a hairpin.

S(W) can be used to predict equilibrium behavior

Now that we have invoked the microcanonical ensemble to form a relationship for
S(W), we can use thermodynamic relations to see how a group of these polymers
134 CHApTER 5: Statistical Mechanics Primer

Figure 5.10 Entropy S as a function

of internal energy W. For the hairpin
W = 0.5Nε
model, polymer entropy is maximal
when W = 0.5N, which corresponds to
the case when half the sites contain a

entropy S
hairpin.

0
internal energy W

Nota Bene would behave at thermal equilibrium. In particular, we know from thermody­
namics at thermal equilibrium,
Thermodynamic equations of
state imply relations between 1 ∂S (W )
state variables. The expression = . (5.20)
1 ∂S T ∂W
= is called an equation of
T ∂W Combining Equation 5.19 and 5.20, we obtain
state because it gives a relation
between state variables
1 ∂S (W )
(macroscopic variables that describe =
the thermodynamic state of a T ∂W
system). Equations of state can be
derived from the fundamental kB  W 1  W 
=− ln − ln  N −  
thermodynamic relation    N   
d W = Td S − P d V + µ d N ,
 W  (5.21)
where P is pressure, μ is chemical kB   
potential, and N is the number of = − ln  
 W
particles. In particular, dW can  N − 
equivalently be written as 

kB 
 ∂W   ∂W  1 
dW =  dS +  dV
 ∂S  V ,N  ∂V  S ,N =− ln  N .
  − 1
W 
 ∂W 
+ dN ,
 ∂N  S ,V Solving for W yields

where the subscripts following the N

W =  . (5.22)
derivative terms denote that those
terms are kept constant. Comparing e kBT
+1
the two expressions, we obtain the
equations of state
Because W = Nh, the fraction of hairpin sites which contain hairpins can be writ­
ten as
 ∂W   ∂W  Nh 1
T = , P = − ,
 ∂S  V ,N  ∂V  S ,N = ε . (5.23)
N
e kBT
+1
 ∂W 
and µ =  .
 ∂N  S ,V
The number of hairpins at equilibrium is dependent
on temperature
Equation 5.23 gives the fraction of sites that contain a hairpin when the polymer
is at thermal equilibrium (Figure 5.11). We see that at very high temperatures,
the fraction asymptotically approaches the value of Nh/N = 0.5. Based on what
we know about the entropy for this polymer, this makes intuitive sense. In par­
ticular, we know that at thermal equilibrium, the polymer will seek a state that
minimizes its free energy, and that in the high temperature limit, the free energy
is dominated by the entropy term. The free energy will be minimized when
entropy is maximal and Nh = N/2, because entropy is maximal when W = 0.5N.
 MiCroCanoniCal EnsemBle 135

Figure 5.11 Fraction of sites

containing a hairpin as a function

containing hairpins (Nh /N )

of temperature. At low temperature,
energy dominates the behavior, and very

fraction of sites
few hairpins occur. At high temperature,
entropy dominates, and hairpins occur
1
– more frequently. In the limit of infinitely
2 high temperature, half of the sites
contain a hairpin.

temperature T

What happens when you decrease the temperature? In the limit of T = 0, we
know that
Nh 1
lim = lim  = 0. (5.24)
T →0 N T →0
e kBT
+1
Because there are no hairpins, W = 0. In other words, in the low-temperature
limit, the free energy is dominated by the internal energy, and the free energy will
be minimized when internal energy is zero. This occurs when no sites contain a
hairpin (for instance, when the polymer is “straight”). The transition between
these two limits is sigmoidal, as can be seen in Figure 5.11.

Equilibrium obtained via the microcanonical ensemble is

identical to that obtained via free energy minimization
In the last section we used the microcanonical ensemble to calculate S(W), and
then used thermodynamic relations to predict the behavior of the polymer when it
is at thermal equilibrium at some constant temperature. Recall that the principle of
free energy minimization states that a closed system at constant temperature will
spontaneously adapt itself to lower its free energy, with thermodynamic equilib­
rium attained when free energy is minimized. This suggests that the equilibrium
behavior obtained in Equation 5.23 using the equation of state Equation 5.20 should
be identical to that obtained via free energy minimization. To confirm this, we must
find the free energy as a function of the number of hairpins that occur. Recall
Equation 5.18, in which we found the dependence of entropy on Nh. Because we
also know the dependence of internal energy on Nh (see Equation 5.14), we can
write an expression for the free energy as a function of the number of hairpins

Ψ( N h ) = N h − kBT ( N ln N − N h ln N h − (N − N h )ln( N − N h )). (5.25)

Now that we have an expression for the free energy in terms of the number of hair­
pins, we can compute the number of hairpins at equilibrium by finding the value
of Nh that minimizes the free energy. To do this, we take the derivative of the free
energy with respect to Nh and set this equal to zero,
∂Ψ ∂W ∂S
= −T = 0. (5.26)
∂N h ∂N h ∂N h
Analyzing the internal energy term in this expression, we find
∂W
= . (5.27)
∂N h
136 CHApTER 5: Statistical Mechanics Primer

For the entropy term,

∂S  Nh 
= −kB ln N h − ln ( N − N h ) = −kB ln  .
∂N h  N − N h  (5.28)

Plugging Equations 5.27 and 5.28 into W and solving for Nh/N,
Nh 1
=  , (5.29)
N
e kBT
+1
which is equivalent to Equation 5.23.

5.5 CanoniCal EnsemBle

In the last section, we derived an expression for the number of hairpins in our
model polymer when it is at thermal equilibrium. Specifically, we used the micro­
canonical ensemble to obtain an expression for S(W) and used this relation in
conjunction with thermodynamic relations to derive an expression for the aver­
age energy/number of hairpins at thermal equilibrium. In this section, we seek to
derive this same expression using an alternative approach for the calculation—
the canonical ensemble. In the canonical approach, we consider an ensemble of
microstates associated with a system at constant temperature. Although both the
microcanoncial and canonical ensembles are theoretical constructs that can be
used to perform the same calculations, the latter may be considered a more “natu­
ral” method for analyzing many problems, because experiments are typically per­
formed at constant temperature rather than constant energy. In addition, we will
see that unlike the microcanonical approach, the canonical approach permits
analysis to be performed in a manner that does not necessitate the enumeration
of the density of states, which can be mathematically difficult.

Canonical ensemble starting from the microcanonical

ensemble
We now derive the relevant relations of the canonical ensemble. In particular, we
derive two important relations, Boltzmann’s distribution and the partition func-
tion. Later, these relations will be used to analyze our model polymer.
Consider a system of interest contained within a heat bath (Figure 5.12). The sys­
tem is in thermal contact with the heat bath, such that it is maintained at constant
temperature (in other words, the bath and the system of interest are in thermal

Figure 5.12 Schematic

demonstrating the system of
interest, heat bath, and total system of
system. In the canonical ensemble, the interest
system of interest and heat bath are in
thermal contact, and we consider
combinations of system of interest and total system
bath microstates such that the energy of
the total system is fixed. heat bath
 CanoniCal EnsemBle 137

equilibrium). Both the system of interest and heat bath have associated micro­
states and internal energies. If we were to consider a polymer (the system of inter­
est) placed in a very large chamber filled with an ideal gas (the heat bath), a system
microstate would be a particular configuration of the polymer, and a bath micro­
state would be a particular state of the gas in which each gas particle had a par­
ticular position and velocity. Both microstates would be associated with a
particular value of internal energy, one for the polymer, and one for the ideal gas.
Let the subscript “s” denote the system of interest and “b” denote the bath. In
addition, let Qs(ms) be the internal energy of the system microstate ms, and
Qb(mb) be the internal energy of the bath microstate mb. If we idealize the system
and bath together as a single entity, we can consider an ensemble of different
combinations of system and bath microstates such that the total internal energy
(the sum of the internal energy of the system and the bath) is Wtot. In this man­
ner, by considering an ensemble of microstates with a fixed value of energy, this
approach is similar to the microcanonical ensemble example presented previ­
ously in this chapter.
Consider the case in which the system is in microstate ms. The system’s internal
energy is Qs(ms), and the internal energy of the heat bath is Wtot − Qs(ms). An
important realization (and a key step in the derivation) is that the probability for
the system to be in microstate ms is proportional to the number of heat bath
microstates with energy Wtot − Qs(ms). This can be expressed as

(
p(ms ) ∝ Ω b Wtot − Qs (ms ) , )
(5.30)

where Ωb is the bath density of states. We can write the right-hand side in the
above expression as a function of entropy. In particular, S = kB lnΩ, so the right-
hand side of Equation 5.30 can be written as
Sb(Wtot −Qs (ms ))

Ω b (Wtot − Qs (ms )) = e kB
. (5.31)

Combining Equations 5.30 and 5.31, we find
Sb(Wtot −Qs (ms ))

p(ms ) ∝ e kB
. (5.32)

Next, require the temperature of our system to remain constant. We can accom­
plish this by making the assumption that the heat bath is very large, such that
changes in energy and entropy of the system do not affect the energy and entropy
of the bath. Mathematically, this implies that for any system microstate ms, Qs/
Wtot  1, allowing us to perform a Taylor expansion. In particular, for very small
displacements Δx between two points x and x0, it can be shown that a “suitably
smooth” function f(x) can be approximated as

f ( x ) ≈ f ( x 0 ) + ∆xf ′( x ), (5.33)

when Δx/x0  1.
In our case, we can Taylor-expand Sb (assuming x0 = Wtot, x = Wtot − Qs, and Δx = −
Qs) using Equation 5.33 as

∂S b
( )
Sb Wtot − Qs (ms ) ≈ Sb (Wtot ) −
∂W
Qs (ms ). (5.34)

But, because at equilibrium,

∂S b 1
= , (5.35)
∂W Tb
138 CHApTER 5: Statistical Mechanics Primer

where Tb is the temperature of the heat bath, Equation 5.34 can be rewritten as
Qs (ms )
Sb (Wtot − Qs (ms )) = Sb (Wtot ) − . (5.36)
Tb
Combining Equations 5.32, 5.36, we finally find
Sb (Wtot ) Qs (ms ) −Qs (ms )
− 1 (5.37)
p(ms ) ∝ e kB kBT
= e kBT
,
Z
where in the right-hand-most relation in Equation 5.37 we have turned the pro­
portionality for the probability into an equality by introducing a normalization
factor,
−Qs (ms )

Z= ∑ ms
e kBT
. (5.38)

Note that the summation in Z is over all possible discrete system microstates ms,
and that we have lumped the exp(Sb(Wtot)/kB) term within Z, as it does not depend
on ms.
Although we have thus far only considered discrete microstates, we will often con­
sider systems that possess a continuous distribution of microstates. In this case, Z
can equivalently be written as
−Qs (m s )

Z=
∫
ms
e kBT
dms . (5.39)

The probability distribution in Equation 5.37 and normalization factor in
Equations 5.38 and 5.39 are the key relations of the canonical ensemble. Equation
5.37 is called Boltzmann’s distribution, and Equations 5.37 and 5.38 are called the
canonical partition function.

Probability distribution from the canonical ensemble gives

Boltzmann’s law
The probability distribution in Equation 5.36, Boltzmann’s distribution (or
Boltzmann’s law), implies that for the system described in the canonical (con­
stant temperature) ensemble, the probabilities of the individual microstates
are not evenly distributed. Instead, the probability goes as the inverse expo­
nential of the energy of that microstate. Interestingly, Boltzmann’s law can be
applied to an amazingly diverse set of systems. For example, it can be used to
predict the distribution of electrons under an electric field, or the diffusion of
molecules or proteins experiencing Brownian motion. As long as the system is
at thermal equilibrium with and can exchange energy, but not mass, with
its surroundings, the microstates will be distributed following a Boltzmann’s
distribution.
It is worthwhile discussing the differences between discrete and continuous prob­
ability distributions, such as those that appear within the sum and integral in
Equations 5.38 and 5.39, as they are associated with different quantities. In par­
ticular, in the discrete case, the probability distribution in Equation 5.38 gives the
probability of microstates with energy Q, while in the continuous case, it gives
the  probability of a range of microstates within a specified interval. Therefore,
unlike the discrete distribution in Equation 5.38, we cannot evaluate the probabil­
ity of a specific microstate in Equation 5.39. Instead, we can calculate the proba­
bility of a range of microstates.
 CanoniCal EnsemBle 139

Example 5.2: Boltzmann distribution (discrete)

Consider a large number of cells migrating on the sur­ Knowing Z, we can now compute that the fraction of
face of a culture dish. Empirically, we observe that the cells with zero velocity is simply the total number of cells
migration velocity of the cells, in an appropriate fixed times the probability that the cell will have zero velocity,
time interval, can exhibit a Boltzmann distribution, with
the energy Q equal to a constant α times the migration 1  0  1
p(v = 0) = exp   = = 0.904
velocity (in Section 10.1, we will show that this may be a Z  kBT  1.106
reasonable model in some cases, as this form of energy
arises due to energetic effects associated with adhesion
1  −α  0.097
energy and viscous dissipation). For simplicity, assume p(v = 1) = exp  = = 0.087
that the velocities can only take discrete values of 0, 1, or Z  kBT  1.106
2 μm/s. Let the energy be constant α = 10−20 W/μm and
assume the temperature is 37°C. Determine the proba­ 1  −α  0.009
p(v = 2) = exp  = = 0.008.
bility of observing a cell migrating at 0, 1, or 2 μm/s. Z  kBT  1.106
Because this is a discrete problem, we first calculate Z
These results show that a Boltzmann distribution is
using the summation:
heavily weighted toward the lower end, and there is a
2 exponential drop-off in probability as you increase the
 −αv 
Z= ∑ exp k T  = 1 + 0.097 + 0.009 = 1.106
v =0
B
energy level being measured. Note that the sum of the
probabilities is not exactly 1, due to rounding.

Example 5.3: Boltzmann distribution (continuous)

Now consider the same plate of cells with migration Now, we can compute the following probabilities
velocities that follow a Boltzmann distribution. However,
1
in this case, assume that the velocities are not discrete 1  −αv  0.386
values, but instead range continuously between 0 and
2 μm/s. We cannot compute the probability of observing
p(0 ≤ v ≤ 1) =
Z ∫ exp k T  dv = 0.424 = 0.912
B
v =0

a cell migrating at a specific velocity, but rather within

2
some specified range. Compute the probability of 1  −αv  0.037
observing a cell with migration velocity in the range 0–1
μm/s or 1–2 μm/s.
p(1 ≤ v ≤ 2) =
Z ∫ exp k T  dv = 0.424 = 0.088.
B
v =1

First, we need to compute Z,

You can see that the probabilities when using the contin­
2 uous distribution decay similarly with energy as com­
 −αv  −kBT   −2α  
Z=
∫ exp k T  dv =
B α   exp  
 kBT 
− 1 = 0.424.

pared with the case using the discrete distribution.
v =0

The free energy at equilibrium can be found using

the partition function
We now turn our attention from the Boltzmann distribution to the other impor­
tant relation obtained from the canonical approach, the partition function Z from
Equations 5.37 and 5.38. At first glance Z looks to be just a normalization constant
for the Boltzmann probability. However, it is an extremely valuable quantity that
can be used to calculate equilibrium behavior. The partition function is related to
the free energy and can be used to calculate the free energy at equilibrium without
having to enumerate the density of states.
Consider the case in which we have a system of interest as defined in the canoni­
cal ensemble. We know that the Boltzmann distribution gives the probability
p(ms) of a microstate with microscopic energy Q(ms). To show that this distribu­
tion is related to the free energy, consider the probability of the system having
140 CHApTER 5: Statistical Mechanics Primer

internal energy W at thermal equilibrium. We shall denote this probability p(W).

We know that for a macroscopic value of W, the system may have multiple micro­
states. The probability of the system having a macroscopic internal energy W can
be calculated as the total number of microstates with energy Q = W, multiplied by
the associated Boltzmann probability,
−W
1 kBT (5.40)
p(W ) = e Ω(W ).
Z
But because

Ω(W ) = e S kB = e ST kBT
, (5.41)

we can see that Equation 5.40 is related to the free energy as

− (W −TS )
1 1 −βΨ (5.42)
p(W ) = e kBT
= e ,
Z Z
where β = 1/kBT and Ψ is the free energy. Equation 5.42 implies that analogous to
how the Boltzmann distribution implies that the probability of a microstate goes as
the inverse exponential of its microscopic energy the probability of a macrostate
(which can be associated with a different number of microstates) goes as the
inverse exponential of its free energy. With this relation in hand, we are now able to
show that Z is related to the free energy.
We previously showed that Z can be obtained by integrating the Boltzmann distri­
bution over all microstates, as in Equations 5.38 and 5.39. But we also know that
for Equation 5.42 to yield the correct probability for p(W),

∫
Z = e − β Ψ(W )dW

(5.43)

or in the discrete case, Equation 5.43,

Z= ∑e − βΨ(W )
. (5.44)
W
Collectively, our analysis shows that Z can be computed in two distinct ways,
using Equation 5.39 or Equation 5.43. The main difference between these two
expressions is that in Equation 5.43, the exponential term contains the free energy
Ψ instead of the microstate energy Q, and that the integral is over all macroscopic
energies W instead of microstates ms.
For very large systems, integrals or sums of the form in Equation 5.43 and Equation
5.44 can be approximated using the saddle point method, as long as the quantity
being integrated or summed over is an extensive quantity. An extensive quantity is
one that scales linearly with the number of particles in the system, whereas an
intensive quantity is one that does not depend on the size of the system. For
instance, mass is an extensive quantity, while temperature is an intensive quan­
tity. Free energy is an extensive quantity, so the above expression can be approxi­
mated using saddle point integration. In particular,
−Ψmin

Z =e kBT
, (5.45)

where Ψmin is the free energy evaluated at which Ψ(W) is a minimum. Rearrang­
ing this expression,

Ψ min = −kB ln( Z ). (5.46)

 CanoniCal EnsemBle 141

Equation 5.46 has important implications, in that it reveals that the partition func­
tion has greater utility than just serving as a normalization constant for the
Boltzmann distribution. In particular, we know that for closed systems, the free
energy is minimized at thermal equilibrium. Therefore, Equation 5.46 implies
that to compute this value of free energy at equilibrium for a given system, one
merely needs to compute the partition function using Equation 5.38.

Advanced Material: Saddle point approximation can be used to approximate

summations of certain exponential quantities

Statistical mechanics deals with determining the over­ summation. It also implies that for integrals of the
all  behavior of systems consisting of many bodies and/ form
or  degrees of freedom. Within treatments in statistical
­mecha­nics, it is often useful to make mathematical approx­
imations when the number of degrees of freedom is very ∫
S = e Ny ( x )dx ≈ e Nymax

large. For example, in statistical mechanics, we often have

to find summations of exponential quantities of the form for very large N. Specifically, y(x) is a function that is
bounded at ± ∞ independent of N, and ymax is its global
n maximum value. Computing the integral in this manner
S= ∑e
i =1
yi N
, is referred to as saddle point integration.
Now, define a new function z(x) = Ny(x). In other words,
where N is a large number and yi is real. Let ymax be the z is an extensive quantity.
maximum value of yi. Because yiN is in the exponential, In this case,
as N is increased, the sum will very quickly become
dominated by the term corresponding to yi = ymax. In S = e zmax .
fact, it can be shown that for a very large N,
Thus, if z(x) is an extensive quantity, we can approximate
S ≈ e Nymax . S using saddle point integration, as long as N is suffi­
ciently large.
In other words, instead of performing the summation,
one simply needs to calculate the largest term in the

The internal energy at equilibrium can be determined using the

partition function
In the last section, we saw that the partition function can be used to calculate the
free energy at equilibrium. We show here that it can also be used to compute the
internal energy. This is demonstrated in the following example. Let’s say we want
to calculate the internal energy W of the system at equilibrium. W can be calcu­
lated by summing over each of the microstate energies Q(m), multiplied by the
probability p(m) of that microstate occurring:

∑ p(m)Q(m) = Z ∑ e
1 − βQ (m )
Q = Q(m), (5.47)
m m
where β = 1/(kBT). Note that in Equation 5.47, the brackets denote the expected
value based on an average over all microstates. Equation 5.47 can equivalently be
written as

∂ 1 ∂ 1 ∂Z
∑ ∂β e ∑e
1 − βQ (m ) − βQ (m )
=− =− =− . (5.48)
Z Z ∂β Z ∂β
m m
142 CHApTER 5: Statistical Mechanics Primer

But we know that the partial derivative of lnZ is defined as

∂ ln Z 1 ∂Z
= . (5.49)
∂β Z ∂β

Therefore,
∂ ln Z
W = Q =− . (5.50)
∂β

This shows that, like the free energy, the internal energy at equilibrium can also be
found from the partition function. Again, we are able to find the equilibrium inter­
nal energy without having to enumerate the density of states.
We now know that if we calculate the partition function, we can calculate Ψ and
W at equilibrium. If we want to know S, it can be calculated easily, because
Ψ = W − TS. Thus, we can specify the thermodynamic state of the system at equi­
librium from its microscopic behavior, as long as we can calculate the partition
function Z. In the next section, we revisit our model polymer to demonstrate use
of these relations.

Using the canonical approach may be preferable for analyzing

thermodynamic systems
Let us reconsider the non-interacting two-level system (the hairpin polymer), but
this time, analyze it using the relations derived from the canonical ensemble. We
know that the energy Q of a given microstate is
N

Q= ∑n ,
i =1
i (5.51)

where ni equals one if a hairpin occurs at hairpin site ni, or zero if a hairpin does
not occur. Combining the microstate energy in Equation 5.51 with the Boltzmann
distribution, we obtain an expression for the probability of a given microstate in
thermal equilibrium,
N

1 − β  ∑ ni (5.52)
p(n1 ,n2 ,..., nN ) = e i =1 ,
Z
where Z is the discrete partition function given in Equation 5.37
N
−β  ∑ ni
Z= ∑e i =1
. (5.53)
m
To compute Z, we need to perform a sum over all possible microstates. To do this,
we perform N summations over each individual hairpin site,
N N
−β  ∑ ni 1 1 1 −β  ∑ ni
Z= ∑e
m
i =1
= ∑ ∑ ...∑ e
n1 = 0 n2 = 0 nN = 0
i =1
. (5.54)

The summation in the exponential in the rightmost relation in Equation 5.52 can
be written out explicitly as
1 1 1

Z= ∑ ∑ ...∑ e
n1 = 0 n2 = 0 nN = 0
− β n1 − β n2
e ... e − β nN . (5.55)

 RanDom WalKs 143

Now, because n1, n2, . . . nN are independent variables, we can separate the sums
1 1 1

Z= ∑
n1 = 0
e − β n1 ∑
n2 = 0
e − β n 2 ... ∑e
nN = 0
− β nN
.

(5.56)

In Equation 5.56, each of the sums are numerically equivalent. Therefore, can
rewrite the above expression as
N
 1

Z =

∑e − β n


= zN , (5.57)
n =0
where
1

z= ∑e
n =0
− β n
= 1 + e−β  (5.58)

Nota Bene
is termed the single partition function. With the partition function, we can calcu­
Partition functions for systems
late the free energy at equilibrium as of non-interacting particles can
be calculated from single
ln Z N ln(1 + e − β  ) (5.59) partition functions. In Equation
Ψ=− =−
β β
5.57 we saw the partition function
was calculated as
and the equilibrium internal energy as Z = ZN,
∂ ln Z e−β  N where Z is the single partition
W = = Nε −β 
= β . (5.60) function. This will be the case for any
∂β 1+ e e +1
system of N identical particles that do
not interact. For instance, in the case
Note that Equation 5.60 is identical to Equation 5.22, which we derived through of the hairpinned polymer, the N
the calculation of S(W) in the microcanonical ensemble (in conjunction with hairpin sites are identical (they all can
thermodynamic relations). Although we performed the same calculation in two undergo the same range of motion)
different ways, we can see that mathematics were quite different, with the use of and they do not interact (whether
the canonical approach avoiding the need to use combinatorics to compute the one site contains a hairpin is not
density of states. For this reason, the canonical ensemble may often be preferable influenced by whether a hairpin
occurs elsewhere.)
for analyzing thermodynamic systems.

5.6 RanDom WalKs

We now leave our discussion of the equilibrium behavior of our model polymer
and discuss another classical topic of statistical mechanics: the random walk.
Random walks are a large class of mathematical problems in which at discrete
points in time a “walker” moves in space a discrete distance. They have application
to a wide range of problems in fields as diverse as physics, chemistry, computer
science, biology, and even economics. In this section we will introduce them in
their most basic form with some simple analysis. We will also show how they relate
to another fundamental statistical process: diffusion. Note that we will revisit the
concept of random walks in Section 7.4 in our discussion of polymer mechanics.

A simple random walk can be demonstrated using soccer

Consider a soccer player standing at midfield on a soccer field, facing one of the
goals. We define midfield to be at location r = 0. The player takes out a coin, and
flips it. If it comes up heads, the player steps forward a distance b toward one goal;
if it comes up tails, the player steps backward toward the opposite goal. This “flip
and step” process is repeated a certain number of times. We will now calculate the
probability of the player being at a certain location after a given number of steps.
In particular, after n random, b-sized steps, what is the probability that the player
144 CHApTER 5: Statistical Mechanics Primer

will end up at location r? Note that if the player has taken more backward steps
than forward steps, r is negative, and vice versa.
If the player takes n steps, and for each step they can only move backwards or for­
wards, then there are 2n different paths that can be taken (although the player may
end up in the same position with many different paths). The number M of those
paths in which only n+ of the steps are forward can be calculated using the bino­
mial coefficient from combinatorics. To see why the binomial coefficient can be
used, imagine that the “objects” are the coin flips, and that the “picks,” or the
selection of an object, are how many heads (or positive steps) occur. Then, if there
are n total flips and n+ of them come out heads, M must be

 n
M =  . (5.61)
 n+ 

Our goal is to calculate the probability of the player being at point r after n flips
and steps. We have the following relationships, n = n+ + n− and r = b(n+ − n−),
where n− is the number of backward steps. Combining these yields
r
n+
b. (5.62)
n+ =
2
And,

 n 
  n!
M (n , r ) =  n + r  = .
b n+ r r (5.63)
 n−
 2  b! b!
2 2
The probability that the player will end up at position r after n steps (p(n, r)) is
simply the number of ways or “paths” the player could use to get to point r divided
by the total number of paths the player might take. M is the total number of ways
to get to r and, as we noted above, there are 2n total paths available to the player.
Therefore, the probability is

M (n , r ) 1 n!
p(n , r ) = = n .
2n 2 n+ r n− r
(5.64)
b! b!
2 2
Equation 5.64 is the exact expression for the probability that the player will be at
a position r on the field after n steps of size b. To get a feel for what the random
walk distribution looks like for different step numbers, we calculate p(n, r) for
various values of n in Figure 5.13. Note that the appearance of the distribution
begins to resemble a normal distribution as n gets larger. This is perhaps not
unexpected, as the position r results from the sum of several independent coin-
flipping events, with the probability of heads or tails the same at each step (in
other words a 50/50 chance). The central limit theorem states that for a suffi­
ciently large number of steps, r will be approximated by a normal distribution,
with this approximation becoming better as the number of steps are increased.
This is an important result, as in many cases we are interested in the behavior of
random walks in the limit as n becomes very large. We can rewrite Equation 5.64
in the limit of large n in the form of a normal or Gaussian distribution,
1 2
2 nb 2
p(n , r ) = e −r . (5.65)
2πnb 2

Equation 5.65 gives the probability distribution for a one-dimensional random
walk for n steps of length b in the limit of large n; the full derivation for Equation
 RanDom WalKs 145

n=1 n=2 n=3 n=4

1.0 1.0 1.0 1.0

0.8 0.8 0.8 0.8

0.6 0.6 0.6 0.6

p(n,r )

p(n,r )

p(n,r )

p(n,r )
0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0

–1 1 −2 0 2 −3 −1 1 3 −4 −2 0 2 4

n=8 n = 16
1.0 1.0

0.8 0.8

0.6 0.6
p(n,r)

p(n,r)

0.4 0.4

0.2 0.2

0.0 0.0
−8 −6 −4 −2 0 2 4 6 8 −16−14 −12−10 −8 −6 −4 −2 0 2 4 6 8 10 12 14 16

Figure 5.13 Parametric plots of Equation 5.65. As n is increased, the distribution can be increasingly approximated as a normal or
Gaussian distribution. For simplicity we have assumed b=1.

Nota Bene
5.65 is in Section 7.4. It can be shown that for a random walk of n steps of size b,
the root mean-square position is Recurrence in random walks is
dependent on the dimension of
r2 = b n. the walk. Although it is beyond our
(5.66)
scope here, it can be shown that the
random walk example we have used
The derivation of Equation 5.66 is left as an exercise for you. Combining Equation here is recurrent. This means that
5.65 and 5.66, we obtain an alternative expression for the random walk probability eventually, the player will visit every
distribution, point on the playing field, even if the
field is infinitely wide. In the context of
1 −r 2 / r 2 games of chance, this property of
p(n , r ) = e . (5.67)
2π r 2

recurrence is sometimes called the
Gambler’s Ruin. The idea is that a
gambler’s funds at any point in time
The diffusion equation can be derived from the random walk can be considered analogous to the
soccer player’s position on the field.
At the molecular level, the process of diffusion is governed by the random motion The flip and step of the soccer player
of particles. It is perhaps not surprising that the behavior of a random walk scaled is the same as each spin of the wheel,
up to the continuum level can be described by the diffusion equation. This rela­ roll of the dice, or hand of cards. No
tionship was first demonstrated by Albert Einstein in a 1905 paper he published matter how large the gambler’s initial
while working in the Swiss patent office. He demonstrated that Brownian motion, bankroll (so long as it is finite), he or
she will eventually go bust, assuming
the random motion of small particles suspended in a fluid, results in a rate of dif­ the bets are always of the same
fusion related to the velocity of the individual particles. His theoretical results fit amount, even if the odds were fair
the observed data so well, that they provided convincing support for the atomic (instead of favoring the house, as it
theory of matter before there was direct evidence of the existence of atoms or usually does). Interestingly, and
molecules. perhaps surprisingly, the property of
recurrence of random walk processes
The one-dimensional diffusion equation is known as Fick’s second law (this law is can be shown to be limited to one-
covered in more detail in Chapter 9.3), and is given as and two-dimensional situations only.
In three dimensions (or higher), the
dC d 2C walker will not visit all available
=D 2, (5.68) coordinates. Random walks of these
dt dx
types are known as transient.
146 CHApTER 5: Statistical Mechanics Primer

where C is the concentration and D is the diffusion coefficient. We now show that
this equation can be derived from a random walk process.
First, let’s recast our statement of the random walk problem slightly. Consider a
particle that is constrained to move in only one dimension. We define its initial
position to be at r = 0. After some time t, we are interested in the probability that
the particle will end up at some new location r. In recasting our random walk, we
replace the number of steps n with time t, which we will initially treat as a discrete
variable. In this case, we are able to assess where the particle is at discrete time
points. In regards to the particle position, we will treat r as a continuous variable.
Thus, the particle can move arbitrary distances. To allow for steps of any distance,
not just steps to the right and left of a specified increment (for instance, we do not
use a fixed step size b), we need to introduce another probability function, p(x).
This function gives the probability distribution for a particle making a step (posi­
tive or negative) of distance x in any time step. We can now state the new problem
in terms of the probability distribution of finding the walker at a certain position,
r, at time t + 1 in terms of the distribution at time t. Specifically,
∞

p(t + 1, r ) =
∫ p(x)p(t , r − x)dx. (5.69)
−∞
This type of relationship is known as a recursion relationship. It defines p at one
point in time in terms of its distribution at the prior time. Next we can perform a
Taylor series expansion for the p(n, r − x) term in the integral about r to obtain
∞
 
∫
dp(t , r ) 1 2 d 2 p(t , r )
p(t + 1, r ) = p( x ) (pt , r ) − x + x + O 3 ( x ) dx
 d r 2 dr 2 
−∞

∞ ∞ ∞
dp(t , r ) 1 d 2 p(t , r )
=
∫
−∞
p( x )p(t , r )dx −
∫
−∞
p( x ) x
dr
dx +
2 ∫
−∞
p( x ) x 2
dr 2
dx + O 3 ( x ),

(5.70)
where O3(x) represents terms of third order or higher in x. Because p(t, r) and its
derivatives do not depend on x, these terms can be removed from the integrals. In
this case, Equation 5.70 can be rewritten as
∞ ∞ ∞
dp(t ,r ) 1 d 2 p(t ,r )
∫
p(t + 1, r ) = p(t ,r ) p( x )dx −
−∞
dr ∫
−∞
p( x ) x dx +
2 dr 2 ∫ p(x)x dx + O (x).
−∞
2 3

(5.71)
Next, we seek to simplify each of the terms on the right-hand side of Equation 5.71.
For the first term, we know that probability distributions possess the property that
integrals over their entire range must be unity. Therefore, for the first term,
∞

∫
p(t ,r ) p( x )dx = p(t ,r ). (5.72)
−∞
The second term in Equation 5.71 can be simplified if we restrict the probability
distribution, p(x), to be isotropic. In other words, there is no bias in the probabil­
ity of a particle to make a step in one direction or another, p(x) = p(−x). This con­
dition also implies that there is no net motion of the particles, owing, for example,
to convection should there be flow in the underlying medium. This restriction
implies that the integral in the second term is zero, thus
∞
dp(t ,r )
dr ∫ p(x)xdx = 0. (5.73)
−∞
 Key ConCepts 147

Finally, we assume that the higher-order terms from the Taylor series expansion
can be neglected, which can be shown to be valid for some small region around r
as long as p is a continuous function. The net result is that
∞
1 d 2 p(t ,r )
p(t + 1,r ) − p(t ,r ) =
2 dr 2 ∫ p(x)x dx.
2
(5.74)
−∞
Now, consider the right-hand side in Equation 5.74. It is the integral over all pos­
sible positions of the particle of the position squared times the probability of being
in that position. This is the definition of the average squared position, a term that
we encountered earlier in this chapter. In Section 7.4, we show that for a random
walk of n steps with fixed step size b, 〈r2〉 = nb2 (see Equation 5.66). It can be simi­
larly shown that for a random walk with variable step size but with an average
squared displacement of Δr2 per step,

r 2 = n∆r ,
2
(5.75)

where n is the number steps. Combining Equations 5.74 and 5.75, we get

t ∆r 2 d 2 p(t , r )
p(t + 1, r ) − p(t , r ) = . (5.76)
2 dr 2
Now, in the limit of a large number of particles, all behaving according to this
equation, the concentration of particles at a given location is proportional to the
probability of finding a single particle at that position. Therefore, we can replace
P(n,r) by the concentration C. Also, if we divide the equation by Δt, the left-hand
side becomes the time derivative of concentration in the continuous limit and
we have

dC n∆r 2 d 2C
= . (5.77)
dt 2∆t dr 2

This is, in fact, the one-dimensional diffusion equation with

n∆r 2 (5.78)
D= .
2∆t
We will build on our treatment of random walks in subsequent chapters. We will
use the random walk to better understand the behavior of polymers in Section
7.4. We will also learn how confining diffusion into a two-dimensional mem­
brane can greatly enhance the kinetics of diffusion-limited biochemical reaction
in Section 9.1.

Key Concepts

• Within statistical mechanics, entropy is directly behavior, whereas the influence of entropy increases
defined as S = kB lnΩ. Ω is the density of states gives with increasing temperature. The competing nature of
the number of microstates associated with a given energy and entropy in equilibrium behavior is
value of a macrostate. captured by a thermodynamic potential called the
• Thermal equilibrium of thermodynamic bodies is free energy, defined as Ψ = W − TS.
governed by two competing phenomena: internal • The principle of minimum total potential energy states
energy, and entropy. At low temperatures, internal that when a structure is subject to mechanical
energy–driven phenomena dominate a body’s loading, it shall deform in such a way as to minimize
148 CHApTER 5: Statistical Mechanics Primer

the total potential energy. The principle of minimum probability of a given microstate is given by the
free energy states that a closed system at constant Boltzmann distribution. The partition function is a
temperature will spontaneously adapt itself to lower normalization factor for the Boltzmann
its free energy. Just as potential energy minimization distribution, and is related to the minimum value of
can be used to find the equilibrium state of free energy. We can specify the thermodynamic state
mechanical systems, free energy minimization can be of the system at equilibrium by calculating the
used to find the equilibrium state of thermodynamic partition function without enumerating the density
systems. of states.
• Different analytical approaches can be used to • Random walks are a class of mathematical problems
determine behavior of thermodynamic systems at in which at discrete points in time a “walker” moves
equilibrium, such as the microcanonical ensemble in space a specified distance. Through its use we can
and the canonical ensemble. In the microcanonical relate certain macroscopic behavior of
ensemble, we analyze microstates at constant energy. thermodynamic bodies to microscopic properties.
Thermal equilibrium can be obtained from relations • When the behavior of random walks is scaled up to
derived from the microcanonical ensemble. the continuum level, their behavior is described by
• In the canonical ensemble, we analyze microstates in the diffusion equation.
a system held at constant temperature. The

Problems

1. We know that the mechanical behaviors of everyday Assume that the amount of this bias is a small number,
rod-like objects, such as steel beams and jump ropes, . Now, what is the probability of each macrostate?
can be predicted through the principle of minimum
total potential energy, but not soft structures such as 3. Consider a strand of DNA which only contains two
biopolymers, whose behavior is significantly influenced different nucleotides, A or T. Using the binomial
by entropy. We know that a transition between energy- coefficient, calculate the number of different possible
and entropy-dominated behavior occurs when W = TS. sequences if the strand is N nucleotides long, and only
Given the statistical mechanics definition of entropy, N1 of the nucleotides are T.
we can estimate a characteristic energy at which this
4. Consider 100 people in a building. Every time a person
transition occurs as kB (why would this be the case?).
moves up one floor, it takes them 1kBT of energy.
Calculate this value of energy assuming a temperature
Assume that given enough time the people distribute
of 37°C.
themselves according to Boltzmann’s distribution.
Now, consider a length of rope and a strand of DNA. How many people are on each floor on average? How
Calculate the strain energy stored in a jump rope tall does the building need to be to accommodate this
bent 180 degrees with a constant radius of curvature average distribution of people? How tall would it need
of 1 m, assuming it can be modeled as an elastic to be if there were 10 people? And if there were 1000
cylindrical beam with a Young’s modulus of 100 MPa, people?
and a radius of 0.5 cm. Do the same for a strand of
5. Spring is in the air and a mosquito is in your apartment.
DNA with the same curvature. Again, assume that the
You are watching it fly and land on a set of stairs.
strand of DNA can be modeled as an elastic cylindrical
Assuming that it needs 1kBT of energy to fly up from
beam, assuming a Young’s modulus of 1.9 GPa and a
one step to the next, what would a Boltzmann’s
radius of 1 nm. How do these values compare to the
distribution predict for the probability of finding it on
characteristic transition energy?
any given step, if there are 10 steps in total?
2. Two dice are rolled, each of which can generate
6. Consider a molecule that has a “home” position and a
a number between 1 and 6. Consider this a
linear restoring force that is proportional to the distance
statistical system. How many microstates are there for
that it is away from this position. In other words, the
this system? How many macrostates are there? What is
molecule acts like it is attached to the tip of a spring.
the density of states, Ω, for each macrostate? What
The restoring force is F = −kx, where x is the distance
is the probability of each macrostate if the dice are
from the home position and k is the spring constant.
rolled many times?
The potential energy is a function of position and can be
Now suppose that the “fair” dice are substituted with expressed as
a pair of “loaded” dice. In the loaded dice there is a
small weight placed under the 1 spot. This makes the
1 2
probability of getting a 6 slightly higher than 1/6, and W = kx .
the probability of getting a 1 slightly lower than 1/6. 2
 AnnotateD ReferenCes 149

We can treat each molecular position of the molecule, 8. Imagine the same box as in Problem 7, except the
x, as a microstate of the system. In this case, assuming partition is constructed differently. Specifically, it does
thermal equilibrium, the probability of finding the not slide; however, energy (for instance, heat) is allowed
molecule in any given position is given by the to transfer freely between the two compartments.
Boltzmann distribution. What is the average energy? How would you expect the temperatures of the two
Hint: a table of integral identities may be useful. gases to be related at equilibrium? Formalize this
Note: your finding is known as the principle of result by showing that entropy is maximized when the
equipartition of energy and holds for the energy temperatures of the two gases are equal.
associated with any parameter that has an associated
9. Consider a mechanosensitive ion channel that can be in
energy that varies with the square of the parameter. For
one of two states: open or closed. Define the energy of
example, it also works for kinetic energy, which varies
the channel using a configuration parameter σ that can
with the square of speed.
be equal to 0 (closed) or 1 (open). The energy is
7. In this problem, we show that for an isolated system,
equilibrium is attained by maximizing the entropy (this E = σε open + (1 − σ )ε closed − τ∆A
is the second law of thermodynamics and is equivalent where εopen and εclosed are the energies of the channel
to stating that equilibrium is attained by minimizing in the open and closed configurations, τ is the tension
the free energy, assuming constant energy). Imagine applied to the channel, and ΔA is the change in area
a box that is isolated from the external environment, of the channel in going from closed to open. Write
meaning that the total energy of the contents within expressions for the partition function, the probability of
the box is constant. The box contains a partition the channel being in an open state, and average energy.
that divides it into two subspaces. The partition is
What is the probability that the channel is open,
constructed such that it can slide freely along the length
assuming that τ = 1, 2, 3, 4, and 5 pN/nm, Δτ = −5kBT,
of the box, similar to a plunger in a syringe; however,
and ΔA = 10 nm2?
it does not allow energy to transfer between the two
subspaces. We now fill both sides of the partition with 10. Consider a two-dimensional model polymer composed
some amount of gas. The partition slides around until of n = 4 rigid segments, with neighboring segments
it comes to rest at its equilibrium position. How would connected via a rotating joint. Assume that the angle
you expect the pressures to be related at equilibrium? between adjacent links θ can only take discrete values:
Now formalize this intuition, by showing that entropy 0° (hairpin), 90°, or 180° (straight between links).
is maximized when the pressures are equal. Hint: we Calculate the value of the density of states for two
know that the total entropy of the system is different end-to-end lengths R = 2 and R = 4. Calculate
the entropy for these same values of R. Which value of
Stot = S1( N 1 ,V1 , E1 ) + S2 ( N 2 ,V2 , E2 ) R is entropically more favorable?

You will need the thermodynamic definition of pressure 11. Consider the “90° polymer” depicted in Figure 5.6. Write
to complete the derivation, a MatlaB module that analyzes a polymer made up of
N segments of unit length and which determines the
p  ∂S  number of configurations possible for each end-to-end
=  . length between 0 and N.
T  ∂V  E ,N

Annotated References

Berg HC (1993) Random Walks in Biology. Princeton University Press. Pande VS (2006) Graduate Statistical Mechanics. (lecture notes
This book, first released in 1983 and expanded ten years later, is an from course number Chem 275, Stanford University, Stanford, CA.)
excellent source for examples of random walks in biology, including The general structure of this chapter is based on a one-semester
diffusion. course on graduate statistical mechanical developed by Dr Pande
at Stanford University. The treatments of specific topics in this chap-
Chandler D (1987) Introduction to Modern Statistical Mechanics.
ter, including enumeration of states via the three-coin system, and
Oxford University Press. A concise introduction to elementary sta-
analysis of the two-level noninteracting system via the microcanon-
tistical mechanics.
ical and canonical ensemble, are based on his course notes, which
Dogic Z, Zhang J, Lau AWC et al. (2004) Elongation and fluctuations are currently in development as a formal text.
of semiflexible polymers in a nematic solvent. Phys. Rev. Lett. 92,
Phillips R, Kondev J & Theriot J (2009) Physical Biology of the Cell.
125503. Reports the formation of hairpin defects in actin filaments
Garland Science. This is an excellent text that gives many examples
(similar to that in our model polymer) suspended in anisotropic
of insights that can be made into biological phenomena through
solutions of aligned rod-like macromolecules.
statistical mechanics. (Note: Problem 6 is adapted from the text of
McQuarrie DA (2000) Statistical Mechanics. University Science Books. Section 5.5.2 in Physical Biology of the Cell; Problem 7 is adapted
A widely used text in introductory statistical mechanics courses. from Section 5.5.2; Problem 8 is adapted from Section 7.1.2.)
This page intentionally left blank
to match pagination of print book
CHAPTER 6
Cell Mechanics in the
Laboratory

I n scientific investigation, the shortcomings of experimental techniques are

often the limiting factor in our ability to gain a better understanding of the
principles underlying phenomena of interest. It comes as no surprise, then,
that the development of novel experimental approaches can have profound
effects on advancing our understanding of cell mechanics and mechanobiol-
ogy. Engineering principles have long been used to better understand the role
of mechanics in regulating biological processes; however, our understanding
of the mechanics of cells has generally lagged behind that of tissues. This is due
in part to the technical challenges associated with investigating mechanical
behavior at the small length scales associated with cells. In this chapter, we pre-
sent an overview of some common experimental approaches for mechanically
manipulating cells and investigating whole-cell mechanical behavior. In par-
ticular, we describe experimental techniques for cell probing, such as magnetic
micromanipulation, atomic force microscopy, and optical trapping. We also
describe approaches for quantifying traction forces in cells and analyzing the
mechanics of cells under physiological mechanical stimuli such as matrix
deformation and fluid shear stress. Finally, we conclude this chapter with a
brief discussion of experimental design. Our goal is to provide a basic under-
standing of how each technique works and approaches for analyzing data
obtained thereby (and, for the interested reader, resources for learning about a
particular technique in more detail). Although we give relatively brief descrip-
tions of each technique, the contribution of these techniques to the field of cell
mechanics should not be underemphasized. Indeed, the use of these
approaches led to most of the insights in cell mechanics and mechanobiology
presented in this text.

6.1 PROBING THE MECHANICAL BEHAVIOR OF

CELLS THROUGH CELLULAR
MICROMANIpULATION
We begin this chapter with techniques for cellular micromanipulation. To better
understand why such techniques could be useful for investigating the mechan-
ics of cells, consider a situation in which we are given a sample of some unknown
material, and asked to measure its stiffness. How could we do this? One approach
is to subject the sample to mechanical loading of known magnitude and meas-
ure the deformation. For instance, if we sought to characterize the stiffness of a
spring, we could hang a weight on it and measure its elongation. Or, if we had a
homogenous, isotropic, linear elastic material, we could determine its Young’s
modulus by measuring its strain under a known stress. But what if we wished to
characterize the mechanical stiffness of an individual cell? Micromanipulation
approaches allow for application of known mechanical forces on the small scale
of the cell.
152 CHAPTER 6: Cell Mechanics in the Laboratory

Known forces can be applied to cells through the use

of cell-bound beads and an electromagnet

Nota Bene In magnetic micromanipulation, localized forces are applied to cells by adhering
ferrous beads, typically on the order of micrometers in diameter, to the cell sur-
Magnetic beads. The name
face, and then applying electromagnetic forces to the beads. In some cases, the
“magnetic beads” is commonly used
to refer to beads used in magnetic beads are introduced into the cell interior by endocytosis. An electromagnet can
micromanipulation experiments. The be constructed by winding wire around a metal core. Shaping the tip of the core to
beads themselves are not usually a chisel (typically, a few hundred micrometers across) results in a strong magnetic
magnetic; they are typically iron- field that is generally laterally homogeneous with a strength that depends only on
bearing beads. distance from the tip.
A representative system for magnetic micromanipulation is shown in Figure 6.1.
Cells that have been seeded in a dish and that have had ferrous beads bound to
their surface are placed on a temperature-controlled stage of an inverted micro-
scope. An electromagnet is held above the dish with the tip immersed in the media.
A cell with a bead attached is located under the microscope, and the tip of the mag-
net is positioned near it. Current is passed through the electromagnet, generating
a magnetic field that causes the bead to be pulled toward the magnet tip. The force
applied is determined by the distance from the bead to the magnet tip. Bead move-
ments are determined microscopically (Figure 6.2) with particle-tracking algo-
rithms (see Section 6.2). To study the behavior of specific molecules, so-called
functionalized (surface modification is explained at the end of Section 6.2.) beads
can be used that have been coated with the protein of interest, which can engage
specific receptors of interest on the cell surface.

The dependence of force on distance from the magnet

tip can be calibrated through Stokes’ law
In magnetic micromanipulation experiments, the force applied to the bead is a
function of the distance to the tip. This relationship is determined in a calibration
procedure. One approach is to suspend the beads in a high-viscosity fluid. When the
magnet is activated, the beads will reach a terminal velocity, the constant velocity at
which the drag forces on the beads balance the magnetic forces pulling on them.
The terminal velocity is related to the force applied on them through Stokes’ law,
F = 3πµDV , (6.1)

Figure 6.1 A magnetic

micromanipulator mounted on a
microscope stage. The electromagnet
is wound with copper wires and placed
in a micromanipulator (in back, on the
right side of the image). The tip is
immersed in the dish containing cells;
the dish is placed on a temperature-
controlled unit. (From, Huang H, Kamm
RD, Lee RT (2004) Am. J. Physiol. Cell
Physiol. 287).
 PROBING THE MECHANICAL BEHAVIOR OF CELLS THROUGH CELLULAR MICROMANIpULATION 153

Figure 6.2 The bead is the dark round object left of center, with the lines drawn in
to make it easier to visualize bead motion. The shadow on the right is the
electromagnet tip. Cells are visible in the background, but typically during experiments the
light is increased to white out the cells and obtain clearer tracking of the bead.
Nota Bene
Constant force is applied to the
magnetic beads. Because force is
where F is the force, μ is the dynamic viscosity, D is the bead diameter, and V is dependent on the distance from the
the terminal velocity. By performing this procedure on beads at various distances bead to the magnet tip, the force
from the electromagnet tip, one can estimate the bead force as a function of dis- applied to a cell-bound bead actually
tance. When using this approach, it is important that flow remain laminar, as increases as the bead is pulled in the
Stokes’ formula is only applicable at very low Reynolds numbers (see Section 4.2). direction in the magnet. In most
cases, the bead displacement is
small enough that the change in
Magnetic twisting and multiple-pole micromanipulators force is can be neglected.
can apply stresses to many cells simultaneously
Although magnetic micromanipulation is useful for investigating the mechanical
behavior of single cells, it is often desirable to apply mechanical loads to many
cells simultaneously. Magnetic twisting is an alternative method that uses mag-
netic beads to apply forces to many simultaneous beads. An electromagnet is
used to magnetize beads temporarily in one direction with a powerful pulse. This Nota Bene
is quickly followed by a pulse in a perpendicular direction, causing a twisting
movement to be applied to the beads to align them with the new field (Figure Magnetic bead types. In magnetic
micromanipulation experiments,
6.3). As the beads twist, their individual magnetic fields add together to produce having the bead lose its
a net magnetic field that is rotated from the original orientation. A magnetome- magnetization is desirable if repeated
ter, which is simply a wire coil attached to a sensitive amplifier, is used to meas- measurements are being taken. That
ure the magnitude of the rotated magnetic field. It is possible to obtain the is, you want to apply some force to
relationship between torque and rotation, in an average sense, for hundreds of the bead, release it, wait some time,
beads simultaneously. and then apply the same force. Here
you want to use paramagnetic beads,
which lose most or all of their
Optical traps generate forces on particles through transfer magnetization once the magnetic
of light momentum field breaks. A paramagnetic material
is also used to construct the
Light is another way to impart forces to cell-bound beads. Perhaps you recall that electromagnet so that you can
quickly control the applied magnetic
photons are massless, but nonetheless carry momentum. Light is able to impart
field. Materials that retain a
momentum to matter with which it interacts. This fundamental property of light is significant portion of the
exploited with high-power lasers to build optical traps (also termed optical magnetization are referred to as
tweezers). ferromagnetic. In general,
paramagnetic materials respond
To understand how an optical trap works, consider an transparent particle at the more weakly to magnetic fields, but
center of a focused laser beam (Figure 6.4). The index of refraction of the particle their ability to de-magnetize is better
is higher than the surrounding environment, such that as light passes through the than ferromagnetic materials.
particle, the beams change direction or bend. The particle is redirecting the light
beam and therefore changing the light’s momentum. By conservation of momen-
tum, there must be a force acting on the particle. The bending of the beams leads
to a force gradient that pushes the bead toward the focal point and into the center

Figure 6.3 Schematic of magnetic

twisting cytometry. After a strong
magnetometer magnetization in the horizontal direction,
a weaker magnetic pulse is applied in
the vertical direction, resulting in an
upward torque on the beads. The extent
of bead motion is measured by a
before after during magnetometer and is reflective of local
magnetization magnetization twist cell properties.
154 CHAPTER 6: Cell Mechanics in the Laboratory

Figure 6.4 Schematic of an optical laser beam

trap. The bead is trapped in a laser
beam, with the equilibrium point in the
center of the beam just downstream of
the focal region. If the bead is displaced,
then there is a net restoring force on the
bead. Attaching the bead to a cell or
other structure allows one to test its
mechanical properties. (Adapted from,
Huang H, Kamm RD, Lee RT (2004) Am. J.
Physiol. Cell Physiol. 287). bead
force

objective

of the beam, where the intensity is highest. There is also a ­scattering force that
Nota Bene results in a force along the direction of the laser beam. The result of these com-
bined forces is that the particle becomes “trapped” within the focused laser beam
A few key figures of the optical
in a location just downstream of the focal point. If the bead is displaced relative to
trap. In the late 1960s, Arthur Ashkin
from Bell Labs pioneered the use of the beam, the forces will act in a restorative manner to move the bead back into
light to manipulate the position of the trapping location, creating a stable equilibrium point.
tiny particles of matter. Eventually
this developed into modern-day
optical trap systems. It is also the
Ray tracing elucidates the origin of restoring forces in
basis of the work of Steven Chu, who optical tweezers
used it to cool atoms by selectively
trapping slow-moving ones, allowing To quantify the origins of the lateral restoring forces, we can use a technique called
faster-moving ones to escape. Dr. Chu ray tracing. It is useful for determining such forces, as long as the particle is large
received a Novel Prize for Physics in compared with the wavelength of light. Consider a spherical bead trapped within
1997 and became United States a laser beam, as in Figure 6.5. When the beam is unfocused it resembles a cylin-
Secretary of Energy in the Obama der. We further assume it to have a constant intensity in the axial direction, but a
administration.
radially decreasing intensity profile in the lateral direction, with the highest inten-
sity in the center.
If we follow ray 1 in Figure 6.5, it enters the left side of the bead and is deflected
rightward. In this rightward deflection, there is transfer of momentum from the
beam to the bead that imparts a force on the bead that is primarily leftward (F1).
Ray 2 is located opposite ray 1 and equivalently imparts a primarily rightward
force (F2). Because ray 2 is equal in intensity to ray 1, F2 is equal in magnitude to
F1, resulting in no net lateral force, if the bead is centered. However, if we displace
the bead to the left, the intensity of ray 2 will be less than that of ray 1 owing to the
radial gradient of light intensity. In this case, the magnitude of F1 is greater than
that of F2, resulting in a net rightward restorative force that pushes the bead back
to the beam center. Because the forces are the result of the intensity gradient, they
are referred to as gradient forces.
Although the lateral gradient force example demonstrates why the bead becomes
trapped in the lateral direction, what about the axial direction? To understand this
phenomenon, consider a focused laser beam as in Figure 6.6. The beam forms a
cone that converges at the focal point. If we follow a ray entering the left side of a
spherical bead located upstream of the focal point, it is deflected rightward upon
entering the bead, and again upon exiting the bead. There is a change of momen-
tum as it enters and exits the bead, because the light entering the bead is not
purely vertical, the shift of the rays toward the centerline means that the exiting
 PROBING THE MECHANICAL BEHAVIOR OF CELLS THROUGH CELLULAR MICROMANIpULATION 155

Figure 6.5 A bead situated within a

laser beam with highest intensities
in the beam center will experience
forces toward the beam center.
Additionally, the bead will experience
forces downstream.

1 2 1 2

F1 F2 F1 F2

net net
force force

maximum
intensity

light has a larger vertical momentum component than the entering light. By sum-
ming the difference in momentum vectors, one can see that there is a net vertical
change in momentum, which is transferred to the bead as an axial force in the
downstream direction. A similar analysis reveals a net vertical change in momen-
tum in the upstream direction if the bead were downstream of the focal point,
resulting in an upstream restorative force. When the lateral and axial restoring
forces are combined with the downstream scattering force, the result is that the
bead becomes trapped just downstream of the focal point.

What are the magnitudes of forces in an optical trap?

We now have gained some intuition about the orientation of the restoring forces
in optical tweezers that facilitate particle trapping. But what determines the mag-
nitude of forces imparted? In general, they are determined by the physical proper-
ties of the bead, the beam, and the medium in which the bead is immersed. Two
different approaches can be used to calculate these forces, depending on the size
of the particle.

Figure 6.6 A focused light beam

interacting with a bead. A bead that
net moves down in the trap will deflect light
force beams toward the bottom, resulting in
an upward net force. Conversely, a bead
moving up will deflect the beams up and
experience a downward restoring force.

net
force
156 CHAPTER 6: Cell Mechanics in the Laboratory

The generalized Lorenz–Mie theory is a ray optics–based model, and a relatively

accurate approach when the wavelength of light is much smaller than the bead
diameter. The mathematics are complicated and beyond the scope of this text.
In the case of a bead smaller than the wavelength of light, a much simpler approach
based on Rayleigh scattering theory can be used. The forces can be separated into
two components: scattering force (in the laser direction) and the force gradient
(the in-plane force), respectively. Using the Rayleigh approximation, the scatter-
ing force when the particle is trapped is
2
 nb 
− 1
128π 5 nm I Rb6  nm 
Fz = 2 , (6.2)
3 c λ  nb
4

 n + 2 
m

where I is the intensity of the trapping light, nb and nm are the indices of refraction
of the bead and medium respectively, a is the radius of the bead, c is the speed of
light, Rb is the radius of the bead, and λ is the wavelength of the light. Upon inspec-
tion of the above expression, we can see that given the scaling of F with a, a small
increase in particle size can yield a large increase in force.
The second component is the gradient force,

 n 2 
Rb3   b  − 1
n I   nm  
F1 = 2π m . (6.3)
c   nb  2

  n  + 2
 m 

The gradient force is dependent on the gradient of the light intensity and not the
absolute intensity. For a typical experiment, the forces are relatively small (on the
order of tens of piconewtons, although some modern configurations can achieve
higher forces).
Recall that the generalized Lorenz–Mie theory and Raleigh scattering can be
used when the wavelength of the beam is much smaller and larger than the beam
diameter, respectively. Often the sizes of particles of interest may be of the same
order of magnitude as the wavelength. For example, green light has a wavelength
of 510 nm; most cellular targets of interest for trapping range from a few microm-
eters to 250 nm. Particles much smaller than this are very hard to trap, because
they are difficult to image using standard optical techniques and generate much
smaller forces. Therefore for many particle sizes and beam wavelengths, the rela-
tionship between force and particle/beam properties is not very well under-
stood. In this case, forces generated by the trap can be determined empirically.
One approach is to capture a bead of known size, and move the trap at increasing
velocity until the capture is lost. Quantifying the fastest velocity before the bead
is lost, allows the maximum force generated by the trap to be calculated using
Stokes’ law.

How does optical trapping compare with magnetic

micromanipulation?
Although optical trapping is similar to magnetic micromanipulation in that both
allow for the manipulation of particles within or bound to cells, there are important
differences between the two that may make one approach more appropriate for a
particular application. In optical trapping, the trapped particle need not be ferrous,
and spatial manipulation of particles can generally be performed in a more precise
manner. Certain customizations allow the independent control of multiple beads
simultaneously. Also, some subcellular organelles and other structures like chro-
mosomes, vesicles, bacteria, and even viruses have the correct optical properties
 PROBING THE MECHANICAL BEHAVIOR OF CELLS THROUGH CELLULAR MICROMANIpULATION 157

that they can be manipulated directly without the need for beads. However, for
some biomechanical studies, optical trapping may not be the best method, as the
forces are generally low. Additionally, force-based manipulation with optical trap-
ping requires feedback technology, because forces are generated based on the rela-
tive position of the bead and the laser beam. Another important difference is that an
optical trap, as the name suggests, is a “trap.” That is, there is a relatively small equi-
librium point and a large gradient of restoration forces around that point. The forces Nota Bene
generated magnetically affect a far greater spatial region and are more homogene-
ous. This allows many beads to be manipulated together, although the manipula- The technical challenges of
bead-based studies. Although
tions are not independent. Finally, it is not clear whether magnetic fields have bead-based studies can lead to
effects on biological tissues, but it is well known that focused lasers can heat biologi- powerful insights, there are important
cal material, and with sufficient power can destroy tissue locally. In fact, this effect technical challenges and concerns
has been exploited to generate what are called laser scissors or laser scalpels, allow- that need to be kept in mind. For
ing one to disrupt parts of a cell selectively. instance, there can be significant
variability in bead–cell adhesiveness.
Often, bead–cell adhesion tends to
Atomic force microscopy involves the direct probing of objects improve slowly over time with the
with a small cantilever bead eventually being internalized to
the cell. Therefore, it may be difficult
Atomic force microscopy (AFM) is an approach for characterizing cells that allows to distinguish the mechanical
for the direct probing of the cells, using a small cantilever. There is no need to behavior of the cell from the
differences in adhesion. Indeed, the
attach a bead or other foreign substance to the cell surface. AFM works by tracking
very act of adhesion may induce an
the displacements of a cantilevered beam brought into contact with the cell or unexpected biological response.
other object of interest. Modern designs are able to measure both the shape of a
cell and its mechanical properties. If one were to imagine a finger as an AFM can-
tilever, it is easy to see how a range of physical attributes of an unknown object
could be inferred, such as mechanical stiffness (by pushing down on the object
and determining how much resistance the object offers), surface geometry (by
running one’s finger over the surface), and spatial heterogeneity in mechanical
stiffness (by probing or poking in several places).
The central component of an atomic force microscope is a cantilever that has a
small tip at the free end. The cantilever and tip are typically constructed of silicon
or a silicon compound. A typical tip has a radius of curvature on the order of
nanometers, which is quite sharp, even at the scale of a cell. In probing a speci-
men, the cantilever is lowered so that the tip contacts the cell surface. As the tip
pushes into the specimen, the cantilever deflects. We saw in Section 3.2 (see
Equation 3.31) that for the case of a linear elastic beam of length L clamped at one
end, EI is the flexural rigidity, the displacement w of the cantilever at any point x
along the beam is a function of the force at the tip, F,

F  x3 x2 
w= − L . (6.4)
EI  6 2 

We can see from Equation 6.4 that the deflection at the tip (x = L) is propor-
tional to the tip force. The displacement at the tip can be related to the force
through an effective spring constant. Once this constant is determined, the
force being applied at the cantilever tip can be readily calculated from the tip
deflection, allowing for simultaneous quantification of deflection and force. In
this way, force-displacement curves can be generated by positioning the tip
over the surface of the specimen and measuring the deflection of the cantilever
as the tip is lowered into the specimen (this mode of AFM use is sometimes
referred to as force readout mode). Such curves are conceptually similar to the
force-deformation curves we discussed in Section 3.2 (see Figure 3.1) and can
be used to infer mechanical properties.

Cantilever deflection is detected using a reflected laser beam

An AFM cantilever is about the same size as a cell (on the order of tens of micro­
meters), and its deflections are smaller (tens of nanometers). These tiny deflections
can be greatly amplified and measured using a geometric trick. A laser is focused on
158 CHAPTER 6: Cell Mechanics in the Laboratory

Nota Bene a reflective area on the back of the cantilever and the lateral motion of the beam is
measured a large distance away (Figure 6.7). A detector is placed in the path of the
Hertz contact. Heinrich Hertz reflected beam, such that a laser spot is generated on the detector. As the cantilever
introduced the solution for
bends, the laser spot on the detector changes its location. In general, the longer the
frictionless contact between two
spheres in 1881 by assuming that the laser beam path, the more the laser spot will move for a given deflection of the can-
contact force could be represented tilever, and the smaller the deflections are that can be detected. Generally, force
by an integral of point forces applied sensitivities using AFM are on the order of piconewtons or less.
over a small patch or region of
contact. With this solution, the
contact between a sphere and a Scanning and tapping modes can be used to obtain cellular
plane was determined by letting the topography
radius of one of the spheres
approach infinity. Similar approaches In addition to characterizing mechanical properties and behavior, AFM can be
can be used to determine the contact used for specimen visualization, in particular for determining surface topogra-
force due to a conical or cylindrical phy. There are several modes whereby this can be achieved. Scanning mode is
indenter.
used to generate a topographical profile of the surface being examined. Here, the
tip is lowered until it just touches the surface of the specimen, then the tip is
scanned, or dragged, across the specimen. The deflection of the cantilever at each
point is used to generate a height profile. To prevent specimen damage, in many
cases a feedback mechanism is incorporated to lower/raise the tip to maintain a
constant force. If the material is homogeneous, then the deflection is g­ enerally a
good representation of the surface geometry. For specimens with heterogeneous
mechanical properties, the topography may be reflective of the surface topogra-
phy as well as the local stiffness. If we were to consider a specimen that had a flat
surface but spatially heterogeneous stiffness, a stiffer region may appear “higher”
than a more compliant region.
A second AFM mode is tapping mode, whereby the tip is actively vibrated up and
down (typically by a piezoelectric actuator coupled to the AFM tip holder) close
to its resonant frequency, which is usually on the order of 10–100 kHz. The oscil-
lating tip is lowered to the specimen surface until it begins to experience van der
Waals forces, resulting in a reduction of the magnitude of the vibration. Similar to
Nota Bene scanning mode, during tapping mode operation, feedback is typically used to
Half-space. A half-space is a
maintain the cantilever oscillation amplitude constant. A major advantage of
mathematical abstraction. It is refers tapping mode over scanning mode is that when the tip contacts the surface, it
to region of space that is infinitely has sufficient oscillation amplitude to overcome “sticking” or adhesion forces
deep and wide, but has a flat surface. between the tip and the sample. It also reduces the chances that a tip will become
It is a reasonably good assumption fouled or coated in cellular debris because, ideally, the tip never comes into
for a cell, because on the scale of an physical contact with the cell. One important consideration with AFM is the
AFM tip, the surface of a cell is nature of the probes, which are sometimes unable to detect overhangs or very
virtually infinite.
steep walls.

A Hertz model can be used to estimate mechanical properties

Once we have obtained an AFM force–displacement profile, the next question is
how can mechanical properties of the sample be inferred? One mathematical
approach is to use the Hertz contact model, which describes contact between a

Figure 6.7 Schematic of AFM

detector detector
operation. The tip at the end of the
cantilever is brought down on a
specimen. As the specimen indents, the
cantilever flexes and deflects the laser laser laser
to a different location as determined by
the detector. It is possible to calculate cantilever
the indentation force using the cantilever
information about cantilever flexure
from the laser spot deflection. (Adapted
from, Huang H, Kamm RD & Lee RT
(2004) Am. J. Physiol. Cell Physiol. 287.) cell cell
 PROBING THE MECHANICAL BEHAVIOR OF CELLS THROUGH CELLULAR MICROMANIpULATION 159

Figure 6.8 Schematic of the

indenter. It is depicted here as a sphere
R of radius R, but as long as the contact
region is circular, the rest of the probe
can have an arbitrary shape. The relation
between d, R, and the material
d properties of the surface is given in
Equation 6.5, with the indenter having
infinite rigidity. The material being
E, ν indented, shown here as a dented
rectangle, is treated as having infinite
depth and width, which is called a
half-space.

rigid sphere and a homogeneous, isotropic, linearly elastic material half-space

Nota Bene
(Figure 6.8). The Hertz solution is
The Hertz model can be used to
4 E 
calculate elastic properties for
F=  3
 Rd , (6.5) specific locations. In many cases,
3  1 − ν2   the assumptions underlying the Hertz
  model may not be adequately met
within cells. For instance, the
where F is the applied force, d is the depth of indentation of the sphere into the structural components of cells may
material, E is the elastic modulus of the specimen, ν is Poisson’s ratio of the mate- not be linearly elastic or isotropic. In
rial, and R is the radius of the sphere/indenter. This mathematical relationship addition, due to the spatial
heterogeneity of cells, a single elastic
assumes that the indenter is very stiff compared with the sample.
modulus is rarely sufficient to capture
One can approximate a pyramidal or other “sharp” AFM tip as a sphere in the the spatial variation in material
Hertz model if the scale of indentation is small enough that the tip’s curvature is properties. Some studies have
focused on using the Hertz model to
evident. Additionally, in some experiments, the cantilever can be customized by
calculate elastic properties for
attaching a bead to the tip. This way, beads of different diameters can be used to specific locations and/or structures,
obtain stiffness at different length scales, and sometimes with different forms of such as stress fibers.
stress (e.g., shear vs normal stress).

Example 6.1: Estimating the elastic modulus of a cell using the Hertz model
Using the Hertz model, estimate the elastic modulus of a you can only obtain about two significant figures) yields
cell using υ = 0.5 (incompressible) and υ = 0.3 (typical of 75 kPa for incompressible materials and 90 kPa for bio-
many biological tissues). The radius of the probe tip is logical tissues. You can see that the choice of Poisson’s
50 nm. ratio does not make a huge difference in our estimate,
which is admittedly on the high side (cells typically in
Using Figure 6.9, we estimate the force for a 100 nm
the  10–30 kPa range by AFM) if you are not targeting
indentation (from 175 nm to 75 nm along the tip
stress fibers.
approach line) to be 1000 pN = 1 nN. Using the formula
E = 100,000(1 − υν) (actually a bit more, but realistically

Figure 6.9 Sample plot of an AFM force-

distribution

displacement curve. The lighter curve shows

the approach of the tip above the sample (an
endothelial cell), moving from right to left. Contact
2000 is made at around 150 nm. Further displacement
force (pN)

200 300 400 to the left shows increasing force to deform the
rupture force (pN) cell, such that at 1 nN, the tip has advanced about
1000 75 nm into the cell. The darker line is the retraction
of the tip. At about 175 nm, the adhesion of the tip
to the cell is broken and there is a sudden
0 “zeroing” of the force. The force magnitude was
collected for a number of cells, and the histogram
of that adhesive-rupturing force is shown in the
−1000
inset. (From, Liu J, Weller GE and Zern B (2010)
0 100 200 300 Proc. Natl. Acad. Sci. USA, 107).
distance (nm)
160 CHAPTER 6: Cell Mechanics in the Laboratory

6.2 MEASUREMENT OF FORCES pRODUCED

BY CELLS
In previous sections, we learned some techniques for applying and analyzing
mechanical forces through micromanipulation. Such forces are considered
exogenous, in that they are generated from outside the cell. In contrast, endog-
enous forces are those generated from within the cell. Endogenous forces can
play critical roles in regulating physiological processes such as cell spreading
or locomotion. In these situations, the cell generates traction forces, or endog-
enous contractions, that are transferred to the substrate through cell adhesion
sites. Traction forces can be clearly demonstrated by plating cells on a flexible
silicon sheet. The traction forces cause small wrinkles to develop near the sites
of adhesion (Figure 6.10). Although traction forces are relatively easily to visu-
alize, determining the magnitude and direction of the forces applied is more
difficult. The focus of this section is to detail two methods for quantifying
­cellular traction forces: traction force microscopy and the use of micropillar
arrays.

Traction force microscopy measures the forces exerted

by a cell on its underlying surface
Nota Bene
In traction force microscopy (TFM), the approach is to characterize the deforma-
The substrate is often a thin film
of a flexible polymer such as tion of a substrate induced by traction forces, and to use the deformation field to
polyacrylamide. Polyacrylamide is calculate the traction forces exerted at each point. Small beads placed within the
chosen not only for its flexibility and substrate act as displacement markers and facilitate the tracking of deformations
flexibility that can be adjusted over a and calculation of the displacement field. Typically, bead positions are deter-
wide range, but also because it mined with the cell attached to the substrate, and again after detachment of the
behaves like an isotropic linear cell using trypsin, a serine protease that degrades cell adhesions. An inverse prob-
elastic material. This greatly simplifies
the analysis. Other gels used in cell
lem is then solved to extract forces generated by the cell. We briefly describe these
culture, such as collagen gels, might two steps in the next sections.
be more representative of
physiological environments, but are Cross-correlation can be used for particle tracking
not generally isotropic and linearly
elastic. They additionally offer spaces An important requirement for TFM is the ability to track the motions of beads. In
for cells to crawl within the gel itself,
which can confound traction
this section we will briefly discuss the practicalities of tracking before fully con-
mapping. sidering TFM. Cross-correlation is a statistical measure that allows the position
of particles to be determined with great precision. Consider a set of two images of
a cell (with attached fluorescent beads) obtained before and after mechanical

Figure 6.10 Image of a cell plated

on a thin deformable membrane.
The cells exert traction forces and locally
wrinkle the membrane in a
demonstration that in order for cells to
spread out, they exert some level of
surface forces. (From, Harris AK, Wild P &
Stopak D (1980) Science 208.)
 MEASUREMENT OF FORCES pRODUCED BY CELLS 161

loading. The images contain unique local patterns of intensity that reflect the
shape and brightness of the fluorescent objects within that locale. These local
intensity patterns can be indentified with image correlation, and changes in posi-
tion and/or shape used infer the deformations that gave rise to these changes.
In an ideal scenario, a bead will be a perfect sphere, which appears as a perfect
circle on our image. We can then locate the center of the sphere, and thus the
bead, exactly. If we repeat the same measurement after the bead moves, we can
obtain the exact displacement vector of the bead. In the laboratory, however,
beads are not perfect spheres and the images will be limited by resolution, illumi-
nation, refraction, and other artifacts. Cross-correlation allows us to determine
locations, and therefore displacements, with high precision by using a mathemat-
ical algorithm to fit two profiles against each other.
To illustrate this approach, consider a cell seeded with k fluorescent beads (Figure
6.11). We obtain fluorescence images of each bead before, during, and after load-
ing. Each image can be represented by an array of numerical intensities that we
assume, for simplicity, are stored in a square array known as the image intensity
function, I(i, j) with 2n + 1 pixels in each direction such that
− n ≤ i , j ≤ n. (6.6)

The intensity pattern for a single bead would be an approximately circular

region of high intensity (the bead) surrounded by a dark region (the background)
(Figure 6.12). The intensity in the circular region is highest in the center and
drops off toward the edges of the bead. As the bead moves, the location of the cir-
cular distribution will shift within the array, but because the bead is rigid, the gen-
eral intensity pattern will not change. In essence, the bead image undergoes a
rigid-body translation. Our goal is to determine the magnitude of the translation.
One approach might be to track the movement of the brightest pixel. However, the
intensity peak can be rather broad, which makes this approach imprecise and
sensitive to noise. Imagine a large collection of pixels that are all near the peak
with similar high-intensity values but with some fluctuations. In this situation, the

Figure 6.11 Fluorescent beads

attached to cells. In this image a
phase-contrast image of cells has been
superimposed on an epifluorescence
image of beads that have been filled
with a fluorochrome. (From, Kwon R,
et al. (2007) J. Biomech. 40)
162 CHAPTER 6: Cell Mechanics in the Laboratory

Figure 6.12 The digital (A) (B) (C)

representation of a circular bead
can be used for quantifying its
displacement. (A) A fluorescent bead
appears as a circle of non-uniform
intensity. Because the sphere is thicker
near the center, there is more
fluorescence emanating from that
location, so the bead tends to be brighter
in the center and dimmer near the edges.
(B) When digitized, the circle is
approximated by pixels that reflect the
intensity. Note that the same intensity bead digitized bead image
pattern is displayed in the digitized
version: lighter near the center and
darker near the edges. (C) The bead itself
can be extracted as a pattern of (D) (E)
intensities called a template. (D) When the
bead moves, the pattern representing
the bead undergoes a rigid-body
translation in the image. (E) With the bead
at its new location within the image, the
location of the pixels representing the
bead has shifted, but the pattern (or
template) remains the same. In actuality,
the intensities will fluctuate over time,
owing to instrumentation and other
factors, but the rough pattern (brighter
near the center) will remain consistent.

Nota Bene
brightest pixel can jump from place to place dramatically. To address this prob-
lem, a statistical image correlation technique can be applied that is actually so
Image distortion. Although the lack precise it can determine the location of the bead with sub-pixel resolution.
of shape change in the intensity
distribution simplifies our analysis, To determine the translation in the intensity pattern between images, we first
methods exist for correlating intensity define a template image K(i, j). The template image is the ideal intensity pattern of
patterns even when they become a bead, which for simplicity we locate at (0, 0). This template could be obtained
distorted.
from the theoretical distribution for an ideal bead, or an actual bead could be
imaged before deformation. In either case, we assume that K(i, j) is 2m + 1 pixels
in each dimension, with m < n. With K defined, a statistical cross-correlation can
Nota Bene
be calculated as a measure of the degree of similarity between K and I on a pixel-
Image texture correlation. An by-pixel basis. A typical statistical correlation measure is
alternative approach for obtaining
discrete displacements that does not m m
require the use of markers such as
beads is image texture correlation.
C (i , j ) = ∑ ∑ I (i + x , j + y)K (x , y),
x =− m y =− m
(6.7)
This approach relies on the matching
of local image texture, in other words
the unique spatial variation in which is a matrix convolution of I and K. The resulting matrix, C(i, j) is known as
intensity surrounding a pixel of the cross-correlation field. It gives the degree of similarity between the intensity
interest, between an undeformed pattern surrounding the pixel (i, j) and K. The i and j that maximize C(i, j) give the
and deformed image. In particular, location at which the intensity pattern of I most closely matches that of K. This
pixels from a subset domain within location is assumed to be the “correct” location of the bead in I. If the bead is
the undeformed image are selected,
originally at (0, 0), then the distance of the bead displacement can then be calcu-
and used as the template for a
cross-correlation calculation within lated from i and j. The location of this peak in C can be computed with an accu-
the deformed image. If the deformed racy better than the size of a pixel by calculating the centroid of the
image is distorted, a transformation cross-correlation field,
(such as a stretch or shear) may need
to be applied to the template before
performing the cross-correlation ∑iC(i , j)
i,j
∑ jC(i , j)
i,j
calculation. The displacement is again ic = , jc = . (6.8)
the one that results in a maximum
in the cross-correlation.
∑C(i , j)
i,j
∑C(i , j)
i,j

 MEASUREMENT OF FORCES pRODUCED BY CELLS 163

Determining the forces that produced a displacement Nota Bene

is an inverse problem
Inverse problems. The field of
inverse problems originated with the
Using cross-correlation, we can determine displacement of beads. The problem
work of physicist Viktor
of determining the set of forces that produced a known displacement field is Ambartsumian, done while he was
known as an inverse problem. In addition to observed data (displacements in still a student in the 1920s. His initial
this case), an inverse problem requires a governing equation known as the for- paper went unnoticed for 20 years
ward model. We need a theory that gives the expected displacements for a given until the mathematics community
surface load. Luckily, an analytical solution exists, known as the Boussinesq developed it into the general field it is
solution, which gives the displacement at the surface of an infinite homogene- today. Conceptually, in each inverse
problem, there is a set of underlying
ous linear elastic half-space owing to a point-load at the surface. One issue that
governing equations that convert a
arises with the half-space assumption is that the substrates used in TFM are set of model parameters or
fairly thin (on the order of 100 μm). Thus, the half-space assumption may seem properties (input) into observable
somewhat suspect. However, in general the method produces acceptably accu- data (output). The inverse problem is
rate results as long as care is taken that the displacements are small relative to to deduce the input from the output.
the substrate thickness. Some typical examples are to
determine the distribution of density
Consider a set of n discrete displacement vectors ui=1,2,. . .,n determined at discrete within a planet from measurements
points xi=1,2,. . .,n. The displacement vector ui at point xi is assumed to arise as a of its gravitational field (using
result of the combined influence of a set of j discrete point force vectors fj located Newton’s law of gravity) or to
at another set of points xj. The Boussinesq solution G(r) for a single force can be determine the epicenter of an
earthquake from seismic waves
(using the wave equation).

(A) bead displacement vector Figure 6.13 Traction force

microscopy takes advantage of the
existence of the surface tractions
exerted by the cell. Plating a cell on a
deformable substrate with embedded
marker beads (A, B) will result in the cell
beads deforming the substrate. Releasing the
cell (using trypsin or waiting until the cell
has moved away) allows the original
location of the beads to be determined.
deformable
A displacement field and the traction
substrate force field can then be calculated (C).
(A, adapted from, Roy P, Raifur Z &
lo Pomorski P, (2002) Nature Cell Biol. 4.
co cell With permission from MacMillan
m Publishers Ltd., on behalf of Cancer
ot
io Research UK.; B, from Munevar S et al.
n
Biophys. (2001) Biophysical Society.)

(B) (C)

1.75 × 10–3
1.00 × 106
164 CHAPTER 6: Cell Mechanics in the Laboratory

used to relate the displacement ui to fj as

m
Nota Bene
Green’s function. The function G(r)
ui = ∑G(r ) f ,
j =1
ij j (6.9)
is also known as a Green’s function.
This approach allows a general
solution to be built up by adding where rij = xi − xj is a distance vector, and
many simpler solutions. It is similar to
a matrix or integral convolution. The 1 + v  (1 − v )r + vrx 
2 2
vrx ry
simple solution is the Boussinesq G (r ) = 3
. (6.10)
solution for a simple force. πEr  vrx ry (1 − v )r + vr 
2 2
y

In the above expression, rx and ry are the x and y component of r and r is | r |. E and
ν are Young’s modulus and Poisson’s ratio of the elastic substrate, respectively.
Computing the above expression for all n displacements gives a simultaneous
equation between m force vectors and n displacement vectors. The goal then
becomes to find the m force vectors that give rise to the n displacement vectors
(Figure 6.13). Although methods for solving the problem are outside the scope of
the text, there are nice reviews in the literature.

Advanced Material: Boussinesq solution

The French mathematician Joseph Boussinesq derived F

the solution of a force applied to the surface of a half-
space in 1885 (Figure 6.14). He started with the solu-
tion of a force applied to an infinite continuum. Then
he calculated the force along the z = 0 plane. He used
the principle of superposition to apply an equal and
opposite force to this to cancel it out, again using the
solution for a point force in an infinite continuum. Figure 6.14 Depiction of the Boussinesq problem of a
When an integral is constructed in this way to solve an force applied to a half-space. The force is applied to the
inhomogeneous differential equation subjected to top surface and the half-space extends infinitely within the
complex boundary conditions, the integrand is termed plane and in the depth.
a Green’s function.

Example 6.2: Traction force microscopy

A special case that does not require extensive numerical  (1 + v) 
modeling is a single particle that displaces owing to a 0
 πEd 
single point force. Specifically, suppose you have a G= 2 
.
force  F and displacement u separated by a distance d  0 (1 − v ) 
(Figure 6.15) and that the force and displacement act in  πEd 
the same direction. Our experimental displacement is
then u = [u, 0]. Our force vector is F = [F, 0]. We can cal- Because u = GF,
culate the Green’s function, (1 + v )
u= F.
πEd
(1 + v )  (1 − v )d + vd 
2 2
0
G= 3  2
. which allows us to solve for F based on our knowledge of
πEd  0 (1 − v )d 
u and d. In the more complex non-co-linear condition,
there would be y-contributions, and we would need to
The off-diagonal terms are zero because we place the invert G rather than having a single linear relationship.
origin at the location of applied force, so there is no
y-component to this problem. Note that the (1, 1) term However, in many traction-force situations, there is
can be simplified to just d2. Further simplification yields more than a single force to deal with. Generally, multiple
 MEASUREMENT OF FORCES pRODUCED BY CELLS 165

forces are active—all of unknown magnitudes, being convolutions in frequency space become multiplica-
applied at unknown locations, being exerted in tions. Fourier transforms should be used with caution,
unknown directions. Consider a bead exposed to two as boundary conditions in frequency space are not
different unknown forces. If the bead moves, it could be clear-cut. Typically, there are no traction forces outside
because of either force or both forces acting in different the cell, but many solutions using Fourier transforms
amounts. To constrain the problem, multiple beads are will generate forces near the border of, but typically out-
required. However, the problem then becomes mathe- side of, the cell.
matically cumbersome because each bead gets an equa-
tion u = GF for each force being applied. Furthermore, d
(A)
is generally unknown, because the force locations are
not proscribed.
bead
To address this there are sometimes logical places to
assume forces are being applied, such as points of cel-
lular adhesion that can be determined by immunocyto-
chemistry. A more general approach is to create a
regularly spaced grid for the locations of force applica-
tion. If the grid has high enough density, a fairly com-
plete picture of the force field can be obtained.
Mathematically, for each grid location a displacement u (B)
and a force F is known. From this one can construct G.
This approach involves large arrays of displacements,
positions, and components of G, for which optimization
u F
routines are applied to extract the traction forces. This
technique is particularly sensitive to small errors in dis- d
placement because the force–displacement relation-
ship drops rapidly with distance away from the location
of the force. Another method is to use Fourier trans-
forms to convert the problem into frequency space. This Figure 6.15 A single force and single displacement
allows the inverse problem to be solved directly, because example. The force and displacement are co-linear.

Microfabricated micropillar arrays can be used to measure

traction forces directly
Traction forces can be quantified by solving an inverse problem. An alternative
method uses the process of microfabrication to create an array of thousands of
micropillars that essentially act as microscopic force sensors.
Microfabrication is the process of fabricating small structures with dimensions
on the orders of micrometers. Particular processing steps—many of which were
developed in the semiconductor industry for the manufacturing of integrated
circuits—are performed in sequence, resulting in a micromachined structure
with desired designs, surface patterns, or topography. Typically, the approach
involves three processing steps: thin-film deposition, photolithography (which
transfers a defined pattern onto material film using light), and etching (the
removal of material in patterns defined through photolithography).
Using these techniques, dense arrays of micropillars (on the order of 1–10 μm in
diameter and 10–100 μm in length) can be manufactured, which allows each pil-
lar to function as a force transducer by acting as a cantilever beam. By treating the
top surfaces of the micropillars with extracellular matrix proteins, cells seeded
onto the arrays adhere to the tops of the pillars and exert traction forces, causing
the pillars to bend (Figure 6.16). The deflection of the pillar tip can be related to
the force exerted on that pillar as in Equation 3.31

F  x3 x2 
w= − L . (6.11)
EI  6 2 

166 CHAPTER 6: Cell Mechanics in the Laboratory

Figure 6.16 An electron micrograph

of a cell sitting on a bed of
microneedles. The traction forces
exerted by the cell result in the pillars
being bent toward the center of the cell.
Because the forces exerted on the
pillars can be modeled by beam
equations, there is no need for the
computational complexity associated
with classic traction-force microscopy.
(From, Tan JL, Tien J & Pirone DM (2003)
Proc. Natl. Acad. Sci. USA 100.)

Because we only measure the displacement at the tip δ at x = L, the above expres-
sion can be simplified to

 3EI 
F =  3 δ. (6.12)
 L 

Example 6.3: Deflection of the micropillars in response to cell traction forces

In Figure 6.16, you can see the deflection of the micropil- The moment of inertia of a circular cross section is
lars in response to cell traction forces. Given that the I =  πR4/4 ≈ 4 × 10−24m4. Thus,
scale bar is 10 μm, the length of the pillars is 11 μm and
the elastic modulus E of the pillar material is 2.5 MPa, 3 (2.5 × 106 PA )(4 × 10 −24 m 4 )(5 × 10 −6 m)
F=
estimate the magnitude of the forces exerted by the 1.33 × 10 −15 m 3
periphery of the cell.
= 1.12 × 10 −7 N ≈ 100 nN.
The displacement of the pillar tips is roughly 5 μm. The
diameter is roughly 3 μm. So This is slightly on the high side but not ridiculously so
(the original article by Tan et al. reports values from 0 to
 3EI 
F =  3 δ. 80 nN).
 L 

Surface modification can help determine how a cell interacts

with its surroundings
Many of the techniques we have just described involve attaching or touching a syn-
thetic structure to the cell, either beads, AFM tips, micropillars, or other structures.
 AppLYING FORCES TO CELLS 167

However, cells typically do not encounter glass, plastic, or silicon in vivo. How the
cells interact with these artificial surfaces is an important consideration. The pro-
cess of surface modification can be a powerful way to modify the artificial material
not only to make the material more physiological or biocompatible, but also to gain
insight into the role of various properties or even specific proteins and molecules.
Techniques are available to alter roughness, charge, surface energy, hydrophobic-
ity, and other physical properties. A particularly useful approach is to functionalize
a surface, which is the process of coating it with functional groups or whole pro-
teins. These can be extracellular matrix proteins (such as fibronectin or collagen)
but they can also be protein fragments, peptides, or antibodies against epitopes of
a particular functional interest. Because the cell’s interaction with the extracellular
matrix can be highly dynamic and time-dependent, selecting the right incubation
time is critical. For small beads, a time interval 15 minutes and 1 hour is often used.
If the incubation time is too brief, the beads will not have sufficient time to bind;
too long, and the cells can endocytose (engulf) the beads, causing loss in receptor
specificity.
This process of functionalization can be combined with the photolithographic
microfabrication techniques described above. Microcontact printing or micropat-
terning involves using a soft polymer such as PDMS (polydimethyl­siloxane) to Nota Bene
“print” a protein onto a surface with a microscopic pattern. A silicon master is first A functional group. A functional
created by coating a chip with a photoresist that can be removed optically on a group is the region or moiety of a
microscopic scale. The resulting pattern is then acid-etched into the silicon. protein that is responsible for its
Multiple PDMS stamps can then be created from the master. The stamp is then particular biochemical characteristic
or activity.
dipped into a protein or other solution and pressed onto glass or another surface.
Microfabrication techniques are used in “lab-on-a-chip” technologies using min-
iature channels, pumps, valves, etc. The small scale of these devices provides
enormous advantages in terms of small sample size, high throughput, accelerated
assay time, and even portability.

6.3 AppLYING FORCES TO CELLS

So far we have examined experimental systems to measure cell mechanical
­properties and the forces that cells produce. Equally important are systems
aimed at investigating how cells sense and respond to mechanical signals.
Although we are ultimately interested in the response of a tissue in vivo, for cells
residing within a tissue, a simple mechanical stimulus such as deformation can
lead to many cellular-level physical signals (such as fluid flow, electric fields, and
more). One attractive strategy is the reductionist approach that is common in
biology (described in Section 2.4). With this strategy the cellular response to sim-
ple mechanical stimuli is examined in vitro. This allows the signal of interest to
be applied in a controlled fashion, and, importantly, many molecular approaches
for modulating gene or protein functions can be applied much more easily in
cell culture.
At the basic level, there are a few simple mechanical stimuli that have received the
most attention, including fluid shear and stretch. We will examine each in turn
and consider the key issues and methods involved in their application.

Flow chambers are used for studying cellular responses

to fluid shear stress
As we will discuss in Section 11.1, several cell types have been shown to respond
to fluid shear stress. Cells that line fluid-transporting vessels, like endothelial and
certain epithelial cells, are perhaps expected to respond to shear, but other cell
types, such as bone and cartilage cells, are also sensitive to shear. As a result, there
is substantial diversity in shear flow application systems, owing to the different
physiological conditions. Another reason for a diversity of flow devices is that
168 CHAPTER 6: Cell Mechanics in the Laboratory

establishing well-controlled flow can be a challenge in itself; there is generally no

way to directly measure the shear stress being applied on the cell monolayer.
Although custom devices might circumvent some of these issues, those devices
must be designed carefully. A basic understanding of fluid mechanics can greatly
ease the development of design specifications based on fluid-flow analysis, such
as that presented in Section 4.3. Before fluid mechanics specific to shear devices
are addressed, we examine one common consideration in using fluid shear
devices.

The transition between laminar and turbulent flow is governed

by the Reynolds number
To design the fluid flow device, one normally uses laminar flow to expose the cells
to a controlled shear stress. Nevertheless, there are numerous studies involving
turbulent flow, such as investigations of how turbulent flow contributes to the
development of atherosclerosis. Although a detailed analysis of turbulent fluid
mechanics is beyond the scope of this text, one consideration is to ensure that no
unintended turbulence occurs. To ensure the flow is laminar, one generally
designs the device so that the Reynolds number (Re) is very low (recall the discus-
sion of the Reynolds number in Section 4.2).
When the Reynolds number is very low, the viscous forces dominate. That means
the fluid tends to move uniformly without much mixing. These flows are termed
laminar owing to the laminar structure of the streamlines. High Reynolds num-
ber flows tend to be inertial: mixing, uneven fluid profiles, and unsteady flows.
These flows are turbulent. The actual transition Reynolds number between lami-
nar flow and turbulence is different for different geometries, but for pressure-
driven pipe flows (known as Poiseuille flow), a Reynolds number of approximately
2000 is a good estimate of the transition point. A Reynolds number of, say, 200 is
generally laminar despite having a relatively higher inertial component. If the
walls of the pipe are very smooth, it is possible to increase the Reynolds number
to even higher values without inducing turbulence, partly because the inertia of
pipe flows is one-dimensional. However, if there is roughness on the pipe walls
that could “trip” the fluid and induce mixing, a fluid element running close to the
wall would move in non-axial directions. This, in turn, causes the adjacent ele-
ments to shift, and soon the entire flow is tumbling. This is deliberately done in
the dimpling of golf balls to induce turbulence during the flight of the golf ball
(Figure 6.17).

Parallel plate flow devices can be designed for low Reynolds

number shear flow
One strategy for applying fluid shear to cells is the parallel plate flow chamber. In
such a chamber, fluid is pumped through a chamber of rectangular cross section.
The height is very small compared with the width, so the fluid profile can be
assumed to occur between two infinite parallel plates. The cells are plated on the
bottom surface.
In designing such a shear stress apparatus, multiple factors need to be consid-
ered. The fixed factors are relatively easily dealt with. The fluid is typically a cell-
compatible medium that can be assumed to have nearly the same density and
viscosity as water. The temperature is controlled by some custom or commercially
available heaters/coolers, or the device is used within a cell culture incubator. The
cells are plated on a surface treated with adhesion molecules or otherwise modi-
fied to be compatible with cell adhesion. To obtain a low Reynolds number, the
chamber dimensions must be carefully selected. To control the shear stress, one
usually controls the velocity. However, this also affects the Reynolds number.
Therefore, there is an interaction between the size of the chamber and the velocity
at which the transition to turbulence will occur.
 AppLYING FORCES TO CELLS 169

Figure 6.17 Dimples in golf balls

induce turbulence and create a
smaller boundary layer that takes
longer to separate. The smaller wake
helps reduce pressure drag on the ball.
Early golfers playing with smooth golf
low- balls found that badly dented balls
pressure tended to fly farther. For many cell
region mechanical applications, though,
laminar flows are required and therefore
smoother surfaces are desired.

flow separates

low-
pressure
region

Nota Bene
Let us look at an approximation that can be used for a quick estimate of the Dynamic versus kinematic
flow  device dimensions. Consider a system using cell culture medium as the viscosity. Viscosity is sometimes a
fluid (ρ = 1000 kg/m3, μ = 1 × 10-3 kg/ms). The Reynolds number is then Vh/ confusing term, as the concept of
(1 × 10-6 m2/s), where V is some characteristic velocity and h is the height of viscous resistance is associated with
the flow chamber. For Re < 1, we then need Vh < 1 × 10-6 m2/s. Next, consider both dashpots (which we will cover in
the shear stress, τ =  μ du/dy. We can approximate this as τ ~ μV/h, so long as Section 6.4) and fluids, but these terms
have different units and definitions. In
the width is large compared with h. A typical shear stress for physiological this text, we refer to dashpots as
modeling of endothelial cells is about 1 Pa. So, V/h = 1000 s-1. Assume we having a viscous friction coefficient
express V in m/s and h in m to keep units simple, then Vh < 10-6 and V = 1000 h. whereas fluids have viscosity.
This means, h2 < 10-9 m, or h < 3 × 10-5 m = 30 μm, which is just about accept- In fluid mechanics, there is a further
able, seeing as cells are typically at least a few micrometers (10-6 m) large. distinction between the dynamic
Although a chamber this short can be difficult to manufacture, generally, the viscosity (which is viscosity that is
defined in Section 4.2, represented by
chamber is designed with the shortest height possible. Calculations based on
μ, and has units of kg/ms) and the
actual flow profiles and relaxing the Reynolds number requirement (to 10–100) kinematic viscosity, which is the
can improve this range somewhat. dynamic viscosity normalized by the
density of the fluid (υ = μ/ρ, with units
of m2/s). The reason for using the
Fully developed flow occurs past the entrance length kinematic viscosity is that it simplifies
some calculations (such as Reynolds
Another consideration of shear flow devices is the entrance length. Laminar flow
number), can be written with three
profiles require space to develop, and the entrance length is the size of this terms instead of four, and represents
region. It changes depending on the velocity and dimensions of the chamber. For a useful ratio that eliminates density
a circular pipe, the entrance length is about 0.06 times the Reynolds number (mass, specifically) from comparisons.
multiplied by the diameter, 0.06 × Re × d. When locating cells to examine, it is That is, denser fluids might appear
best to avoid the entrance length to avoid unexpected or unpredictable shear more viscous because of inertial
stresses. The cross-wise flow profile (perpendicular to both the height and the effects. For this chapter, viscosity will
refer to dynamic viscosity.
flow direction) will also develop over space; this is generally not desired, as the
170 CHAPTER 6: Cell Mechanics in the Laboratory

flow is best modeled as two-dimensional. A rule of thumb requirement is that the

width of the chamber should be at least ten times the height, and the height
should be as small as possible. The circular pipe formula for entrance length can
be used for roughly estimating what the entrance length is for a parallel plate flow
chamber, but if validation is required, empirical measurements of the flow pro-
file using beads in the fluid can be performed.

Cone-and-plate flow can be used to study responses to shear

Another shear flow device uses a cone-and-plate viscometer. A cone is posi-
tioned above a stationary plate and rotated. A constant shear stress on the bot-
tom plate is established, depending on the gap between the plate and cone
bottom and the rotational speed of the cone. The condition of constant shear is
obtained by selecting a very shallow cone angle so that the gap between the cone
and the plate increases from the apex to the periphery. This configuration facili-
tates the controlled application of shear stress because it is directly proportional
to the rate of rotation of the cone. Note that the Reynolds number must still be
kept low, because this configuration can generate turbulence or unsteady transi-
tions if the rotation is too high. Additionally, systems involving moving compo-
nents can require higher engineering precision, because since small machining
errors can have big impacts on the flow. For example, the spacing between paral-
lel plates can be controlled with gaskets sandwiched between the plates. For the
cone-and-plate viscometer, ensuring the cone’s axis is perpendicular to the plate
can be challenging.

Example 6.4: Shear stress in a cone-and-plate viscometer

Show that for a cone-and-plate viscometer (Figure 6.18), from the cone to the bottom plate. Thus, the shear
the shear stress is the same far from the tip of the cone, stress is
if the angle is shallow enough.
τ = µ(du/dz ) = µ (ω r)/h = ( µωr )/(rα ).
Let ω be the rotational speed in radians/second, r the
radius away from the tip of the cone (where it contacts And,
the plate), and α the angle between the cone and the
plate, with the fluid in the gap and the cells plated on τ = ( µω )/α .
the bottom plate. The distance between the bottom plate
and the lower surface of the cone is h, where Note that this is spatially constant, so for a given viscos-
ity, a fixed ω and a fixed α, the shear stress on the bottom
h = r tan α.
plate is the same everywhere. Cone-and-plate viscome-
ters are useful because you can attain good control over
For a very small α, this can be approximated as
the shear stress just by varying the cone rotation rate (ω)
h = rα. and it uses relatively little medium because it is not recir-
culated. However, careful machining and alignment are
If the rotation of the cone is steady, the flow is laminar, crucial, because small errors can result in varying shear
and assuming α is small, we can treat the flow as paral- stresses along the bottom place or operational instabili-
lel plate flow; that is, for each r, the flow profile is linear ties (vibrations, etc.).

z Figure 6.18 Schematic of a

cone-and-plate viscometer. As
ω long as the cone-angle is shallow,
the shear on the bottom surface is
homogeneous.

Bottom flat surface (cells plated on top of this)

 AppLYING FORCES TO CELLS 171

(A) rectangular flow chamber Figure 6.19 Two examples of

parallel plate flow chambers. (A)
oscillatory pump Example of a shear flow setup that
pump conditions the media and allows for
imaging during shear application, and (B)
the actual flow chamber that has a
separation/recirculation region and a
air, 5% CO2 laminar flow region to assess effects of
microscope disturbed flow on the cell monolayer.
reservoir (Adapted from, Chien S (2008) Ann.
Biomed. Eng. 36.)

EC monolayer

EC

(B) step flow chamber

oscillatory pump
pump

air, 5% CO2
microscope
reservoir

flow side view

step
a b c d

disturbed flow direction of

reattachment zone laminar flow

Although we have focused primarily on steady laminar flow, it is worthwhile to

note that oscillatory flows (flows that either halt or reverse), turbulent flows, sepa-
rated flows (flows that develop recirculation regions), and combinations thereof
(Figure 6.19) have also been extensively used to investigate cellular responses to
temporally varying shear stress. These studies have become particularly impor-
tant in the cardiovascular field, in which complex flow patterns have been related
to pathological changes.

Diverse device designs can be used to study responses

to fluid flow
Before we conclude our discussion of devices for studying responses to fluid
flow,  it is worthwhile mentioning that although pressure-driven parallel plate
and  cone-and-plate devices are commonly used designs for investigating flow-
mediated mechanotransduction, there is great diversity in shear flow application
systems. For example, tubes have been used with success, although it becomes
more difficult to image the cells during the application of shear. This geometry
may be more physiologically relevant for modeling blood vessels. Another
approach is to replace a pressure-driven flow with a driven-plate configuration,
172 CHAPTER 6: Cell Mechanics in the Laboratory

known as Couette flow. A treadmill used in lieu of the upper plate of a parallel
plate chamber will generate a linear flow profile, which mitigates problems with
entrance lengths and creates a simpler relationship between the moving wall
velocity and shear stress.

Flexible substrates are used for subjecting cells to strain

Although many cells experience fluid shear (or at least, mechanical shearing of
some sort), other cells primarily experience stretch. Blood vessel smooth muscle
cells, cardiac cells, skin cells, bladder/stomach/intestine, and lung cells are exam-
ples of physiological systems in which stretch is vital. Applying controlled stretch
to a cell directly is problematic, because the cell membrane is difficult or impos-
sible to grasp. In addition, cells generally experience stretch due to deformation in
the surrounding tissue in which they are anchored. The most common way of
stretching cells is to plate them on a flexible substrate, which is then stretched.
This method has the advantage of allowing many cells to be stretched simultane-
ously. Substrate stretch is not always a physiologically representative stretch, as it
is confined to two dimensions and primarily experienced through the basal sur-
face of the cell. This is less of a concern for cells normally found in monolayers,
such as endothelial cells, some epithelial cells, or osteoblasts. However, many key
physiological responses can be activated by this technique. Therefore, substrate
stretch can be useful for studying these responses, even if the stimulus is not quite
physiological.

Confined uniaxial stretching can lead to multiaxial

cellular deformations
In stretching cells, some consideration must be given to how the stretch is applied.
The simplest form of stretch is unconstrained uniaxial, whereby the substrate is
held on two opposing sides and then pulled apart. This lengthens the substrate and
cells in the direction being pulled, but will shorten them in the perpendicular
direction. Additionally, if the ends being pulled are clamped tightly, then the short-
ening will be nonuniform, being most pronounced at the center (Figure 6.20).
This partly constrained system can result in strain heterogeneities. A constrained
uniaxial design overcomes this by constraining the two stationary sides. This
eliminates strain in the perpendicular direction. It is a suitable model for some
physiological systems, such as blood vessels, and can demonstrate cellular reori-
entation responses to stretch quite nicely. However, such designs are trickier to
build, because the sides must be kept fixed in one direction but must move in
another. Further, constrained uniaxial stretch can result in some degree of hetero-
geneity in cells that exhibit strong axial orientations. That is, an elongated cell may
respond differently if strained along its long axis than along its minor axes.

Cylindrically symmetric deformations generate uniform

biaxial stretch
A final, and perhaps the most straightforward, way to expose cells to substrate
stretch is with a uniform-biaxial stretch (Figure 6.21). In this strain field, every

Figure 6.20 A flexible membrane

that is stretched in one direction
will result in a “bowtie” formation
that is non-uniform. The shape is
exaggerated for clarification.
 ANALYSIS OF DEFORMATION 173

lid Figure 6.21 An example design for a

membrane membrane stretch device. The cells
are plated on the upper surface of the
membrane holder ring
membrane, with a lid and rings used to
mobile plate keep the cells immersed in media. The
indenter ring can be pushed up (or the
Teflon membrane pushed down) to stretch the
O-ring silicon O-ring membrane along with cells attached to
the membrane surface. (From, Sotoudeh
motion motion M, Jalali S & Usami S (1998) Ann. Biomed.
Eng. 26.)
indenter
ring

fixed plate

pair of points is expanded by the same percentage strain regardless of orientation.

This strain field has no preferred direction, and each cell will experience the same
strain in all directions. There are several designs for this sort of stretching, includ-
ing using air to push up a dome, but a common configuration uses a cylindrical
piston.

6.4 ANALYSIS OF DEFORMATION

In the previous sections, we learned about how cells can be exposed to forces to
characterize their mechanical properties and responses. One question left open is
how can we use such experimental data to describe the mechanical behavior
observed? Suppose we obtained displacement profiles for two different cell popu-
lations, such as different cell types or cells in the presence/absence of a cytoskel-
etal inhibitor. How could we convert the displacement profiles into something
with which we can compare the mechanical properties of these two populations?
One approach is to compare the raw data alone. We might directly compare bead
displacement. However, this approach is a simplistic comparison. A broader,
potentially more useful, approach is to develop models for a cell whereby param-
eters that are experimentally obtained can be used to make predictions and com-
parisons across different experiments.

Viscoelastic behavior in micromanipulation experiments

can be parameterized through spring–dashpot models
One of the approaches for mechanical modeling of displacement and/or force
profiles generated using micromanipulation is to construct models consisting of
spring and dashpot elements. Springs obey Hooke’s law, which can be written as

F = kx , (6.13)

with the spring constant k a measure of the stiffness of the element. Dashpots are
viscous damper elements that obey Newton’s fluid law

F = ηv , (6.14)
174 CHAPTER 6: Cell Mechanics in the Laboratory

where v is the velocity and η is a parameter known as the viscous friction

coefficient.
In general, a spring or dashpot in isolation is insufficient to model physiological
cellular mechanical behavior. Consider the case where we perform a magnetic
micromanipulation experiment where we obtain the displacement of a cell-
bound bead under a constant force as a function of time. There is no force
applied initally; then we turn on the magnet, generating a near-instantaneous
step-increase in force that remains constant as long as the magnet is turned on.
After a few seconds, the magnet is turned off, resulting in a near-instantaneous
step-decrease in force. If the cell behaved as a purely elastic material, like a
spring, the displacement curve of the bead would resemble the profile of the
forcing function, with an instantaneous displacement when the magnet was
turned on and which would remain constant. When the magnet was turned off,
an instantaneous jump to zero displacement would occur, as seen in Figure
6.22.
On the other hand, if the cell behaved as a purely viscous material, like a dashpot,
the bead’s response to a step up or down in force would appear as a line with non-
zero slope (Figure 6.23). Unlike a spring, the displacement of the dashpot
increases linearly with time with slope η. When the magnet is turned off, the dash-
pot does not return back to its original length. Instead, it exhibits creep, or perma-
nent deformation.
In reality, cells do not exhibit purely elastic or viscous behavior. Rather, they are
viscoelastic, exhibiting time-dependent mechanical behavior that has attributes
of both elastic and viscous materials. Under a constant force, the displacement
does not occur instantaneously or slowly increase at a constant rate, but instead
typically exhibits asymptotic behavior. Similar asymptotic behavior is exhibited
when the force is removed, and in many cases cells exhibit permanent deforma-
tion upon removal of the force. As a result, using a spring or dashpot alone is
insufficient to capture this mechanical behavior.

Combinations of springs and dashpots can be used to model

viscoelastic behavior
One approach to producing more realistic cellular behavior is to combine several
springs and dashpots together. Given the fact that cells exhibit behavior that is
both elastic- and viscous-like, models constructed of combinations of springs and
dashpots in series or parallel can result in mechanical behavior that resembles
experimental observations remarkably well. Three examples of such combina-
tions are a spring and dashpot in series (called a Maxwell body), a spring and

Figure 6.22 The response of a 1.2

spring element to a step force
being applied at 1 second and
turned off at 2 seconds. The
1.0
displacement reaches a constant as
displacement

soon as the force is applied. Note that 0.8

there is no oscillation because there is
no inertial term. 0.6

0.4

0.2

0.0
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50
time (s)
 ANALYSIS OF DEFORMATION 175

3.0 Figure 6.23 At one second, the

force is applied, and the response
element starts “stretching out.”
2.5 Because the force is proportional to the
velocity, a constant force application
displacement

2.0 results in a constant velocity response.

At 2 seconds, the force is released, and
1.5 the element no longer displaces, but
does not recoil because here is no
elastic restoring force.
1.0

0.5

0.0
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50
time (s)

dashpot in parallel (a Kelvin–Voigt body), and a combination of a Kelvin–Voigt

body in series with a dashpot (Figure 6.24). As we will see, the analytical solution
for the displacement of these bodies with time under a constant force can be com-
puted in a relatively straightforward manner.
Consider the combination of a Kelvin–Voigt body in series with a dashpot. A free-
body analysis shows that the force F applied to the system must equal the force on
the right-hand dashpot, and it must equal the sum of the forces in the left-hand
spring and dashpot. If we expand these forces with the governing equations
(Equations 6.13 and 6.14), we obtain
F = η1v1 = kx 2 + η2v2 . (6.15)

The total displacement is the sum of the displacement of the single dashpot and
the Kelvin–Voigt body, x(t) = x1 + x2. Thus, to obtain x(t), we can compute x1(t)
and x2(t) individually, and then add them together to get the displacement of the
entire body. For the single dashpot, we know that x1 = (F/η1)t. For the Kelvin–Voigt
body the total force across the body is equal to the sum of the force within the
spring and dashpot
dx 2
F = kx 2 + η . (6.16)
dt

This is a first-order differential equation, which has the solution

F  − kt  
x 2 (t ) =   . (6.17)
x  1 − e 2  
η

η2 Figure 6.24 A model of cell material

properties using spring and
dashpot combinations. In this
η1 example, a spring and dashpot are in
parallel. Together, they are in series with
F F a second dashpot that allows for plastic
deformation.

x2 x1
176 CHAPTER 6: Cell Mechanics in the Laboratory

Figure 6.25 Response of the cell 2.5

model to a step force applied at 1
second and turned off at 2 seconds.
Note that after the application of the 2.0
force, the response slowly approaches a
straight line. After the force is released

displacement
the recoil is not to zero producing 1.5
residual deformation or strain. This
response profile to a step increase and
decrease in force is a reasonable 1.0
approximation of real-cell response to
magnetic micromanipulation.
0.5

0.0
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50
time (s)

This can be verified by direct substitution, with initial condition x(0) = 0. Summing
the displacements of the single dashpot and the Kelvin–Voigt body,

)
Nota Bene
F F
x (t ) =   t + 
 − kt 
Lumped parameter models. The  η  . (6.18)
 η1  k 1 − e 2
models of the type we have been
discussing are sometimes called
rheological models, because they The behavior of this model is depicted in Figure 6.25. It is very useful, because it
originated from the analysis of is the simplest model that captures both viscoelastic behavior as well as plastic
viscous fluids (see Section 4.4). They
behavior (permanent deformation). To understand this phenomenon, consider
are also sometimes referred to as
lumped parameter models, because what happens as t goes to infinity. The first term (the displacement of the single
they collect spatially distributed dashpot) increases without bound at a constant rate of F/η1. The second term (the
behaviors and lump them into simple displacement of the Kelvin–Voigt body) gives a displacement at t = infinity of F/k.
discrete elements. For the lumped For finite time, the Kelvin–Voigt body never reaches F/k, but approaches it asymp-
parameter model in Figure 6.25, all totically. Finally, it is worthwhile noting that the exponential term, −kt/η2, is
the elastic behavior was captured in sometimes written as −t/τ, where τ =  η2/k is called the time constant. The time
a single equivalent spring stiffness
constant is a measure of how long it takes the exponentially decaying quantity in
parameter (k).
the parentheses to decrease by 1/e, or about 63%, of the initial amplitude.

Example 6.5: Maxwell body: spring, and dashpot in series

Examine a Maxwell body: a spring and dashpot in series Therefore, the dashpot will behave like a rigid structure
(Figure 6.26). and the response of the Maxwell body will be similar to
that of a spring alone at high frequencies.
At very low frequencies, the dashpot velocity will be small,
Figure 6.26 Schematic of a Maxwell body. A Maxwell and so the force sustained by the dashpot will be negligi-
body is a spring and dashpot in parallel. ble. As a result, the dashpot acts as an open structure, and
the spring acts like a rigid structure. Here, Maxwell body
Let the spring have constant k and the dashpot have con-
will be similar to that of a dashpot.
stant η.
B. Quantify your response to part A.
A. If we were to oscillate the tip of the Maxwell body at very Let the stimulus be F(t) = F0sin(ωt), where ω is the fre-
high or very low frequencies, would the response be quency of oscillation and F0 is the amplitude. The
dominated by the spring or the dashpot? response will be x(t) = Asin(ωt +  δ), where δ is the phase
At very high frequencies, the dashpot velocity will gener- lag. We know the force through the spring and dashpot
ally be large. Because F =  ηv, as the velocity goes to are the same (both equal to F(t)). We also know the dis-
­infinity, the force on the dashpot goes to infinity as well. placement of the entire system is the sum of the
 ANALYSIS OF DEFORMATION 177

­ isplacements of the spring and dashpot. Thus, we can

d In this case, δ = −π/2 and F0 =  ηωAd
solve for each individually.
and thus,
For the spring, this is a simple substitution.
x d (t ) = (F0 /ηω )sin(ω t − π /2).
Let x s (t ) = As sin(ωt + δ )
The total displacement is the sum of the spring and dash-
F (t ) = F0 sin(ω t) = kx s (t ) = kAs sin(ωt + δ ). pot components,
So we know δ = 0 for the spring and that As = F0/k. x (t ) = ( F0 /k )sin(ω t) + ( F0 /ηω )sin(ω t − π /2).
Therefore,
As ω goes to infinity, the dashpot (right-hand) term,
xs(t) = (F0/k)sin(ωt).
which varies as 1/ω, goes to zero, and we are left with
For the dashpot, we need a differential equation. only the spring term. Similarly, at very low frequencies,
the dashpot term overshadows the spring term and we
Let x d (t ) = Ad sin(ω t + δ ) , then are left only with the dashpot term. At low frequencies,
the spring does not significantly influence the behavior
F (t ) = F0 sin(ω t) = η(dx d /dt ) = ηωAd cos(ω t + δ ). of the system.

Microscopy techniques can be adapted to visualize cells

subject to mechanical loading
In Section 6.1, we first considered the challenge of determining cellular mechani-
cal behavior, but we limited our discussion to force application at discrete loca-
tions, such as beads or AFM probes. As we described in Section 6.3, it is also
possible to expose cells to mechanical stimuli more representative of their native
environment. Consider, cartilage explants freshly dissected from collected tissue
and subjected to the dynamic compression that they would experience in vivo.
Here, the complex biochemical and mechanical environment of the tissue gives
rise to cellular deformations that are the additive result of several different physi-
cal and/or mechanical phenomena. These might include loads imparted to the
extracellular matrix that are transferred to the cells at the sites of adhesion, fluid
shear, hydrostatic pressure, altered osmotic pressure, etc. However, the deforma-
tional analysis is more complex, because our readout is not simply the displace-
ment at one location. One approach is to observe cells microscopically during
loading. We will discuss here how typical time-lapse image sequences are
obtained and how their deformation is analyzed.
To generate image sequences of cells subjected to mechanical loading, struc-
tures in living cells can be imaged by epifluorescent or confocal microscopy by
using fluorescent stains, or transfecting cells with DNA constructs that encode
for fluorescent proteins. Alternatively, certain structures can be visualized with
techniques that generate image contrast in transparent specimens, such as
phase contrast. By fabricating specialized loading devices that fit on the stage of
microscopes, these microscopy techniques can be adapted in a variety of ways
to visualize cellular or even subcellular deformations. To illustrate, consider the
case of cells subjected to fluid shear stress. Such chambers have been con-
structed to allow for visualization of the cells from the bottom, or in some cases,
from the sides as well. Such multi-plane approaches use mirrors and multiple
optic paths to perform simultaneous imaging in two orthogonal planes
(Figure 6.27) resulting in quasi-three-dimensional imaging without a confocal
microscope.
178 CHAPTER 6: Cell Mechanics in the Laboratory

Figure 6.27 The side-view (side view) (top view)

microscope allows imaging of a cell
from the top and in profile condensor
simultaneously. This is achieved by
having two separate lightpaths and sets
of optics. The mirror allows for a more medium cell
compact design with all optical
components in a conventional
configuration. A rectangular flow
45°
channel is required for side-imaging
(From, Cao J, Donell B, Deaver DR, et al.
(1998) Microvasc. Res. 55.) 45°

objective lens

(top view) (side view)

T=0
Flow

10 µm
T = TP T = Te

Cellular deformations can be inferred from image sequences

through image correlation–based approaches
Once image sequences of mechanically–loaded cells are obtained, they can be
analyzed to infer the cellular deformations that occurred through the use of image
correlation. As we described previously, this can be accomplished either with tex-
ture correlation or by tracking the motion of attached beads. In either case, the
result is a set of displacements at discrete spatial locations. Often, we would like to
know the whole displacement field. This can be obtained through a process of
interpolation. Consider set of k discrete displacement vectors u1,2,. . .,k obtained at
positions x1,2,. . .,k. To construct a displacement field, the discrete displacements
are interpolated to all the interior locations between the x positions. A common
approach is to interpolate two-dimensional discrete displacements linearly to
construct triads, or groups of three displacement vectors that neighbor each other
in space. One algorithm for automatically generating triads is Delaunay triangu-
lation, which maximizes the minimum angle of all the angles of the triangles in
the triangulation (and therefore avoids long aspect ratio triangles), as in Figure
6.28. The details of triangulation approaches are beyond the scope of this text but
can be found in many image-processing treatments.
Once the triads have been specified, the displacement field can be constructed by
linearly interpolating the displacements between each triad. Linear interpreta-
tion over the triangular domains results in a plane determined by these three
locations. The equation of plane for u and v, the displacements in the x- and
y-directions, respectively, are
u( x , y ) = ua x + ub y + uc (6.19)

 ANALYSIS OF DEFORMATION 179

Figure 6.28 Delaunay triangulation

of fluorescent beads. Here we have
applied a Delaunay triangulation
procedure to the cell with attached
beads from Figure 6.11. The number of
beads was kept small to better illustrate
the approach. In practice, more beads
and triangles with lower aspect ratios
would be required for accurate results.
(From, Kwon R, et al. (2007) J. Biomech.
40, 3162–3168.)

and

v ( x , y ) = v a x + v b y + vc , (6.20)

where ua, ub, uc, va, vb, and vc are unknown constants. To solve for ua, ub, and uc,
we form the set of simultaneous equations

ua x1 + ub y1 + uc = u1

ua x 2 + ub y 2 + uc = u2 (6.21)
ua x 3 + ub y 3 + uc = u3 ,

which results in three equations and three unknowns. The constants va, vb, and vc
can be solved for using a similar approach. We segment the image into triangles,
at which the vertices are locations for which we have displacement data. Then, to
estimate the displacement at an arbitrary point, we select the three vertices mak-
ing up the triangle that encloses the point.

Intracellular strains can be computed from displacement fields

Given the displacement fields u(x, y) and v(x, y), it becomes possible to calculate
intracellular strains. In Chapter 3, for small deformations, we derived the infini-
tesimal strains,

du
ε xx =
dx
dv
ε yy = (6.22)
dy

1  du dv 
ε xy = + .
2  dy dx 

180 CHAPTER 6: Cell Mechanics in the Laboratory

Nota Bene
For large deformations, the Green–Lagrange strains can be computed as

Constant strain triangles. Note du 1   du  2  dv  2 

that displacements that are linearly E xx = + +  
interpolated produce a special form dx 2   dx   dx  
of strain. Specifically, the derivatives
dv 1   du   dv  
2 2
in Equation 6.22 result in constant
values, that is not dependent on x E yy = +  +   (6.23)
and y. Thus, the computed strains will dy 2   dy   dy  
result in a discontinuous strain field,
with constant strains within each 1  du dv  1  du du dv dv 
E xy = + + + .
triad/triangle. For this circumstance, a
2  dy dx  2  dx dy dx dy 
higher-order scheme can be used to
interpolate the displacements if a
discontinuous strain field is An alternative approach can be used to compute the Green–Lagrange strain that
undesirable. does not require the computation of the displacement derivatives. Recall from
Section 3.3 that for any differential line segment dX, its deformed length after a
deformation is given by

dx 2 = dXCdX , (6.24)

where C = FTF is the symmetric Cauchy–Green deformation tensor. Assuming a

two-dimensional linear deformation, there are three unknowns to solve for: C11,
C22, and C12. Recalling our triads of displacement vectors, the vertices of each
triad located at x1, x2, and x3 define three line segments in the undeformed state.
Similarly, the points X1 = x1 + u1, X2 = x2 + u2, and X3 = x3 + u3 define the same
three line segments transformed under the deformation described by F. Now we
can compute F directly. Then we can extract information regarding the principal
directions and stretches. Additionally, once C is computed, the Green–Lagrange
strain can computed as E = 1/2(C − I), where I is the identity tensor.

Example 6.6: Deformation gradient analysis

Let us revisit Example 3.6, but instead of determining the where I is the identity matrix. We can then write
principal strains, we will calculate the principal stretches
and principal stretch directions. The analysis is very sim- (U − Iλ ) ν= 0.
ilar, but the final result is different.
This must be true for nontrivial ν (our eigenvectors) and
Recall that the quantity in parenthesis must be non-invertable. That
means the determinant of (U − Iλ) = 0. To set this up,
2.5 0.5 
F= .
0.5 2.5  2.5 − λ 0.5
= 0.
0.5 2.5 − λ
We note that F is symmetric, so that FTF = F2, but we
know FTF = U2 and therefore U = F. Our rotation tensor Some algebra leads to
is the identity matrix and there is no rotation.
To find the principal stretches, we need the eigenvalues 6.25 − 5λ + λ 2 − 0.25 = 0 = λ 2 − 5λ + 6.
of U,
The roots of this equation are λ = 2, 3. Physically that
U ν= λ ν means if a vector is in a principal stretch direction, the
deformation will stretch it out by a factor of 3; if it is in
is the characteristic equation relating our stretch tensor the other principal stretch direction, the deformation
U, the principal stretches, λ, and the principal stretch will stretch it out by a factor of 2. What are the direc-
directions ν. Noticing that we can rearrange the equation tions? We simply substitute back into the characteristic
to read: equation

U ν−I νλ = 0, U ν= λ ν
 BLINDING AND CONTROLS 181

for each λ. We find that the eigenvectors are (1,1) (corre- Typically, the vectors would be expressed in unit form,
sponding to λ = 3) and (1, − 1) (corresponding to λ = 2). but we do not do so here to avoid the use of square roots.
Note that when you are solving for the eigenvectors, you Our result is also consistent with our findings in Example
will not generally be sufficiently constrained to get the 3.6 in that the principal stretch directions are the same as
exact values; so you can guess a value for one component. the principal strain directions.

6.5 BLINDING AND CONTROLS

Biomedical engineering experimentation can be a delicate field because it strad-
dles more traditional engineering, in which results are often more analytical or
quantitative, and biology, where results can be more empirical and experimental.
Because of such contrasts in approaches, there are two concepts common to
experimental biology that engineers should understand to perform and interpret
analysis correctly: blinding and controls.
A control is a duplicate of the experimental condition for which an intervention or
stimulus of interest is left out. If we wished to assess whether the addition of com-
pound A will cause cells to multiply faster, what we could do is to get a dish of cells,
add A to the media, and then measure the cell division rate. We could then com-
pare the measured cell division rate to known cell division rates and see if it is
increased in the presence of A. The problem is that the cells we have been using
may already be dividing faster than average, or perhaps the media we are using to
feed the cells causes them to divide faster, or the incubator where the cells are
being kept possibly creates an environment whereby the cells will divide faster.
The best way to account for all of these effects is to maintain a second dish of cells,
to which we do not add A, using the same cell type, media, and incubator. Then,
we can compare the division rates between the two dishes and assess whether A,
in fact, influences cell division. Although it is accepted that there will be some dif-
ferences (the dishes will not be kept in exactly the same location, will not be han-
dled exactly the same way, etc.), as many conditions as possible should be kept
identical between the two dishes. The ideal experiment is one in which the only
difference between experimental conditions is the factor being tested or exam-
ined. The clinical analog of the control is the placebo, designed to counter the
psychological influence of providing a potential cure to a patient. Patients often
feel better when they believe that they have received a medication (this is called
the placebo effect).
Blinding refers to a process that reduces potential bias by the investigator. The
“Clever Hans” horse is a popular story that illustrates this effect. Essentially, the
horse appeared to be able to perform simple arithmetic. A random person would
pose a question (2 + 3), and Clever Hans would start stamping his foot. At the cor-
rect number of stamps (5), he would stop. However, it was later determined that
the horse would simply watch his owner, and when his owner relaxed (at the right
number), he would stop stamping his foot. Similar bias can exist in experiments,
whereby an investigator who is expecting a certain result may interpret experi-
mental outcomes in favor of preexisting expectations. The best way to avoid this is
to blind the investigator to the conditions, so that measurements are taken with-
out this bias. If we were to give three people vitamins, and three people a placebo
(as a control), and our hypothesis was that vitamins will prevent headaches, we
would need to ensure that the people receiving the substances are not told
whether they are receiving the vitamin or placebo. This would be a blinded experi-
ment, because the participants would not know which substance they received
and could not use that information to influence the results. For instance, knowing
they received the vitamin, they might tend to say they have fewer headaches.
182 CHAPTER 6: Cell Mechanics in the Laboratory

However, our presence might also influence the participants, because we know
who received the placebo and who received the vitamins. Therefore, we may act—
perhaps unconsciously—more pleasantly toward those who received vitamins. To
avoid this error, we would have an assistant interview the participants about their
headaches; this assistant could not know who received which substance. The
patients and their treatments would be encoded so that nobody involved in the
experiment knows who received what. This approach is a double-blind design,
and is the most rigorous method for uncovering true influences.
Sometimes blinding does not work or is impractical. If we were to perform an
operative procedure to relieve pain, the people receiving the operation would
clearly know they had it, compared with patients who were left untreated. To
account for this, sometimes people use what is called a sham. This is a special
control whereby a procedure is mimicked, but a key therapeutic step is omitted.
So in this situation, the control group would have an operation, but without
whatever manipulation was thought to be analgesic. A sham surgery raises
important ethical concerns in human studies, but is a common strategy for ani-
mal investigations.
The use of controls and blinding applies to many of the techniques that are cov-
ered in this chapter. For instance, if you perform an experiment using magnetic
twisting by attaching magnetic beads to cells and then applying a torque, you can-
not simply obtain a readout (stiffness, upregulation of some gene, activation of
some channel, etc.) and decide whether the cells are responding or not. You need
to compare the readout to cells that do not have a similarly applied torque.
However, the addition of the beads may have caused some changes in cell base-
line. The better control is to have a sham in which you add beads but do not torque
the beads. Some investigators will go so far as to complete the sham by adding
beads to cells, placing the cells in the magnetic twisting device, and turning on the
equipment, but will not actually apply the torque. Alternatively, the experiment
can be performed with nonmagnetic beads. In this way, vibrations from the
equipment, stresses and temperature changes from handling and placing the dish
on the apparatus, the magnetic field itself, etc., are all eliminated as potential con-
founding influences.

Key Concepts

• A variety of techniques allow us to measure cell • Beads embedded in a flexible substrate allow
forces and manipulate cells. substrate deformations to be determined. An inverse
• There are various modes of magnetic problem is solved to obtain the applied forces.
micromanipulation (pulling vs twisting). • Cells have both solid and viscous behavior. Models
• Optical traps work by deflection of light and offer that incorporate both behaviors have been developed
contrast to magnetic methods by manipulating bead to describe cellular deformation.
displacement rather than force. Magnetic forces can • There are diverse methods for applying fluid shear
be applied to a large number of beads stress to cells. Consideration must be given not only
simultaneously. to maintaining biological compatibility, but to control
• In atomic force microscopy cells can be probed by of turbulence, entrance length, and flow development.
application of a cantilevered tip against the cell. The • Stretched samples in which strain levels are large can
deflection of the cantilever can be related to the force be analyzed by deformation gradient mathematics. Data
being applied. are generally acquired at discrete points, which are then
• Microfabrication is useful for assaying small numbers interpolated to yield displacement at desired locations.
of cells. Micropillars can be used to measure cell– • Experiments are most useful when only one factor, or
substrate forces based on the deflection of the pillars, one set of factors, is being changed at a time. Use of
which are modeled as beams. controls, blinding, and shams can help ensure the
• Traction-force microscopy is another way to measure readout is most representative of the response to the
cell–substrate forces. factor(s) being examined.
 PROBLEMS 183

Problems

1. Show that a series of n springs in parallel, with spring where the value in the ith row and jth column is the
constants k1, k2, . . ., kn, where ki does not necessarily intensity for pixel (i,j).
equal kj for i ≠ j with a single dashpot of viscous
coefficient η can be reduced to a single spring and You have determined that the template image (K)
dashpot in parallel. should be

2. Below is the response of a bead on a cell experiencing 1 2 3 2 1

a step–displacement at time t = t0. The vertical axis, F, 2 7 8 7 2
represents the force on the bead. Devise a spring–dashpot
3 8 9 8 3
model that is consistent with this response plot, and draw
2 7 8 7 2
the predicted response for a step release in force (i.e.,
force is some value F0 at t < t0 but force = 0 at t ≥ t0. 1 2 3 2 1

F Using Equation 6.7, compute C(3,3) and C(5,5).

7. Below is a figure of a stretch experiment performed

on a piece of tissue. You determine the following
t0 coordinates for the markers on the tissue

3. Consider a spring and dashpot in parallel subject

to an instantaneous step displacement. What is the C′
C
force through the spring? And through the dashpot? B′
Why would such a model be a poor choice for x A B A′
modeling the behavior of a bead subject to a step x
y
displacement?
  y

4. Devise two spring-dashpot models that are initially at

rest, experience a step-increase to a constant force F0 at A = (0, 0)
time t0, then a step-decrease to zero force at time t1 > t0,
such that (1) the stress–strain (or force-displacement) B = (1, 0)
response curves are identical for loading and unloading
C = (0,1)
and (2) the stress–strain (or force-displacement)
response curves are not identical for loading and A′ = (1,1)
unloading. The latter case is called hysteresis.
B′ = (3, 2)
5. Determine the energy required to extend a spring of
spring constant k from equilibrium (x = 0) to a new C ′ = (2, 5)
position x = x1. Next, determine the ratio of energy
required to extend a spring of spring constant k in Determine F, the deformation gradient tensor.
parallel with a dashpot of viscous coefficient η to that of
a spring alone. The answer should depend on the rate of 8. You are working with a deformation gradient tensor
loading, which may be assumed to be constant. What
2 −0.5
happens at low rates? And at high rates? given as F =  . A colleague created a special
1 4 
6. You have just completed imaging a fluorescent bead as program to extract the rotation tensor R (from the
part of a traction force microscopy experiment. Your decomposition F = RU). He input your F and obtained
array of intensity values is  1 1 
 2 2
2 1 0 2 3 1 1 2 1 0 R =   . Is this a plausible rotation tensor?
 1 1 
1 2 1 2 2 3 3 2 1 1 − 2 2 

2 1 2 1 6 4 4 2 0 2
Justify your response.
2 0 4 5 5 6 5 4 2 0
1 0 3 5 7 8 3 5 2 1 Explain why U, the stretch tensor, must be positive
2 1 7 8 9 8 6 4 6 2 definite from a physical perspective (that is not a purely
3 3 4 6 8 7 6 2 2 1 mathematical proof).
2 5 3 5 5 6 2 3 1 1 9. You are given a 2 × 2 tensor C = F TF. Let λ12 and λ22
1 3 2 2 3 2 1 4 2 1 be the eigenvalues of C, with |λ1| > |λ2|. Prove that
1 2 1 1 1 0 2 0 3 0 λ1 is the maximum stretch for an arbitrarily oriented
184 CHAPTER 6: Cell Mechanics in the Laboratory

cell undergoing deformation described by F, and diameter as the membrane. As the piston moves up
that λ2 is the minimum stretch under the same and down, the media in which the cells are immersed
conditions. will also flow back and forth across the cells. Assume
the media has viscosity μ and density ρ and that the
10. Use dimensional analysis to obtain an expression for cells are linearly elastic with elastic modulus 10 kPa
the drag force on a sphere under laminar flow. You must undergoing 5% stretch. Determine the ratio of stretch-
identify the parameters that govern the force and then based stresses to that of fluid shear stresses. You may
obtain a dimensionless expression. Next, show that model the cells as 50 μm × 50 μm × 5 μm rectangular
the obtained expression is consistent with Stokes’ law solids with the smallest dimension being the height. An
(Equation 6.1). order of magnitude ratio is sufficient.
11. Consider the case in which you are devising a shear- 13. Unlike the rounded tips described in this chapter,
based technique to estimate the adhesion force of cells atomic force microscopes may also be equipped with
to a substrate. You have an inlet into a cylindrically sharp tips that are typically pyramidal in shape (for
symmetric outflow as shown below. example, the Berkovich tip). Owing to their geometry,
these tips have modified relationships between the
Q force, E, the indentation depth, and the geometry of
the tip as given by the Hertz model. To model AFM
with a pyramidal tip, it can be useful to use a scaling
relationship instead of trying to get an exact solution.
r (a) First, derive a scaling relationship for the force and
indentation depth of a spherical tip atomic force
microscope. You may assume the radius of the bead
h (R) is very large compared with the radius of the
area being indented (a), and that the radius of the
area being indented is very large compared with
The dashed line represents the cell monolayer. The flow
the indentation depth (δ). Further, you may assume
into the inlet is at volume flow rate Q (m3/s) across
only the linear elastic modulus applies (E). To solve
cross-sectional area A. The height of the outflow region
this, use an energy balance by calculating the strain
is h, and you may assume that h < r and that the flow
energy of the deformation and equating that with
is laminar. At the inlet, you pump fluid at a constant Q,
the work done by pushing the tip in. Because this is
and at the outlet, collect all the cells that detach and
a scaling relation, you may ignore constants.
exit the outflow region around the perimeter. You then
(b) Now, assume the tip is a pyramid instead. Assume
count the cells that have detached. Relate the number
the angle of the tip is α, the indentation depth is
of detached cells to the maximum adhesion force of
δ, and the cell has elastic modulus E (continue to
the cell, assuming that the adhesion force per cell is
ignore shear effects). You may find it easier to start
uniform, that cell–cell interactions can be neglected,
by assuming the tip is actually a cylinder with a
and that each cell is 10 μm in radius.
rounded end with radius a (and assuming that δ is
12. Consider the case in which you are given a stretch much less than a), and then realize that the angle α
device consisting of a membrane that is 10 cm in is related to the “width” of the indenter. Once you
diameter. The membrane is stretched by a piston derive the scaling relationship, you should see that
pushing on its bottom surface, as depicted in Figure there is a mathematical difference between the
6.21. You may assume the piston is nearly the same two types of indentation.

Annotated References

Alessandrini A & Facci P (2005) AFM: a versatile tool in biophysics. Chen CS, Mrksich M, Huang S, Whitesides GM & Ingber DE (1997)
Meas. Sci. Technol. 16, R65–R92. This journal review article summa- Geometric control of cell life and death. Science 276, 1425–1428.
rizes many aspects of biological studies using AFM. This paper describes the use of micropatterning to determine
whether adhesion area or spreading area is more important for
Brown TD (2000) Techniques for mechanical stimulation of cells
maintaining cell viability.
in vitro: a review. J. Biomech. 33, 3–14. This journal review article
describes several different methods for applying stretching and Chien S (2008) Effects of disturbed flow on endothelial cells. Ann.
shear to cells and tissues. Biomed. Eng. 36, 554–562. This summarizes findings of how shear
flow affects endothelial biology in a way that implicates disturbed
Cao J, Donell B, Deaver DR, et al. (1998) In vitro side-view imag-
flows in the development of atherosclerosis.
ing technique and analysis of human T-leukemic cell adhesion
to ICAM-1 in shear flow. Microvasc. Res. 55, 124–37. This paper Goubko CA & Cao X (2009) Patterning multiple cell types in co-
describes the use of side-view imaging to obtain quasi-3D images cultures: a review. Mat. Sci. Eng. C. 29, 1855–1868. This review
of cells during rolling adhesion. article briefly summarizes some basic, micropatterning and then
 ANNOTATED REFERENCES 185

describes the more recent pattern generation for use in multiple sham surgery without insertion of the scope. The patients were
cell patterns. kept unaware of whether they had the surgery. Interestingly, the
control group experienced as much improvement as the surgery
Hoffman BD & Crocker JC (2009) Cell mechanics: dissecting the group.
physical responses of cells to force. Annu. Rev. Biomed. Eng. 11,
259–288. This paper reviews rheological properties of cells with dis- Munevar S, Wang Y & Dembo M (2001) Traction force microscopy of
cussion of various techniques and what the results from different migrating normal and H-ras transformed 3T3 fibroblasts. Biophys J.
techniques may have in common. 80 4, 1744–1757. This article describes traction force microscopy
results applied to cell analysis.
Huang H, Kamm RD & Lee RT (2004) Cell mechanics and mecha-
notransduction: pathways, probes, and physiology. Am. J. Physiol. Neuman KC & Block SM (2004) Optical trapping. Rev. Sci. Instrum. 75,
Cell Physiol. 287, C1–C11.This journal review article outlines a 2787–2809. This paper describes the state of optical trapping and
few experimental methods for studying cell mechanics and some the necessary components for achieving it.
underlying bases for the various approaches taken.
Roy P, Rajfur Z, Pomorski P, Jacobson K (2002) Microscope-based
Janmey PA & McCulloch CA (2007) Cell mechanics: integrating techniques to study cell adhesion and migration. Nat. Cell Biol. 4,
cell responses to mechanical stimuli. Annu. Rev. Biomed. Eng. 9, E91–6. This journal review article discusses several microscopy-
1–34. This journal review article that discusses mostly theoretical based techniques to characterize cell adhesion structures and
and sensory mechanisms for understanding cell mechanics and forces, including FRET and traction force microscopy.
mechanotransduction, but is also notable for presenting a nice Sotoudeh M, Jalali S, Usami S, Shyy JY, Chien S (1998) A strain device
summary in Table 1 that compares the elastic modulus by different imposing dynamic and uniform equi-biaxial strain to cultured cells.
techniques and cells with extensive references. Ann. Biomed Eng. 2, 181–189. This article discusses the design and
Kuo SC (2001) Using optics to measure biological forces and implementation of a stretch device used for cell mechanics studies,
mechanics. Traffic 2, 757–763. This journal review article summa- with validation of the exerted strains.
rizes some key findings for using optical traps and microrheology to Tan JL, Tien J, Pirone DM, et al. (2003) Cells lying on a bed of
study biological systems. microneedles: an approach to isolate mechanical force. Proc. Natl.
Landau LD & Lifshitz EM (1970) Theory of Elasticity, vol. 7, 2nd ed. Acad. Sci. USA 100, 1484–1489. This article describes the use of
Pergamon Press. This book, part of the Course of Theoretical Phys- micropillars (they call it microneedles) to determine traction forces
ics, contains many formulae that are of relevance when trying to using beam-bending analysis.
characterize deformations in solid objects. It does have newer edi- Wang JH & Lin JS (2007) Cell traction force and measurement
tions and the other books in the series cover different topics. methods. Biomech. Model. Mechanobiol. 6, 361–371. This paper
Legant WR, Miller JS, Blakely BL et al. (2010) Measurement of describes traction-force microscopy and a few different techniques
mechanical tractions exerted by cells in three-dimensional matri- for analysis of the results to extract the traction field.
ces. Nat. Meth. 7, 969–971. This is a recent journal paper that Wang N, Tolić-Norrelykke IM, Chen J et al. (2002) Cell prestress. I.
describes an approach for extending traction force microscope stiffness and prestress are closely associated in adherent contrac-
from a two-dimensional monolayer to cells imbedded in a three- tile cells. Am. J. Physiol. Cell Physiol. 282, C606–C616. This journal
dimensional hydrogel. article characterizes cell structures (including those based on the
tensegrity hypothesis) using information from a variety of tech-
Liu J, Weller GE, Zern B, Ayyaswamy PS, Eckmann DM, Muzykantov
niques, including traction force and magnetic twisting.
VR, Radhakrishnan R (2010) Computational model for nanocarrier
binding to endothelium validated using in vivo, in vitro and atomic Weibel DB, Diluzio WR & Whitesides GM (2007) Microfabrication
force microscopy experiments. Proc. Natl. Acad. Sci. USA 38, 16530– meets microbiology. Nat. Rev. Microbiol. 5, 209–218. This journal
16535. This article uses computational methods to calculate the review article summarizes different micropatterning techniques
free energy of binding a functionalized “nanocarriers” to endothe- and their applications to biological systems.
lial cells.
Young EW & Simmons CA (2010) Macro- and microscale fluid flow
Moseley JB, O’Malley K, Petersen NJ et al. (2002) A controlled trial of systems for endothelial cell biology. Lab Chip 10, 143–160. This
arthroscopic surgery for osteoarthritis of the knee. N. Engl. J. Med. paper discusses shear flow design and analysis and presents prin-
347, 81–88. This is a journal article describes a study in humans ciples for shear flow design at the microscale by taking keys from
of arthroscopic knee surgery in which the controls received a macroscale shear flow devices.
This page intentionally left blank
to match pagination of print book
PART II: praCtiCES
This page intentionally left blank
to match pagination of print book
CHAPTER 7
Mechanics of Cellular
Polymers

W e now turn our attention to one of the most important structural compo-
nents of cells: biopolymers. In Section 2.1, we learned that polymers are
linear molecules made up of repeating structures known as subunits. Biopolymers
possess a diverse range of functions within cells. For example, DNA and RNA are
polymers of nucleotides whose primary function is the storage and passage of
genetic information. In Chapter 8, we will learn more about the cytoskeleton, an
interconnected network of polymers that plays a critical role in a variety of func-
tions, such as maintenance of cell shape, mechanical force generation, and intra-
cellular transport. In many cases, the mechanical behavior of these biopolymers
is critical in allowing them to perform the diverse functions they serve.
In this chapter, we focus on obtaining a quantitative understanding of the mechan-
ics of biopolymers. We first describe the molecular structure of three important
cytoskeletal polymers: microfilaments, microtubules, and intermediate filaments.
We follow this with a model of polymerization dynamics in microfilaments and
microtubules. Next, we consider how to characterize a polymer as being flexible
or stiff and discuss three different polymer models: the ideal chain, the freely
joined chain (FJC), and the wormlike chain (WLC). Through this development, it
is our hope that you will have a better understanding of how the mechanics of
biopolymers may dictate their capacity to perform their biological functions.

7.1 BIOPOLYMER STRUCTURE

Microfilaments are polymers composed of actin monomers
Microfilaments (MFs) are polymers composed of the protein actin, a highly con-
served, 42 kDa protein. In its monomeric form, it can be referred to as globular or
G-actin. When G-actin polymerizes, it forms filamentous actin, referred to as
F-actin. F-actin has a double-helix-like structure consisting of two strands that
spiral around the axis of the polymer (Figure 7.1). This results in the appearance
of repeating helix loops, with each loop repeating every 37 nm. The overall diam-
eter of the polymer is approximately 7–9 nm.

F-actin polymerization is influenced by the molecular

characteristics of G-actin
In the cell, there is a highly dynamic and regulated balance between the G and F
forms of actin. This balance is under strict biochemical regulation within the cell.
This regulation allows actin to adopt different morphologies and organization and
perform a variety of cellular functions, such as intracellular force generation dur-
ing cell motility and maintenance of cell shape.
The molecular structure of G-actin can greatly affect the dynamics of its polym-
erization into F-actin. Actin is a polar molecule. Unlike the charge polarity asso-
ciated with water (as we will describe in Section 9.1), here polar means that the
two ends of the monomer are different. The implication of this polarity is that
190 CHAPTER 7: Mechanics of Cellular Polymers

+ end actin molecule

NH2 + end
COOH

37 nm

ATP
(ADP when
in filament)

− end
− end

25 nm
(A) (B) (C)
Figure 7.1 G-actin and F-actin structures. G-actin possesses a (+) and a (–) end as well as binding site for ATP. F-actin has a double-helix-
like structure consisting of two strands that spiral around the axis of the polymer, with each helical loop repeating every 37 nm. (Adapted
from, Alberts B et al. (2008) Molecular Biology of the Cell, 5th ed. Garland Science. Photo courtesy of Roger Craig.)

actin MFs have inherent directionality. This directionality was originally

deduced from the observation that when a MF is decorated by multiple myosin
molecules, the myosins tend to be angled, all in the same direction. This gave
rise to the notation of a pointed (or “ − ”) and a barbed (or “ + ”) end of the poly-
mer and the monomer (Figure 7.2). Owing to this polarity, the polymerization
kinetics of actin, in other words the rates at which monomers are added or sub-
tracted from the end of the polymer, can be very different at the (+) and (−) ends.
Another aspect of actin’s molecular structure that can affect its polymerization
kinetics is its capacity to bind adenosine triphosphate (ATP). Upon binding to
actin, ATP hydrolyzes to adenosine diphosphate (ADP). This hydrolysis happens
relatively quickly for polymeric F-actin, but is quite slow for monomeric G-actin.
As a result, almost all of the G-actin in a cell is of the ATP-bound form. In con-
trast, the longer an actin subunit is polymerized within a MF, the more likely it
will be in the ADP-bound form. In Section 7.2, we explore in detail how actin
polarity and ATP/ADP binding affects its poly­merization kinetics.

Figure 7.2 The two ends of a MF are

denoted as the barbed, or (+), end
and the pointed, or (–) end.

barbed end pointed end

+ end − end
 BIOPOLYMER STRUCTURE 191

Microtubules are polymers composed of tubulin dimers

Another class of cytoskeletal polymers is the microtubules (MTs). The subunits for
microtubule polymers are heterodimers, with each subunit made up of an α and
β isoform of the 55 kDa protein tubulin. The structure of the MT consists of 13
protofilaments joined together in a hollow tubular structure (Figure 7.3). The
protofilaments are offset such that the dimers form a roughly helical structure,
with each helical turn containing 13 subunits. The external diameter of a MT is lumen
approximately 25 nm. This is larger than that of F-actin and contributes to the
increased flexural rigidity of MTs compared with MFs. Biologically, MTs are cen- + end
tral players in several processes such as mitosis, in which they form the mitotic
spindle responsible for segregation of the chromosomes (Figure 7.4). MTs also
contribute to cell shape and migration and forming the structure of cilia and some
flagella.

50 nm
MT polymerization is affected by polarity and GTP/GDP binding
The polymerization kinetics of tubulin are dictated by its asymmetric nature and
its ability to bind guanosine tiphosphate/guanosine diphosphate (GTP/GDP,
analogous to ATP/ADP in actin). Like F-actin, MTs are polar, possessing a rap-
idly polymerizing (+) end and slowly polymerizing (−) end. Once polymerized,
GTP hydrolyzes to GDP. As tubulin polymerizes, there is a region near the
polymerizing end of GTP-tubulin known as the GTP cap (Figure 7.5). Only GTP
tubulin tends to become polymerized because polymers of GDP tubulin are
unstable and easily return to monomeric form. The GTP cap can be thought of − end
as preventing the depolymerization of the entire MT.
microtubule
Periodically, the GTP cap is lost owing to impediments to further polymeriza-
tion or simple random fluctuations. When this happens, the GDP MT cata-
β tubulin heterodimer
strophically depolymerizes until the GTP cap can be reestablished. Micro­tubules
are constantly shifting from a state of slow polymerization to rapid depolymeri- α (= microtubule subunit)
zation in a process known as dynamic instability. Dynamic instability results in
MTs constantly extending and retracting. This is thought to be important for Figure 7.3 Structure of a MT. Each MT
consists of 13 protofilaments joined
the ability of the MTs to find and attach to the kinetochore during mitosis. At
together in a hollow tubular structure.
other times, cytoplasmic MTs tend to organize into a network with their (−) MT subunits consist of a heterodimer of
ends near the nucleus and their (+) ends pointing outward toward the cell α- and β-tubulin. (Adapted from, Alberts
membrane. B et al. [2008] Molecular Biology of the
Cell, 5th ed. Garland Science.)

kinetochore Nota Bene

Taxol blocks dynamic instability.
The activity of certain pharmacologic
therapies is derived from their effect
on the dynamic instability of MTs.
Taxol®, an anti-cancer drug, works by
spindle making the GDP form of tubulin stable.
This blocks dynamic instability and
inhibits mitosis.

Figure 7.4 MTs during mitosis. When cells divide, MTs are responsible for segregating
chromosomes and pulling them to the poles of the cell. They do this by forming two spindles
and attaching to the kinetochores.
192 CHAPTER 7: Mechanics of Cellular Polymers

Figure 7.5 A polymerizing MT with

a GTP cap. As tubulin polymerizes, GTP +
there is a region of GTP + tubulin. The tubulin
GTP cap arises because only GTP +
tubulin tends to become polymerized as
it is an unstable polymer unstable and
easily returns to monomeric form. GTP +
GDP + tubulin
tubulin

Intermediate filaments are polymers with a diverse

range in composition
A third type of cytoskeletal biopolymer is the intermediate filaments. The struc-
ture and function of intermediate filaments are somewhat less well-characterized
than MFs and MTs, perhaps because of their complexity. Over 70 different genetic
sequences appear to code for the various intermediate-filament proteins (Table
7.1). Intermediate filaments are found in a variety of locations within the body
and occur in multiple variants, depending on their location. Vimentin is a com-
monly studied intermediate filament in epithelial cells. Desmin is an intermedi-
ate filament that is more specific for muscle, especially the heart. Lamins are
intermediate filaments that reside in the nuclear envelope, the membrane that
surrounds the nuclei of cells. Keratins are another type of intermediate filament;
there are over two dozen types of keratin (and likely more that have not yet been
studied in detail). In humans, these intermediate filaments reside in the skin but
are also expressed in hair and fingernails. In other animals, keratin is a major
component of horns, hides, and scales.

Intermediate filaments possess a coiled-coil structure

Regardless of protein composition, all intermediate filaments have an α-helical
structure and adopt a “coiled-coil” structure. Unlike actin and MTs, intermediate
filaments are not composed of small globular subunits. Instead, the individual
subunits of intermediate filaments are long α-helical regions of a protein. Two
such proteins are coiled together to form a dimer. Two dimers are then staggered
Nota Bene (in opposite orientations) to form a tetramer. Long chains of tetramers are then
Naming of intermediate coiled together to form the filament (Figure 7.6). The association between fila-
filaments. Although the size of ments arises out of hydrophobic interactions of the coiled regions. Being 10 nm in
intermediate filaments is between diameter, intermediate filaments are larger in diameter than MFs, but smaller
that of MFs and MTs, they actually than MTs. Unlike MFs and MTs, intermediate filaments are not polar and do not
get their name because they were have (+) and (−) ends. In addition, their structure does not facilitate rapid depo-
first described as having a diameter
lymerization, as the filament has to first uncoil before depolymerization can
between MFs and myosin fibers.
occur.

Intermediate filaments have diverse functions in cells

In general, disorders caused by dysfunction of intermediate filaments and related
proteins are quite varied. As a result, a well-defined function for intermediate fila-
ments has not yet been established. However, they are thought to provide mechan-
ical strength to tissues in some cases. They attach to membrane plaque-like
structures known as desmosomes that mediate cell–cell adhesions. Certain defects
in desmosomes appear more pronounced in highly mechanically stressed tissues
expressing them, including skin and cardiac tissue. Some intermediate filaments,
such as keratin, can also attach to the extracellular matrix protein laminin—not to
be confused with the intermediate filament lamin—by integrins using structures
 BIOPOLYMER STRUCTURE 193

called hemidesmosomes. This adhesion occurs in a manner similar to focal adhe-

sions, although the latter rely on MFs. Although hemidesmosomes have roles in
mediating cell–matrix interactions, the regulatory role of hemidesmosomes is not
clearly established.

Table 7.1 ​Intermediate filaments by type, size, chromosome, and distribution Note that there are multiple subtypes for
each class; keratins, for example, consist of 20 different type 1 filaments. (From Omary MB, Coulombe PA & McLean WH.
(2004) N. Engl. J. Med. Copyright Massachusetts Medical Society.)

Location and Type Size kDa Chromosome Cell or tissue Comments

name with associated distribution
gene

Cytoplasmic
Keratins I 40–64 17 Epithelium (keratins Form obligate 1:1 heteropolymers with
9–20); hair (keratins type II; protect from mechanical and
Ha1–Ha8) nonmechanical forms of stress
II 50–68 12 Epithelium (keratins Form obligate 1:1 heteropolymers with
1–8); hair (keratins type I; protect from mechanical and
Hb1–Hb8) nonmechanical forms of stress
Vimentin III 55 10 Mesenchyme Involved in vascular tuning and wound
repair in mice
Desmin III 53 2 All muscle May be important for mitochondrial
positioning and integrity
Glial fibrillary acidic III 52 17 Astrocytes Also found in hepatic stellate cells
protein
Peripherin III 54 12 Peripheral neurons Found in enteric neurons; may be
required for development of a subset of
sensory neurons
Syncolin III 54 1 Muscle (mainly Interacts with α-dystrobrevin
skeletal and cardiac)
Neurofilaments IV 61 (light), 90 8 (light), 8 Central nervous Form obligate 5:3:1 (light:medium: heavy)
(light, medium, and (medium), (medium), 22 system heteropolymers
heavy chains) 110 (heavy) (heavy)
α-Internexin IV 61 10 Central nervous May partly compensate for peripherin in
system peripherin-null mice
Nestin IV 240 1 Neuroepithilial Is also an early developmental marker, as
found in the pancreas
Synemin IV 180 (α) 15 All muscle Found at lower levels than desmin; also
and 150 (β) (ß-isoform mainly in found in astrocytes; probably indentical
(two splice striated muscle) to desmuslin
variants)
Nuclear
Lamins A and C V 62–78 1 Nuclear lamina Arise from a single, differentially sliced
gene
Lamins B1 and B2 V 62–78 5, 19 Nuclear lamina Arise from two different genes
Other
Phakinin (CP49) Orphan 46 3 Lens Forms beaded filaments; deletion in mice
causes lens defect
Filensin (CP115) Orphan 83 20 Lens Forms beaded filaments; deletion in mice
causes lens defect
194 CHAPTER 7: Mechanics of Cellular Polymers

NH2 COOH
(A)
α-helical region in monomer

NH2 COOH
(B)
coiled-coil dimer
NH2 COOH
48 nm

COOH NH2
NH2 COOH
(C)

COOH NH2
NH2 COOH
staggered tetramer of two coiled-coil dimers

(D)

two tetramers packed together

(E)

eight tetramers twisted into a ropelike filament

10 nm

Figure 7.6 Intermediate filament structure. The monomer (A) forms a dimer with another monomer to form a coiled-coiled structure
(B). Two dimers are then staggered to form a nonpolar tetramer (C). The tetramers associate in a staggered manner (D), allowing them to
form the final helical filament structure (E). (Adapted from, Alberts B et al. (2008) Molecular Biology of the Cell, 5th ed. Garland Science.)

7.2 POLYMERIZATION KINETICS

In the previous section, we learned that F-actin and MT polymerization are tightly
regulated intracellular processes. In addition, we learned that actin monomers
and tubulin dimers share two distinct features in their molecular structure that
affect their polymerization dynamics: they are both polar molecules, and they
both possess the capacity to bind ATP/ADP (for actin) or GTP/GDP (for MTs). In
this section, we explore how these structural features affect their polymerization
kinetics.
 POLYMERIZATION KINETICS 195

polymer subunit polymer Figure 7.7 Dynamics between the G

and F forms of actin. Reversible
... + ...
conversion of G-actin to F-actin occurs
F-actin G-actin F-actin with the addition of a monomer to a
polymer at one of its free ends.

Actin and MT polymerization can be modeled as a bimolecular

reaction
Consider a model polymer of some length undergoing the reversible addition of
one subunit, as in Figure 7.7. If we denote the existing polymer before the subunit
addition to be Apoly, and after the addition as Apoly+1, we can write the polymeriza-
tion process as a chemical reaction as
kon
Apoly + G  Apoly +1 , (7.1)
koff

where kon is the rate of the forward reaction (units of s−1 M−1) and koff is the rate
of the reverse reaction (units of s−1). We can write the rate of change of [Apoly+1] as

d  Apoly +1 
= kon  Apoly [G ] − koff  Apoly+1  . (7.2)
dt
When the polymer is not changing length, the time derivative of [Apoly+1],
d[Apoly+1]/dt, is equal to zero. When this occurs, Equation 7.2 implies that

kon  Apoly [G ] = koff  Apoly +1  . (7.3)

Solving for [G] in Equation 7.3, we obtain

 Apoly+1  koff
[G ] = K  where K = . (7.4)
 Apoly  kon

The constant K is known as the dissociation constant of the reaction. It is also
known as the critical concentration, as it is the concentration of subunit at which
the polymer is neither growing nor shrinking.

The critical concentration is the only concentration at which

the polymer does not change length
The critical concentration K has an interesting property for our model polymer:
it is the only subunit concentration at which the polymer does not change
length. To see this, we can write an expression for the elongation rate of the
polymer; in other words, the change in length of the polymer per unit time. We
first note that if δ is the size of the subunit, then the rate of lengthening in units
of length per unit time due to polymer addition is kon[G]δ, and that due to sub-
traction is koff δ. The total elongation rate dL is the sum of these two rates,
dt
dL
= (kon[G ] − koff )δ. (7.5)
dt

A plot of Equation 7.5 can be seen in Figure 7.8. If [G] > K, dL is positive, the poly-
dt
mer is growing, whereas if [G] < K, dL is negative, the polymer is shrinking. The
dt
196 CHAPTER 7: Mechanics of Cellular Polymers

Figure 7.8 The kinetics of dL

polymerization as a function of elongation rate —
(dt)
monomer concentration. If the
monomer concentration [G] is higher
than K, the elongation rate is positive,
and the polymer will grow. If it is below
K, the polymer will shrink. The only elongation
concentration at which the elongation
rate is zero is when [G] = K.

K [G]
shrinking

only point at which dL = 0, in other words when the polymer is neither growing
dt
nor shrinking, occurs when [G] = K. This confirms that the critical concentration
is the only subunit concentration at which the polymer does not change length.

Polarity leads to different kinetics on each end

Thus far we have assumed that the subunits were not polar and that the kinetics
of polymerization were equivalent at each end of the polymer. Now we examine
the implications of subunit polarity. The kinetics of the polymerization reac-
tion are different at the (+) and (−) ends of the polymer. At the (+) end, kon and
koff are higher than at the (−) end. This implies that if an excess of monomers
were rapidly introduced, elongation would occur faster at the (+) end.
Conversely, if the monomers were rapidly depleted, depolymerization would
occur much faster at the (+) end. For our plot of elongation rate in Figure 7.8,
we will have two different lines, one for the (+) end, and one for the (−) end.
Although it is beyond the scope of this book, it can be shown by a principle
called detailed balance that the critical concentration at both ends of a polar
polymer must be the same if it is homogenously composed of a single species
of subunit. If the critical concentration is the same at both the (+) and (−) ends,
the lines will cross the zero axis at the same point. In Figure 7.9, we revise our
plot for elongation rate to reflect that there are now two different lines, one for
the (+) end, and one for the (−) end, each passing through dL v = 0 at [G] = K.
dt

Figure 7.9 Polymerization kinetics dL

differ at the (+) and (−) ends. The elongation rate —
(dt)
kinetics of polymerization are different
at the (+) and (−) ends of the polymer. + end
Polymerization/depolymerization
kinetics tend to be faster at the (+) end
and slower at the (−) end for a given elongation − end
subunit concentration. However, the
critical concentration is the same at
both ends.
K [G]
shrinking
 POLYMERIZATION KINETICS 197

Polymerization kinetics are affected by ATP/ADP

in actin and GTP/GDP binding in tubulin
To complete our discussion, we now seek to incorporate the effects of ATP/ADP
binding for actin and GTP/GDP binding for tubulin. For simplicity, we will refer to
ATP-bound actin and GTP-bound tubulin as T-form subunits, and ADP-bound
actin and GDP-bound tubulin as D-form subunits. Recall from Section 7.1 that
within T-form subunits, the ATP or GTP can undergo hydrolysis, which results in
the conversion of a T-form subunit into a D-form subunit. This hydrolysis has two
effects on the polymerization kinetics of our model polymer. First, recall that the
primary form of the subunit in the cytoplasm is T form; however, a given polymer
may contain both D- and T-form subunits. The rates of polymerization will be dif-
ferent depending on whether the end of the polymer that the subunit is binding
contains a D-form or a T-form subunit. Second, hydrolysis has the effect of lower-
ing the affinity of the subunits to the polymer, which increases the critical concen-
tration for the D form. In Figure 7.10, we take these effects into account in our
elongation rate plots by having one set of (+)/(−) lines for D-form actin, and one
set of (+)/(−) lines for T-form actin. The critical concentration for T-form actin is
denoted KT, and the critical concentration of D-form actin KD is denoted.

Subunit polarity and ATP hydrolysis lead to polymer

treadmilling
Imagine a polymer composed entirely of D-form subunits. The polymer is quickly
submerged in a bath of T-form subunits. If the subunit concentration is higher
than KT, T-form subunits will be added at both ends, with faster growth occurring
at the (+) end. Therefore, at any given point in time, the T-form region on the (+)
end will be longer than that on the (−) end.
Now consider what happens when we lower the subunit concentration, perhaps
by diluting the solution with water. As the subunit concentration is reduced, the
rate of T-form subunit addition slows. In addition, the recently polymerized
T-form subunits begin undergoing hydrolysis. If the rate of hydrolysis is faster
than the rate of T-form addition at the slower (−) end, then at some point in time
all the T-form actin at the (−) end will have been hydrolyzed to D form. However,
because there are more T-form subunits at the (+) end, there will still be some
unhydrolyzed T-form subunits there. The polymer possesses two critical concen-
trations: KT at the (+) end and KD at the (−) end.
The critical concentration has interesting ramifications when the monomer con-
centration is lower than KD but greater than KT. The (−) end will be depolymerizing,

dL Figure 7.10 Polymerization kinetics

elongation rate — for D- and T-form actin. The critical
(dt)
concentration and rates of
+ + polymerization/depolymerization are
− different for the D and T forms of actin.
elongation −

KT KD [G]
shrinking
198 CHAPTER 7: Mechanics of Cellular Polymers

Nota Bene but the (+) end will be polymerizing. The addition of subunits on one end and
subtraction at the other is known as treadmilling. Once D-form subunits are lost
Polymerization force. The peak
from the shrinking (−) end, they release ADP/GDP and bind ATP/GTP to become
force that a single fixed polymerizing
filament can exert on a surface has T-form subunits and are recycled to the growing (+) end. Intracellular concentra-
been derived through tions of G-actin and tubulin are typically within the range of subunit concentra-
thermodynamic considerations to be tions at which treadmilling occurs.

kT  k [G ] 
F = ln  on  .
δ  koff  7.3 PERSISTENCE LENGTH
Using this relation, it has been
estimated that the peak force for We now shift our focus from polymer structure and kinetics to mechanical behav-
F-actin is 9 pN for a typical ior. In analyzing the mechanical behavior of polymers, the appropriate choice of
concentration of G-actin of 50 mM. model is determined in large part by whether the behavior is energy- or entropy-
This is many times the force that one dominated. We learned in Section 7.1 that different biopolymers can have very
myosin motor can generate. different molecular structures. One consequence of this is that under the influ-
ence of thermal forces, these polymers will have very different tendencies to bend
and coil. Consider a strand of DNA approximately 10 μm in length. If we grab it at
the ends, straighten it, and then release it in into a solution at room temperature,
L = 10 mm we would see the strand of DNA begin to take on a convoluted configuration, with
many undulations forming along its length. Now, we do the same for a MT 10 μm
microtubule in length. The MT does not coil. In fact, it may tend to resemble a fairly straight rod
(Figure 7.11).
actin The tendency for these two polymers to take on such different configurations in
the presence of thermal forces can be attributed to differences in energetic and
DNA entropic influences in equilibrium. The molecular structure of MTs results in a
much higher increase in potential energy during bending than that of DNA.
Figure 7.11 Representative The behavior of the MTs is energy-dominated taking on a straight configura-
configurations for a 10-μm segment tion, as this reduces the internal energy and therefore the free energy.
of a MT, actin, and DNA polymer at Conversely, the behavior of DNA is entropy-dominated. The free energy is
room temperature. Under the
reduced by increasing entropy, which occurs when the DNA strand takes on a
influence of thermal forces, the
tendency to take on a coiled versus coiled configuration.
straight configuration varies for each of
the polymers. Persistence length gives a measure of flexibility in a thermally
fluctuating polymer
In this section, we introduce the persistence length, which is a characteristic
length scale that gives a measure of flexibility in a thermally fluctuating poly-
mer. Consider a continuous polymer of contour length L undergoing thermal
fluctuations, as in Figure 7.12. We define a quantity s that runs from zero to L
and gives a parameterization by which each point on the polymer can be iden-
tified. We define the orientation at each point θ(s) as the angle the polymer
makes with an imaginary horizontal line. At s = 0, the polymer end is fixed such
that θ(0) = 0.
Now consider the case where we monitor θ(s) over time, and draw the probability
distribution for θ. If we were to draw the distribution for small s (near the fixed

Figure 7.12 Relevant quantities in free end

the definition of persistence (s = L)
length. The quantity s represents the polymer undergoing
distance along the polymer from 0 to L. thermal fluctuations
The angle the polymer makes with an
imaginary horizontal line at each point s θ(s)
is given by θ(s). fixed end
(s = 0)
 PERSISTENCE LENGTH 199

end), then we would expect that the chances for the polymer to have an orienta-
tion different from θ = 0 would be very small, and therefore the distribution would
be very sharp, with a peak at θ = 0. If we were to find the average cosine of the
angle, we would find that 〈cos θ〉 ≅ 〈cos 0〉 ≅ 1, where the brackets indicate the
time average. In contrast, at larger s, we would expect a much higher chance for
the polymer to have a different orientation than θ = 0, so the distribution would be
wider (Figure 7.13). The distribution becomes wider and wider for larger s until,
at some point, the distribution becomes essentially uniform. Beyond this point,
the polymer orientation is effectively random (i.e., the orientation becomes
uncorrelated from the fixed end), and so 〈cos θ〉 ≅ 0. Therefore 〈cos θ〉 decreases
from 1 to 0 as s gets larger and larger. In fact, we now show that this decrease
occurs exponentially.
Consider the following approximation for the derivative of some function f (s),

df f (s + ∆s ) − f (s )
≅ (7.6)
ds ∆s
derived from the first two terms of a Taylor series approximation. Equation 7.6 is a
good approximation for small values of Δs, as long as f (s) is reasonably smooth. In
our case, f (s) = 〈cosθ ′(s)〉 , where θ ′(s) = θ(s) − θ(0), and

df cos (θ ′(s + ∆s )) − cos (θ ′(s ))

≅ . (7.7)
ds ∆s
Now, if we let Δθ ′(s) = θ ′(s + Δs) − θ ′(s), we can rearrange this to read
θ ′(s + Δs) = Δθ ′(s) + θ ′(s), and we can substitute this expression into the numerator

df cos ( ∆θ ′(s ) + θ ′(s )) − cos (θ ′(s ))

≅ . (7.8)
ds ∆s
Using the fact that Δθ ′(s) and θ ′(s) are independent quantities, we can use the
identity cos(a + b) = cos(a)cos(b) − sin(a)sin(b) in the above expression:

df cos ( ∆θ ′(s ))cos (θ ′(s )) − sin ( ∆θ ′(s ))sin (θ ′(s )) − cos (θ ′(s ))
≅ . (7.9)
ds ∆s
Again, using the fact that Δθ ′(s) and θ ′(s) are independent, we can use the identity
〈ab〉 = 〈a〉 〈b〉, and

df cos ( ∆θ ′(s )) cos (θ ′(s )) − sin ( ∆θ ′(s )) sin (θ ′(s )) − cos (θ ′(s ))
≅ . (7.10)
ds ∆s

Figure 7.13 Probability distribution

short s (sharp of orientation of the free polymer
distribution) end for different values of s. For
short s
very short s, the probability for the
polymer to have an orientation different
probability

from θ = 0 is small, and therefore the

distribution is very sharp. At large s,
long s there is a much higher probability and
long s (wider so the distribution widens. At very large
s, the distribution is essentially uniform.
distribution)
very long s
(uniform)
very long s
0
− 0 
orientation of free end (radians)
200 CHAPTER 7: Mechanics of Cellular Polymers

However, because Δθ ′(s) and θ ′(s) are equally likely to be negative or positive
(symmetric about zero, odd functions like sines average to zero),

df cos ( ∆θ ′(s )) cos (θ ′(s )) − cos (θ ′(s ))

≅ . (7.11)
ds ∆s
Finally, factoring our expression for f (s), we obtain

df  cos ( ∆θ ′(s )) − 1 
≅  f (s ). (7.12)
ds  ∆s 
Nota Bene
Three-dimensional orientation
Notice that the term in the parentheses is a constant, not dependent on s, and in
correlation function. If the motion
of the polymer is three-dimensional, addition it will generally be negative, because 〈cos[Δθ ′(s)]〉 is generally less than 1.
then the orientation correlation We can rewrite the above expression as
function is
df
 −s  ≅ −Cf (s ). (7.13)
cos ∆θ ( s ) = e
 
 2p 
. ds
A solution to this equation is
The different normalization takes into
account the fact that if the polymer is  −s 
allowed to move in three dimensions,   (7.14)
 p 
it can bend in two directions cos ∆θ (s ) = e ,

orthogonal to its long axis, whereas if
it is constrained to move in two where for the rest of the chapter we will use the notation Δθ(s) = θ ′(s), 〈cos
dimensions, it can only bend in one Δθ(s)〉 is called the orientation correlation function, C = 1/ℓp, and ℓp is a nor-
direction. Therefore in two malization factor known as the persistence length. From the above expression,
dimensions, the orientation it is easy to see that the persistence length gives a characteristic length scale
correlation function will decay
half as fast.
over which the orientations of a thermally undulating polymer become mostly
uncorrelated.

Example 7.1: Consider a segment of actin in relation to a segment of DNA

Calculate the length of each segment such that the change The corresponding lengths for actin and DNA are 1.5 μm
in angle between the two ends of each segments is, on and 5-nm, respectively. Physically, this means that on
average, 25 degrees. Assume the persistence lengths for average, a 1.5-μm span of a thermally fluctuating actin
actin and DNA are 15 μm and 50 nm, respectively. polymer would have a difference in angle of 25 degrees
from one end to the other, whereas in DNA this would
Using Equation 7.14 we can solve for s by taking the
occur over a 5-nm span. These dimensions make some
natural logarithm of both sides and isolating s to obtain
sense considering the functions of each polymer; actin
s = −ℓp ln(〈cosΔθ(s)〉). filaments must have the capacity to span significant
lengths within the cell, whereas DNA must have the
We know that a change in angle of 25 degrees corre- capacity to be tightly coiled within the nucleus.
sponds to a value for the orientation correlation function,
〈cosΔθ(s)〉, of 0.9. Using this value, we obtain

s = 0.1ℓp.

Persistence length is related to flexural rigidity for an

elastic beam
In Section 3.2, we learned that bending of an elastic beam is governed by its flex-
ural rigidity, the product of its Young’s modulus and moment of inertia EI. We
now show that if we model a thermally fluctuating polymer as a curvy elastic
beam, its persistence length is proportional to the flexural rigidity of the beam.
 PERSISTENCE LENGTH 201

Consider an elastic three-dimensional beam with flexural rigidity EI bent 180

degrees (π radians) with a constant curvature R. In Chapter 3, you may have com-
puted the elastic energy for this beam as a homework problem; however, for con-
venience we state it here as
EI π
Q= . (7.15)
2R

Because the beam was bent π radians, we can express Equation 7.15 in terms of a
general bend angle θ as
EIθ
Q(θ ) = . (7.16)
2R

Using the fact that the bend angle can be expressed in terms of arc length, θ = s/R,
we can rewrite Equation 7.16 as
EIθ 2
Q(θ ) = . (7.17)
2s

Equation 7.17 describes the internal energy of a beam of arc length s subject to a
constant curvature such that the bend angle is θ radians. We now assume we are
given a polymer of length s that can be modeled as an elastic beam. If we were to
immerse the polymer within a constant temperature heat bath, we could use
Boltzmann’s distribution to find the probability of finding a polymer with bend
angle θ as
1 −Q (θ )/kBT
p(θ ) = e . (7.18)
Z
To compute the partition function Z, we must integrate the exponential term over
two angles, θ and ϕ, to allow for bending in three dimensions. Specifically, Z can
be computed as
2π π

Z=
∫ ∫e
0 0
−Q (θ )/kBT
dφ sinθ dθ ,

(7.19)

where we have taken the integral with respect to a differential element of solid
angle dϕsinθdθ. We now seek to quantify the average amount of polymer curva-
ture. We can do this by computing 〈θ  2〉 as
2π π
1
θ 2
=
Z ∫ ∫e −Q (θ )/kBT
θ 2dφ sinθ dθ . (7.20)
0 0
For small angles, it is left to you to show that the solution to this integral is

2kBTs
θ2 = . (7.21)
EI
Now, we relate the above expression to the orientation correlation function. We
can do this using a Maclaurin series expansion. The Maclaurin series for cos x is
1 2 1 4
cos x = 1 − x + x ... (7.22)
2 24
Using this expansion and a small-angle assumption, we can write the orientation
correlation function as

∆θ 2 (s ) ∆θ 2 (s )
cos ∆θ (s ) ≈ 1 − = 1− . (7.23)
2 2
202 CHAPTER 7: Mechanics of Cellular Polymers

The term in brackets in the right-most relation in Equation 7.23 is the average
value of the difference in angle (squared) between two points separated by a dis-
tance s along the polymer. If the polymer is assumed to be a constant curvature
beam of arc length s, then we can substitute in our expression for 〈θ2〉 found in
Equation 7.23. In this case,

kBT (7.24)
cos ∆θ (s ) ≈ 1 − s.
EI

Using the approximation e−x ≈ 1 − x in the limit of small x, we can write our
expression for the orientation correlation function found in Equation 7.24 for
small s as

−s
−s
cos ∆θ (s ) = e  p ≅ 1 − . (7.25)
p

Relating Equations 7.24 and 7.25, we can see that the persistence length ℓp is
related to flexural rigidity as

EI
p ≡ . (7.26)
kBT

Equation 7.26 is a useful relation, because it allows us to estimate the effective

Young’s modulus of a polymer from its persistence length. We can observe the
thermal fluctuations of a polymer, and calculate cosΔθ(s) along the polymer
length. We do this multiple times and then find the average values at each point,
which gives 〈cosΔθ(s)〉. These points can be fit to an exponential to find ℓp. Finally,
assuming we know the temperature at which the measurements were made, we
can compute the flexural rigidity using Equation 7.26.

Polymers can be classified as stiff, flexible, or semi-flexible

by the persistence length
The persistence length can vary dramatically for different biopolymers. For
example, it is 50 nm for DNA, 15 µm for F-actin, and 6 mm for MTs. The differ-
ence in persistence lengths for these three biopolymers spans more than five
orders of magnitude! Because the persistence length gives a length over which
the orientations of a thermally undulating polymer become uncorrelated, it is a
natural length scale for the classification of a given polymer as being stiff or
flexible.
Let us revisit our example of a thermally fluctuating MT and strand of DNA,
each 10 μm in length. Recall that the MT tends to take on a rod-like configura-
tion, whereas the DNA tends to take on a more coiled configuration. For the
MT, its contour length is much smaller than its persistence length, L  ℓp. We
know that the contour length of the MT is much shorter than the length required
Table 7.2 ​Classification of polymer
flexibility based on persistence for the orientations of the polymer to become uncorrelated. Because the poly-
length mer will tend to resemble a straight rod, the polymer is considered “stiff”. In
contrast, for the strand of DNA, L  ℓp. We know that the contour length is
Type of polymer Persistence much greater than the length required for the orientations to become uncor-
behavior length related, so the DNA will tend to take on a very coiled configuration. Because it
Stiff
takes on a coiled configuration, the polymer is considered “flexible.” We can
ℓp  L
classify polymers of length L and persistence length ℓp as being stiff, flexible, or
Flexible ℓp  L semi-flexible, as in Table 7.2.

Semi-flexible It is important to note that classification of a polymer as being flexible, semi-flex-

ℓp ≈ L
ible, or stiff depends not only on its persistence length, but also its contour length.
 IDEAL CHAIN 203

L = 10 m L = 10 mm L = 10 µm L = 10 nm Figure 7.14 Representative

configurations of MT, actin, and
DNA polymers of various contour
microtubule lengths at room temperature. The
tendency to take on a straight versus
actin coiled configuration depends not only
on persistence length but also on
DNA contour length. A 10-nm-long DNA
segment is much shorter than the
persistence length, and the
configuration will tend to be rod-like. In
For very short DNA segments that are much less than the persistence length, the contrast, a 10-m-long MT has a contour
configuration will tend to be rod-like. In contrast, a very long MT that has a con- length much greater than its persistence
length and will take on a coiled
tour length much greater than its persistence length will take on a very coiled con- configuration.
figuration. This is demonstrated schematically in Figure 7.14.
In the next section, we introduce several common biopolymer models and their
resulting mechanical (force-extension) behavior. These different models differ
substantially in their assumptions. As a result, a given model may be good for
flexible polymers, but not for stiff polymers. As we will see, the appropriateness
of a given model is going to depend, in large part, on the persistence length of the
polymer, the length of the polymer, and the degree to which it is being extended.

7.4 IDEAL CHAIN

The ideal chain is a polymer model for flexible polymers
We start our introduction to biopolymer models with the ideal chain. The name
for this model comes from the fact that in the ideal chain all changes in internal
energy are ignored. It is often used for modeling flexible polymers whose behavior
is dominated by entropy.
Consider a chain of n segments of length b that are connected by freely rotating
joints as in Figure 7.15. Each segment in this model is referred to as a Kuhn seg-
ment, with the length of each segment called the Kuhn length. We will see later in
this chapter that the Kuhn length is related to the persistence length.
If the chain contains n segments, it will contain n − 1 rotating joints, or vertices,
between segments. Let ri be the segment vector from vertex i to vertex i + 1. There
are n of these vectors, and all of them have the same length b. The contour length
of the chain, or the length of the chain if it were completely straight, is nb. In gen-
eral, the chain will not be fully straightened out. We will describe the degree to
which it is straightened out by the end-to-end vector R, which is the sum of all
bond vectors in the chain
N

R= ∑r . i (7.27)
i =1
We would like to say something about the average end-to-end length. However, if
we average the random vector R, we get 0 due to symmetry. In particular, there is
no preferred spatial orientation for our polymer, so any vector R that occurs can

Figure 7.15 The model polymer for

the ideal chain. The polymer consists
R of rigid links of segment length b
connected by freely rotating hinges. The
end-to-end vector R and the bond
vector ri for link i are depicted.

b ri
204 CHAPTER 7: Mechanics of Cellular Polymers

also occur with equal probability as −R. Nonetheless, we can say something about
the average magnitude of R by examining its square

R2 = R ⋅ R

 N
  N 
= 

∑ i =1
ri  ⋅ 
 
∑ r 
j =1
j
(7.28)

N N

= ∑∑ r ⋅ r
i =1 j =1
i j .

We know that for the dot product of two vectors |a⋅b| = |a| |b| cos θab, where |a| and
|b| are the lengths of a and b, and θab is the angle between a and b. If we substitute
this expression into Equation 7.28, we obtain
N N

R2 = ∑∑b
i =1 j =1
2
cosθij

(7.29)
N N

=b 2
∑ ∑ cosθ
i =1 j =1
ij .

Furthermore, we know that the direction of any given segment is independent of
all of the other segments. The value of cosθab between any two segments will range
from −1 to 1. So, the average of cos θab will be 0, except when computed on the
same bond (a = b). In this case θab = 0 and cos 0 = 1. An equivalent statement is
〈cos θij〉 = 0 for i ≠ j, and 〈cos θij〉 = 1 for i = j, or 〈cos θab〉 =  δij. Therefore,
n n

∑∑δ
Nota Bene
R 2 = b2 ij = nb 2 . (7.30)
Kronecker delta. The Kronecker
i =1 j =1
delta δij is a compact notation used
to signify the following relation: This is the mean square end-to-end length of a three-dimensional chain com-
i = j posed of n segments of size b. Notice that just as in the random walk from Section
1, if
δ ij =  5.6, the quantity 〈R2〉 scales linearly with n. Only now, instead of n representing
0, if i ≠ j
the number of steps in the walk, it represents the number of links in the chain.

The probability for the chain to have different end-to-end

lengths can be determined from the random walk
With our model described, we now seek to determine the probability of the chain
having a specific length. This probability will allow us to determine the free energy
and ultimately the force required to maintain the chain at a given length, giving us
the chain’s force–displacement relationship.
The probability distribution function for the ideal chain is based on the random
walk. Let us return again to the one-dimensional random walk. Recall that the
probability of being at position r after n steps was given by Equation 5.64
M (n ,r ) 1 n!
p1d (n ,r ) = = n .
2 n
2 n+ r r
n− (7.31)
b! b!
2 2
and that the Gaussian approximation to this distribution was
1 2
/2 〈 r 2 〉
p1d (n ,r ) = e −r . (7.32)
2π〈r 〉 2

 IDEAL CHAIN 205

Equation 7.32 is for a one-dimensional random walk. We now seek to extend this
relation to three dimensions. To see the connection between one and three dimen-
sions, let us start by defining the end-to-end distance for our three-dimensional
random walk in the coordinate directions, R = Rxex + Ryey + Rzez (where ex, ey, and
ez are unit vectors in the x-, y-, and z-directions). The steps in each direction are
independent. So, the mean square end-to-end distance is the sum of the mean
square end-to-end distances along each of the three coordinate directions, or

R 2 = Rx2 + Ry2 + Rz2

(7.33)
= Rx2 + Ry2 + Rz2 .

Because there is nothing special about our choice of axes, each term in Equation
7.33 must be equal and the mean square end-to-end distance along one direction
is simply one-third of the total mean square end-to-end distance, in other words:

R2
Rx2 = Ry2 = Rz2 =
3
(7.34)
2
nb
= .
3

Because the three components of a three-dimensional random walk along the

three coordinate directions are independent of each other, the three-dimensional
probability distribution function for a random walk can be computed as the prod-
uct of three one-dimensional distribution functions in each of the three direc-
tions. But, we have to be careful. We have to account properly for the fact that the
one dimensional random walk is a projection of a three-dimensional random
walk into one dimension. More specifically, we have to scale the mean square dis-
tance in Equation 7.32 such that 〈r2〉 is equal to the mean square distance in one
dimension for our three-dimensional random walk,

〈r 2 〉 = 〈 Rx 2 〉 = 〈 Ry 2 〉 = 〈 Rz 2 〉 = 〈R 2 〉/3. (7.35)

Because each of these quantities is one-third of the three-dimensional average

squared end-to-end distance, 〈r2〉 = 〈R2〉/3. Combining Equations 7.32 and 7.35,
we obtain

3 2
2〈R2 〉
p1d (n , Rx ) = p1d (n , Ry ) = p1d (n , Rz ) = e −3 Rx . (7.36)
2π〈R 〉
2

The three-dimensional probability distribution is the product of three one-dimen-

sional distributions,

p3d (n ,R ) = p1d (n , Rx )p1d (n , Ry )p1d (n , Rz )

32
 3  −3 Rx 2 2 R 2 −3 Ry 2 2 R 2 −3 Rz 2 2 R 2
=  e e e
 2π R 2 

32
 3  (7.37)
−3( Rx2 + Ry2 + Rz2 ) 2 R 2
=  e
 2π R 2 

32
 3  −3 R 2 2 R 2
=  e .
 2π R 2 

206 CHAPTER 7: Mechanics of Cellular Polymers

Nota Bene We can express Equation 7.37 in terms of the number of segments n and the Kuhn
length b by substituting the relation 〈R2〉 = nb2. We obtain
Excluded volume interactions. In
the ideal chain, we assume that 3/2
segments can overlap. In other  3  2
/2 nb 2
p3d (n ,R ) =  e −3R , (7.38)
words, two segments can occupy the  2πnb 2 
same space at the same time. In
reality, this cannot occur. Models that
restrict segment overlap are said to which is the three-dimensional probability distribution for an ideal chain consist-
account for excluded volume ing of n segments of length b to have an end-to-end vector of R. Because the ideal
interactions. One such model that
accounts for excluded volume effects
chain is based on the above Gaussian distribution, it is also referred to as the
is based on the self-avoiding random Gaussian chain.
walk, a walk that cannot cross a point
it has traced previously.

Advanced Material: Gaussian approximation for the random walk

Here, we show that the probability distribution for a one- n 

dimensional random walk approaches a Gaussian distri- ln(p(n ,r )) = −n ln(2) + ln(n !) − 2 ln  !
2 
bution for large n. To begin, we first take the natural log
of the exact probability given in Equation 7.31 to yield R /2 R /2

∑ n 
∑ ln  2 + 1 − s
n
− ln  + s  +
ln(p(n , R)) = −n ln(2) + ln(n !) 2 
s =1 s =1

n+ R  n−R  n 
− ln  ! − ln  ! . (7.39) = −n ln(2) + ln(n !) − 2 ln  !
 2   2  2 

Now, if a, b, and c are positive integers such that a ≥ b, it  n 
R /2 +s
can be shown that ((a + b)/c)! can be expressed as
b/c
− ∑
s =1
 2
ln 
n


 + 1 − s 
a+b
∏ a 
a 2
!= !  + s  , (7.40)
c c s =1
c
 n 
= −n ln(2) + ln(n !) − 2 ln  ! (7.44)
and ((a − b)/c)! can be written as 2 

a  2s 
a−b ! R /2 1+
c
!= b/c
c . (7.41) − ∑ 
ln  n ,
2s 2 

∏s =1
a
c

 + 1 − s 

s =1  1 − + 
n n   

The third term in Equation 7.39 can be written as where we have divided the numerator and denominator
by n/2 in the last term in the last line.
R /2
n+ R   n!
∏ 2 + s
n We now invoke our large n approximation. In particular,
ln  ! = ln  
 2   2 in the last term of Equation 7.44, any term with n in
s =1
(7.42) the  denominator will go to zero in the limit of large
R/2
n.  Because ln(1 + a) ≅ a for |a|  1, we can approxi-

n 
= ln  ! +
2  ∑ s =1
n
2

ln  + s 


mate the logarithm in the last term of Equation 7.44 as

and the fourth term in Equation 7.39 as  2s 

1+
 n  = ln  1 + 2s  − ln  1 − 2s + 2 
ln     
n+ R 
R /2 2s 2  n n n
n 
∑ ln  2 + 1 − s .
n  1 + + 
ln  ! = ln  ! − n n (7.45)
 2 
(7.43)
2 
s =1 
4s 2
≅ − .
Combining Equations 7.39, 7.42, and 7.43, ln(p(n,R)) can n n  
be expressed as
 IDEAL CHAIN 207

Using Equation 7.45 and the identities Because ln(a) = b and a = eb are equivalent statements,
a
Equation 7.48 can be rewritten as

∑ s = a(a + 1)/2
s =1
(7.46)
1 n! − −
R2 R2
 p(n , R) ≅ n e 2n ≅ Ce 2n , (7.49)
2 (n/2)!(n/2)! 
and
a

∑1 = a ,
s =1
(7.47) where

1 n!
C= .
Equation 7.44 can be expressed as 2n n ! n ! (7.50)
2 2 
n 
ln(p(n , R)) ≅ −n ln(2) + ln(n !) − 2 ln  !
2  We can obtain a simplified expression for C by observing
R /2 that it is a normalization constant such that
∑  n − n 
4s 2
−
∞

∫
s =1
p(n , R)dR = 1. (7.51)
−∞ 
n 
≅ −n ln(2) + ln(n !) − 2 ln  !
2 
In this case, we obtain
R /2 R /2

∑ s + n ∑1
4 2
− ∞ R2

∫e
n −
s =1 s =1 C =1 2n dR = 1 2πn . (7.52)
n  −∞ 
≅ −n ln(2) + ln(n !) − 2 ln  ! (7.48)
2 
 Combining Equations 7.49 and 7.52, we obtain Equation
7.32, which shows its equivalency to Equation 7.31 in the
 R R  limit of large n.
+ 1
4  2   2  R
− +
n 2 n
2
n  R
≅ −n ln(2) + ln(n !) − 2 ln  ! − .
 2  2n

The free energy of the ideal chain can be computed

from its probability distribution function
We will now use the probability distribution for the ideal chain to compute its free
energy. In our model, the polymer has freely rotating joints with no capacity to
store energy. The polymer has zero internal energy regardless of its conformation,
and the free energy is simply Ψ = −TS. In Section 5.2, we learned that entropy can
be calculated from the density of microstates, Ω(R), which is the number of con-
figurations in which the polymer can have an end-to-end vector of R. To calculate
the density of states, we can use the probability distribution for the ideal chain
obtained in Equation 7.37. More specifically, we can take advantage of the fact
that the probability distribution function p3d(n,R) is proportional to the number
of polymer configurations that have an end-to-end vector R. In other words,

p3d (n ,R) ∼ Ω(n ,R). (7.53)

This proportionality arises due to the facts that (1) the integral of p3d(n, R) over
some range gives the exact probability for the polymer to have a end-to-end
vector within that range, and (2) this probability is equal to the number of poly-
mer configurations divided by the total number of all configurations. Using
208 CHAPTER 7: Mechanics of Cellular Polymers

Equation 7.53 we can compute the entropy as

S = kB ln Ω (n ,R) ∼ kB ln p3d (n ,R). (7.54)

Combining Equations 7.38 and 7.54, we obtain the following relation for entropy

  3  3/2 −3R 2 /2nb2 

S ∼ kB ln   2
e . (7.55)
  2πnb  

Equation 7.55 can be simplified as

3 R2
S (n , R ) = − kB 2 + S0 , (7.56)
2 nb
where we have turned the proportionality into an equivalency by accounting for
any terms that do not depend on R in the constant S0. Given an expression for
entropy, we can calculate the free energy for a chain composed of n segments with
an end-to-end vector R as

3 R2
Ψ(n ,R ) = −TS (n ,R ) = kBT 2 + Ψ0 , (7.57)
2 nb
where Ψ0 = −TS0 does not depend on R.

Force is the gradient of free energy in thermodynamic systems

At this point in our development we need to make a short aside to consider how
force and energy are related. Recall from Section 5.1 that for mechanical systems,
the principle of minimum total potential energy states that the equilibrium state
is achieved when the total potential energy is minimized. Recall that in our two-
spring system (Figure 5.2), forces were balanced when potential energy was mini-
mized, in other words when the gradient of potential energy was zero. The equivalency
between force and the gradient of potential energy becomes apparent.
In addition, as we discussed in Section 5.4 for thermodynamic systems, equilib-
rium is dictated by the principle of minimum free energy. This principle states
that equilibrium is achieved when free energy is minimized, in other words forces
are in balance when the gradient of free energy is zero. For thermodynamic sys-
tems, force is equivalent to the gradient of free energy. So, we can use Equation
7.57 to calculate the force necessary to extend the chain. In particular, force can be
found as the gradient of free energy with respect to the end-to-end vector R as

∂Ψ(n , R ) 3kBT
Fx = = Rx
∂Rx nb 2
∂Ψ(n , R ) 3kBT
Fy = = Ry (7.58)
∂R y nb 2

∂Ψ(n , R ) 3kBT
Fz = = Rz
∂Rz nb 2
or,
3kBT
F= R. (7.59)
nb 2
We previously remarked that the Kuhn length b is related to the persistence length.
In Section 7.6 we will show that this relation is

b = 2 p . (7.60)

 IDEAL CHAIN 209

So, we can rewrite Equation 7.59 in terms of the contour length L = nb and the
persistence length as

3kBT
F= R
nb 2
3kBT
= R (7.61)
(nb)b
3kBT
= R.
2 L p

Equation 7.61 gives the force–displacement relationship for an ideal chain. On
inspection, a couple of interesting things can be seen. First, for any non-zero
temperature and non-zero R, the force is non-zero as well. This is perhaps some-
what counterintuitive, because the polymer consists of freely rotating segments
that are unable to store energy. However, these forces are not as mysterious as
they may seem. Physically, the capacity for the polymer to exert a force if the ends
are separated arises out of random collisions from the thermal environment sur-
rounding it, as well as the tendency for the polymer to take on a probabilistically
favorable state. In particular, the maximum number of polymer conformations
can occur with a zero end-to-end distance, and this state is the most entropically
favorable. Straightening the polymer out reduces the entropy and therefore
requires force.
Another interesting aspect about Equation 7.61 is that the force F is linear in R, so
pulling the ends apart with a distance R produces the same force as an elastic
spring with a spring constant of (3kBT)/(2Lℓp). This relationship has been referred
to as describing an entropic spring. The resistance to deformation is derived
entirely due to entropy and depends on temperature. We can see that the stiffness
of the spring is proportional to temperature, so increasing T increases the stiff-
ness. This is in contrast to most engineering materials such as steel, which become
more compliant with increasing temperature.

The behavior of polymers tends toward that of an ideal chain

in the limit of long contour length
In developing the ideal chain model, we assumed that the polymer was made up
of freely rotating links with no capacity to store energy. In this model, the poly-
mer is assumed to have zero internal energy regardless of the configuration. Of
course, in reality, there will be energetic interactions within any polymer. These
interactions can arise, through changes in bond angles and distances between
backbone atoms, or the electrostatic interaction of backbone atoms with each
other from a distance. One may ask whether the ideal chain, which has zero
energy regardless of the conformation, is a good representation of any polymers
in reality?
The simple answer is yes, as long as the polymer is long enough. To gain some
intuition for this notion, recall our discussion of the thermally fluctuating micro-
tubule in Section 7.3. We found that the tendency for the MT to exhibit flexible,
semi-flexible, or stiff behavior depended not only on its persistence length, but
also its contour length. A typical MT found in a cell has a contour length much less
than the persistence length and tends to be found in rod-like configurations.
However, if we were to construct a very long MT that had a contour length much
greater than its persistence length, we would find that it would exhibit greatly dif-
ferent behavior. In particular, the behavior of the very long MT would resemble
that of  a flexible polymer, whose tendency to coil is an entropically driven
phenomenon.
210 CHAPTER 7: Mechanics of Cellular Polymers

The tendency for polymers with large contour lengths to tend toward ideal chain
behavior can be better understood by examining the physical meaning of the
persistence length. We know that at length scales longer than the persistence
length, the orientation correlation between two points of a thermally fluctuating
polymer vanishes. Physically, this means that at this length scale, entropic influ-
ences become dominant over energetic influences. We also know that if a poly-
mer is longer than its persistence length, then for every span of polymer
approximately equal to this length, the orientations of the polymer become
roughly uncorrelated. At very long length scales, the polymer behaves as if it were
composed of many independently fluctuating chain segments, with each seg-
ment being on the order of the persistence length in size. The behavior of most
polymers will be dominated by entropy and tend toward that of an ideal chain
when  the contour length is significantly greater than the persistence length.
In modeling such cases using an ideal chain, a Kuhn segment is not necessarily a
single monomer, but rather a span of polymer roughly equivalent to the persis-
tence length in size.

7.5 FREELY JOINTED CHAIN (FJC)

The FJC model places a limit on polymer extension
Although the ideal chain is an extremely useful model, it has certain limita-
tions. In Section 7.4 we found a relationship for the response of an ideal chain
to tension. The stiffness did not depend on how much it was extended.
Therefore, the model predicts that the stiffness will be constant even as the
chain is extended past its contour length nb! This is clearly unphysical. Although
the ideal chain is useful if the end-to-end distance is much less than the con-
tour length, an alternative model is desirable for cases in which this condition
is not met.
A model that addresses this limitation in the ideal chain is the FJC, which gives
more realistic behavior at long extensions. Similar to the ideal chain, the FJC
consists of n links of length b that are connected to each other by freely rotating
joints. However, in the FJC model, the end-to-end length is constrained such
that it cannot be longer than the contour length. We derive the force extension
behavior here.
Let us revisit the chain in Figure 7.15. For the ith segment, we define the vector ri
as the product of b and the ith unit vector ui, ri = bui. The end-to-end vector is
given by
n

R=b ∑u . i (7.62)
i =1

To simplify the development (and without losing any generality) we will consider
a chain that is being extended in the z-direction. We can simplify the math by
expressing R in spherical coordinates about the z-axis with θi being the angle with
the z-axis and ϕi being the azimuth (Figure 7.16) as

e x ⋅ ui = sinθi cosφ i

e y ⋅ ui = sinθi sin φ i (7.63)

e z ⋅ ui = cosθi ,

where ex, ey, and ez are unit vectors in the x-, y-, and z-directions.
 FREELY JOINTED CHAIN (FJC) 211

z Figure 7.16 Definition of spherical

coordinates for a segment of
length b.
z = b·cos f
b

y = b·sin f sin θ
x = b·sin f cos θ y
b·
sin
θ f

Using these relations, the end-to-end distance along the z-axis can be written as
Rz = e z ⋅ R
n

=b ∑e ⋅ u
i =1
z i
(7.64)
n

=b ∑ cosθ .
i =1
i

The force–displacement relation for the FJC can be

found by the canonical ensemble
With the configurational geometry of the chain defined, we can now calculate the
force it generates with extension. For the ideal chain this calculation was based on
determining the change in free energy associated with a reduction in entropy with
extension. For the FJC we take a different approach. In Section 5.5, we learned that RZ
for a constant temperature system at equilibrium, the probability for a given micro-
state is given by Boltzmann’s distribution. We seek to define a system with a known
potential energy such that we can use Boltzmann’s distribution to find the proba-
bility of a particular polymer configuration at equilibrium.
Consider a system consisting of a FJC immersed in a constant temperature heat
bath. One end of the chain is constrained such that its position is fixed but freely
rotating. At the other end, we attach a small weight that pulls downward with a
constant force of Fz (Figure 7.17). As the chain is extended, the potential energy of
the weight (and therefore the entire system) is decreased by FzRz. We can write the FZ
internal energy as
n
weight
Q = − Fz Rz = − Fz b ∑
i =1
cosθi , (7.65)
Figure 7.17 Schematic
demonstrating the FJC with a
weight attached to its end. The
where in the right-hand-most side we have substituted the expression for Rz found weight subjects the polymer to a
in Equation 7.64. We know that for the canonical ensemble, the probability of a constant downward force Fz. As the
microstate having internal energy Q can be found by the Boltzmann distribution. weight moves downward, the potential
For our system, each microstate can be characterized as the set of bond angles energy of the weight decreases.
212 CHAPTER 7: Mechanics of Cellular Polymers

θ1,2,. . . ,n and ϕ1,2,. . .,n. The Boltzmann probability associated with the microstate
having  energy Q(θ1,2,. . . ,n,ϕ1,2,. . . ,n) is
n

1 κ ∑ cosθi (7.66)
p(θ1 ,θ 2 ,....θ n ,φ 1,φ 2 ,... ,φ n) = e i =1 ,
Z
where
Fz b 2 Fz  p
κB = = . (7.67)
kBT kBT

To calculate the partition function Z, we must integrate the exponential term in

Equation 7.66 over all possible configurations of the chain. To do this, we integrate
over all possible bond angles θ1,2,. . .,n and ϕ1,2,. . .,n as
Z=
n
2π 2π 2π π π π
κ ∑ cosθi
∫ ∫ ∫ ∫ ∫ ∫
φ1 = 0 φ 2 = 0 ... φ n = 0 θ1 = 0 θ 2 = 0 ... θn = 0
e i =1
sinθ1 sinθ 2 ...sinθ n dθ1dθ 2 ... dθ n dφ 1dφ 2 ... dφ n ,

(7.68)
where we have performed the integral with respect to a differential element of
solid angle sinθ dθ dϕ because the bond angles were defined in spherical coordi-
nates. We can rearrange the order of integration in Equation 7.68 to perform the
integrals for each segment in turn as
2π π 2π π

Z=
∫ ∫
φ 1= 0 θ1 = 0
eκ cosθ 1 sinθ1dθ1dφ 1
∫ ∫e
φ 2=0 θ2 =0
κ cosθ 2
sinθ 2dθ 2dφ 2 …

2π π

∫ ∫e
φ n = 0 θn = 0
κ cosθn
sinθ ndθ ndφ n .
(7.69)

Equation 7.69 can be written in a simplified, compact form as

n 2π π

Z= ∏∫ ∫e
i =1 φ i = 0 θi = 0
κ cosθi
sinθi dθi dφ i = z n , (7.70)

where
2π π

z=
∫ ∫e
φ =0 θ =0
κ cosθ
sinθdθdφ (7.71)

is the single partition function. To compute z, we first integrate out θ as
π

z = 2π
∫e κ cosθ
sinθdθ . (7.72)
θ =0
Next, we change variables by letting ρ = cos θ and dρ = −sin θ dθ. Equation 7.72
can be evaluated as
1
eκ − e −κ sinhκ
z = −2π eκρ dρ = 2π
∫ κ
= 4π
κ
. (7.73)
−1
where sinh κ is the hyperbolic sine, sinh κ = (1/2)(eκ − e−κ).
 FREELY JOINTED CHAIN (FJC) 213

We now seek to use the partition function to calculate 〈Rz〉 at equilibrium. To do

this, we use a similar method as in Section 5.5, in which we computed the average
internal energy using the partition function in the canonical ensemble. In particu-
lar, we first start with the relation for 〈Rz〉,

∫
Rz = p(Θ)Rz dΘ , (7.74)
Θ
where for simplicity, we have used Θ in Equation 7.74 to denote integration over
all possible bond angles θ1,2,. . .,n and ϕ1,2,. . .,n. Substituting our expression for p
obtained in Equation 7.68 into Equation 7.74 and using the relation
 n 
n
 ∑
κ  b cosθi 

κ ∑ cosθi i =1 κRz (7.75)
e i =1
=e b =e b ,

we obtain
κRz
1
Rz =
Z
e
∫ b RzdΘ. (7.76)
Θ
Similar to our approach in calculating internal energy in the canonical ensemble,
we can rewrite Equation 7.76 as a logarithm of Z with the following manipulation

∂  bz 
κR
b
Rz =
Z ∫
Θ

∂κ 
e  RzdΘ

κRz
b ∂
=
Z ∂κ
e
∫
Θ
b R z dΘ
(7.77)
b ∂Z
=
Z ∂κ
∂ ln Z
=b .
∂κ
Using Equation 7.77, we can compute the mean extension as

∂ ln Z
Rz = b
∂κ
∂
= bn [ln(sinh κ ) − lnκ + ln 4π] (7.78)
∂κ
 1
= bn  cothκ −  .
 κ

Recall that force is embedded inside κ. Equation 7.78 gives a force-displacement
relationship for the FJC. The expression is a little different than those with which
we are familiar, with because it is an implicit relationship. In other words, although
it does produce a force for every displacement and vice versa, we do not have an
explicit expression for force. However, it is still an entirely valid relationship.

Differences between the ideal chain and the FJC

emerge at large forces
We can get a sense of the force–displacement behavior predicted by the FJC
model by examining some special cases. To do this, we first note that the term in
214 CHAPTER 7: Mechanics of Cellular Polymers

Figure 7.18 Plot of the Langevin 0.9

function. The Langevin function
approaches 1 for large values of x. For 0.8
small values of x, the slope is 0.7
approximately one-third.
0.6

coth(x) − 1/x
0.5
0.4
0.3
0.2
0.1
0.0
0 1 2 3 4 5
x

parentheses in Equation 7.78 is known as the Langevin function. Expanding the

hyperbolic cotangent, the Langevin function ℒ can be expressed as

e x + e−x 1
= − . (7.79)
e x − e−x x

A plot of the Langevin function can be seen in Figure 7.18. Notice that as x gets
very large, ℒ approaches one. This means that even as the force grows indefinitely,
the end-to-end length of the polymer cannot exceed the contour length, nb. Thus,
the FJC overcomes one of the most critical limitations of the ideal chain. When the
value of x is small, the slope of ℒ(X) approaches 1/3. In this limiting case we can
approximate ℒ(x) to be x/3 and

κ F nb 2
Rz ≈ nb = z , (7.80)
3 3kBT

which is that obtained for the ideal chain in Equation 7.61. When the extension
(and force) are very small, the FJC model predicts the same force–displacement
behavior as the ideal chain.

7.6 worm-like chain (WLC)

The WLC incorporates energetic effects of bending
In both the ideal chain and the FJC, we modeled a polymer as rigid segments
connected by freely rotating hinges. With this entirely entropic approach, we
ignored energetic costs from changes in orientations between segments. In this
section, we present the WLC, which incorporates both energetic and entropic
effects associated with bending. In the WLC, polymers are modeled as continu-
ous space curves rather than discrete segments.
A schematic demonstrating the relevant quantities for the WLC can be seen in
Figure 7.19. The configuration of the space curve is given by the vector-valued
function a(s). Similar to our use of s in defining the persistence length, here it
runs along the polymer from 0 to the contour length L and is a parameteriza-
tion by which each position on the polymer can be identified. The vector a(s)
extends from the origin of the coordinate system and ends at some point along
the polymer given by s. For simplicity, the WLC is constrained to be inextensi-
ble (that is, the contour length cannot change), which is enforced by setting
 worm-like chain (WLC) 215

Figure 7.19 Parameterization of the

space curve in the WLC. At each
position s along the curve, a(s) is the
position vector and its partial derivative
s=L is the tangent vector.
s=0

∂a(s)
—
∂s

a(s)

the tangent vector (the first derivative of a(s)) to have a unit magnitude for all
values of s
2
 ∂a(s ) 
  = 1. (7.81)
∂s 
Alternatively, the tangent vector ∂a(s)/∂s is a unit vector.
We now specify the constitutive relations for the model. In the WLC, the chain is
assumed to resist bending deformation through increases in potential energy
with curvature similar to an elastic beam. In Section 3.2, we found the bending
energy for a beam is given by
L 2
EI  ∂ 2a(s ) 
Q(a(s )) =
∫
2  ∂s 2 
ds , (7.82)
0
where the term in parentheses is a vector that describes the local curvature. We
can express Equation 7.82 in terms of persistence length. In particular, in Section
7.3 we showed that for a curvy elastic beam subject to thermal fluctuations, its
persistence length is related to its flexural rigidity as ℓp = EI/kBT. We can recast
Equation 7.82 in terms of persistence length as
L 2
k T   ∂ 2a(s ) 
Q(a(s )) = B p 
2 ∫
 ∂s 2 
ds. (7.83)
0

Example 7.2 DNA looping and lac repressor

Within the bacteria Escherichia coli, the genes lacZ, Experiments show that modifying the operator distance
lacY, and lacA encode for enzymes that are involved in by inserting different-sized fragments of DNA between
metabolizing lactose. Near the lac genes are two bind- the lac repressor binding sites alters repression of the lac
ing sites for a protein called “lac repressor.” The binding enzymes, with the repression of the lac enzymes peaking
sites are separated by a fragment of DNA with a length when the operator distance is about 70 base pairs long,
known as the operator distance. The binding sites must and decreases when the operator distance becomes
be brought together by forming a DNA loop for lac larger or smaller than this value. Calculate the probabil-
repressor to bind. Once bound, lac repressor blocks the ity of looping for very short and very long operator dis-
path of RNA polymerase, preventing the transcription of tances. Speculate why this behavior might occur.
lacZ, lacY, and lacA.
216 CHAPTER 7: Mechanics of Cellular Polymers

We first calculate the probability of looping for very short chain of independently fluctuating segments (each seg-
operator distances. We know that the smaller the contour ment having length on the order of ℓp), and the free
length relative to its persistence length, the more we must energy becomes dominated by the entropy. Taking this
account for energetic influences. We use the WLC model. into account, for very long operator distances, we model
Assuming an operator distance of L loops into a circle of the polymer as an ideal chain. The probability distribu-
radius R, the energy of looping for a WLC is tion for an ideal chain is given by Equation 7.38. A loop
occurs when R = 0, and so
L 2
kBT  p  1  kBT  p L 2π 2kBT  p
Qloop =
2 ∫   ds =
R 2 R 2
=
L
,  (7.84)
 3 
ploop = 
3/2
 3 
=
3/2
, (7.85)
 2πnb 2   2πLb 
0

where in the right-most relation we made use of the fact
that R = L/2π. We know from the Boltzmann distribution where we made use of the fact in the right-most relation
that the higher the value for Qloop, the lower the proba- that L = nb. In this case, ploop → 0 as L → ∞.
bility of looping, ploop. Thus, ploop → 0 as L → 0.
Our models predict that the probability of looping goes
Now let us compute the probability for looping for very to zero for very short and long operator distances. In
long operator distances. As the contour length becomes both cases, repression would be expected to decrease
longer and longer, the polymer begins to resemble a (causing lac mRNA levels to increase).

The force–displacement relation for the WLC can be

found by the canonical ensemble
We now seek to determine a force–displacement relation for the WLC. We know
that for a polymer immersed in a constant temperature heat bath, the probability
of a microstate with a given internal energy is described by the Boltzmann distri-
bution. To calculate the force generated by extension, we construct a system with
a known internal energy such that we can use Boltzmann’s distribution to find the
probability of a particular polymer configuration at equilibrium.
Similar to our approach for the FJC, we consider a system consisting of a WLC
immersed within a constant temperature heat bath, with one end of the chain
fixed in space by a freely rotating hinge, and the other end attached to a small
weight that produces a downward force of Fz. As the chain is extended, the loss in
potential energy due to the decrease in height of the weight is FzRz. For the FJC, we
were able to calculate Rz as a sum over chain segments. For the WLC, which is
continuous curve in space, we can compute this quantity as an integral
L
∂a(s )
Rz (a(s )) =
∫ ∂s
⋅ e zds , (7.86)
0
where ∂a(s)/∂s is the tangent vector, and the dot product with ez is used to obtain
the z component of the tangent vector. The total internal energy of the system for
a given chain configuration of a(s) can be computed as

L 2 L
k T   ∂ 2a(s )  ∂a(s )
Qtot (a(s )) = Q(a(s )) − Fz Rz (a(s )) = B p 
2 ∫2 
 ∂s 
ds − Fz
∂s ∫
⋅ e z ds.
  0 0
(7.87)
Now that we have the internal energy of the system for each configuration a(s), we
can compute the probability of each curve using the Boltzmann distribution as

1 −Qtot (a( s ))/kBT

p(a(s )) = e , (7.88)
Z
 worm-like chain (WLC) 217

where

Z=
∫e −Qtot ( a ( s ))/kBT
da. (7.89)
∀a
In principle, we can compute the average extension under the applied force as

Rz =
∫ p(a(s))R da z (7.90)
∀a

and relate the average extension to the partition function as

∂ ln Z
Rz = kBT . (7.91)
∂Fz

However, determining analytical solutions to these integrals is a lot easier said

than done. They both require the integration over all possible polymer configura-
tions, with each configuration being a curve in space. In fact, an analytical result
for the force–extension relation of the WLC does not exist, except in special limit-
ing cases. However, it has been investigated using computational models, with the
results described by the following interpolation equation

kBT  1  Rz 
−2
Rz  1 (7.92)
Fz =  1 − − + .
p  4  L  4 L 

Differences in the WLC and FJC emerge when they are fitted
to experimental data for DNA
The differences in the force–extension behavior of the FJC and the WLC can be
demonstrated by fitting the two models to force–extension curves for DNA, such
as in Figure 7.20. The measurements were made by attaching one end of a DNA
molecule to a glass surface and the other end to a magnetic bead. The bead was
pulled with a known force Fz, and the extension Rz was measured optically. The
data were fitted to the WLC and FJC using the relation b = 2ℓp = 106 nm. Upon
inspecting the model fits, there are two noticeable trends. First, at large forces, the
extension asymptotically approaches the contour length. Second, the experimental

(A) (B) Figure 7.20 Comparison of the FJC

and WLC. The force–extension behavior
of DNA can be obtained by applying a
force to a magnetic bead tethered by
30 DNA, and measuring the displacement
magnetic of the bead optically (A). Data obtained
from such an experiment can be seen in
bead
Extension (µm)

the adjacent plot (B). The experimental

20 data (squares) were fitted to a WLC (solid
line) and a FJC (dashed line) assuming a
persistence length of 106 nm and a
DNA contour length of 33 μm. At large forces,
the extension asymptotes at a value
RZ 10 equal to the contour length. The
experimental data are better fit by the
WLC at higher forces. (B, from,
Bustamante C, Marko JF, Siggia ED &
0 Smith S (1994) Science. Reprinted with
101 10 2 10 3 10 4 10 5 permission from AAAS.)
Force (IN)
218 CHAPTER 7: Mechanics of Cellular Polymers

data is fit better by the WLC, especially at forces greater than approximately 0.1
pN. The WLC predicts that more force is required to extend the polymer to a given
length than the FJC. The underlying reason for this can be understood by consid-
ering the thermal fluctuations in a polymer as waves of different wavelengths
superimposed on top of each other. In the FJC, wave-like fluctuations are con-
strained to be of wavelength b or higher, because they cannot be shorter than the
length of each link. This constraint is not present in the WLC. Additional force is
required to smooth out these short wavelength fluctuations present in the WLC,
but not the FJC.

Persistence length is related to Kuhn length

In this section, we quantify the 〈R2〉 behavior of the WLC under thermal fluctua-
tions. We shall see that by comparing this behavior to that of discrete chain mod-
els, we will be able to relate persistence length to Kuhn length, in other words the
origins of Equation 7.60.
Let us start by defining the polymer end-to-end vector R for the WLC. We do this
by integrating the tangent vector ∂a(s)/∂s over its length as

L
∂a(s )
R=
∫ ∂s
ds. (7.93)
0

Using Equation 7.93, we can now compute 〈R2〉 as

L L
∂a(s ) ∂a(s ′)
R2 = R ⋅ R =
∫
0
∂s
ds ⋅
∂s ′ ∫
ds ′
0

L L
∂a(s ) ∂a(s ′)
=
∫∫
0 0
∂s
ds ⋅
∂s ′
ds ′ (7.94)

L L
∂a(s ) ∂a(s ′)
=
∫∫ ∂s
⋅
∂s ′
dsds ′.
0 0

Recall that in our parameterization of the WLC, the tangent vector ∂a(s)/∂s was
constrained to be a unit vector. Therefore the magnitude of the product in the
brackets is one and Equation 7.94 can be written as

L L

R 2
=
∫ ∫ cos ∆θ s −s′ dsds ′ , (7.95)
0 0
where Δθs−s′ denotes the angle between the tangent vectors at s and s′. In
Section 7.3 we found that for a thermally fluctuating polymer, the average dif-
ference in angle between tangent vectors is given by the orientation correlation
function, which decays exponentially such that 〈cos(Δθ(s))〉 = e−s/ℓp. We assume
that on average, the difference in angle between any two points on the chain
will decay similarly as the distance between the two point is decreased, or in
other words,
L L

∫∫e (
− s − s ′ )/ p
R2 = dsds ′. (7.96)
0 0
 Key Concepts 219

It is left as an exercise to show that the solution to the double integral in Equation
7.96 is

 −L L
R 2 = 2 2p  e  p − 1 +  . (7.97)
 p 

Consider the case where the contour length is much greater than the persis-
tence length, L  ℓp in Equation 7.98. The exponential term in the brackets goes
to zero, and assuming the L/ℓp term is much greater than one, then

L
R 2 ≈ 2 p 2 = 2 L p . (7.98)
p

If we compare this quantity to the mean squared end-to-end length of nb2 for the
FJC, we obtain a relation between Kuhn length and persistence length,

2 L p = nb 2 = (nb)b = Lb. (7.99)

Equation 7.100 can be simplified to

b = 2 p , (7.100)

which shows the origins of Equation 7.60. Equation 7.100 implies that a discrete
chain with Kuhn length b has a persistence length of b/2 in the limit of b  L.

Key Concepts

• The cellular cytoskeleton primarily consists of three dominated by entropy. The probability distribution
biopolymers: MFs, MTs, and intermediate filaments. function for the ideal chain is based on the Gaussian
The subunits for MFs and MTs are actin monomers approximation to the random walk.
and tubulin dimers, respectively. Intermediate • For thermodynamic systems, force is equivalent to
filaments can be composed of different proteins the gradient of free energy. Separating the ends of an
depending on their location within the body. ideal chain reduces the entropy, and thus requires
• There is a highly dynamic balance between the G force.
and F forms of actin. Owing to G-actin’s polarity and • The mechanical behavior of polymers will tend
the ability for bound ATP to undergo hydrolysis, toward that of an ideal chain when the contour
polymerization kinetics can be very different at the length is significantly greater than the persistence
(+) and (−) ends. Treadmilling occurs when D-form length.
subunits are lost from the shrinking (−) end, become
• The FJC addresses a key limitation of the ideal chain,
T-form subunits, and are recycled to the growing
which is unphysical behavior for large forces. The FJC
(+) end.
is similar to the ideal chain at low extensions, but
• The persistence length gives a characteristic length gives more realistic behavior at long extensions.
scale over which the orientations of a thermally
• The WLC does not have joints but rather treats the
undulating polymer become mostly uncorrelated. A
polymer like a flexible beam with elastic energy.
polymer can be classified as being flexible, semi-
Compared with the FJC, the WLC predicts that more
flexible, or stiff depending on the relation of its
force is required to extend the polymer a given
persistence length to its contour length
length.
• In the ideal chain, all conformational changes in
internal energy are ignored. It is often used for
modeling flexible polymers whose behavior is
220 CHAPTER 7: Mechanics of Cellular Polymers

Problems

1. Using the one-dimensional Gaussian approximation to 8. What is the slope of the Langevin function, L, near
the random walk and step size of b = 1, show that the zero?
root mean square displacement 〈R2〉1/2, is the square
root of the number of steps taken. You may use integral 9. What is the force generated in polymers of spectrin,
identities. actin, and tubulin (ℓp = 15, 15 × 103, and 2 × 106 nm,
respectively) with a 100-nm contour length predicted
2. Compare the root mean square end-to-end distance by the WLC model when the polymer is extended to
〈R2〉1/2 in three dimensions of strands of spectrin, actin, 50 nm?
and MTs (ℓp = 15, 15 × 103, and 2 × 106 nm, respectively)
with 200-nm contour length ( L). 10. What is the effective (tangent) spring constant as a
function of average displacement for a flexible polymer
 −s 
  using the WLC model? What is the approximation in
2 p 
3. Show that in three dimensions cos θ ( s) = e . the limit as displacement approaches zero? And when
displacement approaches the contour length? Hint: the
4. What is the force generated in polymers of spectrin, tangent stiffness is the slope of the force–displacement
actin, and tubulin (ℓp = 15, 15 × 103, and 2 × 106 nm, curve.
respectively) with a 100-nm contour length predicted
by the ideal chain model when the polymer is extended 11. What is the effective (secant) spring constant as a
to 50 nm? And to 150 nm? Why is this not realistic? function of average displacement for a flexible polymer
Assume a temperature of 300 K in your calculations. using the WLC model? What is the approximation in
the limit as displacement approaches zero? And when
5. Assuming a persistence length for DNA of 50 nm, displacement approaches the contour length? Hint: the
determine the force at which the ideal chain and the secant stiffness is the slope of a line from the origin to
FJC differ by 10% in the extension length, assuming they the point on the force–displacement curve.
are used to model the same strand of DNA with a 1-μm
contour length. Which model predicts a longer polymer 12. Assume a persistence length, ℓp = 15, 15 × 103, and
at this force? Assume a temperature of 300 K. 2 × 106 nm for spectrin, actin, and MTs, respectively. For
2π π filaments 1-cm long at T = 300 K, what is the effective
2k BTS
∫ ∫ θ p(U)dφ sinθ dθ =
6. Show that θ 2 = 2 spring stiffness (tangent) at zero displacement and
EI when fully extended?
0 0

assuming that the probability follows a Boltzmann 13. Consider a 30-µm length of DNA such as might be found
distribution and the energy is the strain energy of a in a virus. What force is required to stretch the DNA
beam bent to an angle θ. Hint: the integral you obtain is to an end-to-end displacement x = 10, 20, and 25 µm
quite challenging. Rather than attacking it directly, try at 300 K using first a FJC model and secondly a WLC
the mathematical trick we used for the FJC and WLC. model? Assume ℓp = 50 nm. Comment on the results of
Namely, show that 〈θ2〉 can be expressed in terms of the the two models.
derivative of ln(Z) with respect to E.
14. Conduct a numerical comparison of the FJC and WLC
7. What is the force generated in polymers of spectrin, models. Plot the force–extension behavior for a 1-cm
actin, and tubulin (ℓp = 15, 15 × 103, and 2 × 106 nm, long polymer with a 0.1-mm persistence length with
respectively) with a 100-nm contour length predicted by each model. Is one always higher or lower than the
the FJC model when the polymer is extended to 50 nm? other? Does this make sense? Comment on how the
And to 150 nm? How is this more realistic than the Ideal two compare in the short and long limits. Repeat with
chain model? Hint: the relationship between force and a 1-cm persistence length and a 10-cm persistence
displacement for the FJC model is implicit (you cannot length. How does persistence length affect how the
plug in a value of 〈R〉 and calculate F.) Instead, solve models compare?
it numerically to two significant figures by guessing L L

∫∫e (
2 − s − s ′ ) / p
values of F and interpolating (or a more sophisticated 15. Show that the solution of R = dsds′ is
approach if you like). You can use MATLAB or another 0 0
programming approach. Alternatively, plot the force–  −L 
2l p2  e lp − 1 + L  .
displacement relationship and estimate the point  lp 
from the graph. Comment on your results.
 Annotated References 221

Annotated References

Boal D (2001) Mechanics of the Cell. Cambridge University Press. An ­operator. J. Molec. Biol. 267, 21–29. This article shows dependence
excellent text on cell mechanics with many in-depth treatments of of lac promoter repression on operator distances.
the polymer models covered in this chapter. Parts of the relation of
Omary MB, Coulombe PA & McLean WH (2004) Intermediate fila-
persistence length to Kuhn length were based on developments in
ments and their associated diseases. N. Engl. J. Med. 351, 2087–
this text.
2100. A nice review of intermediate filament biology. Table 1 in this
Bustamante C, Marko JF, Siggia ED & Smith S (1994) Entropic elastic- chapter was obtained from this study.
ity of l-phage DNA. Science 265, 1599– 1600. This paper gives the
Phillips R, Kondev J & Theriot J (2009) Physical Biology of the Cell.
force extension behavior for the WLC and shows the differences in
Garland Science. This textbook gives a more in-depth treatment of
the WLC and freely jointed chain when fit to force–extension behav-
the lac repressor example as well as many interesting applications
ior for DNA.
of the polymer models developed in this chapter.
Howard J (2001) Mechanics of Motor Proteins and the Cytoskel- Rubenstein M & Colby RH (2003) Polymer Physics. Oxford University
eton. Sinauer Associates. A great introduction to cytoskeletal Press. An excellent introduction to polymer physics. Derivation for
polymerization and mechanics. Students interested in learn- the Gaussian approximation to the random walk is based on the
ing more about polymerization kinetics and force generation by treatment developed in this text.
polymerization are referred to this text. The derivation of persis-
tence length, the relation of persistence length to flexural rigid- Spakowtiz AJ (2008) Polymer Physics (lecture notes, from course
ity, as well as some parts of the relation of persistence length to number ChemEng 466, Stanford University, Stanford, CA). Several
Kuhn length were based on the treatments of these topics in of the treatments in this chapter were based on excellent course
this text. notes developed by Dr. Spakowitz for a graduate course on polymer
physics at Stanford University. These treatments include develop-
Muller J, Oehler S & Muller-Hill B. (1996) Repression of lac promoter ments for the ideal chain, the freely jointed chain, and the WLC, as
as a function of distance, phase, and quality of an auxillary lac well as the lac repressor example.
This page intentionally left blank
to match pagination of print book
CHAPTER 8
Polymer Networks and
the Cytoskeleton

I n Chapter 7, we explored the mechanics of individual biopolymers. We now

turn our attention to the mechanics of biopolymer networks, which are depend-
ent not only on the mechanical behavior of the individual biopolymers within, but
also on how these biopolymers are organized into a specific microstructure or
architecture. For example, the microstructure of biopolymer networks such as the
cytoskeleton can vary dramatically from cell to cell in terms of how the biopoly-
mers are oriented with respect to one another, the number of filaments per unit
volume, and how they are cross-linked. This microstructure can also vary sub-
stantially between different locations within individual cells. Because of the radi-
cal effect that these microstructural variations can have on network mechanics,
the relationships between biopolymer network mechanical behavior and micro-
structural properties has become an extremely active area of research. Indeed,
several investigational approaches, including analytical, experimental, and com-
putational approaches have been used to better understand these relationships.
In this chapter, we will discuss the use of scaling approaches to relate microstruc-
ture to mechanical properties, constitutive models of a class of networks called
affine networks, and the mechanics of specific cytoskeletal architectures found in
vivo as they are related to their mechanobiological function.

8.1 POLYMER NETWORKS

Polymer networks have many degrees of freedom
We learned about approaches for modeling the mechanics of individual polymers
in Chapter 7. In principle, these models could be used for mechanical modeling of
polymer networks. To model a network of flexible polymers, one could examine
each of the polymers within the network, determine its geometry, and explicitly
account for each polymer in the network model. Each polymer could be modeled as
an ideal chain with some entropic spring constant and initial end-to-end distance.
In reality, however, this “discrete” approach is rarely feasible. The reason is that pol-
ymer networks (such as within a cell) generally have an enormous number of fila-
ments and an even larger number of degrees of freedom. A typical endothelial cell
contains approximately 10 mg/ml of F-actin. Assuming the average filament con-
tains 100 subunits (and assuming an average weight of ~1 × 10−14 mg per filament),
and the cell volume is on the order of 1 × 10−8 ml (for a polygonal cell 5 μm high,
50 μm long, and 40 μm wide), this would mean that a cell contains approximately
10 million actin filaments. If we use a simplified mechanical model in which the
state of each filament could adequately be described by its center of mass position
(x, y, z), orientation (θ, α), and end-to-end length (L), explicitly modeling each poly-
mer would involve 60 million degrees of freedom.

Effective continuums can be used to model polymer networks

To model polymer networks with large numbers of degrees of freedom, several
approaches can be taken. One could try to use high-performance computing to
224 CHAPTER 8: Polymer Networks and the Cytoskeleton

explicitly account for all the individual polymers in a given network. However, in
most cases, the numbers of degrees of freedom in physiological networks are too
large even for the world’s most powerful supercomputers.
An alternative approach is to represent the network as an effective continuum to
reduce the number of degrees of freedom. What we mean by an effective contin-
uum is that we equate the mechanical contributions of all the discrete polymers
within the network to the mechanical behavior of some equivalent continuum.
Then, the mechanical state of the network can be described through the analytical
framework of continuum mechanics.
Consider an imaginary cube-shaped polymer network. The length of each side of
the network is L. The network is filled with randomly aligned polymers that are
cross-linked to one another (Figure 8.1). We apply a very small uniaxial deforma-
tion to the network in the x-direction by fixing one side of the cube so that it does
not translate in the x-direction, and subjecting the opposite side of the cube to a
uniaxial normal stress σ. The lengths in the y- and z-directions are assumed to be
unconstrained, such that the network can still expand in the y- and z-directions.
Upon deformation, the new length in the x-direction is L1.
Next, we replace the polymer network with an isotropic continuum with Young’s
modulus E = Lσ/(L − L1). The deformation of the entire network under stress σ
can be described by a single homogeneous strain of (L − L1)/L. This simple exam-
ple demonstrates that by assigning appropriate constitutive behavior to the con-
tinuum, we are able to capture the mechanical behavior of the network without
explicitly modeling each individual filament. In the following sections, our overall
focus will be to form effective continuum models for the mechanical behavior of

Figure 8.1 Effective continua can be (A)

used to model the mechanical fixed in
behavior of polymer networks.
undeformed x-direction deformed
(A) An imaginary cube-shaped polymer
network with undeformed side length
L is subjected to a small uniaxial
deformation by fixing one side and σ
subjecting the opposite side to a uniaxial
stress σ, resulting in a new length in the L
x-direction of L1. The lengths in the
y- and z-directions are unconstrained. z
(B) “Replacement” of the network with a L x
fictitious isotropic homogeneous linear y x
L
elastic continuum with Young’s modulus L
E = Lσ/(L − L1). The deformation to the
entire network is described by a single L1
homogeneous strain, ε = (L − L1)/L = E/σ.
(B)

E
σ
L

x L x
L L1 – L
L ε=—
L
L1
 SCALING APPROACHES 225

polymer networks. In doing so, we will seek to identify relationships between the
effective mechanical behavior of networks and their microscopic properties, such
as polymer stiffness, length, density, orientations, and other properties.

8.2 SCALING APPROACHES

We begin our discussion of effective continuum behavior for polymer networks by
deriving scaling relationships between effective network stiffness and microstruc-
tural properties. Scaling approaches can be highly useful for making predictions
of how the effective mechanical properties of a network may change with, for
example, the number of filaments within the network per unit volume. However,
these simplified approaches generally do not yield an explicit constitutive model.
Although this may seem somewhat unsatisfying given the analytical tools we have
developed thus far, such scaling relationships are easier to formulate than consti-
tutive models, and may be useful for a variety of situations, such as interpreting
experimental results during investigations, involving bead twisting/pulling, and
in atomic force microscopy experiments.

Cellular solids theory implies scaling relationships between

effective mechanical properties and network volume fraction
We begin our discussion of scaling approaches with the theory of cellular solids. In
this theory (also referred to as “open foams”), one seeks to develop a model of
network structure, and then use scaling arguments to find the dependence of
effective elastic moduli on network volume fraction, which is the fraction of the
volume of the network taken up by polymer material. For a network with a volume
fraction of 0.10, 10% of its volume contains polymer, and the other 90% of its vol-
ume is empty space. In the cellular solids approach, we first construct a “unit cell”
that is a representative structural unit of our network. This unit cell is assumed to
possess geometric scaling such that results obtained from the analysis of the sub-
unit are valid when scaled up to the size of an entire cell. We then find the amount
of strain the unit cell would experience under an applied stress, and use this
quantity to find the effective elastic modulus. In carrying out these calculations,
we use scaling arguments, as we are only interested in simple relationships (if the
volume fraction of the network doubles, what happens to the apparent stiffness of
the network?). We will ignore all but the most basic parameters and drop con-
stants to not vary with the dependent variables. In our analysis, we will perform
calculations for two different cases: when the network deformation occurs exclu-
sively because of polymer bending, and when the deformation occurs exclusively
because of axial deformation of the polymers.

Bending-dominated deformation results in a nonlinear scaling

of the elastic modulus with volume fraction
We start with a unit cell that is a representative subunit of our network and con-
structed such that compression of the network results in bending (Figure 8.2). The
length of the beam members making up the sides of the cube in the center of the
unit cell is assumed to be L, and the beams themselves are assumed to have a char-
acteristic radius R. In this configuration, adjoining cells are coupled such that
transmission of loads occurs at member mid-spans, resulting in a bending moment.
We begin our analysis by recalling Equation 3.31, which describes the deflection
of a cantilevered beam,

F  x3 x2 
w=  − L .
EI  6 2
226 CHAPTER 8: Polymer Networks and the Cytoskeleton

Figure 8.2 Unit cell undergoing F

bending deformations. (A) In the
absence of force, the beam members of F
(A) (B)
the unit cell have undeformed length L
and radius R. (B) In the presence of
force, bending of cross members occurs
owing to transmission of loads at
member mid-spans, resulting in a δ
2R
bending moment.
L

F
F

The displacement at the tip (x = L) is,

− FL3
w= .
3EI

This relation can be used as the basis of the scaling relationship for the mid-span
displacement in our unit cell, δ, because

FL3 FL3
δ~ 4
~ , (8.1)
EbR EbR 4

where F is some load applied to the subunit, Eb is the elastic modulus of the beam,
and the moment of inertia, I, scales as R  4. We have omitted many constants from
this expression, but those are fixed values with which we are not concerned. We
want to see how the two parameters are related, not an actual magnitude. We can
rearrange Equation 8.1 to solve for force as

E b R 4δ
F~ . (8.2)
L3

We now wish to calculate the apparent stress applied to the unit cell and the
apparent strain that the unit cell experiences. To do this, we consider the case in
which surrounding the unit cell is a hypothetical “cube” (Figure 8.3). The length

Figure 8.3 Unit cell undergoing axial (A) (B)

deformations. (A) The coupling F F
between unit cells is modified slightly
such that transmission of force results in F F
axial deformations of beam members. 2R
δ
(B) In the presence of force, vertical L
beams undergo a change in length.
L

F F

F F
 SCALING APPROACHES 227

of each side of the cube is L. The apparent stress is the total force applied across a
surface of the cube divided by the area L2, and the apparent strain is the total
deflection of the cube divided by the side length L. This apparent stress can be
found by dividing the force in Equation 8.2 by the area of the face of the unit cell
over which the force is acting,

F E b R 4δ
σ~ ~ . (8.3)
L2 L5

The apparent strain can be found as the displacement δ divided by the unde-
formed length of the unit cell L,
δ
ε~ . (8.4)
L

Using Equations 8.3 and 8.4, the apparent elastic modulus of the unit cell can be
found by dividing the apparent stress by the apparent strain,

σ R4
E~ ~ Eb 4 . (8.5)
ε L

Recall that volume fraction is the fraction of the volume taken up by polymer
material. We can recast Equation 8.5 in terms of this quantity, which we will
denote as ρvol. For the unit cell, the volume of the beams scale as ~LR2, whereas
the volume of the unit cell scales as L3. In this case,

LR 2 R 2
ρvol ~ ~ 2. (8.6)
L3 L

Combining Equations 8.5 and 8.6 yields the final expression

E ~ E b ρ 2vol. (8.7)

Equation 8.7 implies that when deformation of the cellular solid occurs due to
bending, the apparent macroscopic modulus scales linearly with the polymer
modulus but quadratically with the polymer volume fraction. This is an interest-
ing result, as it suggests that the stiffness of the network is more dependent on
changes in the polymer concentration rather than on the stiffness of the polymers
themselves.

Deformation dominated by axial strain results in a linear

scaling of the elastic modulus with volume fraction
We now perform a similar analysis to find out how the scaling relationship in
Equation 8.7 would be altered if the deformation of the cellular solid was assumed
to occur exclusively due to axial deformation of the beam members. To do this, we
first slightly modify the configuration of the unit cell to that in Figure 8.3. By modi-
fying the coupling, transmits loads from adjoining cells in such a manner that
deformation occurs exclusively due to axial strain in the members. It can be shown
that the force within each beam scales as

E b R 2δ
F~ , (8.8)
L

and that the apparent elastic modulus scales as

E ~ E b ρ vol. (8.9)
228 CHAPTER 8: Polymer Networks and the Cytoskeleton

Nota Bene The derivations of Equations 8.8 and 8.9 are left as student exercises. Equation 8.9
implies that when deformation of the cellular solid occurs due to axial strain, the
Experimental measurements of effective elastic modulus of the network scales proportionally to both the cytoskel-
cytoskeletal volume fraction. We
etal modulus and the polymer volume fraction. By assuming a different mecha-
have seen that the cellular solid
model can predict very different nism of deformation at the microstructural level, we find different dependencies
scaling behaviors depending on of the network stiffness on the polymer volume fraction.
whether the deformation is assumed
to occur due to member bending or
axial deformation. One may ask how
The stiffness of tensegrity structures scales linearly with
these two models compare with member prestress
experimental investigations of the
relationship between cell stiffness We conclude this section with a discussion of tensegrity structures. The term
and cytoskeletal polymer volume tensegrity, which is a contraction of tensional integrity, was originally developed
fraction, and subsequently, which by Buckminster Fuller to describe a structural principle that had been invented by
model is more “correct”. Kenneth Snelson. Tensegrity structures are developed from linear structural ele-
Unfortunately, a conclusive answer to ments that are designed to withstand either tension only (such as strings or cables)
this question has yet to be attained,
owing in large part to the technical
or compression only (such as rigid struts). A critical aspect of a tensegrity struc-
difficulties associated with measuring ture is that none of the elements experience bending moments. This lack of bend-
cytoskeletal volume fraction within a ing moment is typically achieved architecturally by using freely rotating joints at
live cell. all connections. The resulting structure derives its stability from preload or pre-
stress (Figure 8.4).
Donald Ingber first suggested that the cytoskeleton might function as a tensegrity
structure. In this model, large-diameter microtubules act as rigid compression
elements, and the microfilaments and intermediate filaments act as flexible ten-
sile elements. There are some compelling arguments why a tensegrity structure
would be advantageous. Tensegrity structures possess the capacity to deform sig-
nificantly without large deformations of the individual components. That is, the
tension and compression elements rearrange themselves in response to some
load, but the individual elements do not need to stretch or compress by as much
as the entire structure. Such elemental rearrangements also underlie the capacity
for tensegrity structures to translate a deformation in one location into an equally
large (or larger) displacement some distance away, termed action at a distance.
However, a consensus of opinion on whether the cytoskeleton functions as a
tensegrity structure has not yet been formed, and this topic remains a controver-
sial one within cell mechanics.
Scaling relationships for tensegrity structures have been formed. Though its deriva-
tion is outside the scope of this book, it can be shown that in the existence of a pre-
stress, the effective network elastic modulus scales (under certain conditions) as

Figure 8.4 Tensegrity art. Entitled

Mozart by Kenneth Snelson, this
sculpture is made up of tension-only
cables and compression-only struts.
 AFFINE NETWORKS 229

Fp ρ vol
E~ , (8.10)
r2

where Fp is the contractile force within the members generating the prestress.
Equation 8.10 implies that the stiffness of a tensegrity network scales linearly
with the polymer volume fraction, which is perhaps not surprising, given that
(1) the elements within tensegrity structures cannot sustain bending, and (2) a
similar scaling occurs in the axially loaded cellular solid model. Equation 8.10
also implies that the network stiffness scales linearly with the prestress of the
network.

Advanced Material: The controversy over tensegrity

One reason for the controversy over the tensegrity the  tensegrity hypothesis. Nuclear motion has been
hypothesis is that there is evidence both in support of observed in response to a small bead being pulled far
and against this notion. In contrast to the tensegrity from the nucleus when engaging certain receptors, but
model, it is generally accepted that microtubules and not others. These results are interpreted as supporting
actin filaments cannot be classified as strictly compres- the tensegrity model’s action-at-a-distance concept.
sion- and tension-bearing elements, respectively. Prestress in actin filaments has been demonstrated using
Extension of lamellipodia occurs by axial compression of a combination of photobleaching and laser-scalpel tech-
actin fibers, and microtubules in tension are, in part, nology. The tensegrity hypothesis will likely continue to
responsible for pulling chromosomes apart during mito- be controversial for many years to come.
sis. On the other hand, there is evidence supporting

8.3 AFFINE NETWORKS

In the previous section, we focused on deriving simple scaling relationships
between effective stiffness and network density. Such relationships can be very
useful for obtaining order-of-magnitudes estimates of cytoskeletal stiffness
based on experimental measurements of filament density. However, because
these relationships only give the scaling behavior of stiffness with polymer den-
sity, rather than an explicit expression for network stiffness, they are generally
inadequate for mechanical modeling of cells. In this section, we present several
formulations for constitutive models of polymer networks. As we will see, such
constitutive models can be in the form of a direct relationship between stress and
strain, or between strain energy and strain (from which stress/strain relation-
ships can be calculated).

Affine deformations assume the filaments deform as if they

are embedded in a continuum
In formulating relationships between network stiffness and microstructural
properties, an important consideration is how the filaments deform in response
to an applied load. The most common approach is to assume an affine deforma-
tion. An affine deformation is one that can be described by an affine transforma-
tion, which consists of some linear transformation (such as rotation, shear,
extension, or compression) superimposed with a rigid translation. If we let the
position of any point within the undeformed network be given by the vector A,
and its position within the deformed network at some time t be given by a(A,t).
An affine deformation is one that can be described in the form

a ( A , t ) = F (t ) A + c ( t ) , (8.11)

230 CHAPTER 8: Polymer Networks and the Cytoskeleton

Nota Bene
where F(t) is the linear transformation, and c(t) is the rigid-body translation. Note
that if F or c are functions of A, F = F(A,t) or c = c(A,t), then the deformation in
Validity of the affine Equation 8.11 ceases to be affine. Therefore, affine transformations are sometimes
approximation. The validity of the referred to as homogeneous.
affine approximation for different
classes of networks, and in particular, Under an affine deformation, the polymers deform as if they are embedded in a
for networks composed of semi- continuum. Consider an imaginary cubic network subjected to a uniaxial load,
flexible polymers such as the
and assume that the filaments undergo a deformation that is affine. For all fila-
cytoskeleton, is an active area of
research. The conditions under which ment segments oriented similarly between two cross-link points, the stretch is the
deviations from affine behavior occur same. Similarly, at all cross-links in which the two cross-linked polymers are simi-
are a complex problem that depends larly oriented, the change in angle between each cross-linked pair will also be the
on several factors, such as the same. Importantly, the affine assumption greatly simplifies making a constitutive
microstructural details of the model as it really does not matter how many filaments are in the network, because
network, the applied load, and the they are assumed to be all deforming in the same manner by the deformation
length scale being considered.
prescribed by F and c.

Flexible polymer networks can be modeled using

rubber elasticity
In this section, we focus on an approach for analyzing biopolymer networks that
originates from the field of rubber elasticity. The microstructures of rubbers are in
some ways similar to the cytoskeleton. Rubbers are constituted of polymers that
form a cross-linked network structure. The principal difference between rubber
and, say, the actin cytoskeleton is that the individual rubber molecules are highly
flexible. We learned in Chapter 7 that F-actin’s persistence length is on the order
of micrometers, similar to its contour length in vivo. The mechanics of actin are
governed by both energetic and entropic influences. In contrast, rubber mole-
cules have contour lengths much greater than their persistence lengths. The
mechanics of individual rubber molecules are entropy-dominated. Owing to this
entropy-dominated behavior, rubber-like materials have several unique mechan-
ical properties and exhibit some very interesting behaviors. They are highly
deformable and show almost complete recoverability when the load is released.
In addition, the stiffness of rubber increases with temperature, which is the oppo-
site of what occurs in most engineering materials.
To begin our analysis of rubber, consider a network of randomly oriented, cross-
linked polymers. We assume that the network is shaped like a rectangle, with
dimensions Lx, Ly, and Lz. Because rubber networks consist of highly flexible
polymers whose mechanical behavior is entropy-dominated, we will assume that
the polymers within the network can be modeled as ideal chains. Recall from
Section 7.4 and Equation 7.56 that we can write the entropy of the ideal chain as

3kBR 2
S (R ) = − + S0 , (8.12)
2nb 2

where S0 is a constant.
For a single chain, the change in free energy associated with a change in initial
end-to-end vector R = [X,Y,Z] to a new end-to-end vector r =[x,y,z] is
3kBT
∆Ψ = −T(S (r ) − S (R )) = + (r 2 − R 2 ). (8.13)
2nb 2

Also recall from Equation 7.30 that the mean square end-to-end length for the ideal
chain is ∙R2∙ = nb2. This implies that if the network was synthesized by a sequential
process of polymerization followed by cross-linking, then the mean end-to-end
length for each chain would be roughly nb2, assuming that the polymers were given
time to equilibrate before cross-linking, and that the density of the polymer solution
was low enough that substantial inter-polymer volume exclusion interactions did
 AFFINE NETWORKS 231

not occur. Because the chains in the network are randomly oriented, this implies
that the mean squared end-to-end distances in the x-, y-, and z-directions must
equal each other, and therefore,

R2 nb 2
R 2
x = R 2
y = R 2
z = = . (8.14)
3 3

Next, we deform the network such that the lengths in the x-, y-, and z-directions
are now lx, ly, and lz. The deformation of the network is given by

rx = λ x Rx

ry = λ y Ry (8.15)

rz = λz Rz ,

and

lx ly l
λx = , λy = , λz = z (8.16)
Lx Ly Lz

are the stretch ratios in the x-, y-, and z-directions. In this treatment, we will
assume the deformations are homogeneous (and affine). This means that the
stretch of the individual chains between cross-links are spatially independent,
and can be described by the stretch of the bulk network (or, in other words, by the
relations above). Therefore, on average, the mean squared lengths of the chains in
the x-, y-, and z-directions after deformation will be

rx2 = λ x2 Rx2

ry2 = λ y2 Ry2 (8.17)

rz2 = λz2 Rz2 .

The change in free energy of a single chain will be

∆Ψ =
3kBT
2nb 2
(
r 2 − R2 )
=
3kBT 2
2nb 2
(
rx + ry2 + rz2 − Rx2 + Ry2 + Rz2 )
(8.18)
3k T
(
= B 2 λ x2 Rx2 + λ y2 Ry2 + λz2 Rz2 − Rx2 − Ry2 − Rz2
2nb
)
=
2
(
kBT 2
λ x + λ y2 + λz2 − 3 , )

where we have used the relation ∙X 2∙ =  ∙Y 2∙ =  ∙Z 2∙ = nb2/3. The total average change
in free energy for the entire network ∙ΔΨnet∙ can be computed as the product of the
change in free energy for a single polymer, the number of polymers per unit vol-
ume ρn (which is also called the number density), and the total volume of the
network V = LxLyLz as

ρ nVkBT 2
∆Ψ net =
2
(
λ x + λ y2 + λz2 − 3 . ) (8.19)

232 CHAPTER 8: Polymer Networks and the Cytoskeleton

Now, consider the case of uniaxial stretching in the x-direction. We fix one end,
and load the opposite end such that the deformed length in the x-direction is

l x = λ Lx. (8.20)

In the y- and z-directions, there is contraction due to Poisson’s effect. Now, for
simplicity, assume that the network as a whole is incompressible. We can calcu-
late the stretch in the y- and z-directions. In particular, if we equate the volumes of
the undeformed and deformed networks,
Lx Ly Lz = λLx λ y Ly λz Lz . (8.21)

For Equation 8.21 to hold,
1
λ y = λz = . (8.22)
λ

Therefore,

ρnVkBT  2 2 
∆Ψ net = −  λ + − 3 . (8.23)
2  λ 

We can find the force in the x-direction as

∂∆Ψ net  1
fx = = ρnAkBT  λ − 2  , (8.24)
∂l x  λ 

where A = LyLz is the undeformed cross-sectional area (the proof of this is left as
an exercise). Now, we can find the axial stress in the x-direction as

fx  1
σ = = ρnkBT  λ − 2  . (8.25)
A  λ 

In the limit of small strains, it can be shown that

 1
 λ − 2  ≈ 3ε . (8.26)
λ

Combining Equations 8.25 and 8.26, we obtain

σ = 3ρnkBT ε . (8.27)

The results of our analysis show that if we assume small deformations and that
the network is linearly elastic, isotropic, and incompressible, then the Young’s
modulus of our network is E = 3ρnkBT. There are several things worth noting.
First, the stiffness of the network scales linearly with the density of filaments,
ρn. This scaling of stiffness with density is similar to the scaling obtained in
Section 8.2, in which we assumed that all deformations in our lattice of beams
were axial, but not similar to the scaling obtained when we assumed the defor-
mations occurred because of bending. Perhaps this is not unexpected, because
in ­rubber-like networks, the ideal chains behave as entropic springs, which do
not generate force when bent. Second, we can see that the stiffness of the net-
work increases with increasing temperature, unlike most engineering materi-
als. This occurs because in a network of entropic springs, the stiffness of the
network is derived entirely from entropic costs of straightening out the indi-
vidual polymers. In particular, the polymers become straighter and have fewer
configurations available. Increasing the temperature increases the cost of this
reduction of entropy, and as such the material’s elastic modulus effectively
increases.
 AFFINE NETWORKS 233

Anisotropic affine networks can be modeled using strain

energy approaches
In the last section, we derived the effective Young’s modulus for a rubber-like net-
work of flexible polymers. In the derivation, we assumed that the filaments were
isotropically aligned (that is, they had no preferred orientation). Further, we
assumed that the network was incompressible. However, in some cases, we may
wish to relax these assumptions.
An isotropic, linearly elastic material has two independent elastic constants. The
Young’s modulus and Poisson’s ratio. In Section 3.2 we found that, in general
form, Hooke’s law can be expressed as a matrix equation such that

σ = Cε (8.28)

where σ and ε are 6 × 1 vectors. Therefore, in general, C is a 6 × 6 matrix. However,
Nota Bene
it turns out that C must also be symmetric, so even for a fully anisotropic linearly
elastic material there are only 21 independent elastic moduli, Symmetry of C. The reason that C
is symmetric may not be immediately
clear. Consider a complicated
 C11 C21 C13 C14 C15 C16 
composite material with lots of fibers
 C21 C22 C23 C24 C25 C26  aligned in the z-direction. In this
  anisotropic case, why would the
 C31 C32 C33 C31 C32 C33 
stress in the x-direction resulting
C =  ,
 (8.29) from a strain applied in z-direction be
equal to the stress in the z-direction
 C411 C42 C43 C44 C56 C46 
  when the same strain is applied in
 C51 C52 C53 C54 C55 C56  the x-direction? One requirement for
 C C62 C63 C64 C65 C66  an elastic material it that it has a
61
strain energy function
σε cεε ∂ 2w
where Cij = Cji. In this section, we demonstrate a calculation of these 21 elastic w(ε ) = = . So, C = ,
2 2 ∂ε 2
moduli for a network of anisotropically aligned elastic rods undergoing an affine
which is symmetric, because the
deformation. However, before we do this, we will first discuss computation of
order or differentiation does not
elastic moduli from the strain energy density, because this approach will be useful matter.
in several developments in following sections.

Elastic moduli can be computed from strain energy density

For linearly elastic materials, the elastic moduli relate stress and strain. For a rod
composed of an isotropic linearly elastic material, the Young’s modulus E gives
the stiffness of an object in resisting a uniaxial stress, and relates the stress σ that
results from a uniaxial strain ε,

σ = Eε . (8.30)

In Section 5.1 we introduced the notion of strain energy. For convenience, we

denoted the strain energy density as dW, but here we will reserve W for the total
strain energy and use a lower-case w for strain energy density. In Equation 5.6 we
showed that the strain energy density, or the strain energy per unit undeformed
volume, is one-half the component-wise product of stress and strain,
σε
w= .
2 (8.31)

Combining Equations 8.30 and 8.31, we can see that the strain energy density w
can be formulated in terms of E and ε as

1 2
w= Eε . (8.32)
2
234 CHAPTER 8: Polymer Networks and the Cytoskeleton

Equation 8.32 implies that if we know the relationship between strain energy den-
sity and strain, we can compute the Young’s modulus E for the rod. In particular,
we see that E can be calculated by differentiating the strain energy density as
∂ 2w ∂ 2 1 2 Eε 2
=
∂ε 2 ∂ε 2
(8.33)
∂E ε
=
∂ε

= E.
In this simple example, we assumed that the rod was composed of an isotropic lin-
early elastic material. However, it can be shown that this relation holds for fully ani-
sotropic linearly elastic materials as well. Consider a continuum that is subjected to
some infinitesimal, homogeneous deformation described by the strain tensor. This
deformation results in a stress state described by the corresponding stress tensor.
Similar to the approach of determining the Young’s modulus by differentiating the
strain energy density with respect to strain, it can be shown that, more generally, all
21 independent elastic moduli of C can be computed as

 ∂ 2w ∂ 2w ∂ 2w ∂ 2w ∂ 2w ∂ 2w 
 ∂ε xx ∂ε xx ∂ε xx ∂ε yy ∂ε xx ∂ε zz ∂ε xx ∂ε xy ∂ε xx ∂ε yz ∂ε xx ∂ε xz 
 
 ∂ 2w ∂ 2w ∂ 2w ∂ 2w ∂ 2w ∂ 2w 
 
 ∂ε yy ∂ε xx ∂ε yy ∂ε yy ∂ε yy ∂ε zz ∂ε yy ∂ε xy ∂ε yy ∂ε yz ∂ε yy ∂ε xz 
 ∂ 2w ∂ 2w ∂ 2w ∂ 2w ∂ 2w ∂ 2w 
 
 ∂ε zz ∂ε xx ∂ε zz ∂ε yy ∂ε zz ∂ε zz ∂ε zz ∂ε xy ∂ε zz ∂ε yz ∂ε zz ∂ε xz 
C= .
 ∂w ∂ 2w ∂ 2w ∂ 2w ∂ 2w ∂ 2w 
2

 ∂ε xy ∂ε xx ∂ε xy ∂ε yy ∂ε xy ∂ε zz ∂ε xy ∂ε xy ∂ε xy ∂ε yz ∂ε xy ∂ε xz 
 
 ∂w ∂ 2w ∂ 2w ∂ 2w ∂ 2w ∂ 2w 
2

 ∂ε yz ∂ε xx ∂ε yz ∂ε yy ∂ε yz ∂ε zz ∂ε yz ∂ε xy ∂ε yz ∂ε yz ∂ε yz ∂ε xz 
 
 ∂ 2w ∂ 2w ∂ 2w ∂ 2w ∂ 2w ∂ 2w 
 
 ∂ε xz ∂ε xx ∂ε xy ∂ε yy ∂ε xz ∂ε zz ∂ε xz ∂ε xy ∂ε xz ∂ε yz ∂ε xz ∂ε xz 
(8.34)

Equation 8.34 is a very useful result, as it implies that we can determine the effec-
tive elastic moduli of the network as long as we can formulate the functional
dependence of strain energy density with strain.

Example 8.1: The cytoskeletal modulus determined from strain energy density
If you examine the cytoplasm of a very thin cell, in a  C11 C12 C13 0 0 0 
region far from the nucleus, you might see the major fila-  C12 C11 C13 0 0 0 
ments of the cytoskeleton lying horizontally, in loose lay-  
ers. If you are given the strain energy density of such a  C13 C13 C33 0 0 0 
section as, C= 0 0 0 C44 0 0 
 
 0 0 0 0 C44 0 
w = 0.5C11(𝜀xx2 +  𝜀yy2) + 0.5C33(𝜀zz2) + C12𝜀xx𝜀yy + C13𝜀zz  
(𝜀xx +  𝜀yy) + 0.5C44(𝜀xy2 +  𝜀yz2) + 0.25(C11 − C12)  𝜀xz2,  0 0 0 0 0 C11 − C12 
 2 
calculate the stress–strain relationship and determine
This form of a material is associated with transverse iso­
the general type of symmetry, if any, that exists in this
tropy, which means that the material has one set of prop-
model.
erties in one longitudinal direction and another set in
Using the tensor from Equation 8.35, we can fill in the any transverse direction within a plane perpendicular to
components as, the longitudinal direction.
 AFFINE NETWORKS 235

Elastic moduli of affine anisotropic networks can be

calculated from appropriate strain energy density
and angular distribution functions
After discussing the relationship between strain energy density and elastic mod-
uli, we now return our attention to computing the effective elastic moduli for an
anisotropic network undergoing an affine deformation. Consider a network of
cross-linked elastic rods. For simplicity, we assume that these rods are all cylin-
drical and of the same length and cross-sectional area. The network is contained
within a cube of side length L, with total volume Vnet = L3. Cross-links are
assumed to be freely rotating (namely that there is no energetic cost for changes
in angle between filaments), and deformations are assumed to be affine.
Assume that each rod has an orientation that is specified by a unit vector n with
components

 nx 
n =  ny  . (8.35)
 
 nz 

We can construct an angular probability density function, ω(n), which gives the
distribution of filaments over the unit sphere (which we will annotate as S2 based
on convention). The angular probability density function has the property that

∫ ω(n)dS = 1,
S2
(8.36)

which normalizes the total probability to be one. Next, consider our network
under an imposed load. Recall that the deformation is assumed to be affine.
The rods will freely rotate relative to one another and undergo some degree of
stretch. For an elastic rod with Young’s modulus Erod and volume Vrod, the total
strain energy in deforming the rod by stretching it by a factor of λ is

Vrod Erod
Wrod (λ ) = (λ − 1)2 . (8.37)
2

The degree to which each filament undergoes axial deformation will depend on
the network deformation and the orientation of the filament. In the case of
small deformations, the stretch for a filament oriented in the direction n under
an imposed strain of 𝜀 is approximately

λ(ε , n) = nT εn. (8.38)

For a network with rods oriented with angular probability density ω(n), the total
strain energy of the network is

Vrod Erod N
Wnet (ε ) =
∫W
S2
rod (λ (ε , n))ω (n)NdS =
2 ∫ (n ε n − 1) ω (n)dS ,
S2
T 2
(8.39)

where N is the total number of rods. The strain energy density can be calculated by
dividing the strain energy by the undeformed volume

ρvolErod
Wnet (ε ) =
2 2
∫
(nT εn − 1)2 ω (n)dS , (8.40)
S
236 CHAPTER 8: Polymer Networks and the Cytoskeleton

where ρvol = NrodVrod/Vnet is the volume fraction of rods within the network.

Equation 8.40 gives the dependence of strain energy density with strain. Using
this expression, each of the 21 independent elastic moduli can be computed using
Equation 8.34. For most angular distributions, the integral and derivatives in the
above expression are difficult to solve analytically but can be solved computation-
ally using numerical methods. Upon inspection of Equation 8.40, it is noteworthy
that the elastic moduli will have a linear dependence on polymer density, similar
to the linear dependence on density observed in our analysis of rubber-like net-
works and our analysis of cellular solids in which we assumed deformation was
stretching-dominated.

8.4 BIOMECHANICAL FUNCTION AND

CYTOSKELETAL STRUCTURE
Up to this point, our focus has been on the formulation of relationships between
effective mechanical behavior and microstructural network properties. In this
section, we shift our focus to analyzing the mechanics of specific cytoskeletal
structures, in particular, the red blood cell’s cytoskeleton and the cross-linked
actin bundles within filopodia. Although the role of the distinct architectures
found in these structures in facilitating or regulating biological function is an
active area of research, in this section we seek to use relatively simple analyses to
explore whether the architectures found in these structures are optimally suited
to perform their biological roles.

Filopodia are cross-linked bundles of actin filaments involved

in cell motility
We begin our analysis of structures with a discussion of filopodia. During cell
migration (a process that will be discussed in detail in Chapter 10), several
cytoskeletal alterations occur, particularly within two distinct subregions of the
cell: the leading edge (the part of the cell that is moving forward) and the trailing
edge (the part which follows the main bulk of the cell). Filopodia are dynamic,
cross-linked bundles of actin filaments at the leading edge of crawling cells.
Morphologically, they resemble little fingers, and they are thought to act as “feel-
ers” as the cell moves (Figure 8.5). These bundles of actin quickly grow out from
the leading edge, pushing on the membrane, typically at rates on the order of
~0.1 μm/s to form rod-like structures, and then retract. The total time for exten-
sion and retraction is ~100 s. Filopodia are on the order of ~0.2 μm in diameter
and 1–5 μm in length. Within a single filopodium, 20–30 actin filaments are
aligned parallel with their barbed ends facing the membrane and cross-linked by
the protein fascin. At the ends of the filaments are capping proteins, so named
because they cap the actin filaments to prevent further growth.
The way in which cells are able to form filopodia and regulate their structure is an
active area of investigation. Even fundamental questions about their structure,
such as how densely cross-linked the actin filaments are, are unknown. In this sec-
tion, we explore the role of mechanics in regulating their structure. In particular,
we seek to determine the maximum length a filopodium may be before it under-
goes mechanical failure. The motivation for this analysis comes from two observa-
tions: filopodia are slender structures; and they are generally straight (see Figure
8.5). This implies that filopodia may be structured in a specific manner to resist
buckling. We will perform this analysis under two extreme conditions: (1) when no
cross-linking is present, or (2) when a high degree of cross-linking is present.

Actin filaments within filopodia can be modeled as elastic

beams undergoing buckling
To begin, we first consider whether to treat filipodium actin filaments as ener-
getically or entropically dominated polymers (or both). In Section 7.3, we learned
 BIOMECHANICAL FUNCTION AND CYTOSKELETAL STRUCTURE 237

(A) (B) fascin

capping protein
actin-related protein 2/3 complex
F-actin

1–5 µm

–0.2 µm

20–30 actin filaments

per filopodium

Figure 8.5 Cytoskeletal structure within filopodia. (A, top) Electron micrograph of the actin cytoskeleton within a lamellipodium
of a cell. A dense network of actin filaments can be seen. Several filopodia can be seen protruding from the edge. (A, bottom) Close-up
of a filopodium. The aligned bundle of actin filaments within the filopodium can be seen. (B) Schematic of several important proteins
within filopodium. Note that the actin filaments are bundled together by the cross-linking protein fascin. (A, from, Mejillano MR et al.
(2004). Cell.)

that the persistence length of F-actin is on the order of ~10 μm. Filopodia are
much shorter than this, ~1 μm in length. Because the persistence length of actin Nota Bene
is an order of magnitude larger than the average length of the filaments within Image-based modeling of
the filopodia, we neglect entropic contributions and analyze the filaments as cytoskeletal networks. The
elastic beams. Since we are interested in the buckling behavior of filopodia, we angular distributions of filaments for
seek a relation between buckling loads and beam mechanical properties and several cases of cytoskeletal
networks have been quantified using
geometry. Such a relation was found in Section 3.2 when we analyzed the buck- image-processing analysis. Such
ling behavior of an elastic beam of length L with one end fixed, and with an axial angular distributions could be
force F applied at the free end (Figure 8.6). We found that for such a beam it will incorporated into the model
buckle if F exceeds the buckling force (Equation 3.38) described in Equation 8.40 to
simulate the mechanical behavior of
π 2 EI such networks based on realistic
Fbuckle = , (8.41) microstructures obtained from
4 L2 imaging.

where E is the Young’s modulus of the beam, and I is the moment of inertia.

Advanced Material: Cross-links and “dangling ends”

Within polymer networks such as the cytoskeleton, an occurs through the cross-links such that bending and
individual polymer will typically have many cross-links stretching of polymers occurs between cross-links only.
along its span. However, near the ends of the polymer, the Because the free end is ”dangling”, it will likely not undergo
ends will be “dangling”: in other words, there will be a free stretch or bending. In Equation 8.40, for simplicity we
segment near the ends of the polymer that is not bounded assume that the free ends satisfy the affine deformation
on both sides by a cross-link. This free segment contrib- assumption. Though this is likely unphysiological, the
utes to the density of the network but in general, not its contributions of these free ends to the total strain energy is
strain energy. The reason is that the load transmission small.
238 CHAPTER 8: Polymer Networks and the Cytoskeleton

The membrane imparts force on the ends of filopodia

F
With our relation for buckling force identified, we need an estimate of the force
imparted on the ends of filopodia. During filopodium formation, actin bundles
F exert a force against the membrane to form a local protrusion. We idealize the
actin bundle as a cylinder of radius r (Figure 8.7). The force exerted by the mem-
brane, Fmem, can be approximated by making an imaginary cut through the cyl-
inder and membrane together. The resultant force of the membrane is equal to
the product of the surface tension, N, of the membrane and the circumference of
the cylinder,
Figure 8.6 Schematic Fmem = 2πRN . (8.42)
demonstrating buckling of beam
with one end fixed and a force F
applied at the free end. The beam Let us apply Equation 8.42 to a real-life situation. For neutrophils, the surface ten-
will buckle if F exceeds Fbuckle. sion as measured through micropipette aspiration is approximately 35 pN/μm,
resulting in a force of 22 pN for a filopodium 100 nm in radius. However, this force
is likely conservative, as it does not take into account membrane bending and
breaking of membrane–cortex links, both of which will contribute resistance to
protrusion. Experimentally, ~50 pN of force has been found to be necessary to
form filopodium-like membrane tethers within neutrophils; therefore, we will
assume that Fmem = 50 pN.

r
The maximum filopodium length before buckling in the
absence of cross-linking is shorter than what is observed
in vivo
Using Equation 8.42 and our estimate of Fmem, we can now calculate the maxi-
mum filopodium length before buckling occurs. We first assume that there is no
Figure 8.7 Protrusion of membrane cross-linking within the bundle. Assume that there are n = 30 actin filaments
by a cylindrical bundle. The force within the bundle, and each filament has a radius of Ractin = 3.5 nm and Young’s
exerted on the protrusion by the
membrane is proportional to the radius
modulus of Eactin = 1.9 GPa. Each filament will experience a compressive force
of the cylinder. Fmem/n, so the maximum filopodium length before buckling occurs is

π 2 Eactin I actin
Lnocl =
4 Fbuckle

πR 4actin (8.43)
π 2 Eactin
= 4 .
Fmem
4
n

Plugging in the appropriate values gives length before buckling occurs, Lnocl, is
Lnocl = 0.57 μm. Therefore, for a filopodium containing 30 actin filaments that are
not cross-linked, it can only grow to be 0.57 μm long before it will buckle. This
length is substantially shorter than the 1–5 μm long filopodia observed in vivo,
implying that in the absence of cross-linking, filopodia do not possess adequate
mechanical integrity to extend. Next, let us examine how this length is altered
when we take cross-linking into account.

Cross-linking extends the maximum length before buckling

Consider a situation in which we are performing an analogous analysis for a bun-
dle of cylindrical steel beams. To simulate a high degree of cross-linking, one
could weld all the steel beams together into an approximately cylindrical-shaped
bundle. As more beams were welded together, the bundle would begin to resem-
ble a cylinder. Its mechanical behavior would more and more resemble a single
solid cylindrical steel beam with the same radius as the bundle. Motivated by this
 BIOMECHANICAL FUNCTION AND CYTOSKELETAL STRUCTURE 239

thought experiment, as a first approximation, we will assume that a highly cross-

linked bundle of actin will behave as a single elastic beam of effective radius rbundle
(Figure 8.8). To calculate this radius, we know that for a bundle of n filaments, the
total cross-sectional area of F-actin is

( )
2
nπR 2actin = π nRactin . (8.44) Figure 8.8 Two different
approximations for analyzing the
filopodium. In the first (left) we assume
The term in parenthesis on the right-hand side of Equation 8.44 yields the radius no cross-linking. In the second (right) we
of a cylinder with the same cross-sectional area of n actin filaments, which is the assume that the bundle is tightly
effective radius of the bundle, cross-linked, resulting in behavior similar
to that of a single, large cylinder.

Rbundle = nRactin . (8.45)

Now, the moment of inertia of the bundle becomes

πR 4bundle
I bundle =
4
(8.46)
( ).
4
π nRactin
=
4

Finally, assuming Fbuckle = Fmem, then

π 2 Eactin I bundle
Lcl =
Fbuckle
(8.47)
( )
4
π nRactin
π Eactin
2

= 4 .
4Fmem

Substituting the appropriate values gives Lcl = 3.2 μm. In other words, a highly
cross-linked filopodium, can grow to be 3.2 μm before buckling. Comparing
Equations 8.43 and 8.47, we find that the ratio of Lcl/Lnocl is n . This indicates that
the more filaments that are in a filopodium, the more pronounced the difference
in the buckling lengths between the cross-linked and un-cross-linked filopodia.
What can be learned from our analysis? We have already learned that filopodia of
1–5 μm are commonly observed in vivo. Our analysis show that without cross-
linking, filopodia will buckle before they are able to reach these lengths. However,
if they are highly cross-linked, they are much more suited to resist buckling at that
length. Filopodia much longer than Lcl (on the order of tens of micrometers) have
been observed on occasion. This suggests that other mechanisms may act in con-
cert with cross-linking to stabilize the actin bundles within filopodia, such as con-
straints on transverse deformations by the membrane sheath surrounding the
bundle.

Is the structure of the red blood cell’s cytoskeleton functionally

advantageous?
We now turn our attention to another example in which cytoskeletal structure
may play an important role in biological function, the highly structured red blood
cell cytoskeleton. The mechanical behavior of red blood cells is extremely impor-
tant to their function. To deliver oxygen to different parts of the body, these cells
 240 CHAPTER 8: Polymer Networks and the Cytoskeleton

(A) must be able to squeeze through tiny capillaries (many of which have diameters
that are smaller than the cells themselves), and then return to their original
shape. This behavior depends in large part on the cytoskeletal mechanics of
these cells.
In red blood cells, the cytoskeletal network is bound to the membrane in a two-
dimensional network. The main constituent of the network is a polymer called
spectrin. The spectrin polymers are bound together at distinct vertices, or “junc-
tions”. At each vertex is a junctional complex consisting of several proteins
(including F-actin) that serve to cross-link the spectrin polymers together, as
well as anchor the cytoskeletal network to the membrane. There are also protein
complexes along the length of the spectrin polymer that anchor it directly to the
membrane. Typically, six spectrin polymers radiate from each junction. Such a
junction is considered to exhibit sixfold connectivity. Fourfold connectivity, or
four spectrin polymers radiating from each junction, has also been observed
(B)
Q1 (Figure 8.9).
The question we seek to address is what are the structural and functional impli-
cations of these different connectivities? As the analytical and numerical
tools  for understanding the mechanics of the red blood cell’s cytoskeleton
become more sophisticated, the impact of such distinct microstructures on the
mechanical behavior of red blood cells is becoming better understood.
However, we can also learn a lot from simple analyses. In the following section,
we estimate the mechanical properties of a model red blood cell’s cytoskeleton
from the mechanical behavior of its polymers and its microstructure. By seeing
how these properties change with sixfold and fourfold connectivity, we also
can see why, for red blood cells, sixfold connectivity may be functionally
advantageous.

Figure 8.9 Electron micrographs of

the red blood cell’s cytoskeleton Thin structures can be analyzed using the two-dimensional
with different connectivities. shear modulus and the areal strain energy density
(A) Fourfold connectivity. (B) Sixfold
connectivity. (From, Byers TJ et al. (1985). The red blood cell’s cytoskeleton is thin, only ~100 nm thick, which is much smaller
Proc. Natl. Acad. Sci. U. S. A.) than its micrometer-sized lateral dimensions. For such structures, we usually are
not interested in how quantities such as stress and strain change with depth. It is
mathematically convenient to treat them as two-dimensional by integrating out
the depth. To do this, consider a three-dimensional block with length l, height h,
depth d, and shear modulus G undergoing shear (Figure 8.10). Recall from
Section 3.2 that when we can shear the block by displacing the top surface by a
small amount 𝛿, which results in the sides of the block making an angle of γ with
an imaginary vertical line. We know that

∂u ∂v
γ = +
∂y ∂y
δ
= (8.48)
h
= tan γ

≈γ,

where the last line is the small angle approximation. Equation 8.48 implies that for
small shear deformations, the (engineering) shear strain is simply the angle
formed with the vertical, γ. The shear modulus G relates the shear stress τ to (engi-
neering) shear strain γ,

τ = Gγ . (8.49)

 BIOMECHANICAL FUNCTION AND CYTOSKELETAL STRUCTURE 241

(A) l τ Figure 8.10 Treatment of a thin

δ three-dimensional structure
subjected to shear as a
two-dimensional object. (A) A thin
block of height h, length l, depth d, and
shear modulus G is subjected to shear
h G γ strain γ by imposing a shear stress τ,
resulting in the displacement of the top
surface by a small amount δ. (B) For
simplicity, we treat the thin structure as
a two-dimensional structure by
integrating out the depth. The structure
d possesses a shear modulus per unit
ns depth of Ks, and undergoes a shear
δ strain γ under a shear force per unit
(B) length ns.

h Ks γ

In a manner analogous to Equation 8.33, the strain energy density for an isotropic
linearly elastic material under shear strain γ is
1 2 (8.50)
w= Gγ ,
2
implying that

∂ 2w (8.51)
G= .
∂γ 2
If we assume that the depth d is very small, we can treat the block as a two-
dimensional structure. Instead of a shear stress (with units of force per area), we
can define a shear force per unit length ns such that

ns = τd , (8.52)

assuming that the shear stress is constant through the depth. The shear force per
unit length can then be related to the engineering shear strain as

ns = K sγ , (8.53)

where Ks is the shear modulus per unit depth,

K s = Gd . (8.54)

Similar to Equation 8.50 the areal strain energy density, or the strain energy per
unit undeformed area (as opposed to volume) is
1
Wa = K sγ 2 (8.55)
2
and the areal strain energy density can be differentiated to obtain the shear mod-
ulus as
∂ 2Wa (8.56)
Ks = .
∂γ 2
242 CHAPTER 8: Polymer Networks and the Cytoskeleton

(A) Sixfold connectivity facilitates resistance to shear

Using Equation 8.56, we will now calculate Ks for the red blood cell’s cytoskeleton
assuming sixfold or fourfold connectivity. Our strategy is to find the total change
in strain energy W for a given shear strain γ, divide by the undeformed area to
ksp /2 obtain wa, and finally find Ks using the above expression.
To compute a relation for areal strain energy density, we first need to model the
(B) polymers and microstructure. In a cell, spectrin polymers have a contour length of
L = 200 nm, and a persistence length of ℓp = 15 nm. The junctions are separated
by a distance of 75 nm. Because ℓp ⪡ L and the spectrin polymers are not fully
ksp /2 stretched out, we can analyze the cytoskeleton as a network of entropic springs
with spring constant ksp = (3kBT)/2(Lℓp) (Equation 7.60), connected at the junc-
ksp /2 tions by freely rotating cross-links.
To simplify our analysis, we analyze a unit cell of the network instead of an entire
network. Consider a sixfold network of springs with constant ksp. If we use an
equilateral triangle of springs as the unit cell of the network, then we can see in
(C) Figure 8.11 that each pair of neighboring triangles contributes a spring between
each pair of junctions, resulting in two springs of stiffness ksp (or, equivalently, a
single spring of stiffness 2ksp) between each pair of junctions. We can correct this
ksp by making the stiffness of our springs in our equilateral triangle equal to ksp/2.
Now, consider an equilateral triangle of springs, each with side length R0. The tri-
angle has a height of

2
R 
Figure 8.11 Schematic h = R02 −  0 
demonstrating justification for  2
choice of unit cell. For our analysis, (8.57)
we assume that the unit cell is an 3 R0
equilateral triangle with spring = .
constant ksp/2 (A). The spring constant 2
is ksp/2 when this unit cell is
superimposed with similar unit cells
If we now displace the top vertex by a small amount 𝛿 such that the triangle under-
(B). The result is a sixfold network with goes shear strain γ (Figure 8.12), then
spring constant ksp (C).
δ
tan γ =
h
(8.58)
2δ
= .
3 R0

But tan γ ≃ γ for small γ, so Equation 8.58 can be rewritten as

3 R0γ
δ = . (8.59)
2

Figure 8.12 A simple model of δ R0/2 δ

sixfold symmetry. This simple model is
composed of an equilateral triangle of
springs undergoing shear strain γ by
displacing the top surface by a small R0
amount δ. h
R
γ

R0
 BIOMECHANICAL FUNCTION AND CYTOSKELETAL STRUCTURE 243

Under this deformation, the left and right diagonal springs lengthen and shorten,
respectively, while the bottom spring is unchanged. The deformed length R of the
left diagonal spring can be found geometrically,

2
R 
R = h2 +  0 + δ 
 2 

2
 3 R0   R0 
2
=   +  + δ 
 2  2

(8.60)
= R 20 + R0δ + δ 2

R 20δ R 20δ 2
= R 20 + +
R0 R 20

δ δ2
= R0 1 + + 2.
R0 R 0

Because 𝛿 is small, we ignore higher-order terms in 𝛿, and Equation 8.60 becomes

δ
R ≈ R0 1 + . (8.61)
R0

We can simplify this even further by noticing that

2
 δ  δ δ2
 1 + 2 R  = 1 + R + 4 R 2
0 0 0
(8.62)
δ
≈1+ .
R0

Combining Equations 8.61 and 8.62, R can be rewritten as

2
 δ 
R ≈ R0  1 +
 2 R0 
(8.63)
δ
≈ R0 + .
2

This implies that, to first order, the left diagonal spring increases in by 𝛿/2. It can
be shown similarly that the right diagonal spring shortens by the same amount.
We know that the change in strain energy for a spring with constant k, original
length R0 and new length R is

k( R − R0 )2
Wsp = .
2 (8.64)

The total change in strain energy is the sum of the change in strain energy for all
three springs. Assuming a spring constant of ksp/2 for our network, this expression
244 CHAPTER 8: Polymer Networks and the Cytoskeleton

becomes

W = Wspleft + Wspright + Wspbottom

1  ksp  1  ksp 
(
( R0 + δ 2) − R0 ) (
( R0 − δ 2) − R0 )
2 2
= + +0
2  2  2  2  (8.65)

kspδ 2
= .
8

The areal strain energy density can be found by dividing W by the undeformed
area of the triangle,

W left
wa =
Asp

kspδ 2
= 8
1
R0h (8.66)
2
kspδ 2
= 8 .
1 3 R0
R0
2 2

We can write the areal strain energy density in terms of γ by substituting our rela-
tionship between 𝛿 and γ found in Equation 8.59 into Equation 8.66, which becomes

3kspγ 2 (8.67)
wa = .
8

Finally, the shear modulus of the network Ks is

∂ 2wa
Ks =
∂γ 2
(8.68)
3ksp
= .
4

Upon inspection of Equation 8.68, there are several interesting things to note.
First, there is no explicit dependence of the shear modulus on the number, den-
sity, or volume fraction of filaments. This is because sixfold connectivity necessi-
tates a fixed relationship between filament length L and the filament density. In
particular, if we model each spectrin polymer as a rod with length L and cross-
sectional area A, and we assume that the cytoskeletal network/membrane has a
depth d, then the volume fraction scales as

ρ ∼ ( AL)/( L2d ) ∼ A/( Ld ). (8.69)

In other words, the volume fraction scales linearly with (1/L). Because ksp ~ 1/L,
this implies that Ks ~ ρ. Another interesting observation is that our prediction for
Ks compares remarkably well with experiments. Substituting b = 30 nm, T = 300
K, L = 200 nm for spectrin yields Ks = 0.9 μN/m. Experimental measurements of
Ks of red blood cell cytoskeletons in which the membrane was removed showed
 BIOMECHANICAL FUNCTION AND CYTOSKELETAL STRUCTURE 245

average values of the shear modulus to be Ks = 2.4 μN/m. This is roughly three
times our predicted value, but still is in agreement to an order of magnitude. What
are some potential sources of error? One may be our approximation of spectrin as
an ideal chain. Remember that when R approaches L, the stiffness of a real chain
will increase, while the stiffness of an ideal chain will not. Therefore, the Gaussian
approximation is best when L ⪢ R, whereas for the red blood cell cytoskeleton, L
is only a little more than twice that of R.

Fourfold connectivity does not sustain shear as

well as sixfold
We will now calculate Ks for a fourfold network. Here, our unit cell is a square
­lattice of springs, each with spring constant ksp/2 and undeformed length R0 δ
(Figure 8.13). If we displace the top surface of the square by a small amount 𝛿,
then the top and bottom springs do not change length, and the deformed lengths
of the left and right springs are
R0 R
R ≈ R 20 + δ 2
(8.70)
≈ R 20 ,

if we ignore higher-order terms. What this means is that, to first order, the left
and right springs do not change length. This implies that the change in strain Figure 8.13 A simple model of
energy will be effectively zero, and so Ks = 0. This result arises because the stiff- fourfold symmetry. For fourfold
ness in shear comes from springs changing in length. However, if the springs do symmetry we utilize a simple square
not change length, then the network cannot resist shear, and so it has no effec- network of springs undergoing shear by
displacing the top surface by a small
tive stiffness. The result of this analysis is that sixfold connectivity is advanta- amount δ.
geous for red blood cells compared with fourfold connectivity, because sixfold
connectivity allows resistance to shear. Such resistance is important in allowing
red blood cells to squeeze through tiny capillaries and return to their original
shape.

Example 8.2: Nonlinear fourfold connectivity

Instead of dropping the second-order term in the four- energy density is the total strain energy divided by the
fold connectivity model, assume that we keep it. Does area, or
the analysis change?
1 ksp δ 4 ksp 4
We assume that the deformation is small compared ws = = γ ,
with the side of the square; that is, 𝛿 ⪡ R0. Therefore, we R 20 8 R 20 8
can write
where γ is the shear strain and is very small compared
δ2 with one. The shear modulus is then
R 2 = R02 + δ 2 , so R ≈ R0 + .
2 R0 3
Ks = ksp γ 2 .
2
The total strain energy is then
If you compare this expression to the expression for the
1 ksp 1 ksp k δ4
W = ( R − R0 )2 + 0 + ( R − R0 )2 + 0 = sp 2 sixfold network, you will see that fourfold connectivity
2 2 2 2 8 R0 stiffness is tiny, in part because the γ2 term is very small.
The modulus here is not constant, because it is greater as
for the contributions from the left, top, right, and bottom you shear the network more, but its magnitude is essen-
sides (see Figure 8.13) and substituting for R. The strain tially negligible.
246 CHAPTER 8: Polymer Networks and the Cytoskeleton

Key Concepts

• Polymer network models aim to capture the • Filament cross-linking can have a substantial impact
aggregate behavior of large numbers of individual on the mechanical behavior of a network independent
filaments. They can be used to model cell behavior of filament properties and density.
under different underlying assumptions. • Strain energy can be used to calculate the moduli for
• Even if explicit relationships between mechanical affine materials, even if they are not isotropic.
properties and network parameters cannot be found, Anisotropic network behavior can be estimated from
scaling relationships can be identified. The cellular filament orientation.
solids model predicts a squared modulus–density • The geometric arrangement of polymers can alter the
relationship for bending dominated filament loading resistance of the macrostructure to deformation.
and a linear one for axial loading. Sixfold symmetric (triangular) networks tend to be
• Tensegrity models have no bending and rely on stiffer in shear than fourfold (square) symmetric
tensile and compressive elements forming a stable networks.
structure. They predict linear scaling of modulus with
both density and pre-load. They also predict the
action-at-a-distance phenomenon.

Problems

1. For a rubber-like network of flexible polymers loaded in

uniaxial tension show that the force required to stretch microtubule is a solid rod of radius r and they are all
1 tightly cross-linked to one another. Assume that there
the material by lx is ρAkT( λ − 2 ). are enough microtubules that the circumference of the
λ
primary cilia can be approximated by the sum of the
2. Show that in the small strain limit λ = 1 + 𝜀 and
1 individual microtubule diameters.
( λ − 2 ) ≈ 3ε .
λ 5. Use your result in Problem 4 to calculate the buckling
3. Compute the buckling length of a primary cilium. length of a primary cilium.
The primary cilium is made up of nine microtubules
arranged in a ring as shown. Assume that each 6. Consider a polymer network made up of a cubic lattice
microtubule is not linked to its neighbors and therefore of repeating junction points that are connected by
buckles independently. Be sure to state what values polymer filaments.
you are assuming for any constants such as Young’s
modulus, membrane tension, radius and moment of F
inertia of the microtubules.

R
r

Approximate the behavior of each polymer as an ideal

4. For the example above, use the parallel axis theorem to chain with n links of Kuhn length b at temperature T.
Further assume that the equilibrium length of each
show that the bending moment of inertia of a collection molecule is R0, such that R0 = b 2n/3 .
nR 4  2n3  (a) W
 hat would the macroscopic Young’s modulus, E
of n microtubules is  n + π 2  assuming each
4 and Poisson ratio, ν, be?
 Annotated References 247

(b) W hat happens to the Young’s modulus if we 8. Discuss the following statements as they support or
increase the temperature? Why? oppose tensegrity as a model.
(c) Now, repeat your answer to (a) and (b), but this (a) The notion of action at a distance applies to a
time assume that the polymers have a long simple beam embedded in a wall. Push down
persistence length compared with their contour on the beam at the free end and there must
length. Model them as a network of rods with E be a reaction at the embedded end. If the wall
of 1 GPA and diameter of 10 nm. Assume that the is flexible enough, you will see substantial
junctions are freely rotating pin joints. deformations at considerable distance from the
(d) For your answer in (c), how would you expect the applied load.
Young’s modulus of the material to scale with the (b) If you sever a single actin filament within a living
modulus of the polymer and volume fraction of cell, the filament clearly retracts, and the local
the polymer. Only present a verbal argument. No structure of the cell changes within a short
calculations are required. time. However, within a fixed cell, severing an
(e) How would you expect the shear modulus of the actin filament does not result in any dramatic
material to scale with the modulus of the polymer changes.
and volume fraction if the junctions were not
pinned (they are bound such that the angles the 9. Assuming a linearly elastic isotropic material with
polymers make with each other is fixed). three-dimensional deformations (small deformations),
write the strain energy density in terms of strains and
7. Examine the network response of fourfold versus constants E and Poisson’s ratio.
sixfold connectivity for a normal stress, instead of shear
stress. Is there an advantage to one versus the other 10. Using an approach similar to that in the bending-
under normal stress? You may assume the normal dominated cellular solids model, derive the
stress results in similar strains for both cases. relations in Equations 8.8 and 8.9.

Annotated References

Boal D (2001) Mechanics of the Cell. Cambridge University Press. An Mogilner A & Rubinstein B (2005) The physics of filopodial protru-
excellent source for detailed analyses of red blood cell cytoskeletal sion. Biophys. J. 89, 782–795. Examines the mechanics and spatial-
mechanics. The analysis of sixfold and fourfold connectivity was temporal dynamics of filopodia. The analysis of filopoidal buckling
based on developments in this text. was based on treatments in this study.
Byers TJ & Branton D (1985) Visualization of the protein associa- Rubenstein B & Colby RH (2003) Polymer Physics. Oxford University
tions in the erythrocyte membrane skeleton. Proc. Natl. Acad. Sci. Press. Gives a detailed treatment of the ­elasticity of rubber net-
USA 82, 6153–6157. This references contains excellent images of works.
the erythrocyte spectrin cytoskeleton, including fourfold and sixfold Satcher RL & Dewey CF (1996) Theoretical estimates of mechani-
connectivity. cal properties of the endothelial cell cytoskeleton. Biophys. J. 71,
Evans E & Yeung A (1989) Apparent viscosity and cortical tension of 109–118. Applies cellular solid theory to estimate the elastic moduli
blood granulocytes determined by micropipette aspiration. Biophys. of the cytoskelton. This article details some of the scaling analysis
J. 56, 151–160. Gives membrane tension estimates for use in filopo- and experimental support used in the cellular solids model.
dia buckling analysis. Shao EY & Hochmuth RM (1996) Micropipette suction for measuring
Kamm RD (2005) Molecular, Cellular, and Tissue Biomechanics (lec- piconewton forces of adhesion and tether formation from neutro-
ture notes from course number 20.310, Massachusetts Institute of phil membranes. Biophys. J. 71, 2892–2901. This article gives esti-
Technology, Cambridge, MA). Several of the treatments in this chap- mates of force required to form filopodium-like membrane tethers
ter were based on course notes developed by Dr. Kamm for this within neutrophils.
graduate course. Specific material includes the scaling analysis of Stamenović D & Coughlin MF (1999) The role of prestress and
cellular solids and several problems. architecture of the cytoskeleton and deformability of cytoskeletal
Kwon RY, Lew AJ & Jacobs CR (2008) A microstructurally informed filaments in mechanics of adherent cells: a quantitative analysis.
model for the mechanical response of three-dimensional actin net- J.  Theor. Biol. 201, 63–74. An excellent review of scaling analyses
work. Comput. Meth. Biomech. Biomed. Eng. 11, 407–418. Gives for cellular solids and tensegrity structures. The scaling relationship
details of the nonanisotropic affine network model, as well as an for tensegrity structures was obtained from this study.
extension of the model that accounts for some degree of non-affine Wang N, Naruse K, Stamenović D et al. (2001) Mechanical behav-
deformations. ior in living cells consistent with the tensegrity model. Proc. Natl.
Acad. Sci. U. S. A. 98, 7765–7770. This journal article details one
Mejillano MR, Kojima S, Applewhite DA et al. (2004) Lamellipodial
of the most compelling experiments that supports the tensegrity
versus filopodial mode of the actin nanomachinery: pivotal role
hypothesis.
of the filament barbed end. Cell 118, 363–373. This reference
assesses the role of various actin-binding proteins on the forma- Warren WE & Kraynik AM (1997) Linear elastic behavior of a low-
tion of lamellipodia and filopida and proposes a model whereby density kelvin foam with open cells. J. Appl. Mech. 64, 787–794. This
lamellipodial versus filopodial organization can be selected via journal article provides some very comprehensive mathematical
regulation of these proteins. analysis of cellular solids.
This page intentionally left blank
to match pagination of print book
CHAPTER 9
Mechanics of the Cell
Membrane

T he cell membrane is much more than a passive “bag” separating the cyto-
plasm from the environment. It is a heterogeneous, regulated barrier that
allows for both active and passive transport of substances between the inside and
outside of the cell. Its mechanical integrity is fundamental to its barrier function.
Bending and stretching of the membrane is centrally involved in ­exocytosis and
vesicle budding, as well as in fusion and viral invasion. It also incorporates struc-
tures for interacting with the extracellular matrix, other cells, and various com-
pounds in solution. Proteins trapped in the membrane allow for signaling between
the inside and outside of the cell. Understanding the biology and mechanics of
the membrane can foster an appreciation of the complexity of this fascinating cel-
lular component. In this chapter, we discuss the cell membrane’s structural
organization, how the two-dimensionality of the membrane limits diffusion, and
how the barrier function can be understood. We end with a formal treatment of
the cell membrane’s mechanical function.

9.1 MEMBRANE BIOLOGY

Water is a polar molecule
Cells exist in an aqueous environment, and understanding the structure and
function of the cell must include an understanding of how molecules, particularly
the lipid bilayer, behave in aqueous conditions. Water molecules have an electric
polarity, unusual for such a small molecule. The flanking hydrogens are not co-
linear, but form an oblique angle with the oxygen (the H−O−H angle is roughly
105 degrees). Because oxygen is more electronegative than hydrogen, this leads to Nota Bene
a net shift in the electric charge and the resulting molecule develops a polarity The surface tension of a fluid
based on this incomplete separation of charges (Figure 9.1). derives from intermolecular
attractions. Deep within a fluid, a
Water molecules can form hydrogen bonds among themselves by aligning the given molecule is attracted from all
charges between different molecules. This gives rise to many of water’s unique sides by these forces. However, at the
properties, including its high surface tension, the decrease in density when water surface, the molecule is only
freezes, and the hexagonal symmetry of snowflakes. It is important to note that attracted from molecules on one
these bonds are not permanent; water tends to rearrange itself all the time; how- “side,” resulting in the molecule being
ever, at any given time, water molecule interactions are heavily influenced by pulled toward the interior. Therefore,
most fluids will form spherical
hydrogen bonding. Indeed, as water freezes, the molecules are less free to re-form
shapes, small droplets that minimize
associations and, as a result, the molecules can no longer “fit” as well together the exposed surface. We will see in
(they tend to form fixed hexagonal crystals, which is why snowflakes tend to be Section 10.1 that it takes energy to
six-sided). The net result is that water expands as it freezes, and the resulting create new surfaces, and as a result,
decreased density is enough to make ice float on liquid water. The proclivity for two surfaces have an adhesion
the formation of hydrogen bonds allows a remarkably wide range of substances to energy (the energy required to
be dissolved in water. Water is referred to as the universal solvent because of this separate the surfaces) that is
identical to the surface tension.
property and the ubiquitous presence of this molecule on Earth.
250 CHAPtER 9: Mechanics of the Cell Membrane

Figure 9.1 A water molecule (A) (B)

represented in the typical “ball and
stick” model. (A) The oxygen atom has
two pairs of electrons that occupy the
“upper region” (not shown). The two
pairs of electrons and the two hydrogen
atoms are located roughly at the corners
of a tetrahedron. (B) The sizes of the
actual water molecule components. The
“bending” of the water molecule gives it
a polarity, with the negative charge
closer to the oxygen atom. (From,
Alberts B, Johnson A, Lewis J et al. (2008)
Molecular Biology of the Cell, 5th ed.
Garland Science.)

Cellular membranes form by interacting with water

Substances that are easily dissolved in water are known as hydrophilic (water lov-
ing). Those that repel or are repelled by water are known as hydrophobic (water
fearing). Hydrophobic molecules tend not to be electrically polarized and thus do
not readily form hydrogen bonds. When placed in an aqueous environment they
tend to cluster or form bonds with themselves, as a result of being excluded by the
surrounding water. In contrast, hydrophilic molecules tend to be polar, charged or
easily dissociated into charged subunits. Complex molecules may exhibit regions
or domains that are hydrophobic, while other regions are hydrophilic. These mol-
ecules are known as amphiphilic. This property gives rise to much of the diverse
biochemical behavior that is at the root of the function of biological membranes
and fundamental to cell biology.
Hydrophobic domains tend to be long hydrocarbon regions made up of repeating
CH2 units. Hydrophilic domains contain charged or polar regions. Charged
regions can be anionic (negatively charged) like phosphates or sulfates, or they
may be cationic (positively charged) like amines. Polar groups can be side chains
such as alcohols and alkyls. Many biological membranes, including the cell mem-
brane, are made up of phospholipids with a hydrophilic head region and a hydro-
phobic tail region. In an aqueous environment, the phospholipids self-assemble
in structures to “isolate” the tail regions from water while keeping head regions in
Nota Bene aqueous contact. As the concentration of phospholipids is increased, they begin
Micelle instability. When the lipid
to self-assemble into complex structures to protect the tail regions. The simplest of
concentration rises to the point that these, known as a micelle, is ball shaped, with the tails facing inward and away
lipids begin to encounter one from the surrounding water. Micelles can also take on ellipsoidal or even cylindri-
another, they will initially fuse to form cal shapes. As concentration rises, the micelle geometry eventually becomes
larger micelles. Eventually the unstable, and a layered structure is formed with the head regions facing outward.
water-excluding hydrophobic core This can take the form of a bilayer sheet or a small volume of water (and perhaps
becomes too large to be supported,
other trapped substances dissolved in the water) may be sequestered, forming a
and a bilayer results.
liposome (Figure 9.2).

Figure 9.2 Lipid configurations micelle bilayer

depend on concentration. As their
concentration increases, lipids begin to
associate into spherical micelles, lipid
thereby hiding their hydrophobic tails
from the surrounding water. At even
hydrophilic
higher concentrations, a bilayer forms. hydrophobic
 MEMBRANE BIOLOGY 251

The saturation of the lipid tails determines some properties

of the membrane
The phospholipid’s hydrophobic (nonpolar) region is a tail consisting of two long
carbon chains (Figure 9.3). Typically, one of these is saturated, meaning that each
carbon is associated with two hydrogens and two carbons (one on either side).
The other chain can consist of a few unsaturated carbons, meaning they form
double bonds with adjacent carbons and can only accept a single hydrogen per
carbon atom. Saturation leads to a more viscous behavior and a higher degree of
self-association (saturated fats, like butter, tend to be solid at room temperature,
whereas unsaturated fats, like vegetable oils, tend to be liquid). So it is possible to
vary some of the fluid properties by controlling the amount of saturation of the
lipids. In the same way that margarine is partially hydrogenated, or partially satu-
rated, vegetable oils can be modified to become solid at higher temperatures
compared.
From a mechanics point of view, the reason saturation is important has to do with
steric, or spatial, properties. A fully saturated hydrocarbon chain will have some
“zig-zag” in its length, but will be roughly straight (that is, the amplitude of the zig-
zag is small compared with its overall length). Additionally, each bond is free to
rotate, allowing for increased packing. The introduction of a double carbon bond
creates a large bend in the chain. Double bonds also do not allow rotation, thereby
further inhibiting packing. The result is that unsaturated phospholipids are not as
tightly packed, diminishing the strength of intermolecular interactions and result-
ing in lower viscosity and a lower freezing point.

The cell membrane distinguishes inside and outside

The cell membrane is a lipid bilayer, also known as the plasma membrane,
physically separating the extracellular biochemical environment from the cell’s

CH2 N+(CH3)3 Figure 9.3 The lipids in a cell

CH2 membrane consist of two fatty acid
O chains attached to a polar head
O P O
_ group. Typically, one tail is fully
saturated and the other consists of a
O
varying number of carbon double bonds
CH2 CH CH2 hydrophilic head (making it unsaturated). The properties
O O of the unsaturated tail are important for
C O C O hydrophobic tails determining the membrane fluidity,
CH2 CH2 partly because of the bend in the tail
CH2 CH2 (exaggerated in this depiction). (From,
CH2 CH2 Alberts, Johnson A, Lewis J et al. (2008)
CH2 CH2 Molecular Biology of the Cell, 5th ed.
CH2 CH2
Garland Science.)
CH2 CH2
CH2 CH2
CH2 CH cis double bond
CH2 CH
CH2 CH2
CH2
CH2
CH2
CH2
CH2
CH2 CH2
CH2 CH2
CH2 CH2
CH2 CH3

CH3
252 CHAPtER 9: Mechanics of the Cell Membrane

internal fluid volume, known as the cytoplasm. However, the plasma mem-
brane is not a total barrier, and certain molecules can diffuse across it. This
semipermeable behavior is critical for its proper function. The barrier proper-
ties of the plasma membrane are partly a result of the thin hydrophobic layer in
the interior tail region of the bilayer. Charged or polar molecules are less likely
to cross this region. The plasma membrane also forms a barrier against large
molecules. The barrier function of the plasma membrane is further increased
by the activity of flippases, which are proteins that concentrate negatively
charged phosphatidyl serine to the inner leaflet, forming an additional charged
boundary. Permeability of the cell membrane to a given substance is hard to
predict, but is generally a function of size, charge, and solubility. Water can
cross the plasma membrane and does so by osmosis, the relatively slow transfer
from the low solute concentration side (hypotonic) to the high solute concen-
tration side (hypertonic). In general the cell’s cytoplasm is hypertonic. Water
movement across the plasma membrane is also facilitated by aquaporins, small
protein complexes that form tiny holes in the membrane to regulate the pas-
sage of water molecules.

The fluid mosaic model of the cell membrane describes

its physical properties
The molecular arrangement of the bilayer is a result of the attraction of the phos-
pholipid head groups to water and the repulsion of the tail groups. This also pro-
Nota Bene duces an effective mutual attraction between the tail groups. Although an
The mechanics of the fluid individual phospholipid can move about laterally within the bilayer, there is
mosaic model. In the sections that great resistance to its moving outside of the bilayer, which would expose the tails
follow, we will formulate a mechanics to water. In terms of the mechanical properties, this results in very low shear
description of the in-plane shear and ­stiffness, but high resistance to areal expansion. Indeed, the shear stiffness of
areal behavior of the membrane. This the membrane is so low; it essentially behaves as a fluid within the membrane
will allow us to quantify just how plane. This view of the membrane as a fluid constrained to two dimensions was
“fluid” the membrane is.
advanced by Singer and Nicolson in the 1970s and is termed the fluid mosaic
model.
The practical implication of this model is that within the membrane itself, mol-
ecules such as proteins freely diffuse in the two-dimensional plane. This greatly
simplifies and accelerates protein–protein interactions. There are also com-
partmentalizations in the membrane that can trap proteins in a “bin” or micro-
domain. Occasionally, the protein obtains sufficient energy (from thermal
fluctuations) to break through a barrier and enter another compartment. This
phenomenon was only observed with the development of small-molecule
tracking techniques and fast-imaging cameras.

9.2 PHOSPHOLIPID SELF-ASSEMBLY

With our introduction to membrane biology in hand, the remainder of this chap-
ter will be more quantitative in nature and will focus on membrane formation,
mechanics, and function. We previously learned that when amphiphilic mole-
cules such as phospholipids are placed into an aqueous environment, above a
critical concentration, they self-assemble into organized structures. The concen-
tration at which this aggregation occurs is called the critical micelle concentration
(CMC). Experimentally, this concentration has been found to be highly sensitive
to the molecular structure of the amphiphiles. The resulting shape of the self-
assembled cluster (that is the spherical micelle or bilayer sheet) has also been
found to be dependent on phospholipid molecular structure. In this section, we
investigate the dependency of CMC and aggregate shape on various molecular
attributes of phospholipids.
 PHOSPHOLIPID SELF-ASSEMBLY 253

Critical micelle concentration depends on amphiphile

molecular structure
Experiments have demonstrated that the CMC is significantly decreased when
the hydrocarbon tails are extended by adding more carbon atoms, or when the
molecules are synthesized with two tails instead of one. Why would this be? One
approach to understanding this phenomenon is to use a simplified model with
only two idealized phases; a condensed phase in which all the amphiphiles are
part of a single aggregate surrounded by water molecules and a surrounding
aqueous phase. In the latter, the individual amphiphiles are dispersed through-
out the solution in a dilute manner. We are able to approximate their behavior as
that of an ideal gas. The advantage of this idealized treatment is its simplicity,
however, some accuracy is sacrificed. In reality, the system will not be con-
strained to exist in these two states. There will be a spectrum of intermediary
states with some molecules aggregated and others dispersed throughout the
solution.
We start by considering a condensed state consisting of a spherical micelle of one-
tailed phospholipids with hydrocarbon chain length l immersed in an aqueous
solution. The chain length is equal to the product of the number of carbon atoms
in the chain nc and the average bond length between carbon atoms lc, l = lcnc. We
consider the condensed state to be the reference state, with zero free energy (zero
free energy only occurs with zero internal energy [W ] and zero entropy [S]). Next,
we calculate the free energy of the aqueous state. In this state, the hydrophobic
tails of the molecules are now exposed to water, and there is an increase in energy
per molecule that can be approximated as the interfacial energy between the
water and the exposed hydrophobic hydrocarbon region. Let γint be the interfacial
energy per unit hydrocarbon chain length (this energy/length can be roughly
approximated from measurements of surface tension and an estimation of the
effective radius of the hydrophobic region). Then for each molecule, this energy is
equal to

W = γ intl = γ intlcnc . (9.1)

To calculate the entropy we assume that in this state, the amphiphiles are suffi-
ciently dispersed to behave like an ideal gas. The entropy per molecule for an ideal
gas is

5  ρ 3 
Sideal = kB  − ln  3/ 2  
, (9.2)
2  (2πmkBT )  

where ρ is the number of molecules per unit volume, m is the mass of each mol-
ecule, and  is Planck’s constant. Now, we can write the free energy per molecule
in the aqueous state as

5  ρ3 
Ψ = W − TSideal = γ intlcnc − kBT  − ln  3/ 2  
. (9.3)
2  (2πmkBT )  

Remember that this free energy in the aqueous state was constructed with the
condensed state defined as the reference state with zero free energy. So, the tran-
sition between the condensed and aqueous states occurs when Ψ = 0, or when

5  ρ 3 
γ intlcnc = kBT  − ln  3/ 2  
. (9.4)
2  (2πmkBT )  

254 CHAPtER 9: Mechanics of the Cell Membrane

The density ρ at this transition gives the critical density of molecules at which
aggregation occurs. Solving for ρ, we obtain
 γ int lc nc 
 5/ 2 − k T  (9.5)
ρ = A e B
,

where

A = (2πmkBT )3/2  −3 . (9.6)

Although the mass m depends on nc, for physiological values of γint, lc, and nc, the
exponential term typically decreases much faster than A with nc. Therefore, we can
see that for constant values of γint and lc, increasing the number of carbon atoms nc
decreases the aggregation density. Intuitively, this seems reasonable, as having
longer tails would increase the interfacial energy of the molecule with water and
make it more energetically unfavorable to remain in the aqueous state. Similarly,
phospholipids with two tails have a lower aggregation density compared with their
single-tailed counterparts, owing to their higher interfacial energy. What is per-
haps not as intuitive is the profound impact that increasing nc can have on lower-
ing ρ. The predicted exponential scaling of ρ with nc, has, in fact, been observed
experimentally. Specifically, increasing nc by x atoms decreases ρ by a factor of ex.
If we were to extend the hydrocarbon chain by just two atoms (let us say from 10 to
12), this would result in an almost tenfold decrease in the aggregation density.

Example 9.1: Lipid packing as a function of tail length

Consider a lipid having a carbon tail of length nc = 8 mol- In addition, we know that the fold change in the expo-
ecules. Calculate the fold change in the critical micelle nential term is
density if the length of tail was increased to nc = 16 mole-
 8γ l 
cules. Assume lc = 0.1 nm, γint = 10 kBT/nm, and that the 
5/ 2 − int c 
k T 
mass of the 8-carbon tail lipid is 200 g/mol. e B

 16γ int lc 
= e8 ≈ 3000
5/ 2 −
 k T 
First, we must calculate the mass of the longer lipid. We e B

know that the molecular mass of carbon is 12 g/mol.

Because the longer lipid contains 8 more carbons, the We see that increasing the length of the lipid tail by 8 car-
mass of the longer lipid is 200  + (8 × 12) ≃ 300 bons will result in an approximately 3000 × 0.5 = 1500-
g/mol. Using Equation 9.6, we know that the fold change fold decrease in the critical micelle density.
in A for the two different lipids is (2003/2)/(3003/2) ≃ 0.5.

Aggregate shape can be understood from packing constraints

In the previous section, we investigated the dependency of CMC on amphiphile
molecular structure. In our treatment, we did not consider the shape of the clus-
ter. The tendency for certain amphiphiles to form certain geometries can, in prin-
ciple, be investigated through free energy calculations; however, such calculations
are very difficult to formulate. Nonetheless, by examining constraints associated
with molecular packing, we can gain insight into the tendency for particular
shapes to be preferred.
Consider a spherical micelle of radius R, each with effective head group surface
area Ah. In addition, assume that the effective hydrocarbon chain volume is Vc.
Because the interior of the micelle is hydrophobic, it cannot contain voids that
would potentially enclose water. Therefore, the entire volume of the micelle is
occupied by the hydrocarbon chains. The number of molecules n in the micelle
can be calculated as either

4πR 2 (9.7)
n=
Ah 
 MEMBRANE BARRIER FUNCTION 255

from the surface area, or

4πR 3
n= (9.8)
3V 

from the volume. Setting Equations 9.7 and 9.8 equal, we see that for a micelle
with a given radius R, Vc and Ah must be related as

R = 3Vc /Ah . (9.9)

Now, let l = lcnc be the hydrocarbon chain length. We assume that the interior of
the micelle cannot contain voids, owing to the enormous energy that creation of a
vacuum would require. This implies that the radius of the micelle must be less
than or equal to the hydrocarbon chain length, R ≤ l. Equation 9.9 becomes

Ah /Ae ≥ 3, (9.10)

where Ae = Vc/l is the effective cross-sectional area of the hydrophobic region.

Equation 9.10 implies that, based on packing constraints, a spherical micelle
requires the effective area of the head group to be more than threefold greater
than the hydrophobic region.
For phospholipids with progressively smaller Ah/Ae ratios, the preferred micelle
shape would transition from a spherical micelle to a bilayer as this ratio approaches
unity. Consider a large flat bilayer with width w, height h, and thickness t. In this
case, packing requires the number of amphiphiles to be

n = wh/Ah (9.11)

based on surface area or

n = wht /2Vc (9.12)

based on volume. Setting Equations 9.11 and 9.12 equal we find

Ah = 2Vc /t . (9.13)

The thickness must be less than or equal to twice the hydrocarbon chain length, or

t ≤ 2l . (9.14)

In this case,

Vc /l = A0 ≤ Ah . (9.15)

This indicates that when the effective area of the hydrophobic region approaches
that of the head region, based on packing, a bilayer shape is preferred.

9.3 MEMBRANE BARRIER FUNCTION

We now turn our attention from investigation of membrane formation to that of
membrane function. We previously learned that one of the primary roles of the
plasma membrane is to separate the extracellular biochemical environment from
the cytoplasm. However, it is only a semipermeable barrier, selectively allowing
certain molecules to diffuse across it. How can we quantitatively assess the barrier
properties of the membrane?
In this section, we analyze membrane barrier function using an approach
based on the diffusion equation, which was derived using the random walk in
256 CHAPtER 9: Mechanics of the Cell Membrane

Section 5.6. The connection between random molecular processes (the ran-
dom walk) and the net macroscopic or continuum-level behavior (the diffusion
equation) was one of Albert Einstein’s many important contributions. Owing to
entropy and Brownian motion, there are many applications of random walks in
biology. In this section we are going to apply the continuum level description
to understand how molecular transport works at the level of the cell and the
role of the membrane in that process. We begin by revisiting Fick’s first and
second laws of diffusion.

The diffusion equations relate concentration to flux

Nota Bene per unit area
Sinks are not only abstractions. What does diffusion imply about the number of particles that might pass through
The use of a cell as a sink is not just a boundary such as the cell membrane? Using the continuum-level description
academic. Many dyes used in cell the flux (from the Latin word fluxus meaning flow). J is given by
biological experimentation have
modifications that allow the dye to ∂C
cross the cell membrane, but are J = −D . (9.16)
modified by enzymes within the cell ∂x
so they cannot escape. One such
modification is the acetoxymethyl This is Fick’s first diffusion equation. It can be generalized to higher dimensions as
ester (AM ester). This modification is
hydrophobic and allows the
J = − D∇C , (9.17)
fluorescent molecule (if small enough)
to enter the cell. Once inside the cell,
native esterases cleave this group off where ∇ is the gradient operator defined as
which results in the fluorescent
molecule becoming charged and  ∂( • ) ∂( • ) ∂( • ) 
∇( • ) =  , , .
trapped within the cell.  ∂x ∂y ∂z 

Advanced Material: Flux and the random walk

Fick’s first law can also be obtained from analysis of the

J =
( N ( x ) − N ( x + d )) ,
flux in a discrete random walk. In one dimension, let us 2 A f ∆t
say we have N(x) particles at location x and N(x + d) par-
ticles at location x + d. To find the flux between these two
locations, consider that in a time step Δt, half of the par- where Af is the area through which the flux is moving. We
ticles at x will move rightward (crossing the boundary) then perform some simple manipulations,
and half of the particles at n + d will move leftward
(crossing the boundary in the other direction). If we are d 2 1  N (x + d) − N (x) 
J =−  .
interested in the number of particles moving to the right 2∆t d  Af d
(in the positive x-direction) we get number of particles
moving right per time step = 0.5(N(x) − N(x + d)). The difference inside the square brackets is analogous to a
To find the flux (which is a rate), we need to divide by the concentration gradient. If we take the limit as d approaches
time step Δt as well as some area that the particles are zero, we find Fick’s first law (see Equation 9.17), with the
crossing. This yields diffusion constant D being d2/2Δt.

Example 9.2: Discrete diffusion example: capture time

Consider particles that are freely diffusing until they Imagine a molecule that is freely diffusing along the
meet a sink, wherein they become captured. How long x-axis, but with a sink at a particular location. Once the
would we need to wait for a diffusing particle to reach a particle reaches that location, it is “captured.” We want
given sink from some specified starting location? to know how long we would wait before the molecule is
 MEMBRANE BARRIER FUNCTION 257

captured. At a position x, let the mean time of capture be The above equation describes the wait time for a single
Tc(x). Let us start with the discrete random walk and particle. To generalize to higher dimensions, you can
then extend it to a differential form. Assume that I wait a replace the second term (d2Tc/dx2) with the Laplacian of
small time Δt, the molecule will be at x + b or x − b with appropriate dimension. To solve this equation one needs
equal probability. We can write this recursive relationship boundary conditions. For this example, let us assume
that the absorber is a sphere of radius a located within a
Tc ( x ) = ∆t + 0.5(Tc ( x + b) + Tc ( x − b)).
larger impermeable sphere of radius b (a  b). On the
surface of the absorber, the wait time is obviously zero
Rearranging, (Tc(a) = 0). At the second boundary, Tc has zero gradi-
ent. We leave it as an exercise to show that integrating
over the known volume the mean wait time is
0 = 2∆t + (Tc ( x + b) − Tc ( x )) − (Tc ( x ) − Tc ( x − b))
b3
Tc = .
3Da
2∆t 1 1
0= + (Tc ( x + b) − Tc ( x )) − (Tc ( x ) − Tc ( x − b)).
b b b The mean capture time is
Letting b approach zero, we obtain b2  b 
Tc = ln
2D  a 
2∆t 1  dTc ( x ) dTc ( x − b) 
0= +  −  .
b2 b  dx dx for a particle in a two-dimensional surface with a small
circular absorber of radius a. Therefore, as the distance
Letting b approach zero once more increases, diffusion limits on kinetics increases much
more slowly if the reaction is confined to a membrane
1 d 2Tc
0= + . than if it needed to occur in free space.
D dx 2

Fick’s second law shows how spatial concentration changes

as a function of time
In Section 5.6 we considered a collection of particles to be moving independently,
in discrete time and space. We showed that if each moved either to the right or left Nota Bene
by a distance b at each point in time, we obtained the one-dimensional diffusion Capture time at an impermeable
equation (or Fick’s second law), boundary. It may seem strange that
the assumption of an impermeable
dC d 2C boundary results in a condition on
=D 2, (9.18) wait time gradient. One way to see
dt dx this is to consider a repeating lattice
of absorbers. At the midpoint
where C is the concentration and D is the diffusion coefficient. We can relate D to between two absorbers, there will be
the molecular behavior of the particles and specifically that no net flow, because particles are
equally likely to end up at either one.
nb 2 So, this is effectively an impermeable
D= , (9.19)
boundary. Also, the wait time at the
2∆t
midpoint reaches a maximum
where Δt is the time step in which a particle moves a distance b. because moving in either direction
will shorten the distance to one of
We can also generalize this to higher dimensions by writing the absorbers and reduce the wait
time. Therefore, the gradient is zero.
∂C
= D∇ 2C , (9.20)
∂t
where ∇2 is the Laplacian of C defined as

∂ 2C ∂ 2C ∂ 2C
∇ 2C = + + . (9.21)
∂x 2 ∂y 2 ∂ z 2
258 CHAPtER 9: Mechanics of the Cell Membrane

Fick’s second law tells us how spatial concentration changes as a function of

time. In essence, there is a net transport of particles from high concentration to
low concentration. The rate of transport depends on the diffusion coefficient in
addition to the concentration distribution. This transport is referred to as a flux
(from the Latin word fluxus, meaning flow) and is usually given in terms of amount
of flow per unit area.

Example 9.3: Continuous diffusion example: total flux

With Fick’s first and second laws revisited, we can now To model the sink we assume that the concentration
begin to quantitatively assess diffusive behavior of a par- of  particles at the surface is zero. This provides our
ticular molecule through a membrane. For example, one boundary conditions, C(R = a) = 0 and C(R → ∞) = C0.
could quantify the current, or integrated flux through a Fick’s first diffusion law (Equation 9.17) in spherical
region. We start with the assumption that all of the space coordinates is
outside a given sink is filled with a collection of particles.
How many particles per unit time are crossing the sink ∂C
boundary when the system reaches steady state? This is J = −D ,
∂R
analogous to the situation in which a cell with a mem-
brane permeable to some substance is incubated in a
large volume of a solution of that substance. We wish to which has a solution of the form
know how much of the molecule is crossing the cell
boundary, assuming that once a cell consumes the com- a
pound it is metabolized. J ( R) = − DC 0 .
R2
For simplicity, we will assume that the cell is a sphere
of  radius a. If we are at steady state, then we can
The net “current” of molecules passing through the
apply  Fick’s second diffusion equation, with the time
sphere is then I = JA, where A is the area of the boundary.
derivative set to zero. In spherical coordinates, this
At R = a,
becomes

1 ∂  2 ∂C  a
R  = 0. I = − DC 0 4πR 2 = −4πDC 0a.
R 2 ∂R  ∂R  R2

Example 9.4: Fluorescence recovery after photobleaching and molecular mobility

Diffusion can be used to assist in biological microscopy attached to a protein with high mobility, though, pro-
experiments to assess the mobility of molecules within teins attached to unbleached molecules will diffuse in,
cells. Because it is difficult to see individual molecules, and the dark region will recover and become brighter
there is no simple way to determine whether, say, a par- again. Alternatively, if the protein is well-anchored (low
ticular receptor is free to move around the membrane or mobility), then the dark spot will persist. This can  be
is anchored in place. quantified, based on the work of Axelrod and Webb.
The assumptions for this approach are that the motion
One type of experiment that can provide some informa-
of molecules is purely diffusive (there is no convec-
tion is fluorescence recovery after photobleaching
tion), and that the photobleached region is disk shaped.
(FRAP). In this technique, the target protein is tagged
with a fluorescent molecule. After continuously excit- As a function of radius, Fick’s second law is
ing a small region, the fluorescent molecule will
become “bleached”, that is, lose its fluorescent proper-
ties. This region becomes darker compared with the ∂C ( R , t )
= D∇ 2C ,
surrounding regions. If the fluorescent molecule is ∂t
 MEMBRANE MECHANICS I: IN-PLANE SHEAR AND TENSION 259

with boundary conditions C(∞, t) = C∞ and initial condi- has a rough idea what D is, one can design an experi-
tions C(R,0) = C0(R) the initial photobleach profile. ment to time the FRAP.
Typically,
Let us work a numerical example. Assume that αP0C0 = 1
C ( R , 0) = C 0e −T ϕ ( R ) (fluorescence units), D = 10−11 m2/s, and Tφ(0) = 0.1.
Determine the half-life of the photobleached region
relative to F(0), if the photobleaching region is a disk of
where T is the time interval for the bleaching and φ(R) radius 5 μm.
is a scaled excitation intensity from the laser (φ = 0 for
R > w, where w is the disc radius, and φ = constant for Because Tφ(0)  1, we can apply the solution for F(t),
R ≤ w. The constant depends on the laser power, a which gives us
scaling factor, and the characteristic size of the laser
beam.)  1 
1−
F (t ) =   4tD   .
However, what we really want is the diffusion constant  20  1 + 2  
D, and what we really measure is the fluorescence   w 
profile
Substituting the provided numbers,

∫
F (t ) = αϕ ( R)C( R ,t )dA , 
F (t ) =  1 −
1 
,
 (20 + 32t ) 
where α is a scaling factor that further corrects for
imaging attenuation and the integral is over the pho- 1 19
with t in seconds. We can quickly get F(0) = 1 − =
tobleached area. The solution for F(t) can be found 20 20
with Fourier transforms and series analysis. If one (this is not much photobleaching at all). To get to half of
uses  the assumption that Tφ(0) is much less than this value, we solve
unity  (slight bleaching) then the solution can be
expressed as 39  1 
= 1−
40  (20 + 32t ) 
 Tϕ (0) 
1−
F (t ) = (α P0C 0 )   t , for t. One gets t = 5/8 s, or a little over half a second. This
 21 +   is very fast and unlikely to be something that would be
  τd  
useful using FRAP as modeled. One can either increase
the intensity of illumination to photobleach out the sam-
where P0 is the laser power and τd = w2/4D is the charac- ple more (in which case the expression no longer holds,
teristic time for diffusion. By measuring F(t), one can and one needs to use a different approach) or use a
obtain D for the protein of interest. Alternatively, if one smaller photobleaching area.

9.4 MEMBRANE MECHANICS I: IN-PLANE SHEAR

AND TENSION
We have investigated various aspects of membrane formation (amphiphile self-
(A) (B)
assembly) and function (acting as a barrier). The rest of this chapter focuses on
membrane mechanics. As we have discussed, for energetic reasons, each lipid
bilayer has an inherent optimal microstructure with some optimal spacing
between the lipid molecules. Any perturbation to this optimal arrangement dis-
turbs this energetically favorable microstructure. The lipid bilayer exhibits an
inherent resistance to deformation. If a portion of membrane is stretched, the
lipid molecules are pulled apart, requiring energy in the presence of water. If the
membrane is bent, the molecules on the outer layer are separated and those on
the inner layer are compressed, again storing deformational (strain) energy Figure 9.4 A region of membrane
(Figure 9.4). However, as we will see, the lipid molecules are relatively free to exposed to either tension (A) or
move about within the bilayer. bending (B).
260 CHAPtER 9: Mechanics of the Cell Membrane

Nota Bene With this as our motivation, the goal of this section is to develop the equations
that describe the mechanics of the continuum behavior of the cell membrane. As
Differential geometry is needed
this section is mathematics-heavy, we will consider systems of increasing levels of
to express the mathematics of
shells. The reason we restrict most of complexity to develop some degree of intuitive understanding before presenting
our analyses to plates is that in a the general governing equations.
“full-blown” analysis, equations must
be developed in a curved system,
rather than the “ordinary” planar
Thin structures such as membranes can be treated as
Euclidian coordinate systems with plates or shells
which you are probably familiar. The
mathematical study of curved The defining characteristic of a membrane is its very thin structure relative to
coordinate systems is known as the overall size of the cell. The cell membrane lipid bilayer has a thickness of
differential geometry. It is remarkably approximately 7 nm. The cell itself is typically on the order of micrometers. By
similar in its use in shell mechanics restricting the dimensionality of the problem, certain kinematic assumptions
and general relativity. At its heart is the can be made that will greatly simplify the problem from that of general three-
definition of the covariant derivative,
which is the derivative along the
dimensional continuum mechanics. In other words, we can approximate the
curved tangents of the shell’s surface. membrane as a curved two-dimensional structure, or, from Chapter 3, a shell.
It is defined in terms of Christoffel You might be more familiar with a plate, or a flat two-dimensional structure. In
symbols. They represent an additional this section, for simplicity, we will assume that the domain or subdomain of the
term in the derivative that occurs problem is flat (that is, a plate). One exception is that in certain cases we will
because of the curved coordinate consider spherical shapes (shapes that can be described by radii in the x- and
space and greatly simplify complex
y-directions).
calculations. However, they are quite
beyond the scope of this treatment.

Advanced Material: Einstein and curved coordinate systems

The mathematics to describe behavior in curved coor- coordinates, and when they occur in pairs, a summa-
dinate systems was developed to a great extent by tion is implied. The square of the magnitude of a vector
Einstein. He needed this description to cast the formu- v would be denoted as vivi = v1v1 + v2v2 + v3v3  = vxvx + 
lation of general relativity theory in which space is vyvy + vzvz. This is a very powerful and compact way to
actually curved owing to the gravitational effect of represent complex vector and tensor manipulations.
mass. He developed a notational system (known as Any advanced treatment of continuum mechanics
inidical notation) in which coordinates are denoted beyond the level used in this text will generally adopt
with numbers (1, 2, or 3) rather than letters (x, y, and z). this notation out of expediency.
In his system, indices are introduced to represent the

As before, our continuum mechanics analysis involves three different parts, kin-
ematics, constitutive equations, and equilibrium. For the material model we are
again going to assume generalized Hookean behavior. We will consider equilib-
rium in distinct parts, one for the in-plane forces and one for the out-of-plane
forces. Let us begin our discussion of membrane mechanics by making kinematic
assumptions that take advantage of the thin shell assumption.

Kinematic assumptions help describe deformations

Similar to our approach to beams, we introduce a kinematic construct to describe
the deformation. In the case of the beam, we did this in terms of planar collections
of points perpendicular to the neutral axis. For shells, we are going to do this in
terms of linear collections of points through the shell perpendicular to the shell’s
midline surface. Our kinematic assumptions are that these lines
1. remain straight,
2. do not stretch, and
3. remain normal to the midline surface.
 MEMBRANE MECHANICS I: IN-PLANE SHEAR AND TENSION 261

These kinematic assumptions suggest a particular way of expressing the deforma- Nota Bene
tion of the shell. In particular, it makes sense to talk about the deformation due to
The Kirchhoff hypothesis. The
motion in the plane of the shell (we assume this is the x–y-plane) and motion due to
shell kinematic assumptions are also
transverse displacements and the subsequent rotation of the normals (Figure 9.5). known as the Kirchhoff hypothesis.
This is analogous to our separation of axial deformation and bending deformation The lines perpendicular to the shell
when we considered a linear element. We will denote the total deformation as utot surface are known as directors.
and vtot in the x- and y-directions, respectively. They are simply the sum of the in-
plane deformation and the displacement due to rotation of the shell’s mid-surface.
Our kinematic assumption amounts to just assuming that the deformations take a
particular form that is a simple extension of the beam kinematics,
dw
u tot ( x , y , z ) = u( x , y ) − z
dx
dw (9.22)
ν tot ( x , y , z ) = ν ( x , y ) − z
dy

w tot ( x , y , z ) = w( x , y ).

From these, we calculate strain from Equations 3.48, 3.49, and 3.52

du tot du d 2w 1  du tot dν tot  1  du dν  d 2w

ε xx = = − z 2 ε xy =  +  =  +  −z
dx dx dx 2  dy dx  2  dy dx  dxdy

dν tot dν d 2w 1  du tot dw tot  1  du dw d 2w d 2w 

ε yy = = − z 2 ε xz =  +  =  − −z + 
dy dy dy 2  dz dx  2  dz dx dxdz dx 2 

dw tot 1  dν tot dw tot  1  dν dw d 2w d 2w 

ε zz = =0 ε yz =  +  =  − −z + .
dz 2  dz dy  2  dz dy dydz dy 2 
(9.23)
At this point we have used two of our three kinematic assumptions. Namely, that the
perpendiculars remain straight and that they are normal to the surface. However, we
did not use the assumption that they do not stretch. This condition implies that the
thickness of the shell does not change, and as a result, any strain terms involving z
must be zero. We already know that εzz =  0, but this condition implies that εzx =  εzy = 0
as well. The final form of our kinematics or strain-displacement relation is

du d 2w 1  du dν  d 2w
ε xx = − z 2 ε xy =  + − z
dx dx 2  dy dx  dxdy

dν d 2w (9.24)
ε yy = − z 2 ε xz = 0
dy dy
ε zz = 0 ε yz = 0.

Figure 9.5 The kinematic

assumptions of shell mechanics.
Transverse displacement within the
plane causes the perpendicular lines to
move but remain parallel to each other.
Bending causes them to rotate.

undeformed transverse displaced rotation

262 CHAPtER 9: Mechanics of the Cell Membrane

Advanced Material: Small rotation assumption

In this development we have assumed that the defor- not  be a valid assumption, particularly for cells. One
mations are small enough that our infinitesimal strain more advanced approach, known as von Kármán shell
measures are adequate. However, what may not be theory, assumes infinitesimal strains, but moderate
obvious is that we have also assumed that the rotation rotations. This introduces an extra quadratic terms in
of the normals is also infinitesimal. Often, this may the strains.

εip εθ Notice that the in-plane strain components (εxx, εyy, εxy) contain two parts, one
constant through the thickness, and one that varies linearly through the thickness.
These correspond to the strains due to in-plane motion of the normal lines, εip, and
strains due to rotations of the normal lines, εq (Figure 9.6). The in-plane strains are
edge +
related to tension and shear within the plane of the shell. The linear strains are
related to out-of-plane bending. Therefore, shells have three different deforma-
tional modes, in-plane tension (or compression), in-plane shear, and bending. The
total deformation is simply the linear superposition or sum of these three modes.
Figure 9.6 Decompositon of
deformation modes. As a result of our
kinematic assumption, the deformation A constitutive model describes material behavior
of a shell can be separated into a
component that is constant through the As in the past, we are going to assume a generalized Hookean material response
thickness, εip, and a component that (Equation 3.58). Owing to the simplified kinematics, it takes on a particularly sim-
varies linearly through the thickness, εθ. ple form:
σ xx = 2 µε xx + λ (ε xx + ε yy )

σ yy = 2 µε yy + λ (ε xx + ε yy ) (9.25)
τ xy = 2 µε xy .

We have characterized our kinematic behavior, and given u, v, and w, we can

determine strain, and through Hooke’s law, stress. What remains is to determine
what implications equilibrium holds for stress.

Nota Bene The equilibrium condition simplifies for in-plane tension and shear
The two meanings of the term
“membrane.” As we briefly When we considered kinematics, we talked about the deformation modes.
mentioned at the end of Section 3.4, Similarly, we can decompose our treatment of equilibrium by considering the dif-
in structural mechanics terms, a ferent loading modes in turn. Specifically, it is helpful to consider in-plane load-
curved two-dimensional structure ing separately from bending. Starting with in-plane loading, this mode is restricted
that can support bending moments to be within the plane of the shell locally. Although the shell may bend, we assume
is called a shell. A curved two-
that there is no energy, stress, or moment associated with this bending. This is
dimensional structure that cannot
withstand bending is called a strange. What does it mean to say we will allow bending, but ignore bending
membrane. Unfortunately, this is a stresses? Actually, this is not as unusual as you might think. Consider a thin plastic
very different use of the word from bag like you might get from the grocery store. The plastic in these bags is so thin
its biological meaning. So be alert to (about 30 μm) that they bend with the slightest touch. Although there is bending,
the context in which these terms are the resultant moments are so small that they can effectively be ignored. In addi-
used. Mechanical analysis of tion, since the lipid bilayer acts as a two-dimensional fluid, the two leaflets can
membranes can be thought of as a
generalization of the analysis of
slide over each other further reducing the resistance to bending.
strings to two dimensions and finds We will consider two particular cases of in-plane loading, shear and tension. Let
application in the analysis of rubber us start our consideration of equilibrium by applying it to an imaginary free body.
sheets and drums.
An infinitesimal element of the membrane subjected to in-plane forces is illus-
trated in Figure 9.7.
As we learned in Chapter 3, we need to apply the equilibrium conditions to
forces rather than to stresses. To accomplish this, we must define resultant
forces in terms of stress. The resultant forces are the equivalent force that results
 MEMBRANE MECHANICS I: IN-PLANE SHEAR AND TENSION 263

z Figure 9.7 An infinitesimal element

of a membrane depicting the forces
Nxx and Nyy.
x y

Nxx
Nyy

from the net action of the stresses acting on each edge of the element. Remember Nota Bene
that stress is force per unit area. So, just as in the rod or column in Section 3.2,
Pressure assumption. In this
the resultant force on each edge is the integral of the stress over the area of the
analysis we only consider the
edge. However, there are some important differences. First, because the shell is situation in which the pressure is in
two-dimensional, we must introduce the relevant x- and y-components denoted the −z-direction ( px = py = 0). We can
by subscripts. Secondly, rather than deal with the total force resultant on an always construct a coordinate
edge, N, it is more convenient to divide by the edge width, b, such that n = N/b. system with x and y aligned in the
The concept of resultant force per unit length (width) can be thought of as a plane of the membrane. We are
generalization of surface tension that we discussed in Section 1.3. We define the ignoring the effect of rotation of the
normals, which would introduce what
resultant forces per length to be
is known as a follower force and
h /2 h /2 h /2 greatly complicates the math.

∫ ∫ ∫τ
However, as long as the rotations are
nxx = σ xx dz nyy = σ yy dz nxy = xy dz. (9.26) small, we can assume that the
− h /2 − h /2 − h /2
 pressure acts only in the z-direction.

Note that sometimes (particularly in the structural mechanics literature) these are
called the stress resultants.
With these definitions, we are now ready to derive the equilibrium conditions by
constructing, as usual, a free-body diagram. If we consider an element of the
membrane (Figure 9.8) that is not accelerating, equilibrium tells us that the

z Figure 9.8 A free-body diagram for

an infinitesimal element of a
membrane, dx × dy. Note that
because the membrane can curve, it is
not restricted to remain in the x–y-plane.
nyy nxx The slopes and rate of change of slopes
are denoted on the interior of the shell.
nyx nxy
dw dw
— —
dx dy

pz
dw d2w dw d2w
—+— dx —+— dy
dx dx2 dy dy2
x y
dw d2w dw d2w
— + — dx — + — dy
dx dxdy dy dxdy

dnxx dnyy
nxx + — dx nyy + — dy
dx dnxy dnyx dy
nxy + — dx nyx + — dy
dx dy
264 CHAPtER 9: Mechanics of the Cell Membrane

resultant forces must sum to zero. This situation is slightly more complicated,
because the membrane is not flat and the force resultants therefore change not
only with respect to dx and dy, but also dz.
Figure 9.8 is more complex than we are used to. Owing to the curvature of the
membrane, we now have to consider the local slopes. At the x = 0 edge, the
slope of the surface in the y-direction is dw/dy. At the x = dx edge, we must also
account for the rate of change in the slope. On this edge, the slope of the sur-
face in the y-direction is even more complicated, dw/dy + (d2w/dy2) dy. For the
summations in the x- and y-directions, we ignore this effect, however, for the
summation in the z-direction, it cannot be ignored. If you are careful, you
should find,

 dnxx   dnyx 
∑ f x ⇒ − nxx dy +  nxx + dx  dy − nyx dx +  nyx + dy  dx = 0
 dx   dy 

 dnyy 
dy  dx − nxy dy +  nxy +
dnxy 
∑ f y ⇒ − nyy dx +  nyy + dx  dy = 0
 dy   dx 

dw  dnxx   dw d 2w 
∑ fz ⇒ − nxx dy +  nxx +
  dx + dx 2 dx 
dx  dy 
dx  dx
(9.27)
dw  dnxy   dw d 2w 
− nxy dy +  nxy + dx  dy  + dx
dy  dx   dy dxdy 

dw  dnyy   dw d 2w 
− nyy dx +  nyy + dy  dx  + dy
dy  dy   dy dy 2 

dw  dnyx   dw d 2w 
− nyx dx +  nyx + dy  dx  + dy + pzdxdy = 0.
dx  dy   dx dxdy 

Notice that if we divide by dx and dy, the terms simplify significantly to

dnxx dnxy
∑ fx = 0 ⇒ + =0
dx dy
dnyx dnyy
∑ fy = 0 ⇒ + =0
dx dy (9.28)

 dw dw   dw dw 
d  nxx + nxy d  nxy + nyy
 dx dy   dx dy 
∑ fz = 0 ⇒ + + pz = 0.
dx dy

The first two equations of Equation 9.28 are analogous to the stress equilibrium
condition we derived in Equation 3.46, applied to a two-dimensional membrane
structure. However, the third equation is new and shows how the force equilib-
rium condition in the transverse direction relates membrane tension to pressure.
We can further simplify by writing out the derivatives

d 2w d 2w d 2w
nxx 2
+ 2nxy + nyy 2 + pz = 0. (9.29)
dx dxdy dy

We can express Equation 9.29 in terms of curvature, just like we did for the beam,
except that now there are curvatures in different directions.

nxxκ xx + 2nxyκ xy + nyyκ yy + pz = 0, (9.30)

 MEMBRANE MECHANICS I: IN-PLANE SHEAR AND TENSION 265

where
d 2w d 2w d 2w
κ xx = 2
, κ xy = , κ yy = . (9.31)
dx dxdy dy 2

In summary, equilibrium requires that the following conditions on the stress

resultants be met:

dnxx dnxy
+ =0 Nota Bene
dx dy
Moment balance. When we applied
dnxy dnyy the equilibrium equations, we did not
+ =0 (9.32) consider the implications of the
dx dy
moment summation. Particularly
useful is the observation that the
d 2w d 2w d 2w
nxx + 2n xy + n yy + pz = 0. in-plane shear resultants must be
dx 2 dxdy dy 2 symmetric, nxy = nyx, which is a

consequence of the requirement
To gain a deeper understanding of this equation, we will take a closer look at this that the moments about the z-axis
expression and elaborate it for two special cases, the case of planar shear and sum to zero.
equibiaxial tension.

Equilibrium simplifies in the case of shear alone

In the case of shear loading, the applied tension in one direction, say nxx, is larger
than in the orthogonal direction ( y-direction). The simplest case is one of “pure
shear,” which occurs when nxx = −nyy. In pure shear, a direction of maximum shear
occurs at a 45 degree angle to the direction of load application and will have a
magnitude of (nxx − nyy)/2. Even if the loading is not pure shear, there is generally
some amount of shear, as long as nxx and nyy are not exactly equal (if they are, we
have equibiaxial tension, which is considered next). Recall from the constitutive
equation for a general Hookean material (Equation 3.57), that the shear stresses
and strains are decoupled from the normal stressed and strains. This means that
the shear behavior (or deformational mode) is not influenced by anything else
and can be analyzed on its own. Even for complex loading we can create a condi-
tion of pure shear by subtracting the equibiaxial component,

 n − nyy 
n xx = nxx −  xx 
 2

 n + nyy  (9.33)
n yy = nyy −  xx 
 2
n xy = nxy .

The equilibrium condition (Equation 9.33) is not very informative. It requires that
a case of pure shear can only exist if the shear resultant is constant. However, we
can apply the constitutive equation to gain some insight, τxy = γxy. We can reformu-
late this in terms of stress resultants. Because nxy = τxyh

nxy = Gγ xy h
(9.34)
= K Sγ xy .

We have introduced a new constant KS = Gh called the membrane shear modulus,
which has units of force per unit length.
What are typical values of the membrane shear modulus for cells? The cell mem-
brane of a red blood cell has a typical value of KS = 6 × 10−6 – 9 × 10−6 N/m = 6–9
pN/μm. Why is it so small? As we have described, the lipid bilayer is formed of
molecules that are tightly bound to the layer, but easily move within the plane of
266 CHAPtER 9: Mechanics of the Cell Membrane

the layer. This behavior is sometimes called a “two-dimensional liquid” because,

like a liquid, if the membrane is exposed to shear it will simply flow to a new con-
figuration until the shear is gone. For the bilayer, we can generally ignore the
effects of shear. Another property of the lipids that make up the cell membrane is
that they have a relatively large volume expansion or bulk modulus. That is, they
are incompressible. As we will see in the following analysis, the in-plane bulk
modulus of the lipid bilayer is also high, further supporting a two-dimensional
fluid description for the membrane.

Equilibrium simplifies in the case of equibiaxial tension

In a state of equibiaxial tension the membrane experiences tensile forces that are
equal in both directions, nxx = nyy = n, while the shear force vanishes, nxy = 0. We
will further assume that n is the same throughout the membrane; that it is inde-
pendent of position and direction. At first glance this may appear to be a very
peculiar and aggressive assumption. It would seem only to be valid for very spe-
cific artificial conditions. Surprisingly, however, this is a common and important
condition for bilayer membranes. Why? As we have seen, any non-equal compo-
nent in nxx and nyy will result in a shear in some plane. Because of the mem-
brane’s low viscosity and ability to flow under shear, it will reorganize such that
the shear is again zero. As long as our boundary conditions are applied displace-
ments, we know that nxy is zero everywhere. If this is the case, equilibrium
(Equation 9.32) requires that dnxx/dx = dnyy/dy = 0. The membrane tension must
remain constant or the membrane will flow from regions of low tension to high
tension until equilibrium is again satisfied. This is actually a very good assump-
tion for the behavior of the lipid bilayer. It is equally applicable to other two-
dimensional fluids, such as soap bubbles.
In this particular, but important, special case, we can simplify our description sig-
nificantly. The question of force equilibrium in the x- and y-directions is auto-
matically satisfied because the gradients are zero. In the transverse (z) direction,
equilibrium reduces to the law of Laplace which we are familiar with from Section
1.3, but now generalized to allow curvatures in different directions,
 d 2w d 2w 
n 2 + + pz = 0. (9.35)
 dx dy 2 

Note that the mixed term has dropped out because of the condition that nxy = 0.
Curvature is related to the local radius according to

d 2w 1
2
= κ xx =
dx Rx
(9.36)
d 2w 1
= κ yy = .
dy 2 Ry

Therefore,

 1 1 
pz = n  + . (9.37)
 Rx Ry − 

If the membrane is spherical in shape, we recover Equation 1.1 for the oil-drop
model of the cell,
2n
+ p z = 0. (9.38)
R
Notice that the law of laplace is typically expressed as 2n = pz , because the
R
­curvature and pressure are oriented in different directions such that pz = –p.
 MEMBRANE MECHANICS II: BENDING 267

Areal strain can be a measure of biaxial deformation

In Section 1.3, we briefly introduced areal strain and the areal expansion modulus
in our treatment of micropipette aspiration experiments. These quantities are par-
ticularly relevant when characterizing deformations under equibiaxial tension.
We revisit them here to develop their origins more completely. In equibiaxial ten-
sion, any collection of points on the membrane will be deformed into a new state
that is similar in shape to the original one, but will have its area increased or
decreased. Similar to our definition of strain (the ratio of change in length to origi-
nal length), we can define areal strain to be the change in area over the original
area ΔA/A. We can relate areal strain to linear strain by considering a small square
element of dimension L on each side. Its initial area A = L2 will increase to A + ΔA
= (L + ΔL)2 when it is deformed. However, this can be written in terms of linear
strain, because ΔL =  εL, A + ΔA = (1 +  ε)2L2. Therefore,

∆A = (1 + ε )2 L2 − L2 (9.39)

and

∆A/A = (1 + ε )2 − 1 = 2ε + ε 2  2ε (9.40)

by assuming that the term involving the square of strain will be in general much
smaller than the others and can be neglected.
We can now define the material property areal expansion modulus for equibiaxial
tension in terms of the tension and areal strain,

n
KA = . (9.41)
(∆A/A)

KA is a measure of the resistance of a membrane to in-plane biaxial stretching. It is

left as an exercise for the student to show that it can be expressed as a function of
Young’s modulus and Poisson’s ratio

Eh
KA = . (9.42)
2(1 − ν )

Typical values of the area expansion modulus for lipid bilayers are in the range of
KA = 0.1−1.0 N/m. The cell membrane of red blood cells, for example, has an
areal expansion modulus of approximately KA = 0.45 N/m (= 450,000 pN/μm).
This value is many orders of magnitude greater than the membrane shear modu-
lus. The bilayer is often treated as effectively inextensible. Typically, a lipid bilayer
membrane can only tolerate 4–6% areal strain before rupture. The large resist-
ance to areal change can be attributed to the changes in energy associated with
exposing the hydrophobic core of the lipid bilayer to water as the spacing between
the individual molecules is increased. Experimental approaches for measuring
KA are discussed in Section 9.6.

9.5 MEMBRANE MECHANICS II: BENDING

In Section 9.4 we considered a two-dimensional curved structure that does not
experience bending moments (a structural membrane). Now we are going to add
bending. A two-dimensional structure might be flat, in which case we are talking
about a plate, or curved, in which case we are talking about a shell.
Recall that owing to its fluidity, bending resistance of the plasma bilayer is neg-
ligible for curvatures typical in a cell. It is reasonable to ask why we even worry
about bending moments. Cells can, in fact, have substantial bending stiffness.
268 CHAPtER 9: Mechanics of the Cell Membrane

spectrin ankyrin How could the red blood cell maintain its characteristic bi-concave structure
rather than acting as a flaccid bag of cytoplasm?
The answer to this paradox is that cells derive their bending stiffness from
underlying cytoskeletal structures that support the bilayer. In the red blood cell,
a network of the polymer spectrin exists directly under the bilayer. It is bound to
the bilayer through periodic inclusions of ankyrin, a linker protein that attaches
membrane integral membrane to membrane-spanning proteins (Figure 9.9). As we discussed in Section 8.4,
protein band 3 the spectrin network takes on a characteristic triangular–symmetric form
Figure 9.9 Complexes of (Figure 9.10).
membrane-spanning proteins and Another way in which cells develop bending rigidity is by covering themselves
linker proteins are able to provide
with a coating of proteoglycans known as the glycocalyx. The glycocalyx can be
the cell membrane bending
stiffness. as thick as 0.5 μm. It is made up of long bottle-brush-like molecules that are
highly bifurcated and covered with a high negative charge density. This causes
intermolecular repulsion in the glycocalyx and tends to attract water molecules.
The net effect is a tendency of the glycocalyx to be highly resistant to bending. In
terms of bending, the bilayer tends to function primarily as a chemical barrier
with the structural integrity being dominated by the cytoskeleton and glycoca-
spectrum lyx (Figure 9.11).

anchors In bending the kinematics are governed by membrane rotation

Analysis of bending moments in a curved shell is actually quite advanced
and requires a working knowledge of differential geometry and the mathematics
Figure 9.10 The spectrin network of of curved spaces. So, we simplify our analysis by considering a flat plate as classi-
the red blood cell forms a cally described by the Kirchhoff plate equation, which we will see is a fourth-
characteristic triangular network. order differential equation in the transverse displacement. Again, the plate
The ankyrin and the spectrin networks equation is a result of the traditional continuum equations (the kinematics, the
are unique to red blood cells. However, constitutive equations, and the equilibrium equations) combined with the defi-
other animal cells also support their nition of the stress resultants. Essentially, this is a two-dimensional extension of
bilayer with a network of membrane-
associated actin. This layer of peripheral,
the development we considered previously for beam bending in Section 3.2.
dense actin is known as “cortical” actin Let us begin with the kinematics. We have already covered the Kirchhoff assump-
and typically forms a quadrilateral tions and the decomposition of strain into a part that is constant through the
repeating network.
thickness and a part that varies linearly though the thickness. The former do not
produce any bending moments, so, we can drop them from Equation 9.22. For the
pure bending that remains, our kinematic assumption gives us u and v in terms of
the slope of the z-displacement w,

dw dw
u = −z υ = −z . (9.43)
dx dy

From this we can calculate the strains from their continuum definitions, Equations
3.48, 3.49, and 3.52,

∂u ∂ 2w
ε xx = =− 2z
∂x ∂x
Figure 9.11 Cortical actin stiffens
the cell membrane. In this fluorescent ∂v ∂ 2w
confocal micrograph, the actin forms a ε yy = =− 2z (9.44)
∂v ∂y
dense band around the periphery of the
cell. See www.garlandscience for a color
1  ∂u ∂ v  ∂ 2w
version of this figure. (Courtesy of ε xy =  +  =− z.
Andrew Baik, from the laboratory of X. 2  ∂y ∂ x  ∂x ∂y

Edward Guo.)
As before, we can use our definition of curvature to obtain

ε xx = κ xx z ε yy = κ yy z ε xy = κ xy z. (9.45)

 MEMBRANE MECHANICS II: BENDING 269

Linear elastic behavior is assumed for the constitutive model

Next, we use the relevant constitutive equations, that is, the equations relating
stress and strain. Again, we will assume a generalized Hookean material behavior
and that the material properties and thickness are constant (homogeneous). In
particular we need to relate the in-plane normal stresses σxx and σyy and the in-
plane shear stress σxy to the corresponding strains ε or curvatures κ. Because of
our kinematic assumptions, all of the z-direction strains are zero: εzz =  εxz =  εyz = 0.
From Equation 3.57 we obtain,

σ xx = 2 µε xx + λ (ε xx + ε yy ) = 2 µzκ xx + λz(κ xx + κ yy )

σ yy = 2 µε yy + λ (ε xx + ε yy ) = 2 µzκ yy + λz(κ xx + κ yy ) (9.46)

τ xy = 2 µε xy = 2 µκ xy z.

Equilibrium places conditions on resultant forces

and moments
Now we are at the point at which we need to employ the equilibrium equation.
However, equilibrium cannot be applied to stresses or strains directly. So, we
need to consider resultant forces again. In this situation we are considering bend-
ing only, so we need to derive the stress moment resultants just as we did in Section
3.2 for beams. This will be a little more complex because of the two-dimensional
nature of the situation, but it is quite analogous. For the beam, there was only one
moment resultant. For the plate, we actually have three, denoted mxx, myy, and
mxy. It is no surprise that they are obtained by integrating the corresponding
stresses through the thickness of the plate,

h /2 h /2
Nota Bene
mxx ( x , y ) =
∫
− h /2
σ xx ( x , y , z )z dz m yy ( x , y ) =
∫
− h /2
σ yy ( x , y , z )z dz
The units of the moment
resultants are the same as for
h /2 force. Remember that for our

∫σ (9.47) distributed force resultants, the units

mxy ( x , y ) = xy ( x , y , z )z dz. were force per unit length, similar to
− h /2 surface tension. Normally, a moment
has units of force times length. So, a
distributed moment has units of
With the moment resultants defined, what can equilibrium tell us? Imagining a force times length per unit length, or
small element exposed to moments is a little more complex than for in-plane just force.
force resultants, but it is conceptually the same. Figure 9.12 depicts the edges of a
small element of the plate and the moment resultants at the edges with a double
arrow symbol.

Figure 9.12 The moment resultants

on the edges of the surface either
directly (left) or with the double
arrow symbology (right). The double
arrow implies a moment around the axis
of the arrow following the right-hand rule.
myy
mxx mxx
z myx
mxy myx myy
mxy

x y
270 CHAPtER 9: Mechanics of the Cell Membrane

Just as with the beam, we cannot satisfy equilibrium conditions in terms of

moments without considering shear. We must introduce another resultant force
that we neglected before, the transverse shear force. We neglected it before
because it does not play a role in membrane mechanics, but it is important in
plates. The transverse shear force is the z-direction force that is produced from the
qy x–z- and y–z-shear stresses (Figure 9.13).
qx
z
We mathematically define the shear resultants by integrating through the thickness,
x y h /2 h /2

Figure 9.13 The distributed shear

force resultants on the edges of the
qx ( x , y ) =
∫σ
− h /2
xz ( x , y , z )dz qy ( x , y ) =
∫σ
− h /2
yz ( x , y , z ) dz. (9.48)

surface.

Now we are finally ready to apply the equilibrium equations. In the membrane anal-
ysis, we used force equilibrium to derive the governing equations and moment equi-
librium about the z-axis to show a symmetry condition. For plate bending, we need
to add the moment equilibrium about the x- and y-axes. Also, force equilibrium in
the x- and y-directions does not apply because there are no forces in those directions.
So, let us start with the force equilibrium in the z-direction. From Figure 9.14 we can
deduce the force equilibrium condition in the z-direction. Some simple manipula-
tion (dividing by dx and dy and canceling terms) implies

dqx dqy
+ + pz = 0. (9.49)
dx dy

Similarly for the moments

dmxx dm yx
∑ Mx = 0 ⇒ + − qx = 0
dx dy
(9.50)
dm yy dmxy
∑ My = 0 ⇒ + − q y = 0.
dy dx

Also, we can use the z-axis moment condition to show myx = mxy.

Figure 9.14 Free-body diagram of z

resultants on an isolated portion of
a plate depicting shear and
moment resultants. Notice that the
moment resultants on the edges of the qy qx
surface are represented with the double myx mxy
arrow symbology. myy

pz mxx
dqx dqy
x qx + — dx qy + — dy y
dx dy
dmxx dmyy
mxx + — dx myy + — dy
dx dy

dmyx
myx + — dy
dmxy dy
mxy + — dx
dx
 MEMBRANE MECHANICS II: BENDING 271

These are the equilibrium equations for a plate. They are a coupled set of first-
order differential equations. You may remember that it is often possible to
replace a coupled set of lower-order differential equations with a single higher-
order differential equation. Indeed, this is the case here. We can eliminate the
transverse shear components to find a second-order equation in the moments.
It is left as an exercise for you to show that Equations 9.49 and 9.50 can be com-
bined to yield

d 2mxx d 2mxy d 2m yy
2
+2 + + p z = 0. (9.51)
dx dxdy dy 2

This is the classical equilibrium equation for a plate in terms of the distributed
moments and pressure.
Now we have all the components we need to find the governing equations for
plate bending. First, we combine our definition of moments (Equation 9.47) with
our constitutive law (Equation 9.46). Combining the integrals yields
h /2

mxx =
∫σ
− h /2
xx zdz = K B (κ xx + νκ yy )

h /2
(9.52)
m yy =
∫
− h /2
σ yy zdz = K B (κ yy + νκ xx )

h /2
1 − ν2
mxy =
∫
− h /2
σ xy zdz = K B
1+ν
κ xy ,

where

Eh 3
KB = . (9.53)
12(1 − ν 2 )

We can now insert this new expression for moments in terms of curvature into our
equilibrium Equation 9.51 to yield

 d 2κ xx d 2κ xy d 2κ yy 
pz = K B  +2 + (9.54)
 dx 2
dxdy dy 2 

or in terms of displacement

 d 4w d 4w d 4w 
pz = K B  4 + 2 2 2 + . (9.55)
 dx dx dy dy 4 

This is the fourth-order differential equation describing plate bending relating the
transverse displacement to applied pressure.
This completes our development and derivation of the plate-bending equation.
However, let us try to get an intuitive feel for bending at a cellular level. The bend-
ing stiffness, KB, of lipid bilayers such as the red blood cell membrane is on the
order of 10−19 N/m. This is an exceedingly low value. Remember that KA was on
the order of 0.1 or 1 N/m. Of course, the units of these quantities are different.
Nonetheless, there are about 20 orders of magnitude between typical bending and
expansion forces. Bending forces in a bilayer are small even compared with shear.
Recall the shear stiffness, KS, is on the order of 10−6 N/m. Bending remains 13
orders smaller than this. We knew this was the case (although perhaps not how
272 CHAPtER 9: Mechanics of the Cell Membrane

dramatically different these force regimes are), because we know resistance to

bending of the cell membrane comes from membrane-associated structural pro-
teins and not the bilayer. A support network of spectrin or cortical actin can
increase the effective KB dramatically.

Which dominates, tension or bending?

We have treated in-plane loading and bending as two separate conditions.
Certainly, in reality, both occur simultaneously. Nonetheless, in many cases, one
will dominate. To see this more clearly, we can formulate a description for com-
bined loading,

 d 2w d 2w   d 4w d 4w d 4w 
n 2 + 2 
+ KB  4 + 2 2 + + pz = 0. (9.56)
 dx dy   dx dx dy dy 4 

In general this equation can be difficult or impossible to solve analytically, but

numerical solutions are certainly available. We can also determine which loading
mode will dominate the solution and simplify in that way. We can estimate the
rough relative magnitude of forces originating from each term, membrane surface
tension due to n, and bending stiffness due to KB. Assume that w is the transverse
displacement that occurs over a characteristic length λ. The membrane term will
scale with the quantity nw/λ2. Similarly, the bending term will scale with KBw/λ4.
The ratio of these two quantities, KB/[nλ2] gives us a good indication of whether
tension or bending is dominant for a given situation.

KB
1 ⇒ tension dominates
nλ 2
(9.57)
KB
1 ⇒ bending dominates
nλ 2

For a typical cell we might have KB = 10−18 N/m, n = 5 × 10−5 N/m, and λ = 1 μm.
Therefore, KB/[nλ2] = 0.02, indicating that, in-plane effects are more important
than bending.

9.6 MEASUREMENT OF BENDING RIGIDITY

We conclude this chapter with a discussion of the measurement of material
properties of membranes. In Section 1.3, we saw that micropipette aspiration can
be used to measure the areal expansion modulus KA in lipid vesicles or in cells.
Can a similar approach be used to measure the bending rigidity KB? The answer
is “yes,” and conveniently, one can measure both properties simultaneously
within the same experiment. To understand how bending rigidity can be ascer-
tained in micropipette experiments, we must revisit the concept of thermal
undulations.

Membranes undergo thermal undulations similar to polymers

In Section 7.3 we introduced the concept of thermal undulations in polymers and
saw that entropy can play a significant role in their mechanical behavior.
Membranes also undergo such fluctuations. In polymers, we characterized these
thermal fluctuations using the persistence length, a characteristic length scale
over which the segment of a polymer under the influence of thermal forces
remains relatively straight. To review, the persistence length was derived from the
orientation correlation function f (s) = <cos Δθ (s) > , where θ (s) is the angle the
 MEASUREMENT OF BENDING RIGIDITY 273

polymer makes with an imaginary horizontal line at each point s, and Δθ (s)=
θ (s) − θ (0) (see Equation 7.14). In particular, we found that

 −s 
  (9.58)
p 
cos ∆θ (s ) = e  .

The orientation correlation function exponentially decreases from 1 to 0 as s gets

larger and larger. In relating the persistence length to our continuum description
of a beam, we found that it is related to its flexural rigidity EI as

EI
p = . (9.59)
kBT

For surfaces such as membranes, thermal undulations can also be described

using an orientation correlation function with some modifications. In particular,
differences in orientations at two different points on the membrane r1 and r2 can
be characterized by the dot product of the normals at these points, n(r1) n(r2).
•

If we let Δr = ||r1 − r2||, then the membrane persistence length can be specified as

n(r1 ) i n(r2 ) = e − ∆r / p . (9.60)

The persistence length in membranes can also be related to continuum models, in

particular,
4 πK B

 p ∼ be 3kB T , (9.61)

where b is a characteristic length scale associated with the intermolecular spacing

within the membrane. The derivation is beyond our scope here, but Boal (2001)
has provided a detailed step-by-step treatment. As in a polymer, the persistence
length in membranes increases with increasing resistance to bending and
decreasing temperature. However, unlike in polymers, ∙p scales exponentially
with these parameters instead of linearly.

Membranes straighten out with tension

Just like polymers, membranes are subject to thermal undulations in a manner
that is dependent on bending rigidity. In polymers, we showed that “coiled” ver-
sus “straight” configurations could be characterized through differences
between the end-to-end length and contour length. In addition, we found that
“coiled” configurations were entropically favored; however, through the appli-
cation of force, the polymer could be extended. Analogously, in membranes,
“wrinkled” configurations are entropically favored, and these wrinkles can be
smoothed out through the application of tension. Furthermore, a quantitative
relationship exists.
Consider a surface defined through a height function over the xy-plane, h(x,y).
We assume that there are no overlaps in the surface, that is, h is unique at each
point (x,y). Let A be the contour area (the true area) of the surface and Aproj be
the projected area in the xy-plane. Aproj can be shown to depend on the tension,
n, and bending rigidity KB,

AkBT  π 2 /b 2 + n/K B 
Aproj (n) = A − ln . (9.62)
8πK B  π 2 /A + n/K B 

274 CHAPtER 9: Mechanics of the Cell Membrane

Note that Aproj goes to zero when kBT/KB goes to zero, indicating that no wrinkles
will occur in the limit of zero temperature or infinite bending rigidity. Aproj also
goes to zero in the limit as n/KB gets arbitrarily large.
From this we can compute the amount of tension required to increase the pro-
jected area from a tensionless reference state,

Ak B T  1 + nA/π 2 K B 
Aproj (n) − Aproj (0) = ln (9.63)
8πK B  1 + nb 2 π 2 /π 2 K B 

or

k BT
∆A/A = ln(1 + nA/π 2 K B ), (9.64)
8πK B

where ΔA = Aproj(n) − Aproj(0), and we have assumed that A >> b2, the square of the
intermolecular spacing. The above expression gives the “effective” areal strain
associated with smoothing out thermal undulations under a given tension n. This
expression can be superimposed with the zero temperature areal strain given by
the areal expansion modulus (that is, the strain associated with increasing the
intermolecular distance),

k BT
∆A/A = ln(1 + nA/π 2 K B ) + n/K A . (9.65)
8πK B

Example 9.5: Measurement of KB and KA in a single experiment

How does Equation 9.65 behave in the high and low ten- Consider an experiment in which a lipid vesicle is being
sion limits and how might this be used to determine KA subjected to micropipette aspiration. In the absence of
and KB? tension, entropic forces induce wrinkles in the mem-
brane. Upon application of a low tension, these entropic
In the low tension regime, Equation 9.65 becomes
forces are overcome, the wrinkles are smoothed, and the
response is given by the low tension expression. The
k BT areal strain is linear with ln(n); fitting the slope gives KB.
∆A/A ≈ ln(1 + nA/π 2 K B ) .
8πKB Once the membrane has been pulled taut, the stiffness
dramatically increases, owing to the force associated
with increasing the intermolecular spacing between the
In the high tension regime, n dominates over ln(n), so, as lipids in the bilayer. Here, the response is given by the
n approaches zero, high tension expression, and the areal strain is linear
with n. KA is also easily derived from the curve fit.
∆A/A ≈ n/K A .

Advanced Material: Pipette aspiration measurement of areal expansion modulus

An alternative approach to measuring the areal expansion where [Ci] is the molar concentration of each spe­cies
modulus is using a small-bore pipette. One can determine present. From the law of Laplace one can relate posmotic
the osmotic pressure across the membrane as a function to n, and the areal extension can be determined from the
of solute concentration and temperature as amount of membrane drawn into the pipette.

posmotic = kBT ∑[C ],

i
i
 Problems 275

Key Concepts

• The amphiphilic nature of cell membrane • The continuum mechanical behavior of the membrane
phospholipids allows them to self-assemble into the can be decomposed into in-plane and bending
bilayer. The packing behavior of the phospholipids components. In-plane behavior can be further
contributes to the structures formed. decomposed into equibiaxial and shear responses.
• Microdomains within the membrane enhance • Derived moduli, including KA, KS, and KB, characterize
chemical kinetics. the areal, shear, and bending stiffnesses-respectively.
• Saturated and unsaturated fatty acid chains • The quantity KB/nλ2 is indicative of tension-dominated
contribute to determining membrane properties. or bending-dominated behavior.
• The fluid mosaic model describes the two- • The bilayer is relatively flexible in shear and bending
dimensional fluid behavior of the bilayer. but can be stiff with respect to areal tension.
• The membrane is semipermeable and has a critical • Entropic membrane undulations can be described
barrier function that can be described with the statistically and are related to continuum properties.
diffusion equation.

Problems

1. (a) In one dimension, the mean squared displacement (a) If you had n channels in the cell surface, where n is
of a particle undergoing a random walk varies small, what is the total current flux into the cell?
with the square root of t (time). How does the (b) If the entire sphere is acting as a sink with current
mean squared displacement vary with time in two flux I0, then the total current flux if you had n
dimensions? And in three dimensions? Why? channels—for any n—is I = I0/[1 + πb/na]. Is this
(b) L et us say you defined the diffusion velocity as the consistent with your answer from (a)? Justify your
root mean squared displacement divided by time, response.
for a given diffusion coefficient. If you do this, what (c) Using the information in (b), design a cell with
happens as t approaches 0? How can you tell if radius b = 10 μm with n channels (sinks) of radius
experimental data you are analyzing are affected a = 1 nm so that its current flux through all of the
by this problem? channels is half of the current flux if the entire cell
surface was a sink (that is, find n). You may neglect
2. For typical small molecules in water, the diffusion the local curvature in calculating the total area
coefficient is D ≃ 10−5 cm2/s. Small ions typically pass of the channels. Next, find the fraction of the cell
through channels diffusively. You can assume that the surface area that is covered with channels for this
channels are only wide enough for a single ion to pass to occur. What does this result tell you about the
through at a given time. ability of a cell to acquire needed compounds by
(a) E
 stimate the time it takes for a small ion to cross diffusion through channels?
the channel, assuming there is no interference
from other ions, and the channel has the same 4. Consider the transport of a solute through a thin
length as the thickness of a cell membrane. blood vessel. The blood vessel has inner radius R1, the
(b) N
 ow assume someone is building a scaled-up endothelial layer increases the radius to R2, and then
macro-cell whose radius is about 1 m. How the outer layer (matrix) further increases the radius
long would it take the same small ion to cross to R3. If the diffusion coefficient of the endothelial
the macro-channel? Is diffusion an effective layer for the solute is D1 and the diffusion coefficient
mechanism for large-scale processes? of the matrix is D2, find the total (steady-state) solute
flux out of the blood vessel, assuming the blood
3. Consider a hole of radius a in an infinitely large wall in concentration is effectively constant at C0 inside the
space. If the concentration far away from one side of the blood vessel and is zero everywhere outside (the solute
disk is C0, and is zero everywhere on the other side, the is immediately consumed). Keep in mind you need to
current flux of freely diffusing molecules passing through use the cylindrical form of the diffusion equation(s).
this hole is given by I = 4DaC0. This can be a model for
a single channel in a membrane (assuming the radius, 5. Using the same setup as in Problem 4, model the
a, is much smaller than the radius, b, of the cell itself). blood vessel as a hollow cylinder of inner radius R1
We have derived the current flux of freely diffusing and outer radius R3 but consisting of only one material
molecules passing through a sphere as I = 4πDC0b. of diffusion coefficient D, with the same flux out of
276 CHAPtER 9: Mechanics of the Cell Membrane

the vessel. Find D in terms of D1, D2, and the given increases to C0 > 0. You may assume that inside the cell,
parameters. at x → ∞, the concentration C = 0 forever.
(a) Write Fick’s second law of diffusion for this setup
6. You are developing a drug delivery system consisting of
for t > 0.
a small sphere of inner radius R, wall thickness h, and
(b) Define a new variable α so that
wall diffusion coefficient D. Assume that the internal drug
concentration is constant at C and the concentration x
α = .
outside the sphere is zero everywhere (used up 4 Dt
immediately) for the time interval we are interested in.
(a) C
 alculate the total drug flux out of the sphere. Rewrite the equation you derived in (a) in terms of
(b) N
 ext, assume you have a flat wall of the same α and show that the equation can be expressed as
thickness h and the same surface area as the inside ∂ 2C ∂C
of the sphere in the original problem statement. = 2α = 0
∂α 2 ∂α
Further, assume the concentrations and diffusion
coefficients are analogous to the spherical setup. You may need to use the chain rule.
Calculate the total drug flux through the wall. (c) W  rite the boundary conditions for the new
(c) F
 ind a criterion for h and R in terms of the other equation in (b) using the same constraints you
parameters such that the difference between determined for (a).
your answers for parts (a) and (b) is less than 1%. (d) Find the solution for the differential equation in (b),
Would something the size of a typical cell with the using the boundary conditions you found in (c). You
membrane being the “wall” qualify, based on your may want to express the solution in terms of the
criterion? error function (erf(x)).
(e) What happens at the limit of the solute flux as t
7. Assume the cell membrane has a fixed thickness h and
approaches infinity? Is it the same as the steady-
a diffusion coefficient D for some solute of interest.
state diffusive flux? Why or why not?
You may model the cell membrane locally as a flat wall.
(f) Someone makes the following statement:
The external concentration of the solute is C and is 0
“The solution derived for the time-dependent
everywhere inside the cell (consumed immediately).
concentration profiles in (d) suggest that if I have
(a) D
 evelop an expression for the concentration inside two materials of different diffusion coefficients,
the membrane at steady state, making sure to with everything else being the same as in the
identify the boundary conditions you used. Sketch original problem statement, then the concentration
this concentration profile. profiles in the two materials will be identical, only
(b) A
 t some time after this steady state has been shifted in time. That is, it will never be the case that
achieved, the cell changes the diffusion coefficient for t > 0 one material will exhibit a concentration
to a new D′ < D (for example, by closing some profile that the other will never exhibit.” Support or
previously open channels). You then wait for a new disprove this statement.
steady state to be achieved. List what you expect
to change with the new diffusion coefficient D′, 11. Is it possible to have water climb 1 km in a capillary tube
assuming every other parameter remains the same. on Earth, using only surface tension? You may control the
diameter of the tube (which does not have to be constant
8. You are given a finite number of identical (finite-volume) in space), the tube’s angle to the vertical (but only the
tanks separated by geometrically identical membranes. vertical height counts), and how deep the water the end
You may set the initial concentration of the solute in each is submerged (but only the height above the surface
tank (and assume that the concentration is always uniform counts). You may also assume that water is a continuum
inside the tank itself) as well as the diffusion coefficient so that you can make the diameter arbitrarily small
of each membrane (but once you have chosen the D for without worrying about molecular effects (or even that
a membrane, it must remain fixed forever). Configure the the diameter may be smaller than a water molecule). If
tanks and membranes such that the tank with the highest your answer is yes, what is the shape of the tube?
initial solute concentration will receive a flux of the solute
at some point in time (before equilibrium). 12. If you have a water droplet and compress it slightly,
resulting in an ellipsoid (technically an oblate spheroid),
b3 b3 and release the compressing force, show that the
9. Show that Tc = is the solution for Tc =
3Da 3Da droplet tends to return to a spherical shape and not to
for the conditions specified in Example 9.2. distend further, using free-body diagram arguments.

10. We wish to model the cell locally as a flat wall of 13. The lipid bilayer consists of fatty acids that have varying
thickness h and diffusion coefficient D. Because we degrees of hydrogenation. Recall that fully saturated
are looking very closely at the cell, we will treat it as lipid chains will be relatively straight (zig-zag, actually).
a semi-infinite media, such that the cell starts at x = 0 Do you expect unsaturated lipid chain regions to have
and increases to infinity. Initially, the concentration higher or lower viscosity compared with the saturated
of some solute is zero everywhere. At time t = 0, the region? How about regions containing primarily partially
concentration of the solute outside the cell suddenly hydrogenated fatty acids?
 Annotated References 277

14. Show that 20. In the text, the claim was made that the bending
stiffness, KB, of the bilayer was much smaller than
 dw dw   dw dw  for the cell, because of the support provided by the
d  nxx + nxy d  nxy + nyy
 dx dy   dx dy  cytoskeleton. Conduct a literature search and write a
+ + pz = short report (no more than a quarter to half of a page)
dx dy
to support or refute this claim.
d2w d2w d2w
nxx 2
+ 2nxy + nyy + pz 21. Suppose you have a 1 μm cell with a membrane
dx dxdy dy 2
thickness of 10 nm made of an incompressible material
with a Young’s modulus of 108 Pa. Estimate the change
15. In the text we mentioned typical strains at rupture for
in pressure necessary to inflate the cell from an initial
a bilayer. Using reasonable values given in the text,
collapsed state in which the aspect ratio (height/
estimate the typical range of maximum membrane
diameter) is 0.2, to one in which the aspect ratio is one
tension and cytosolic pressure a cell of diameter 1 μm
with the large diameter approximately equal to 10 μm.
could withstand.
(This is roughly equivalent to the geometry change
16. Consider the cortex of a red blood cell. Develop a associated with a red blood cell going from its normal
model to predict the membrane mechanics of a two- state to spherical.) Assume that bending stiffness
dimensional array of cross-linked spectrin polymers dominates over tension.
arranged in a simple rectangular lattice. The lattice
22. Consider a plate experiencing a moment in one
spacing is Le. Each one of the polymers has a persistence
direction (Mxx = M Myy = Mxy = 0).
length ℓp, contour length of L, and diameter d.
This is similar to the beam-bending problem we
Using optical tweezers, researchers have found that the
analyzed in Chapter 3. In that case we found that
persistence length of spectrin is 10 nm. Furthermore,
the contour length of each spectrin molecule is 200 nm,
and Le is approximately 70 nm. d2w
M = EI .
(a) A
 ssuming that the deformations are small, dx 2
what regime (ideal chain, semiflexible polymer,
continuum mechanics) are these polymers in Does the plate theory produce the same equation for
and why? What model would be appropriate to this special case? If not, why is it different? What further
describe their behavior? assumption would be needed for the plate theory to
(b) N
 ow derive an expression for the membrane area reproduce the beam theory?
expansion modulus KA for the model. Eh3
23. Show that KB = .
17. Show that τxy = Gγxy = E/(2(1 + ν))γxy and KS = Gh = 12(1 − ν2 )
Eh/(2(1 + ν)).
24. Show that, taken together, the moment–curvature
18. Derive an expression for the areal expansion modulus relation (Equation 9.52) and moment equilibrium
for a membrane in terms of the membrane thickness, equation (Equation 9.51) result in the classical plate
Young’s modulus, and Poisson’s ratio. bending equation in terms of displacements,

19. Show that Equations 9.49 and 9.50 can be combined to  d2κ xx d2κ xy d2κ yy 
yield the equilibrium equation for a plate. pz = KB  +2 +
 dx 2 d xd y dy 2 

Annotated References

Axelrod D, Koppel DE, Schlessinger J et al. (1976) Mobility measure- Evans E & Rawicz W (1990) Entropy-driven tension and bend-
ment by analysis of fluorescence photobleaching recovery kinetics. ing elasticity in condensed-fluid membranes. Phys. Rev. Lett.
Biophys. J. 16, 1055–1069. This research article covers the theoreti- 64, 2094–2097. Gives expression for “effective” areal strain
cal basis underlying fluorescence recovery after photobleaching in associated with smoothing out thermal undulations under a given
order to characterize diffusion in 2-dimensional systems. tension.
Berg HC (1983) Random Walks in Biology. Princeton University Press. Helfrich W (1975) Out-of-plane fluctuations of lipid bilayers. Z. Natur-
Presents a wide range of biological problems that can be analyzed forschung. C 30, 841–842. Early work demonstrating that out-of-
through the random walk. Specifically our treatment of diffusion is plane fluctuations of membranes lead to a decrease in effective
motivated from this book. area and alters stretching elasticity. Gives expression for projected
surface area for a given tension n and bending rigidity.
Boal D (2001) Mechanics of the Cell. Cambridge University Press. An
excellent text on cell mechanics with many in-depth treatments of Kamm RD (2005) Molecular, Cellular, and Tissue Biomechanics
shell mechanics. Particularly useful for this chapter is the statistical (­lecture notes from course number 20.310, Massachusetts Insti-
treatment of membrane behavior. tute of Technology, Cambridge, MA). Several of the treatments in
278 CHAPtER 9: Mechanics of the Cell Membrane

this chapter were based on these graduate course notes. Specific Singer SJ & Sackmann E (1972) The fluid mosaic model of the
material included in-plane and bending behavior of plates and structure of cell membranes. Science 175, 720–731. One of
shells and several problems. the earliest descriptions of the fluid mosaic model of the cell
membrane.
Kooppel DE, Axelrod D, Schlessinger J et al. (1976) Dynamics of fluo-
rescence marker concentration as a probe of mobility. Biophys J. 16, Timoshenko S & Woinowsky-Krieger S (1959) Theory of Plates and
1315–1329. An early description of FRAP used to measure lateral Shells. Engineering Society Monographs. A defining text on the
diffusion in the cell membrane. mathematics of plate and shell mechanics.
Simons K & van Meer G (1988) Lipid sorting in epithelial cells. Bio-
chemistry 27, 6197–6202. An early description on microdomains in
lipid membranes.
CHAPTER 10
Adhesion, Migration,
and Contraction

T hus far, our focus has primarily been on cell biomechanics; in other words,
the study of the mechanical behavior of cells. In these treatments, we used
theoretical and experimental mechanics to understand better the mechanical
behavior of cells or their structural constituents. Ultimately, we are interested in
using this knowledge to understand how mechanics governs cellular function.
We now make a conceptual turn, shifting our focus from cell biomechanics to
cellular mechanobiology; in other words, aspects of biology in which mechani-
cal force is generated, imparted, or sensed, leading to alterations in biological
function. Because mechanics plays a critical role in many different physiologi-
cal and pathological processes, a complete treatment of cellular mechanobiol-
ogy is more than we can address here. In this chapter, we aim to give the reader
a firm foundation upon which to build by focusing on three vital cellular pro-
cesses in which mechanics dictates biological function: adhesion, migration,
and contraction.

10.1 ADHESION
Cells can form adhesions with the substrate
We begin this chapter with a discussion of cell–substrate adhesion, which is vital
for many biological processes, including tissue cohesion, repair, inflammatory
responses, and growth. Adhesion is also important for engineering applications,
such as seeding new constructs with cells for implantation.
Among the best-characterized cell–substrate adhesive molecules are integrins.
As we first described in Section 2.2, integrins are extracellular matrix protein
receptors that are tightly bound to the plasma membrane. They are dimers com-
posed of α and β subunits, each of which is found in one of several isoforms. Each
isoform is denoted with subscripts. The different isoform dimerizations allow for
the extracellular component of the integrins to bind to different extracellular
matrix components such as collagen or fibronectin. For example, α5β1-integrin is
composed of an α5 subunit and a β1 subunit and is a fibronectin receptor, specifi-
cally targeting the arginine–glycine–aspartate (RGD) peptide sequence in
fibronectin. The intracellular components of integrins are often structurally cou-
pled to the cytoskeleton in an indirect manner through one or more intermedi-
ary molecules. The intracellular component of the integrin may also be a major
site of enzymatic activity.
Integrins can be assembled into discrete adhesive plaques called focal adhe-
sions. As they are assembled, an intracellular molecular complex is also created.
This complex consists of many proteins, each of which has various roles in intra-
cellular biochemical signaling and forms structural links with the actin cytoskel-
eton (Figure 10.1). It is currently unclear what, if any, advantage is associated
with multiple focal attachments rather than a more distributed scheme. However,
their formation and destruction are highly regulated. It has been shown that
microtubules can disrupt focal adhesions locally at microtubule leading tips. It is
280 CHAPTER 10: Adhesion, Migration, and Contraction

Figure 10.1 Focal adhesions are (A) focal

molecular complexes involved in adhesion nucleus actin
cell–substrate adhesion.
Photomicrograph of focal adhesions in
adherent cells. The cells have been
immunostained for the focal adhesion
protein talin (encircled dot). The actin
cytoskeleton and nuclei have also been
stained. (B) A pictorial representation of
a focal adhesion demonstrating many of
the key proteins comprising this
molecular complex. Proteins within the
complex have various roles, including
functioning as intracellular signaling
molecules and forming structural links
with the actin cytoskeleton. (A, see
www.garlandscience.com for a color
version of the figure. B, adapted from,
Kamm R & Lang M Molecular, Cellular,
and Tissue Biomechanics.
Massachusetts Institute of Technology)

(B)
extracellular
fibronectin matrix

α β
integrin
cell membrane

Ras cytoplasm
SOS
Grb2
FAK
Src
lin
xil Kinase
Pa Pa
in
Tal in Zyxin xil
lin P130
ul culin in
nc in
lin

V s
Vi n
Ta

Te

focal adhesion

signals that control

cellular activities
actin
α-actinin

thought that that the various components of focal adhesions play different roles
in adhesion, activation, and maintenance of mechanical competence, although
the exact roles of each component are still being defined. Because they play
important roles in both structural integrity and intracellular signaling, they are a
major candidate for mechanotransduction sensors (covered in Section 11.2).
 ADHESION 281

Fluid shear can be used to measure adhesion Nota Bene

strength indirectly
Cell adhesion and muscular
dystrophy. Another ubiquitous cell
It is clear that adhesions play vital roles in cellular function. Therefore, it is of great
adhesion protein is the glycoprotein
interest to quantify their adhesive strength. Adhesive strength can be assessed dystroglycan. The dystroglycans exist
both directly and indirectly. Indirect measurements do not allow for the quantifi- with α, β, γ, and δ subunits. In skeletal
cation of actual forces required to detach cells from a given substrate, but they do muscle, the β subunit binds to actin
allow for relative comparisons of adhesive strength between multiple samples. via dystrophin, allowing the
dystroglycan complex to function as
One approach for indirectly measuring adhesive strength is to measure the level a transmembrane linkage between
of fluid shear stress required to detach adherent cells from a surface. If a large the extracellular matrix and the
enough shear stress is applied, the cells will begin to detach. As we saw from cytoskeleton. Mutations that result in
Equation 6.12, one way to apply a spatially varying shear stress over a uniform cell impaired binding of actin by
population is through a centralized flow that spreads out radially from a central dystrophin result in the muscle-
wasting disease muscular dystrophy.
inlet. The farther the cell is from the center, the lower the flow rate and the lower
the shear stress experienced by the cell. If cells are plated on a coverslip and
placed in a radial flow chamber there will be a circular patch of cleared cells. The
radius of the cleared area is inversely proportional to the adhesive forces of
the cells on the coverslip.
Parallel plate flow chambers can also be used for indirectly measuring adhesive
strength. Recall from Section 4.2 that we obtained the flow profile within infinitely
wide parallel plates. It can be shown that the surface shear stress on the bottom of
a parallel plate flow chamber is
µV0
τ = , (10.1)
h

where V0 is the velocity at a height h above the surface and μ is the fluid viscosity. Nota Bene
For a pressure-driven parallel plate configuration, Equation 10.1 becomes
Plate washing. An easy method for
6 µQ indirectly measuring adhesion
τ = , (10.2) strength is plate washing. In this
bh 2
approach, the same numbers of cells
are plated in multiple wells and
where Q is the volume flow rate, b is the chamber width, and h is the chamber allowed to attach for some fixed time
height. Different shear stresses can be applied to the same cell population by (typically short, such as 30 min). The
increasing the flow rate with time, or by constructing a chamber with a spatially wells are then washed several times,
varying geometry, such as one whose width changes along the direction of flow. and the remaining adhered cells in
each well are counted.

Detachment forces can be measured through direct cellular

manipulation
In contrast with the use of fluid shear stress for indirectly assessing adhesive
strength, the forces required to detach adherent cells from their substrate can also
be measured through direct cellular manipulation. One approach involves the
use of micropipette aspiration (Figure 10.2A). Suction pressure within a micropi-
pette is used to hold a cell and place it in contact with a surface with interest, such
as a functionalized bead. After binding of the cell to the surface, the cell is pulled
away from the surface, and the critical force required to separate the cell from the
surface is measured with a force transducer. This setup can be modified to meas-
ure cell–cell adhesion strength by using two micropipettes instead of one (Figure
10.2B). The two cells are brought into contact with each other, intercellular adhe-
sions are allowed to form, and the force required to separate the cells is measured.
This technique is particularly useful for cells that function normally in the absence
of a substrate, such as white blood cells.
For adherent cells, the lack of a substrate and a spread-out morphology may affect
processes associated with adhesion. Therefore, detachment forces in a non-
spread configuration may not be representative of physiological conditions. To
address this problem, microplate manipulation can be used (Figure 10.2C).
282 CHAPTER 10: Adhesion, Migration, and Contraction

Figure 10.2 Schematic depicting the (A)

use of (A) single pipette, (B) double
pipette, (C) microplate, and
(D) atomic force microscopy to
measure adhesion strength. The
arrows indicate the direction of motion
to attach/detach the cell to the probe or
to another cell. (B)

(C)

(D)

Nota Bene Microplate manip­ulation is similar to micropipette aspiration. Instead of using a

Dispase can be used to measure
vacuum within a micropipette to hold a cell in place, one uses a small island to
cell–cell dissociation. An assay which the cell adheres. This technique can be modified to measure cell–cell adhe-
unique to cell–cell adhesion is the sion through the addition of a second microplate manipulator. However, the force
dispase-dissociation assay. Cells are required to break the cell–cell adhesion must be smaller than the force required to
plated and allowed to grow to break the cell–matrix adhesion. Similar to microplates, atomic force microscopy
confluence. They are then detached has been used for measuring cell adhesion using the built-in cantilever to meas-
from the substrate using dispase,
ure the force and functionalizing the cantilever tip to facilitate cell attachment
which typically targets cell–substrate
receptors but not cell–cell receptor (Figure 10.2D).
interactions. The result is that the Another scheme is to use micropipette aspiration and micromanipulation to peel
cells are released from the substrate
an edge of the adherent cell off a substrate and then pull the cell back. This has the
as a sheet. This sheet of cells is then
vigorously mixed so that shearing advantage of being a more direct measurement of cell–substrate adhesive force.
forces separate the cell sheet into However, peeling is less a measure of whole-cell adhesive force than of individual
smaller sections. The size of the molecular strength in sequence. We consider this in more quantitative detail in
aggregates is a surrogate for cell–cell the following section.
adhesion.

The surface tension/liquid-drop model can be used to describe

simple adhesion
nLG We now shift our focus from experimentally measuring adhesive strength to
θ mathematical models of cell–substrate adhesion. As we will see, such models
nSG nSL can provide a great deal of insight into such mechanobiological functions as
the factors that govern whole adhesion strength or how cells control their mor-
Figure 10.3 Surface tensions for a phology through the expression of adhesive molecules. We begin this section
drop partly wetting a surface. This with a discussion of the energy of adhesion of liquid drops in contact with a
schematic depicts a close-up of the point solid surface. We have already discussed an application of the liquid-drop
of contact between the droplet and the model in Section 1.3, in which we found that, during micropipette aspiration,
substrate. At this point, there are three
some cells behave as a liquid droplet with a particular surface tension. In this
surface tensions between the solid,
liquid, and gas phases: one tangent to section, we will see that the liquid-drop model can also be integrated with
the solid–gas interface (nSG), one tangent force- or energy-based models of adhesion to gain insight into processes such
to the solid–liquid interface (nSL), and one as peeling behavior.
tangent to the liquid–gas interface (nLG).
The angle between the gas–liquid To begin our treatment, consider a liquid drop partly wetting a surface, as depicted
interface and the substrate is θ. in Figure 10.3. Recall that surface tension arises from an imbalance of molecules
 ADHESION 283

at the interface of two different phases. At the edge of the liquid drop, there will be
three surface tensions between the solid, liquid, and gas phases: one tangent to
the solid–gas interface (nSG), one tangent to the solid–liquid interface (nSL), and
one tangent to the liquid–gas interface (nLG). Assuming the drop is at equilibrium,
we can perform a horizontal force balance at the liquid-drop edge,

−nSG + nSL + nLG cosθ = 0. (10.3)

We now seek to calculate the energy required to “lift” the drop from the surface.
Physically, surface energy is the energy required to generate a new surface, and
arises due to the disruption of intermolecular bonds when new surfaces are cre-
ated. For a liquid, surface tension (which has units of force per unit length) and
the surface energy density (surface energy per unit area, which also has units of
force per unit length) are identical. The energy density of adhesion is

J LG = nLG + nSG − nSL . (10.4)

Substituting into Equation 10.3, we can rewrite Equation 10.4 as

J = n(1 + cosθ ), (10.5)

where J = JLG and n = nLG. Equation 10.5 is sometimes referred to as Young’s equa-
tion or the Young–Dupré equation. It implies that the energy density required to
“lift” the drop from its surface is dependent on only two factors, the surface ten-
sion n, and the angle that the membrane makes with the surface.
It is possible to derive a relation similar to Equation 10.5 by considering the internal
force within membrane receptors directly. It can be shown that the relationship
between adhesion energy density J and membrane tension n is
ρbFR L
= n(1 + cosθ ), (10.6)
2

where ρb is the receptor bond density, FR is the force that would rupture the recep-
tor–ligand bond, and L is the critical bond length at which the receptor is stretched
with force FR. Notice that this is essentially a reformulation of Young’s equation
(Equation 10.5).
From the form of Equation 10.6 we can make some predictions about the capacity
for cells to control the degree to which they spread. Consider a cell that is at adhe-
sive equilibrium as a half-dome on a flat surface. We can treat the cell as a liquid
drop with contact angle π/2 and membrane tension n = J0.
Equation 10.5 implies that, the adhesion energy density is J = J0. One might sup-
pose that to spread out more toward a flat pancake shape, the cell would be
required to increase the adhesion energy density indefinitely, but this is not the
case. In particular, if we increase the adhesion energy density by a factor of q, then
Equation 10.6 predicts that

qJ 0 = n(1 + cosθ ). (10.7)

But because we know n = J0, Equation 10.7 simplifies to

q = (1 + cosθ ). (10.8)

This implies that q is at most two, which occurs when θ is equal to zero. In other
words, doubling the energy adhesion density is sufficient to spread the cell to the
maximum extent allowed by the membrane tension. This example suggests that a
cell may only need to adjust its adhesion energy density by small amounts to pro-
duce substantial changes in spreading (for instance, by adjusting its receptor
bond density as in Equation 10.6).
284 CHAPTER 10: Adhesion, Migration, and Contraction

Advanced Material: Young’s equation can be derived using energy considerations

Equation 10.5 was derived using a combination of a force Combining Equations 10.9, 10.10, and 10.11, and solving
balance and energy considerations. One can also derive Equation 10.12, leads directly to Equation 10.3.
this relation solely using energy considerations. Consider
a droplet to be a truncated sphere of radius R, parame-
terized by the angle θ (Figure 10.4). A “cone” with vertex
angle θ will truncate the part of the sphere that we use to
model the drop, resulting in a contact angle of θ as well.
θ
Using our relation for energy density of adhesion given
in Equation 10.4, the energy of the drop can be written as

E = (nSL − nSG ) ASL + N LG ALG , (10.9)

where
ALG
ASL = πR sin θ (10.10)
2 2
θ
is the area of the flat surface with which the drop is in ASL
contact and
Figure 10.4 A droplet of water approximated as a
ALG + 2πR (1 − cosθ ) (10.11)
2 truncated sphere. The angle of truncation, or the
vertex angle of the cone, is θ. This angle is also the contact
angle of the droplet on the surface. The area of the solid–
is the total area of the “dome shaped” liquid–gas interface. liquid interface is “underneath” the droplet and the area of
To obtain Equation 10.3, we can use the fact that at equi- the liquid–gas interface is the “upper surface” of the
librium there is no change in energy with respect to ASL, droplet.

∂E
= 0. (10.12)
∂ ASL 

Adhesive peeling can be modeled using continuum mechanics

In the last section, we investigated the dependence of adhesion energy density
on membrane surface tension and geometry. In our analysis, we assumed that
the detachment of the cells from the surface occurs instantaneously. However,
when a cell is detached from a surface, the process often does not occur all at
once, but progressively. This process of peeling can greatly alter the force required
to detach a cell from its substrate. Consider the free-body diagram of a small sec-
tion of a membrane immediately adjacent to the attached section, as shown in
Figure 10.5.

Figure 10.5 A section of membrane n

being peeled off a surface. The adherent released
region to the left is still attached,
whereas to the right it has released. The
relevant balance of forces occurs
between the membrane tension, n, the
adhesive force per unit length, Fa, and
internal bending moments.

n fa
 ADHESION 285

As we have done in the past, we could create a free-body diagram for Figure 10.3.
However, we have already analyzed a more general case. Note that this situation is
entirely analogous to a plate bending owing to a distributed pressure. Recall
Equation 9.37,

 d 2w d 2w   d 4w d 4w d 4w 
n 2 + 2 
+ KB  4 + 2 2 + + Fa = 0, (10.13)
 dx dy   dx dx dy dy 4 

where KB is the bending modulus and we have replaced pz with Fa to indicate

adhesion force. Assuming that the peeling is occurring in the x-direction, we have

 d 2w   d 4w 
n  2  + K B  4  + Fa = 0. (10.14)
 dx   dx 

We can model the adhesion force with simple linear spring-like behavior such
that the force in a single bond is given by

 Fm 
 y 0 < y < lm
Fb =  lm  , (10.15)
 0 y < 0, y > lm


where Fb is the force for an individual bond, Fm is the maximum force generated
by the bond before bond breakage, lm is the length of the bond at the maximum
force, and w is the displacement (Figure 10.6). If w > lm, then the bond has bro-
ken and is no longer capable of sustaining force. Next, we observe that the total
adhesion force is simply the sum of the individual bond forces:

Fa = nb Fb , (10.16)

where nb is the area density of bonds attaching the cell to the surface. We can use
energy density of adhesion by defining

nb Fmlm
J = , (10.17)
2

the work done in extending all the bonds in a unit area from zero length to their
maximum length.

Fb Figure 10.6 The plot of the adhesion

force modeled as a linear spring.
For y less than the maximum extension
lm, the bond acts as a spring with spring
constant Fm/lm. At lm, however, the bond
Fm breaks and no restoring force is present.
Unlike a classical spring, the bond
cannot be compressed, so that if w is
less than zero, there is also no force
generated.

w
lm
286 CHAPTER 10: Adhesion, Migration, and Contraction

Advanced Material: Energy density of adhesion

To determine the energy density of adhesion, we first spring. For our example, the spring constant k is the slope
use the relationship that work equals force times dis- (Fm/lm) and x is the maximum length of the spring lm.
tance and that this work goes entirely into the adhesion Substituting yields W = 0.5 × Fm × lm. Because this is for a
energy density. Force, however, is not constant. It varies single bond, we can convert it to the desired energy density
with the length according to Equation 10.15. Therefore by multiplying by the number of bonds per unit area nb,
the work for a single bond is the integral of the force–­
displacement curve, or W = 0.5 × k × x2 for a classical W = 0.5 × nb × Fm × lm = J.

Introducing this energy allows us to rewrite the adhesion force density as

F   2J   y   2J 
Fa = nb  m  y =     =  2  y (10.18)
 lm   lm   lm   lm 

by using appropriate substitutions. Substituting into Equation 10.14, we obtain an
expression for adhesion in terms of y alone, with the remaining terms based on
physical parameters. Although this model is fairly explicit and can be (and has
been) used to model peeling, it relies on knowing parameters that are typically
difficult to measure. For this reason, many studies of adhesion rely on strict com-
parisons from direct force measurements rather than fitting to parameterized
models. Also, bond strength is known to depend on the level of prestress, which
this treatment ignores.

Adhesion energy density can be obtained through

consideration of strain energy
Previously, we obtained relations for adhesion energy density by considering sur-
face tensions associated with a liquid drop, as well as peeling of a two-dimen-
sional continuum. An alternative approach is to consider strain energy following
adhesion, because the energy associated with adhesion must overcome the
energy associated with cellular deformation.
Consider a spherical cell of radius R. After adhesion, the cell retains a mostly spher-
R
ical morphology, except for the base region contacting the surface (Figure 10.7). At
this region, the cell and substrate form a circular region of contact with radius a,
which is assumed to be much less than R. The deformation of the cell is assumed to
a occur solely because of the adhesion process.
Now consider the case in which we model the cell as a homogeneous isotropic
Figure 10.7 A mostly spherical cell linear elastic continuum with Young’s modulus E. It can be shown through Hertz
attached to a flat surface. The radius contact theory (described in Section 6.1) that the strain energy associated with
of the cell is R. At the contact surface, the this deformation scales as
cell and substrate form a circular region
of contact with base radius a, which we Ea 5 (10.19)
assume to be much less than R.
Wdef ∝ .
R2

The adhesion energy scales as

Wadh ∝ Ja 2 , (10.20)

where J is the adhesion energy density. Assuming that the strain energy is bal-
anced by the adhesion energy, then Wdef ~ Wadh. We can relate the right handed
sides of the expression in Equations 10.19 and 10.20 as
Ea 5 (10.21)
∝ Ja 2 ,
R2
 ADHESION 287

which simplifies to

JR 2 (10.22)
a∝ 3 .
E

This is only a scaling relationship, and is only valid for a relatively small cellular
deformation occurring upon adhesion. Still, it allows us to predict some interest-
ing scaling behavior. Consider what happens to adhesion contact if the cell radius
decreases by a factor of two, but all other parameters remain constant. A reduc-
tion in R by a factor of two would result in an approximately 1.6-fold reduction in
adhesion contact radius. If the elastic modulus increases by a factor of two the
adhesion contact radius decreases by approximately 1.3-fold. We can see that the
contact radius of the cell is more sensitive to changes in its radius than changes in
its stiffness, a somewhat unexpected result.

Targeting of white blood cells during inflammation involves

the formation of transient and stable intercellular
adhesions
We conclude our discussion on adhesion by analyzing the mechanics of cell–cell
adhesions. Intercellular adhesions are involved in a variety of biological pro-
cesses. For example, intercellular adhesions are critical for the formation of
impermeable or semipermeable cell linings within tissues. They are also neces-
sary for the formation of structures that facilitate direct intercellular communica-
tion, such as gap junctions, specialized channel pairs that allow for the flow of
signaling molecules from one cell to another. Cell–cell adhesion in known to be
mediated by many transmembrance proteins, including cadherins, junctional
adhesion molecules, some integrins, and desmosomal proteins, each of which
has associations with many other proteins.
One of the best-characterized processes associated with intercellular adhesion is
the interaction of white blood cells with blood vessel walls during inflammation.
White blood cells are key players in the body’s immune response to infection. By
flowing with blood though the circulatory system, they have rapid access to almost
all of the tissues and compartments of the body. However, we will see that they are
able to exit the bloodstream quickly and target specific locations by processes
mediated by intercellular adhesions (Figure 10.8).
One may observe the formation of two distinct adhesions in neutrophils interact-
ing with a vessel wall (Figure 10.9). Initially the cells of the endothelium become
activated. Endothelial activation is marked by the expression of a class of recep-
tors called selectins that bind glycoproteins coating the surface of the white blood
cell. Neutrophils that bump into the activated endothelial cells are held in contact
with them by the selectins. However, the selectin bonds have relatively low affinity
and are not strong enough to fully resist the hemodynamic drag force. Rather, the
Nota Bene
neutrophil begins to roll along the endothelial layer with selectin bonds forming
and breaking transiently. Selectin nomenclature. The three
types of selectins are named after
In the second step, the neutrophil becomes activated by signals released by the the cell types in which they were first
endothelial cells. Activation of the neutrophil is marked by expression of integ- characterized. E-selectin is found on
rins on the neutrophil membrane. Unlike selectins, which bind to the glycopro- the surface of endothelial cells, and
L-selectin on leukocytes. P-selectin
tein layer coating the cell, integrin-mediated intercellular adhesion is much was first found on platelets, but
stronger and eventually halts the neutrophil rolling, resulting in a so-called firm later discovered to be expressed
adhesion. During this halting period, formation of tethers have been observed by endothelial cells as well. The
(Figure 10.10). These tethers are small elastic strands that serve to help arrest the feature they all have in common is
neutrophils. After arrest, the neutrophils then migrate to the junctions between that they all bind to sugar molecules
endothelial cells and perform paracellular migration, squeezing between the in the glycoprotein coating of the
target cell.
cells to exit the bloodstream.
288 CHAPTER 10: Adhesion, Migration, and Contraction

Figure 10.8 In vivo image of

neutrophils passing through a 20.0 µm
blood vessel. The neutrophils are in
various stages of rolling and attachment.
(Courtesy of Gustavo Menezes; see
www.garlandscience.com for a color
version and movie of the figure.)

neutrophils

Nota Bene Kinetics of receptor–ligand binding can be described with the

Mediators of firm adhesions. The law of mass action
integrins involved in firm adhesion
have been identified as α4β1 and Given that neutrophil adhesion to an endothelial cell layer involves two types of
αLβ2. They bind transmembrane interaction, transient binding mediated by selectins and more stable bonds
members of the immunoglobulin ­mediated by integrins, we now seek a means to characterize the difference in
superfamily (which also includes kinetics between these interactions. We will start by adopting a descriptive
antibodies), vascular cellular approach taken from chemical kinetics. A ligand–receptor pair can be in a bound
adhesion molecules (VCAMs), (B) or unbound (L + R) state. We arbitrarily consider the rate at which binding is
intercellular adhesion molecules
occurring to be the “forward” or “ + ” direction of the reaction and therefore call
(ICAMs), and nerve cellular adhesion
molecules (NCAMs). the binding rate constant k+ and the unbinding rate constant k−. The reaction can
be written as analogous to Equation 7.1,
k

L +R
+
 B. (10.23)
k
−

Figure 10.9 Schematic endothelial neutrophil neutrophil firm

demonstrating the stages of activation trapping activation adhesion
neutrophil adhesion in a blood and rolling
vessel during an inflammatory
reaction.

endothelial layer

selectin unactivated integrin

selectin ligand activated integrin
intercellular adhesion molecule (ICAM)
 ADHESION 289

(A) rolling time < 1 min (B) rolling time > 4 min Figure 10.10 Neutrophil tethering at
various fluid shear stresses on a
surface coated with P-selectin.
(A) and (C) show the tether formation
<1 min of attachment, and (B) and
(D) show tethering after >4 min of
attachment, so the tethers are more
2 dynes/cm2 fully developed (From, Ramachandran
V et al. (2004) Proc. Natl. Acad. Sci. USA
with permission from the National
Academy of Sciences.)

flow
(C) (D)

8 dynes/cm2

As we saw in Section 7.2, the kinetics of this reaction can be described quantita-
tively through the so-called law of mass action. Specifically, the assumption,
which can be obtained from statistical mechanics by calculating the rate at which
reactants would randomly collide, is that the rate at which a reaction occurs is
linearly related to the rate constant and the concentration of the reactants. In our
example of two reactants, the forward reaction occurs at a rate of

k+ [L][R] (10.24)

and the reverse at

k−[B], (10.25)

where the square brackets indicate concentrations. If the reaction is at or near

equilibrium, the forward and reverse reactions occur at nearly the same rate, or

k+ [L][R] = k−[B]. (10.26)

The equilibrium constant, or dissociation constant, is defined as the ratio of the

two reaction constants, analogous to Equation 7.4,
k− [L][R]
Kd = = . (10.27)
k+ [B]

The dissociation constant has the unique property that when the concentration of
ligand is equal to Kd, half of the receptors will be bound. This can be better under-
stood by computing the fraction of bound receptors as
[L]
[B]
[B] [B] [L]
%bound = = = . (10.28)
[B]+[R] [B] [L] +[R] [L] [L]+K d
[B] [B]
290 CHAPTER 10: Adhesion, Migration, and Contraction

Figure 10.11 The percentage of 100

bound reactants plotted as a
function of ligand concentration.
When the ligand concentration is Kd,

percentage bound
exactly half of the receptors will be in
the bound state.
50

Kd [L]

Notice that when the [L] = Kd, %bound = 0.5. A plot of Equation 10.28 can be seen
in Figure 10.11.

The Bell model describes the effect of force on

dissociation rate
Although the dissociation constant is a useful way to describe the behavior of
ensembles of bonds on long timescales, it does not take into account the effects
of force. Consider the kinetics of force-induced bond rupture between a single
receptor–ligand pair. Characterization of such kinetics can be performed using
experimental approaches such as micromanipulation or atomic force micros-
copy. After a single bond is identified, a controlled force can be applied in what
is known as a force clamp configuration. This allows for the generation of force–
displacement curves (Figure 10.12). At  the single-molecule scale, entropic
influences become extremely important. If we were to probe a single ligand–
receptor interaction, even in the absence of an applied force, thermal forces will
eventually cause the bond between the two molecules to dissociate. In general,
the time that a bond exists decreases as the force across the bond increases.
In 1978 George Bell advanced a theory to describe how force would affect the rate
of dissociation. He assumed that when a bond is subjected to loading, once it is
ruptured the receptor and ligand move too far away from each other to rebind (in
other words k+ = 0). This means that the relevant kinetics can be described only by

Figure 10.12 A typical force– F (pN)

displacement curve from a single-
molecule binding experiment. 80
Notice that initially some force is
required to push the receptor and ligand
together. Then, as the bond begins to be
pulled apart, a tension force is evident
that was not present previously. If the
force is the same as before contact, no 40
bond is formed. Alternatively, if the single
adhesion fails in a series of steps, it bond multiple
suggests that more than one bond is bonds
formed. Once an isolated bond is
identified, the investigator can apply a
contact
fixed force and measure the bond
lifetime. 0
no bond

t
 ADHESION 291

the rate of dissociation, k−, and does not depend on k+. Next, he assumed that
force has an exponential influence on bond rupture such that the rate of dissocia-
tion in the presence of a force F across the bond is
σF
(10.29)
k− = k e0 kBT
− ,

where k−0 is the rate of dissociation in the absence of force, kB is Boltzmann’s con-
stant, T is absolute temperature, and σ is a constant that characterizes the influ-
ence of force. Notice that the form of this relationship is very similar to the
Boltzmann probability derived in Section 5.5. Because the denominator in the
exponent has units of energy, σF must also have units of energy, and therefore σ
must have units of length. Indeed, it can be shown that σ is the bond extension
that is associated with its minimum potential energy configuration. Equation
10.29 can also be expressed as
F
kBT (10.30)
k− = k e , FB =
0 FB
− .
σ

FB is a characteristic measure of bond strength, although not in the traditional

sense, because bond dissociation is always occurring at any force level. Rather, it
is a measure of the influence that force has on unbinding kinetics. It is noteworthy
that in the Bell model, the time evolution of the concentration of bound mole-
cules is also described by an exponential relation. Specifically, since only bond
rupture is considered and rebinding is not allowed,

d [B]
= − k − [ B ]. (10.31)
dt
The solution of this differential equation is an exponential of the form

[B(t )] = [B(0)]e− k t . −
(10.32)

There are actually two exponentials in Bell’s model, one describing the influence of
force on the rate of dissociation, and another describing the time evolution of [B].

Shear enhances neutrophil adhesion—up to a point

From what we have discussed so far, we would expect that the rate at which bonds
are broken increases with the force across the bond. Indeed, for most bonds, this is
the case. In other words, k− ∝ F. However, in experiments in which neutrophil
adhesion and rolling was examined quantitatively, a surprising result was observed.
As flow increased, the cells were actually observed to adhere better. This was
reflected in the number of adherent cells and the number of tethers per cell increas-
ing with flow, and rolling velocity decreasing with flow (Figure 10.13). In fact, up to
a certain value of wall shear stress, adhesion seems to increase linearly. Beyond
this shear threshold, adhesion again decreases linearly with flow, as expected.
For some time the mechanism behind the shear threshold phenomenon was
unknown, but it seemed to suggest the existence of catch bonds. The term “catch
bond” is meant to suggest that some pulling is required for the bond to engage
or “catch.” Without the pull, the bonds tend to float around engaging and disen-
gaging randomly. In our everyday life, this might be similar to picking up a
shopping bag with a hooked finger. Perhaps the receptor binding involves the
engagement of hook-shaped molecules? Such bonds have been shown to exist
experimentally at the single-molecule level. P-selectin exhibits catch behavior
up to approximately 25 pN. Beyond this level, lifetime dramatically decreases in
a behavior termed slip. Such bonds are sometimes called catch-slip, reflecting
the two types of relationship between bond lifetime and force.
292 CHAPTER 10: Adhesion, Migration, and Contraction

Figure 10.13 Plot of the number of

adherent cells as a function of

number of adherent cells

shear stress during neutrophil
adhesion and rolling. Up to a certain
shear stress, the number of adherent
cells increases with increased shear.

1 2 3 4
shear stress (dynes/cm2)

10.2 MIGRATION
Cell migration can be studied in vitro and in vivo
Cell migration is a fundamental aspect of cellular behavior that is central to pro-
cesses of immunity, regeneration, repair, inflammation, and cancer, to name a
few. In vitro assays, such as tracking motility or movement through a polymer gel,
are powerful reductionist approaches. However, in vivo strategies are becoming
more common. One elegant study was performed on heart transplant patients. Of
the people who had heart failure significant enough to require transplantation,
there were eight males who received hearts from female donors. After the male
patients died (mostly within a year, some after nearly 2 years), their female hearts
were assayed for the Y chromosome. It was found that a substantial fraction of
organ cells (up to 10%, depending on the type of cell) were from the male recipient
(turning the organ into a “chimeric tissue”, Figure 10.14). Although the source of
those cells is not known, such changes can influence other readouts, including
migration.

Cell locomotion occurs in distinct steps

Before we discuss quantitative approaches for analyzing cell locomotion, we
must first introduce some background biology of this process. Cell locomotion

FPO
Figure 10.14 A section of heart
showing the presence of a Y
chromosome. The bright spot (see
arrow) is a Y chromosome detected by
in situ hybridization in a female heart
transplanted into a male patient. (From
Quaini F et al. (2002). New Engl. J. Med.
346. With permission from the
Massachusetts Medical Society.)
 MIGRATION 293

can generally be delineated as a series of four steps: protrusion, attachment,

translocation, and release (Figure 10.15). Protrusion occurs when the cell
extends actin-rich projections outward from its current location. There are sev-
eral such projections that can generally be classified by their gross morphol-
ogy. For example, lamellipodia are flat, broad, veil-like extensions containing
highly branched actin ­networks at the leading edge of crawling cells. As we
learned in Section 8.4, filopodia are finger-like extensions containing cross-
linked actin bundles. The process of protrusion has been likened to an explora-
tory process by which the cell “feels” new surfaces and “determines” the
direction of motion. The second step is attachment, which occurs when the cel-
lular extension forms a stable adhesion with the surface such that it can be
used as an anchorage point for subsequent locomotion. Although this adhe-
sion is generally stable, it need not be permanent, as such adhesions are often
observed to be transient and withdrawn after the third step, translocation. In
this step, the cell moves in the direction of attachment. Because this process
involves a substantial amount of cell motion, it is often characterized by a large
degree of actin–myosin contractile activity. The fourth step is release, or detach-
ment of the trailing end of the cell. This release is often associated with with-
drawal of adhesions formed during attachment. However, it can sometimes be
more forcible. If the cell’s motion is faster than the release of the focal contacts
in the trailing end, little bits of cellular material can be broken off and left
behind. Though we have delineated cell locomotion as four sequential steps, it
is important to note that the cell can be undergoing more than one step at any
given time.

Protrusion is driven by actin polymerization

The generation of intracellular force is critical for cellular locomotion, particularly
during the processes of protrusion and translocation. Translocation is accepted to
be largely governed by activation of actin–myosin contraction, a process that we
will discuss in great detail in Section 10.3. However, the molecular mechanisms
driving cellular protrusion are less clear.
In Figure 10.16, a schematic depicts what are believed to be the primary molecu-
lar mediators of protrusion of lamellipodia. A key member of this process is the
Arp2/3 complex, which consists of two primary molecules, Arp2 and Arp3, as well
as several other binding molecules. Interestingly, the Arps (short for actin-related
proteins) are structurally similar to actin and can serve as nucleation sites for the
creation and extension of new filaments. In lamellipodia, Arp2/3 serves to initiate
new actin branches on existing actin filaments, resulting in the formation of a

Release Translocation Attachment Protrusion Figure 10.15 Schematic showing

the major steps for cell migration.
actin filament actin/myosin focal adhesion actin
In this depiction, the cell is migrating
contraction and contraction formation polymerization toward the right.
focal adhesion
sacrifice movement

retraction filopod lamellipod

edge

contracting new focal

focal adhesion actin/myosin adhesion
about to be lost filaments
294 CHAPTER 10: Adhesion, Migration, and Contraction

6 Growing filaments push

membrane forward
2 Produce active
1 Extracellular GTPases
stimuli and PIP2
5 Barbed
ends elongate
70°
7 Capping protein
3 Activate terminates elongation
WASP
4 Activate Arp2/3
complex to initiate
new filaments
Activated GTPases
and PIP2 8 Aging

WASP/Scar
ADF/cofilin
11 Pool of ATP–actin
ATP–actin
bound to profilin
ADP–actin
9 ADF/cofilin severs
Profilin and depolymerizes
ADP–actin filaments
Capping protein
Arp2/3 complex

10 Profilin catalyzes
exchange of ADP
for ATP

Figure 10.16 Stages for protrusion of the cell membrane in lamellipodia. In this model, external stimuli (1) result in the
activation of GTPases (2), which activate Wiskott–Aldrich syndrome protein (WASP). (3). This complex activates the Arp2/3 complex (4),
resulting in the formation of a new actin branch on a pre-existing actin filament. This newly formed actin branch grows by
polymerization at the barbed end of the filament (5) from a pool of profilin-bound actin monomers, pushing the membrane forward (6).
Elongation of actin filaments is terminated by capping protein (7), which binds to the growing ends of polymering actin filaments.
Older ADP-bound filaments (8) are severed by actin-depolymerizing factor (ADF)/cofilin (9). Depolymerized actin undergoes
dissociation of ADP and binds ATP (10), upon which ATP–actin re-binds profilin, renewing the pool of ATP-bound actin monomers
available for assembly (11). (Adapted from Pollard TD (2003) Nature. 442.).

branching actin network. Multiple sites for actin extension permit a relatively uni-
form growth pattern during protrusion to be achieved. We will see in the next sec-
tion that although it is generally accepted that actin branches serve to push the
membrane forward by polymerization, the exact physical mechanism by which
this may occur is still relatively unclear.

Actin polymerization at the leading edge: involvement of

Brownian motion?
A major question about the capacity for actin polymerization to give rise to pro-
trusion at the leading edge is how monomers can be added to the filament ends
when they are physically blocked by the membrane? Although it is possible that a
polymerizing actin filament may generate sufficient force to deform the cell mem-
brane locally, this does not address how physical space is created between the
filament end and the membrane such that there is room for monomers to be
added to the growing end of the filament.
Several mechanisms have been proposed to explain how actin monomer addition
can occur at the leading edge of a protruding membrane. One such mechanism is
often referred to as the Brownian ratchet (Figure 10.17). Here, entropic forces cause
 MIGRATION 295

Figure 10.17 The Brownian ratchet

depends on thermal fluctuations. As
G-actin the membrane randomly fluctuates,
nothing if the it shifts closer to the tip of the
fluctuation
happens actin filament, nothing happens because
leftward the actin monomers are blocked by the
membrane. As the membrane shifts
away and opens a gap, the actin
monomers can extend the actin
filament. This has the effect of
F-actin generating a local protrusion.
fluctuation extension
rightward of F-actin
G-actin

Brownian
fluctuation
of membrane
Nota Bene
Somebody stole my actin! Listeria
monocytogenes is an intracellular
the local cell membrane to undergo small thermal fluctuations. If a pool of free actin bacterium responsible for the disease
listeriosis, the most common cause
monomers is available at the leading edge, when the membrane shifts forward, space
of death associated with food-borne
is temporarily generated that allows polymer extension. This polymerization pre- pathogens. As an intracellular
vents the membrane from shifting back. As this ratcheting process is repeated, the bacterium, it resides within the
resulting actin polymerization leads to the development of a protrusion. cytoplasm of another cell, remaining
relatively hidden from the immune
An alternative hypothesis that has been advanced is related to the Brownian ratchet system. To maximize its capacity to
in that the actin filaments themselves rather than the membrane are undergoing move within the cell, the bacteria has
thermal fluctuations. The filaments randomly fluctuate in end-to-end length. developed a mechanism whereby it
When the end-to-end length decreases slightly (through bending), space is tempo- “hijacks” the actin polymerization
rarily generated that allows for polymer extension (Figure 10.18). When the fila- machinery in the cell and rides on a
ment straightens, this generates a protrusive force that extends the membrane. wave of polymerizing actin that it
induces. This polymerizing actin can
often be viewed as a “comet-tail”
Cell motion can be directed by external cues structure behind the bacterium. The
polymerizing actin propels the
We have considered the process of cell migration as if it were an aimless process. bacterium to the membrane and
However, it is clear that, cell migration can be directed to specific locations. generates a protrusion. If the infected
Chemical compounds that direct cell migration are called chemoattractants. If cell is adjacent to another cell, the
bacterium can then enter the second
there is a chemical gradient in a chemoattractant, cells may sense these gradients cell by endocytosis or a similar
and direct their migration toward the source. One of the most prominent exam- mechanism. This method of
ples of directed cell migration toward a chemoattractant is the targeting of patho- propagation has the advantage that
gens by neutrophils. The pathogens release compounds that the neutrophil the bacterium can travel from cell
recognizes as foreign. Upon activation, the neutrophils crawl about, targeting to cell without ever leaving the
local pathogens. Upon contacting the pathogens, the neutrophils engulf them, intracellular environment.
removing the source of chemoattractant.

Figure 10.18 Membrane protrusion

formation from polymer thermal
fluctuations. In this mechanism, the
membrane is considered mostly static,
whereas the actin filament undergoes
fluctuation thermal fluctuations, resulting in flexing
or
F-actin in filament of the filament. This flexing creates more
space at the tip of the filament, allowing
for polymerization to extend the
G-actin filament. Then, when the filament
straightens out again, it generates a
protrusive force to generate a
extension protrusion.
296 CHAPTER 10: Adhesion, Migration, and Contraction

Figure 10.19 Plot showing amoeba 1.0

chemotaxis “accuracy” versus

accuracy of chemotaxis (K)

chemoattractant (in this case, cyclic
AMP [cAMP]) concentration. The 0.8
accuracy is measured between 0 and 1,
with 0 representing random motion and 0.6
1 representing fully directed motion.
Cells demonstrate the ability to move
fairly unerringly toward the source of the 0.4
chemoattractant given the right
concentration. Note the bi-phasic
0.2
response; at higher concentrations, it is
possible that enough chemoattractant
exists to engage all cell receptors, 0.0
thereby “blinding” the cell to existing –12 –11 –10 –9 –8 –7 –6
chemical gradients. (From, Fisher PR, log cAMP concentration (M)
Merld R, & Gerisch G, (1989) J. Cell Biol.,
Rockefeller University Press.)

The physical and molecular mechanisms driving chemoattraction have yet to be

fully established, though several proposed mechanisms have been put forth. It
has been proposed that cells may sense local chemical gradients across the cell
surface through the differential activation of receptors from one end of the cell to
the other. When one considers that cells are typically only tens of micrometers
across, this implies that very small differences in activation from one end of the
cell to other must be distinguished. This is a substantial challenge and what the
direction-sensing mechanisms might be is still unclear.
Another possible mechanism is that the cell randomly travels over a given dis-
tance in many different directions, allowing it to sample the local chemoattractant
concentration over a fairly large area. It is not yet known whether cells correct
themselves after moving in the direction of decreasing concentration or if they
sense whether an increase might be improved upon in a different direction. It is
important to note that such a mechanism requires some degree of random migra-
tion, suggesting that at times they will be incorrectly oriented. In contrast, most in
vitro experiments suggest that in the right conditions, cells move fairly unerringly
toward the source of the chemoattractant (Figure 10.19).

Cell migration can be characterized by speed and

persistence time
In investigating the process of directional cell migration, one question is how do
we characterize the degree to which a cell tends to travel in the same direction as
high low
persistence persistence it migrates through space? It is often useful to characterize cell trajectories using
similar approaches to characterizing a diffusing particle (Section 5.6) or a fluctu-
low ating polymer (Section 7.4). Cellular trajectory can be quantified in terms of per-
speed
sistence time (P) and speed (S) (Figure 10.20). In the context of cell migration,
speed refers to how fast the cell is moving, whereas persistence time refers to the
time a cell spends moving in a given direction. That is, a cell might exhibit large
summed displacements with little net translation, which typically correlates with
high low persistence times.
speed
To better understand the influence of persistence time on cell migration, consider
a cell migrating in one dimension with speed S and a change in direction per unit
time of λ. The directional persistence time of the cell, defined as the time per
Figure 10.20 Schematic
demonstrating persistence versus direction change, is P = 1/λ. It can be shown using a random walk model that the
speed. The traces show different cell differential equation governing the time dependence of the mean square distance
trajectories for a given time period. of the cell 〈d 2〉 is
Speed is a measure of how fast a cell is
moving. Persistence is related to how ∂2 d 2 2∂ d
2

likely it is to continue to move in the + = 2S . (10.33)

same direction. ∂t 2 P ∂t
 MIGRATION 297

The particular solution of Equation 10.33 is

d 2 = S 2 Pt + C1 + C 2e −2t / P , (10.34)

where the constants C1 = −C2 = −2S2P2 are found by applying the initial condi-
tions, 〈d 2〉 = 0 at t = 0, and d〈d 2〉/dt = 0 at t = 0 (the cell initially has no bias). In
two dimensions,

( (
d 2 = S 2 Pt − P 2 1 − e −t / P . )) (10.35)

Note that in the limit for t ⪢ P

d 2 = 2S 2 Pt = 2Dt , (10.36)

where D = S2P has the units of m2/s and can be considered to be an effective “dif-
fusion coefficient” of the cell.
To determine the values of S and P experimentally, one can acquire multiple cell
paths for different cells with various time intervals, and compute an average mean
square distance for each time interval. With these data established, the speed and
persistence can be calculated using standard least squares fitting algorithms. In
this approach, it is assumed that every cell is migrating under similar mathemati-
cal patterns. Some examples of speed and persistence relationships for various
cell types are presented in Figure 10.21.

1000 Figure 10.21 Speed–persistence

relationships for different cell
types. Note that there is a roughly
microvessel inverse relationship between speed and
endothelial cells persistence. It also noteworthy that
neutrophils have high speed for low
persistence times, suggesting a need to
smooth move quickly to reach an infection site
muscle cells but also to change directions quickly to
100 target pathogens. (From Lauffenburger D
fibroblasts & Linderman JJ (1993). With permission
persistence (min)

from Oxford University Press, New York.)

melanoma
cells
alveolar
macrophages

10

neutrophil
leukocytes

1
0.01 0.1 1 10 100
speed (µm/min)
298 CHAPTER 10: Adhesion, Migration, and Contraction

Directional bias during cell migration can be obtained from

cell trajectories
Persistence time can be used to directly compare the migration behavior of differ-
ent cell types. One limitation of persistence time is that it does not take into
account whether there is bias in the direction in which a cell travels, under the
influence of a chemoattractant. How would we quantify such a tendency?
A relatively simple method is to determine the position of a cell at set time
intervals, fit straight-line segments between the points, and determine the
change in angle from one segment to another. The average change in angle can
then be computed over all the segments. This quantity can be used to distin-
guish left- or rightward bias during cell migration. Also, it can be used as a fairly
simple means to obtain cell speed and a rough indication of persistence during
migration.

Example 10.1: Migration paths for two different cells

Consider the trajectories of two cells shown in Figure angles of −102° between 0 < t < 2δ t and −99.2° between
10.22. Both start at (1, 0), but one cell follows the x-axis 1 < t < 3δ t. The average angle is −100.6°. The large magni-
(cell 1) while the other does not (cell 2). Determine the tude and negative sign indicate a prominent leftward bias
average speed of each cell, and show quantitatively that in its migration.
cell 2 has a leftward bias in its migration.
First, we consider the cell speeds. They are not constant y
throughout their trajectories. For each successive time
interval of δ t, cell 1 moves 1 unit, then 0.8 units and
(–0.2, 0.4) (1, 0.3)
then 1.1 units. The average speed for cell 1 is then
2.9/3δ t = 0.97/δ t. The average speed for cell 2 can be
determined similarly, and is 0.99/δt. The speed of cell 2 x
is marginally higher than that of cell 1. (0, 0) (1, 0) (1.8, 0) (2.9, 0)

Now we seek to demonstrate that cell 2 has a leftward bias. (0.8, –0.5)
We do this by calculating the change in angle the cell
makes at each successive time interval. In doing so, we Figure 10.22 Migration paths for two different cells. The
define the angle relative to a line parallel to the line seg- values of x in the plot refer to cell positions obtained at time
ment associated with the previous time iteration. Leftward intervals separated by a constant, δ t. The (x, y) coordinates of
changes in orientation result in negative angles, and right- each point is given in parentheses.
ward changes result in positive angles. For cell 2, we find

10.3 CONTRACTION
In the final section of this chapter, we discuss the generation of intracellular
contractile forces. When we talk about cellular contraction, one of the first
examples that comes to mind is muscle cells, which possess specialized struc-
tures for the generation of contractile forces. However, non-muscle cells also
possess the capacity to generate contractile force, such as during locomotion, or
in the generation of traction forces. In this section, we give overviews of the
molecular structures within muscle and non-muscle cells that allow for con-
tractile force generation, as well as experimental measurements of force genera-
tion in both cases. We also present molecular-based mathematical models of
contraction that allow us to gain mechanistic insight into a range of observed
phenomena.
 CONTRACTION 299

Muscle cells are specialized cells for contractile

force generation
One of the most important examples of cellular force generation is when cells act
as an ensemble and generate force within muscle. In this section, we will con-
sider how muscles generate force motivated first by cardiac heart muscle and
then skeletal muscle. Both cardiac and skeletal muscles are examples of striated
muscles. All other muscles, including blood vessels, and the reproductive gastro-
intestinal, and respiratory tracts, are smooth muscles. The basic contractile unit
of a striated muscle fiber is called the sarcomere (Figure 10.23). It consists of
several key structures, including the anchoring z-discs (which are non-contrac-
tile structural proteins), a bundle of myosin filaments forming a “thick” filament,
and actin filaments that are anchored to the z-discs. In striated muscles the
z-discs and M-line (between the z-discs) of the sarcomeres align, giving the mus-
cles a banded appearance. In smooth muscles, this alignment does not occur
and, generally, smooth muscles are able to undergo much larger amounts of
elastic stretch.

Studying cardiac function gave early insight into

muscle function
The foundations of understanding of muscle force generation were laid from
basic insights into the pumping action of the heart. These occurred in the early
twentieth century, long before we understood the molecular mechanism of
actin–myosin force generation or even the physiology of skeletal muscle. The
physiologist Ernest Starling, in 1914, anticipated the relationship between force
and sarcomere length when he stated, “the mechanical energy set free in the
passage from the resting to the active state is a function of the length of the fiber.”
Specifically, using data collected from intact hearts, he constructed diagrams of
the cardiac cycle. He realized that during the diastolic filling phase of the cycle,
cardiac muscle fibers are passively stretched, and that during the systolic ejec-
tion phase the fibers are actively contracting. Therefore, volume can be inter-
preted as fiber length, and pressure as fiber tension. The Frank–Starling law (or
perhaps more properly the Frank–Starling mechanism) is the observation that
an increase in the filling volume of the heart (reflected in a higher-end diastolic
volume) causes more stretching of the heart muscle fibers and a more forceful

sarcomere

thin filament z-disc

thick
filament

M-line

pointed end
barbed end
Figure 10.23 A schematic of a sarcomere, the basic contractile unit of a muscle fiber. The z-discs (which are commonly
visible in pictures of muscles) anchor the actin filaments. The myosin bundle (thick filament) with the myosin heads sticking out
(in opposite directions on either end) is caged inside the actin (thin) filaments. (Adapted from, Bray D (2000) Garland Science,
New York.)
300 CHAPTER 10: Adhesion, Migration, and Contraction

ejection and a higher systolic pressure. This led him to the realization that the
tension a sarcomere is able to generate must increase with its initial stretch or
length (Figure 10.24).

The skeletal muscle system generates skeletal forces for

ambulation and mobility
Because of the relative ease of isolating skeletal muscle, much of what we under-
stand about muscle contraction and actin–myosin interaction is derived from the
study of skeletal muscle. Using isolated muscles it is possible to quantify the force
a muscle generates as a function of excitation, relative length, and contraction
velocity. Isotonic contractions occur when muscle tension is held constant and
length is changed (if the muscle is acting to support or raise a fixed weight).
Isotonic contractions might be concentric (the muscle is shortened) or eccentric
(the muscle is lengthened). Isometric contractions occur when the muscle is held
to a fixed length. Activation can take the form of a single electrical stimulation that
produces a transient force known as a twitch, or continual stimulation can pro-
duce a maximal force in what is termed a tetanic contraction.
The two most important mechanical parameters affecting muscle force genera-
tion are length and velocity (Figure 10.25). It is perhaps not surprising that the
force generated by a muscle depends on length such that if the muscle is stretched
or shortened too much, the ability to generate force is compromised. In terms of
velocity, the maximum force is generated when the muscle is not allowed to con-
tract, and it is reduced as velocity is increased.

The Hill equation describes the relationship between muscle

force and velocity
The classic hyperbolic relationship between muscle force and velocity is the Hill
equation,

( F + a)(v + b) = k , (10.37)

where F is force, v is velocity, and a, b, and k are constants. This equation captures
critical aspects of the force–velocity relationship. As force increases, velocity
decreases. There is a maximum force at zero velocity and, vice versa, a maximum
shortening velocity at zero force. Between these two extremes the relationship is
hyperbolic and has been shown to provide a remarkably accurate fit to experi-
mental data.
Interestingly, Hill did not arrive at his famous equation by making measurements
of force and shortening velocity directly. He actually made measurements of the
heat liberated by muscle as it shortened. These measurements were based on his

Figure 10.24 A schematic of the

cardiac cycle. The Frank–Starling law Psystole systole
states that as the heart’s stroke volume
increases owing to more diastolic filling,
the muscle contracts with more force, ↑EDV, ↑Psystole
leading to a higher systolic pressure.
pressure

This is depicted by the blue line with a

higher EDV, resulting in a higher systolic
pressure, Psystole.
End-Diastolic-Volume (EDV)
diastole
volume
 CONTRACTION 301

(A) (B) Figure 10.25 Length and velocity

control the ability of muscle to
generate force. (A) Plot of force versus
isometric length. A biphasic dependence of force
contraction on length is observed. (B) Plot of force
versus velocity. Force generation
increases with decreasing velocity.
force

force
Maximum force generation occurs
during isometric contraction, or when
the muscle is held to a fixed length.
Maximum contraction velocity occurs
with zero force.

length velocity
maximum
contraction
velocity

observation that the heat released is proportional to shortening distance xsh

through a proportionality constant he denoted as “a”. Next, he considered the Nota Bene
total energy liberated, El, by the muscle during a contraction. This energy contains
two parts, the heat released (shortening distance times the proportionality con- Hill’s measurements. The English
physiologist Archibald Hill collected
stant) and the mechanical work done (force times the shortening distance), or meticulous data describing the
relationships between muscle force,
El = Fxsh + axsh . (10.38) lengths, and velocity. He made these
measurements by quantifying the
Experimentally he found that the rate at which this energy is liberated decreases heat liberated by muscle contraction
linearly with force, to a precision of 10−3 °C. He also
advanced the equation that bears
dxsh his name to quantify the relationship
(F + a) = (F + a)v = −bF + c . (10.39) between muscle force and
dt contraction velocity. His 1938 paper
is a classic in terms of the scientific
This finding can be easily rearranged to yield, contribution, but also the elegance
of the writing is a product of a

( F + a )(v + b ) = c + ab = k. (10.40) bygone era.

Non-muscle cells can generate contractile forces within

stress fibers
We now shift our focus from contractile force generation in muscle to non-muscle
cells. Although the basic molecular machinery responsible is similar, the molecu-
lar composition and organization of the structures can be quite different. Stress
fibers are long thick bundles of actin that were first identified in cells cultured on
adhesive surfaces. They span the interior of the cell, interconnecting focal adhe-
sions. Stress fibers are involved in cell motility and can be observed aligned in the
direction of movement in slowly migrating cells, but, surprisingly, are transversely
aligned in the fast-moving epidermal keratocyte. They are known to have impor-
tant physiological functions, such as participating in wound closure by fibroblasts
and maintaining blood vessel integrity by endothelial cells. Stress fibers can be
induced to form with the application of either biochemical or biomechanical
stimulation.
The capacity for stress fibers to exert contractile forces resulting in a pre-stress and
pre-strain can be demonstrated experimentally. Consider an experiment in which
cells are cultured on a flexible substrate that has been pre-stretched in one direc-
tion (Figure 10.26), and stress fibers that are oriented in the same direction. If one
then releases the substrate stretch by a small amount, such that the resulting con-
tractile strain in the stress fibers is less than its pre-strain, the fiber will retain a
302 CHAPTER 10: Adhesion, Migration, and Contraction

Figure 10.26 If cells are cultured on

an elastic membrane that is
allowed to uniaxially contract, the
stress fibers will exhibit buckling if
the contraction is large enough. The
length of the fiber under tension is
denoted as Rs, length of a straight, but
unloaded fiber is L0, and once it has
buckled, it has an end-to-end length of
Rb and a contour length of Lc. (From
Costa KD, Hucker WJ & Yin FC (2002).
Cell Motil. Cytoskeleton. 52.)

F
Rs

L0

L c(=L 0)

Rb

“straight” morphology. However, if the release strain is large such that it exceeds
the stress fiber pre-strain, the fiber will exhibit a “wiggly” or “buckled” morphol-
ogy. An alternative approach to demonstrate pre-stress in stress fibers is to sever a
stress fiber using a high-powered laser, or disrupt a focal adhesion anchoring a
stress fiber. In both cases, the free end of the stress fiber can be observed to con-
tract, again demonstrating that the stress fiber is under a pre-strain.

Stress fiber pre-strain can be measured from buckling behavior

The buckling of stress fibers can be used to estimate the level of pre-strain. Recall the
Nota Bene experiment in Figure 10.26, consider a stress fiber oriented in the same direction of
Contractile force in stress fibers.
substrate pre-stretch. This stress fiber, is assumed straight with an end-to-end length
The contractile force in pre-stressed Rs. We slowly contract the substrate, decreasing the pre-stress in the fiber.
stress fibers has been quantified Immediately before the stress fiber is found to take on a “buckled” morphology, we
using both traction force microscopy measure its end-to-end length and term this L0, the contour length of the stress fiber
and atomic force microscopy. in the absence of a pre-stress. As we contract the stress fiber even further, the stress
Tensions on the order of 4 nN have fiber assumes a “buckled” morphology with some end-to-end length Rb that is less
been reported.
than R0. Using these quantities, the pre-stretch in the fiber can be found as
Rs
λf = (10.41)
R0

and the pre-strain as 1 −  λf. It is noteworthy that λf can be calculated in an alterna-
tive manner that does not require the determination of Rs. Consider the case
where we know Rb and the stretch ratio of the substrate, λs. λs is related to Rs and
Rb as
Rs
λs = . (10.42)
Rb

Next we define the tortuosity as

Rb
T = , (10.43)
R0
 CONTRACTION 303

and then λf can equivalently be determined as

λf = T λs . (10.44)

Using this approach, the stress fiber pre-strain of cultured human endothelial cells
has been estimated to be as high as 25%, although it is highly heterogeneous. Nota Bene
Myosin variants. There is no single
Myosin cross-bridges generate sliding forces within myosin protein, but rather a huge
superfamily. There are over 40
actin bundles different genes and about 100
different protein products that are
In the previous section, we discussed various structures in muscle and non- grouped into 18 major variants
muscle cells that allow for the generation of contractile forces. In these cases, known as classes. This immense
although the structures were distinct in molecular makeup and organization, the diversity is generally the result of
same basic molecular mechanism was responsible for driving contraction: myo- various splice variants. Myosin II is
sin movement along actin. Actin and myosin together form a cellular force gen- the type found in muscle; the tails of
eration system that is critical to an amazingly wide variety of cells and tissues. this myosin allows it to form polymer
filaments. Other myosins remain as
Given that both muscle cells and non-muscle cells possess the capacity to gener- monomers and are known as
ate contractile forces using actin and myosin, the question arises, how does this unconventional myosins because
occur? Although it is well established that myosin functions as a molecular motor myosin was first identified in muscles.
in converting chemical energy into mechanical work by moving along actin, many Most of the variability is found in the
details have yet to be elucidated. However, the cross-bridge model in which collec- tail region, associated with binding to
tions of myosin molecules link adjacent actin polymers together provides a different cargos or assembling into
fibers, whereas the head regions tend
remarkably accurate model of actin–myosin interactions both within and outside
to be much more conserved. All
sarcomeres. myosins move toward the + or
In the cross-bridge model, myosin molecules go though several distinct stages or barbed end of the actin polymer,
except myosin VI, which moves
configurations (Figure 10.27). The process is cyclical in nature and at completion backward!
the myosin has returned to its initial configuration. In the first stage, myosin is
tightly bound to actin, and its ADP/ATP (adenosine diphosphate/adenosine
triphosphate) binding domain is not occupied. This is known as the rigor state,
because no sliding or motion can occur. To leave the rigor state, myosin must bind
a molecule of ATP. When this occurs, myosin releases from actin, and ATP is

Cross-bridge model Figure 10.27 The myosin

configurations of the cross-bridge
+ − model. In this case the actin molecule
is sliding from right to left.

ATP
ATP
rigor state

ADP•Pi

power stroke ADP•Pi

ADP
ADP Pi
304 CHAPTER 10: Adhesion, Migration, and Contraction

hydrolyzed into ADP. This releases chemical energy, allowing myosin to change
conformation into what is known as a cocked configuration. The hallmark of this
configuration is that the myosin head has moved toward the (−) or barbed end of
the actin polymer. In the third stage, myosin rebinds to the actin, but because it is
in the cocked configuration, it rebinds at a site more toward the (−) end. In the
fourth stage, rebinding of actin induces the myosin to release its phosphate. In
this process, it also undergoes the power stroke. It returns to the uncocked con-
figuration, and in the process drags the actin molecule toward the center of the
myosin cross-bridge. The power stroke is where mechanical work is produced. In
the fifth and final stage, the spent molecule of ADP is released. This returns the
myosin to the rigor state.

Myosin molecules work together to produce sliding

Nota Bene An interesting aspect of actin–myosin contraction is that the myosin heads are not
able to bind just anywhere along the polymer. In fact, each actin subunit has only
Rigor mortis occurs with lack of one binding site for the myosin heads. As we learned in Section 7.1, actin subunits
ATP. A familiar phenomenon occurs
polymerize in a helical structure. The pitch of this helix has been measured to be
when ATP generation is suddenly
halted in the body. Actin–myosin Δ = 36 nm; in other words, the polymer completes one full turn of the helix for every
interactions will be halted in one of 36 nm along its length. Interestingly, the distance that the myosin molecule stretches
two steps: just before the power when it is cocked in preparation for a power stroke, δ, is only around 5 nm. Because
stroke, or just after the power stroke, binding sites can only occur once in each full twist, it would be impossible for a
before the myosin head can myosin molecule to reach far enough forward to grab another binding site each
disengage. In either case, there may time through the cross-bridge cycle. A simple explanation for this apparent paradox
be some residual contraction,
depleting the available ATP stock,
is that many myosin molecules must work together to produce sliding. At any given
followed by a failure of relaxation point in time, most are not engaged with actin at all, whereas a few are bound and
because the actin–myosin tugging on the actin molecule, much like the legs of a millipede.
interactions are locked in place. An
example of this occurring is during The implication of this cooperative behavior is that any given myosin molecule,
rigor mortis after death. The body spends most of its time bound to ATP and, as a result, unbound from actin. For
may feel stiff because of the example, let the average time per cycle that the myosin is not bound to actin be
contraction of muscles that are toff, and the time that myosin is bound to actin be ton. We might estimate the frac-
subsequently locked in place owing tion of time myosin is engaged (ton/ttotal), a quantity we will call the duty ratio,
to exhaustion of the available ATP
supply.
rduty, by assuming that is it proportional to the ratio of δ and Δ,

δ 5 nm
rduty = = = 0.14. (10.45)
∆ 36 nm

However, even this seems to require much more frequent binding to actin than is
actually observed. In fact, myosin seems to be quite a “lazy” molecule. The ration-
ale for this statement comes from considering the force generation capacity of
entire actin–myosin bundles. The total tension should be approximately the num-
ber of cross-bridges multiplied by the force generated by each cross-bridge. The
number of cross-bridges can be estimated, and the time-average force per cross-
bridge is a function of ton and toff,

t on Fon + t off Foff

F = . (10.46)
t on + t off

However, the force generated when myosin is not engaged to actin must be zero
(<Foff > = 0). Therefore,

t on
F = Fon = rduty Fon (10.47)
t on + t off
 CONTRACTION 305

or

δ t on F
rduty = = = . (10.48)
∆ t on + t off Fon

The power-stroke model is a mechanical model of

actomyosin interactions
Nota Bene
Given what we know about the mechanisms of actin–myosin contraction, we now The famous Huxleys. Sir Andrew
analyze the mechanical aspects of this process. To do this, we use the power-stroke Huxley is credited with developing
model, which quantitatively describes the molecular mechanics. This model the power-stroke model, the
treats the myosin heads as simple linear springs (Figure 10.28) with an effective paradigm that established the
spring constant of km. The model considers three possible positions of the myosin equations upon which our
spring. The relaxed position is the zero-force position. The spring moves forward quantitative understanding of
actin–myosin force generation is
by a distance δ+ when it is cocked and, consequently, where it binds to actin. based. However, this contribution is
Finally, the model does not assume that the actin is released at the zero position. overshadowed by his work on the
Rather, it assumes that the spring may be compressed by some amount δ- past the Hodgkin–Huxley model of nerve
zero position before it is released. This phenomenon is sometimes referred to as conduction, for which he was
drag (Figure 10.29). awarded the Nobel Prize in
Physiology or Medicine in 1963. This
We now compute the average force being generated by the spring. Recall that established the action potential, and
〈Fon〉 is the average force over one cross-bridge cycle when myosin is bound to suggested the existence of ion
actin. This force occurs as the spring travels from the δ+ position, when it first channels. Remarkably, his half-brother
Aldous Huxley is also very well
binds, through the zero position, to the δ- position, where it unbinds. The aver- known. He wrote about the
age force is dehumanizing aspects of scientific
progress and was the author of Brave
New World.
km (δ + − δ − ) (10.49)
Fon = .
2

relaxed binding sites Figure 10.28 Schematic diagram of

the three power-stroke model
configurations. In the relaxed position
there is no force and the myosin is not
engaged with actin. Myosin is modeled
myosin
as a simple linear spring. The spring
moves forward by a distance δ+ when it
is cocked and imparts a force of kmδ+ on
cocked the actin filament. In the “compressed”
position, the spring compresses by
force δ− past the zero position before
releasing actin.

δ+

compressed

δ–
306 CHAPTER 10: Adhesion, Migration, and Contraction

[F] Combining this with our expression for duty ratio and making the observation
that the total stroke distance δ =  δ+ +  δ-, we obtain
Ft
F = rduty Fon

δ km
approximately = (δ + − δ − )
hyperbolic ∆ 2
km (10.50)
= (δ + + δ − )(δ + − δ − )
2∆
vmax v km 2 k m 2
= δ+ − δ− ,
Figure 10.29 The force–velocity 2
 ∆ 2∆

F+ F−
relationship predicted by the
power-stroke model. Although not
strictly hyperbolic, for physiological where the first term is the positive elastic force, F+, and the second term is the
values of the physical parameters, the negative drag force, F−.
power-stroke model is similar to the
prediction of Hill’s equation. Let us consider the form of Equation 10.50 in a bit more detail and see if we can
relate it to sliding velocity. The first term depends on the distance that the myosin
heads move forward when they are cocked. As we described, most of the myosin
head groups are already bound with ATP and therefore cocked at any point in
time. Therefore, there is no reason to expect that this would depend on velocity.
On the other hand, the second term depends on the distance the heads over-
shoot the zero force position before releasing, in other words the distance the
myosin springs get compressed during each cycle. It would be reasonable for this
quantity to depend on sliding velocity. In fact, this compression distance should
increase linearly with velocity. If we introduce the release time as a new param-
eter tr, we obtain,

km
F = F+ − (vt r )2 . (10.51)
2∆

Notice that the first term is the maximum positive force the myosin is capable of
generating. It is a constant, depending only on structural parameters. Also, note
that the maximum force occurs when the velocity is zero, consistent with the
concept of a stall force, at which point the second term (the drag force) is also
zero. The power-stroke model is consistent with some important features of the
force–velocity behavior we have observed, a maximum force at zero velocity and
a maximum contraction velocity that occurs when the force is zero. However, the
drag predicted with this model increases with the square of the sliding velocity.
Upon inspection of Equation 10.51, we can see that this prediction is not consist-
ent with the predictions of the Hill model, which does not have a square depend-
ence on velocity. To resolve this issue, let us again consider the parameter Δ, the
distance separating the myosin-binding sites on actin. As we alluded to previ-
ously, the minimum possible value of Δ is the physical spacing of the sites, which
we will now term ds. From the duty ratio argument, we expect Δ to exceed ds by
some amount, but the myosin heads and the actin-binding sites must align at
some point. Therefore,

∆ = nds , (10.52)

where n is an integer. To estimate what n might be, consider that each binding site
has an effective “sweet spot” of width wbs. If, in order to bind, the myosin head
needs to be within this width, the time available for binding is inversely related to
velocity such that the time available to form a bond with the binding site is

wbs
t bs = . (10.53)
v
 CONTRACTION 307

Further, we assume that while the myosin is in the sweet spot, it will bind actin at
a constant rate of k+ such that

d[M ]
= − k + [ M ]. (10.54)
dt
Note that Equation 10.54 is entirely analogous to Equation 10.31, which described
the time evolution of adhesive bonds in the Bell model. As before, the solution to
Equation 10.54 is an exponential of the form

[ M ](t ) = [ M ](0)e− k t . +
(10.55)

From the perspective of the individual myosin molecule, the probability of transi-
tioning from an unbound to a bound state in the time available for binding is

p(t bs ) = 1 − e − k+ t bs
w
(10.56)
− k+ bs
=1−e v .

The probability that a myosin head will bind in the time available is an exponen-
tial. To find the total probability over the period, we must integrate from 0 to tbs.
t bs

∫k e ( −e ) = 1 − e
− k+ t t bs − k+ t − k+ t bs
p(t bs ) = + dt = . (10.57)
0
0
You can see from this expression that if the velocity is very high, this probability
gets quite small, and the myosin heads would pass by many potential binding
spots without binding. In this limiting case Δ ⪢ d and n is large. In other words,
relatively few of the binding sites are occupied. Because each binding site is sepa-
rated from the next by a distance δ, and the distance between an occupied site is
Δ, then the probability of a given head being bound is also given by

1 ds
p(t bs ) = = . (10.58)
n ∆

Setting Equation 10.56 equal to Equation 10.58 we obtain

ds
∆= w (10.59)
− k+ bs
1−e v

and

F =
km 2
2∆
( 2
δ+ − δ − )
kmδ +2  δ2
=  1 − −2  (10.60)
2∆  δ+ 

kmδ +2   v 2t −2 
wbs
− kon
=  1−e v
  1 − .
2ds   δ +2 

Equation 10.59 gives behavior that resembles the hyperbolic relationship observed
experimentally by Hill. To see this, consider some realistic values for the parame-
ters. Molecular measurements of myosin predict a value of km of roughly 5 pN/nm
and t− = 0.6 ms. As discussed, molecular structures suggest a binding-site spacing
of ds = 36 nm and myosin stroke distance of δ+ = 5 nm. The term k+ would change
308 CHAPTER 10: Adhesion, Migration, and Contraction

depending on ATP concentration, but from statistical mechanics a theoretical

maximum limit is the rate at which ATP will encounter its myosin-binding site
(~21 s−1) if there is an abundance of ATP. wbs is also difficult to measure, but again
a theoretical maximum would be the binding site spacing, ds. As can be seen in
Figure 10.29, for these parameters, the force velocity curve is not precisely hyper-
bolic, but within a physiological range it is quite close.

Key Concepts

• Cell adhesion involves multiple proteins interacting to • Cell migration involves a series of coordinated
form structures such as focal adhesions in cell–matrix processes, including protrusion, adhesion,
adhesion and similar adhesive plaques with cell–cell displacement/contraction, and release/retraction.
adhesion. Brownian ratchet processes occur at the interface of
• Adhesiveness is difficult to model precisely without polymerizing actin and membrane, and depend on
knowing the number and energy of receptors. thermal fluctuations forming a gap for the insertion
However, lumped parameter models based on energy of a new monomer.
of adhesion and surface tension can be quite • Actin–myosin interactions generate cellular force in
informative. an ATP-dependent fashion. They are responsible for
• During inflamation, neutrophils attach to vessel walls force generation in smooth and striated muscles as
in stages characterized by endothelial cell activation, well as non-muscle cells in stress fibers and during
selectin-mediated rolling, neutrophil activation, and cellular motility.
integrin-mediated firm adhesion. • The Hill equation describes the inverse hyperbolic
• The Bell model describes the effect of force on bond relationship between force and velocity.
kinetics with an exponential increase in off-rate with • Cross-bridge models describe the cyclic series of
force. states that myosin goes through in converting ATP to
• Catch bonds exhibit increased adhesion with load. force. The duty ratio of myosin is the fraction of time
Catch-slip bonds exhibit decreased adhesion once a that it is engaged with actin, which is surprisingly low.
certain threshold force is reached. • The power-stroke model describes the actin–myosin
• Cell migration can be observed in vivo but is more interaction mechanics at a molecular level by treating
easily quantified in vitro. The net migration of a cell myosin as a simple spring that is shortened during
is the result of the speed and persistence. They cocking. It predicts force–velocity relationships in
can be quantified individually from cellular terms of cocking distance, drag, and the size and
trajectories. frequency of actin-binding sites.

Problems

1. During cell migration, a cell does work that is typically 2. Consider a cell attached to a surface, as shown below.
dissipated into viscous losses within and outside the Other than the flattened contact region, the cell is
cell. In vitro, the outside of the cell is typically much less spherical with radius R. You have deployed a shear
viscous than the cytoplasm, so the losses outside the stress device to determine the adhesion strength of the
cell can be neglected. Treat the cell as a thin pancake cell to the surface, by gradually increasing the shear
of height h and radius R, and treat the cytoplasm as force applied to the cell. The adhesion force per unit
an incompressible Newtonian fluid of viscosity μ and area is denoted Fa, exerting its force roughly uniformly
density ρ. The cell attaches to the substrate with an over the adhesion area, which is approximately Ra in
adhesion force per unit area of Fa (you may assume this diameter. The shear stress device consists of a chamber
is uniform across the cell’s basal surface). If the velocity with an upper platform moving with velocity v, which
of the cell v is constant, derive an expression for v in is increased very slowly so that the flow pattern can be
terms of the other parameters and determine whether considered fully developed Couette flow at any time.
the cell would move faster or slower if it were to thin by The chamber height is h. At some critical velocity vc the
a small amount (h decreases). cell detaches.
 ProbLems 309

persistence of the cell? What if you sample every

tenth of a second (0.1 s)?
6. Consider a spherical cell. If you split the sphere with a
flat plane, you get two components as shown below
(the component that is labeled B has been inverted for
clarity).

(A)

(B)

Show that the ratio of adhesion energies for

The fluid has viscosity μ and density ρ, and the flow can
cell component A and cell component B can be
be considered laminar. Use a scaling relationship based θ
on (a) dimensional analysis and (b) a (scaling) moment expressed as tan2 , where θ is the contact angle
2
balance to derive an expression for vc based on the of cell component A to the surface. You may treat
provided parameters. (c) Show that the expression the cell components as droplets without receptor
obtained from (b) can be derived using the functional interactions.
form from (a).
7. From geometric arguments and the form of Equation
3. In discussing actin–myosin contraction, we briefly 6.5, derive Equation 10.19.
described what occurs during rigor mortis. Examine this
process more closely by considering the following: 8. Consider the cell migration process, and focus on the
(a) At which stage of the contraction cycle does rigor Brownian ratchet model as applied to the membrane
mortis occur? or the actin filaments (or both). What would you expect
(b) Why does this make the body stiff? If you just to happen to cell migration as the temperature is
lie there and do not move around, are you increased or decreased? Why is a fever advantageous
considerably less stiff than a dead body? But are for neutrophils?
your muscles not in the same position?
9. Derive the probability of a single myosin head
(c) After a brief time, rigor mortis dissipates. This
binding (Equation 10.55) from the sweet spot
occurs fast enough that experts can sometimes tell
argument.
when death occurred based on the body stiffness.
Why does rigor mortis dissipate? 10. In terms of the Hill equation parameters, a, b, and k,
 d y 4
 d y 
2 what is the maximum shortening velocity?
4. (a) From Equation 10.14,  EI 4 + Fa = T  2   ,
 dx  dx  
11. Typical values for the Hill parameters for frog
determine a critical peel length (along the muscle are a = 37.49 mN/mm2, b = 0.317 mm/s,
membrane) at which the bending component of and k = 47.14 mN/mm2 s. Plot the predicted
the membrane roughly balances the stretching force–velocity curve and determine the isometric
component of the membrane. contraction force and maximum contraction
(b) Assume that stretching is much more dominant velocity.
than bending. Derive an expression for the
membrane tension as the cell is being peeled. 12. For the power-stroke model (Equation 10.59) prepare
a force–velocity plot for typical values. What is the
5. (a) Sketch out sample trajectories for a cell with the
maximum velocity? Calculate work 〈F〉Δ, working
four combinations of speed and persistence (high
distance δ, and duty ratio rduty = (δ/Δ) as a function of
and low of each). Alternatively, you may use Figure
sliding speed, v.
10.20 as a reference. Discuss the differences
among the different combinations. 13. Consider stress fiber actin-myosin activity. What
(b) Assume a cell travels along a sine wave would happen if the fiber was cut with a laser?
(y = sin(2πt)) at constant velocity. If you sample How would δ+ and δ− change. What would happen
every half second (0.5 s), what is the speed and to drag?
310 CHAPTER 10: Adhesion, Migration, and Contraction

Annotated References

Ananthakrishnan R & Ehrlicher A (2007) Forces behind cell move- Huxley AF & Niedergerke R (1954) Structural changes in muscle
ment. Int. J. Biol. Sci. 3(5), 303–317. This journal review article pre- during contraction: interference microscopy of living muscle fibres.
sents a nice overview of cell adhesion and migration as they work Nature 173, 971–973. Early description of muscle striations and how
together to move cells. they change with contraction.
Bell GI (1978) Models for the specific adhesion of cells to cells. Johnson KL (1985) Contact Mechanics. Cambridge University Press.
Science 200, 618–627. The original presentation of Bell’s model of A good reference for Hertz contact and other contact problems in
the effect of force on adhesion bonds. mechanics.
Boal D (2002) Mechanics of the Cell. Cambridge University Press. Kamm RD and Lang M. Molecular, Cellular, and Tissue Biomechan-
Boal’s text covers many aspects of cell mechanics mathematically, ics (course material from course number 20.310, Massachusetts
including some of the adhesion derivations presented in this chapter. Institute of Technology, Cambridge, MA available online at http://
Costa KD, Hucker WJ & Yin FC (2002) Buckling of actin stress fibers: ocw.mit.edu). This course covered basic principles of biomechan-
a new wrinkle in the cytoskeletal tapestry. Cell Motil. Cytoskel. 52, ics, including scaling analysis and engineering analysis, as applied
266–274. This article is the source for the method of determining to biological molecules, cells and tissues. Much of the class material
stress fiber pre-stress by culturing on a flexible substrate. served as inspiration for many topics throughout this chapter. Some
of the lecture materials can be found online in OpenCourseWare.
Deguchi S, Ohashi T & Sato M (2005) Evaluation of tension in actin
bundle of endothelial cells based on preexisting strain and tensile Kendall K (2001) Molecular Adhesion and Its Applications. Kluwer
properties measurements. Mol. Cell. Biomech. 2, 125–33. Direct Academic/Plenum Publishers. This textbook presents many funda-
measurement of stress fiber pre-stress using cantilevers. mentals of adhesion with a section on cell adhesion. Some of the
basic laws of adhesion are presented in this book.
Dembo M, Torney DC, Saxman K & Hammer D (1988) The reaction-
limited kinetics of membrane-to-surface adhesion and detachment. Lauffenburger D & Linderman JJ (1993) Receptors: Models for Bind-
Proc. R. Soc. Lond. B. 234, 55–83. Some of the first clear evidence ing, Trafficking and Signaling. Oxford University Press. This text cov-
suggesting that catch bonds might exist. ers both biology and mathematics of cell interactions, migration,
adhesion and signaling. The speed–persistence relationship and
Dillard DA & Pocius AV (2002) The Mechanics of Adhesion. Elsevier membrane peeling presented in this chapter is based, in part, on
Science. Dillard’s text provides some mathematical background the material in this book.
in adhesion. It complements but also overlaps some of Kendall’s
material. Dillard’s material includes the discussion of the JKR Maheshwari G & Lauffenburger DA (1998) Deconstructing (and
described in Section 10.1. reconstructing) cell migration. Micro. Res. Tech. 43, 358–368 This
journal article reviews many aspects of cell migration, including
Evans EA (1985) Detailed mechanics of membrane-membrane chemical gradients, path tracing, and the various steps involved in
adhesion and separation I&II. Biophys. J. 48, 175–192. An early cell motion.
report of micropipette cell–cell adhesion experiments and peeling
analysis. Marshall BT, Long M, Piper JW et al. (2003) Direct observation
of  catch bonds involving cell-adhesion molecules. Nature 423,
Evans EA & Calderwood DA (2007) Forces and bond dynamics in cell 190–193. More recent direct observations of the existence of catch
adhesion. Science 316, 1148–1153. A review of the role of forces in bonds.
cellular adhesion.
Pollard TD, Blanchoin L & Dyche Mullins R (2000) Molecular mecha-
Finger EB, Puri KD, Alon R et al. (1996) Adhesion through L-selectin
nisms controlling actin filament dynamics in nonmuscle cells. Annu.
requires a threshold hydrodynamic shear. Nature 379, 266–268. A
Rev. Biophys. Biomol. Struct. 29, 545–576 Pollard’s review article
quantitative characterization of cell rolling via selectin binding.
discusses various aspects of actin remodeling and dynamics in the
Hill AV (1938) The heat of shortening and dynamics constants of context of the leading edge of a cell undergoing migration.
muscles. Proc. R. Soc. Lond. B 126, 136–195. This elegant paper
Pollard TD (2002) The cytoskeleton, cellular motility and the reduc-
describes the original Hill experiments describing measurements
tionist agenda. Nature 422, 741–745. This insight article (similar to
of changes in temperature of 1 × 10−3 °C. It is the seminal presen-
a focused review) discusses what’s known about the role of the
tation of the Hill model.
cytoskeleton in describing cell mobility.
Holmes JW (2006) Teaching from classic papers: Hill’s model of
muscle contraction. Adv. Physiol. Edu. 10, 67–72. A description of Ramachandran V, Williams M, Yago T, et al. (2004) Dynamic altera-
a simulation-based teaching module that takes students through tions of membrane tethers stabilize leukocyte rolling on P-selectin.
Hill’s original reasoning process, eventually allowing them for for- Proc. Natl. Acad. Sci. USA 101, 13519–13524. This research article
mulate his model. describes the behavior of leukocytes as they contact a surface
coated with P-selectin. The cell velocities and tether formations are
Howard J (2001) Mechanics of Motor Proteins and the Cytoskeleton. discussed.
Sinauer Associates. A great introduction to cytoskeletal polymeriza-
tion and mechanics. This book provides a solid foundation of the Tanner K, Boudreau A, Bissell MJ & Kumar S (2010) Dissecting
mechanics of the cytoskeleton and motor protein force generation. regional variations in stress fiber mechanics in living cells with laser
Much of the development for the cross-bridge model, as well as nanosurgery. Biophys. J. 99, 2775–2783. This article describes the
some problems, are adapted from this text. use of high-power lasers to disrupt stress fibers.

Huxley AF (1957) Muscle structure and theories of contration. Tees DF & Goetz DJ (2003) Leukocyte adhesion: an exquisite bal-
Prog. Biophys. Biophys. Chem. 7, 255–318. Original presentation of ance of hydrodynamic and molecular forces. News Physiol. Sci. 18,
the power-stroke model of force generation in terms of molecular 186–190. Review of balance and interplay between the forces of
parameters. flow and adhesion in leukocytes.
CHAPTER 11
Cellular
Mechanotransduction

I n this chapter, we focus on the process of mechanotransduction, in other words

the process by which a cell transduces a mechanical stimulus into a specific cel-
lular response. In biology, cellular signal transduction is typically defined as the
process by which a cell senses and responds to a chemical stimulus. For example,
binding of a ligand to a cell receptor may lead to a conformational change in that
receptor, triggering an intracellular signaling cascade that ultimately alters cellu-
lar function. Mechanotransduction can be viewed as a completely analogous pro-
cess to chemical signal transduction in that application of mechanical forces may
induce a conformational change in mechanosensing molecules that activate an
intracellular signaling pathway, which leads to altered function. As discussed in
Section 1.1, an incredibly diverse range of tissues and organs depend on the cel-
lular sensation of extrinsic mechanical signals for proper development or func-
tion. Disruptions in this process have been implicated in contributing to the onset
or progression of several critically important diseases.
We will discuss cellular mechanotransduction mechanisms in a variety of cell Nota Bene
types. This chapter is divided into four parts, organized to reflect the sequentially
occurring stages of cellular mechanotransduction. First, the cell must be exposed Mechanotransduction can occur
to mechanical loading. In other words, force must be transmitted from the envi- without cells. Mechanotransduction
at the tissue level does not
ronment to the level of the cell. Second, the force must be detected by mechano- necessarily require cellular
sensitive molecules. That is, these loads must be transmitted to a molecule or a mechanotransduction. Bone
complex of molecules that can activate (that is, undergo conformational change). possesses piezoelectric properties,
Third, a conformational change in these molecules must be sufficient to activate leading some to hypothesize that
intracellular signaling pathways (Figure 11.1). Finally, a cascade of intracellular skeletal adaptation to mechanical
signaling events occur that ultimately result in altered cellular function. In the first loading may be mediated by
sensation of electric fields by bone
section, we provide a survey of various mechanical stimuli that regulate cell func-
cells. Skeletal mechanotransduction
tion. Next, we discuss various structures that have been proposed to transmit the may not require the sensation of
loads arising from these stimuli to putative mechanically sensitive molecules. In mechanical loads by the cells
the third section, we discuss ways in which force-induced conformational changes themselves. Another cell-free
in proteins may allow the conversion of mechanical forces into biochemical mechanism involves loading of
events. Finally, we conclude the chapter by discussing mechanisms by which extracellular matrix proteins like
intracellular signaling events ultimately alter cellular function. As we will see, this fibronectin causing unfolding and
uncovering of cryptic binding sites.
is an active area of investigation, with many molecular aspects of the process not
fully defined.

11.1 MECHANICAL SIGNALS

We begin our discussion of cellular mechanotransduction with a survey of
mechanical signals that have been demonstrated to alter cellular function. We
first introduced the notion of fluid flow as a regulatory signal in cells in Section
6.3. The body is filled with fluid: air, blood, cerebrospinal fluid, etc. When fluid-
filled tissues and organs are subjected to mechanical loads that generate pres-
sure gradients, fluid flow results, subjecting the cells within to several physical
signals. These signals include altered transport of nutrients and/or signaling
312 CHAPTER 11: Cellular Mechanotransduction

Figure 11.1 Schematic of a bone (A)

cell subjected to fluid flow,
demonstrating that
flow
mechanotransduction occurs in
sensor
four fundamental steps. (A) The first
step involves transfer of load from the
environment to the cell. (B) Second, the
load is transmitted to a “flow-sensing”
molecule, resulting in a conformational
change. (C) Third, an intracellular
signaling cascade is initiated, by altered
levels of a second messenger molecule. (B) fluid flow
(D) Finally, the procession of the
intracellular signaling events leads to
altered cellular function, specifically new flow
bone formation. sensor

(C) fluid flow

flow
sensor

second messenger

(D) fluid flow

flow protein further

sensor kinase propagation
second messenger of signal new
bone

molecules, streaming potentials, and fluid shear stress. As we will see, cells in a
variety of tissues sense and respond to fluid flow, even in tissues in which trans-
port of fluid is not a primary function.

Nota Bene
Vascular endothelium experiences blood-flow-mediated
Lining cells respond and adapt to shear stress
enhance their function. The
capacity for lumen-lining endothelial The vascular system is one of the earliest systems in which fluid flow was impli-
and epithelial cells to respond to flow cated as an important signal for regulating cell behavior. This is due in large part
indicates that these vessels are not
merely a collection of pipes and
to the fact that endothelial cells (cells lining the vessels) exhibit dramatic,
tubes meant for transport. They are observable changes in morphology in response to fluid shear stress. In particu-
dynamic, responsive systems of lar, when exposed to steady flow, they typically adopt a morphology that is
different cell types that respond to aligned in the direction of flow (Figure 11.2), although this can be dependent
external stimuli. This adds to the on the location of the cells within the vasculature. Endothelial cells isolated
growing body of evidence that the from heart valves tend to realign perpendicular to the direction of shear stress
body has few (or perhaps no) passive
(Figure 11.3). In vitro studies investigating the responses of endothelial cells to
components.
flow have revealed the capacity for fluid shear to not only regulate all morphology,
 MECHANICAL SIGNALS 313

but also to regulate the levels of a variety of other regulatory molecules such as
vasodilators/vasoconstrictors and growth factors. Recirculant, or “disturbed,”
flows have been found to result in monolayer disruptions and alterations in cell
division that have been linked to sites of atherosclerotic plaque formation in
humans (Figure 11.4). This suggests that the onset of atherosclerosis may be
linked to flows that are unsteady. This hypothesis has ample support from in
vitro and in vivo experiments,. Disturbed flows tend to occur near vascular
bifurcations, and these bifurcations are where plaques tend to form. Fluctuating,
fluid shear stress may contribute to the development and progression of
atherosclerosis.

Lumen-lining epithelial cells are subjected to fluid flow

Similar to endothelial cells lining blood vessels, lumen-lining epithelial cells such
as those lining kidney tubules are subjected to fluid shear stress. In kidneys, the
ability of epithelial cells to sense urine flow is fundamental to proper kidney func-
tion. A link between renal flow sensing and polycystic kidney disease (PKD), Static
which is characterized by the formation of renal cysts, has been established and
will be discussed in Section 11.2. Flow sensing has similarly been implicated in
maintaining liver health. Cholangiocytes, epithelial cells that line the lumen of the
liver bile duct, are exposed to passive movement of bile. These cells have been
found to sense and respond to fluid flow by a mechanism that appears to be simi-
lar to that of the kidney.

Fluid flow occurs in musculoskeletal tissues

Fluid flow also regulates cells in tissues that do not transport fluid as part of Flow 15 min
their primary function. Bone tissue incorporates a fluid-filled network of voids
(lacunae) and channels (canaliculi). Within the lacunae reside bone cells
called osteocytes that extend long, slender processes through the canaliculi to
neighboring osteocytes. Mechanical loads associated with habitual loading
(ambulation—crawling, walking or running) generate pressure gradients

Flow 24 h

Figure 11.2 ​Pulmonary endothelial

cells adapt their actin cytoskeleton
to flow. The cells in these figures have
been stained with fluorescently
flow conjugated phalloidin, which binds to
actin. Under flow, the cells first quickly
Figure 11.3 Endothelial cells exposed to flow. These cells are cultured from porcine reinforce their actin cytoskeleton, as
aortas (left) and aortic valves (right) exposed to 20 dynes/cm2 of shear flow for 48 hours. seen in the middle panel. More gradually,
They exhibit vastly different realignment properties demonstrating that cell source is a they reorganize their cytoskeleton and
critical factor in the mechanobiologic response. See www.garlandscience.com for a overall morphology in the direction of
color version of this figure. (from Butcher JT et al. (2004) Arterioscler. Thromb. Vasc. flow. (From, Birokov KG et al. (2002) Am.
Biol. 24.) J. Respir. Cell Mol. Biol. 464.)
314 CHAPTER 11: Cellular Mechanotransduction

Figure 11.4 Regions near

bifurcations are more prone to
atherosclerotic formation (shaded proximal
black). Several groups hypothesized
that because of recirculant or other
disturbed flow patterns near these
regions, the endothelial cells alter their mid-proximal
behavior, leading to atherosclerotic
formations. (From, Thubrikar MJ et al.
(1995) Ann. Thorac. Surg. With
distal
permission from Elsevier)

Figure 11.5 Mechanical loading of

bone induces interstitial fluid flow
within the lacunar–canalicular
system in which osteocytes reside.
When bone is loaded fluid flows from
regions of compression to regions of
tension.

osteocyte canaliculus lacuna

force

w
fluid flo
compression

tension

force
 MECHANICAL SIGNALS 315

bleached cell Figure 11.6 Fluorescence recovery

after photobleaching in a single
lacuna within a mouse femur. By
monitoring increases in the recovery
rate in the presence of mechanical
loading, one may determine whether
enhanced transport of unbleached dye
prebleach postbleach t = 32 s t = 64 s t = 96 s t = 128 s due to convective flow is occurring.

within the lacunar–canalicular system that drives flow from areas of compres-
sion to areas of tension (Figure 11.5), exposing osteocytes to fluid shear stress.
An important distinction of osteocyte flow sensing compared with luminal flow
sensing is that although luminal flow (in blood vessels or within kidney tubules)
can be readily observed, experimental measurement of bone flow is very diffi-
cult. This is due to the technical challenges associated with the ­microscopic
fluid spaces of the lacunar–canalicular system (~0.1–1 μm). However, the
recent use of fluorescence recovery after photobleaching (FRAP) has provided
some insight (Figure 11.6). In this approach, a fluorescent tracer is injected
into the animal and allowed to equilibrate within the lacunar–canalicular sys-
tem. A bone surface is exposed, and a laser scanning confocal microscope is
used to photobleach a single lacuna. Infilling of unbleached tracer molecules
from surrounding lacunae and canaliculi results in gradual recovery of fluo-
rescence in the photobleached lacuna. In the presence of flow, infilling is
enhanced by convective transport (the motion of tracer carried by the fluid).
We can determine whether flow is occurring by comparing transport rates in
the presence or absence of loading. Further, the convective velocity can be
estimated by mathematical modeling. Several experiments in cultured bone
cells have demonstrated that they sense and respond to levels of fluid flow
predicted to occur in habitual activities of daily living. Fluid flow within the
musculoskeletal system is not limited to bone. Like bone, articular cartilage is
a mechanosensitive tissue that is well known to adapt its properties to the
loading it experiences. Within cartilage, chondrocytes are embedded in an
extracellular matrix consisting primarily of proteoglycans, collagens, and
water. With mechanical loading, a variety of physical signals are generated
including hydrostatic pressure, matrix strain, and fluid flow. Several experi-
ments in cultured chondrocytes as well as in perfused cartilage and tissue-
engineered constructs indicate that fluid flow is a potent regulator of cartilage
metabolism.

Fluid flow during embryonic development regulates the

establishment of left–right asymmetry
We have so far discussed the role of fluid flow in regulating tissues and organs in
the adult. However, the regulatory role of fluid flow is also important during
development. In the mammalian embryo, the generation and subsequent sens-
ing of flow is critical for the establishment of left–right asymmetry. During gas-
trulation, a triangular-shaped indentation called a node appears on the surface
of the embryo. The cells lining the surface of the node are ciliated, and some of
these cilia are motile (Figure 11.7). They move in a vortical pattern, generating Figure 11.7 Electron micrograph of
a leftward flow of extra-embryonic fluid (termed nodal flow). As will be dis- primary cilia (highlighted by white
cussed in more depth later in this chapter, this nodal flow is sensed by cells on arrows) within the embryonic node.
the left margin of the node, resulting in an asymmetric intracellular Ca2+ signal. These cells sense flow that is critical to
Superimposing a rightward flow (by loading the embryos in a flow chamber) establishing the right/left directions in
the embryo (From, Nonaka S et al. (1998)
can reverse the left–right asymmetry such that the left and right sides of the
Cell. 95.)
body are exchanged.
316 CHAPTER 11: Cellular Mechanotransduction

Advanced Material: Fluid shear stress or chemotransport?

Fluid flow exposes cells to several different physical sig- where Q is the flow rate, μ is the fluid viscosity, b is the
nals, and it is often desirable to distinguish which of chamber width, and h is the chamber height. Equation
these signals are driving flow-induced cellular responses. 11.1 implies that one can subject cells to the same shear
Flow exposes cells to convective chemotransport, in stress but different flow rates by adjusting the viscosity
other words the carriage of nutrients and/or signaling of the flow media (this can be achieved by adding neu-
molecules by the fluid motion. It also exposes cells to tral dextran to the media). The same shear stress would
fluid shear stress. Because these signals are coupled, one result from a doubling of the fluid viscosity and halving
question is how do we determine whether cells are being of the flow rate, or a halving of the viscosity and a dou-
stimulated by chemotransport or fluid shear? bling of the flow rate. If cell responsiveness increases
with flow rate but not shear stress the cells are respond-
One way to isolate the effects chemotransport in parallel
ing to chemotransport. Alternatively, if the responsive-
plate flow chambers is to alter shear stress parametri-
ness of cells increases with shear stress but not flow
cally while keeping flow rate constant, and vice versa.
rate, it suggests that the cells are responding to fluid
Recall from Equation 10.2 that, in these chambers, shear
shear.
stress is related to flow rate as
6 µQ
τ = (11.1)
bh 2

Strain and matrix deformation function as regulatory signals

We now turn our attention from fluid flow as a regulatory signal to substrate or
matrix strain. As we learned in Section 3.2, whenever a solid material is loaded (or
exposed to stress), it deforms and undergoes strain. As the matrix deforms, the
cells embedded in these tissues experience loads at the sites of adhesion to the
matrix; as a result, they are subjected to strain (Figure 11.8). We present some
examples in which substrate or matrix strain function as critical mediators of bio-
logical processes.

Smooth muscle cells and cardiac myocytes are subjected to

strain in the cardiovascular system
We previously learned that, within blood vessels, endothelial cells sense and
respond to hydrodynamic forces arising from blood flow. Strain is another regula-
tory signal within the cardiovascular system. Pulsatile changes in blood pressure
induce oscillatory stretch within the walls of blood vessels, whose major cellular
component is vascular smooth muscle cells. Under normal circumstances, these
cells do not come into contact with blood directly. However, they are sensitive to
stretch. Mechanical stretch alters several functions in vascular smooth muscle
cells, such as cell alignment, migration, proliferation, and apoptosis. Stretch also
regulates secretion and production of several paracrine and endocrine factors.

Figure 11.8 Mechanical coupling of

cell to matrix. Because the cell is well
connected within the extracellular
matrix, deformation of a matrix will
transmit loads to the cells residing
within. This will result in cellular strain.
 MECHANICAL SIGNALS 317

Cardiac myocytes are also subjected to stretch-induced stimulation. Under hyper-

tension, this stimulation is increased, and this overstimulation is believed to lead
to cardiac hypertrophy. Elucidation of relevant regulatory pathways could have
therapeutic implications. We know that pressure overload leads to altered
mechanical stimulation of cardiac myocytes and hypertrophy, which is unhealthy
in the long term. Current therapies focus on preventing pressure overload, but
this is not always possible. A better understanding of how this pathway becomes
mechanically activated may reveal new targets for pharmacological inhibition of Nota Bene
hypertension-induced hypertrophy.
Smooth muscle cells respond to
fluid shear. Shear stress has been
Cellular strain in the musculoskeletal system is dependent shown to inhibit the migratory and
on tissue stiffness proliferative behavior of smooth
muscle cells. In vivo, shear stresses
Given the function of the musculoskeletal system to support and generate forces experienced by the endothelial cells
are transmitted to the smooth
during movement, cells within these tissues are also subjected to mechanical
muscle cells directly underneath
strain. Articular cartilage is subjected to cyclic compression during ambulation, them. Because smooth muscle and
which deforms the matrix and the cells residing within. Because such loading is endothelial cells are in contact and
important physiologically there are extensive efforts being made to characterize likely receive and transmit
these deformations both experimentally and numerically. The local mechanical mechanical and biochemical stimuli
environment of chondrocytes is determined in large part by the interaction of the to and from each other, co-culture
chondrocyte with its pericellular matrix, which possesses a high level of type VI mechanotransductive studies are a
promising approach. Similarly, an
collagen and proteoglycans (the combination of the chondrocyte and its pericel- active area of research is examining
lular matrix is sometimes referred to as a chondron). When cartilage is com- the response of endothelial cells to
pressed, chondrocytes have been demonstrated to undergo changes both in stretch. It turns out that stretch and
shape and volume. These deformations are nonuniform, owing in large part to shear stress can be competing
mismatches in stiffness between the cell and pericellular matrix. factors, with endothelial cells
sometimes orienting in an
“undecided” fashion if they
The lung and bladder are hollow elastic organs that are experience contrary input from
regulated by stretch different mechanical stimuli.

Sensation of deformation within elastic, hollow organs such as the lung and blad-
der is critical for their proper function. Within the lung, breathing generates cyclic
matrix stretch. In vitro, stretch has been shown to regulate lung cell growth,
remodeling of the cytoskeleton, activation of signaling molecules, and phospho-
lipid secretion. Within the bladder, during the storage phase, stretch stimulates
afferent neurons. Bladder epithelium is also sensitive to stretch, as release of
adenosine triphosphate (ATP), acetylcholine, and nitric oxide occurs under
mechanical stimulation. Elucidation of these mechanosensitive pathways may
provide potential pharmacological targets for future treatment of lung or bladder
dysfunction.

Cells can respond to hydrostatic pressure

In addition to fluid shear and substrate strain, cells can sense and respond to
hydrostatic pressure. The cells within bone, articular cartilage, intervertebral
discs, and the cardiovascular system are exposed to cyclically varying hydrostatic
pressure. Compared with fluid shear and substrate strain, the body of studies
investigating the mechanisms by which cells sense pressure is much smaller. This
is perhaps due, in part, to the fact that cells are fluid-filled structures, and their
compressibility under pressure is typically small. Red blood cells have been esti-
mated to undergo a change in cellular volume of only about 0.1% when subjected
to a relatively high 10 MPa load (note that typical peak systolic blood pressures of
120 mmHg corresponds to 0.02 MPa, or about three orders of magnitude less).
It may also be due to the insensitivity in some cell types when exposed to physio-
logical pressures. Within the bone marrow cavity, pressures are typically on
the  order of 10 mmHg but may be on the order of 100 mmHg during impact
loading. However, several in vitro studies have demonstrated that bone cells are
318 CHAPTER 11: Cellular Mechanotransduction

unresponsive to such pressures. It is interesting to note that, despite the apparent

insensitivity of cells to physiological pressures, almost all cells undergo a similar
alteration in cytoskeletal structure and morphology under high hydrostatic pres-
sure (>100 MPa). When subjected to these pressures, the cytoskeleton rapidly dis-
assembles and the cells become rounded. In many instances, apoptotic pathways
become activated, and cell death occurs.

Advanced Material: Parasitic flows

Because cells must be immersed in culture media during To isolate the effects of substrate strain versus fluid flow,
substrate stretch experiments, some degree of fluid investigators exploited the dependence of strain on sub-
movement (sometimes referred to as parasitic flow) is strate thickness during four-point bending. In particular,
unavoidable. Careful attention must be paid to minimize the beam-deflection equation for four-point bending is
parasitic flow, as it can severely confound interpretation
td
of experimental results. A classic example can be found ε= ( L − 1.33α ), (11.2)
in the field of bone mechanotransduction. Bone is a rela- α
tively stiff material, so the strains it is subjected to are where ε is strain on the substrate surface, t is the sub-
relatively small: on the order of hundreds to thousands strate thickness, d is the displacement, L is the length
of microstrains (abbreviated as με). Currently, the con- between inner supports, and α is the length between
sensus of the scientific community is that these small the inner and outer supports. By seeding cells on slides
strains are generally insufficient to stimulate bone cells of different thicknesses, investigators were able to apply
in vitro, but this view was not always accepted. For many different strains while keeping displacement rate (and
years, strains of one thousand με were believed to be suf- parasitic flow) constant, and vice versa. They found that
ficient to stimulate bone cells because in vitro cell- the cells were insensitive to changes in strain, but were
straining systems were based on bending (Figure 11.9A). highly sensitive to the displacement rate. This shows
Because bending resulted in a large displacement of the that cells were responding to fluid flow rather than sub-
substrate, these systems generated a significant amount strate strain. Because of these and other similar studies,
of parasitic fluid flow. In particular, the faster the dis- alternative approaches to cell stretching that minimize
placement through the media, the higher the levels of parasitic flows have been developed (Figure 11.9B).
fluid flow to which the cells were subjected.

(A)

(B)
Figure 11.9 Modalities for inducing substrate strain in vitro. (A) Cells are seeded on a substrate subjected to four-point
bending. The cells undergo a large degree of movement within the culture medium, exposing them to parasitic flow. (B) Cells are
subjected to substrate strain in a manner that reduces parasitic flow.

11.2 MECHANOSENSING ORGANELLES AND

STRUCTURES
In the last section, we reviewed different types of mechanical stimuli that act as
regulatory signals in cells in a variety of tissues and organs. We now shift the focus
to cellular structures and organelles that have been implicated in sensation of
mechanical forces. Much of what we have learned in the next few sections has
been a direct result of the in vitro reductionist approaches discussed in the previ-
ous section. As we will see, cells may possess very specialized structures to facili-
tate load transmission. We begin our survey with one of the most elegantly
designed structures for this purpose, which is found in hair cells in the inner ear.
 MECHANOSENSING ORGANELLES AND STRUCTURES 319

fixed tectorial membrane Figure 11.10 Inner ear hair cells. The
hair cells of the inner ear are responsible
for hearing. They transduce vibrations of
the cochlear basilar membrane into
stereocilia nerve impulses.

vibrating basilar membrane

It is also worth noting that, by virtue of the topic, this chapter is more biological
and less quantitative than the others in this text.

Stereocilia are the mechanosensors of the ear

The inner ear is one of the best-characterized mechanosensitive structures in the
mammalian body. Its function is to convert physical vibrations into the electro-
chemical nerve impulses that mediate hearing. The eardrum receives sound
waves and transmits the vibration to three bones (hammer, anvil, and stirrup),
which in turn transmit the vibrations into the cochlea (Figure 11.10). Inside the
cochlea, there is a basilar membrane, which has varying frequency sensitivities.
For a given frequency, a specific part of the membrane resonates and undergoes
more motion than other parts of the membrane (frequency range 20  Hz to 20  kHz).
Within the cochlea reside mechanosensory cells (hair cells) that transduce this
membrane motion into nerve impulses. Hair cells have evolved a remarkable
structure to sense this movement. Bundles of stereocilia (hair bundles) project
from the apical surface and are linked together by coupling structures. The longest
stereocilia are in direct contact with the membrane. When the membrane reso-
nates as a result of sound, the hair bundle deflects, leading to an influx of cations
into the stereocilia. Mechanosensing in these cells can be incredibly fast. Frog
cochlear hair cells have been shown to produce a nerve impulse in 40 μs. In con-
trast, visually evoked potentials in the retina occur on the order of tens of millisec-
onds, or several orders of magnitude slower.

Example 11.1: Deflection of hair-cell bundles

Consider the hair-cell bundles depicted in Figure 11.10. directly. However, a reasonable estimate would be the
Recall from Figure 1.1 the arrangement of distal tip links diameter of a typical molecule that can pass through the
that mechanically open membrane ion channels. Given channel pore. Five nanometers is a reasonable guess and
that the stiffness of hair-cell bundles has been measured likely accurate within an order of magnitude. Because the
to be about 500  μN/m, estimate the tip force required to stiffness of hair-cell bundles has been measured to be
open a channel. about 500 μN/m, a reasonable estimate for the tip force
would be 2.5 pN. For comparison, the binding force of a
The amount of channel displacement required for its
single integrin molecule is on the order of tens of
activation, in other words to allow the transition from a
piconewtons.
closed to an open configuration, has not been measured
320 CHAPTER 11: Cellular Mechanotransduction

Figure 11.11 The mechanosensing (A) (B)

of touch. Mechanosensory structures
and cells involved in touch sensation in
hairy (A) and non-hairy (B) skin all
involve nerves that terminate in
specialized effector cells. These include
slowly adapting pressure-sensing Merkel
cells (grey), rapidly adapting vibration-
sensing Pacinian corpuscles (black), and
Meissner’s corpuscles (blue) that are
nerve endings sensitive to light touch.

Specialized structures are used in touch sensation

Touch, the sensory experience resulting from force transmission to the skin, is one
of the five major senses by which humans are able to experience their environment.
Of these five senses, the cellular and molecular mechanisms underlying touch are
the least well understood. Touch sensation is initiated by mechanical deformation
of nerve endings within specialized structures. In hairy skin, oval-shaped Merkel
cells envelope hair follicles, allowing them to be activated by hair movement (Figure
11.11). Alternatively, nerve endings may be free, or encapsulated within structures
called corpuscles. Meissner’s corpuscles are a type of nerve ending involved in the
sensation of light touch. They consist of flattened support cells arranged in horizon-
tal lamellae and encapsulated by a connective-tissue capsule. Within the corpuscle,
a single nerve fiber meanders between the lamellae. Other similar structures
include Pacinian corpuscles, which are involved in sensing vibration, and Ruffini
corpuscles, which are slow-response mechanoreceptors activated by skin stretch.

Primary cilia are nearly ubiquitous, but functionally mysterious

A projecting structure that is well suited for mechanosensing in non-sensory
cells is the primary cilium. It is a rod-like structure that projects from the surface
of the plasma membrane into the extracellular environment. Primary cilia are
found in almost all cells, and despite having been initially described near the
turn of the century, their function has only recently begun to be elucidated.
Unlike motile cilia, primary cilia are solitary and nonmotile. However, like motile
cilia, primary cilia are microtubule-based structures with axoneme cores. The
axonemes of primary cilia are composed of nine microtubule doublets, but lack
the central pair of doublets that motile cilia have. They are described as having a
9 + 0 (as opposed to 9 + 2 for motile cilia) microtubule architecture. By extending
into the extracellular space and possessing mechanical characteristics that allow
flow-induced bending, they are uniquely situated to serve a mechanosensory
role. Several investigators have demonstrated that cilia passively bend under
fluid flow and recoil after cessation of flow (Figure 11.12). Ciliary bending by
micropipette suction is sufficient to increase intracellular Ca2+ concentration
in kidney epithelial cells. Cilia have also been implicated as sensors of fluid flow in
kidney, liver, bone, blood vessels, and the embryonic node.
Because flow-induced deflections would result in the generation of membrane
tension that could serve to open stretch-activation ion channels, efforts have been
made to identify channels that localize to the cilium. One such channel is the
poly­cystin-1/2 complex. Polycystin-1 is an integral membrane protein encoded
by the gene PKD1, and polycystin-2 is a cation channel encoded by the gene PKD2
that heterodimerizes with polycystin-1. Mutations in either PKD1 and PKD2
result in autosomal dominant polycystic kidney disease (PKD). Polycystin-1 and
polycystin-2 co-distribute in renal primary cilia, and loss of polycystin-1 and pol-
ycystin-2 function disrupts flow sensing in kidney cells, suggesting that aberrant
 MECHANOSENSING ORGANELLES AND STRUCTURES 321

Figure 11.12 ​Primary cilia bending

no flow with flow. This side view of a bone cell
and its primary cilium in the absence
(top) and presence (bottom) of fluid flow.
Deflection of the cilium under flow can
primary cilium be observed. (From Gefen A (2011)
Cellular and Biomolecular Mechanics
and Mechanobiology. Copyright
permission Springer Science, New York.).

substrate

fluid flow

substrate

mechanosensing due to dysfunction in this complex may contribute to PKD

pathogenesis.

Cellular adhesions can sense as well as transmit force

Adhesion sites, or plaques, are thought to serve as mechanosensory structures
because they would likely experience high stress under both substrate strain and
fluid shear. Most connective tissues are composed primarily of structural poly-
mers such as collagen or fibronectin. Cells embedded within these tissues are
anchored within this fiber network at discrete sites of adhesion (sometimes
referred to as focal contacts or focal adhesions). If the fiber network were
deformed, the cells would also deform, with the loads transferred from the net-
work to the cell through these focal contacts (see Figure 11.8). A variety of linker
proteins mechanically couple focal adhesions to the cell cytoskeleton, so, during
fluid shear, loads would be transmitted from the apical surface of the cell, through
the cytoskeleton to points of cell–substrate attachment (see Figure 11.1).
Several proteins within adhesion complexes that are directly involved in this load-
bearing path also possess signaling functions. Such proteins would be ideally
suited to undergo force-induced conformational changes that could serve to
transduce mechanical force into an intracellular biochemical signal. Such pro-
teins include focal adhesion kinase, vinculin, talin, tensin, paxillin, and others.
Cell–cell adhesions have also been suggested as sites of mechanotransduction,
because they incorporate signaling molecules and would likely be sites of high
stress. Cell–cell junctions are particularly important in endothelial cells, where
they form a tight barrier that prevents vessel leakage. In these cells, PECAM-1
localizes to the cell–cell junction, and is rapidly activated upon exposure to shear
322 CHAPTER 11: Cellular Mechanotransduction

stress. There is also evidence that cadherins, the primary component of adherens
junctions, and vascular endothelial growth factor receptor 2 (VEGFR-2), the
receptor for VEGF, forms a complex with PECAM-1 that is activated by fluid shear.

The cytoskeleton can sense mechanical loads

As described in Section 7.1, the cytoskeleton is composed of three components:
actin filaments, microtubules, and intermediate filaments. Of these components,
the actin cytoskeleton has received much attention as a mechanosensing struc-
ture. There are several ways in which the actin cytoskeleton may contribute to
mechanosensing. First, it may play an indirect role, transmitting loads to sites of
mechanotransduction. In cells exposed to fluid flow, the cytoskeleton could serve
to transmit shear forces to mechanotransduction sites, such as focal adhesions,
where force-induced conformational changes could occur. Alternatively, the actin
cytoskeleton has been proposed to serve a mechanosensory function by trans-
mitting loads to the nucleus and directly manipulating DNA, exposing previ-
ously hidden transcription sites to initiate the synthesis of transcription factors or
other proteins. Structural nuclear matrix proteins including nesprin and lamins
have been implicated as being required for sensing of some mechanical loads.
In terms of direct mechanosensing, the cytoskeleton may itself serve as a structure
in which mechanical forces are transduced into a biochemical signal. Loading may
cause elements of the network to slide or deform, liberating trapped signaling mol-
ecules or exposing (or hiding) sites of enzymatic activity. Indeed several signaling
molecules as well as other structural molecules have been found to associate with
the actin cytoskeleton. Structural rearrangements of the network under load could
lead to activation or inactivation of intracellular signaling by bringing two mole-
cules closer together or farther apart (Figure 11.13). Microtubules have been
shown to buckle when loaded with external compressive loads, leading to loss of
the ATP cap and entry into the depolymerization catastrophe. Many of the cytoskel-
etal transduction mechanisms are more theoretical than accepted. It is a compel-
ling research direction because rarely are so many proteins with both signaling and
structural roles in such close proximity.

Example 11.2: Microtubule buckling

One proposed model of microtubule-based mechano­ value (Table 3.3) for a 10 μm cell, the buckling load for
transduction is that externally applied forces cause each molecule is
buckling that acts as an intracellular signaling trigger.
What magnitude of pressure could be detected by such a (3.142)2 360 × 10 −25 Nm 2
Fb = ≈ 3.5 pN. (11.4)
mechanism? (10 × 10 −6 m)2

To estimate an upper bound on sensitivity, consider a
cell responding to an external pressure. Ignoring the If we have 1000 microtubules in a typical cell, the result-
contributions of the cytoplasm, imagine that this pres- ing pressure would be
sure is supported by a microtubule network radiating
from the perinuclear region to the cell periphery like the 14nN
Pb = ≈ 45Pa. (11.5)
spokes of a bicycle wheel. Recall that the buckling load is 4π(5 × 10 −6 m)2
given by Equation 3.38 as
To lend some scale, typical blood pressures are on the
π 2 EI order of tens of millimeters of mercury or kilopascals.
Fb = 2 . (11.3) Microtubule bucking has the potential to be an extremely
L
sensitive mechanism. Of course, in reality this is the the-
Experimental measurements of flexural rigidity for oretical lower bound on the detectable pressure. Fluid
microtubules are highly variable and span orders of pressurization of the cytoplasm would be expected to
magnitude ranging from 1 to 200 pN/μm2. Using a typical carry a significant amount of external load.
 MECHANOSENSING ORGANELLES AND STRUCTURES 323

(A) Figure 11.13 Potential roles of the

cytoskeletal network in
mechanotransduction. These include
indirect force (A) indirect force transmission, (B)
transmission exposure (or hiding) of signaling
domains, release (or capture) of
signaling molecules with load, and (C)
even simple buckling-induced
(B) exposure of site depolymerization. Mechanotransduction
involves converting a physical signal to a
biochemical one, so the cytoskeleton is
a natural candidate because the
network brings structural and signaling
roles together.
release of signal

actin-binding
protein

(C)

Mechanosensing can involve the glycoproteins covering

the cell
The glycocalyx, mentioned briefly in Section 9.5 as a source of bending stiffness, is
a layer of membrane-bound macromolecules that surrounds several cell types. Its
composition and thickness depend on tissue type, and, even within similar tis-
sues, on anatomical location. It consists of a meshwork of proteoglycans, glyco­
saminoglycan side chains, and associated proteins from the surrounding fluid
that bind this meshwork. Physically, it is ideally suited to serve a mechanosensory
role, as it acts as the interface between the extracellular fluid and the cell mem-
brane/cytoskeleton, particularly in cells that respond to fluid flow. In addition,
because the proteoglycans, with their numerous side chains, project into the
luminal region, they have the potential to deflect under shear flow. Through their
associations with the apical membrane/cytoskeleton, they may indirectly con-
tribute to mechanotransduction by transmitting forces to the cell. Conversely, the
glycocalyx could serve to decrease cell sensitivity to fluid flow by shielding the
324 CHAPTER 11: Cellular Mechanotransduction

Figure 11.14 Longitudinal (A) and (A) (B)

transverse (B) view of pericellular
matrix surrounding an osteocyte
process in bone. The cell process and
the mineralized bone tissue are both
dark. The light-colored space between
the two is occupied by a glycoprotein
gel as well as discrete linker proteins.
(Courtesy of Lidan You, University of
Toronto.)

membrane from shear stresses. So far, studies investigating the role of the glycoca-
lyx in flow sensing have focused primarily on endothelial cells.
Though the role of the glycocalyx in flow sensing has been studied primarily in the
context of endothelial mechanotransduction, membrane-associated proteins
and glycoproteins have also been proposed as facilitators of mechanotransduc-
tion in bone. Within the lacunar–canicular system, osteocyte processes are sur-
rounded by a pericellular matrix that is believed to be similar in composition to
the endothelial glycocalyx (Figure 11.14). In the presence of fluid flow, drag forces
are generated on the pericellular matrix that are transmitted to the cell surface.
Theoretical calculations predict that these forces are much higher than those gen-
erated by fluid shear alone.

The cell membrane is ideally suited to sense mechanical loads

Nota Bene Although evidence for cytoskeletal mechanosensing is relatively scarce, the role
of the membrane in mechanosensing is much better developed. The membrane is
Membrane-spanning regions. an ideal site, as it contains specific structures that confer sensitivity to the envi-
Membrane-spanning regions are
commonly α-helices and can often
ronment while maintaining the integrity of the intracellular components. Recall
be identified by looking for a series of that the bilayer represents a hydrophobic layer of a specific thickness (about
hydrophobic amino acids of the 7 nm). Any protein that contains a hydrophobic region, known as membrane-
correct length from a genetic spanning regions, of the same thickness can become embedded within the mem-
sequence. This can be a critical step brane. Proteins such as receptors that have binding or activation domains on both
in estimating a protein’s structure the extracellular and intracellular sides of the cell can facilitate signaling through
and function from sequence
the cell membrane by conformational changes that occur when a ligand binds or
information alone. In the case of
multiple membrane-spanning dissociates from one side. Analogously, changes in the physical and chemical
domains, the protein will fold back on properties of the lipid bilayer after exposure to mechanical force can play a critical
itself, with each membrane transition role in supporting mechanotransduction by potentially inducing conformational
forming intracellular or extracellular changes in membrane proteins.
loops. If a membrane-spanning
protein has an odd number of Perhaps one of the best-studied mechanisms of mechanosensing, membrane
membrane-spanning domains, the tension, is well known to contribute to increased probability of ion channel open-
carboxy (C)- and amino (N)-terminal ing. Stretch-activated channels have been studied in detail in membrane patches
ends of the protein will be on or fragments of bilayers anchored to the end of a micropipette. Pipette pressure
opposite sides of the membrane. can be changed to regulate the areal strain in the patch, while opening and closing
Likewise, if the protein has an even
number of membrane-spanning
of the channels is assayed by measuring electrical conductance.
domains, the C and N termini will be Other than tension, alterations in other membrane physical properties may be
on the same side of the membrane, involved in mechanosensing. Shear-induced alterations in the membrane may
either intracellularly or extracellularly.
initiate mechanotransduction through enhancing/inhibiting interactions
between signaling molecules. Several studies have demonstrated that membrane
fluidity can be altered by fluid flow. Membrane fluidity refers to the changes in
 MECHANOSENSING ORGANELLES AND STRUCTURES 325

membrane viscosity that may arise from local ionic transients coupled with aggre-
gation of proteins. Alterations in membrane fluidity may induce conformational
changes in transmembrane proteins, or enable a change in the rate of interaction
of membrane proteins. In particular, because proteins with membrane-spanning Nota Bene
domains float within the cell membrane, they can be thought of as being confined Small GTPases regulate
within a two-dimensional space. This confinement can greatly alter their chemi- cytoskeletal behavior. G-proteins
cal kinetics. Proteins are much more likely to encounter and interact with each are a particular type of GTPase.
other if they are constrained to the two-dimensional space of the membrane than Another type of GTPase linked to
if they must find each other in the full three-dimensional cytoplasm. Mechanically many mechanobiological functions
includes the small GTPases,
induced changes in membrane fluidity may serve to alter normal interactions consisting of Rho, Rac, Rap, and
within this space, or allow aggregation of new groups of proteins that breach the others. These GTPases are activated
typical compartmentalization. when bound to a GTP. Knocking out,
or mutating, various GTPases leads to
Finally, membrane-bound proteins may also act as direct mechanical receptors alterations in cell morphology, actin
independent of changes in membrane physical properties. Fluid shear has been organization, migratory capability,
shown to be sufficient to activate heterotrimeric G-proteins reconstituted into spreading, and adhesion. Though
liposomes, even in the absence of any other potential mechanotransducing mol- these molecules have been
ecules. As described briefly in Section 2.2, G-protein-coupled receptors are mem- implicated in the control of several
brane-spanning receptors that are linked to G-proteins. When activated, the pathways related to motility and
adhesion, the role of these small
receptor undergoes a conformational change that allows the guanosine triphos-
GTPases in mechanotransductory
phate (GTP) to exchange for the guanosine triphosphate (GDP) on the G-protein, pathways is not well understood.
resulting in activation of the G-protein.

Lipid rafts affect the behavior of proteins within the membrane

As discussed previously, protein interaction kinetics within the membrane may
be enhanced by their confinement within a two-dimensional space. These inter-
actions may be enhanced even further by the inherent heterogeneity in fluid
properties within the membrane. There are separate phases or domains within
the lipid bilayer. One such microdomain, 20–200 nm in size, occurs because of the
presence of cholesterol, and other lipids known as sphingolipids. These microdo-
mains were termed lipid rafts in the 1970s. The hydrophobic tails of cholesterol
have a slightly different structure than the rest of the membrane phospholipids.
Specifically, their tail domains are a bit longer. This causes cholesterol to self-
aggregate into small regions with lower fluidity than the bulk of the phospholip-
ids. Proteins suspended in these domains tend to remain within them and have
difficulty passing laterally in and out of the raft. The reaction rates of proteins con-
fined to lipid rafts can be higher as a result of having a locally higher reactant con-
centration. These types of lipid raft are referred to as planar to distinguish them
from a second type of lipid raft that occurs because of the action of caveolin.
Caveolin is a small protein with cytoplasmic C and N termini. The termini are
linked, causing an increase in curvature (Figure 11.15). In caveolin-rich areas of

(A) (B) Figure 11.15 Electron micrograph of

a thin section (A) and freeze-
fracture (B) of the cell membrane.
The protein caveolin causes the
membrane to bend, forming infolded
regions. These microdomains form
subregions within the membrane with
altered chemical kinetics. (From,
Anderson RGW (1998) Annu. Rev.
Biochem. 67.)

0.2 µm 0.1 µm
326 CHAPTER 11: Cellular Mechanotransduction

Nota Bene
the membrane, tiny invaginations of the membrane known as caveolae (Latin for
“small cave”) occur. Both forms of lipid raft have been demonstrated to mediate
Membrane structure. Lipid rafts mechanotransduction in a variety of cell types, although the specific mechanism
are not the only structure within the
has not been established. However, similar to the mechanisms described above in
cell membrane. Barriers within the
membrane can form from interaction the membrane, it has been proposed that mechanical force could alter the phys-
with the actin cytoskeleton. The part icochemical properties of the lipid rafts, leading to the direct activation of indi-
of the actin cytoskeleton that vidual signaling molecules residing within them, or facilitating interactions
underlies and supports the between molecules.
membrane is sometimes referred to
as the cortical cytoskeleton. It is
periodically attached to the
membrane via anchoring proteins. 11.3 INITIATION OF INTRACELLULAR SIGNALING
Also, there are large cell–cell
signaling structures such as synapses In the first section we provided a survey of the types of mechanical stimulus
and gap junctions and cell–cell that cells are exposed to, and in the second section we discussed structures in
mechanical connections known as
cells that transmit loads from the environment to mechanically sensitive mol-
desmosomes. There are also cell–
substrate interactions, such as focal ecules. This section focuses on the final event of mechanotransduction, load-
adhesions and hemidesmosomes, induced protein conformational changes and the generation of an intracellular
which are responsible in part for signaling cascade. In particular, we discuss two potential mechanisms by which
attaching the cell to a surface. mechanical force can be transduced into a biochemical signal at the molecular
level, opening of mechanosensitive ion channels, and exposure of cryptic bind-
ing sites.

Ion channels can be mechanosensitive

Many membrane-bound proteins are involved with cellular exchange of mole-
cules between the cytoplasm and the extracellular environment. Ion channels
are complexes of proteins that form small passages in the membrane for ion
Nota Bene flow. Channels can be selective based on species, size, charge, and chemical
Pumps. We learned that channels interactions. They are passive transport mechanisms in the sense that they only
are passive transporters of ions, allow ion passage in the direction of their chemical gradient (from high to low
allowing their passage in the concentration). Examples of ions that move through channels are hydrogen,
direction of their chemical gradient.
calcium, sodium, chloride, and potassium. Channels can be gated and change
To move ions against their chemical
gradient, active proteins termed configurations from an open to a closed state and back. Many biochemical fac-
pumps or transporters are required. tors can affect the gating characteristics of a channel. As we have touched upon
These proteins require chemical in this chapter, mechanical signals can alter channel behavior, allowing the
energy for their activity. This energy transduction of forces into a biochemical event. In particular, mechanosensi-
can come in the form of ATP or can tive  ion channels are membrane-bound, pore-forming proteins that open
involve the simultaneous travel of in response to mechanical forces. Mechano-neurosensing in both hearing and
another species along its chemical
touch response is almost exclusively due to mechanically gated ion channels.
gradient. The latter are sometimes
referred to as exchangers. They are thought to mediate mechanosensation in several nonsensory cells
as well.
The identification of mechanosensitive ion channels has generally relied on one
of two experimental approaches. In the first approach, membrane tension is
applied to cells or pieces of isolated membrane containing candidate channels.
Such an approach led to the first demonstration of mechanical gating of one of
the most well-characterized mechanosensitive ion channels, the bacterial mech-
anosensitive channel of large conductance (MscL). It has been shown that when
there is a large pressure imbalance between the outside and inside of a bacte-
rium carrying this channel, the MscL opens and allows the bacterium to jettison
some of its innards and relieve the pressure. In the second approach, randomly
generated genetic mutations are screened for alterations in specific touch
responses (such as gentle touch or pressure sensing). The specific mutated genes
responsible for the changed response are identified through approaches such as
linkage mapping. Examples of mechanosensing mutants found with screens
include touch-insensitive mutants in the worm Caenorhabditis elegans, defects
in bristle mechanosensation in Drosophila mela­nogaster mutants, and lateral
 INITIATION OF INTRACELLULAR SIGNALING 327

line mechanosensation mutants in zebrafish. Three classes of mechanosensitive

ion channel have been identified to mediate mechano-neurosensing: epithelial
sodium channel (ENaC); transient receptor potential (TRP) channel; and the
two-pore-domain potassium channel protein.

Example 11.3: Is membrane tension sufficient to open a channel?

To see if bilayer tension is a feasible mechanism for regu- approximately 3.5 nm. A lower bound on the change in
lating ion channel opening, estimate how much work is area would be
performed on a channel opening under membrane sur-
∆A = πr 2 = 38.5 nm 2 . (11.7)
face tension.
The channel alamethicin is known to increase its effec- In Section 1.5 we estimated the surface tension in a neu-
tive two-dimensional area when it transitions from trophil to be about 35 pN/μm. The associated work is
closed to open. If the membrane is exerting a surface
W = 38.5 nm 2 × (10 −9 m/nm)2 35 pN/µm
tension, n, on the channel when this happens, there will
(11.8)
be work done. Specifically, (10 −12 N/pN) (µm/10 −6 m) = 1.35 × 10 −21 J

W = n∆A, (11.6) or about one zeptojoule. Although this is a tiny amount of

energy, it is on the same scale as the channel activation
where ΔA is the change in area. Next, we can estimate the energy, which has been estimated at 14kBT or 50 zJ. Thus,
change in area. Thioredoxin is a molecule known to pass particularly at higher membrane tensions, this may be an
through alamethicin channels and has a radius of important mechanotransduction mechanism.

Conceptually, the mechanism whereby membrane tension could lead to chan-

nel opening is attractive, owing to its simplicity. Because the channel is embed- Nota Bene
ded within the bilayer, it exists in a two-dimensional environment. In-plane The MscL Channel. Although much
membrane tension could directly pull on the proteins that make up the channel insight into mechanosensitive ion
and cause it to open. Of course, on this scale entropic effects need to be consid- channels has been gained through
ered. Channel kinetics are often described in terms of the probability of the chan- studying the MscL channel, it is
important to note that the channel is
nel being in one of two configurations, open or closed, with a transition or found only in prokaryotes. Because
activation energy required to move between the two. Open and closed transi- the cell wall of bacteria is reinforced
tions can be observed directly in experiments examining single channels, or the with a structural component known
aggregate effects of many channels can be superimposed to modulate mem- as peptidoglycan, there are important
brane conductance. differences between this channel and
those found in eukaryotic bilayers.

Hydrophobic mismatches allow the mechanical gating of

membrane channels
Another level of complexity in bilayer–protein interactions involved in mechani-
cal gating of membrane channels is hydrophobic mismatch. This phenomenon
involves the mechanical deformation of the membrane after protein insertion.
When the thickness of the hydrophobic region of the membrane protein is differ-
ent than the thickness of the bilayer, it will cause localized distortion, including
squashing, stretching, and/or tilting of the lipid chains (Figure 11.16). This effec-
tively results in a mechanical coupling of the lipids to the protein. In this situation,
in-plane stretching of the bilayer will lead to changes in membrane thickness,
potentially altering the conformation of the membrane protein. In support of this
hypothesis, some channels are known to get shorter when they are in the open
configuration. They would be a better fit within a membrane that had been
thinned due to stretching. In fact, experimental evidence has shown that when
certain channels are placed in a thinner bilayer, they are more likely to be in the
open configuration.
328 CHAPTER 11: Cellular Mechanotransduction

Figure 11.16 Hydrophobic

mismatch. When a hydrophobic
protein or a protein domain that has the favored
same thickness as a membrane is
inserted, it is energetically favored.
However, if it is longer or shorter it will
distort the adjacent bilayer, increasing
the free energy. unfavored

unfavored

Example 11.4: Can membrane thinning contribute to channel opening?

Estimate the increase in free energy incurred by mem- when the membrane is stretched just to the point of rup-
brane thinning and how much hydrophobic area this ture. This corresponds to an areal strain of roughly 3%.
would expose in a membrane-bound protein. However, for the volume to be preserved, the membrane
will need to thin by a corresponding 3% or 0.15 nm for a
Consider the energetics of the hydrophobic mismatch
5 nm thick membrane. Next we need to determine how
ion-channel mechanism. When a hydrophobic protein
much hydrophobic area this would expose on the pro-
is transferred from an organic solvent to an aqueous
tein. For the MscL channel, it is roughly cylindrical in
environment, the free energy increases by an amount
shape, with a 5 nm diameter. So, 2.4 nm2 of area is
pro­portional to the exposed area, roughly 17 mJ/m2. So
exposed to water, corresponding to an increase in free
the energy available to open a channel would be roughly
energy of 40 zJ. This is quite comparable to the channel
the free energy increase due to the hydrophobic surface
activation energy.
exposed. Again considering an upper bound, the maxi-
mum membrane thinning will occur at the lytic limit or

Advanced Material: Incompressibility of bilayer leads to thinning with stretch.

Hydrostatic pressures (σxx =  σyy =  σzz) have been applied where

as high as 100 atmospheres (107 N/m2) with no signifi-
cant effect on lipid density. Compressibility in response σ xx + σ yy + σ zz ε xx + ε yy + ε zz
σh = and ε d = . (11.10)
to hydrostatic pressures is quantified by the bulk modu- 3 3
lus EB, which is the ratio of the hydrostatic stress and the
dilatational strain. Results from these studies have produced estimates of
bilayer bulk modulus on the order of 1010 N/m2.
σh Essentially, for physiological pressures the bilayer is
EB = (11.9) incompressible. Therefore any in-plane stretching
sd
resulting in an area increase is accompanied by a cor-
responding decrease in thickness.

Mechanical forces can expose cryptic binding sites

In addition to force-induced activation of membrane ion channels, another
molecular mechanism that has the potential to transduce mechanical loads is
force-induced exposure of cryptic (that is, hidden) binding sites. Consider a pro-
tein coupled to or within a load-bearing structure, such as a focal adhesion or the
cytoskeleton. When the cell is loaded, mechanical forces are transmitted to the
protein and may cause conformational changes (such as partial unfolding) that
 INITIATION OF INTRACELLULAR SIGNALING 329

binding ligand Figure 11.17 Cryptic binding sites

can act as mechanotransducers.
Force-induced unraveling of a protein
can expose cryptic binding sites. This
allows them to become enzymatically
active or to bind their ligands.

hidden cryptic
binding site

expose new binding sites at which chemical reactions can occur (Figure 11.17).
Certain structural motifs exist that may facilitate force-induced conformational
changes in a predictable and force-dependent manner. Some proteins that
mechanically link integrins to the cytoskeleton (and would likely be exposed to
large forces during fluid shear and/or cell stretching) have been found to possess
repeating sequences of structural motifs or modules (such as α-actinin and talin).
The mechanical stability of these modules would dictate the sequence in which
the protein unravels under force, with the more unstable modules unfolding
before the stable ones. These intermodule differences in mechanical stability
could result in sequential exposure of cryptic binding sites with increasing load
magnitudes. A molecule that undergoes such sequential conformational changes
under increasing loads could allow the cell to sense force magnitude.

Bell’s equation describes protein unfolding kinetics

Protein folding and unfolding occurs on very small length scales, so entropic
effects need to be considered. In fact, folding and unfolding are more stochastic
than deterministic. In Section 10.1 we described how Bell’s model can be used to
describe adhesive bond rupture and breakage. Perhaps, not surprisingly, Bell’s
equation is also effective in describing protein unfolding. Specifically, the kinetic
rate of unfolding is given by

F ∆x
k = k0 e , (11.11)
kBT

where k0 is the unfolding rate of an unloaded protein, F is the force, and Δx can be
related to the extension of the protein with unfolding but is more correctly termed
the effective energy barrier width.

Molecular conformation changes can be detected fluorescently

The capacity for force to induce conformational changes through the mechanisms
described above has primarily been investigated using one of two approaches. In
the first approach, molecular dynamics can be used to directly simulate the molec-
ular rearrangements that occur during an unfolding trajectories. The second
approach is to use genetically encoded fluorescent sensors to investigate these
phenomena experimentally. In this approach, a biosensor is constructed by fus-
ing two different fluorescent proteins to a candidate mechanosensing protein.
The fluorescent proteins are chosen such that the emission spectrum of one of the
molecules (donor) overlaps the excitation spectrum of the other molecule (accep-
tor). If the proteins are sufficiently close to one another and the donor molecule is
excited, a physical phenomenon called Förster resonance energy transfer (FRET)
occurs. Specifically the donor and acceptor are nonradiatively coupled, leading to
excitation of the acceptor. For any donor–acceptor pair, the FRET efficiency is
dependent on the distance between the two as well as their relative orientation to
330 CHAPTER 11: Cellular Mechanotransduction

Figure 11.18 FRET biosensors can CFP excitation

change fluorescence depending on
proximity or orientation. For
CFP emission
construction of a FRET biosensor, YFP excitation
fluorescent proteins must be chosen YFP emission
such that the emission spectrum of one
of the molecules (in this case cyan
fluorescent protein (CFP)) overlaps the
excitation spectrum of the other
molecule (in this case yellow fluorescent
protein (YFP)). FRET efficiency changes,
depending on the relative distance and 350 400 450 500 550 600 650
orientation between the two molecules. wavelength (nm)

photon
FRET

CFP YFP

Nota Bene one another (typically, FRET occurs at distances of 10 nm or less). By monitoring
changes in FRET efficiency (the ratio of acceptor–donor emission intensity) in a
Enzymes strain their substrates. cell under mechanical load, one can probe whether mechanical forces can be
It is well known that enzymes exert
strain on their substrates upon
transformed into a conformational change in the candidate protein of interest
binding, thereby catalyzing the (Figure 11.18).
underlying reaction. Straining the
enzymatic substrate could therefore
potentially regulate enzyme activity.
In this way, mechanical force has
11.4 ALTERATION OF CELLULAR FUNCTION
been hypothesized to modulate
enzymatic activity independently of After the initial molecular mechanotransduction event, an intracellular biochem-
exposure of cryptic binding sites. ical signaling cascade must be initiated if these conformational changes are ulti-
However, this mechanism has been mately to lead to alterations in cellular function. In Section 2.2, we provided a
relatively unexplored. survey of intracellular signaling mechanisms and pathways. Many of these have
been shown to be activated by mechanical stimulation. In general, the investiga-
tion of specific signaling pathways is driven, in large part, by physiological rele-
vance for a given cell type. In vascular endothelial mechanotransduction, we
might be interested in a pathway that ultimately leads to regulation of vascular
tone. In bone cells, we might be interested in a pathway that has been shown to
lead to increased mineralization. Many detailed reviews of mechanically regu-
lated intracellular signaling pathways are available (many of which are specific for
certain cell or tissue types); we will not try to recapitulate them here. Rather, we
focus on responses to mechanical stimuli that appear common across the many
cell types. We note that, the material presented is somewhat general, but these are
very active areas of research.

Intracellular calcium increases in response to

mechanical stress
One of the most commonly investigated intracellular signaling systems in mecha-
notransduction is intracellular calcium signaling. Calcium signaling is important
during mechanotransduction both in excitable cells (in other words, cells that
exhibit action-potential propagation such as nerves and cardiac myocytes) as well
as non-excitable cells (epithelial cells, bone cells, chondrocytes, etc.). One reason
for its widespread investigation in mechanotransduction is that it is a ubiquitous
 ALTERATION OF CELLULAR FUNCTION 331

second messenger molecule, capable of eliciting a wide range of effects that can Nota Bene
affect many different downstream signaling molecule pathways.
Intracellular calcium and cell
Intracellular Ca2+ signaling can be visualized in real time through the use of fluo- mechanics. A variety of processes
rescent Ca2+ indicators (calcium ionophores or chelators). In vitro, such dyes can associated with the generation or
be loaded into cells relatively easily, and, when combined with fluorescent imag- resistance to mechanical force
ing, allow the visualization of the initiation and propagation of intracellular Ca2+ depend on intracellular calcium.
Muscle contraction depends on
waves in response to mechanical stimuli. Mechanical activation of intracellular
calcium interacting with the troponin/
calcium signaling, in general, occurs very quickly (generally on the order of sec- tropomyosin system for actin–myosin
onds after the initial mechanical stimulus), and the calcium waves can spread contraction to occur. Further, calcium
throughout the cell rapidly (on the order of milliseconds). Mechanically induced is required for certain types of cell
intracellular Ca2+ elevation is thought to be mediated by the rapid opening of adhesion, such as the calcium-
mechanically sensitive calcium channels in the cell membrane, or release of cal- dependent adhesion molecule
cium from intracellular Ca2+ stores. Once elevated, intracellular calcium then cadherin.
goes on to activate other, downstream pathways.

Nitric oxide, inositol triphosphate, and cyclic AMP, like Ca2+, are
second messenger molecules implicated in mechanosensation
Although intracellular calcium signaling is the most well-studied of the second Nota Bene
messenger systems, several other second messenger molecules have been Ionophores and chelators. An
shown to exhibit rapid changes soon after exposure to mechanical stimulation. ionophore allows the transfer of ions
Recall from Section 2.2 that second messengers can be broadly classified into through a hydrophobic barrier such
one of three categories: hydrophobic molecules associated with the cell mem- as the bilayer, in which they would
normally be insoluble. Ionophores
brane, dissolved gases, and molecules that do not freely cross lipid membranes. typically encase the ion in a polar
Like Ca2+, cyclic adenosine monophosphate (cAMP) is a second messenger mol- interior while exposing a hydrophobic
ecule that falls into the latter category. cAMP is synthesized from ATP by a pro- exterior to the outside. A chelator
cess that is catalyzed by adenylyl cyclase. Activation of adenylyl cyclase is typically forms multiple stable bonds
generally associated with activation of G-protein-coupled receptors. Levels of with metal ions, inactivating them
cAMP can also be regulated by cyclic nucleotide phosphodiesterases, which from their normal function or effect.
It is derived from the Greek word for
degrade cAMP. cAMP has been shown to be rapidly regulated after exposure to
“lobster claw”, Chelè.
mechanical loading in a variety of cell types. Adenylyl cyclase exists in several
different isoforms whose expression can be highly tissue-specific. The activity of
many of these isoforms can be modulated by Ca2+, allowing cross-talk between
the Ca2+ and cAMP pathways.
Recall from Section 2.2 that inositol triphosphate (IP3) is a second messenger
molecule associated with the cell membrane. It is synthesized by hydrolysis of
the molecule phosphatidylinositol 4,5-bisphosphate (PIP2) by phospholipase
C, a phospholipid that is localized in the plasma membrane. Upon cleavage of
PIP2 to IP3 by phospholipase C, IP3 diffuses to the endoplasmic reticulum and,
after binding to IP3 receptors, initiates intracellular calcium release. Several
studies have implicated IP3 in mediating mechanically stimulated intracellular
Ca2+ release.
The dissolved gas nitric oxide can be a potent second messenger because of its
ability to quickly diffuse though the cytoplasm and across lipid membranes.
Shear-induced nitric oxide production in endothelial and bone cells has been
widely observed. Production of nitric oxide is catalyzed by nitric oxide synthase
(NOS). In mammals, the endothelial isoform of NOS, eNOS (also known as
NOS-3) is a primary regulator of vascular tone. In particular, eNOS-derived
nitric oxide has been shown to be a potent vasodilator, relaxing smooth muscle
within blood vessels.
It is noteworthy to point out that second messenger molecules like those described
above have often been used as primary outcome measures for assessing the
mechanosensory role of a particular protein. In general, implicating a molecule as
having a mechanosensory function by inhibitory strategies can be more straight-
forward when assaying a cellular response that occurs fairly rapidly after sti­
mulation. This is because after long exposures to loading, numerous molecular
332 CHAPTER 11: Cellular Mechanotransduction

Nota Bene
interactions often occur, and thus it becomes harder to interpret whether effects
on mechanosensation caused by inhibition of a particular molecule are because
Parallel activation of signaling the molecule in question has a direct mechanosensory function, or whether it is
pathways. Although several studies
merely involved somewhere “downstream” in the signaling pathway.
have implicated calcium in mediating
mechanosensing in a variety of cells,
in many cases these cells retain the
capacity to exhibit some
Mitogen-activated protein kinase activity is altered after
mechanically induced responses exposure to mechanical stimulation
even after complete inhibition of
loading-induced intracellular Ca2+ Downstream of second messengers, protein–protein signaling cascades involv-
activation. This suggests that there ing phosphorylation and dephosphorylation can be critical (see Section 2.2). A
are several signaling mechanisms particularly prominent signaling mechanism often activated within minutes
acting in parallel with the Ca2+ after exposure to mechanical stimulation is the group of the mitogen-activated
response.
protein kinases (MAP kinases). MAP kinase phosphorylation in response to most
(including mechanical) stimuli can be highly dynamic, occurring soon after
stimulation and remaining in the phosphorylated state for many minutes. One
well-studied MAP kinase is extracellular-signal-regulated kinase 1 and 2
(ERK1/2), which regulates cell growth and differentiation. Another is c-jun
Nota Bene N-terminal kinase (JNK), which is a so-called stress-activated protein kinase.
Note that “stress-activated”, in this context, refers to systemic stresses and not
MAP kinases and cell growth. As
the name suggests, the MAP kinases necessarily mechanical ones (such as, heat shock or chemical shock). Because
can be activated by pro-proliferative JNK is implicated in apoptosis and inflammation, it is particularly of interest in
signals (mitogens). Though MAP the study of chronic conditions such as atherosclerosis, which is likened to a
kinase activation is associated with slow-acting inflammation (similar to how metal rusting is a slow combustion
cell growth and division, in several process).
cases mechanically stimulated cells
may exhibit rapid MAP kinase One interesting aspect of MAP kinases is that they are capable of directly activat-
activation but no downstream ing transcription factors, the DNA-binding proteins that control gene transcrip-
alteration in proliferation. The factors tion. ERK1/2 is known to activate the transcription factor Elk1. JNK has been
determining whether mechanically shown to activate several transcription factors, including c-Jun, Elk1, SMAD4,
induced MAP kinase activation leads
to tissue growth are unclear. This is
ATF2, and NFAT1. By possessing the capacity to be quickly activated and to
an important issue to clarify because, directly regulate transcription factors, MAP kinases are ideally suited to partici-
in some cases, such growth is pate in pathways involved in early gene expression. Indeed, several components
undesirable. When the smooth of the MAP kinase pathway have been implicated in the expression of primary
muscle cells lining the vasculature response genes, in other words genes whose expression is altered soon after stim-
are exposed to altered mechanical ulation and that do not require de novo protein synthesis.
stresses, they may proliferate in
response, leading to an undesirable
thickening of the vascular wall. Mechanically stimulated cells exhibit prostaglandin release
Another well-characterized response of cells to mechanical stimulation is prosta-
glandin release. Prostaglandins are lipid compounds that are enzymatically
derived from fatty acids and mediate several cellular functions. There are several
types of prostaglandin, of which prostaglandin E2 (PGE2) is arguably the most
well studied in cellular mechanotransduction. PGE2 is synthesized from arachi-
donic acid, which is derived from membrane phospholipids. This arachidonic
acid is converted to prostaglandin G2 and subsequently prostaglandin H2 by the
enzyme cyclooxygenase (COX). A final isomerization step converts prostaglandin
H2 into the biologically active PGE2. COX exists in constitutive (COX-1) and induc-
ible (COX-2) isoforms and is considered the rate-limiting enzyme in the PGE syn-
thesis process. Several studies have demonstrated that mechanical stimulation
activates COX2 gene expression, elevates COX-2 protein levels, and induces PGE2
release into the extracellular environment. Upon its release, PGE2 may initiate
signaling cascades in an autocrine fashion or in other cells by binding PGE2
receptors on the cell surface, such as the receptor for PGE2 (EP2).

Mechanical forces can induce morphological changes in cells

The cascade of early biochemical responses discussed above ultimately results in
functional alterations downstream. One that can be readily observed is altered
 ALTERATION OF CELLULAR FUNCTION 333

cell morphology. It has been widely demonstrated that when cells are subjected to
substrate stretch or fluid shear (generally over an extended period, on the order of
hours or sometimes days), cells may actively remodel their cytoskeletons and
change their gross shape. As described in Section 11.1, endothelial cells subjected
to fluid flow have been observed to align in the direction of flow. These cells have
also been demonstrated to align perpendicular to the direction of uniaxial stretch.
Other cells, such as smooth muscle cells and fibroblasts may exhibit alignment
parallel or perpendicular to major stretch directions, depending on the type of cell
used, the plating conditions, and the exact nature of the stretch (such as percent-
age strain and applied frequency).
In some cases, cells may exhibit no gross changes in morphology but may undergo
extensive cytoskeletal remodeling. Mechanically induced formation of actin stress
fibers has been observed in many cells. The stress fibers can exhibit a preferential
alignment relative to the primary direction of stretch or flow. It has been proposed
that these alignments may serve to minimize intracellular stresses (Figure 11.19).

Mechanical stimulation can induce extracellular

matrix remodeling
In Section 1.1, we learned about several examples in which local matrix remode-
ling occurs in response to mechanical stimuli. When bone cells do not experience
proper mechanical stimulation, bone formation ceases and bone resorption is
initiated. Another example is onset of osteoarthritis after changes in mechanical
signals experienced by chondrocytes. Given these examples, it is unsurprising
that mechanoregulation of matrix generation and matrix degradation pathways
are of particular importance. A prominent degradation mechanism is through
regulation of matrix metalloproteinases, proteins that cleave and degrade matrix
molecules. Generally, mechanical stresses are thought to provoke a “protective”
response (down-regulation of matrix metalloproteinases) in cells where the
immediate extracellular environment is structurally reinforced. This serves to
reduce mechanical loads on the cells within. However, this protective response
may not necessarily be physiologically beneficial. For a heart subjected to high

20x magnification Figure 11.19 Fluorescent staining of

(A) the actin cytoskeleton in flowed
bone cells. Cells were either kept in
static control conditions (A) or steady
flow (B) to show that flow exposure
induces stress fiber formation.
Notice that this stress fiber formation is
control (no flow) induced in the absence of any preferred
fiber orientation. (From, Malone AM
et al. (2007) Am. J. Physiol. Cell. Physiol.)

(B)

steady flow (1.2 Pa)

334 CHAPTER 11: Cellular Mechanotransduction

blood pressure, fibrosis may be induced. In this case, the heart can become much
stiffer, resulting in increased workload for the heart muscles.

Cell viability and apoptosis are altered by difference processes

Cell viability is another functional outcome that can be altered by mechanical
stimuli. Cell death can occur by two distinct processes, necrosis or apoptosis.
Necrosis is cell death due to cell damage, and is a process that can be considered
unexpected or accidental. In contrast, apoptosis is cell death due to programmed
events. Unlike necrosis, apoptosis is a normal preprogrammed event, and does
not induce an inflammatory response. It has been demonstrated in a variety of
cell types that mechanical stimulation can alter cell apoptosis. It has been shown
that endothelial cells may exhibit decreased apoptosis with fluid shear, although
this may depend on the flow profile. Shear resulting from disturbed flows or oscil-
latory shear stresses has been found to be less effective than steady shear or pul-
satile shear in reducing cell apoptosis.

Key Concepts

• The sensing of mechanical signals by cells, or cellular adhesions, the nucleus, and primary cilia have all
mechanotransduction, is critical to many aspects of been implicated as potential sites of
physiology and understanding disease. Four distinct mechanosensing.
phases are involved: conversion of tissue or organ- • Mechanosensitive ion channels are a major class of
level loads into cell-level physical signals; force mechanosensitive molecules. They are known to
detection by mechanosensitive molecules undergoing respond to membrane tension as well as to thinning
a conformational change; activation of due to hydrophobic mismatch.
intracellular signaling systems; and altered cell
• Other proteins can change their enzymatic potential
metabolism.
by buckling, unfolding, and exposure of cryptic
• Many cellular-level physical signals have been shown binding sites. FRET is a powerful way to detect such
to be potent regulators of cell metabolism, including changes fluorescently.
fluid flow, stretch, and pressure.
• Second messenger signaling including IP3, cAMP, and
• Specialized excitable cells mediate our senses of Ca2+ signaling have been shown to be activated by
touch and hearing. They are some of the best- mechanical signals. This leads to activation of protein
understood mechanosensing cells and often have signaling cascades such as MAPK signaling, and
highly sensitive cellular structures capable of intercellular signaling such as PGE2. Ultimately, these
exquisite sensitivity. cascades lead to altered gene expression,
• In nonexcitable cells, structures such as the modification of the extracellular matrix, and changes
glycocalyx, cell membrane, cytoskeleton, focal in cell viability.

Problems

1. Estimate the force required to remove a membrane- Would that be sufficient to overcome the activation
bound protein from the bilayer if it has a cylindrical energy of a channel?
hydrophobic region that is 2 nm in diameter and 5 nm in
length. You can do this by assuming that the mechanical 3. Figure 11.11 depicts a primary cilium undergoing
work done by the pulling force is equal to the change in bending. From the figure, estimate the magnitude of
free energy caused by exposing the hydrophobic domain deflection and strain in the membrane surrounding
to water. You will also need to assume that 5 nm of the cilium at its base. You may assume that the
displacement is sufficient to remove the protein. cilium diameter is 200 nm. Finally, use the result
from Problem 8.4 to estimate what force applied
2. Consider the bending hair bundle from Example 11.1. to the tip would produce this magnitude
How much energy was put into deflecting the tip? of bending.
 Annotated References 335

4. One of the roles of the cytoskeleton is to connect gradient? How does this compare with the channel
the nucleus to the rest of the cell. This force activation energy? Assume a typical resting potential
transmission is thought to be involved in cellular of −70 mV.
mechanotransduction. If a cell is exposed to an
externally imposed deformation, describe how you 6. In Example 11.3, the change in free energy for a
would expect the nucleus to deform in response, if (a) channel moving from a closed to an open configuration
the cytoskeleton is removed, (b) the cell cytoplasm acts was estimated to be 14kBT. Using the Boltzmann
like a continuum solid without a cytoskeleton per se, equation, predict the fraction of the time that the
or (c) the cytoskeleton acts like a tensegrity structure, channel would be open owing only to thermal
with normal cytoplasm. Specifically, how would the fluctuations. How would this change if you account for
magnitude of the nuclear deformation compare with the work done by membrane tension as the channel
that of the cell overall? opens? Assume that the cell is at its rupture or lytic
strain (3%) and a typical areal expansion modulus, KA ,
5. Ion channels are sometimes gated by the passage of 0.5 N/m.
of ions themselves. Specifically, they can lose
conductance as ions flow through them. One putative 7. For a bilayer with an areal expansion modulus
mechanism is that the ions moving through the KA = 1.0 N/m, what would the stiffness in the transverse
channel supply the energy to close the channel. direction be? In other words, for a force applied through
How much energy is associated with a monovalent the thickness, what is the slope of the force–deflection
ion moving through the channel along its electrical curve?

Annotated References

Anderson RG (1998) The caveolae membrane system. Annu. Rev. tion examined from a genetics perspective, particularly in sensory
Biochem. 67, 199–225. Gives a comprehensive review of the cell cells.
biology of caveolae.
Eyckmans J, Boudou T, Yu X & Chen CS (2011) A hitchiker’s guide to
Birukov KG, Birukova AA, Dudek SM et al. (2002) Shear stress- mechanobiology. Dev. Cell 21, 35–47. A summary of mechanotrans-
mediated cytoskeletal remodeling and cortactin translocation duction mechanisms, primarily focused on non-sensory cells.
in pulmonary endothelial cells. Am. J. Respir. Cell Mol. Biol. 26,
453–464. This journal research article describes the response of Gefen A (2011) Cellular and Biomolecular Mechanics and Mechano-
pulmonary endothelial cells to applied shear stress. The study biology. An edited text describing recent advancements in cell and
also describes the effects of certain GTPases and early response molecular mechanics and mechanobiology.
mechanoresponsive pathways to shear. Hamill OP & Martinac B (2001) Molecular basis of mechanotransduc-
Brown TD, Bottlang M, Pedersen DR & Banes AJ (1998) Loading tion in living cells. Physiol. Rev. 81, 685–740. A review of molecular
paradigms—intentional and unintentional—for cell culture mecha- mechanotransdcution mechanisms with a focus on the membrane
nostimulus. Am. J. Med. Sci. 316, 162–168. A numerical analysis of and channels; the primary source for the material on hydrophobic
parasitic flows in bending-based systems for studying the strain mismatch and channel-membrane coupling.
response in vitro. Jacobs CR, Temiyasathit S & Castillo AB (2010) Osteocyte mecha-
Chalfie M (2009) Neurosensory mechanotransduction. Nat. Rev. nobiology and pericellular mechanics. Annu. Rev. Biomed. Eng. 12,
Mol. Cell. Biol. 10, 44–52. Comprehensive review of molecular and 369–400. Provides a comprehensive review of mechanosensory
cellular mechanisms of mechanosensing in sensory cells. structures and mechanisms identified in bone cells.

Chancellor TJ, Lee J, Thodeti CK & Lele T (2010) Actomyosin ten- Knothe Tate ML, Steck R, Forwood MR & Niederer P (2000) In vivo
sion exerted on the nucleus through Nesprin-1 connections influ- demonstration of load-induced fluid flow in the rat tibia and its
ences endothelial cell adhesion, migration, and cylic strain-induced potential implications for processes associated with functional
reorientation. Biophys. J. 99, 115–123. Provides recent evidence of adaptation. J. Exp. Biol. 203, 737–745. Describes quantification of
the role of nucleus loading and nuclear matrix proteins in mecha- loading-induced flow in bone.
notransduction.
Kooppel DE, Axelrod D, Schlessinger J et al. (1976) Dynamics of
Corey DP & Hudspeth AJ (1979) Response latency of vertebrate hair fluorescence marker concentration as a probe of mobility. Biophys.
cells. Biophys. J. 26, 499–506. Early data on the molecular mecha- J. 16, 1315–1329. An early description of fluorescence recovery
nism of hair cell mechanosensing, particularly the incredibly fast after photobleaching used to measure lateral diffusion in the cell
response time. membrane.

DeBakey ME, Lawrie GM & Glaeser DH (1985) Patterns of athero- Kung C (2005) A possible unifying principle for mechanosensation.
sclerosis and their surgical significance. Ann. Surg. 201, 115–131. Nature 436, 647–654. Provides a concise summary of the roles of
This journal article presents an analysis of common atherosclerotic mechanosensitive ion channels in touch sensation and hearing.
lesion development sites as noted in clinical examinations. The arti-
Malone AM, Batra NN, Shivaram G et al. (2007) The role of actin
cle further discusses the clinical aspects of atherosclerosis, includ-
cytoskeleton in oscillatory fluid flow-induced signaling in MC3T3-
ing classification, progression and recurrence of disease.
E1 osteoblast. Am. J. Physiol. Cell Physiol. 292, C1830–C1836. Dem-
Ernstrom GG & Chalfie M (2002) Genetics of sensory mecha- onstrated differential cytoskeletal remodeling in cells exposed to
notransduction. Annu. Rev. Genet. 36, 411–53. Mechanotransduc- static and dynamic fluid flow.
336 CHAPTER 11: Cellular Mechanotransduction

Nauli SM, Alenghat FJ, Luo Y et al. (2003) Polycystins 1 and 2 mediate evidence that fluid flow is a critical cell-level physical signal in bone
mechanosensation in the primary cilium of kidney cells. Nat. Genet. mechanobiology.
33, 129–137. A discussion of the polycystins and their putative role
in primary-cilium-based mechanosensing. Reilly GC, Haut TR, Yellowley CE et al. (2003) Fluid flow induced PGE2
release by bone cells is reduced by glycocalyx degradation whereas
Nishiyama M, Shimoda Y, Hasumi M et al. (2010) Microtubule depo- calcium signals are not. Biorheology 40, 591–603. Supplies some of
lymerization at high pressure. Ann. N. Y. Acad. Sci. 1189, 86–90. A the only evidence that the cellular glycocalix is critical for mecha-
demonstration that high hydrostatic pressure can induce microtu- nosensing in bone cells.
bule depolymerization in vitro.
Simons K & van Meer G (1988) Lipid sorting in epithelial cells. Bio-
Nonaka S, Tanaka Y, Okada Y et al. (1998) Randomization of left-right chemistry 27, 6197–6202. An early description on microdomains in
asymmetry due to loss of nodal cilia generating leftward flow of lipid membranes.
extraembryonic fluid in mice lacking KIF3B motor protein. Cell 95,
p. 829–837. Seminal study demonstrating that flow generated by Tabouillot T, Muddana HS & Butler PJ (2011) Endothelial cell mem-
nodal cilia is critical for left–right determination. brane sensitivity to shear stress is lipid domain dependent. Cell.
Mol. Bioeng. 4, 169–181. Direct evidence that microdomains in the
Olsen B (2005) Nearly all cells in vertebrates and many cells in inver- bilayer, including rafts and caveolae, are modified with fluid shear
tebrates contain primary cilia. Matrix Biol. 24, 449–450. An editorial stress and are potentially involved in mechanosensing.
on the ubiquitous nature of primary cilia and their potential physi-
ological function. Vogel V & Sheetz M (2006) Local force and geometry sensing regu-
late cell functions. Nat. Rev. Mol. Cell. Biol. 7, 265–275. Provides a
Owan I, Burr DB, Turner CH et  al. (1997) Mechanotransduction in concise review of potential mechanisms by which force-induced
bone: osteoblasts are more responsive to fluid forces than mechan- conformational changes may lead to exposure of cryptic binding
ical strain. Am. J. Physiol. 273 (3 Pt 1), C810–815. Provides evidence sites.
that osteoblasts subjected to four-point bending were responsive to
fluid flow rather than substrate strain. Wang Y, McNamara LM, Schaffler MB & Weinbaum S (2007) A model
for the role of integrins in flow induced mechanotransduction in
Qin YX, Lin W & Rubin C (2002) The pathway of bone fluid flow as osteocytes. Proc. Natl Acad. Sci. USA 104, 15941–15946. Describes
defined by in vivo intramedullary pressure and streaming potential a proposed mechanism for mechanosensing of flow by integrins.
measurements. Ann. Biomed. Eng. 30, 693–702. Some of the earliest
Abbreviations

Chapter 1 GTP guanosine triphosphate

IRDS infant respiratory distress syndrome MF microfilament
RBC red blood cell MT microtubule
WLC wormlike chain
Chapter 2
AFM atomic force microscopy Chapter 9
cAMP cyclic adenosine monophosphate AM acetoxymethyl
cGMP cyclic guanosine monophosphate CMC critical micelle concentration
DAG diacylglycerol FRAP fluorescence recovery after photobleaching
IP3 inositol triphosphate
NO nitric oxide Chapter 10
CO carbon monoxide RGD arginine–glycine–aspartate
ER endoplasmic reticulum VCAM vascular cellular adhesion molecule
GDP guanosine diphosphate ICAM intercellular adhesion molecule
GEF guanine nucleotide exchange factor NCAM nerve cellular adhesion molecule
GPCR G-protein-coupled receptors Arp actin-related protein
GTP guanosine triphosphate
DAPI 4′,6-diamidino-2-phenylindole Chapter 11
GFP Green fluorescent protein VEGFR-2 vascular endothelial growth factor receptor 2
STM scanning tunneling microscopy PIP2 phosphatidylinositol 4,5-bisphosphate
SDS sodium dodecyl phosphate PKD polycystic kidney disease
PAGE polyacrylamide gel electrophoresis ATP adenosine triphosphate
PCR polymerase chain reaction GTP guanosine triphosphate
siRNA small inhibitory RNA GDP guanosine diphosphate
SSR site-specific recombinase MscL mechanosensitive channel of large conductance
Cre cyclic recombinase ENaC epithelial sodium channel
TRP transient receptor potential
Chapter 4 FRET Förster resonance energy transfer
SI Système International d’Unités cAMP cyclic adenosine monophosphate
IP3 inositol triphosphate
Chapter 6 NOS nitric oxide synthase
AFM atomic force microscopy eNOS endothelial isoform of nitric oxide synthase
TFM traction force microscopy MAP mitogen-activated protein
PDMS polydimethylsiloxane ERK1/2 extracellular-signal-regulated kinase 1 and 2
JNK c-jun N-terminal kinase
Chapter 7 PGE2 prostaglandin E2
ADP adenosine diphosphate COX cyclooxygenase
ATP adenosine triphosphate EP2 prostaglandin receptor for PGE2
FJC freely jointed chain CFP cyan fluorescent protein
GDP guanosine diphosphate YFP yellow fluorescent protein
List of variables and units

Chapter 1 ε strain
γ shear strain ε strain vector
σ tensile stress εa axial strain
τ shear stress εt transverse strain
A deformed surface area εxy, εyx, εxz components of strain
A o undeformed surface area θ angle of twist
d membrane thickness θxx rotation angles
ΔP Patm – Ppip κ local curvature
FP resultant force due to pressure λ stretch ratio
Ft resultant force due to tension ν Poisson ratio
k spring constant ϕ eigenvalue
kB Boltzmann’s constant σ stress
Lpro protrusion length σ stress vector
n surface tension σyx, σzx, σzy components of stress
Patm environmental pressure τ shear stress
Pcell cellular pressure a acceleration
Pi internal pressure a deformed vector
Po external (outside) pressure A area
Ppip pipette pressure A undeformed vector
R radius of pressure vessel C FTF, right Cauchy-Green deformation tensor
Ra cell radius at one end D compliance matrix
Rb cell radius at other end E Young’s modulus
R o undeformed radius e strain vector
Rpip pipette radius E Green–Lagrange strain
Rpro protrusion radius F force
V volume F deformation gradient
G shear modulus
Chapter 3 h beam width
α angular acceleration I second moment of inertia
γ shear strain I identity
δ transverse displacement J polar moment of inertia
ΔL column extension k spring constant
ΔR change in radius of curvature kB Boltzmann’s constant
ΔS change in beam length L length
 LIST OF VARIABLES AND UNITS 339

M moment L length
m mass m mass
n surface tension n surface tension
P pressure P pressure
P′ transformed vector Re Reynolds number
P original vector t time
Q rotation matrix u fluid velocity
R radius x, y, z spatial coordinates
RU or VR rotation and stretch components of polar
decomposition Chapter 5
Sx resultant force
γ shear strain
SXx components of Sx σ stress vector
SXy components of Sx τ shear stress
SXz components of Sx A cross-sectional area
u, v, w displacements β 1/kBT
v eigenvector b distance traveled in each step, Kuhn length
v1, v2, and v3 principal directions D diffusion coefficient
w beam displacement  energetic cost per hairpin
x spring deformation E Young’s modulus
x deformed vector k spring constant
X undeformed vector kB Boltzmann’s constant
x, y, z spatial coordinates L contour length
m microstate
Chapter 4 n surface tension
α constant scaling exponent n, n+ number of flips, number of flips that come
β constant viscous factor out heads
δ phase lag N number of particles
γ shear strain Nh number of hairpin sites
γ* complex shear strain p(m) probability of microstate m
μ viscous coefficient q heat
μeff effective viscosity Qs(ms) energy of microstate ms
ω frequency R end-to-end length
ρ density S entropy
σ stress vector t time
τ* complex shear stress T temperature
τ shear stress V volume
ξ structural damping coefficient Y helmholtz free energy
E elastic modulus or storage modulus z single partition function
E* complex modulus Z partition function
g acceleration due to gravity
G* complex shear modulus Chapter 6
h height α angle
–
i imaginary unit, √–1 γ shear strain
k spring constant δ phase lag
kB Boltzmann’s constant ε strain
340 LIST OF VARIABLES AND UNITS

εxy, εyx, εxz components of strain kon, koff rate of reaction

η viscous friction coefficient K dissociation constant
λ stretch ratio ℓp persistence length
μ dynamic viscosity ℒ Langevin function
ν Poisson’s ratio L length
ρ density n surface tension
τ time constant, η2/k P probability or pressure
τ shear stress Ploop probability of looping
υ kinematic viscosity Qloop energy of looping
σ stress vector R end-to-end distance or radius
ω rotational speed ri segment vector
c speed of light R end-to-end vector
C FTF, right Cauchy-Green deformation tensor s arc length
F deformation gradient S entropy
G(r) Green’s function t time
h height v elongation/shrinking
I intensity of the trapping light x, y, z direction
I identity tensor z single partition function
k spring constant Z partition function
kB Boltzmann’s constant
L length Chapter 8
l wavelength of light γ shear strain
n surface tension δ transverse displacement
nb nm indices of refraction λ stretch ratio
R radius ρn number of polymers per unit volume
Re Reynolds number ρvol volume fraction
t time σ stress vector
V velocity τ shear stress
w displacement Ψ free energy
x deformed vector wa areal strain energy density
X undeformed vector A area
d depth
Chapter 7 E Young’s modulus
γ shear strain F force
θ angle G shear modulus
σ stress vector I second moment of inertia
τ shear stress kB Boltzmann’s constant
Ψ free energy ksp spring constant
Ω(R) density of microstates K s membrane shear modulus
b Kuhn length ℓp persistence length
cosΔθ(s) orientation correlation function L length
E Young’s modulus P pressure
I mass or first moment of inertia r vector
k spring constant R vector
k B Boltzmann’s constant t time
 LIST OF VARIABLES AND UNITS 341

V volume S entropy
w beam displacement t thickness
x, y, z spatial coordinates T temperature
T time
Chapter 9 utot, vtot total deformation
α scaling factor w transverse displacement
γ shear strain W energy
γint interfacial energy W density of states
ε strain x, y, z spatial coordinates
θ angle
λ length Chapter 10
ρ number of molecules per unit volume γ shear strain
σ stress vector Δt time interval
τ shear stress Δ pitch of polymerized actin helical structure
𝜑(R) excitation intensity θ angle that a membrane makes with a
surface
Ψ free energy
λb stretch ratio of substrate
 Planck’s constant
μ fluid viscosity
A area
n membrane tension
c concentration
ρ density
D diffusion constant
σ stress vector
E Young’s modulus
τ shear stress
EI flexural rigidity
a constant
G shear modulus
A area
J flux
ds binding-site spacing
k spring constant
D diffusion constant
kB Boltzmann’s constant
E Young’s modulus or energy
KA areal expansion modulus
F force
KB bending stiffness
F− negative drag force
KS shear stiffness
F+ positive elastic force
l hydrocarbon chain length
h height
lc average bond length between carbon
atoms J adhesion energy density
ℓp persistence length k spring constant
L length/dimension k− unbinding rate constant
m mass of molecule k+ binding rate constant
M, m moments kB Boltzmann’s constant
n force kd equilibrium constant
n surface tension kon rate of reaction
nc number of carbon atoms in chain KB bending modulus
nc number of molecules n surface tension
N number nb area density of bonds
P pressure p probability
P0 laser power R radius or end-to-end length
q heat S speed
R radius t time
342 LIST OF VARIABLES AND UNITS

toff average time per cycle that myosin is not σ stress vector
bound to actin τ shear stress
ton time per cycle that myosin is not bound to ΔA change in area
actin
Δx effective energy barrier width
tr release time
A area
T temperature or tortuosity
d displacement
v or V velocity
EB bulk modulus
w displacement
F force
wbs width of binding site ‘sweet spot’
Fb buckling load
W density of states
k spring constant
W work
kB Boltzmann’s constant
Wadh adhesion energy
KA areal expansion modulus
Wdef strain energy associated with deformation
n surface tension
x maximum length of spring
P pressure
x, y, z spatial coordinates
Q flow rate
xsh shortening distance
R radius
T temperature
Chapter 11 W density of states
γ shear strain W work
ε strain x, y, z spatial coordinates
Index

absolute pressure 91 definition 20 nuclei 30

accommodation 30 protein definition 20–22 polymerase chain reaction 45
acetylcholine 317 amoeba chemotaxis 296 ball and stick models 249, 250
acrylamide 41, 42, 160 amphiphiles 253–254, 255 barbed polymer ends 190, 196,
actin 129–130, 189 analytical gels 42 197–198
contraction 301–308 angular distribution functions 235–236 barrier functions, membranes 255–259
filopodia 236–237, 238–239 anionic regions 250 base parameters 110, 111
kinetics 194, 195, 197 anisotropic affine polymer bases 22–24
membranes 268 networks 233 bead-based studies 152–155, 157
migration 293–295 ankyrin 268 cross-correlation 160–161, 162
morphological changes 333 annealing temperatures 44–45 Delaunay triangulation 178, 179
myosin interactions 303–308 antibodies 39–40, 42–43 beams
persistence length 198, 203 anti-cancer drugs 191 beam equations 66–68
pulmonary endothelial cells 313 anvil bone 319 buckling loads 67–68
stress fibers 301–304 apoptosis 334 definition 84
action at a distance 228 areal expansion moduli 13–15, equilibria 64
action potentials 305 267, 274 filopodia 236–237
adenine (A) 22–24 areal strain kinematics 62–63
adenosine diphosphate (ADP) biaxial membrane deformation 267 persistence length 200–202
biopolymer structure 190 thin structure analyses 240–241 Bell models 290–291
contraction 303–304 arrays, traction force 165–166 Bell’s equation 329
polymerization kinetics 194, 197 arteries 4 bending
secondary messengers 33 aspiration see micropipette aspiration cytoskeleton proteins 66
adenosine triphosphate (ATP) atherosclerotic formations 313, 314 membranes 267–274
biopolymer structure 190 atmospheric pressure 91 polymer networks 214–216, 225–227
contraction 303–304 atomic force microscopy (AFM) second moment of inertia 65
mechanotransduction 317 40–41, 157–159, 282 Bernoulli’s equation 97
polymerization kinetics 194, ATP see adenosine triphosphate biaxial deformation 172–173, 266, 267
197–198 attachment, migration 293 bifurcations 313, 314
rigor mortis 304 autocrine signals 31 binding sites, mechanotransduction
secondary messengers 33 autofluorescence 38 328–329
adenylyl cyclase 331 average squared position, Bingham plastic fluids 98
adhesion 279–291, 292 definition 147 biochemical signal communication
key concepts 308 axial deformation 226–227, 267 30–31
mechanotransduction 321–322 axial loading 68 bioinformatics 49
ADP see adenosine diphosphate axial strain 227–228 bladder 317
Aequorea victoria 40 axial tension 266 blood cells
affine polymer networks 229–236 axoneme cores 320 adhesion 287, 288, 289
AFM see atomic force microscopy cytoskeleton 239–240, 241
agarose gel 42 bacteria micropipette aspiration 9, 10
alamethicin channels 327 dimensional analyses 110, 111, 114 pathogens 7
ambulation 300, 301 drag forces 110, 111 spectrin networks 268
amino acids lac repressors 215–216 blood-flow-mediated shear stress
codon sequences 25, 26 mechanical structure 7 312–313, 314
combinatorics 25 migration 295–296 blood vessels 287, 288
344 Index

blotting 42–43 computer simulations 9 cytometry 153

Boltzmann, Ludwig 124, 125, 138 concentration cytoplasm 192, 193, 251–252
Boltzmann’s distribution 136, 138, critical 195–196 cytosine (C) 22–24
139, 142–143, 216 diffusion equations 256 cytoskeleton
Boltzmann’s law 138 Fick’s second law 257–259 deformable body networks 58, 62, 66
bone cells 4, 6–7, 311–315, 324, 333 concentric contractions 300 fluorescence imaging 39
Boussinesq solution 163, 164 condensation reactions 21 mechanotransduction 322, 323,
bowtie membrane formations 172 condensers 37–38 325, 333
brightfield illumination 37–38 cone-and-plate flow 170–171 pathogens 7
Brownian motion 119–120, 145, confined uniaxial stretching 172 polymer networks 223–247
294–295 constant energy microstates 132–133 affine 229–236
Buckingham Pi Theorem 111 constant strain triangles 180 biomechanical function 236–245
buckling constitutive cyclooxygenase (COX-1) key concepts 246
beam equations 67–68 enzymes 332 scaling relationships 225–229
filopodia 236–237, 238–239 constitutive equations 62, 65, 73–74 pulmonary endothelial cells 313
mechanotransduction 322 constitutive models 262, 269 red blood cells 239–240, 241
constitutive relations 102–103
calcium continua 223–225, 229–230, 284 D-actin 197
channels 5–6, 33 equations 68–69 DAG see diacylglycerol
mechanotransduction 330–332 equilibria 123 dashpots 173–177
reporter dyes 32, 33, 40 vector notation 74–75 dedifferentiation 28
cAMP see cyclic adenosine continuity equations 102–103 deformable body networks 55–78
monophosphate continuous Boltzmann deformations
canaliculi 313, 314, 315 distributions 139 correlation-based approaches
cancer cells 8 continuous diffusion 258 178–179
canonical ensembles 127, 136–143, contour lengths 129, 209–210 gradient analyses 180–181
211–213, 216–217 contraction 279, 298–308 kinematics 260–262
cantilevers 66–67, 157–158 contrast illumination 37 laboratory cell mechanics 173–181
carbon monoxide (CO) 32 convective acceleration 101 mechanics 78–83
α-carbons 20 coordinate rotations 75–76 mechanotransduction 316
cardiac function 299–300 correlation-based approaches 178–179 polymer networks 225–228, 229–230
cardiac myocytes 316–317 cortical actin 268 shear 113
cardiovascular system 4, 316–317 Couette flow 171–172 degrees of freedom 120, 224
cartilage 4 COX see cyclooxygenase enzyme dehydration synthesis 21
cationic regions 250 creep 174 Delaunay triangulation 178
Cauchy–Green deformation tensors 80 Crick, Francis 25–27 denaturing 41–42, 44–45
cellular solids theory 225 critical concentrations 195–196 density of states 124–126, 127, 133
central dogma of molecular biology critical micelle concentration (CMC) deoxyribonucleic acid (DNA)
25–27, 29 252–255 correspondence 25–27
CFP see cyan fluorescent protein cross-bridge actomyosin models definition 20, 22–24
cGMP see cyclic guanosine 303–308 directional properties 24
monophosphate cross-correlation 160–162 freely joined polymer chains
channels 5–6, 33, 324, 326–327, 328 cross-linked filopodia 236–237, 217–218
characteristic equations 76 238–239 gel electrophoresis 41–43
characteristic velocity 94–95 cross-wise flow profiles 169–170 lac repressors 215–216
chelators 331 cryptic binding sites 328–329 persistence length 198, 203
chemoattractants 295, 296 curved coordinate systems 260 polymerase chain reaction 43–45
chemotaxis 296 cyan fluorescent protein (CFP) 40, 330 transcriptional regulation 28–29
chromosomes 292 cyclic adenosine monophosphate worm-like polymer chains 217–218
cilia 5, 315, 320, 321 (cAMP) 32, 36, 296, 331–332 desmosomes 192–193
c-jun N-terminal kinase (JNK) 332 cyclic guanosine monophosphate detachment forces 281–282
CMC see critical micelle concentration (cGMP) 32, 36 detailed balance principle 196
CO see carbon monoxide Cyclic Recombinase (Cre) 48 diacylglycerol (DAG) 32, 36
cochlea 319 cyclooxygenase (COX) enzymes 332 dichroic mirrors 38
codon sequences 25, 26 cylinders, torsion 60, 61 differential interference microscopy
coiled-coil structures 192 cylindrically symmetric deformations (DIC) 38
combinatorial enumeration 133 172–173 differentiation 28
complex moduli 106–108 cytokines 31 diffusion 145–147, 256–257, 258
 Index 345

dimensional analyses 110–115 mechanotransduction 312–314, Euler’s Formula 107

dimensionless parameters 111–113 321–322, 324 Hill 300–301
dimers, tubulin 191 migration 297 membranes 256–257
directional bias, migration 298 endothelial isoform of NOS (eNOS) of motion 54–55
discrete Boltzmann distributions 139 331 Navier’s 102–103
discrete diffusion 256–257 end-to-end length 129–131, 204–206 Navier–Stokes 98–103, 115
disease frameworks 4–8 energy protein unfolding 329
dispase, cell–cell dissociation 282 adhesion 284, 286–287 random walks 145–147
displacement 71–73 affine polymer networks 233–236 solid mechanics 54–55, 62, 65–69,
adhesion 290 anisotropic affine polymer networks 73–75
atomic force microscopy 159 233 thermodynamics 134
deformable body networks 62–63 barrier widths 329 vector notation 74–75
digital representation 162 critical micelle concentration Young’s 284
freely joined polymer chains 253–254 equilibria
211–213 elastic deformations 122–123 adhesion 289–290
intracellular strains 179–180 elastic moduli 233–234 deformable body networks 62, 64,
inverse problems 163–165 microcanonical ensembles 132–133, 69–71
worm-like polymer chains 216–217 135–136 free energy 128–129
dissociation constant 195, 282, statistical mechanics primers hairpin number 134–135
289–291 129–131 membranes 262–266, 269–272
distortion, images 162 thin structure analyses 240–241 microcanonical ensembles 133–136
double pipettes, adhesion 281, 282 worm-like polymer chains 214–216 partition functions 139–141
drag 305, 306 eNOS see endothelial isoform of NOS statistical mechanics primers 123,
drugs 191 ensembles 128–129, 133–136, 139–141
dsRED see red fluorescent protein canonical 127, 136–143, 211–213, strain energy minimization 123
dynamic instability 191 216–217 Escherichia coli 215–216
dynamic viscosity 93, 169 equilibria 135–136 etching 165
dystrophy 281 grand canonical 127 Euler buckling loads 68
microcanonical ensembles 127, Euler’s Formula 107
ear 5–6, 319 131–136 excluded volume interactions 206
eccentric contractions 300 statistical mechanics primers 127, exomes 49
eigenvalues 76 131–136 exons 25–26, 49
eigenvectors 76 entrance lengths 169–170 expansion moduli 13–15
Einstein, Albert 145, 260 entropy experimental biology 36–45, 46–50
elastic beams 200–202, 236–237 combinatorial enumeration 133 extension processes 44
elastic behavior 269, 301–302 critical micelle concentration extension ratios 80
elastic components 104–106 253–254 extensive quantities 140
elastic deformations 122–123 hairpin polymers 133, 134 extracellular matrix remodeling
elastic flexible polymers 230–232 statistical mechanics primers 333–334
elastic moduli 107–108, 159, 225–228, 124–128, 129–131 ex vivo, definition 20
233–236 enzymes
elastic organs 317 mechanotransduction 330, 331, 332 F-actin 189–190, 195
elastic solids 16 membranes 252 factorials 133
electromagnets 151–159 transcriptional regulation 28–29 fascin protein 236
electron micrographs epifluorescence 38, 39, 40, 161 fatty acids 24, 251
embryonic primary cilia 315 epithelial cells 312 femur 315
filopodia 236, 237 epithelial sodium channel (ENaC) 327 fibroblasts 297
mechanotransduction 325 equations Fick’s first law 256
micropillar arrays 165–166 adhesion 284 Fick’s second law 145–146, 257–259
neutrophils 15 beams 66–68 filaments 189–190, 229–230
electrophoresis 41–43 Bell’s 329 filopodia 236–237, 238–239
embryonic development 315–316 Bernoulli’s 97 finite deformations see large
ENaC see epithelial sodium constitutive 62, 65, 73–74 deformations
channel continuity 102–103 firm adhesion 287, 288
endocrine signals 31 continuum 68–69 first law of thermodynamics 124
endoplasmic reticulum 29, 33 contraction 300–301 FJC see freely joined chains
endothelial cells 6 diffusion 145–147, 256–257 flexible polymers 202–204, 230–232
adhesion 287 dimensionless parameters 111–113 flexible substrates 172
346 Index

flexural rigidity 65, 200–202 fourfold connectivity 245 green fluorescent protein (GFP) 40
flippases 252 fractures 19–20 Green’s function 164
flow chambers 167–168 fragility fractures 6 Green–Lagrange strain 80
fluid flow Frank–Starling law 299–300 GTP see guanosine triphosphate
diverse device designs 171–172 FRAP see fluorescence recovery after guanine (G) 22–24
mechanotransduction 311–313, photobleaching guanine nucleotide exchange factor
315–316 free-body diagrams 12, 263–264, 270 (GEF) 36
parallel plate 104 free energy guanosine diphosphate (GDP) 36, 191,
fluid mechanics 89–117 critical micelle concentration 194, 197
dimensional analyses 110–115 253–254 guanosine triphosphate (GTP) 36, 191,
fluid statics 89–92 ideal chain polymers 207–209 192, 194, 197, 325
key concepts 115 microcanonical ensembles 135–136
Navier–Stokes equations 98–103 partition functions 139–141 hair cells 5–6, 319
Newtonian fluids 92–98 statistical mechanics primers hairpin polymers 132–134, 142–143
rheological analyses 103–110 128–131, 135–136, 139–141 half-space 158–159, 164
fluid mosaic model 252 freely joined chains (FJC) 189, hammer bone 319
fluids 210–214, 217–218 hearing 5–6
incompressible fluids 90 frequency-dependent changes 108–110 heart 4, 292, 299–300
mass conservation 93–94 FRET see Förster resonance energy Helmholtz free energy 128, 129
Newtonian fluids 92–98 transfer hemidesmosomes 192–193
non-Newtonian behavior 97–98 friction coefficients 173–174 hemodynamic forces 6
fluid shear fully developed flow 169–180 Hertz contact models 158–159
adhesion 281 functional groups 167, 168 high hydrostatic pressure 90
flow chambers 167–168 Fura-2 fluorescent dyes 33, 40 Hill equations 300–301
mechanotransduction 316, 317 fusion, membrane 8 Hodgkin–Huxley model 305
fluorescence 33, 38–40 Hooke’s law 37, 57–58, 73, 76,
conformation changes 329–330 G-actin 189–190, 195 173–174
cross-correlation 160–161, 162 gap junctions 287 hormones 31
Delaunay triangulation 178, 179 gating see channels human disease frameworks 4–8
morphological changes 333 gauge pressure 91 Huxley, Aldous/Andrew 305
fluorescence recovery after Gaussian approximations 204–205, hydrolysis 21, 197–198
photobleaching (FRAP) 206–207 hydrophilic domains 250, 251
258–259, 315 Gaussian chains 206 hydrophobic domains 250, 251, 327,
fluorochrome 161 GCPRs see G-protein-coupled 328
flux 256, 258 receptors hydrostatic pressure 89–90, 91, 92,
focal adhesions 36 GDP see guanosine diphosphate 317–318
force GEF see guanine nucleotide exchange hyper/hypotonic concentrations 252
adhesion 281–282, 290–291 factor
atomic force microscopy 159 gel electrophoresis 41–43 ideal chain polymers 189, 203–210,
contraction 299–302, 303–308 genes, definition 20, 25 213–214
freely joined polymer chains genetics, in situ studies 48 image-based modeling 236, 237
211–213 genomes 27 image correlation-based approaches
hydrostatic pressure 89–90 genomic data, bioinformatics 49 178–179
ideal chain polymers 208–209 genotypes 27–29 image distortion 162
laboratory cell mechanics 152–156, GFP see green fluorescent protein image texture correlations 162
160–173 Gibbs free energy 129 incompressibility, micropipette
mechanotransduction 321–322, glycoproteins 323–324 aspiration 14
332–333 Golgi apparatus 29 incompressible fluids 90
membranes 269–272 G-protein-coupled receptors (GCPRs) indenters 158–159
optical traps 154, 156 36 indices of refraction 37
power-stroke models 305, 306 gradients inductible cyclooxygenase (COX-2)
statistical mechanics primers deformation 180–181 enzyme 332
119–120 ideal chain polymers 208–209 inertia 65, 239
worm-like polymer chains 216–217 large deformations 78–79, 80–82 infinitesimal strain 72
Förster resonance energy transfer Newtonian fluids 93 inflammation 287, 288, 289
(FRET) 329–330 optical traps 154, 156 informatics 49
forward model 163 grand canonical ensembles 127 inner ear hair cells 319
forward primers 44 gravity 4, 89–90 inorganic compounds 20
 Index 347

inositol triphosphate (IP3) 32, 36, lab-on-a-chip technologies 167 MAP see mitogen-activated protein
331–332 laboratory cell mechanics 151–185 kinase activity
in-plane shear and tension 259–267 blinding 181–182 mass action law 288–290
inside-out signaling 36 cellular micromanipulation 151–159 mass conservation 93–94
in situ studies 20, 48 controls 181–182 matrices 76, 316, 324, 333–334
integration 49, 92 deformation analyses 173–181 Maxwell bodies 174–175, 176–177
integrins 36, 192–193, 287 force application 167–173 mechanical loading 177, 178
intensive quantities 140 force measurements 160–167 mechanical stress 56
intermediate filaments 192–193, 194 key concepts 182 mechanosensitive channel of large
internal energy lac repressors 215–216 conductance (MscL) 326, 327
equilibria 141–142 lacunae 313, 314, 315 mechanotransduction 311–336
mechanical state changes 123–124 Lamé constants 75 adhesion 280
microcanonical ensembles 135–136 lamellipodia 293, 294 function alterations 330–334
partition functions 141–142 laminar flow 94–95, 168, 169–170 hearing 5–6
statistical mechanics primers laminin 192–193 intracellular signaling 326–330
120–124, 133–136, 141–142 Langevin functions 213–214 key concepts 334
interstitial fluid flow 313, 314, 315 Laplace law 12–13, 14 mechanical signals 311–318
intracellular strains 179–180 large deformations organelles 318–326
inverse problems 163–165 gradients 78–79, 80–82 steps 311, 312
in vitro studies 20, 292, 318 mechanics 78–83 structures 318–326
in vivo studies 287, 288, 292 rotation 82–83 medial fractures 19–20
ion channels see channels strain 80–82 melanoma cells 297
ionophores 331 stretch 79–80, 82–83 melting temperatures 44–45
IP3 see inositol triphosphate large factorials 133 member prestresses 228–229
isometric contractions 300 laser beams 157–158 membranes 249–278
isotropic hydrostatic pressure 91 lateral fractures 19–20 adhesion 284
Law of Laplace 12–13, 14 barrier functions 255–259
jellyfish 40 law of mass action 288–290 bending 267–274
JNK see c-jun N-terminal kinase lenses, definition 36 biology 249–252
Johannsen, Wilhelm 27–28 leukocytes 297 bowtie formations 172
see also neutrophils contraction 301–302
Kelvin-Voigt bodies 174–176 ligands 8, 30, 31, 32, 288–290 definition 84
key concepts light microscopes 37–38 elasticity 269, 301–302
adhesion 308 light momentum 153–154 fusion 8
cell biology fundamentals 50 linear elastic behavior 269 in-plane shear and tension 259–267
contraction 308 linear scaling 227–228 key concepts 275
fluid mechanics primers 115 linings 7, 312, 313 mechanotransduction 324–326,
frameworks 16 lipids 250, 251–255, 325–326, 332 327, 328
laboratory cell mechanics 182 liposomes 250 micropipette aspiration 11, 12
mechanotransduction 334 liquid-drop models 11–12, 282–284 phospholipid self-assembly 252–255
membranes 275 liquid drops 16 polymer networks 238
migration 308 Listeria bacteria 7, 295 receptors 30
polymer cellular mechanics 219 loading modes 83–84 second messengers 33
polymer networks 246 locomotion see migration stretch devices 172–173
solid mechanics primers 84 Lorenz-Mie theory 156 traction forces 160
statistical mechanics primers 147–148 loss moduli 108–110 two meanings 262
kinases 34, 332 low hydrostatic pressure 90 Merkel cells 320
kinematics 60, 62–63, 71–73 lumen 10, 312, 313 messenger RNA (mRNA) 25, 26,
membrane deformations 260–262 lumped parameter models 176 29–30
membrane rotation 268 lungs 7, 297, 313, 317 metalloproteinases 333–334
viscosity 169 lysosome 29 metastasis 8
kinetics lytic limits 328 MFs see microfilaments
adhesion 288–290 micelles 250, 252–255
polymerization 194–198 macrophages 297 microcanonical ensembles 127,
protein unfolding 329 macroscopic systems 125–126, 127 131–136
Kronecker delta 204 macrostates 124–126, 127–128 microcontact printing 167
Kuhn lengths 203, 206, 208–209, magnetic micromanipulation 151–159 microfabricated micropillar arrays
218–219 magnetic twisting 153 165–166
348 Index

microfilaments (MFs) 189–190 neurotransmitters 31 PCR see polymerase chain reaction

micromanipulation 151–159, 173–174 neutrophils 297 pericellular matrices 324
micropatterning 167 adhesion 287, 289, 291, 292 permanent deformation 174
micropillar arrays 165–166 electron micrographs 15 persistence length 198–203, 218–219
micropipette aspiration 9–16 in vivo imaging 287, 288 persistence time 296–297
areal expansion moduli 13–15 micropipette aspiration 11, 15 phase lag 105–106
cell influx 15 polymer networks 238 phenotypes 27–29
elastic solids 16 Newtonian fluids 92–98, 173–174 phosphatases 34
hydrostatic pressure 91 Newton’s fluid law 173–174 phosphatidylinositol 4,5-biphosphate
Law of Laplace 12–13, 14 Newton’s law 54 (PIP2) 331
liquid-drop models 11–12 Newton’s second law 99–102 phospholipids 252–255
liquid drops 16 nitric oxide (NO) 32, 317, 331–332 see also membranes
setup 9–11 nitric oxide synthase (NOS) 331 phosphorylation 34, 332
surface tension 11, 12, 13–15 nodal flow 315 photobleaching 258–259, 315
microplates 281–282 noise 37 photolithography 165
see also plates non-interacting two-level systems 132, PIP2 see phosphatidylinositol
microscopic strain energy 122 142–143 4,5-biphosphate
microscopy 37–41 nonlinear fourfold connectivity 245 pipettes 274, 281, 282
atomic force 40–41, 157–159, 282 nonlinear scaling 225–227 placebos 181–182
light 37–38 non-Newtonian behavior 97–98 plane sections remain plane
mechanical loading 177, 178 normal stresses 56–57 assumptions 62–63
traction force 160–161, 163, NOS see nitric oxide synthase plasma membranes 33, 251–252
164–165 no-slip conditions 93 plasmids 30
microstates 124–126, 127–128, notation 74–75, 78, 79–80, 204 plates 84, 104, 260, 270, 281–282
132–133 nuclear fluorescence imaging 39 pluripotent cell types 28
microtubules (MTs) 191, 192, 195, nuclear intermediate filaments 192, pointed polymer ends 190, 196,
198, 203, 322 193 197–198
migration 279, 292–298, 308 nuclei 29 Poiseuille flow 168
minimum potential energy 121 nucleic acids 22–24, 25, 41–43 Poisson’s ratio 68, 73, 267
mismatches, hydrophobic 327, 328 nucleotides, see also deoxyribonucleic polar decomposition 82
mitochondria 29 acid; ribonucleic acid polarity 191, 192, 197–198, 249, 250
mitogen-activated protein (MAP) polyacrylamide 160
kinase activity 332 open foams 225 polyacrylamide gel electrophoresis
mitosis 191 optical traps 153–157 (PAGE) 41, 42
Möbius, Karl 29 order-of-magnitude analyses 110 polycystic kidney disease (PKD)
molecular biology, fundamentals organelles 29–30, 318–326 320–321
19–30 organic compounds 20 polymerase chain reaction (PCR)
moment 64–65, 239, 265, 269–272 orientation correlation functions 200 43–45
monomers 20, 189, 195–196 orthogonal principal directions 77 polymerase enzyme 28–29
motility, filopodia 236 orthonomal matrices 82 polymerization 194–198, 293–295
MscL see mechanosensitive channel of osculating circles 63 polymers
large conductance osteoblasts 7 Brownian motion 120
MTs see microtubules osteocytes 7, 313, 314, 315, 324 cellular mechanics 189–221
multiaxial cellular deformations 172 osteoporotic fractures 6 biopolymer structure 189–193,
multiple-pole micromanipulators 153 outside-in integrin signaling 36 194
multiplicity of states 126 freely joined chains 189, 210–216,
multipotent cell types 28 packing constraints 254–255 217–218
muscle cells 33, 297, 299–301, PAGE see polyacrylamide gel ideal chain 189, 203–210
316–317 electrophoresis key concepts 219
muscular dystrophy 281 paracrine signals 31 persistence length 198–203,
musculoskeletal system 313–315, 317 parallel plate flow 104, 168–169 218–219
mutually perpendicular principal parallel signaling pathway polymerization kinetics 194–198
directions 77 activation 332 worm-like chains 189, 214–219
myocytes 316–317 parasitic flows 318 classification 202–203
myosin 303–308 particle tracking 160–162 DNA 22–24
partition functions 136, 138, flexible 202–204
Navier’s equations 102–103 139–142, 143 migration 295
Navier–Stokes equations 98–103, 115 pathogens 7 monomer subunits 20
 Index 349

networks 223–247 principal directions 76–77 definition 20, 22–24

affine 229–236 principal strains 77–78 directional properties 24
biomechanical function 236–245 principle of minimum free energy 128 gel electrophoresis 41
free-body diagrams 55 principle of minimum total potential inhibitory 47–48
key concepts 246 energy 121 reductionist experiments 47–48
scaling relationships 225–229 probability distributions 119, 138, transcriptional regulation 28–29
non-interacting two-level 207–208 ribosomes 25, 26, 29
systems 132 probes 20, 42–43 rigid-body mechanics 53, 55–56
polysaccharides 24 promoters 20, 28–29 rigidity 272–274
protein definition 20–22 proportional limits 57–58 rigor mortis 304
RNA 22–24 prostaglandin release 332 rivers 93
semi-flexible 202–203 proteins rotation 60, 61, 82–83, 131, 262, 268
stiff 202–203 antibodies 42 rubber elasticity 230–232
temperature-dependence 129–131 complex formation 35 ryanodine receptors 34
thermal undulations 272–273 conformation changes 329–330
polypeptides 20–21 definition 20–22 saddle point approximations 141
polysaccharides 24 deformable body networks 58, 62, 66 sarcomeres 299–300
potential energy 120–123 filopodia 236 sarcoplasmic reticulum 33
power laws 108–110 folding 22 scaling 114, 115, 225–229
power-stroke models 305–308 gel electrophoresis 41–43 scalpels 157
preload stresses 228–229 MAP kinases 332 scanning atomic force microscopy
preparatory gels 42 mechanotransduction 323–326, modes 158
pressure 329–330 scattering theory 156
absolute 91 membranes 268 scissors 157
atmospheric 91 polymerase chain reaction 43–45 SDS see sodium dodecyl phosphate
deformable body networks 56, 58–59 receptors 30–36 SDS-PAGE see sodium dodecyl
gauge 91 signaling cascades 34 sulphate–polyacrylamide gel
gravitational forces 89–90 tweaking 29 electrophoresis
mechanotransduction 317–318 unfolding 329 secondary antibodies 39–40
micropipette aspiration 12 see also channels secondary messengers 36
Newtonian fluids 95 proteomics 49 second messengers 32–33, 34,
primary antibodies 39 protrusion 293–294, 295 331–332
primary cilia 315, 320, 321 P-selectin 287, 289 second moment of inertia 65
primers 44 pumps see channels selectins 287, 289
fluid mechanics 89–117 pure shear 59 semi-flexible polymers 202–203
dimensional analyses 110–115 semi-fluid substances 108
fluid statics 89–92 radioactivity 42–43 shams 182
key concepts 115 random walks 143–147, 204–207, 256 shape modes 83–84
Navier–Stokes equations 98–103 Rayleigh scattering theory 156 shear
Newtonian fluids 92–98 ray tracing 154–155 adhesion 281, 291, 292
rheological analyses 103–110 RBCs see red blood cells cell deformations 113
solid mechanics 53–87 Re see Reynolds number cone-and-plate flow 170–171
deformable body networks 55–78 receptors 8, 30–36, 288–290 deformable body networks 59–60,
free-body diagrams 53–55 recursion relationships 146 76–78
key concepts 84 red blood cells (RBCs) 7, 9, 10, entrance lengths 169–180
large deformations 78–83 239–241, 268 flow chambers 167–168
loading modes 83–84 red fluorescent protein (dsRED) 40 fourfold connectivity 245
rigid-body mechanics 53 reductionist experiments 46–48 mechanotransduction 312–314,
shape modes 83–84 reflected laser beams 157–158 316–317
structural elements 83–84 refraction 37 membranes 259–267, 270
statistical mechanics 119–149 regulatory signals 316 Newtonian fluids 93
canonical ensembles 136–143 resolution 36 parallel plate flow devices
entropy 124–128 reverse primers 44 168–169
free energy 128–131 Reynolds number (Re) 94–95, sixfold connectivity 242–245
internal energy 120–124 112–115, 168–169 thinning fluids 98
key concepts 147–148 rheological analyses 103–110 thin structure analyses 240–241
microcanonical ensembles 131–136 ribonucleic acid (RNA) shells 84, 260
random walks 143–147 correspondence 25–27 signaling cascades 31, 32, 34, 35
350 Index

signaling pathways 30–31, 35–36, storage moduli 107–110 tension 11, 12, 259–267, 272–274, 327
311–318, 326–330, 332 strain 73–74 tensors 59
silly putty 104, 105 adhesion 286–287 tetanic contractions 300
similitude experiments 113–114 affine polymer networks 233–236 texture correlations 162
single partition functions 143 anisotropic affine polymer TFM see traction force microscopy
single pipettes 281, 282 networks 233 thermal cyclers 45
sinks 256 biaxial membrane deformation 267 thermodynamic systems
site-specific recombinase (SSR) 48 deformable body networks 57–65, canonical approaches 142–143
sixfold connectivity 240, 242–245 68, 71–73, 76–78 equations of state 134
skeletal muscle contraction 300, 301 elastic deformations 122–123 first law 124
skeleton 4 elastic moduli 233–234 free energy minimization 128–129
sliding forces 303–308 equilibria 123 ideal chain polymers 208–209
small deformations strain 72 flexible substrates 172 Thermus aquatiqus 45
small inhibitory RNA (siRNA) 47–48 large deformations 80–82 thin-film deposition 165
smooth muscle cells 297, 299, mechanotransduction 316–317, 318 thin-walled cylinders 60
316–317 polymer networks 227–228 thioredoxin channels 327
sodium dodecyl phosphate (SDS) thin structure analyses 240–241 three-coin systems 124–126
41, 42 streams 93 three-dimensional orientation
sodium dodecyl sulphate– stress 73–74 correlation functions 200
polyacrylamide gel deformable body networks 56–60, thymine (T) 22–24
electrophoresis (SDS-PAGE) 64–65, 69–71, 76–78 time 256–259, 296–297
41, 42 fibers 301–304 torsion 60, 61
solid cylinder torsion 61 flow chambers 167–168 totipotent cell types 28
solid mechanics 53–87 mechanotransduction 312–314, 316 touch sensation 320
deformable body networks 55–78 membranes 263 traction force 160–161, 163, 164–166
free-body diagrams 53–55 moment resultants 269 traction force microscopy (TFM)
key concepts 84 polymer networks 228–229 160–161, 163, 164–165
large deformations 78–83 protein kinase 332 transcription 25, 26, 27, 28–29
loading modes 83–84 strain energy 123 transduction 5–6, 31, 32
rigid-body mechanics 53–55 symmetry 71 transient intercellular adhesions 287,
shape modes 83–84 stretch 79–80, 82–83, 172–173, 288, 289
structural elements 83–84 317, 324 transient receptor potential (TRP)
Southern blotting 42–43 striated muscle cells 299 channels 327
spectrin polymers 240, 268 structural elements, solid mechanics transition regions 95
speed see velocity primers 83–84 translation 25, 26, 27
spring systems 121, 131, 173–177, 285 substrates 160, 172, 279–280, 318, 330 translocation 293
SSR see site-specific recombinase sugars 24 transverse strains 68
stable intercellular adhesions 287, surface tension 11–15, 249, 282–284 treadmilling 197–198
288, 289 symmetric deformations 172–173 TRP see transient receptor potential
statistical mechanics 119–149 channels
canonical ensembles 136–143 T-actin 197 trusses 84
entropy 124–128 tapered micropipette aspiration 14 tubulin 191, 197
free energy 128–131 tapping atomic force microscopy tumors 8
internal energy 120–124 modes 158 turbulence 94–95, 168
key concepts 147–148 Taxol anti-cancer drug 191 tweezers 153–157
microcanonical ensembles 131–136 Taylor series 69–70, 146–147, 199 twitch forces 300
random walks 143–147 temperature two-dimensional shear moduli
steady flow 92–93 Brownian ratchets 294–295 240–241
stem cells 28 hairpin number 134–135 two-level systems 132, 142–143
stereocilia 319 membranes 272–273 two-pore-domain potassium channel
stiffness persistence length 198–200 protein 327
deformable body networks 57–58 polymerase chain reaction 44–45 two-spring systems 121
mechanotransduction 317 polymers 120, 198–200 tyrosine kinases 34
membranes 268 statistical mechanics primers 129–131
polymer networks 202–203, 228–229 thermal cyclers 45 ultimate stress/strength 58
stirrup bone 319 see also thermodynamic systems uniaxial force 172, 301–302
Stokes’ law 152–153 tensegrity 228–229 uniform biaxial stretches 172–173
 Index 351

universal solvents 249 viscoelastic cells 9, 173–177 white light illumination 37–38
uracil (U) 22–24 viscoelastic materials 104–108 worm-like chains (WLC), polymers
viscometers 170–171 189, 214–219
vascular endothelium 312–313, 314 viscosity 104–106, 169, 173–174
vector notation 74–75 Voigt notation 74–75 Y chromosomes 292
velocity volume 14, 93, 94, 206, 225–228 yellow fluorescent protein (YFP)
contraction 300–301, 305, 306 40, 330
migration 296–297 walls, definition 84 Young’s equation 284
Newtonian fluids 93, 94–95 warping 61 Young’s moduli 233, 234, 237, 267
power-stroke models 305, 306 water 89–90, 249, 250
vesicles 29 wavelengths 37, 38 zero temperature limits 129
viability of cells 334 white blood cells 11, 287
viruses 8 see also neutrophils
This page intentionally left blank
to match pagination of print book