CHAPTER

Correlative two-photon
and serial block face
scanning electron
microscopy in neuronal
12
tissue using 3D near-
infrared branding maps
Robert M. Lees*, Christopher J. Peddiex, Lucy M. Collinsonx,
Michael C. Ashby*, Paul Verkade*, 1
*University of Bristol, Bristol, United Kingdom
x
The Francis Crick Institute, London, United Kingdom
1
Corresponding author: E-mail: [email protected]

CHAPTER OUTLINE
Introduction ............................................................................................................ 246
1. Rationale ........................................................................................................... 251
2. Methods ............................................................................................................ 251
2.1 Multiphoton Microscope Setup ............................................................. 251
2.2 Initial Functional and/or Structural Two-Photon Imaging......................... 252
2.2.1 Principle........................................................................................... 252
2.2.2 Materials .......................................................................................... 254
2.2.3 Protocol ........................................................................................... 255
2.3 Near-Infrared Branding ........................................................................ 258
2.3.1 Principle........................................................................................... 258
2.3.2 Materials .......................................................................................... 261
2.3.3 Protocol ........................................................................................... 262
2.4 SBF-SEM Sample Preparation and Imaging ........................................... 263
2.4.1 Principle........................................................................................... 263
2.4.2 Materials .......................................................................................... 265
2.4.3 Protocol ........................................................................................... 266
Concluding Remarks ............................................................................................... 273
Acknowledgments ................................................................................................... 274
References ............................................................................................................. 274

Methods in Cell Biology, Volume 140, ISSN 0091-679X, http://dx.doi.org/10.1016/bs.mcb.2017.03.007 245

© 2017 Elsevier Inc. All rights reserved.
246 CHAPTER 12 Correlative two-photon and serial block face SEM

Abstract
Linking cellular structure and function has always been a key goal of microscopy, but
obtaining high resolution spatial and temporal information from the same specimen is a
fundamental challenge. Two-photon (2P) microscopy allows imaging deep inside intact
tissue, bringing great insight into the structural and functional dynamics of cells in their
physiological environment. At the nanoscale, the complex ultrastructure of a cell’s
environment in tissue can be reconstructed in three dimensions (3D) using serial block
face scanning electron microscopy (SBF-SEM). This provides a snapshot of high
resolution structural information pertaining to the shape, organization, and localization
of multiple subcellular structures at the same time. The pairing of these two imaging
modalities in the same specimen provides key information to relate cellular dynamics
to the ultrastructural environment. Until recently, approaches to relocate a region of
interest (ROI) in tissue from 2P microscopy for SBF-SEM have been inefficient or
unreliable. However, near-infrared branding (NIRB) overcomes this by using the laser
from a multiphoton microscope to create fiducial markers for accurate correlation of 2P
and electron microscopy (EM) imaging volumes. The process is quick and can be user
defined for each sample. Here, to increase the efficiency of ROI relocation, multiple
NIRB marks are used in 3D to target ultramicrotomy. A workflow is described and
discussed to obtain a data set for 3D correlated light and electron microscopy, using
three different preparations of brain tissue as examples.

INTRODUCTION
Light microscopy (LM) is invaluable to cell biologists as a tool for obtaining
dynamic information about cellular function and structure, especially with the
aid of fluorescent reporters or dyes. Multiphoton (MP) fluorescence microscopy
allows the visualization of these cellular dynamics inside living tissue at greater
depth than is possible with conventional fluorescence microscopy (e.g., confocal
microscopy) (Zipfel, Williams, & Webb, 2003). This is achieved by employing
lower energy, longer wavelength light in the near-infrared part of the spectrum
(w700e1400 nm). To excite a fluorophore, near simultaneous absorption of
multiple, lower energy photons are required instead of a single, higher-energy
photon in single photon fluorescence microscopy. The probability of MP excitation
is extremely low in comparison with single photon excitation. To overcome
this, the laser is repeatedly pulsed in ultra-fast (100 fs) bursts to increase the
photon density. Because photon density falls away with distance from the
focal plane, the chance of obtaining MP excitation is effectively zero outside
that focal plane, conferring inherently high three-dimensional (3D) resolution
(Zipfel et al., 2003). MP microscopy is advantageous to a cell biologist
because the near-infrared light is refracted less as it passes through the tissue.
With a sample that refracts very little light, it is possible to routinely achieve
 Introduction 247

imaging depths of around 300 mm, with a strongly emitted signal that raises in
excess of 500 mm. Although there are other examples of MP excitation, this chap-
ter is focused on two-photon (2P) excitation, the absorption of two photons, using
an MP microscope.
Neuroscience research can utilize 2P microscopy to aid in studies of the
mammalian brain that were previously limited to smaller, semitransparent model
organisms (Svoboda & Yasuda, 2006). It is possible to image the cortical neurons
of mice longitudinally in vivo through cranial windows or thinned skull with either
genetically encoded or virally expressed fluorescent reporters (Holtmaat et al.,
2009; Yang, Pan, Parkhurst, Grutzendler, & Gan, 2010). Electrophysiological
studies can also be aided in the case of genetically encoded calcium indicators
(GECIs) (Akerboom et al., 2012). GECIs are used to look at intracellular Ca2þ
transients in neurons that are a proxy measure of membrane depolarization and
hence neuronal cell activity. Another use of 2P microscopy in neuroscience
is for the uncaging of neurotransmitter (e.g., glutamate; Shoham, O’Connor,
Sarkisov, & Wang, 2005). 2P excitation is used to uncage the neurotransmitter
with high spatial resolution, and the response from the cell is used to map receptors
on the neuron or to induce activity in specific target cells (Ashby & Isaac, 2011).
These techniques have contributed to the understanding of both structural
and functional neuronal dynamics within the spatial resolution limits of 2P
microscopy.
There are technical challenges to labeling and imaging multiple different
fluorophores with 2P in tissue. This includes the lack of fluorophores with
sufficiently separate 2P excitation spectra and the difficulty of obtaining a good
signal-to-noise ratio deep in tissue. Therefore, it has become routine to fix samples
and probe for extra information through immunochemistry using other forms
of LM (often with resectioning of the sample). This is especially useful for
identifying cell types and protein localization post hoc. However, LM becomes
limited by spectral and spatial resolution when identifying, quantifying, and
measuring multiple different subcellular structures that are abundant in the tissue
volume (such as synapses and mitochondria). Problems also occur in penetration of
the tissue with antibodies and therefore obtaining good signal from higher depths.
There are advances in this area, including the reduction of light scattering in whole
tissue samples by clearing (Renier et al., 2014) and serial sectioning of tissue for
easier probing, reprobing and imaging by array tomography (Micheva, O’Rourke,
Busse, & Smith, 2010).
Electron microscopy (EM) can be used to overcome the resolution limits of LM
to be able to distinguish subcellular structures that are a few nanometers apart.
Structures can be identified and have their function inferred by determining protein
composition using immuno-EM. As a tool for neuroscience, EM has been used to
characterize the ultrastructure of neurons for decades, resulting in a comprehensive
understanding of structural identity but with no dynamic information. Pairing
neuronal dynamics from 2P microscopy with the ultrastructural environment
248 CHAPTER 12 Correlative two-photon and serial block face SEM

provides unprecedented insights into structureefunction relationships. However,

correlating 3D light and electron microscopy in the same tissue sample is difficult.
Until recently, the only option for reconstructing tissue volumes at ultrastruc-
tural resolution was through serial section transmission EM (ssTEM) methods.
This involves a highly trained individual cutting and staining perfect serial
sections, acquiring images, and subsequently aligning them in a laborious
workflow. However, by automating the sectioning, imaging, and alignment
simultaneously, time and human error can be mitigated. Both serial block face
scanning electron microscopy (SBF-SEM) and focused ion beam scanning electron
microscopy (FIB-SEM) achieve this goal. They work by destructively sectioning
or milling off the top layer of the tissue and imaging each serial block face (Peddie
& Collinson, 2014). It is conceivable to go from having a fixed brain tissue
specimen to a registered 3D ultrastructural volume within a week. Because FIB-
SEM and SBF-SEM are destructive techniques, they are at a disadvantage in
comparison with ssTEM methods (which have also become more automated in
recent years; Kasthuri et al., 2015). The choice of technique depends on the volume
to be analyzed and the resolution required; at present, SBF-SEM routinely
produces a much larger volume than FIB-SEM, albeit with lower Z-resolution.
The obtainable voxel size for SBF-SEM is currently <20 nm in the Z-axis and
<5 nm in X/Y-axis, whereas FIB-SEM can achieve an isotropic voxel size of
5 nm. In comparison, ssTEM reaches a pixel size of <0.5 nm and a reliable slice
thickness of only 50 nm. The resolution limits of these systems are sufficient for
identification of synaptic structures, making them attractive for use in neurosci-
ence research, especially connectomics (Helmstaedter, 2013).
One of the biggest technical challenges with 3D correlated light and electron
microscopy (CLEM) is finding a reliable and efficient method for relocating the
region of interest (ROI) between imaging modalities. This problem is not novel
and has been approached from many different angles. CLEM in tissue cannot
benefit from finder grid coordinate systems as in cell culture. However, other po-
tential solutions include, immunogold labeling the structures of interest, diamino-
benzidine (DAB) precipitation in the fluorescently labeled cell, and the use of
nearby vasculature as a fiducial marker. These techniques have their uses for
particular samples but present problems for relocation of an ROI in thick tissue.
Membrane structure can become compromised in immunogold labeling due to
steps of permeabilization that requires detergents. Additionally, antibodies can
take a long time to penetrate thicker tissue specimens; as paraformaldehyde fixa-
tion is reversible over time, there will be further loss of structural integrity during
long antibody incubations. Immunogold labeling also only improves identification
and not relocation of the ROI. DAB precipitation can be useful in particularly thin
samples where it is possible to see the precipitate with a brightfield microscope af-
ter resin embedding. The ROI can then be targeted quickly during ultramicrotomy
prior to EM (Knott, Holtmaat, Trachtenberg, Svoboda, & Welker, 2009). However,
 1. Rationale 249

DAB can reduce the contrast intracellularly at the ROI, which is especially bad for
SBF-SEM/FIB-SEM imaging where high contrast is required. Finally, the use of
vasculature as a fiducial marker to relocate an ROI can be extremely useful
when coupled with X-ray imaging using a micro-CT (Karreman et al., 2016).
The X-ray images show contrast where blood vessels are; this can be used to target
ultramicrotomy with great success. However, vasculature is subject to physiolog-
ical variability and so cannot be user defined to target the ROI. There are also
efforts to produce new genetically encoded tags for CLEM that can be resolved
by both LM and EM, which are covered in a previous volume of this book (Hodgson,
Nam, Mantell, Achim, & Verkade, 2014). These may soon be reliable enough for
LM imaging deep in tissue with 2P microscopy.
To circumvent the problems of existing approaches, a technique was developed
that uses the near-infrared laser of a 2P microscope to make fiducial marks in the
tissue, termed near-infrared branding (NIRB; Bishop et al., 2011; Bishop, Nikic,
Kerschensteiner, & Misgeld, 2014). These branding marks can be used to target
relocation by placing them near to the ROI. NIRB marks are formed by increasing
laser power to the point where damage of the tissue occurs. Cumulative buildup of
light exposure results in microbubbles pushing the tissue apart to form a hole over
the area where the laser is scanned. The damage caused is restricted to within a few
micrometers of the focal plane, for this reason branding appears to be an MP process.
The edges of the branding marks are fluorescent in nature, with an emission
spectrum similar to that of the tissue autofluorescence (Bishop et al., 2011). This
is presumably due to the creation and accretion of autofluorescent molecules formed
during the branding process.
Branding has repeatedly been shown to work in fixed neuronal tissue, as well
as kidney and lymphoid tissue (Bishop et al., 2011; Grillo et al., 2013; Maco,
Holtmaat, Jorstad, Fua, & Knott, 2014; Mostany et al., 2013). NIRB marks can
be customized to fit any ROI through the arbitrary scan settings of most micro-
scopy software. These marks can then be relocated in the embedded tissue during
ultramicrotomy, where they appear unstained in the semithin sections in contrast to
the surrounding stained tissue. This allows efficient and accurate relocation of the
ROI on the block face for 3D EM. Subsequent imaging and reconstruction of the
ROI has been demonstrated using both FIB-SEM and ssTEM techniques (Bishop
et al., 2014; Grillo et al., 2013; Maco et al., 2014). NIRB is therefore of remarkable
value when correlating volumes of light and electron images to efficiently target
relocation of an ROI and to provide a fiducial marker between modalities. This
chapter will discuss considerations that should be made during the workflow and
describe in detail a protocol for achieving efficient relocation of an ROI using
3D NIRB maps (Fig. 1).
250 CHAPTER 12 Correlative two-photon and serial block face SEM

FIGURE 1
Workflow diagram outlining steps required for correlative two-photon and serial block face
scanning electron microscopy using near-infrared branding. Steps to the left of the timeline are
involved in processing, and to the right of the timeline are imaging-related steps. The total
duration of the workflow is w10 days. EM, electron microscopy; LM, light microscopy; NIRB,
near-infrared branding; ROI, region of interest; SBF-SEM, serial block face scanning electron
microscopy; TEM, transmission electron microscopy.
 2. Methods 251

1. RATIONALE
2P microscopy allows cell biologists to use light microscopy as a tool to obtain
structural and functional dynamics of cells inside their physiological environment.
3D EM techniques can obtain ultrastructural information from the same ROI.
However, the relocation of the exact ROI for 3D EM imaging is an unattractive
and tedious process. Efficient and accurate relocation can be accomplished by using
3D NIRB maps, which also act as fiducial markers for correlation of the two imaging
modalities.
NIRB was first demonstrated by Bishop et al. (2011) and also documented in
the previous CLEM edition of this book for use with ssTEM and FIB-SEM in brain
tissue (Bishop et al., 2014; Maco et al., 2014). For relocation of a single synapse
in a 1  109 mm3 piece of brain tissue, NIRB marks can reliably reduce the search
area to 1  103 mm3, a millionfold decrease. The technique does not compromise
the preservation or contrast of ultrastructure at the ROI, which is a problem with
other approaches. Previously, the use of a single set of NIRB marks in one plane
was used to relocate the ROI. Here, the efficiency and accuracy of relocation are
increased by creating a 3D NIRB map. The map is formed of multiple fiducial
NIRB marks that aid the relocation of an ROI throughout ultramicrotomy. It is
therefore possible to target ultramicrotomy efficiently and is especially useful
where relocation can take a long time, e.g., larger block faces and ROIs deep in
the tissue.
This chapter shows that correlation of 2P microscopy with SBF-SEM imaging
can be achieved with relative ease, in comparison with previous techniques, using
common equipment present in most EM labs. The reduction in expert skill and
time afforded by SBF-SEM should increase the appeal of this CLEM workflow to
cell biologists.

2. METHODS
2.1 MULTIPHOTON MICROSCOPE SETUP
The same microscope is used for imaging and subsequent branding. The consider-
ations that need to be made when choosing components for the setup are outlined
here:
• The laser must be mode-locked (repetitive, ultrafast pulsing; required for 2P
excitation), have a wavelength suitable for sample excitation, and produce
enough power to create branding marks at the sample (see Near-infrared
branding). A single, tunable laser is used here, set at 910 nm for imaging green
fluorescent protein (GFP) in the sample and 800 nm for branding. The use of
two laser beams, one dedicated to excitation of the sample and one for branding
(set at around 800e900 nm), would make the branding process quicker by not
having to tune a single laser multiple times.
252 CHAPTER 12 Correlative two-photon and serial block face SEM

• The choice of objective lens is important for high-resolution imaging and to

achieve well-defined branding marks. An objective with a higher numerical
aperture (N.A.), and hence smaller point spread function, allows for tighter
confinement of NIRB marks. This is particularly useful for outlining smaller
structures (e.g., a single dendritic spine). Additionally, the objective must have a
long working distance to image and brand within the tissue at the depth of the
ROI (working distance will depend on the depth of the ROI). Water-immersion
objectives with an N.A. of 1.0 or more are used here, this gives enough
control to make sure the marks can be clearly defined by the user. A lower-
magnification objective (e.g., 4, 0.1 N.A. air) is also required for brightfield
imaging to identify landmarks on the tissue.
• Finally, the most important aspect of the imaging setup is the ability to control
both the X- and Y-axis movement of the laser to achieve the correct size/shape
of NIRB mark. It is preferred to have servomotor-controlled galvanometers
(servo mirrors) rather than resonant. Resonant scanning can be used, but
segmented linescans are not possible, as one of the mirrors cannot have its
position user defined. Rotation of the field of view (FOV) can be used to create
shapes with resonant mirrors. This becomes a highly laborious process in
comparison with servo mirrors, which can have a user-defined scanning path.
The preferred software functionality on top of normal image acquisition is an
arbitrary linescan function. However, with some software this may not be
possible. Therefore, as with resonant scanning, the rotation of the FOV can be
used to create shapes with NIRB marks. The branding process is discussed in
more detail later on (see Near-infrared branding).

2.2 INITIAL FUNCTIONAL AND/OR STRUCTURAL TWO-PHOTON

IMAGING
2.2.1 Principle
Below, three examples of sample preparations from different experiments are
used to illustrate the scope of applications for CLEM in brain tissue. The first
preparation is for an electrophysiological recording (Fig. 2A). Here, an acute
400 mm thalamocortical brain slice is taken from a neonatal mouse (postnatal
day 6/7) and whole-cell patch clamp electrophysiology is used to record from a
single neuron while filling it with fluorescent dye from the patch pipette. In this
scenario, it is necessary to correlate the ultrastructure of both the cell and the
surrounding cells (particularly their synaptic inputs) to the response from the
recorded cell. The second preparation is a fixed tissue slice (Fig. 2B), where a
400 mm thalamocortical slice is taken from an adult transgenic mouse expressing
tdTomato in layer 4 of the barrel cortex. CLEM is required here to identify cell
typeespecific effects on synaptic vesicle pool size, synapse size, and number
etc. The last sample preparation is for intravital imaging (Fig. 2C), in which an
adult mouse is fitted with a cranial window and has fluorescent proteins expressed
by viral transduction in the somatosensory cortex. This is the most common
 2. Methods 253

(A) (B) (C)

(A') (B') (C')

FIGURE 2
Initial functional and/or structural two-photon (2P) imaging. Representative images of
sample preparation for 2P imaging of (A) slice electrophysiology, (B) fixed section from a
whole brain, and (C) intravital imaging through a cranial window. (A0 ) Representative 2P
image of patched cell (yellow arrowhead ¼ soma) filled with Alexa Fluor 594 hydrazide from a
glutamate uncaging experiment. Responses from the cell due to discrete glutamate
uncaging along the dendrite are shown (magenta boxes). (B0 ) Representative 2P image of
transgenic neurons expressing tdTomato (yellow arrowheads) in layer IV of the mouse barrel
cortex. (C0 ) Representative 2P image showing axonal structural imaging of cytosolic green
fluorescent protein in layer I of the mouse somatosensory cortex. Addition (green
arrowheads) and loss (red arrowheads) of axonal boutons is indicated in repeated
longitudinal imaging (inset).

scenario in which NIRB is used for CLEM. Typically, dendritic spines or axonal
boutons that have their structure imaged longitudinally (over time) are relocated
to assess changes in size and corresponding ultrastructure of both the labeled
and surrounding cells. These examples are not limiting, but show a variety of cases
in which NIRB can be utilized.
To identify the ROI during subsequent 2P imaging, high-magnification structural
images are acquired after functional imaging and directly before fixation, if not
already collected. These are also used to correlate to the final SBF-SEM data set.
To ensure that the structure of interest can be picked out from the neuropil
in SBF-SEM images, it is essential to optimize the sample preparation. The fluores-
cence labeling should not be homogenous or dense and the structure should be
clearly identified from morphological features (e.g., crossing processes, branches,
vicinity to other cell bodies or the vasculature). This is especially relevant for
254 CHAPTER 12 Correlative two-photon and serial block face SEM

ubiquitous structures, e.g., axons, but not so important for cell bodies, as there will
not be many present within the SBF-SEM imaging volume.
High-magnification 2P structural images are acquired before the tissue is fixed,
as the fluorescence signal will be maximal. This is because fixation with paraformal-
dehyde and glutaraldehyde will introduce autofluorescence and quenching of the
endogenous fluorescence, decreasing signal to noise. It is key to consider this
when choosing fluorescent tags/dyes as they will be affected differently by fixation.
GFP, tdTomato, and Alexa Fluor 594 maintain fluorescence well in this protocol.
Where it is not possible to acquire images before fixation, or there is a need to
have high temporal correlation between the LM and EM, images should be acquired
as soon after fixation as possible.
Low-magnification brightfield images are collected to record the tissue
topography. If the tissue is larger than the FOV, a set of images can be acquired
and stitched together to create a mosaic image of the tissue. An obvious landmark,
such as a ruffle/tear in the tissue or vasculature pattern, is used as the origin for
coordinating the position of the ROI for relocation (Fig. 4). If there is no obvious
landmark, one can be introduced through cutting of the tissue on the corner or edge.
To prevent further structural changes to the tissue, it is fixed as soon as possible
after the final LM images are acquired. Fixation of whole tissue is a relatively slow
process, therefore it is important to consider carefully the osmolarity and pH of any
solutions being perfused to avoid compromising the preservation of ultrastructure.
To avoid any potential problems, fresh fixative with the correct pH is made before
use. It should be noted that CaCl2 is added in this protocol to help preserve lipid
membranes.
For intravital imaging, vasculature of the brain surface is used as a fiducial
marker when relocating ROIs. However, cardiac perfusion removes the contrast
created by red blood cells and therefore a blood vessel marker must be used.
Here, DiI (D-282, Invitrogen), a lipophilic fluorescent dye that is also visible
using brightfield illumination, is used during perfusion to bind to the endothelial
cell membrane of the vasculature (Fig. 4E).
Regardless of fixation technique, the tissue is eventually sectioned to <400 mm
thick and ideally as thin as 100 mm, as long as this still encompasses the ROI. This
is because thicker tissue sections are more likely to ruffle during staining and
embedding steps, decreasing the chance of successful flat embedding.

2.2.2 Materials
2.2.2.1 Equipment
1. MP microscope (e.g., Bruker Ultima Intravital, Leica TCS SP8 MP, Scientifica
VivoScope etc.)
a. Ti:sapphire laser (e.g., Spectra-Physics Mai Tai DeepSee; Newport Spectra-
Physics, UK)
b. Low-magnification objective (e.g., 4, 0.1 N.A.)
c. High-magnification objective (1.0 þ N.A.)
 2. Methods 255

2. Fume hood (for fixation)

3. Vibrating tissue slicer (e.g., Leica VT 1200; Leica Microsystems)
4. Double-edged razor blades (for vibrating tissue slicer)
5. Fine paintbrushes
6. pH meter (to pH perfusates)
7. Glass microscope slides
8. Silicone grease
9. Plastic syringe (10 mL) w/snipped 200 mL pipette tip attached (for silicone
grease)
10. Coverslips (#1.0, rectangular, 22  40 mm)
11. Plastic Pasteur pipettes
12. Perfusion/dissection apparatus (e.g., Fine Science Tools)
a. Small hemostats
b. Large blunt scissors
c. Medium sharp scissors
d. Blunt forceps
e. Small scoop
f. Butterfly cannula w/tubing
g. 10 mL syringes (for PBS and OPTIONAL DiI perfusate)
h. 50 mL syringe (for fixative)
i. 2/3-way tap
13. Small, precision pen-grip drill w/drill bit (e.g., dental drill; OPTIONALdfor
drilling implant from cranial window imaging)

2.2.2.2 Reagents
1. PBS (0.01 M; 137 mM sodium chloride, 2.7 mM potassium chloride, and 10 mM
phosphate buffer, pH 7.4)
2. Glutaraldehyde (25% EM-grade)
3. Paraformaldehyde (16% EM-grade)
4. Sodium cacodylate
5. Calcium chloride (C3306, SigmaeAldrich)
6. ddH2O
7. DiI (D-282, Invitrogen; OPTIONALdfor labeling vasculature)
a. D-Glucose (OPTIONALdfor DiI diluent)
b. Ethanol (OPTIONALdfor DiI stock)

2.2.3 Protocol
2.2.3.1 Sample preparation
1. Sample preparation for imaging is followed according to previously published
methods for slice electrophysiology or in vivo imaging (Edwards, Konnerth,
Sakmann, & Takahashi, 1989; Holtmaat et al., 2009), the discussion of which
is beyond the scope of this chapter. All animal experiments were performed
according to the UK Animal (Scientific Procedures) Act, 1986.
256 CHAPTER 12 Correlative two-photon and serial block face SEM

(A) (B) (C)

(D) (E) (F)

(G) (H) (I)

FIGURE 3
Remounting tissue after fixation. Tissue is postfixed at 4 C for at least 2 h (A). A rectangular
silicone grease well is created on a microscope slide (C) using a syringe attached to a
pipette tip (B). A fine paintbrush is used to handle the tissue after fixation (D) and place it in
the well (E). The tissue is immersed in a small amount of buffer (PBS or cacodylate) (F)
and a coverslip is placed on top (G), gentle pressure is applied on each side to spread the
grease evenly (H). After imaging, the coverslip is removed using a razor blade to lift it away (I).

a. OPTIONAL (tissue not requiring live imaging): The relevant fixation steps
are followed (steps 5e8). The tissue section is washed in 0.01 M PBS and
then mounted for imaging in a few drops of 0.01 M PBS inside a silicone
grease well (Fig. 3).

2.2.3.2 Initial 2P imaging

2. OPTIONAL (for live tissue): Carry out functional 2P imaging (Fig. 2).
3. Low-magnification brightfield images are acquired to record the correct
orientation of the tissue, identify landmarks, and indicate the relative position of
the ROI (4 objective w/0.1 N.A.; Fig. 4).
4. A Z-stack of 2P images is acquired from the tissue surface to the ROI (910 nm,
1024  1024 pixels, 60 objective w/1.0 N.A., 1 mm steps). Afterward, another
stack is acquired encompassing only the structure of interest using a higher
optical zoom.
 (A) (B) (C)

(D) (E)

FIGURE 4
Relocation of region of interests (ROIs) after fixation using landmarks. (A) Representative
low-magnification brightfield image of a fixed thalamocortical slice from a juvenile mouse
(yellow arrowheads ¼ anchor harp marks). (B) Brightfield image of the region outlined in
A (dashed white square). Harp marks (yellow arrowhead), the hippocampus (top-left), and
the barrels of the barrel cortex (yellow, dashed rectangle) are used to relocate a cell that
was previously patched and filled during an uncaging experiment. (C) Two-photon (2P)
maximum z-projection of a cell filled with Alexa 594 hydrazide. (D) Representative brightfield
image of the vasculature pattern seen through a cranial window positioned over the
somatosensory cortex during intravital imaging. The intersection of the dashed lines indicates
the origin (inset; gray dashed lines outline vasculature) that is used to record the relative
positions of ROIs during longitudinal imaging. Representative 2P images of axonal structure
at ROIs 1 and 7 are shown (bottom), achieved through viral expression of cytosolic green
fluorescent protein. (E) Brightfield image of a fixed tangential slice from under the cranial
window in D (Note: this is the second in a series of slices, hence the vasculature has been
sliced through on the right side; scale bar ¼ 500 mm). The origin is clearly visible (inset;
epifluorescence image of DiI-labeled vasculature). Representative multiphoton images of
axonal structure at ROIs 1 and 7 after fixation are shown (bottom; yellow arrowheads ¼
identification of same boutons before and after; scale bars ¼ 10 mm). Note: the focal plane is
not at exactly the same angle.
258 CHAPTER 12 Correlative two-photon and serial block face SEM

2.2.3.3 Fixation
CARE: Fixatives should be used in a fume hood to avoid inhalation and personal
protective equipment used.
5. OPTIONAL (for acute slices): Immediately after imaging, a 400 mm brain
slice is briefly washed three times in 2 mL of cold, fresh fixative (2.5%
glutaraldehyde, 2% paraformaldehyde, 2 mM CaCl2, 0.15 M sodium cacodylate
in ddH20, pH 7.4) using a fine paintbrush to transfer the tissue. Afterward, the
tissue is postfixed for at least 2 h in 2 mL fresh fixative at 4 C (Fig. 3A).
6. OPTIONAL (for whole animal): An adult mouse is put under deep, terminal
anesthesia. Cardiac perfusion is then performed first with 2e3 mL 0.01 M PBS
at 5 mL/min, followed by 20e30 mL fresh, cold fixative (2.5% glutaraldehyde,
2% paraformaldehyde, 2 mM CaCl2, 0.15 M sodium cacodylate in ddH20, pH
7.4) at 5e10 mL/min. Note: Filter solutions with a 0.22 mm filter prior to
perfusion. Tissue is harvested and postfixed in 5 mL fresh fixative at 4 C.
a. OPTIONAL (to visualize vasculature): After exsanguination, 10 mL DiI
(120 mg/mL) is perfused after PBS and prior to the fixative.
i. To make DiI working solution, 200 mL of 6 mg/mL DiI in 100% EtOH is
diluted in 10 mL of a 1:4 mix of 0.01 M PBS and 5% glucose (wt/vol in
dH2O) according to the protocol from Li et al. (2008).
b. OPTIONAL (to obtain tissue under cranial window): After perfusion and
before harvesting tissue, the animal is decapitated and the cranial window
implant is left on during postfixation (2 h in 10 mL fresh fixative at 4 C).
Subsequently, the window and skull are removed using a precision hand drill
and forceps to expose the imaged brain region, but leaving the head bar
intact. 0.01 M PBS is continuously applied using a Pasteur pipette to keep
the area from drying. The head is then mounted in the imaging head-fix
device (in-house; not shown) and a 100e400 mm slice is cut using a
vibratome at the same angle as 2P imaging in 0.01 M PBS. The slice is then
postfixed for a further 2 h in 2 mL fresh fixative at 4 C.
7. After postfixation, a 100e400 mm tissue section is cut from the appropriate
region of the brain in 0.01 M PBS using a vibratome. Tissue sections are then
left in fixative until required for branding (<24 h).

2.3 NEAR-INFRARED BRANDING

2.3.1 Principle
Branding is carried out soon after fixation (within 24 h) to avoid further changes to
the ultrastructure of the tissue and quenching of fluorescence. The tissue section is
washed in buffer before mounting to reduce some of the fluorescence quenching that
is caused by residual fixatives. During mounting, light pressure is applied around the
coverslip to spread the silicone grease and trap the tissue to prevent it from drifting
during imaging (Fig. 3). The grease can spread out further during the imaging
session; therefore care is taken to ensure the coverslip remains flat throughout.
 2. Methods 259

The aim of branding is to make relocation of the ROI efficient and accurate
during ultramicrotomy of the embedded tissue. This is achieved by creating a
3D map of NIRB marks targeting the ROI (Fig. 5A). The Ti:sapphire laser is
used to create NIRB marks in the focal plane. As mentioned previously, using a
higher N.A. objective lens produces more tightly confined NIRB marks in all three
dimensions (greater than 1.0 N.A. is recommended).
Across different samples, the same parameters for NIRB may produce differing
results, therefore branding is tested in a discrete (nonprecious) area of the tissue first.
The accumulated effect of light exposure over time appears to cause branding of the
tissue; a trade-off between the number of lines scanned, the pixel dwell time, and the
laser power is needed to control the spread of damage. To increase the thickness of
an NIRB mark, one of these parameters is increased, keeping all others the same
(Fig. 5B). It is easiest to alter the number of lines scanned as this gives the greatest
dynamic range. Settings are adjusted to achieve an NIRB mark with dimensions of
2e4 mm in the X/Y-axis and 5e15 mm in the Z-axis, measured by the distance
between the centers of the fluorescent edges in 2P images. The fluorescent edges
are not stained during heavy metal staining later in the protocol; therefore, the
thickness of the line is made greater than 2 mm to make identification easier in
semithin sections.
Sometimes, the NIRB mark is incomplete, not creating a fluorescent mark all
the way along the defined linescan (Fig. 5C, C0 , G and G0 ). This may be due to
the heterogeneity of the tissue and certain parts being more susceptible to branding.
As a rule of thumb, if the mark is not visibly fluorescent then the physical branding
was unsuccessful and should be repeated. Alternatively, branding with high settings
can spread damage much further than intended. Small incremental changes to
settings are used to prevent this. The creation of large bubbles (10 mmþ) that escape
to the surface of the tissue sometimes occurs, potentially due to a weakness in
the tissue and escaping microbubbles from the NIRB process. These obscure the
ability to image the area below the bubble; however, they normally clear with
time (Fig. 5E and F).
Note that when changing optical zoom, the pixel dwell time is manually adjusted
(if it is not automatically), to account for the change in pixel size. For example,
repeating a linescan that covers the same physical distance at 1 zoom and 3
zoom produces a much greater branding effect at 3 zoom if all other parameters
are constant, due to the increase in pixels that make up the line. For this reason it
is best to start with the branding mark closest to the ROI (at a high optical zoom),
after testing in a nonprecious area first.
The NIRB mark closest to the ROI acts as a fiducial for correlation, therefore it is
vital to be able to capture it in the SBF-SEM volume. The distance between the closest
mark and the ROI is <10 mm in the Z-axis to avoid unnecessary imaging with SBF-
SEM. This mark is also tight around the ROI in the X/Y-axis to include it in the
SBF-SEM imaging FOV. Here, the final FOV is 25.7 mm and so the NIRB mark
closest to the ROI is no larger than this. With a higher N.A. objective, the distance
between the NIRB marks can be reduced to as little as 4 mm without damaging the
260 CHAPTER 12 Correlative two-photon and serial block face SEM

(A)

(B) (C) (C')

(D) (D') (D'')

(E) (F) (G) (G')

(F')

(F'')

FIGURE 5
Creation of fiducial marks using a Ti:sapphire laser. (A) Schematic of a tissue cross
section with a 3D NIRB map to the cell of interest; NIRB marks are made at regular intervals in
the Z-axis from the surface to the ROI in a concentric pattern. (B) Representative 2P
fluorescence image of NIRB marks showing successively thicker marks with increasing
pixel dwell time (numbers listed are in ms/pixel). (C) 2P image of two concentric, asymmetrical
NIRB marks in the same Z-plane. (C0 ) Brightfield image of the same NIRB marks from C
(red arrowhead ¼ DiI-labeled vasculature; yellow arrowhead ¼ incomplete NIRB mark).
(D, D0 , D00 ) 2P images of three branding marks created at different levels in the tissue, outlining
 2. Methods 261

structure within the ROI (see Bishop et al., 2011). If NIRB marks are placed too close
to the ROI, damage may occur. This is because the thickness of a single mark can vary
considerably, due to the heterogeneity of the tissue (Fig. 5E and F).
Asymmetry of the branding mark is important to be able to accurately know
the orientation of the ROI when correlating the SBF-SEM data set. A segmented
linescan function is most suitable for creating branding marks such as this, because
multiple straight NIRB marks can be created in a single linescan. In cases where a
segmented linescan is not possible, a single straight linescan is used and repeated
with rotation of the FOV to create an asymmetric shape.
The 3D map of NIRB marks is used as a reference during sectioning to keep track
of how close the current block face is to the depth of the ROI (Fig. 7H). Each mark is
easily differentiated by size or shape to avoid confusion during sectioning. A good
guideline is to have NIRB marks every 10e15 mm in the Z-axis, so that one is
always visible in each semithin section during ultramicrotomy. The use of a
cross-shaped NIRB on the tissue surface (Fig. 5D) helps to relocate the ROI during
sectioning if flat embedding is unsuccessful. The center of the cross will usually
appear in one section, even if the tissue is embedded at an angle; this helps to
identify where the center of the ROI is in the embedded block. Note that just as
with MP excitation, branding requires higher power at increasing depths to combat
the effects of increased light refraction.
A high-resolution map of the branding marks made from 2P Z-stack images is
important to obtain the relative positions and sizes of the marks, which are referred
to while sectioning the embedded tissue. To image the NIRB marks, 910 nm is used
(Fig. 5C), as GFP is also excited at this wavelength. However, increasing to 1040 nm
produced a stronger signal from the NIRB marks (not shown). The marks can also be
seen with brightfield illumination (Fig. 5C0 ), which is useful to aid microdissection
of large pieces of tissue (>5 mm  5 mm).

2.3.2 Materials
2.3.2.1 Equipment
1. Power sensor (e.g., Thorlabs S120C photodiode power sensor, Thorlabs)
2. MP microscope
a. Ti:sapphire laser (e.g., Spectra-Physics Mai Tai DeepSee; Newport Spectra-
Physics, UK)

=-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
a dendrite of interest (inset of D00 ; zoom of dashed box). (E) 2P images of two asymmetrical,
concentric branding marks around an axon from a dye-filled neuron (red arrowhead ¼ soma),
two large bubbles (yellow arrowheads) are visible. (F, F0, F00 ) 2P images showing a large bubble
(yellow arrowhead) shrinking over time. (G) A representative 2P fluorescence image of an
incomplete NIRB mark with correlated SBF-SEM data (G0 ) showing the lack of damage to
the ultrastructure in the photobleached area. 2P, two-photon; NIRB, near-infrared branding;
ROI, region of interest; SBF-SEM, serial block face scanning electron microscopy.
262 CHAPTER 12 Correlative two-photon and serial block face SEM

b. Low-magnification objective (4, 0.1 NA. air)

c. High-magnification objective (1.0 N.A.þ)
3. Glass microscope slides
4. Coverslips (#1.0, rectangular, 22  40 mm)
5. Silicone grease
6. 10 mL syringe w/snipped 200 mL pipette tip attached (for silicone grease)
7. Plastic Pasteur pipettes

2.3.2.2 Reagents
1. PBS (0.01 M; 137 mM sodium chloride, 2.7 mM potassium chloride, and 10 mM
phosphate buffer, pH 7.4)

2.3.3 Protocol
2.3.3.1 Remounting tissue for imaging
1. The tissue section is washed and remounted in 0.01 M PBS inside a
silicone grease chamber made on a glass microscope slide and sealed with a
coverslip (Fig. 3). Low-magnification brightfield reference images of the
tissue are used to orient the slice correctly (correct surface facing upward;
Fig. 4A and E).
2. The ROI is then relocated relative to the previously chosen fiducial marks and
centerd with 2P microscopy using the high-resolution structural stack as a
reference (Fig. 4C and E).
3. High-magnification 2P Z-stack images are taken of the structure of interest at this
stage, before NIRB (910 nm, 1024  1024 pixels, 60 objective w/1.0 N.A.,
1 mm steps).

2.3.3.2 Near-infrared branding

4. The segmented linescan function is selected (if possible) and an asymmetrical
shape of 20 mm along the longest edge is made in the tissue at 5 mm superficial
to the structure of interest (30 mm depth; Fig. 5D00 ), using settings previously
tested in an area of the tissue away from the ROI [800 nm, 150 mW (average
power as read from power meter at the sample) and 20 ms pixel dwell time].
NOTE: Some drifting may occur during imaging, be sure to recenter on the ROI
before branding each mark. CAUTION: High laser power may damage the
PMTs, turn them off before commencing the linescan.
5. A second, asymmetrical branding mark of 60 mm along the longest edge is
created at the same depth (Fig. 5C).
6. Subsequent branding marks are made every 10e15 mm in the Z-axis from the
ROI to the surface, with each mark having a progressively longer side length
(Fig. 5A and D0 ). The final branding mark is in the shape of a cross, <2 mm
below the tissue surface (Fig. 5D). NOTE: higher power is needed when
branding deeper in tissue.
 2. Methods 263

7. A high-resolution Z-stack is acquired (910 nm, 1024  1024 pixels, 60

objective w/1.0 N.A., 1 mm steps) of the 3D NIRB map with 2P microscopy.
Higher optical zoom is used to acquire Z-stack images around the structure of
interest, encompassing the closest NIRB mark.

2.4 SBF-SEM SAMPLE PREPARATION AND IMAGING

2.4.1 Principle
Sample preparation steps for thick tissue involve lengthy, repeated rounds of en
bloc heavy metal staining to produce strong contrast and tissue conductivity for
SEM imaging. This staining is sufficient to identify postsynaptic densities, synaptic
vesicles, and mitochondria. Some en bloc staining steps require extra care. Uranyl
acetate may precipitate in the tissue if the tissue is not washed thoroughly with
ddH2O to remove all cacodylate ions from the buffer. Further, for lead aspartate
staining, the lead is removed from the solution if a white precipitate forms prior
to incubation. In this case a new lead aspartate solution should be made up.
Thoroughly cleaning glassware may help when making up lead aspartate solution,
to prevent small impurities causing precipitation.
Resin choice is very important, because certain resins are more susceptible
to charging and beam damage during SEM imaging, which effects the stability of
ultrathin sectioning (Fig. 8H). Here, Durcupan ACM resin is used, as recommended
by Deerinck, Bushong, Thor, and Ellisman (2010).
Care should be taken to ensure the specimen remains flat during embedding.
Thicker tissue sections (400 mm) will ruffle, possibly due to greater strain on the
tissue from staining and dehydration steps. This results in the plane of sectioning
not matching the imaging plane. If this is the case, relocating the branding marks
in the embedded tissue is difficult as they do not appear flat in the semithin sections.
Therefore, it is recommended to keep slices as thin as possible before starting this
part of the protocol.
After embedding and polymerization, the resin block is trimmed and semithin
sections collected and imaged to relocate the NIRB marks (Fig. 7AeG). It is
common for the block to be extremely brittle at this stage. Here, a cutting speed
of 2.0 mm/s and section thickness of 1 mm for semithin sectioning were used.
However, if major artifacts are seen in the semithin sections (tears, holes, etc.), it
is recommended to reduce cutting speed to 1.5 mm/s and section thickness to
500 nm. Trimming the block face to become smaller and smoothing the sides of
the pyramid also increase cutting success (Fig. 7F).
It is vital to inspect every semithin section at this stage to avoid cutting through
the NIRB marks. The branding marks appear as holes in the tissue in semithin
sections with no obvious bordering effect (as seen in 2P). Staining of semithin
sections is required when contrast between the tissue and resin is too low to identify
NIRB marks under brightfield illumination. Omitting the staining step is beneficial
because it eliminates loss of sections during washing and reduces the time required
to prepare each semithin section. If flat embedding of the tissue is successful, the
264 CHAPTER 12 Correlative two-photon and serial block face SEM

marks are easily identified by their shape, but each mark only propagates through
5e10 mm of semithin sections (depending on size of NIRB marks). If unsuccessful,
or sectioning is done at an incorrect angle (relative to the branding plane), then small
parts of the NIRB marks appear in multiple semithin sections (10e20 mm), making
identification harder.
On identification of an NIRB mark in semithin sections, 2P images are referred
to, to calculate how far the block face is from the ROI (Fig. 7H). Once the angle and
depth of the block face are calculated relative to the ROI, sectioning is sped up to
reach the final NIRB mark. The tissue is checked using TEM with a small sample
of ultrathin sections to confirm preservation of ultrastructure.
For SBF-SEM, the trimmed pyramid containing the ROI is detached from the
main bulk of the resin. To prevent loss of the pyramid, it is covered with a small
piece of parafilm, prior to pulling the razor blade through the base of the pyramid.
The block is sputter-coated with 2 nm platinum to improve conductivity. After
insertion into the SBF-SEM microtome, the diamond knife is aligned, and the
surface polished using 100 nm cuts, taking care not to cut into the ROI. The door
is closed and pumped to a pressure of either w5e10 Pa (in a variable pressure
SEM) or 103 Pa (in a high-resolution SEM). See Russell et al. (2017) for an in-
depth description of this process.
Once at vacuum, the imaging and cutting conditions for the SBF-SEM run are set
according to the structures to be resolved (Peddie & Collinson, 2014). The FOV, the
size of the detector, and the resolution are inextricably linked. A pixel resolution of
<5 nm is needed to resolve synaptic structures, which limits the FOV to 25e30 mm
when using a detector with 8192  8192 pixels. Axial resolution is determined
by the section thickness and the accelerating voltage that controls the interaction
volume of the electron beam with the sample, and thus the depth from which the
backscattered electrons are detected. 50 nm section thickness is routine, 25 nm
section thickness can be achieved on most well-prepared samples (in our experi-
ence), and 10 nm section thickness is possible with ideal specimens and environ-
mental conditions (Russell et al., 2017). Here, a 25.7 mm  25.7 mm  25 mm
(XYZ) volume is used to encompass the ROI with a voxel size of
3.1 nm  3.1 nm  50 nm (XYZ). With this resolution it is possible to identify
synaptic vesicles, mitochondria, and postsynaptic densities (Fig. 9F).
Individual .dm4 images are large, in the region of 250 MB, and a serial imaging
run of 500e1000 images can easily reach 250e500 GB. The image stack is batch
converted to a .tiff stack in Digital Micrography (Gatan Inc) or in ImageJ (FIJI
package; Schindelin et al., 2012) using the BioFormats importer (Linkert et al.,
2010) for easier handling. Large .tiff stacks are handled in FIJI and other software
using a virtual stack to reduce RAM load. Contrast variation is reduced across the
volume by equalizing the histogram and smoothing the images with a Gaussian filter
to increase contrast (Fig. 8I).
Correlation of the SBF-SEM data set to the 2P images is not straightforward. The
NIRB mark present in the EM data is used to align the rotation of the 2P images.
The high-resolution 2P images of the ROI are used to measure the rough distance
 2. Methods 265

between the plane of NIRB and the plane of the structure of interest (Fig. 9AeC).
Finally, the structure of interest is identified based on morphology (e.g., crossing
processes, branching, vasculature, cell bodies, etc.).
On identification of the structure of interest, manual segmentation can be carried
out using Amira (FEI) or TrakEM2 (ImageJ plugin; Cardona et al., 2012) to make a
3D model. Both have guides available online or within the software itself to cover
the initial processing and segmentation followed by reconstruction, display,
and measurements. For Amira, the user should refer to the software guide
(https://www.fei.com/software/amira-user-guide/), alternatively the FEI YouTube
channel is populated by some video guides (https://www.youtube.com/channel/
UC33gA9Z-FtCccsj29Z-c2gw). For TrakEM2, the user should refer to (http://
imagej.net/TrakEM2_tutorials).

2.4.2 Materials
2.4.2.1 Equipment
1. Fume hood
2. 60 C oven
3. Fine paintbrushes
4. Glass vials (7e10 mL; e.g., G100, TAAB rolled rim vials)
5. Plastic Pasteur pipettes
6. Tissue rotator (e.g., R050, TAAB rotator, 2 rpmdR050)
7. Measuring cylinders/tubes (for making solutions)
8. 10þ mL syringes and syringe filters (0.22 mm)
9. Molecular sieves (3 Ådfor anhydrous ethanol)
10. Ultramicrotome (e.g., Leica EM UC7 or RMC Powertome)
11. Cocktail sticks
12. Glass microscope slides
13. ACLAR sheets (50 and 200 mm thickness; e.g., AGL4458, Agar Scientific)
14. Metal block/weight (w500 g)
15. Razor blades (single and double edged)
16. Glass (for glass knives, e.g., AGG336, Agar Scientific)
17. Glass knife maker (e.g., LKB 7800 KnifeMaker)
18. Specimen floater (for collecting semithin sections)
19. Brightfield microscope (e.g., Leica DM1000 LED)
20. Copper grids (e.g., AGG2500C, Agar Scientific)
21. TEM (e.g., FEI Tecnai T12 120 kV; JEOL JEM-1400 120 kV)
22. Parafilm
23. Aluminum pins for SBF-SEM specimen mounting (10e006002e50, EM
Resolutions, UK)
24. Sputter coater
25. SEM (e.g., Zeiss Sigma VP FESEM)
26. 3View system (Gatan)
27. Amira 6.0.0 software (FEI)
28. FIJI ImageJ package (2.0.0-rc-43/1.51 g)
266 CHAPTER 12 Correlative two-photon and serial block face SEM

2.4.2.2 Reagents
1. Sodium cacodylate
2. Calcium chloride (C3306, SigmaeAldrich)
3. ddH2O
4. Osmium tetroxide
5. Potassium ferrocyanide
6. Thiocarbohydrazide (TCH)
7. Uranyl acetate
8. L-Aspartic acid (A9256, SigmaeAldrich)
9. Lead nitrate
10. Potassium hydroxide
11. Ethanol
12. Propylene oxide
13. Durcupan ACM epoxy resin kit
14. Cyanoacrylate glue
15. Conductive epoxy glue for 3View sample mounting (CW2400, Farnell
element14, UK)
16. Toluidine blue
17. Borax

2.4.3 Protocol
CARE: Staining and embedding steps are carried out in a fume hood using personal
protective equipment (gloves, goggles, lab coat), the majority of the chemicals used
are harmful or toxic.
NOTE: All incubation steps at room temperature (RT) or 4 C are done in a tissue
rotator (2 rpm; Fig. 6C). All steps are carried out in the same glass vial, transferring
chemicals using a plastic Pasteur pipette, unless otherwise noted. A new pipette is
used for each chemical. Chemicals are not applied directly on to the tissue and
the tissue is not manipulated directly at any stage, as it becomes extremely brittle
after osmication.

2.4.3.1 Staining, dehydration, and flat embedding

1. After fixation, the tissue is transferred to a small glass vial (Fig. 6A) using a fine
paintbrush and washed 5  3 min in cold 0.1 M cacodylate buffer containing
2 mM CaCl2.
2. The tissue is incubated in 2 mL reduced osmium (equal parts 3% potassium
ferrocyanide/0.3 M cacodylate/4 mM CaCl2 and 4% aqueous osmium
tetroxide) for 1 h at 4 C.
a. During the osmium incubation, TCH solution is made. 0.1 g of TCH is
added to 10 mL ddH2O and dissolved at 60 C for 1 h. The solution is
agitated gently by hand to aid dissolving. Afterward, it is filtered using a
0.22 mm syringe filter (Millipore) before use.
3. Wash 5  3 min w/ddH2O at RT.
4. Filtered TCH is added and the tissue incubated for 20 min at RT.
5. Wash 5  3 min w/ddH2O at RT.
 2. Methods 267

(A) (B) (C) (D)

(E) (G)

(F)

(H) (I) (J) (K) (L)

(M) (N) (O) (P) (Q)

FIGURE 6
“Megametal”staining and flat-embedding fixed tissue. (A, B, D) Staining is done inside glass
vials using a tissue rotator for each incubation step to provide gentle agitation (C; 2 rpm). A
plastic Pasteur pipette is used for transferring chemicals in and out of the vial during
incubations and washes, making sure not to touch the tissue (A). (E, F, G) A flat-embedding
chamber is created from ACLAR sheets (thickness and number of spacers depends on
thickness of tissue). After the final resin incubation step the tissue is gently floated to the surface
using a wooden cocktail stick, taking care not to touch the tissue (H) until it can be lifted out (I).
The tissue is placed in the embedding chamber (J) and excess resin is removed using
tissue paper (K). A small amount of fresh resin is applied (L) and the top ACLAR sheet is stuck
down using cyanoacrylate glue (M) and weighed down with a heavy metal block (N). After
polymerization the ACLAR sheets are removed using a thin razor blade (O) and gentle peeling
(P), finally the base ACLAR sheet is peeled away from the tissue with care (Q).

6. 2% osmium tetroxide (in ddH2O) is then added and the tissue is incubated for
30 min at RT.
7. Wash 5  3 min w/ddH2O at RT.
8. The tissue is incubated in 1% uranyl acetate (aqueous in ddH2O; 0.22 mm filter
before use) overnight (16 h) at 4 C.
268 CHAPTER 12 Correlative two-photon and serial block face SEM

(A) (A') (B) (C)

(D) (E) (F) (G)

(H)

FIGURE 7
Relocating the ROI in resin-embedded tissue by correlating semithin sections and 2P images.
(A) Brightfield image of a tissue slice with landmarks indicated (yellow arrowheads). After
resin embedding, the tissue is oriented to match (A0 ). A thin layer of cyanoacrylate glue (B) is
used to stick the flat-embedded tissue to a blank resin stub (C and D). The block face is
trimmed to a trapezoid-faced pyramid that is focused roughly on the area containing the ROI
 2. Methods 269

9. Wash 5  3 min w/ddH2O at RT.

10. After the uranyl acetate is washed out, the tissue is incubated in Walton’s lead
aspartate for 30 min at 60 C.
a. To make Walton’s lead aspartate, 0.066 g of lead nitrate is added to 9 mL
0.03 M aspartic acid stock and then pH is adjusted to 5.5 using 1 M
potassium hydroxide, then incubated at 60 C for 30 min. NOTE: No
precipitate should form.
11. Wash 5  3 min w/ddH2O at RT.
12. The tissue is then dehydrated through serial 20%, 50%, 70%, 90% (in ddH2O),
100% EtOH, and 100% EtOH (anhydrous) incubations for 5 min each and
finally fresh 100% EtOH (anhydrous) for 10 min.
13. Tissue is then incubated in propylene oxide for 10 min while Durcupan ACM
resin is made. NOTE: Use a glass Pasteur pipette for propylene oxide, as
plastic is corroded by it.
a. To make Durcupan ACM resin, 11.4 g part A (epoxy resin), 10 g part B
(hardener), 0.3 g part C (accelerator), 0.05e0.1 g part D (dibutyl phthalate)
are added sequentially on top of each other, mixing well after each part is
added.
14. Tissue is transferred through serial incubations of 25%, 50%, and 75%
Durcupan in propylene oxide for 2 h each.
15. Finally, the tissue is incubated in 100% Durcupan overnight and changed to
fresh 100% resin in the morning for 2 h.
16. The tissue is then carefully removed from the resin using a cocktail stick
(Fig. 6H and I) and transferred to a flat-embedding chamber (Fig. 6J), excess
resin is removed (Fig. 6K) and a drop of fresh resin added (Fig. 6L). The resin
chamber is sealed with cyanoacrylate glue (Fig. 6M) and placed in a 60 C
oven to polymerize for 72 h. A relatively heavy metal block (Fig. 6N) is placed
on top to keep the specimen flat.
2.4.3.2 Relocating the region of interest in embedded tissue
17. After polymerization of resin, the flat-embedding chamber is retrieved from the
oven and the layers of ACLAR are removed using a thin razor blade (Fig. 6O).
Each ACLAR layer is peeled away from the resin (Fig. 6P and Q) and the
tissue is imaged using a brightfield microscope to determine the correct face to
begin sectioning from (Fig. 7A0 ).

=--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
(E, F). Semithin sections are cut from the block face using a glass knife (G) and stained
with toluidine blue. (H) The 2P Z-stack images of the branding marks are used as a reference for
correlating brightfield images of the stained semithin sections (arrowheads ¼ indicating
edges of NIRB marks). The cross is just visible on the tissue surface (top row). The second
branding mark is visible in multiple section (two shown; middle row). Sectioning is stopped when
the NIRB mark closest to the ROI is identified (bottom row). Note: these branding marks are not
all asymmetrical, this example was shown as a good example of relocation. It is recommended
to use asymmetrical marks for correlation. 2P, two-photon; ROI, region of interest.
270 CHAPTER 12 Correlative two-photon and serial block face SEM

18. A thin layer of cyanoacrylate glue is used to attach the resin-embedded tissue to
a flat-topped blank resin stub (Fig. 7C), making sure the tissue face containing
the ROI faces upwards. The glue is allowed to polymerize for at least 2 h
before attempting any sectioning.
19. The stub is mounted in an ultramicrotome and a razor blade is used to outline
the rough area of the ROI on the block face (Fig. 7E). The block face is then
trimmed to a trapezoid-faced pyramid (Fig. 7F).
20. Semithin sections (0.5e1.0 mm) are cut from the block face using a glass knife
at 2.0 mm/s until tissue is present in the sections. Each section containing
tissue is transferred to a clean glass microscope slide and dried. Every section
is imaged using a brightfield microscope until the first branding mark is
relocated on the tissue surface (Fig. 7H). The block face is then roughly
trimmed around this region.
a. OPTIONAL (to increase contrast of tissue): Sections are stained with 1%
toluidine blue (in 1% borax).
21. Sectioning is continued until each subsequent NIRB mark is relocated
(Fig. 7H). Once the final NIRB mark is relocated, the block face is trimmed
tightly (<1  1 mm) around it (Fig. 8A).
22. OPTIONAL (to check preservation of ultrastructure): A few ultrathin sections
are cut, mounted on a copper grid, and imaged using a TEM without
counterstaining.

2.4.3.3 SBF-SEM preparation and imaging

23. The block is covered with parafilm and the top 2 mm of the block is trimmed off
with a razor blade. The trimmed piece is mounted on an aluminum pin using
conductive epoxy resin and baked overnight at 60 C. Subsequently, the block
is sputter-coated with 2 nm platinum (Fig. 8B).
24. The pin is mounted in the 3View (Fig. 8C and D) and the block face
polished with 100 nm cuts (Fig. 8E), before closing the chamber door and
pumping to 5e10 Pa of nitrogen gas. The block is imaged during approach
cuts to relocate the ROI (Fig. 8G). The ROI is centered in the FOV, and
parameters are set for SBF-SEM imaging and sectioning. The SEM is operated
at an accelerating voltage of 2 kV with high current mode active, a 20 mm
aperture, and chamber pressure of w5 Pa. A per pixel dwell time of 2 ms is
used with a slice thickness of 50 nm. Images are acquired at 8192  8192
pixels (horizontal frame width of 25.7 mm, reported pixel size of 3.1 nm) and
indicated magnification of 10,000. The entire volume comprises 500 slices,
totaling 16,512 mm3.

2.4.3.4 Image processing and reconstruction

25. .dm4 images from the 3View are initially batch processed using FIJI software,
reading files with the “Bioformats” plugin, and converting them to eight-bit
.tiff files using the “Batch Convert” plugin.
 2. Methods 271

(A) (B) (C) (D)

(E) (F) (H)

(I)
(G)

FIGURE 8
SBF-SEM sample preparation and imaging setup. The final block face (A) is
accurately squared-off using a glass knife and ultramicrotome, then subsequently stuck to
an aluminum pin and sputter-coated (B). The block is inserted into the SBF-SEM
microtome (C and D). Initial cuts are made to polish the block face before the door is closed
(E). Images are collected during approach cuts (G; inset, zoom of dashed box). The ROI
is relocated using the final semithin section as a reference (F; yellow arrowhead ¼ corner
of NIRB). (H) Representative SBF-SEM images of flaking (left and middle) and
charging artifacts (right). (I) Conversion of raw .dm4 files with varying relative histograms
(top row) are converted to .tiff files and contrast-enhanced as well as smoothed with
a Gaussian filter (bottom row). NIRB, near-infrared branding; ROI, region of interest;
SBF-SEM, serial block face scanning electron microscopy.

a. OPTIONAL (for large files): Files are downsampled to reduce file size
to make them easier to handle. This is used when doing rough
correlation and tracing of structure without fine subcellular structural
detail. Alternatively, large file formats (Amira) or virtual stacks (FIJI) are
used to handle large data sets.
272 CHAPTER 12 Correlative two-photon and serial block face SEM

(A) (B) (C) (D)

(E)

(F) (F') (F'')

FIGURE 9
Correlation of 2P and SBF-SEM images and identification of synaptic structures. (A) A
2P image of a cell of interest (filled arrowhead) before branding and (B) after branding
(empty yellow arrowhead ¼ reference cell body, magenta arrowhead ¼ section of
branding mark). (C) One slice from the SBF-SEM data set for correlating to the 2P images in
A and B. (D) 3D reconstruction of the cell of interest in Amira, overlaid with the 2P
fluorescence image. (E) 3D reconstruction in Amira of a dendrite of interest with
orthogonal planes from an SBF-SEM data set, showing filopodia-like spines at synapses
(inset). (F, F0, F00 ) Three consecutive 50 nm sections from an SBF-SEM data set with 4 nm
pixel size. 2P, two-photon; ax, axon; de, dendrite; m, mitochondrion; psd,
postsynaptic density; SBF-SEM, serial block face scanning electron microscopy;
sv, synaptic vesicles.
 Concluding remarks 273

26. To normalize the intensity across the stack of images, the .tiff files are further
processed using the “Enhance Contrast” plugin in FIJI with a value of 0.3%
(Fig. 8I).
27. To remove noise from the SBF-SEM images that result from nitrogen gas (used for
variable pressure), a Gaussian blur filter in ImageJ with a sigma value of 2 is used.
28. The SBF-SEM images are correlated to 2P images using fiducials e.g., cell
bodies, vasculature, branding marks (Fig. 9AeC).
29. Structures of interest are subsequently manually reconstructed in FIJI software
using TrakEM2 or in Amira 6.0.0 software (Fig. 9D and E).

CONCLUDING REMARKS
Correlating light and electron microscopy images of intact tissue allows cell
biologists to link cellular dynamics to the fine ultrastructural detail of a cell’s
environment. This technique has previously been unappealing for cell biologists
carrying out 2P microscopy in tissue due to the difficult and laborious nature of
relocating specific regions of interest. However, 3D EM techniques can make
correlating tissue volumes in LM and EM a lot more accessible, allowing for the
precise measurement of the geometry and organization of thousands of subcellular
structures in the volume. It is clear that NIRB coupled with 3D EM (FIB-SEM or
SBF-SEM) dramatically reduces the human time and skill required to achieve 3D
CLEM. Here, the NIRB method is refined to go from a live sample to a correlated
SBF-SEM data set in under 10 days.
There are additional ways of improving the efficiency and accuracy of this
technique. The greatest technical challenge is a reliable flat-embedding technique.
Sectioning is labor intensive and prone to errors, unnecessary time is added to the
protocol if NIRB marks do not appear flat on sectioning, as identification is
difficult. It may be possible to improve the accuracy of targeted ultramicrotomy
using alternative NIRB patterning to that outlined here. Other improvements
include, additional labeling for extra fiducials (e.g., cell nuclei), DAB precipitation
of the NIRB marks to make them more obvious in semithin sections (Knott et al.,
2009) and micro-CT imaging to highlight vasculature and potentially NIRB marks
to target ultramicrotomy (Karreman et al., 2016). However, these refinements
require more equipment, expertise, and time.
NIRB has mostly been limited to neuronal cell biology; marking ROIs from
intravital imaging (Bishop et al., 2011; Blazquez-Llorca et al., 2017; Grillo et al.,
2013; Karreman et al., 2016; Maco et al., 2014; Mostany et al., 2013). Here,
some suggestions for the use of NIRB and 3D CLEM in neuronal tissue, outside
of cranial window imaging, are given. NIRB has also been attempted in nonneuronal
tissue, but at limited depths (Karreman et al., 2016). Therefore, it is currently open to
be adapted for use in nonneuronal applications. More widespread utilization of 3D
CLEM in tissue will bring powerful structural evidence to functional studies,
allowing cell biologists to ask questions that are not otherwise possible.
274 CHAPTER 12 Correlative two-photon and serial block face SEM

ACKNOWLEDGMENTS
RML is funded by a Wellcome Trust PhD studentship. This work was supported by the Francis
Crick Institute, which receives its core funding principally from Cancer Research UK
(FC001999), the UK Medical Research Council (FC001999), and the Wellcome Trust
(FC001999). In addition, this research was supported by the MRC, BBSRC, and EPSRC
under grant award MR/K01580X/1 to LMC and Peter O’Toole (York University). Equipment
used in MCA’s laboratory was partially funded by the Medical Research Council (MR/
J013188/1) and EUFP17 Marie Curie Actions (PCIG10-GA-2011-303680). We thank the
Wolfson Bioimaging Facility for their support and expertise, and MRC funding of a
preclinical in vivo functional imaging platform for translational regenerative medicine.

REFERENCES
Akerboom, J., Chen, T.-W., Wardill, T. J., Tian, L., Marvin, J. S., Mutlu, S., … Looger, L. L.
(2012). Optimization of a GCaMP calcium indicator for neural activity imaging. Journal
of Neuroscience, 32, 13819e13840. http://dx.doi.org/10.1523/JNEUROSCI.2601-12.2012.
Ashby, M. C., & Isaac, J. T. R. (2011). Maturation of a recurrent excitatory neocortical
circuit by experience-dependent unsilencing of newly formed dendritic spines. Neuron,
70, 510e521. http://dx.doi.org/10.1016/j.neuron.2011.02.057.
Bishop, D., Nikic, I., Brinkoetter, M., Knecht, S., Potz, S., Kerschensteiner, M., & Misgeld, T.
(2011). Near-infrared branding efficiently correlates light and electron microscopy.
Nature Methods, 8, 568e570. http://dx.doi.org/10.1038/nmeth.1622.
Bishop, D., Nikic, I., Kerschensteiner, M., & Misgeld, T. (2014). The use of a laser for
correlating light and electron microscopic images in thick tissue specimens. In
T. Muller-Reichert, & P. Verkade (Eds.), Correlative light and electron microscopy:
Vol. 124. Methods in cell biology (pp. 328e338). Academic Press.
Blazquez-Llorca, L., Valero-Freitag, S., Rodrigues, E. F., Merchán-Pérez, Á.,
Rodrı́guez, J. R., Dorostkar, M. M., … Herms, J. (2017). High plasticity of axonal
pathology in Alzheimer’s disease mouse models. Acta Neuropathologica Communica-
tions, 5, 14. http://dx.doi.org/10.1186/s40478-017-0415-y.
Cardona, A., Saalfeld, S., Schindelin, J., Arganda-Carreras, I., Preibisch, S., … Douglas, R. J.
(2012). TrakEM2 software for neural circuit reconstruction. PLoS One, 7, e38011. http://
dx.doi.org/10.1371/journal.pone.0038011.
Deerinck, T. J., Bushong, E. A., Thor, A., & Ellisman, M. H. (2010). NCMIR methods for 3D
EM: A new protocol for preparation of biological specimens for serial block face scanning
electron microscopy (Online). https://ncmir.ucsd.edu/sbem-protocol
Edwards, F. A., Konnerth, A., Sakmann, B., & Takahashi, T. (1989). A thin slice preparation
for patch clamp recordings from neurones of the mammalian central nervous system.
Pflugers Archive, 414, 600e612. http://dx.doi.org/10.1007/BF00580998.
Grillo, F. W., Song, S., Ruivo, L. M. T.-G., Huang, L., Gao, G., Knott, G. W., … De Paola, V.
(2013). Increased axonal bouton dynamics in the aging mouse cortex. Proceedings of the
National Academy of Sciences of the United States of America, 110, E1514eE1523. http://
dx.doi.org/10.1073/pnas.1218731110.
Helmstaedter, M. (2013). Cellular-resolution connectomics: Challenges of dense neural circuit
reconstruction. Nature Methods, 10, 501e507. http://dx.doi.org/10.1038/nmeth.2476.
 References 275

Hodgson, L., Nam, D., Mantell, J., Achim, A., & Verkade, P. (2014). Retracing in correlative
light electron microscopy: Where is my object of interest? In T. Muller-Reichert, &
P. Verkade (Eds.), Correlative light and electron microscopy: Vol. 124. Methods in cell
biology (pp. 1e21). Academic Press.
Holtmaat, A., Bonhoeffer, T., Chow, D. K., Chuckowree, J., De Paola, V.,
Hofer, S. B., … Wilbrecht, L. (2009). Long-term, high-resolution imaging in the mouse
neocortex through a chronic cranial window. Nature Protocols, 4, 1128e1144. http://
dx.doi.org/10.1038/nprot.2009.89.
Karreman, M. A., Mercier, L., Schieber, N. L., Solecki, G., Allio, G.,
Winkler, F., … Schwab, Y. (2016). Fast and precise targeting of single tumor cells in
vivo by multimodal correlative microscopy. Journal of Cell Science, 129, 444e456.
http://dx.doi.org/10.1242/jcs.181842.
Kasthuri, N., Hayworth, K. J., Berger, D. R., Schalek, R. L., Conchello, J. A., Knowles-
Barley, S., … Lichtman, J. W. (2015). Saturated reconstruction of a volume of
neocortex. Cell, 162, 648e661. http://dx.doi.org/10.1016/j.cell.2015.06.054.
Knott, G. W., Holtmaat, A., Trachtenberg, J. T., Svoboda, K., & Welker, E. (2009). A protocol
for preparing GFP-labeled neurons previously imaged in vivo and in slice preparations for
light and electron microscopic analysis. Nature Protocols, 4, 1145e1156. http://
dx.doi.org/10.1038/nprot.2009.114.
Li, Y., Song, Y., Zhao, L., Gaidosh, G., Laties, A. M., & Wen, R. (2008). Direct labeling and
visualization of blood vessels with lipophilic carbocyanine dye DiI. Nature Protocols, 3,
1703e1708. http://dx.doi.org/10.1038/nprot.2008.172.
Linkert, M., Rueden, C. T., Allan, C., Burel, J.-M., Moore, W.,
Patterson, A., … Swedlow, J. R. (2010). Metadata matters: Access to image data in the
real world. The Journal of Cell Biology, 189, 777e782. http://dx.doi.org/10.1083/
jcb.201004104.
Maco, B., Cantoni, M., Holtmaat, A., Kreshuk, A., Hamprecht, F. A., & Knott, G. W. (2014).
Semiautomated correlative 3D electron microscopy of in vivoeimaged axons and
dendrites. Nature Protocols, 9, 1354e1366. http://dx.doi.org/10.1038/nprot.2014.101.
Maco, B., Holtmaat, A., Jorstad, A., Fua, P., & Knott, G. (2014). Correlative in vivo 2-photon
imaging and focused ion beam scanning electron microscopy: 3D analysis of neuronal
ultrastructure. In T. Muller-Reichert, & P. Verkade (Eds.), Correlative light and electron
microscopy: Vol. 124. Methods in cell biology (pp. 339e361). Academic Press.
Micheva, K. D., O’Rourke, N., Busse, B., & Smith, S. J. (2010). Array tomography: High-
resolution three-dimensional immunofluorescence. Cold Spring Harbor Protocols,
2010. http://dx.doi.org/10.1101/pdb.top89.
Mostany, R., Anstey, J. E., Crump, K. L., Maco, B., Knott, G., & Portera-Cailliau, C. (2013).
Altered synaptic dynamics during normal brain aging. Journal of Neuroscience, 33,
4094e4104. http://dx.doi.org/10.1523/JNEUROSCI.4825-12.2013.
Peddie, C. J., & Collinson, L. M. (2014). Exploring the third dimension: Volume electron
microscopy comes of age. Micron, 61, 9e19. http://dx.doi.org/10.1016/j.micron.2014.01.009.
Renier, N., Wu, Z., Simon, D. J., Yang, J., Ariel, P., & Tessier-Lavigne, M. (2014). iDISCO: A
simple, rapid method to immunolabel large tissue samples for volume imaging. Cell, 159,
896e910. http://dx.doi.org/10.1016/j.cell.2014.10.010.
Russell, M. R. G., Lerner, T. R., Burden, J. J., Nkwe, D. O., Pelchen-Matthews, A.,
Domart, M.-C., … Collinson, L. M. (2017). 3D correlative light and electron microscopy
of cultured cells using serial blockface scanning electron microscopy. Journal of Cell
Science, 130, 278e291. http://dx.doi.org/10.1242/jcs.188433.
276 CHAPTER 12 Correlative two-photon and serial block face SEM

Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M.,
Pietzsch, T., … Cardona, A. (2012). Fiji: An open-source platform for biological-image
analysis. Nature Methods, 9, 676e682. http://dx.doi.org/10.1038/nmeth.2019.
Shoham, S., O’Connor, D. H., Sarkisov, D. V., & Wang, S. S.-H. (2005). Rapid neurotrans-
mitter uncaging in spatially defined patterns. Nature Methods, 2, 837e843. http://
dx.doi.org/10.1038/nmeth793.
Svoboda, K., & Yasuda, R. (2006). Principles of two-photon excitation microscopy and its
applications to neuroscience. Neuron, 50, 823e839. http://dx.doi.org/10.1016/
j.neuron.2006.05.019.
Yang, G., Pan, F., Parkhurst, C. N., Grutzendler, J., & Gan, W.-B. (2010). Thinned-skull
cranial window technique for long-term imaging of the cortex in live mice. Nature
Protocols, 5, 201e208. http://dx.doi.org/10.1038/nprot.2009.222.
Zipfel, W. R., Williams, R. M., & Webb, W. W. (2003). Nonlinear magic: Multiphoton
microscopy in the biosciences. Nature Biotechnology, 21, 1369e1377. http://dx.doi.org/
10.1038/nbt899.