Correlative Light and Electron MIcroscopy (Thomas Müller-Reichert and Paul Verkade

Series Editors
Leslie Wilson
Department of Molecular, Cellular and Developmental Biology
University of California
Santa Barbara, California
Paul Matsudaira
Department of Biological Sciences
National University of Singapore
Singapore
Methods in Cell Biology
VOLUME 111
Correlative Light and Electron Microscopy
Edited by
Thomas Müller-Reichert
Medical Theoretical Center,
TU Dresdsen,
Germany
Paul Verkade
Wolfson Bioimaging Facility,
Schools of Biochemistry and Physiology & Pharmacology,
University of Bristol,
Bristol,
United Kingdom
AMSTERDAM • BOSTON • HEIDELBERG • LONDON

NEW YORK • OXFORD • PARIS • SAN DIEGO
SAN FRANCISCO • SINGAPORE • SYDNEY • TOKYO
Academic Press is an imprint of Elsevier
Academic Press is an imprint of Elsevier
The Boulevard, Langford Lane, Kidlington, Oxford, OX51GB, UK

32, Jamestown Road, London NW1 7BY, UK
Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands
225 Wyman Street, Waltham, MA 02451, USA
525 B Street, Suite 1900, San Diego, CA 92101-4495, USA
First edition 2012
Copyright © 2012 Elsevier Inc. 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 electronic, mechanical, photocopying, recording or otherwise without the prior written
permission of the publisher
Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford,
UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email: permissions@elsevier.com.
Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/
locate/permissions, and selecting Obtaining permission to use Elsevier material
Notice
No responsibility is assumed by the publisher for any injury and/or damage to persons or property as
a matter of products liability, negligence or otherwise, or from any use or operation of any methods,
products, instructions or ideas contained in the material herein.
Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses
and drug dosages should be made
ISBN: 978-0-12-416026-2
ISSN: 0091-679X
For information on all Academic Press publications

visit our website at store.elsevier.com
Printed and bound in USA
12 13 14 10 9 8 7 6 5 4 3 2 1
CONTRIBUTORS
Numbers in parentheses indicate the pages on which the author’s contributions

begin.
Marius Ader (75), Center for Regenerative Therapies, TU Dresden, Fetscherstraße

105, Dresden, Saxony, Germany
Hannah E.J. Armer (357), Imaging Suite, Institute of Ophthalmology, University
College London, London, UK
Felix J.B. Bäuerlein (259), Department of Molecular Structural Biology, Max Planck
Institute of Biochemistry, Am Klopferspitz 18, Martinsried, Germany
Patrizia Boccacci (95), MicroScoBio Research Center, Department of Experimental
Medicine (DIMES), Department of Physics (DIFI) and Department of Informatics,
Bioengineering, Robotics and Information Science (DIBRIS), University of Genoa,
Genoa, Italy
F. Braet (1), Australian Centre for Microscopy & Microanalysis, The University of
Sydney, NSW, Australia; School of Medical Sciences (Discipline of Anatomy and
Histology) —The Bosch Institute, The University of Sydney, NSW, Australia
John A.G. Briggs (235), Structural and Computational Biology Unit, European
Molecular Biology Laboratory, Meyerhofstr. 1, Heidelberg, Germany; Cell Biol-
ogy and Biophysics Unit European Molecular Biology Laboratory, Meyerhofstr. 1,
Heidelberg, Germany
Edward Brown (175), Department of Biochemistry, School of Medical Sciences,
University of Bristol, University Walk, Bristol, UK
Mandy Büchner (223), Medical Theoretical Center (MTZ), Medical Faculty Carl Gus-
tav Carus, University of Technology Dresden, Fiedlerstraße 42, Dresden, Germany
Daniel J. Bumbarger (203), Department of Evolutionary Biology, Max Planck
Institute for Developmental Biology, Spemannstrasse 37/ IV, Tübingen, Germany
Andrew J. Bushby (357), The Nanovision Centre, School of Engineering and Mate-
rials Science, Queen Mary University of London, London, UK
D. Cheng (1), School of Medical Sciences (Discipline of Anatomy and Histology) —
The Bosch Institute, The University of Sydney, NSW, Australia
Lucy M. Collinson (357), Electron Microscopy Unit, London Research Institute,
Cancer Research UK, London, UK
Katia Cortese (95), MicroScoBio Research Center, Department of Experimental
Genoa, Italy
Thomas J. Deerinck (139), National Center for Microscopy and Imaging Research,
University of California, San Diego, La Jolla, CA, USA
xi
xii Contributors
Clare L. Dennison (307), Technology Facility, Biology Department, University of

York, York, UK
Alberto Diaspro (95), MicroScoBio Research Center, Department of Experimental
Genoa, Italy; IFOM, Fondazione Istituto FIRC di Oncologia Molecolare, Milan,
Italy; LAMBS-IFOM, MicroScoBio, Department of Physics, University of Genoa,
Italy; Nanophysics, Istituto Italiano di Tecnologia, Genoa, Italy
Dominic Eberle (75), Center for Regenerative Therapies, TU Dresden, Fetscher-
straße 105, Dresden, Saxony, Germany
Mark H. Ellisman (139), National Center for Microscopy and Imaging Research,
University of California, San Diego, La Jolla, CA, USA; Department of Neurosci-
ences, University of California, San Diego, La Jolla, CA, USA
Benjamin D. Engel (259), Department of Molecular Structural Biology, Max Planck
Gunar Fabig (75), Center for Regenerative Therapies, TU Dresden, Fetscherstraße
Claire-Lise Forestier (59), INSERM, Institute Pasteur, 25 Rue du Docteur Roux,
Paris, France
Maria Cristina Gagliani (95), MicroScoBio Research Center, Department of
Experimental Medicine (DIMES), Department of Physics (DIFI) and Department of
Informatics, Bioengineering, Robotics and Information Science (DIBRIS), University
of Genoa, Genoa, Italy
Philippe Gasser (325), Electron Microscopy ETH Zurich – EMEZ, ETH Zurich,
Switzerland
Ben N.G. Giepmans (157), Department of Cell Biology, University Medical Center
Groningen (UMCG), University of Groningen, A. Deusinglaan 1, Bldg 3215, room
749, AV Groningen, The Netherlands
Markus Grabenbauer (117), Department of Systems Cell Biology, Max-Planck-
Institute for Molecular Physiology, Otto-Hahn-Str. 11, Dortmund, North Rhine-
Westphalia, Germany
Maja Günthert (325), Electron Microscopy ETH Zurich – EMEZ, ETH Zurich,
Switzerland
Bruno M. Humbel (59), Electron Microscopy Facility, University of Lausanne,
Biophore, Lausanne, Switzerland
M. Huynh (1), Australian Centre for Microscopy & Microanalysis, The University of
Sydney, NSW, Australia
Erik M. Jorgensen (283), Howard Hughes Medical Institute and Department of
Biology, University of Utah, Salt Lake City, UT
Marko Kaksonen (235), Cell Biology and Biophysics Unit European Molecular
Biology Laboratory, Meyerhofstr. 1, Heidelberg, Germany
Ruby Kalicharan (157), Department of Cell Biology, University Medical Center
Contributors xiii
K. Kobayashi (1), Australian Centre for Microscopy & Microanalysis, The Univer-
sity of Sydney, NSW, Australia
Irina Kolotuev (203), Institut de Génétique et Développement de Rennes, UMR
6290 CNRS, Université Rennes 1, Faculté de Medecine/1. Fédération de Recherche
BIOSIT, UMS 3480 Université de Rennes 1, Campus santé. 2 avenue du Pr Leon
Bernard CS34317, Rennes Cedex, France
Susanne Kretschmar (75), Center for Regenerative Therapies, TU Dresden,
Fetscherstraße 105, Dresden, Saxony, Germany
Jeroen Kuipers (157), Department of Cell Biology, University Medical Center Gron-
ingen (UMCG), University of Groningen, A. Deusinglaan 1, Bldg 3215, room 749,
AV Groningen, The Netherlands
Wanda Kukulski (235), Structural and Computational Biology Unit, European
Molecular Biology Laboratory, Meyerhofstr. 1, Heidelberg, Germany; Cell Biology
and Biophysics Unit European Molecular Biology Laboratory, Meyerhofstr. 1,
Heidelberg, Germany
Thomas Kurth (75), Center for Regenerative Therapies, TU Dresden, Fetscherstraße
Michel Labouesse (203), Institut de Génétique et de Biologie Cellulaire et Moléculaire,
1 rue Laurent Fries, Illkirch Cedex, France
Céline Loussert (59), Electron Microscopy Facility, University of Lausanne, Bio-
phore, Lausanne, Switzerland
Falk Lucas (325), Electron Microscopy ETH Zurich – EMEZ, ETH Zurich, Switzerland
Miriam Lucas (325), Electron Microscopy ETH Zurich – EMEZ, ETH Zurich,
Switzerland
Alberto Luini (21), Telethon Institute of Genetics and Medicine, Naples, Italy; Insti-
tute of Protein Biochemistry, Naples, Italy
Judith Mantell (175), Department of Biochemistry, School of Medical Sciences,
University of Bristol, University Walk, Bristol, UK; Wolfson Bioimaging Facility,
School of Medical Sciences, University Walk, Bristol, UK
Giovanni Mariggi (357), Vascular Biology Laboratory, London Research Institute,
Cancer Research UK, London, UK
Ian E.G. Morrison (307), Technology Facility, Biology Department, University of
York, York, UK
Thomas Müller-Reichert (223), Medical Theoretical Center (MTZ), Medical
Faculty Carl Gustav Carus, University of Technology Dresden, Fiedlerstraße 42,
Dresden, Germany
Hidetoshi Nishiyama (307), JEOL Ltd., Advanced Technology Division, Akishima,
Tokyo, Japan
Peter J. O’Toole (307), Technology Facility, Biology Department, University of
York, York, UK
Andrea Picco (235), Cell Biology and Biophysics Unit European Molecular Biology
Laboratory, Meyerhofstr. 1, Heidelberg, Germany
Jürgen M. Plitzko (259), Department of Molecular Structural Biology, Max Planck
xiv Contributors
Elena V. Polishchuk (21), Institute of Protein Biochemistry, Naples, Italy

Roman S. Polishchuk (21), Telethon Institute of Genetics and Medicine, Naples,
Italy
K.R. Ratinac (1), Australian Centre for Microscopy & Microanalysis, The Univer-
sity of Sydney, NSW, Australia
Alexander Rigort (259), Department of Molecular Structural Biology, Max Planck
John M. Robinson (37), Department of Physiology and Cell Biology, Ohio State
University, Columbus, OH, USA
Chikara Sato (307), Biomedical Research Institute, National Institute of Advanced
Industrial Science and Technology, Tsukuba, Ibaraki, Japan
Ulrike Schnell (157), Department of Cell Biology, University Medical Center Gron-
ingen (UMCG), University of Groningen, A. Deusinglaan 1, Bldg 3215, room 749,
AV Groningen, The Netherlands
Martin Schorb (235), Structural and Computational Biology Unit, European Molec-
ular Biology Laboratory, Meyerhofstr. 1, Heidelberg, Germany
Yannick Schwab (203), Institut de Génétique et de Biologie Cellulaire et Molécu-
laire, 1 rue Laurent Fries, Illkirch Cedex, France
Thom Sharp (175), Department of Biochemistry, School of Medical Sciences, Uni-
versity of Bristol, University Walk, Bristol, UK
Xiaokun Shu (139), Department of Pharmaceutical Chemistry, University of Califor-
nia, San Francisco, CA, USA
Klaas A. Sjollema (157), Department of Cell Biology, University Medical Center
Gina E. Sosinsky (139), National Center for Microscopy and Imaging Research,
University of California, San Diego, La Jolla, CA, USA; Department of Neurosci-
ences, University of California, San Diego, La Jolla, CA, USA
Martin Srayko (223), Department of Biological Sciences, University of Alberta,
Edmonton, Canada
Mitsuo Suga (307), JEOL Ltd., Advanced Technology Division, Akishima, Tokyo,
Japan
Carlo Tacchetti (95), MicroScoBio Research Center, Department of Experimental
Genoa, Italy; IFOM, Fondazione Istituto FIRC di Oncologia Molecolare, Milan,
Italy; Experimental Imaging Research Center, Scientific Institute San Raffaele,
Milan, Italy
Toshihiro Takizawa (37), Department of Molecular Anatomy, Nippon Medical
School, Tokyo, Japan
P. Thordarson (1), School of Chemistry, The University of New South Wales, NSW,
Australia
Jan van Weering (175), Department of Biochemistry, School of Medical Sciences,
University of Bristol, University Walk, Bristol, UK
Contributors xv
Paul Verkade (175), Department of Biochemistry, School of Medical Sciences,

University of Bristol, University Walk, Bristol, UK; Wolfson Bioimaging Facility,
School of Medical Sciences, University Walk, Bristol, UK; Department of Physiol-
ogy and Pharmacology, School of Medical Sciences, University Walk, Bristol, UK
Giuseppe Vicidomini (95), Nanophysics, Istituto Italiano di Tecnologia Genoa, Italy
Elizabeth Villa (259), Department of Molecular Structural Biology, Max Planck
Shigeki Watanabe (283), Howard Hughes Medical Institute and Department of Biol-
ogy, University of Utah, Salt Lake City, UT
Susanne Weiche (75), Biotechnology Center, TU Dresden, Tatzberg 47-49, Dresden,
Saxony, Germany
Sonja Welsch (235), Structural and Computational Biology Unit, European Molecu-
lar Biology Laboratory, Meyerhofstr. 1, Heidelberg, Germany
Roger Wepf (325), Electron Microscopy ETH Zurich – EMEZ, ETH Zurich, Swit-
zerland
Silke White (223), Medical Theoretical Center (MTZ), Medical Faculty Carl Gustav
Carus, University of Technology Dresden, Fiedlerstraße 42, Dresden, Germany
Ina Woog (223), Medical Theoretical Center (MTZ), Medical Faculty Carl Gustav
Carus, University of Technology Dresden, Fiedlerstraße 42, Dresden, Germany
Andrew Yarwood (307), JEOL UK Ltd., Jeol House, Watchmead, Welwyn Garden
City, Herts, UK
PREFACE
Introduction to Correlative Light and Electron Microscopy
Correlative microscopy, as an approach to combine methods of different modalities,

has become increasingly important over the past years. With the advent of GFP and the
accompanying improvements in light as well as in electron microscopy technology, it
is the combination of both techniques in Correlative Light and Electron Microscopy
(CLEM) that has generated the most attention. The rise in the importance of CLEM
is reflected by a steady increase in publications related to the fusion of these imaging
techniques. In 2002 approximately 10 studies were published that employed CLEM,
whereas over 60 were published in 2011. Today, the term CLEM is applied to a number
of approaches, all having in common that the imaging and analysis of the same sample
employs both methods. Earlier studies, however, did not apply this combination of
methods in the strict sense of using the same specimen, but these approaches are still
very valuable as highlighted in some chapters.
The oldest set of papers on “correlative microscopy” that can be found in PubMed
was published in 1960 (Godman et al., 1960a and b). In these studies light level his-
tochemical staining patterns, published in the first paper, were correlated with elec-
tron microscopic observations and presented in the accompanying second. Again, the
authors did not examine the same structures with both types of microscopes. The earli-
est example of CLEM on the same specimen was published 15 years later (Abandowitz
and Geissinger, 1975). Interestingly, the sample was first imaged by scanning electron
microscopy, followed then by light microscopy interferometry to ascertain dry mass.
The approach of examining the same samples and/or objects has been carried forward
to the present day in a number of variations. For a more complete historical perspective
the reader is referred to the reference lists of the individual chapters in this volume as
well as those in the earlier book by Hayat (1987).
One of the most common approaches of CLEM involves the use of fluorescence
light followed by electron microscopy on the same cells. Webster et al., (1978) exam-
ined microtubules using immunofluorescence microscopy and subsequent transmission
electron microscopy in whole mount, detergent extracted PtK2 cells. Another early
example of CLEM on single specimens utilized correlative immunofluorescence and
electron microscopy on Epon sections (Rieder and Bowser, 1985). In this study, virally
infected AC-20 cells were labeled with antibodies prior to Epon embedding and then
sections were cut for subsequent fluorescence, phase contrast, and electron microscopy.
Imaging of the same exact structures by fluorescence and electron microscopy was
further explored later when the enzyme myeloperoxidase was detected via immuno-
fluorescence using the bi-functional reagent FluoroNanogold (Takizawa et al., 1998).
Another important advancement in the field of CLEM involved combinating live GFP
imaging with subsequent electron microscopy of transport carriers inside the cell
xvii
xviii Introduction to Correlative Light and Electron Microscopy
(Polishchuk et al., 2000). Since then a number of groups have succeeded in adopting
this strict application of CLEM.
Additionally, the use of alternating semi-thick and thin sections for parallel LM and
EM analysis was developed further by Schwarz and colleagues (Schwarz and Humbel,
2007). Ultrathin cryosections were collected on Formvar-coated EM grids. Fluores-
cence images were generated and then the section was subjected to a silver enhance-
ment reaction for EM-level Nanogold visualization. Later, the thin-section approach
was modified by embedding tissue in LR White resin. Immunofluorescence labeling
was carried out directly on collected thin sections, which could then be examined by
scanning electron microscopy to provide ultrastructural detail (Micheva and Smith,
2007).
The motivation to perform CLEM can be compared when one contrasts the pros and
cons of previous versus current approaches. Independent from the methodologies of
these specific techniques, one wishes to bridge the gaps, thus merging the advantages
of both microscopy ‘worlds’ to optimize the information gained. The combination
should ideally enhance both quantity and quality of information over applying either
technique separately.
Alternatively, rather than overlaying fluorescence signal on an EM image of the same
sample post fixation/embedding, many current CLEM approaches try to capture the
dynamics of cells by live-cell light microscopy and then process the sample for ultra-
strucural analysis. As previously summarized (McDonald, 2009), the rational here is:
I, to combine contextual information from light microscopy with the resolution of EM;
II, to increase EM sample size and throughput;
III, to locate a rare event and/or structure; and
IV, to observe a dynamic process within a known region of interest.
Along these lines, we have employed CLEM to visualize specific processes during
intracellular traffic (Verkade, 2008; van Weering et al., 2010), or to analyze intermedi-
ate stages of both centriole duplication and cell division (Pelletier et al., 2006; Guizetti
et al., 2011). Our own research motivated us to present a number of different CLEM
approaches within a single volume of Methods in Cell Biology.
Clearly, not all current CLEM approaches could be described within this volume,
however, we attempted to discuss the most applicable and interesting combinations
of techniques/imaging modes. These approaches include the following: the use of
either plastic and/or Tokuyasu cryo-sections for CLEM (Kobayashi et al., Polishchuk
et al., Takizawa and Robinson, Lousset et al., Fabig et al., Cortese et al.); the appli-
cation of multifunctional marker molecules for both light and electron microscopy
(Grabenbauer, Ellisman et al., Sjollema et al.,); the use of cryo-fixation as a
starting point for CLEM (Brown et al., Kolotuev et al., Woog et al., Kukulski
et al.); the combination of high-end light and/or electron microscopy, such as
FIB-SEM for specimen preparation (Rigort et al., Lucas et al.) or imaging (Bushby et
al); the advantages of super-resolution light microscopy for structural studies (Wata-
nabe and Jorgensen); and lastly, the integration of light and electron microscopy into
one instrument (Morrison et al.).
Introduction to Correlative Light and Electron Microscopy xix
In parallel to this volume of Methods in Cell Biology, we have designed an annual

practical EMBO course on CLEM, where students can gain hands-on experience in
applying combinatorial approaches to their specific research questions. It is our hope
that this practical course, in concert with this MCB volume, will stimulate further
crosstalk between light and electron microscopy leading to the development of new
correlative approaches in cell biology.
Thomas Müller-Reichert, Paul Verkade

Dresden/Bristol
March 31st, 2012
Acknowledgments
The authors would like to thank Shaun Gamble (Elsevier) for his help in bringing
this volume of Methods in Cell Biology to completion.
References
Abandowitz, H. M., & Geissinger, H. D. (1975). Preparation of cells from suspensions for correlative scan-
ning electron and interference microscopy. Histochemistry, 45, 89–94.
Godman, G. C., Morgan, C., Breitenfeld, P. M., & Rose, H. M. (1960). A correlative study by electron
and light microscopy of the development of type 5 adenovirus. II. Light microscopy. J Exp Med, 112,
383–402.
Guizetti, J., Schermelleh, L., Mantler, J., Maar, S., Poser, I., Leonhardt, H., Muller-Reichert, T., & Gerlich,
D. W. (2011). Cortical constriction during abscission involves helices of ESCRT-III-dependent filaments.
Science, 331, 1616–1620.
Hayat, I. (1987). Correlative Microscopy in Biology. London: Academic Press.
McDonald, K. L. (2009). A review of high-pressure freezing preparation techniques for correlative light and
electron microscopy of the same cells and tissues. J Microsc, 235, 273–281.
Micheva, K. D., & Smith, S. J. (2007). Array tomography: a new tool for imaging the molecular architecture
and ultrastructure of neural circuits. Neuron, 55, 25–36.
Pelletier, L., O’Toole, E., Schwager, A., Hyman, A. A., & Muller-Reichert, T. (2006). Centriole assembly in
Caenorhabditis elegans. Nature, 444, 619–623.
Polishchuk, R. S., Polishchuk, E. V., Marra, P., Alberti, S., Buccione, R., Luini, A., & Mironov, A. A. (2000).
Correlative light-electron microscopy reveals the tubular-saccular ultrastructure of carriers operating
between Golgi apparatus and plasma membrane. J Cell Biol, 148, 45–58.
Rieder, C. L., & Bowser, S. S. (1985). Correlative immunofluorescence and electron microscopy on the same
section of epon-embedded material. J Histochem Cytochem, 33, 165–171.
Schwarz, H., & Humbel, B. M. (2007). Correlative light and electron microscopy using immunolabeled resin
sections. Methods Mol Biol, 369, 229–256.
van Weering, J. R., Brown, E., Sharp, T. H., Mantell, J., Cullen, P. J., & Verkade, P. (2010). Intracellular
membrane traffic at high resolution. Methods Cell Biol, 96, 619–648.
Verkade, P. (2008). Moving EM: The Rapid Transfer System as a New Tool for Correlative Light and
Electron Microscopy and High Throughput for High-Pressure Freezing. Journal of Microscopy, 230,
317–328.
Webster, R. E., Osborn, M., & Weber, K. (1978). Visualization of the same PtK2 cytoskeletons by both
immunofluorescence and low power electron microscopy. Exp Cell Res, 117, 47–61.
CHAPTER 1
Imaging Fluorescently Labeled Complexes

by Means of Multidimensional Correlative
Light and Transmission Electron
Microscopy: Practical Considerations
K. Kobayashi*, 1, D. Cheng†, M. Huynh*, K.R. Ratinac*,
P. Thordarson‡ and F. Braet*, †
* Australian Centre for Microscopy & Microanalysis, The University of Sydney, NSW 2006, Australia
† Schoolof Medical Sciences (Discipline of Anatomy and Histology)—The Bosch Institute, The University of
Sydney, NSW 2006, Australia
‡ School of Chemistry, The University of New South Wales, NSW 2052, Australia
Abstract
I. Introduction
A. Need for Correlative Microscopy
B. Determining the Effects of Drugs on Cells: an Application for CLEM
II. Rationale
III. Methods
A. Cell Culture for Correlative Light and Electron Microscopy Imaging
B. Live-Cell Microscopy and Confocal Laser Imaging
C. Sample Preparation for Relocation Studies by Electron Microscopy
D. 2-D and 3-D Transmission Electron Microscopy
E. Data Processing and Correlative Analysis
IV. Instrumentation and Materials
1 Present
address: Department of Applied Biology and Chemistry, Faculty of Applied Biosciences,
Tokyo University of Agriculture, Tokyo, Japan.
METHODS IN CELL BIOLOGY, VOL 111 0091-679X

Copyright 2012, Elsevier Inc. All rights reserved. 1 http://dx.doi.org/10.1016/B978-0-12-416026-2.00001-7
2 K. Kobayashi et al.
V. Discussion
A. Making the Most of This Approach
B. A 4-Dimensional Future
C. Conclusion
Acknowledgments
References
Abstract
These days the common ground between structural biology and molecular biology
continues to grow thanks to the biomolecular insights offered by correlative micros-
copy, even though the vision of combining insights from different imaging tools has
been around for nearly four decades. The use of correlative imaging methods to dissect
the cell’s internal structure is progressing faster than ever as shown by the boom in the
number of methodological approaches available for correlative microscopy studies,
each designed to address a specific scientific question. In this chapter, we will pres-
ent a relatively straightforward approach to combining information from fluorescence
microscopy and electron microscopy at the supramolecular level. The method com-
bines live-cell and/or confocal laser microscopy with classical sample preparation for
transmission electron microscopy (TEM), thereby allowing the integration of dynamic
details of subcellular processes with insights about the organelles and molecular
machinery involved. We illustrate the applicability of this multidimensional correlative
microscopy approach on cultured Caco-2 colorectal cancer cells exposed to fluores-
cently labeled cisplatin, and discuss how these methods can deepen our understanding
of key cellular processes, such as drug uptake and cell fate.
I. Introduction
A. Need for Correlative Microscopy

Different types of microscopes have different sample requirements, spatial resolu-
tion limits, contrast mechanisms, and associated analytical techniques. For example,
light- and laser-based microscopy techniques have great flexibility in accommodating
specimens under ambient conditions. Thus, one thinks it is trivial to observe cellular
and subcellular processes over minutes or hours, or even days, with a suitable “live-cell
imaging” system. This kind of dynamic imaging, of course, can provide rich detail on
fundamental processes in biological systems, such as cellular proliferation and death or
cell migration. Live-cell imaging is especially powerful when coupled with expression
of fluorophores within the cells, or with other fluorescent labeling methods, to provide
dynamic contrast and help to localize the events of interest to a specific subpopula-
tion of cells or within specific organelles (Giepmans, Adams, Ellisman, & Tsien, 2006;
Tsien, 2003). Despite this power and versatility, which have resulted in this large fam-
ily of techniques becoming the “workhorse” of biological and biomedical laboratories
worldwide, light- and laser-based microscopy techniques are generally limited in spatial
1. CLEM of Fluorescently Labelled Drug Complexes 3
resolution to approximately 0.2 µm, due to the wavelength of light (see the right end of
the scale bar in Fig. 1(A)). Recent developments in super-resolution optical microscopy
have found ways to get around this diffraction limit, and likely will continue to increase
the level of spatially resolved detail that can be obtained under dynamic conditions
(Jones, Shim, He, & Zhuang, 2011; Schermelleh, Heintzmann, & Leonhardt, 2010). Yet
these techniques still leave uncertainty as to the cells’ ultrastructure.
When a far higher resolution is needed, such as when one wants to see supramo-
lecular detail, transmission electron microscopy (TEM) becomes the method of choice.
By virtue of using high-energy electrons, which have an extremely short wavelength,
TEM allows ready observation of molecular detail in biological systems, easily achiev-
ing spatial resolutions down to 1–3 nm (see the left end of the scale bar in Fig. 1(A)).
It should be noted, however, that the final achievable resolution depends critically on
the sample preparation technique used and, to a lesser extent, on the type of TEM
used (Massover, 2011; Sander & Golas, 2011). The use of electrons requires that new
contrast mechanisms be considered when preparing samples; it also provides access to
various analytical techniques, such as elemental analysis via energy-dispersive X-ray
spectroscopy, energy-filtered imaging, or spectral mapping with electron energy loss
spectroscopy (Müller, Aebi, & Engel, 2008). However, the downside of these tech-
niques is that the samples must be extremely thin (on the order of the resolution limit
for light microscopy) to achieve electron transparency and need to be stable under high-
vacuum conditions, necessitating fixing or freezing of living tissues. Consequently,
what TEM makes up for in spatial resolution, it lacks in observing dynamic processes
in real-time, being restricted to taking a series of high-resolution “snapshots” of key
stages in biological processes. Ongoing developments in “environmental” or “4D”
electron microscopy may eventually open up opportunities for obtaining more details
on dynamics (Zewail, 2010), but routine TEM will always be limited to “snapshots.”
Evidently, neither of these two microscopy approaches can provide all the informa-
tion that a cell biologist might desire; yet, to fully understand how complex biologi-
cal structures function, we must integrate knowledge of their dynamic behavior and
ultrastructure. Recognition of these limitations led researchers during the latter part
of the last century to suggest and eventually develop correlative light and electron
microscopy (CLEM). The principle of CLEM is to obtain complementary informa-
tion about the same sample by means of two or more distinct types of microscopy and
then merge those distinct streams of information together to obtain a far richer depth
of insight into biological processes than is possible from a single mode of microscopy
alone (Geissing, 1974; Rieder & Bowser, 1985; Simmons, Pawley, & Albrecht, 1990;
Watari & Herman, 1965). Because of light microscopy’s strengths in dynamic imag-
ing of biological processes and electron microscopy’s ability to resolve the molecular
machinery that drives those processes, their combined use has become increasingly
important for our understanding of the structure and function of cells and tissues at the
molecular level (Fig. 1(A)) (McDonald, 2009; Pelletier, O’Toole, Schwager, Hyman,
& Muller-Reichert, 2006). In this chapter, we show how CLEM can disclose the final
subcellular destination of a pharmaceutical and give insights into the biological effects
of such compounds.
Fig. 1 Scheme illustrating the concept of correlative multidimensional light (3-D) or confocal laser optical imaging (4-D) with electron
microscopy (2-D or 3-D) imaging. (A) Depending on the sample source (e.g., tissue slices, organoid-like cell systems, cell cultures) and
the advanced sample preparation approaches applied (i.e., chemical or physical fixation), researchers can collect dynamic structural and
molecular information on the cells’ behavior down to the supramolecular level. High-end multidimensional live-cell imaging systems
also allow the collection of large datasets over multiple positions (i.e., different cells within the same experiment), thereby increasing
the statistical rigor of the observations. Full CLEM information is obtained when multidimensional real-time optical imaging data
(left, green) are integrated with electron microscopy imaging (right, blue), via advanced sample preparation and relocation approaches
(middle, gray box). [Reprinted and modified with permission from (Jahn et al., 2012) and (Su, Nykanen, et al., 2010).] (B) Schematic
drawing of the two mainstream labeling approaches presently available to perform CLEM studies. A choice must be made at the start of
the experimental design as to whether the same markers or labels will be visually identified across the different microscopy platforms
(i.e., combinatorial labeling, top), or whether different types of markers or labels will be used for each imaging platform (i.e., noncom-
binatorial labeling, bottom). In the combinatorial approach, the probe is followed at both the light microscopy and electron microscopy
levels, but often this requires the use of postlabeling procedures. Such methods typically involve enhancement of one molecular label
(e.g., GFP-photo-oxidation) after the dynamic experiments are completed to allow the detection of the enhanced label by TEM. In
contrast, noncombinatorial labels are only used during the light or laser microscopy, after which samples are prepared directly for sub
sequent electron microscopy studies. The ability to use double- or even multiple-labeling approaches for one or both types of micros-
copy is the major advantage of the noncombinatorial approach. For more information on the available correlative-labeling approaches,
we refer interested readers to expert review papers (Giepmans, 2008; Robinson, 2001).
B. Determining the Effects of Drugs on Cells: an Application for CLEM

Accelerated development of potential pharmaceutical compounds is driving demand
for high-throughput screening approaches that can assess the effects of newly designed
drugs. In addition, there is a further push to determine precisely how such compounds
affect cell state and fate over time at high spatial resolution (Hsieh, Huang, Wu, Shieh, &
Lee, 2009). Of course, high-throughput biomolecular screening assays are readily avail-
able to the scientist to monitor cell behavior with time (Sharma, Ando, Daub, Kaye, &
Finkbeiner, 2012; Szymański, Markowicz, & Mikiciuk-Olasik, 2012); these often form
the entry point for pharmacological studies to assess overall cellular effects of new drugs
and/or to gain insight into the affinity of the new compound for its subcellular target.
Even though this screening approach allows one to capture multichannel fluorescence
images on a large cell population, so that different cellular events can be monitored
with high statistical accuracy, it frequently has limited ability to resolve fine subcel-
lular details. To address this shortfall, specific quantitative microscopy solutions have
been developed recently that can image large numbers of cells in multiple dimensions
at the best spatial resolution possible with optical microscopy (Conrad et al., 2011).
Although this high-throughput, quantitative microscopy approach is exceptionally
well suited to studying dynamic processes in living cells, the link to three-dimensional
(3-D) (sub-)nanometer information is missing.
As briefly introduced above, CLEM can potentially deliver this information because
of its proven ability to spatially resolve structure–function detail at the molecular level
(Mironov & Beznoussenko, 2009; Polishchuk & Mironov, 2001). Initially, the success
of original CLEM approaches was critically dependent on the hard work and persis-
tence of the individual to relocate the region of interest, thereby usually limiting it to
cross-correlating a single event in a given experiment. Currently, CLEM approaches are
available that allow one to investigate the structure–function relationships within mul-
tiple cells within a single experiment; for reviews see Jahn et al., 2012; Su, Nykanen,
et al., 2010. The CLEM approach described in this chapter has arisen from the com-
bination of our past experience in this field, the latest advances in optical technologies
and imaging instruments, and the improvements in software for high-speed image pro-
cessing and data analysis. Together, these recent advances in CLEM methodology can
finally meet the need for screening large numbers of cells, for statistical rigor, while
revealing the dynamic effects, and alterations to fine structure, caused by pharmaco-
logical compounds. Elegantly illustrating the potential of this multicell approach is the
recent study by Guizetti et al. (Guizetti, Mäntler, Müller-Reichert, & Gerlich, 2010),
which combined correlative time-lapse fluorescence microscopy with subsequent serial-
section electron microscopy to screen large number of HeLa cells during cell division.
II. Rationale
The approach that we outline in this chapter aims to fill a void in the current arsenal
of CLEM methods, namely finding a method that can achieve quality outcomes at rela-
tively low cost and with a minimum of consumables and effort in sample preparation,
thereby allowing imaging across multiple cells in reduced time. Our framework offers
an attractive alternative for anyone who wants to correlate light or fluorescence informa-
tion with TEM data (Fig. 1(A)), via noncombinatorial labeling (Fig. 1(B)), which uses
classical sample preparation methods for time-lapse live-cell microscopy (Su, Whan,
Empsen, Soon, & Braet, 2010) and electron microscopy investigation of cells in culture
(Biazik, Jahn, Su, Wu, & Braet, 2010). We will illustrate the ease of this method on the
well-established Caco-2 colorectal cancer cell line (Jahn, Biazik, & Braet, 2011) and
demonstrate the usefulness of CLEM in monitoring the uptake and effects of cisplatin on
cellular structure and function over length scales and multiple dimensions (Fig. 2(A)).
We present our method as two routes, depending on the availability of instruments in
a given microscopy center, that use essentially the same sample preparation steps to
Fig. 2 Schematic overview of the CLEM approach. (A) Diagram of the general experimental design in which we apply the two
c orrelative routes (B or C) to monitor fluorescent compounds. In this setup, cancer cells are exposed to cisplatin tagged with
FITC (green) and to LysoTracker (red), and monitored with the microscopy platforms outlined under B and C. Both the cross-correlative
imagingroutes are based on classical sample preparation procedures for real-time data collection from a light or laser microscope and
then fixation, ultramicrotomy, and staining for subsequent TEM investigation. (B) Cross-correlative 3-D live-cell imaging followed by
2-D TEM. (C) Combination of data from 4-D confocal laser imaging with data from 3-D TEM, where t0 is the start of the experiment
and tc is the “critical time” (or the end point) of the light or laser data-capture process.
cross-correlate structure–function information. The first route is based on 3-D live-cell

(X, Y, and t) imaging followed by 2-D TEM (X and Y) studies (Fig. 2(B)). The second,
and more preferable, route combines 4-D multiphoton imaging (X, Y, Z, and t) with
3-D TEM (X, Y, and Z), but requires a larger investment of time as well as higher
end instrumentation (Fig. 2(C)). Irrespective of which route one chooses, both CLEM
approaches allow one to glimpse the cells’ interior in real-time and then subsequently
integrate ultrastructural detail.
III. Methods

Caco-2 cells obtained from the American Type Culture Collection (ATCC) were
c ultured in 75 cm2 tissue-culture flasks maintained in 10 ml of complete cell-culture
media comprising advanced Dulbecco’s Modified Eagle Medium (DMEM), fetal bovine
serum (FBS, 10%), L-glutamine (2.5 mM), and antibiotic/antimycotic solution (100 U/ml
penicillin, 100 µg/ml streptomycin, and 25 µg/ml amphotericin B). Cells were incubated
at 37°C under CO2 (5% in air) and subsequently subcultured until 70% confluency was
reached (Jahn et al., 2011). For CLEM experiments, semiconfluent cell monolayers were
trypsinized for 5 min at 37°C with a trypsin–EDTA (ethylenediaminetetraacetic acid) solu-
tion and then centrifuged at 2,000 rpm for 4 min. Next, after extensive washing in complete
medium, cells were seeded in carbon-coated MatTek glass-bottom cell culture dishes (see
below) at a dilution of 1:5 and cultured for at least 24 h before live-cell experimentation.
Prior to seeding the cells into the cell culture dishes, we created a carbon-based grid
pattern on the specimen surface of the dishes for ease of relocating specific cells. The
markers were produced according to the method of Paul Verkade (Brown, Mantell,
Carter, Tilly, & Verkade, 2009), and the carbon coating must be thick enough so that
the finder grid pattern is clearly visible to the naked eye (Fig. 3(A)). Briefly, the method
involves placing copper finder grids (we used two) on the base of a 35 mm glass-
bottom MatTek dish and sputter coating the dish to achieve a ~45 nm layer of carbon.
Next, the copper finder grids were removed and the dishes were rinsed thoroughly with
ethanol to remove unwanted carbon deposits. The dishes were subsequently allowed
to dry in an oven at 60°C for 48 h. This drying step is crucial to retain the grid pattern
throughout the successive sample preparation procedures. Prior to plating the cells, the
carbon-coated dishes were sterilized with UV light for 20 min and then complete cell
culture medium at 37°C was placed into the dishes to equilibrate for at least 2 h before
the cells were seeded as described above. Note that this equilibration step is essential,
as it primes the dishes for optimal cell attachment and subsequent cell culture.

It is imperative for live-cell imaging that all components of the microscope and the
sample should be fully equilibrated and accurately controlled at physiological condi-
tions, otherwise there will be a decrease in the overall cell viability as well as a continual
A B
C D
E F
Fig. 3 Practical steps involved in the relocation of cells of interest for cross-correlative microscopy. The
method depends on a carbon-coated pattern. (A) Two carbon-based finder grid imprints with a thickness of
~45 nm created on a MatTek dish with a standard carbon-coater machine. This results in a clearly visible
pattern when imaged under the light microscope (inset). (B) Combined DIC and fluorescence image of
Caco-2 cells cultured on top of this carbon footprint. The green fluorescence corresponds to cisplatin–FITC,
while the red fluorescence represents the LysoTracker probe after 150 min of incubation. Note the carbon
footprint is clearly visible (i.e., the light-colored grid pattern in the background). (C) Sample after two-step
embedding (see text). (D) A capsule after manual detachment from the MatTek dish. (E) Specimen surface
of detached sample. An intact carbon footprint, as previously seen on the MatTek dish, remains visible as
on the surface of the resin block (inset). (F) Light microscopy image of a semithin resin section stained with
toluidine blue, revealing the corresponding area as shown in B. The white arrows point out a dividing cell
that contains both fluorescent probes. Typically, ultrathin sections were then retrieved for further TEM inves-
tigations and cross-correlative analysis.
thermal drift in the Z-direction (and to a lesser degree in the X- and Y-directions) of
the motorized stage. Thus, to allow thermal equilibration of the cell culture dishes,
we transferred the dishes containing the cells to the sample stage of the microscope
and allowed them to equilibrate for at least 30 min before the fluorescent-labeled drug
complexes were added to the dish.
After equilibration, and to enable fluorescence studies, the Caco-2 colorectal cancer
cells were treated for different times with cisplatin (25 µM) conjugated to fluores-
cein isothiocyanate (FITC) (Molenaar, Teuben, Heetebrij, Tanke, & Reedijk, 2000);
we term this “cisplatin–FITC.” Furthermore, in order to simultaneously follow the
dynamic behavior of lysosomes, LysoTracker Red DND-99 was added at a final con-
centration of 50 nM (Lemieux, Percival, & Falgueyret, 2004). The rationale for the
use of both probes is the strong indication that a defective lysosomal apparatus may
cause acquired cisplatin-resistance in malignant cells (Chauhan et al., 2003). After
30 min incubation with both fluorescent compounds and before live-cell light and/or
confocal laser imaging were started, the dishes were carefully rinsed with prewarmed
(37°C) phosphate-buffered saline (PBS) to ensure removal of unbound cisplatin–FITC
and LysoTracker probe. Unbound fluorescent compounds adversely affect the overall
image quality, so careful washing is important to collect high-quality light or laser data
for subsequent cross-correlative analysis at the electron microscope level. Immediately
thereafter, fresh medium (37°C) was added and live-cell imaging was started right
away (Fig. 4(A–H)). Live-cell experiments of the fluorescently tagged cells were per-
formed on an Olympus CellR live-cell microscope. For each experiment, a glass lid was
placed over the MatTek dish to allow local delivery of CO2 (5% premixed in air) to the
cells, and still enable optimal differential interference contrast (DIC) imaging of the
cells grown on the relocation marks (Fig. 3). Note that, throughout imaging, we used
advanced DMEM without phenol-red indicator to avoid phototoxic effects that can
decrease the overall cell viability (see Section V).
To begin, we acquired overview images for each finder grid area with a 10× objective
(UPLAN Super Apochromat NA 0.3) by using the multiple image alignment (MIA) soft-
ware package to record the areas of interest (Fig. 3(A)). Typically, up to eight different
stage positions per finder grid were randomly selected by using a 40× objective (UPLAN
Super Apochromat NA 0.9; Fig. 3(B)). We collected time-lapse DIC and fluorescence
images every 5 min for up to 3 h, detecting the cisplatin–FITC and LysoTracker red sig-
nals by using filters sets for FITC (Ex 492/18, Em 510–550) and for Texas red (TXRED)
(Ex 572/23, Em 595–700), respectively. This generated 3-D data of X and Y over time (t).
At the end of imaging, the cells, still attached to the culture dish, were rinsed five times
with cacodylate buffer (0.1 M sodium cacodylate, 0.1 M sucrose at pH 7.4) and imme-
diately fixed with 2% glutaraldehyde in cacodylate buffer for 1 h at room temperature.
To acquire 3-D spatial data (i.e., “Z-stacks”) after live-cell imaging, we carefully
transferred the glutaraldehyde-fixed samples to a confocal laser microscope (Olympus
FV1000). Each position previously imaged on the live-cell microscope was relocated
with the aid of the carbon fiducial marks and then Z-stacks of the previously imaged
regions were recorded by optically sectioning through the cells (Fig. 5(B–D)). In gen-
eral, it is possible to image the cells on a confocal imaging platform that is set up for
Fig. 4 Correlative imaging with 3-D live-cell microscopy (A–H) and 2-D TEM (J–L). In this experiment, Caco-2 colorectal cells were
incubated with cisplatin–FITC (green fluorescence) and LysoTracker (red fluorescence). (A–H) The dynamic behavior of the fluores-
cently labeled compounds was monitored by using live-cell time-lapse recording for up to 150 min. It is quite evident that cisplatin
is dispersed in discrete patches over the cytoplasm (small arrowheads) during the early stages of uptake, whereas distinct cisplatin
accumulation can be seen (large arrowheads) toward the end of the experiment. Note that E is a DIC image at the end of the experiment,
and F and G are the corresponding fluorescence images, while H is the combined DIC and laser image (i.e., E + F + G). (I) shows the
corresponding area of interest of a toluidine-blue-stained semithick section; this area was then relocated (J) and cross-correlated (K) at
the TEM level. Cross-correlative analysis reveals that both fluorescent compounds end up in different subcellular compartments after
two-and-a-half hours of treatment. Lysosomes (arrow) are apparent within the perinuclear area (pN) while the anti-cancer drug (arrow-
head) is largely associated with the nuclei (N). This experiment illustrates how CLEM can contribute to our understanding of the way
that platinum-based anticancer drugs are taken up by cells, and the intracelullar routes that might explain multidrug resistance or drug
sensitivity (Stewart, 2007; Wang & Lippard, 2005). The corresponding insets in J–L show the perinuclear area, as it is known that the
nuclear compartment is the end destination of cisplatin. (L) provides another example of cross-correlative analysis at a different position
within the same experiment. Note the extensive membrane blebbing (arrow) that is typical for apoptosis. The lysosomal marker (red
fluorescence) reveals a strong association with these subcellular structures that appear during programmed cell-death. Scale bars in A–I
are 50 µm and in J–L are 5 µm.
live-cell imaging from the start of the addition of the fluorescent compounds (see Fig.
2(C)), giving 4-D optical data (X, Y, Z, and t; data not shown).

After laser data had been collected on fixed samples, the glutaraldehyde-fixative
solution was removed from the dishes and cells were briefly washed twice with sodium
cacodylate buffer for 5 min at each wash. Subsequently, cells were subjected to post-
fixation with 1% osmium tetroxide for 1 h at 4°C in the dark and then immediately
Fig. 5 Combining live-cell imaging (A–C) with 3-D electron microscopy (G–J). In this experiment, Caco-2
cells were exposed to cisplatin–FITC (green fluorescence) and LysoTracker (red fluorescence). At the end of
the time-lapse imaging (150 min), cells were fixed and then analyzed by confocal laser microscopy (D) A: DIC;
B: cisplatin–FITC; C: LysoTracker; D: combined Y–Z-fluorescent information. The white circles in A–D denote
the subcellular area of interest to be further analyzed with TEM (E, arrowhead). Note that this intermediate
electron-dense structure is located in the close vicinity of a nuclear invagination (arrow). (F) Closer examina-
tion reveals the presence of rough endoplasmic reticulum (RER; arrowheads) close to the vesicular-like struc-
ture. However, this vesicular compartment does not disclose any additional structural insights about the fine
architecture when imaged under 2-D TEM. (G) The area depicted under F was next subjected to tomographic
analysis, resulting in an aligned 3-D TEM dataset. (H) The segmented dataset with structural features of interest
outlined; the RER is shown with dark-orange lines and the vesicular-like cell inclusion in deep purple. Note
that the combined fluorescent information shown in B and C was merged within this tomographic slice (slice 61
out of 120), showing the cross-correlation between light and electron microscopy information. (I) and (J) show
the corresponding 3-D models of the dataset depicted under E–H, illustrating the RER, vesicular-like cell inclu-
sion, and the neighboring nuclear membrane (ice-blue color). Tomographic reconstruction and corresponding
ultrastructural analysis reveals that the vesicular complex is made up from a larger vesicle, which harbors the
cisplatin and the lysosomal marker, and smaller vesicles, which are devoid of any fluorescence. This configura-
tion is strongly indicative of the lysosomal degradation process in which the anticancer drug is caged within
the cells’ “digestive tract” (i.e., the larger vesicle or secondary lysosomes) and the associated smaller vesicles
most likely are residual bodies. This is an example in which fluorescence microscopy provides an indicative
glimpse while TEM 3-D modeling discloses the entire subcellular picture. Scale bars in A–D are 20 µm; in E,
2 µm; and in F, 500 nm.
washed twice with distilled water for 5 min each. Next, the fixed cells were dehydrated
through a gradient series of ethanol for 10 min each (i.e., 70, 80, and 90% and then
three times at 100%). Infiltration of the sample with resin (Epon) was achieved with
the following mixtures and incubation times: 25% Epon in ethanol for 4 h, 50% Epon
in ethanol overnight, and 100% Epon for 8 h at room temperature. (If it is necessary to
keep the samples at this stage for longer than 8 h, the samples should be stored at 4°C
to minimize self-polymerization of the resin and degradation of ultrastructural detail.)
With resin infiltration complete, we removed the final change of resin and then
used a site-specific approach to embed the cells of interest along with the carbon-film
relocation markers. To achieve this, a thin layer (∼5 mm) of fresh Epon was applied to
the bottom of the dish and then a cylindrical embedding capsule, opened at both ends,
was placed on top of each of the grid patterns, such that the patterns were centered in
the smaller ends of the capsules. The resin in the dish was then allowed to polymerize
overnight at 60°C. After this first stage of polymerization, the two capsules were filled
with more resin and allowed to polymerize for an additional 24 h at 60°C; this second
polymerization step produced a reasonable sized block for handling and ultramicrotomy
(Fig. 3(C)). Finally, the embedding capsules were gently peeled away from the dish,
while the resin was still warm (Fig. 3(D); Hanson, Reilly, Lee, Janssen, & Phillips,
2010). Note that if this step cannot be performed immediately, the dish should be briefly
placed on a heating plate to rewarm the resin before gently peeling the capsules away
from the glass surface; failure to do so will result in damage to the block faces.
The resulting block faces contained the monolayer of embedded cells as well as
the surface imprint of the carbon finder-grid pattern (Fig. 3(E)). The block face was
trimmed with a razor blade, retaining the regions of interest previously captured by
light and/or laser microscopy. Regions of interest were easy to find by observation of
the carbon fiducial marks with either the binoculars of the ultramicrotome or a stereo
microscope (Fig. 3(E)). The trimmed block face was then sectioned with an ultrami-
crotome (Leica Ultracut-7), yielding sections of thicknesses ranging from 70 to 250 nm for
subsequent TEM studies. In some instances, semithick sections (200 nm) were also
collected on glass slides and stained with 0.5% toluidine blue for 1 min on a hot plate
(Fig. 3(F); Wisse et al., 2010). TEM sections were collected on either 200 mesh grids
or formvar-coated slot grids (with a single window or three windows) for electron
tomography. The use of slot grids is highly beneficial in achieving successful cross-
correlative relocation because larger areas can be viewed. This significantly reduces the
chance that the grid bars will obscure areas of interest.
Before examination, the samples on the copper grids were stained with saturated
uranyl acetate (in 50% ethanol) and Reynold’s lead citrate for 10 min each, and were
washed thoroughly with water in between steps to minimize stain deposits. For electron
tomography, gold particles with a diameter of 10 nm were deposited directly onto the
sections before recording the tilt series. Deposition involved dipping the mounted TEM
sections in a 1:100 aqueous solution of colloidal gold for 1 min and then washing the
grids by dipping it three times in double-distilled water. Excess water was removed by
gently blotting the grid on filter paper. The gold particles were used as fiducial markers
for the subsequent calculation of the tomograms.

For TEM investigations, areas of interest were examined with either a JEOL 1400
TEM at 120 kV or a JEOL 2100 TEM at 200 kV (JEOL, Tokyo, Japan). The JEOL
1400 was used to obtain 2-D images for correlation with data acquired by live-cell
microscopy, whereas the JEOL 2100 was the preferred platform for 3-D recording (i.e.,
electron tomography) for correlation with images obtained by confocal laser micros-
copy. Digital Micrograph (Gatan) was used for data acquisition and storage of electron
micrographs on both platforms.
Before images were collected, the regions of interest were relocated within the TEM
with the aid of the corresponding DIC images and/or the information provided by the
toluidine-blue-stained sections. From a practical point of view, it is relatively easy to
relocate cells grown in a semiconfluent manner, especially by starting the examination
of the sections within the TEM at low magnification. For this purpose, characteristic
cellular landscapes, such as unusual cell shapes, intercellular spaces, and cell inclu-
sions, can be used to help relocate the region of interest within the electron microscope
(Fig. 4).
For transmission electron tomography (Fig. 5(G–J)), tilt series were recorded with
a double-tilt holder (JEOL, EM-31630) in an automated manner under SerialEM soft-
ware (Boulder Lab for 3D Electron Microscopy of Cells) with incremental steps of
1° over the range from −60° to +60°. All subsequent data processing steps were con-
ducted with IMOD (Kremer, Mastronarde, & McIntosh, 1996). The electron-dense gold
nanoparticles, which were deposited onto the samples (see part C), were used as fiducial
markers to align each of the images in the double-tilt series. Subsequently, tomograms
were computed by using resolution-weighted back-projection and combined into tomo-
grams with eTOMO. 3-D contour models of the cellular structures were produced by
computer-assisted tracking of high-contrast lines, combined with manual segmenting
within the tomograms by using 3dMOD (Biazik, Jahn, & Braet, 2011; Braet et al., 2009).

Calibration of the digital images and the image-capture process was done with
“magnification calibration” standards provided by the microscope suppliers. For gen-
eral image processing and correlative analysis at matching pixel resolution, digital data
obtained from two microscopes were transferred to ImageJ for color adjustment and
figure assembly by using the “replace colour” and “duplicate layer/merge” functions
(Abramoff, Magelhaes, & Ram, 2004).

Instrumentation: Emitech K950C carbon coater, Heraeus labofuge 200, Neubauer
chamber.
Materials: Caco-2 cell line (ATCC, Item No. HTB-37), 35 mm glass-bottom culture
dishes (MatTek, P35G-1.0-14-C), parafilm (Crown Scientific, 9910001), tissue culture
plasticware (Corning Life Sciences), and copper finder grids (ProSciTech, GCU200F2).
Reagents: Ethanol (Merck, Cat No. 4.10230, CAS # 64-17-5), advanced DMEM
media (Life technologies, 12491023), antibiotic/antimycotic (Life technologies,
15240104), heat-inactivated fetal bovine serum (FBS; Life technologies, 10100-147),
L-glutamine (Sigma-Aldrich, G7513), phosphate buffered saline (PBS; In Vitro Tech-
nologies, IVT3001302), and trypsin-EDTA (Sigma-Aldrich, 59430C).

Instrumentation: Olympus Live CellR IX81 microscope equipped with fully enclosed
temperature-controlled incubator (Solent Scientific) with CO2 enrichment system, MT20
illumination system, and Märzhäuser motorized stage that allows multipoint position-
ing and MIA; portable incubator (LabIVF, LE-INC960GL); Olympus FV1000 confocal
microscope.
Materials: Glass “top lid” (MatTek, P24GTOP-1.5-F), Olympus live CellR software
(FV10-ASW 1.7), “xcellence rt imaging” software with “Multiple Image Alignment”
(MIA).
Reagents: Cisplatin conjugated to FITC (cisplatin–FITC) was synthesized as out-
lined in Molenaar et al., 2000; DMEM, no Phenol Red (Life Technologies, 31053-
028); LysoTracker Red DND-99 (Life Technologies, L-7528).

Instrumentation: Heating oven (Binder WTC, Tübingen, Germany); Leica Ultracut-7
ultramicrotome (Leica, Vienna, Europe).
Materials: Standard embedding capsules (ProSciTech, RB001 BEEM®).
Reagents: Copper grids, 200 mesh square (ProSciTech, GCu200); copper slot grid
(ProSciTech, GCU1M2 M); copper grids, three slots (ProSciTech, GCU2M3X1 M); eth-
anol; glutaraldehyde (ProSciTech, C001); lead nitrate (Sigma-Aldrich, 228621); osmium
tetroxide (ProSciTech, C010); procure 812 embedding kit Epon resin (ProSciTech,
C038); sodium citrate (ProSciTech, C077); sodium cacodylate (ProSciTech, C020);
sodium hydroxide (Sigma-Aldrich, S8045); sucrose (Sigma-Aldrich, S9378); toluidine
blue (ProSciTech, C078); uranyl acetate (ProSciTech, C079-F).

Instrumentation: JEOL 1400 and JEOL 2100 transmission electron microscopes
(JEOL, Tokyo, Japan).
Materials: DigitalMicrograph (v.1.82.366); SerialEM-software (v.3.1); IMOD full
package (v.4.1, Boulder Laboratory for 3-Dimensional Electron Microscopy of Cells
and the Regents of the University of Colorado, USA).
Reagents: Fiducial markers (Sigma Chemicals, 10 nm immunogold-labeled IgG).

Instrumentation: Home-built 64-bit computer (16 Gigabytes RAM 1600 MHz, Intel
i7-2700 K Quad Core 3.5 GHz CPU) running Windows 7 Professional Software Package.
Materials: ImageJ (v1.45s, NIH, USA).
V. Discussion
The methodological protocol outlined here is the culmination of previous CLEM
approaches that we used to answer questions about the role of the actin cytoskeleton in
different cell types and disease models (Braet et al., 2007; Jahn, Barton, & Braet, 2007;
Jahn & Braet, 2008, Jahn et al., 2009). As one of the first elegant examples of CLEM on
whole-mounted cells with noncombinatorial labeling and classical EM sample prepara-
tion, the initial paper by Peachey et al. (Peachey, Ishikawa, & Murakami, 1996) gave
us the methodological inspiration to demonstrate, in 2002, that filamentous actin is not
needed to maintain individual fenestrae structures in rat liver sinusoidal endothelial
cells (Braet et al., 2002), in contrast to what many researchers had previously postu-
lated. This unambiguous outcome was only made possible by combing fluorescence
details from multiple-labeled, whole-mounted cells with data from scanning electron
microscopy (SEM). More recently, we gathered new insights on cell fate and the state
of cancer cells that were subjected to anti-actin and anti-cancer compounds by employ-
ing correlative 3-D time-lapse fluorescence and SEM studies (Su, Whan, et al., 2010).
The high-resolution detail allowed us to identify the onset of micro-membrane struc-
tures in response to actin changes. These structures probably were the first preapoptotic
changes preceding the larger membrane blebbing that is so typical of the final stages of
apoptosis (Su, in preparation), but they would have been impossible to see with fluores-
cence microscopy alone. Despite the power of these SEM-based CLEM techniques, the
current need to investigate the final destination and effects of drug compounds at the
suborganelle level obviously necessitated the use of either thin or semithick sections of
cells for TEM examination. This scientific need led to the development of the CLEM
approach described in this chapter (Figs. 4 and 5).
A. Making the Most of This Approach

We have presented two options for collecting real-time information on living cells
under well-controlled conditions (37°C, 100% relative humidity, and 5% CO2). The
choice of the optical platform largely depends on the demands of the research: whether
one wants to follow multiple areas of the sample in 2-D (X, Y ) for prolonged times
(hours to days) by using monochromator technology at relative high capture-speeds
(live-cell imaging) or to collect accurate 3-D (X, Y, Z ) information on the dynamic
behavior of cells for shorter experimentation times (minutes to hours) by confocal laser
microscopy. An obvious, though essential, component needed to collect meaningful
cross-correlative live-cell observations is a dedicated tissue-culture incubator typically
present on live-cell microscopes or a microscopic cell chamber typically present on
confocal laser imaging platforms to ensure tight control of the physiological envi-
ronment. Such an “environmental” device is absolutely essential for the subsequent
collection of cross-correlative electron microscopy data. Fluctuations in temperature,
osmotic pressure, and pH that can occur in an uncontrolled environment are known to
have severe effects on cell viability and structure, making correlative interpretation
difficult, if not impossible (Shotton, 1993).
As previously discussed (Su, Whan, et al., 2010), precautions must be taken during
live-cell imaging to avoid light-induced cell stress. Such precautions include the use of
minimal light or photon exposure, which can be achieved by combining high-sensitivity cam-
eras with optical-dimming technology. It is also essential to use a cell-culture medium that
is suitable for time-lapse experimentation. The main requirement here is that the medium
should be devoid of phenol-based indicators, which are well-known photon absorbers
and hence contribute directly to light-induced phototoxicity. Based on our overall expe-
rience, we recommend live-cell microscopy with monochromator optical technology if
one wants to correlate full time-lapse observations over prolonged times (from more than
one hour up to several days) with TEM data (Fig. 4). On the other hand, confocal imag-
ing is superior when the dynamic process to be studied occurs in less than an hour and at
high lateral resolution (Fig. 5). Confocal microscopy also has the advantage that volume
data, with a typical Z-resolution on the order of 350–500 nm, can be retrieved and this
is something that live-cell platforms cannot deliver (Cox, 2007). However, it should be
noted that confocal and multiphoton microscopy approaches are relatively prone to
inducing phototoxicity, especially compared to the minimal-light-exposure technology
on dedicated live-cell platforms (Hoebe et al., 2007; Papkovsky, 2010).
The advantage of using relocation marks as reported previously (Brown et al., 2009)
was a major step toward relocating a relative large number of cells (i.e., typically between
10 and 15 cells) within a single correlative experiment (Fig. 3). The use of fiducial car-
bon footprints on glass-bottom cell-culture dishes allowed us to record the regions of
interest at the end of the light or laser microscopy study and then use the marks, which
remained present in the embedded material, to carefully select the correct regions for
ultramicrotomy (Fig. 3(D, E)). Given that the light or laser microscopy experiments and
all subsequent sample preparation steps can be performed within the same dish (Fig.
3(B–D)), previously imaged cells can be easily relocated through the eye pieces of the
ultramicrotome during sectioning (Fig. 3(E–F)). The use of the “cellular landscape” as
seen under the TEM at intermediate magnification (e.g., Fig. 4(J) and Fig. 5(E)) can then
be used to guide the microscopist to the precise location for higher resolution imaging
and a full supramolecular view of the cell’s interior (e.g., Fig. 4(K) and Fig. 5(J)).
As a consequence of these advances in relocation markers and in multipositioning
sample stages on optical microscopes, we typically end up with more marked regions
of interest than can actually be studied in full detail by the TEM. Consequently, care-
ful choices have to be made to obtain a full picture of the biological process to be
studied, without being (i) overwhelmed by the amount of new information available
or (ii) being distracted by artificial structural information introduced during the sample
manipulation steps (Braet & Ratinac 2007). To help avoid these problems, one needs a
good knowledge and understanding of the sample of interest, and must also be able to
apply high-quality EM sample preparation techniques to the sample. This knowledge

and these skills should be brought to bear prior to commencing time-consuming serial
sectioning and 3-D TEM exercises. We should emphasize that expertize in classical
EM sample preparation is foundational—decades of studies with these procedures have
proven their utility for addressing innumerable ultrastructural questions, despite occa-
sional problems with artifacts. Still, should concerns arise over whether any observed
“new” feature might be artifactual, alternative correlative sample preparation proce-
dures should be implemented. This “good practice” can provide a higher degree of
confidence about the structures of interest as the observations from one method can be
compared to those from the other method(s). For this purpose, we recommend com-
paring results obtained by classical preparation procedures for electron microscopy
(i.e., chemical fixing and staining) with those achieved by physical fixation (i.e., high-
pressure freezing and freeze substitution; Braet, Bomans, Wisse, & Frederik, 2003;
McDonald, 2009; Verkade, 2008). Nonetheless, the reader should bear in mind that,
while cryogenic sample preparation is particularly suited for combinatorial immuno-
gold-labeling studies (Brown et al., 2009; Mayhew, 2011; Plitzko, 2009), physical fixa-
tion also has specific limitations, such as small sample size.
B. A 4-Dimensional Future

Irrespective of the current state of CLEM approaches, much can be expected in the
future when it becomes possible to add an extra dimension (i.e., time) to TEM. As evident
from Fig. 1(A), cross-correlative studies are currently at their best when 4-D light optical
data are integrated with static 3-D TEM information. Although TEM sample prepara-
tion approaches offer a “snapshot” of features, the dynamic aspects are still absent. This
limitation inherently complicates full correlation across imaging platforms and requires
skills in correctly assembling the pieces of the structural biology puzzle to reconstruct the
cells’ content and dynamic behavior over the relevant length scales. The ultimate vision
for CLEM is “hand-in-hand” 4-D light microscopy and 4-D TEM, including all the sam-
ple preparation steps in one run and within one working day. While this vision seems far
away, and may not actually be fully achievable due to the limitations of electron optics,
we have already glimpsed the extraordinary possibilities that integrated fluorescence
microscopy and TEM imaging can bring to the scientist (Agronskaia et al., 2008). Much
can be also expected from the latest advances in specialized holders and other devices,
such as the recently reported technology for observing “living” (or at least “active”) bio-
logical material within a native, liquid environment in the TEM (Degen, Dukes, Tanner,
& Kelly, 2012). This is another example of technological advances that will undoubtedly
open up new possibilities in closing the temporal and resolution gap in biological CLEM.
C. Conclusion
We have described here a simple approach to investigate structure–function detail
across different microscopy platforms and multiple dimensions. It is based on noncom-
binatorial labeling, the use of carbon-coated relocation markers, and routine sample
preparation protocols for electron microscopy. This approach allows researchers to

monitor the uptake of fluorescently labeled drugs in real-time and then to combine the
obtained results with TEM data. Today, CLEM has become a rather standard micros-
copy approach, and one that nicely entwines recent technological advances in molecular
and structural biology methods. With time, correlative microscopy is certain to appear
in more and more publications that will demonstrate how this contemporary approach
to visual thinking continues to facilitate groundbreaking discoveries in cell biology.
Acknowledgments
The authors acknowledge the facilities, as well as technical and administrative
assistance from staff, of the AMMRF at the Australian Centre for Microscopy &
Microanalysis of the University of Sydney. The authors are also indebted to E. Kable,
K. Jahn, and D. Barton for continuous support. We wish to thank the Australian Research
Council (ARC) for funding some of the research reported herein through the “Link-
age Infrastructure, Equipment and Facilities” funding scheme (Grants LE0775598,
LE0883030, and LE100100010).
References
Abramoff, M. D., Magelhaes, P. J., & Ram, S. J. (2004). Image processing with image. Journal of Biophotonics
International, 11, 36–42.
Agronskaia, A. V., Valentijn, J. A., van Driel, L. F., Schneijdenberg, C. T., Humbel, B. M., van Bergen en
Henegouwen, P. M., et al. (2008). Integrated fluorescence and transmission electron microscopy. Journal
of Structural Biology, 164, 183–189.
Biazik, J. M., Jahn, K. A., Su, Y., Wu, Y. N., & Braet, F. (2010). Unlocking the ultrastructure of colorectal
cancer cells in vitro using selective staining. World Journal of Gastroenterology, 16, 2743–2753.
Biazik, J. M., Jahn, K. A., & Braet, F. (2011). Caveolae and caveolin-1 in reptilian liver. Micron, 42, 656–661.
Braet, F., & Ratinac, K. R. (2007). Creating next-generation microscopists: structural and molecular biology
at the crossroads. Journal of Cellular and Molecular Medicine, 11, 759–763.
Braet, F., Spector, I., Shochet, N. R., Crews, P., Higa, T., Menu, E. , et al. (2002). The new anti-actin agent
dihydrohalichondramide reveals fenestrae-forming centers in hepatic endothelial cells. BMC Cell
Biology, 3, 7.
Braet, F., Bomans, P. H.H., Wisse, E., & Frederik, P. M. (2003). The observation of intact hepatic endothelial
cells by cryo-electron microscopy. Journal of Microscopy, 212, 175–185.
Braet, F., Wisse, E., Bomans, P., Frederik, P., Geerts, W., Koster, A. , et al. (2007). Contribution of high-
resolution correlative imaging techniques in the study of the liver sieve in three-dimensions. Microscopy
Research and Technique, 70, 230–242.
Braet, F., Riches, J., Geerts, W., Jahn, K. A., Wisse, E., & Frederik, P. (2009). 3-D organization of fenestrae
labyrinths in liver sinusoidal endothelial cells. Liver International, 29, 603–613.
Brown, E., Mantell, J., Carter, D., Tilly, G., & Verkade, P. (2009). Studying intracellular transport using
high-pressure freezing and correlative light electron microscopy. Seminars in Developmental Biology,
20, 910–919.
Chauhan, S. S., Liang, X. J., Su, A. W., Pai-Panandiker, A., Shen, D. W., Hanover, J. A., et al. (2003).
Reduced endocytosis and altered lysosome function in cisplatin-resistant cell lines. British Journal of
Cancer, 88, 1327–1334.
Conrad, C., Wünsche, A., Tan, T. H., Bulkescher, J., Sieckmann, F., Verissimo, F. , et al. (2011). Micropilot:
automation of fluorescence microscopy-based imaging for systems biology. Nature Methods, 8, 246–249.
Cox, G. (2007). The digital image. In G. Cox (Ed.), Optical techniques in cell biology (pp. 77–83). London:
CRC Press.
Degen, K., Dukes, M., Tanner, J. R., & Kelly, D. F. (2012). The development of affinity capture devices—a
nanoscale purification platform for biological in situ transmission electron microscopy. doi:10.1039/
C1032RA01163H. RSC Advances in Press.
Geissing, H. D. (1974). A precise stage arrangement for correlative microscopy for specimens mounted on
glass slides, stubs or EM grids. Journal of Microscopy, 100, 113–117.
Giepmans, B. N., Adams, S. R., Ellisman, M. H., & Tsien, R. Y. (2006). The fluorescent toolbox for assessing
protein location and function. Science, 312, 217–224.
Giepmans, B. N. (2008). Bridging fluorescence microscopy and electron microscopy. Histochemistry and
Cell Biology, 130, 211–217.
Guizetti, J., Mäntler, J., Müller-Reichert, T., & Gerlich, D. W. (2010). Correlative time-lapse imaging and
electron microscopy to study abscission in HeLa cells. Methods in Cell Biology, 96, 591–601.
Hanson, H. H., Reilly, J. E., Lee, R., Janssen, W. G., & Phillips, G. R. (2010). Streamlined embedding of
cell monolayers on gridded glass-bottom imaging dishes for correlative light and electron microscopy.
Microscopy and Microanalysis, 16, 747–754.
Hoebe, R. A., Van Oven, C. H., Gadella, T. W. J., Jr., Dhonukshe, P. B., Van Noorden, C. J.F., & Manders, E.
M.M. (2007). Controlled light-exposure microscopy reduces photobleaching and phototoxicity in fluo-
rescence live-cell imaging. Nature Biotechnology, 25, 249–253.
Hsieh, C. C., Huang, S. B., Wu, P. C., Shieh, D. B., & Lee, G. B. (2009). A microfluidic cell culture platform
for real-time cellular imaging. Biomedical Microdevices, 11, 903–913.
Jahn, K. A., & Braet, F. (2008). Monitoring membrane rafts in colorectal cancer cells by means of correlative
fluorescence electron microscopy (CFEM). Micron, 39, 1393–1397.
Jahn, K., Barton, D., & Braet, F. (2007). Correlative fluorescence- and scanning, transmission electron
microscopy for biomolecular investigation. In A. Méndez-Vilas, & J. Díaz (Eds.), Modern research and
educational topics in microscopy (pp. 203–211). Badajoz: Formatex Press.
Jahn, K. A., Barton, D. A., Su, Y., Riches, J., Kable, E. P.W., Soon, L. L., et al. (2009). Correlative fluores-
cence and transmission electron microscopy imaging of the actin cytoskeleton of whole-mount (breast)
cancer cells. Journal of Microscopy, 235, 282–292.
Jahn, K. A., Biazik, J. M., & Braet, F. (2011). GM1 expression in Caco-2 cells: characterisation of a funda-
mental, passage-dependent transformation. Journal of Pharmaceutical Sciences, 100, 751–3762.
Jahn, K. A., Barton, D. A., Kobayashi, K., Ratinac, K. R., Overall, R. L., & Braet, F. (2012). Correlative
microscopy: providing new understanding in the biomedical and plant sciences. Micron, 43, 565–582.
Jones, S. A., Shim, S. H., He, J., & Zhuang, X. (2011). Fast, three-dimensional super-resolution imaging of
live cells. Nature Methods, 8, 499–508.
Kremer, J. R., Mastronarde, D. N., & McIntosh, J. R. (1996). Computer visualization of three-dimensional
image data using IMOD. Journal of Structural Biology, 116, 71–76.
Lemieux, B., Percival, M. D., & Falgueyret, J. P. (2004). Quantitation of the lysosomotropic character of
cationic amphiphilic drugs using the fluorescent basic amine Red DND-99. Analytical Biochemistry,
327, 247–251.
Massover, W. H. (2011). New and unconventional approaches for advancing resolution in biological trans-
mission electron microscopy by improving macromolecular specimen preparation and preservation.
Micron, 42, 141–151.
Mayhew, T. M. (2011). Mapping the distributions and quantifying the labelling intensities of cell compart-
ments by immunoelectron microscopy: progress towards a coherent set of methods. Journal of Anatomy,
219, 647–660.
McDonald, K. (2009). A review of high-pressure freezing preparation techniques for correlative light and
electron microscopy of the same cells and tissues. Journal of Microscopy, 235, 273–281.
Mironov, A. A., & Beznoussenko, G. V. (2009). Correlative microscopy: a potent tool for the study of rare or
unique cellular and tissue events. Journal of Microscopy, 235, 308–321.
Molenaar, C., Teuben, J. M., Heetebrij, R. J., Tanke, H. J., & Reedijk, J. (2000). New insights in the cellular
processing of platinum antitumor compounds, using fluorophore-labeled platinum complexes and digital
fluorescence microscopy. Journal of Biological Inorganic Chemistry, 5, 655–665.
Müller, S. A., Aebi, U., & Engel, A. (2008). What transmission electron microscopes can visualize now and
in the future. Journal of Structural Biology, 163, 235–245.
Papkovsky, D. B. (2010). Live cell imaging: Methods and protocols. Munich: Springer-Verlag Gmbh.
Peachey, L. D., Ishikawa, H., & Murakami, T. (1996). Correlated confocal and intermediate voltage electron
microscopy imaging of the same cells using sequential fluorescence labeling, fixation, and critical point
dehydration. Scanning Microscopy, 10, 237–247.
Pelletier, L., O’Toole, E., Schwager, A., Hyman, A. A., & Muller-Reichert, T. (2006). Centriole assembly in
Plitzko, J. M. (2009). Correlative cryo-microscopy: from cellular territories to molecular landscapes.
Abstracts of Papers of the American Chemical Society, 238, 86.
Polishchuk, R. S., & Mironov, A. A. (2001). Correlative video light/electron microscopy. Current Protocols
in Cell Biology, 4, 4.8.
Rieder, C. L., & Bowser, S. S. (1985). Correlative immunofluorescence and electron-microscopy on the same
section of epon-embedded material. The Journal of Histochemistry and Cytochemistry, 33, 165–171.
Robinson, J. M. (2001). Biological labeling and correlative microscopy. Acta Histochemica, 103, 261–264.
Sander, B., & Golas, M. M. (2011). Visualization of bionanostructures using transmission electron micro-
scopical techniques. Microscopy Research and Technique, 74, 642–663.
Schermelleh, L., Heintzmann, R., & Leonhardt, H. (2010). A guide to super-resolution fluorescence micros-
copy. The Journal of Cell Biology, 190, 165–175.
Sharma, P., Ando, D. M., Daub, A., Kaye, J. A., & Finkbeiner, S. (2012). High-throughput screening in
primary neurons. Methods in Enzymology, 506, 331–360.
Shotton, D. M. (1993). Electronic light microscopy: Techniques in modern biomedical microscopy. New
York, NY: Wiley, John & Sons, Incorporated.
Simmons, S. R., Pawley, J. B., & Albrecht, R. M. (1990). Optimizing parameters for correlative immuno-
gold localization by video-enhanced light microscopy, high-voltage transmission electron microscopy,
and field emission scanning electron microscopy. The Journal of Histochemistry and Cytochemistry, 38,
1781–1785.
Stewart, D. J. (2007). Mechanisms of resistance to cisplatin and carboplatin. Critical Reviews in Oncology/
Hematology, 63, 12–31.
Su, Y., Nykanen, M., Jahn, K. A., Whan, R., Cantrill, L., Soon, L. L., et al. (2010). Multi-dimensional cor-
relative imaging of subcellular events: combining the strengths of light and electron microscopy. Bio-
physical Reviews, 2, 121–135.
Su, Y., Whan, R., Empsen, C., Soon, L., & Braet, F. (2010). Multidimensional live cell imaging of cancer-
mediated events. In A. Méndez-Vilas, & J. Díaz (Eds.), Microscopy: Science, technology, applications
and education (pp. 2050–2061). Badajoz: Formatex Press.
Szymański, P., Markowicz, M., & Mikiciuk-Olasik, E. (2012). Adaptation of high-throughput screening in
drug discovery-toxicological screening tests. International Journal of Molecular Sciences, 13, 427–452.
Tsien, R. Y. (2003). Imagining imaging’s future. Nature Reviews Molecular Cell Biology, 4, S16–S21.
Verkade, P. (2008). Moving EM: the rapid transfer system as a new tool for correlative light and electron
microscopy and high throughput for high-pressure freezing. Journal of Microscopy, 230, 317–328.
Wang, D., & Lippard, S. J. (2005). Cellular processing of platinum anticancer drugs. Nature Reviews Drug
Discovery, 4, 307–320.
Watari, N., & Herman, L. (1965). Correlative light and electron microscopy of bat islets of Langerhans in
hibernating and nonhibernating states. American Zoologist, 5, 678.
Wisse, E., Braet, F., Duimel, H., Vreuls, C., Koek, G., Olde, D. S., et al. (2010). Assessment of fixation meth-
ods for the study of human and other livers by electron microscopy. World Journal of Gastroenterology,
16, 2851–2866.
Zewail, A. H. (2010). Four-dimensional electron microscopy. Science, 328, 187–193.
CHAPTER 2
Visualizing Live Dynamics

and Ultrastructure of Intracellular
Organelles with Preembedding
Correlative Light-Electron Microscopy
Roman S. Polishchuk*, Elena V. Polishchuk† and Alberto Luini*,†
* Telethon Institute of Genetics and Medicine, Naples, Italy
† Institute of Protein Biochemistry, Naples, Italy
Abstract
I. Introduction and Rationale
II. Materials
III. Methods
A. Cell Transfection, Observation, and Fixation
B. Immuno-labeling
C. Embedding
D. Sectioning
E. Serial-Section Analysis and 3D Reconstruction
IV. Discussion
V. Summary
Acknowledgments
References
Abstract
One of the very effective methods to perform correlative light-electron microscopy

(CLEM) is to combine video imaging of live cells with immuno-electron microscopy.
This technique can thus provide detailed, high-resolution characterization of dynamic
intracellular organelles. The use of green fluorescent protein (GFP)-tagged chimeras
allows the movements and/or behavior of intracellular structures in a live cell to be
followed, which can then be fixed at the moment of interest. The subsequent immuno-
electron microscopy analysis reveals the three-dimensional (3D) architecture of the
same structure, together with the precise identification of the GFP-labeled protein
22 Roman S. Polishchuk et al.
pattern. The process resembles taking a high-resolution snapshot of an interesting

and/or rare live event. Conceptually, it consists of a switch of wavelengths, from that
of photons to that of electrons, with the associated huge gain in resolution. In this
respect, CLEM can be considered as the first, and probably one of the most power-
ful, super-resolution microscopy techniques. This switch, however, requires complex
manipulations of the sample. Considering that CLEM is a very valuable but technically
challenging and time-consuming method, accurate protocols are needed to simplify
the efforts of researchers who are willing to apply this method for their own purposes.
Here, we present a detailed description of the preembedding CLEM procedures that
explains the know-how and the “tricks of the trade” that are involved in carrying out
the crucial steps of CLEM.
I. Introduction and Rationale
Correlative microscopy emerged decades ago when morphologists started to com-

pare structural characteristics of cells or tissues across the light microscopy and the
electron microscopy (EM) levels (Hayat, 1987). A few studies in fixed specimens even
showed how exactly the same structure (such as an actin filament or a secretory gran-
ule) appears under light microscopy and EM (Svitkina, Verkhovsky, & Borisy, 1995;
Takizawa, Suzuki, & Robinson, 1998). However, correlative light-electron microscopy
(CLEM) has been used fairly rarely in the past, probably because the ability to correlate
two static images, one fluorescent and one under EM, is of limited interest.
A critical step forward, however, took place about 12 years ago, as a result of the
pervasive use of green fluorescent protein (GFP)-based video microscopy and the revo-
lutionary impact this technique had in cell biology. In several cases, a few moments
of time-lapse video were enough to resolve an issue that years of microscopy of fixed
cells had failed to resolve (Lippincott-Schwartz, Roberts, & Hirschberg, 2000; Pelham,
1997). Subsequently, however, it became frustratingly clear that spectacular, GFP tech-
nology has its own limits. The relatively low resolution of light microscopy was a major
drawback of GFP-based imaging, at least in the (many) cases in which the crucial struc-
tural changes occur beyond the resolving power of light (Pelham, 1997).
This prompted the cell biology field to search for super-resolution technologies to
move beyond this resolution barrier. Therefore, CLEM was conceived through the inte-
gration of GFP-based video microscopy with the huge resolving power of EM. The
process resembles taking a high-resolution snapshot of an interesting and/or rare live
event. Conceptually, it consists of a switch of wavelengths, from that of photons to that
of electrons, obviously with the consequent huge gain in resolution. CLEM can thus
be considered as one of the first, and probably still the most powerful, super-resolution
microscopy technique. This wavelength switch, however, requires complex manipula-
tions of the sample.
The first description of this technique was reported by Polishchuk et al. (2000), in a
study which demonstrated that transport carriers operating between the Golgi complex
and the plasma membrane differ from the “expected” small round vesicles. Instead,
2. Pre-embedding Correlative Light-Electron Microscopy 23
these transport organelles appeared as significantly larger structures, with a complex

tubular–saccular architecture (Polishchuk et al., 2000). Importantly, in addition to the
monitoring of GFP-labeled organelles in vivo, CLEM also allowed their visualization
in the context of the intracellular space that is filled with neighboring structures, such
as cytoskeleton elements, mitochondria, and membrane compartments.
The productivity of this novel CLEM procedure appeared to be surprisingly high,
as it resulted in a burst of papers on the characterization of the 3D ultrastructure of
transport carriers that operate at the endoplasmic reticulum (ER)-to-Golgi interface
(Marra et al., 2001; Mironov et al., 2003) and at the exit of the Golgi toward either the
different cell-surface domains (Polishchuk, Di Pentima, Luini, & Polishchuk, 2003;
Polishchuk, Di Pentima, & Lippincott-Schwartz, 2004) or the endo-lysosomal system
(Polishchuk, San Pietro, Di Pentima, Teté, & Bonifacino, 2006). CLEM has also been
successfully applied to the analysis of endocytic structures (Caplan et al., 2002) and to
the morphology of the Golgi complex during mitosis (Altan-Bonnet, Phair, Polishchuk,
Weigert, & Lippincott-Schwartz, 2003). Given that the structures of interest in all of
these studies were immuno-labeled for EM before the inclusion of the cells in resin,
this type of CLEM was defined as “preembedding” (Verkade, 2008).
In 2002, the applications for CLEM were extended to the cryosectioning technique
(Oorschot, de Wit, Annaert, & Klumperman, 2002), and then applied to examine the
kinetics and localization of the lysosomal membrane protein LAMP1 (van Rijnsoever,
Oorschot, & Klumperman, 2008). However, up to this point, CLEM of the individual
intracellular structure observed in live cells required chemical fixation for its ultra-
structural EM characterization. Although modern fixation protocols usually preserve
the ultrastructure quite well, under certain conditions, some fine structure details can
indeed be lost (Dahl & Staehelin, 1989). To avoid this, Verkade and colleagues pro-
posed a new tool to rapidly freeze an object of interest at the end of its in vivo obser-
vation by using high-pressure freezing (Brown, Mantell, Carter, Tilly, Verkade, 2009;
Verkade, 2008). They used this high-pressure freezing technique with endocytic trac-
ers, to monitor the dynamics and structure of multivesicular bodies and the fusion of
endosomal structures (Verkade, 2008). Furthermore, the cryofixing CLEM protocol
was employed to characterize centriole assembly in Caenorhabditis elegans embryos
during mitosis (Müller-Reichert, 2007).
Among the various improvements to the CLEM technique, we believe that the gen-
eration of hybrid probes (i.e., probes that can be visualized by both light microscopy
and EM) is of particular significance. An example is seen in the photoconvertion of
GFP fluorescence into an electron-dense signal (Grabenbauer et al., 2005), designated
as the GFP recognition after bleaching (GRAB) technique. Based on GFP bleaching
and photo-oxidation of 3,3′-diaminobenzidine into an electron-dense precipitate, this
GRAB technique allowed the analysis of the distribution of N-acetylgalactosaminyl-
transferase-2-GFP within the Golgi stack (Grabenbauer et al., 2005). However, due
to the low photo-oxidation potential of GFP, this method has failed to achieve broad
applicability. To circumvent this problem, a new fluorescent flavoprotein with excep-
tional photo-oxidation properties, known as mini singlet-oxygen generator (miniSOG),
was engineered from Arabidopsis phototropin 2. MiniSOG can be fused to proteins
instead of GFP, and therefore provides a novel opportunity for CLEM improvement
(Shu et al., 2011). The initial use of miniSOG-tagged proteins (such as actin, his-
tone 2B, connexin43, and SynCAM, to name but a few) allowed efficient correlation
between different intracellular structures at the light microscopy and EM levels, with
very specific labeling, good preservation of ultrastructure, and excellent EM contrast.
Despite the extensive evolution of CLEM (Brown et al., 2009), its original pre-
embedding recipe optimized in our laboratory certainly remains a simple and reliable
way to “see” a single intracellular event of interest (e.g., Fig. 1). Having said this, our
protocol contains a large number of critical steps that we would like to share further
with the scientific community. Therefore, we present here our CLEM procedure, which
comprises the following stages (Fig. 2):
(1) DNA transfection
(2) Observation of structures labeled with GFP in live cells
(3) Fixing and immuno-labeling for EM, followed by embedding in EPON
(4) Identification of the cell of interest in the resin block, and attachment of the
special EPON-made capsule as a support for the flat MatTek surface
(5) Block cutting
(6) EM analysis
(7) 3D reconstruction.
II. Materials
(1) Thirty-five-millimeter, glass-grid-bottomed dishes (MatTek Corporation, P35G-

2-14-C-grid).
(2) HEPES: 200 mM, pH 7.3.
(3) Paraformaldehyde (PFA): 4% solution in 200 mM HEPES, pH 7.3. Add less than
100 mL of 200 mM HEPES into a glass beaker with a magnetic stirring bar, and
warm to about 70°C. Weigh out 4.0 g PFA and add this to the beaker. Mix for a
maximum of 10 min, to have a clear/transparent solution; adjust the final volume
to 100 mL with 200 mM HEPES. Using a funnel with a piece of soft filter paper,
filter the solution. Store at 4°C.
(4) Mixture of 4% PFA and 0.05% glutaraldehyde (GA): To 10 mL of 4% PFA solu-
tion in 200 mM HEPES (pH 7.3) in a plastic container, add 62.5 µL of 8% GA to
obtain a mixture with the required final concentrations.
(5) Phosphate-buffered saline (PBS, pH 7.4) containing NaCl, KCl, Na2HPO4,
KH2PO4.
(6) Blocking/permeabilizing solution: 0.5% bovine serum albumin (BSA), 0.1%
saponin, 50 mM NH4Cl in PBS. Weigh out 0.5 g BSA, 0.1 g saponin, 0.27 g
NH4Cl. Dissolve these in PBS to yield a final volume of 100 mL.
(7) Primary polyclonal antibody against GFP (Abcam, Cat No. AB 290-50).
(8) Anti-rabbit Fab’ fragment coupled to 1.4 nm gold particles (Molecular Probes,
Cat No. 2004, anti-rabbit nanogold).
(9) Gold enhancement (Nanoprobes, Cat No. 2113, Gold enhancement for EM).
Fig. 1 Example of CLEM analysis of the ultrastructure of GFP-GGA1 transport carriers. (A) HeLa cells were transfected with a plas-
mid encoding GFP-GGA1, incubated for 30 min with TRITC-dextran (to discriminate between endocytic and biosynthetic structures),
and imaged using confocal fluorescence microscopy. The area of interest (outlined by the box corresponding to the inset) contains GFP-
GGA transport carriers that were fixed during the acquisition of the time-lapse sequence (numbered from 1 to 3; indicated by arrows).
(B) Sequence of time-lapse frames (B1–B4) shows the movements of the three transport carriers shown in panel A and inset, from the
juxtanuclear area toward the cell periphery. The cell of interest was quickly fixed during the course of observation (B4), and all the three
of these transport carriers were found again (see inset in panel A). (C) The cells were then prepared for immuno-gold EM with an anti-
GFP antibody. This low magnification EM image shows the same cell and the same area of interest (box) as in panel A. (D) Appearance
of these three transport carriers labeled with an anti-GFP antibody, as shown in four serial thin sections (D1–D4). The white arrowhead,
the black arrow, and the white arrow indicate the same three transport carriers as shown in the box and inset in panel A (numbered from
1 to 3, respectively). (E–G) Higher magnification of all the three of these transport carriers allows the spikes of the clathrin coat at their
surface to be appreciated. (H, I) Three-dimensional reconstructions of two of these GFP-GGA1-containing transport carriers, which
correspond to structures 1 and 3 in panels A–E and G.
(10) Phosphate buffer: 100 mM, pH 6.8: Prepare solution A with 200 mM of dibasic
sodium phosphate. Weigh out 35.61 g of Na2HPO4•2H2O and dissolve it in water,
to make 1.0 L. Prepare solution B with 200 mM of monobasic sodium phosphate.
Weigh out 31.21 g of NaH2PO4•2H2O and dissolve it in water, to make 1.0 L. Mix
24.5 mL of solution A with 25.5 mL of solution B, and make up to a total volume
of 100 mL with water.
(11) Osmium tetroxide (OsO4): 2% solution in water.
(12) Potassium ferrocyanide (K4[Fe(CN)6]•3H2O): 3% solution in water.
(13) Thiocarbohydrizide (CH6N4S): 1% solution in water.
(14) Uranyl acetate (UO2(CH3COO)2•2H2O): 0.5% solution in water.
(15) Ethanol (C2H5OH): 100%.
(16) Epoxy resin (EPON): Take 33.6 g EPON (Fluka, Cat No. 45345), 21.0 g dodecenyl
succinic anhydride (DDSA; Fluka, Cat No. 45346), and 16.8 g methyl nadic anhy-
dride (MNA; Fluka, Cat No. 45347) together in a test tube. Heat the tube in an oven
for 2–3 min at 60°C, and mix gently. Add 0.96 g 2,4,6-tris(dimethylaminomethyl)
Fig. 2 The stages of CLEM. The sequence of the main steps during the CLEM procedure. Black arrows indi-
cate the structure of interest, and grey arrows indicate the transition from one step of the procedure to the other.
For further details, see main text.
phenol (DMP-30; Sigma, Cat No. 45348) and mix gently again. It is possible to
freeze the EPON in aliquots at −20°C, and to store it for a long time before use.
(17) Slot grids covered with carbon-formvar supporting film-Cu-slots (Electron
Microscopy Science, FCF 2010 Cu slots).
(18) Pick-up loop (Agar, Cambridge, England).
(19) Tweezers.
(20) Diamond knife.
(21) Hydrofluoric acid: 40%.
(22) Special EPON-made resin blocks. Prepare EPON resin blocks before starting
the CLEM procedure. Fill the embedding capsules (Polysciences, Inc., Cat No.
08408-50) with fresh EPON and polymerize for 24 h in an oven at 60°C. After
polymerization, remove the EPON blocks from the capsules.
III. Methods
A. Cell Transfection, Observation, and Fixation

Our CLEM method starts with seeding the cells of interest into the 35 mm Petri
dishes (MatTek) that have a glass coverslip attached with a photo-etched grid on its
surface. The coordinated grid (see example in Fig. 3) allows the cell of interest to be
tracked quite easily during the different stages of the CLEM procedure. It is important
to note that growing cells to high confluence tends to mask the grid on the coverslip,
and therefore will complicate the detection of the cell observed in vivo. Hence, quite
a low cell density is recommended for the best observations (100,000 cells per 35 mm
Petri dish are enough; e.g., for Cos7 or MDCK cells).
The day after, the cells have to be transfected with the GFP-tagged chimera of inter-
est. Alternatively, stably transfected cell lines can also be used for CLEM. The day after
Fig. 3 The coordinated grid in the CLEM procedure. (A) Scheme of the overall layout of the coordinated grid of the MatTek Petri
dishes. The area outlined by the small rectangle corresponds to the images displayed in panel B. (B) Distribution of GPI-GFP-transfected
cells on the coordinated grid, under fluorescence and phase contrast, and their overlay. The arrow indicates the cell of interest, and the
grey line in overlay image shows the optimal size and position of the pyramid to trim for subsequent serial sectioning. (See color plate.)
transfection, observe the GFP-expressing cells under the light microscope equipped
with a setup that allows acquisition of time-lapse series of images. Before observation,
it is very important to clean the bottom of the MatTek dish with 100% ethanol, as any
dirt can affect the visualization of the coordinated grid. If the grid is hardly visible with
the 60× oil-immersion objective, then switching to the 10× or 20× objectives can help
understand the exact location of the cell of interest. Draw the position of the cell of
interest on the grid map (available from MatTek) and grab the image with the cell posi-
tion on the MatTek grid in phase contrast (or DIC) and fluorescence channels.
Having found the structure of interest, video record it until the moment you want to
catch for EM analysis. At this point, fix the cells very rapidly with 1 mL of the mixture of
4% PFA and 0.05% GA prepared in 200 mM HEPES, pH 7.3. We add the fixative mix-
ture directly into the culture medium using a pipette. It is important to make sure that the
cell of interest does not change its position in the Z-axis, and that the structure of interest
remains in the focal plain during the fixation. Focus drift during the fixation procedure
usually happens even due to minimal differences in the temperature between the cell cul-
ture medium and the fixative. In this case, the use of a hardware-automated focus device
is recommended, to keep the structure and the cell of interest in focus during the fixation
step of the CLEM procedure. It is important to note that immediately after addition of the
fixative, the GFP fluorescence decays significantly. Therefore, to keep the structure of
interest visible, the charge coupled device (or photomultiplier) gain has to be increased
at the moment the GFP signal starts to go down. As soon as the intracellular structures
stop any movement, stop grabbing time-lapse images and carefully change the solution,
adding again 2 mL of the mixture of 4% PFA and 0.05% GA; then leave the cells for
10 min at room temperature. Next, “wash” the cells with 4% PFA once, to remove the
residual GA (as residual GA might affect the recognition of GFP by the antibody), and
then postfix the cells with 4% PFA for 30 min.
B. Immuno-labeling
In the past, we used both immuno-gold and immuno-peroxidase protocols to label
structures of interest (Polishchuk et al., 2000, 2003), for its recognition under the elec-
tron microscope. However, since a few years, satisfactory secondary antibodies con-
jugated with horse radish peroxidase (HRP) have not been available on the market.
Moreover, the immuno-peroxidase method allows the protein epitopes within the lumen
of membrane-enclosed compartments to be labeled effectively, while the staining of
the cytosolic protein domains frequently results in extensive diffusion of the electron-
dense product from the place of the HRP-driven chemical reaction (Polishchuk et al.,
2000). In contrast, the immuno-gold protocol with enhancement of ultrasmall particles
provides reliable and reproducible labeling, independent of the compartmentalization
of antigen, and in addition, it provides an opportunity to quantify the local densities of
proteins of interest. Therefore, we recommend the use of this immuno-EM protocol for
the CLEM procedure.
To start the immuno-labeling, wash the cells three times with PBS. Then, incubate
the cells with the blocking/permeabilizing solution (0.5% BSA, 0.1% saponin, 50 mM
NH4Cl), for 20–30 min, and subsequently add the primary polyclonal antibody against
GFP, diluted 1:250 in blocking/permeabilizing solution. It is worth noting that instead
of using an anti-GFP IgG, an antibody against the protein of interest can be used to
label the GFP-tagged chimera observed in living cells. However, such an antibody
should be checked for its ability to continue to recognize the antigen after fixation with
the 4% PFA/0.05% GA mixture. Usually, 200 µL of the diluted antibody is enough to
cover the center of the Petri dish where the gridded coverslip is attached. Incubate the
cells with the primary antibody for 1 h at room temperature, and then overnight at 4°C.
The day after, wash the cells six times with PBS, and then add the secondary antibody,
the anti-rabbit Fab’ fragment coupled to 1.4 nm gold particles (diluted 1:50 or 1:100 in
blocking/permeabilizing solution), and leave for 2 h.
After this incubation with the antibodies, the gold-enhancement reaction has to be
performed, to increase the size of the 1.4 nm gold particles. The GoldEnhance mixture
is prepared immediately before use, as outlined below.
Gold enhancement reaction:
(1) Mix 1 part (e.g., 3 drops) of component A (enhancer, green cap) with 1 part (e.g.,
3 drops) of component B (activator, yellow cap), and wait for 10 min.
(2) Add 1 part (e.g., 3 drops) of component C (initiator, magenta cap).
(3) Add 1 part (e.g., 3 drops) of component D (buffer, white cap). Three drops of
each solution are enough for one 35 mm Petri dish from MatTek.
Add the mixture to the cells and leave them for 7–10 min. Start observing the cells
using a conventional bright-field light microscope. If you see the cells changing color
after only a few minutes, immediately stop the reaction by washing the cells three
times with PBS. This is very important to achieve gold particles with a homogeneous
size and to avoid their aggregation into big clumps. The Gold enhancement procedure
is successful when the cells became violet-gray in color. At this point, the cells can be
processed for embedding directly or washed three times in PBS and stored in 1% GA
in HEPES for several days.
C. Embedding
After the immuno-labeling, the cells have to be embedded in resin to allow their
further sectioning into serial slices for analysis under the electron microscope. Wash
the immuno-labeled cells three times with PBS, and postfix the cells with a mixture of
2% OsO4 and 100 mM phosphate buffer (pH 6.8) (1 part OsO4 plus 1 part phosphate
buffer) for 25–30 min on ice. Then wash the cells three times with water, and incubate
with 1% thiocarbohydrizide diluted in H2O, for 5 min. Wash the cells again, three times
with water, and add a mixture of 2% OsO4 and 3% potassium ferrocyanide (1:1) and
leave for 25 min.
Following the OsO4 postfixation, wash the cells three times with water and then
incubate them overnight at 4°C in 0.5% uranyl acetate diluted in H2O. We prefer to
stain the cells with 0.5% uranyl acetate before inclusion in the resin. Although uranyl
acetate staining can also be carried out on sections, performing this procedure before
the inclusion of cells in the resin allows reduced manipulation with slot-collected serial
sections, which help to prevent their eventual damage.
After the 0.5% uranyl acetate staining, wash the cells six times with water, and
dehydrate the specimens by passing them through sequential solutions of 50%, 70%,
and 90% ethanol, every 10 min, and then through 100% ethanol for 30 min, changing
the 100% ethanol solution three times (every 10 min). Then add the mixture of EPON
and 100% ethanol (1:1) to the cells and leave for 2–4 h at room temperature. Change
the mixture for pure EPON and leave for 2–4 h at room temperature. We strongly rec-
ommend the use of EPON (and not other resins) for the embedding. Both Spurr and
Araldite react with the plastic of the dish, and therefore they are not suitable for the
embedding procedure. Finally, incubate the specimens at 60°C in an oven for 24 h, to
allow the EPON resin to polymerize. It is important to note that after the addition of
EPON and the resin polymerization, the coordinated grid on the glass bottom of the
MatTek dish is not visible anymore. To observe your cells on the grid again, the glass
needs to be removed with hydrofluoric acid. So, take a plastic beaker with few mil-
liliters of 40% hydrofluoric acid in it, and with the help of tweezers, place the MatTek
dish inside the hydrofluoric acid for 30 min. After that, take the MatTek dish out of the
beaker (always using tweezers), wash it in 200 mM HEPES and then in water, clean it
with a piece of soft paper, and dry it. A replica of the coordinated grid will be visible
again in the surface of the resin at the bottom of the MatTek dish.
After dissolving the glass, the special EPON-made resin block (see Section 2 on
Materials) has to be attached to the region where your cell of interest is located (see
also Fig. 2). First use a conventional inverted microscope to locate the cell of interest
on the grid. While observing the cell of interest under the microscope, put a mark near
the cell on the surface of the resin in the MatTek dish using a STABILO OHPen uni-
versal marker. Then add the drop of resin on to the top of the mark, and stick the EPON
block on to it, placing it at the top of the drop. Incubate this adjusted MatTek dish with
the attached EPON block in an oven at 60°C for an additional 24 h.
D. Sectioning
The sectioning is the most tricky part of the CLEM procedure, and usually requires
previous experience in the cutting of thin sections and in ultramicrotome use. To pre-
pare the specimen for sectioning, check whether the EPON block and resin in the
MatTek dish are fully attached. If so, break the wall of the MatTek dish with pliers and
try also to eliminate excess resin attached to the EPON block (always using the pliers).
Be careful to not touch or damage the area of interest. Then, put the resin block into the
holder of the ultratome and examine it under a stereomicroscope. When the position of
the cell is found, first trim down to quite a large pyramid that has the cell of interest at
its center. Then, by rotating the glass-knife stage, align the bottom edge of the pyramid
parallel to the knife edge. The pyramid of optimal size should contain just few cells
with the cell of interest in the center (see Fig. 3). Bring the sample as close as possible
toward the glass knife, but do not touch the surface of the pyramid. Adjust the gap
(which is visible as a bright band of shadow, if all three of the lamps of an ultratome
are switched on) between the knife-edge and the surface of the sample. The gap has to
be identical in width between the uppermost and lowermost edges of the sample during
the up and down movement of the resin block.
When you are sure of the correct orientation of the pyramid, trim the excess resin to
make the surface of the pyramid smaller. If the cell of interest is the only one present
on the pyramid, this will help you to find it easily under the electron microscope. The
length of the pyramid should be less than 0.5 mm and its height should be less than
0.1 mm. This will allow the collecting of more serial sections on the slot, due to their
small size. The top and the bottom sides of the pyramid should be as parallel as pos-
sible, to achieve a straight ribbon of slices during the sectioning procedure. Align the
pyramid and the diamond knife again. Then start to cut serial sections with 60–70 nm
intervals, trying not to lose the first sections. Usually these are very important for
identification of the structure of interest. Pick up the consecutive ribbons of the serial
sections on the Cu slots (Electron Microscopy Science, FCF 2010, Cu slots), using the
perfect loop.
E. Serial-Section Analysis and 3D Reconstruction

In our experience, identification of the structure of interest under the electron micro-
scope is usually greatly facilitated by its immuno-labeling and by the presence of some
fiducial markers, namely:
(1) Other GFP-labeled organelles that were observed in vivo together with the struc-
ture of interest
(2) Cell nucleus
(3) Particular features of cell shape and/or cell protrusions.
Having the fluorescent and phase contrast (or DIC) images of the cell also signifi-
cantly helps to identify and track the structure of interest in the serial sections. So to
detect the structure of interest, place the slot grid with the first serial sections under the
electron microscope and find the cell of interest using the traces of the coordinates on
the first few sections. After detecting the correct cell, try to find the structure of interest
(that was observed under confocal microscope), and take images of this structure in the
subsequent serial sections.
The image acquisition and overall analysis of the structure of interest can also be
accelerated through the use of automated XY stage in the electron microscope (which
is usually available with modern electron microscopes). Such a stage allows the saving
of the position of the structure of interest in each serial section during the EM observa-
tion, and then to rapidly return to this position for image acquisition or evaluation of
particular details of its morphology. Different designs of 3D reconstruction software
(e.g., Amira, Volocity, Imaris) can be used further to align the images of the organelle
of interest and to build its 3D model.
If the electron microscope is equipped with a tomography stage, thicker (200–300 nm)
serial sections can be used in combination with EM tomography to analyze the 3D
architecture of the structure of interest. Usually such a tomography approach gives
more accurate reconstruction of 3D organization of organelles (with 3–4 nm resolution)

than the usual serial sections.
IV. Discussion
Despite its apparent complexity, the above outlined protocol for preembedding
CLEM remains probably the simplest, cheapest, and most robust way to take an EM
snapshot of a particular structure or a rare event observed in a live cell. Even such
transient events as fusion of post-Golgi transport carriers with the plasma mem-
brane have been captured for EM analysis using this approach. Indeed, despite the
advances in cryofixation technology, the time gap of 4–5 s between the last frame of
a time-lapse sequence and the moment of cell freezing still represents an obstacle for
CLEM analysis of rapidly moving objects in a live cell (Verkade, 2008). Given that
the speed of movement of intracellular structures can be as high as dozen microns per
second (Lippincott-Schwartz et al., 2000; Polishchuk et al., 2000), the few seconds
of transfer from the microscope to the freezing stage can be sufficient for significant
changes in organelle positioning and overall intracellular distribution. In contrast,
preembedding CLEM allows this gap to be avoided, as the immobilization and final
location of the structure of interest are recorded during the time-lapse observation.
However, it is definitely advisable to check whether the fixation conditions used
preserve the morphology of the structures being examined by comparing them in
samples fixed either chemically or using an “artifact-free” method such as high-
pressure freezing (Müller-Reichert, Srayko, Hyman, O’Toole, & McDonald, 2007;
Verkade, 2008).
Additional advantages of this preembedding CLEM protocol are that it does not
require costly EM equipment and that it can be performed in a very basic EM labora-
tory. When the user does not have access to fast-freezing, freeze substitution, and low
temperature embedding systems, preembedding CLEM remains the only way to com-
bine the analysis of the same structure in a live cell and under the electron microscope.
Furthermore, a routine ultramicrotome and a simple electron microscope (even without
a tomography stage) would be sufficient for preparation and analysis of serial plastic
sections, respectively, to study the ultrastructure of the organelle of interest.
On the other hand, however, an inability to label more than one protein for the EM
investigation as well as the limited 3D resolution of the serial sectioning approach
leaves significant room for improvement and optimization of this CLEM technique.
Firstly, the development of new probes that are suitable for transformation of the fluo-
rescent signal under the light microscope to an electron-dense signal in the EM stage
exemplifies the strongest demand of CLEM technology in general. According to a com-
mon consensus of CLEM users, such probes would significantly reduce the length of
the CLEM procedure on the expense of immuno-labeling. Unfortunately, photocon-
version of GFP fluorescence into a dark EM signal (Grabenbauer et al., 2005) did
not provide the expected benefits for cell biologists (see Section I), while the newly
engineered miniSOG protein (Shu et al., 2011) has yet to prove its ability to survive
upon extensive illumination during time-lapse experiments in live cells.
Secondly, on the light microscopy side, an attractive idea would be to combine CLEM
with the novel rapidly developing super-resolution light microscopy approaches, such
as STED or STORM, for example (Toomre & Bewersdorf, 2010). The higher resolu-
tion power of these advanced imaging technologies will allow identification of novel
structural features in live organelles that would be worth to capture and investigate with
CLEM. On the other side, via comparisons of light microscopy and EM images of the
same structure, CLEM will serve to verify how far these super-resolution technologies
can go beyond the resolution limits of light microscopy. In this direction, the resolution
power of 4Pi microscopy has already been tested using our CLEM protocol (Perinetti
et al., 2009).
Thirdly, given that the EM retracing of the cell and the structure of interest fre-
quently takes significant time and requires extensive EM expertize, the automation of
this procedure though to simplify significantly the whole CLEM process. Recently,
a software algorithm based on an innovative image processing toolbox was indeed
reported to help significantly in the correlation of light microscopy and EM images of
immuno-labeled cryosections (Vicidomini et al., 2008). Moreover, work of companies
like FEI and Zeiss on the development of stages, holders, and software support is help-
ing in the observation of the same point within specimens at both the light microscopy
and EM levels. The integration of the above correlation technologies with live-cell
imaging still needs to be achieved to make the CLEM protocol less complicated.
Finally, we believe that development of dual-beam EM machines for 3D imaging
of large volumes in either cells or tissues will allow the trickiest parts of the CLEM
procedure to be circumvented; i.e., the cutting of the serial sections and their collec-
tion on the slots. Modern dual-beam microscopes provide the opportunity to remove
5–6 nm layers from the surface of a specimen using focused ion beam milling, com-
bined with the outstanding imaging of back-scattered electrons with high-resolution
scanning EM. This enables the user to reconstruct relatively large volumes of bio-
logical specimens (up to several hundreds of µm3) with isotropic 5–6 nm resolution
without the use of the ultramicrotome. Of note, even such tiny structures as synaptic
vesicles in brain tissue have been perfectly resolved and reconstructed using this new
3D technology (Knott, Rosset, & Cantoni, 2011; Sisková, Page, O’Connor, & Perry,
2009). Such resolution levels appear to be sufficient to analyze the architecture of
various intracellular structures and organelles, and therefore it appears to be worth of
integrating this into the CLEM procedure to bypass the preparation of the serial sec-
tions, and their further analysis under the electron microscope and alignment for 3D
reconstruction.
Thus, we hope that the combination of advanced light microscopy methods with
new fluorescent probes and the quickly developing 3D EM approaches will increase
the sensitivity and range of the questions that can be answered using the CLEM tech-
nology. This should thus make the use of CLEM more appealing to a number of scien-
tists in the cell biology field.
V. Summary
CLEM remains a unique technology to visualize structures of interest or specific/

rare events in a live cell and then to capture them for detailed EM analysis. Therefore,
instead of having just a static EM picture of the sliced organelles, the cell biologist can
analyze the specific structure under the electron microscope knowing in detail its previ-
ous history. The protocol outlined here provides a detailed guide for time-lapse imaging,
fixation, labeling, and EM processing of CLEM specimens, to achieve reliable results
within a relatively short time and to avoid frustrating and costly drawbacks. Therefore,
we believe that this method should be expanded through the cell biology community,
and that it serves as a platform for further CLEM developments and improvements.
Acknowledgments
We would like to thank C.P. Berrie for critical reading of the manuscript. This work
was supported by Telethon Grants GTF08001, TGM11CB4 and by AIRC Grant IG
10233.
References
Altan-Bonnet, N., Phair, R. D., Polishchuk, R. S., Weigert, R., & Lippincott-Schwartz, J. (2003). A role
for Arf1 in mitotic Golgi disassembly, chromosome segregation and cytokinesis. Proceedings of the
National Academy of Sciences of the United States of America, 100, 13314–13319.
high-pressure freezing and correlative light–electron microscopy. Seminars in Cell & Developmental
Biology, 20, 910–919.
Caplan, S., Naslavsky, N., Hartnell, L. M., Lodge, R., Polishchuk, R. S., Donaldson, J. G., et al. (2002).
A tubular EHD1-containing compartment involved in the recycling of major histocompatibility complex
class I molecules to the plasma membrane. The EMBO Journal, 21, 2557–2567.
Dahl, R., & Staehelin, L. A. (1989). High-pressure freezing for the preservation of biological structure:
theory and practice. Journal Of Electron Microscopy Technique, 13, 165–174.
Grabenbauer, M., Geerts, W. G. C., Fernadez-Rodriguez, J., Hoenger, A., Koster, A. J., & Nilsson, T. (2005).
Correlative microscopy and electron tomography of GFP through photooxidation. Nature Methods, 2,
857–862.
Hayat, M. A. (Ed.), (1987). Correlative microscopy in biology: Instrumentation and methods. New York:
Academic Press.
Knott, G., Rosset, S., & Cantoni, M. (2011). Focussed ion beam milling and scanning electron microscopy
of brain tissue. Journal of Visualized Experiments, 53, doi:10.3791/2588. pii: 2588.
Lippincott-Schwartz, J., Roberts, T. H., & Hirschberg, K. (2000). Secretory protein trafficking and organelle
dynamics in living cells. Annual Review of Cell and Developmental Biology, 16, 557–589.
Marra, P., Maffucci, T., Daniele, T., Tullio, G. D., Ikehara, Y., Chan, E. K., et al. (2001). The GM130 and
GRASP65 Golgi proteins cycle through and define a subdomain of the intermediate compartment. Nature
Cell Biology, 12, 1101–1113.
Mironov, A. A., Mironov, A. A., Jr., Beznoussenko, G. V., Trucco, A., Lupetti, P., Smith, J. D., et al. (2003).
ER-to-Golgi carriers arise through direct en-bloc protrusion and multistage maturation of specialized ER
exit domains. Developmental Cell, 5, 583–594.
Müller-Reichert, T., Srayko, M., Hyman, A., O’Toole, E. T., & McDonald, K. (2007). Correlative light and
electron microscopy of early Caenorhabditis elegans embryos in mitosis. Methods in Cell Biology, 79,
101–119.
Oorschot, V., de Wit, H., Annaert, W. G., & Klumperman, J. (2002). A novel flat-embedding method to pre-
pare ultrathin cryosections from cultured cells in their in-situ orientation. The Journal of Histochemistry
and Cytochemistry, 50, 1067–1080.
Pelham, H. R. (1997). Membrane transport. Green light for Golgi traffic. Nature, 389, 17–19.
Perinetti, G., Müller, T., Spaar, A., Polishchuk, R., Luini, A., & Egner, A. (2009). Correlation of 4Pi and
electron microscopy to study transport through single Golgi stacks in living cells with super resolution.
Traffic, 10, 379–391.
Polishchuk, R. S., Di Pentima, A., & Lippincott-Schwartz, J. (2004). Delivery of raft-associated, GPI-
anchored proteins to the apical surface of polarized MDCK cells by a transcytotic pathway. Nature Cell
Biology, 6, 297–307.
Polishchuk, E. V., Di Pentima, A., Luini, A., & Polishchuk, R. S. (2003). Mechanism of constitutive export
from the Golgi: bulk flow via the formation, protrusion, and en bloc cleavage of large trans-Golgi net-
work tubular domains. Molecular Biology of the Cell, 14, 4470–4485.
Polishchuk, R. S., Polishchuk, E. V., Marra, P., Alberti, S., Buccione, R., & Luini, A. (2000). Correlative
light-electron microscopy reveals the tubular-saccular ultrastructure of carriers operating between Golgi
apparatus and plasma membrane. The Journal of Cell Biology, 148, 45–58.
Polishchuk, R. S., San Pietro, E., Di Pentima, A., Teté, S., & Bonifacino, J. S. (2006). Ultrastructure of
long-range transport carriers moving from the trans-Golgi network to peripheral endosomes. Traffic, 7,
1092–1103.
Shu, X., Lev-Ram, V., Deerinck, T. J., Qi, Y., Ramko, E. B., Davidson, M. W., et al. (2011). A genetically
encoded tag for correlated light and electron microscopy of intact cells, tissues and organisms. PLoS
Biology, 9, e1001041.
Sisková,, Z., Page, A., O’Connor, V., & Perry, V. H. (2009). Degenerating synaptic boutons in prion disease:
microglia activation without synaptic stripping. American Journal of Pathology, 175, 1610–1621.
Svitkina, T. M., Verkhovsky, A. B., & Borisy, G. G. (1995). Improved procedures for electron microscopic
visualization of the cytoskeleton of cultured cells. Journal of Structural Biology, 115, 290–303.
Takizawa, T., Suzuki, K., & Robinson, J. M. (1998). Correlative microscopy using FluoroNanogold on ultra-
thin cryosections. Proof of principle. The Journal of Histochemistry and Cytochemistry, 46, 1097–1102.
Toomre, D., & Bewersdorf, J. (2010). A new wave of cellular imaging. Annual Review of Cell and Devel-
opmental Biology, 26, 285–314.
van Rijnsoever, C., Oorschot, V., & Klumperman, J. (2008). Correlative light-electron microscopy (CLEM)
combining live-cell imaging and immunolabelling of ultrathin cryosections. Nature Methods, 5, 973–980.
Vicidomini, G., Gagliani, M. C., Canfora, M., Cortese, K., Frosi, F., Santangelo, C. , et al. (2008). High data
output and automated 3D correlative light-electron microscopy method. Traffic, 9, 1828–1838.
CHAPTER 3
Correlative Fluorescence and Transmission

Electron Microscopy in Tissues
Toshihiro Takizawa* and John M. Robinson†
* Department of Molecular Anatomy, Nippon Medical School, Tokyo 113-8602, Japan
† Department of Physiology and Cell Biology, Ohio State University, Columbus, OH 43210, USA
Abstract
I. Introduction
II. Correlative Microscopy: Reporter Systems
A. 3,3’-Diaminobenzidine Tetrahydrochloride
B. Quantum Dots
C. Colloidal Gold
D. FluoroNanogold
E. Integrated Fluorescence and Electron Microscopy
III. Correlative Microscopy of Tissues
A. Ultrathin Cryosections
B. Array Tomography
C. Serial Block Face Scanning Electron Microscopy
IV. Conclusions
References
Abstract
Correlative microscopy has meant different things over the years; currently, this
term refers to imaging the same exact structures with two or more imaging modali-
ties. This commonly involves combining fluorescence and electron microscopy. Much
of the recent work related to correlative microscopy has been done using cell culture
models. However, many biological questions cannot be addressed in these models, but
require instead the 3-dimensional organization of cells found in tissues. Herein, we
discuss some of the issues related to correlative microscopy of tissues including the
major reporter systems presently available for correlative microscopy. We present data
from our own work in which we have focused on the use of ultrathin cryosections of
tissues as the substrate for immunolabeling to combine immunofluorescence and elec-
tron microscopy of the same sub-cellular structures.
38 Toshihiro Takizawa et al.
I. Introduction
The term correlative microscopy, as used herein, refers to imaging the same exact
biological structure (e.g., a sub-cellular structure) using two or more imaging modalities.
To be beneficial, the data obtained from correlative microscopy should be more informa-
tion rich than that obtained from any individual imaging platform. A number of different
types of correlative microscopy have been used to address biological questions. How-
ever, the most commonly used type of correlative microscopy combines fluorescence
and transmission electron microscopy. There are several recent reviews dealing with dif-
ferent aspects of correlative fluorescence and electron microscopy (Brown & Verkade,
2010; Cortese, Diaspro, & Tacchetti 2009; Jahn et al., 2012; Mironov & Beznoussenko,
2009; Modla & Czymmek, 2011; Murphy et al., 2011; Robinson & Takizawa, 2009;
Vicidomini et al., 2010).
A number of fluorescent markers have been applied to correlative microscopy or have
the potential to be useful for such studies. These include fluorescent proteins such as green
fluorescent protein, a large variety of small fluorescent molecules, and quantum dots. The
advent of these various “tracers” has been a boon to studies using live cell imaging to mon-
itor dynamic cellular processes in real time (reviewed in Caplan, Niethammer, Taylor, &
Czymmek, 2011; Jyoti, Simons, & Simons, 2004; Lippincott-Schwartz & Patterson, 2003;
Yuste, 2005). Correlative light and cryo-electron tomography is now a developing area of
research (Kukulski et al., 2011; Lučić, Leis, & Baumeister, 2008; Patla et al., 2010).
While imaging with fluorescence microscopy and thin-section transmission elec-
tron microscopy (TEM) that currently predominates, other forms of correlative micros-
copy exist. For example, combining live-cell phase contrast optics with thin section
TEM has been important in studying mitosis (Kapoor et al., 2006; Rieder & Cassels,
1999). In this case, morphological features were tracked over time in living cells; cells
were then fixed and processed for observation by TEM. In this way, the same features
observed in living cells were correlated with higher resolution images from electron
microscopy. In another example, fibrinogen receptor behavior was monitored by com-
bining video-enhanced differential interference contrast light microscopy of colloi-
dal gold-labeled fibrinogen in living platelets with high-resolution scanning electron
microscopy (SEM) (Olorundare, Simmons, & Albrecht, 1992). The live cell images
were then correlated with the SEM images. Live-cell imaging of macrophages cul-
tured on formvar-coated EM grids exposed to a fluorescent marker of endocytosis was
combined with whole mount TEM (Luo & Robinson, 1992). Following live-cell image
acquisition, the macrophages were fixed and subsequently incubated in a cerium-based
enzyme cytochemical reaction medium for the localization of acid phosphatase activ-
ity (Robinson & Karnovsky, 1983). It was found that the fluorescent probe reached the
acid phosphatase-positive lysosomal compartments. The electron-dense cytochemical
reaction product was readily detected in macrophage whole mounts and did not require
thin sectioning. These examples serve to demonstrate that many forms of correlative
microscopy are available.
A limitation of correlative microscopy utilizing optical microscopes has been on
spatial resolution, the so-called diffraction limit. However, recent years have seen
3. Correlative Fluorescence and Transmission Electron Microscopy in Tissues 39
d ramatic increases in the resolving power in fluorescence microscopy. Several differ-

ent approaches to increasing resolution in fluorescence microscopy have been utilized.
The major types of super-resolution microscopy are structured illumination micros-
copy (Gustafsson, 2000), stimulated emission depletion microscopy (Westphal & Hell,
2005), photoactivation localization microscopy (Betzig et al., 2006), and stochastic
optical reconstruction microscopy (Rust, Bates, & Zhuang, 2006). Variations of these
methods have also been developed. A fuller discussion of super-resolution fluorescence
microscopy is beyond the scope of this chapter, but it will be interesting to see how
super-resolution fluorescence microscopy contributes to correlative microscopy going
forward. Commercial versions of these technologies are available and rapid progress
in many areas is to be expected. Indeed, ultrathin sections of Caenorhabditis elegans
were imaged by super-resolution fluorescence microscopy and subsequently by elec-
tron microscopy in a correlative study (Watanabe et al., 2011).
While super-resolution fluorescence microscopy provides important new tools for
biologists, there are still limitations to this technology. Despite these improvements
the electron microscope continues to have the capability to resolve structures smaller
than those resolvable by super-resolution fluorescence microscopy. Another important
consideration is that with fluorescence microscopy one only observes the fluorescence
signals. What is missing is the reference space in which the fluorescence can be placed.
Griffiths (2001) refers to the reference space as the unlabeled structures in the cell or
tissue that are not seen and do not contribute to the analysis in fluorescence microscopy.
Combining fluorescence microscopy and electron microscopy in a correlated manner is
an ideal way to understand fluorescence signals within their reference space.
II. Correlative Microscopy: Reporter Systems
Light from fluorescence sources, observable in the fluorescence microscope, is

not detected in the electron microscope. Thus, to combine fluorescence and electron
microscopy and to achieve correlative microscopy, there needs to be (A) a means to
convert the fluorescence signal into an electron dense product observable with the elec-
tron microscope or (B) a probe that possesses elements detectable with each type of
microscope. We discuss briefly four different approaches to this issue.
A. 3,3’-Diaminobenzidine Tetrahydrochloride
Graham and Karnovsky (1966) introduced 3,3’-diaminobenzidine tetrahydrochlo-
ride (DAB) as a probe for ultrastructural localization. There has been widespread use
of DAB in optical (brightfield) and transmission electron microscopy in the interven-
ing years. This method relied upon DAB precipitating at sites of peroxidase activity.
This could be for detecting endogenous peroxoidatic activity (e.g., myeloperoxidase
in neutrophils) or exogenous peroxidatic activity (e.g., horseradish peroxidase conju-
gated to antibodies for immunolabeling). Indeed, this approach continues to be in use.
In the localization of peroxidase activity, the precipitated DAB has a brownish color
detectable with brightfield microscopy. What makes DAB so useful is that following
exposures to osmium tetroxide the brown precipitate is converted to an electron dense
product known as osmium black. Thus, in its own right, DAB can be used for correla-
tive microscopy: brightfield optical and electron microscopy of the same specimen.
Indeed, this approach was used to study carrier structures moving between the Golgi
and the plasma membrane in cultured cells (Polishchuck et al., 2000).
It was also found that some sources of fluorescent light (e.g., the fluorescent dye
Lucifer yellow) could also precipitate DAB through a photooxidation reaction (some-
times referred to as photoconversion) (Maranto, 1982). This was a major innovation
and enabled fluorescence microscopy and electron microscopy to be combined to image
the exact same structures. Fluorescence-based photooxidation of DAB greatly adds to
our capacity to carry out correlative microscopy (e.g., Dantuma, Pijnenburg, Diederen,
& Van der Horst, 1998; Deerinck et al., 1994; Gaietta et al., 2002; Grabenbauer et al.,
2005; Harata, Ryan, Smith, Buchanan, & Tsien, 2001; Lübke, 1993: Martin & Pagano,
1994; Meisslitzer-Ruppitsch et al., 2008). Aspects of the chemistry of photooxidation
of DAB by the dye eosin have been reported (Natera, Massad, Amat-Guerri, & Garcia,
2011). A recent development in this area was the introduction of a genetically encoded
tag for proteins called miniSOG (this stands for mini Singlet Oxygen Generator) (Shu
et al., 2011). This is a 106 amino acid fluorescent peptide derived from Arabidopsis
phototropin 2. Proper illumination of miniSOG in the presence of DAB leads to for-
mation of a DAB precipitate that can be imaged by electron microscopy after osmium
tetroxide treatment and embedding the sample in resin. This is potentially a major find-
ing and may, as the authors suggest, do for electron microscopy what Green Fluorescent
Protein did for fluorescence microscopy.
B. Quantum Dots
Nanometer-sized colloidal semiconductor crystalline particles, known as quantum
dots, have received attention for their potential use in correlative fluorescence and elec-
tron microscopy. A number of fluorescence-based bioassays, including fluorescence
microscopy have been developed (Chan, Maxwell, Gao, Bailey, & Nie, 2002; Sutherland,
2002; Watson, Wu, & Bruchez, 2003). Quantum dots are potentially useful due to their
robust fluorescence. The color of fluorescence emitted from quantum dots is related to
their size (Bruchez, Moronne, Gin, Weiss, & Alivisatos, 1998). To be useful as probes
in microcopy, conjugation of biological molecules (e.g., antibodies, lectins) that retain
biological activity is necessary. Successful conjugation of a number of biomolecules to
quantum dots has been achieved (Alivisatos, Gu, & Larabell, 2005).
Quantum dots are bifunctional probes that, in addition to being fluorescent, can be
observed in the electron microscope due to their electron density (Alivisatos et al.,
2005). Therefore, quantum dots have the potential to be useful probes in correlative
fluorescence and electron microscopy. Live cell and whole animal imaging has been
achieved using quantum dots. Tumor vasculature in animals has been imaged using
quantum dots (Ackerman, Chan, Laakkonen, Bhatia, & Ruoslahti, 2002). Whole animal
imaging as well as fluorescence and electron microscopy have been reported using
quantum dots coated with different polymers (Ballou, Lagerholm, Ernst, Bruchez, &
Waggoner, 2004). A comparison of quantum dots as probes for correlative microscopy
in fluorescence, conventional, and energy-filtered transmission electron microscopy
has been reported (Nisman, Dellaire, Ren, Li, & Bazett-Jones, 2004). The utility of
biological labeling with quantum dots has been reviewed (Deerinck, Giepmans, Smarr,
Martone, & Ellisman, 2007; Giepmans, Deernick, Smarr, Jones, & Ellisman, 2005).
A major concern with particulate probes relates to their ability to penetrate into
cells or tissues. The most commonly used particulate probe in biological labeling,
colloidal gold, penetrates poorly into fixed cells and tissues (Takizawa & Robinson,
1994). Quantum dots, on the other hand, exhibit modest penetration into fixed and
permeabilized cells when coated with antibody (Deerinck, 2008). This issue has also
been addressed in living cells. Using mercaptopropionic acid-functionalized quantum
dots, EC0156 cells were incubated with 16 nm quantum dots with or without antibody
coating for 24 h. In the presence of the nuclear stain DAPI, the uncoated quantum dots
reached the nucleus of these cells while the antibody-coated quantum dots did not enter
the nucleus (Xu et al., 2008). While quantum dots appear to penetrate better than col-
loidal gold particles, the degree of penetration should be a consideration when using
these probes. Correlative fluorescence microscopy and scanning transmission electron
microscopy of quantum dot-labeled cells maintained in liquid have also been reported
(Dukes, Peckys, & de Jonge, 2010).
C. Colloidal Gold
Colloidal gold particles have been used extensively for biological labeling and elec-
tron microscopy with both SEM and TEM (Roth, 1996). It would seem that colloidal
gold particles might be useful for correlative fluorescence and electron microscopy if
a fluorophore was added to the colloidal gold. However, this approach has not been
widely used since colloidal gold tends to quench fluorescence (De Brabander, Geuens,
Nuydens, Moeremans, & De May, 1985; Goodman, Park, & Albrecht, 1991). Dulkeith
et al. (2005) report that gold nanoparticles quench fluorescence by phase-induced
radiative rate suppression. In immunolocalization studies, an alternative to having the
fluorophore close to the gold particle (where quenching will occur) is to conjugate the
primary antibody to the colloidal gold particle and then to detect the primary antibody
with a fluorophore-labeled secondary antibody (Kandela, Bleher, & Albrecht, 2007,
2008). Spacing the fluorophore away from the gold particle in this manner preserves
some of the fluorescence signal. While this enables use of colloidal gold and fluores-
cence for correlative microscopy, the potential for quenching and the poor penetration
of colloidal gold places limitations on this approach.
D. FluoroNanogold
Ultrastructural immunocytochemistry has benefited greatly from the use of colloidal
gold probes (Roth, 1996). An important feature of colloidal gold is that the particles can
be made in the laboratory in a variety of sizes (Frens, 1973; Geoghegan & Ackerman,
1977: Slot & Geuze, 1981). Particle sizes most commonly used are 5, 10, and 15 nm.
Having particles of different sizes permits double and triple labeling experiments. How-
ever, it has been shown in a number of studies that there is an inverse relationship
between the size of colloidal gold particles and labeling efficiency [5 nm > 10 nm > 15 nm]
(e.g., Robinson, Takizawa, & Vandré, 2000). As noted above, colloidal gold has not
been widely used for correlative microscopy.
Gold cluster compounds have been developed as an alternative labeling probe to
colloidal gold particles (Hainfeld, 1987; Hainfeld & Furuya, 1992). These probes,
known as Nanogold, are chemical compounds and not metallic particles like colloidal
gold. In immunocytochemistry, the labeling efficiency obtained with Nanogold was
greater than that observed with colloidal gold. Additionally, we found that Nanogold
penetrates further into biological samples than colloidal gold, which penetrates poorly
(Takizawa & Robinson, 1994). Nanogold is very small, 1.4 nm, and is not readily vis-
ible in standard electron microscopy of cells or tissues. Thus, for routine imaging the
size of Nanogold needs to be increased. Fortunately, this can be accomplished with
autometallographic methods (e.g., silver enhancement). This does not represent a major
problem since this procedure is relatively rapid and reproducible. Silver enhancement
methods have been reviewed (Danscher, Hacker, Hauser-Kronberger, & Grimelius,
1995; Hacker et al., 1995).
The question of whether Nanogold could be used for correlative fluorescence and
electron microscopy has been addressed. Conjugation of fluorophores to Nanogold
labeling reagents has been successful (Powell et al., 1997). It was found that in this
probe there was no quenching of fluorescence; this represents a major advantage com-
pared to colloidal gold (Powell, Halsey, & Hainfeld, 1998). The fluorophore-labeled
Nanogold labeling reagents are known as FluoroNanogold (FNG). Successful use of this
bifunctional reagent for correlative fluorescence and electron microscopy was achieved
using ultrathin cryosections as the labeling substratum in a “proof of principle” experi-
ment (Takizawa, Suzuki, & Robinson, 1998). In this case, immunocytochemistry was
used to localize the enzyme myeloperoxidase (MPO) in human neutrophils (Fig. 1). In
brief, ultrathin cryosections were collected on formvar-coated EM “finder grids” and
incubated with a primary antibody to MPO and subsequently with the FNG secondary
labeling reporter. The EM grid was mounted between a glass microscope slide and
cover glass and fluorescence microscopy images were collected. The temporary slide
preparation was disassembled and the grid was incubated in the silver enhancement
solution to render the Nanogold visible. The dried grids were then examined with the
electron microscope and the same exact structures examined by fluorescence micros-
copy were located, with the aid of the finder grid. The fluorescence and EM images
were then correlated. In so far as we know, this was the first time that fluorescence and
electron microscopy of the same objects was achieved in an ultrathin section.
E. Integrated Fluorescence and Electron Microscopy

While not a reporter system, another technological approach to correlative fluores-
cence and electron microscopy involves integrating a fluorescence microscope into an
Fig. 1 Myeloperoxidase (MPO) was localized in an ultrathin cryosection of isolated human neutrophils using
a primary antibody to MPO followed by a FNG-labeled secondary antibody. The section was initially imaged
by fluorescence microscopy followed by electron microscopy after a silver-enhancement reaction was con-
ducted. (A) Lower-magnification fluorescence image is shown. An eosinophil (e) and arrow serve as reference
points and grid bars are indicated (*). (A’) The same exact portion of the ultrathin section is viewed by electron
microscopy at the same magnification. Scale bar = 5 µm. (B) A portion of the cell indicated by the arrow in
panel A is imaged at higher magnification. A group of five fluorescent objects are indicated. (B’) The same exact
portion of the ultrathin cryosection was imaged at the same magnification by electron microscopy. The same
five structures shown in panel B are membrane-bounded organelles detected by the presence of silver-enhanced
FNG. These correspond to the so-called azurophil granules of these cells. Myeloperoxidase-negative granules
are indicated by arrowheads. Scale bar = 500 nm. [Reproduced by permission from Robinson et al. (2000). For
experimental details see Takizawa et al. (1998).]
electron microscope (Agronskaia et al., 2008). Thus, the same instrument is used for
both fluorescence and TEM; this is accomplished by rotating the specimen into the two
types of beams. With this approach, the fluorescence signal was correlated with cell or
tissue ultrastructure; no label was observed in the electron microscope. While this is an
intriguing system, a fuller discussion is beyond the scope of this chapter.
III. Correlative Microscopy of Tissues
Correlative microscopy studies cited in the preceding parts of this chapter were
heavily weighted toward analysis of single cells, freshly isolated or maintained in cell
culture. Cell culture model systems provide important insights for our understanding
of cellular structure and function. However, not all important questions are address-
able in these simpler model systems. Indeed, many questions can only be approached
when the cells of animal and human tissues retain their native 3-dimensional organiza-
tion. Since animal and human tissues are considerably more complex than single cells,
some methods used with cultured cells may not apply to tissues. The use of animal and
human tissues has more experimental constraints than does the use of cultured cells.
An obvious example is that of molecular manipulations such as tagging a target protein
with a fluorescent protein; these manipulations are readily done in cell culture and in
some experimental animals but cannot be done with humans for ethical reasons.
As noted, correlative microscopy studies in tissues are less numerous than such stud-
ies in cells. There are, however, excellent correlative fluorescence and electron micro-
scopic studies. We will mention a few of these to illustrate different routes taken by
several investigators. Green fluorescent protein (GFP) expression in proliferating neu-
rons was accomplished using a Moloney murine virus-based system (Toni et al., 2007).
The GFP-positive neurons were subsequently injected with the fluorescent dye Lucifer
yellow; the DAB photoconversion reaction was then conducted. The tissue was then
treated with osmium tetroxide for visualization at the ultrastructural level. Another
example of using GFP and DAB photoconversion in nervous tissue is from the work
of Nikonenko, Boda, Alberi, and Muller (2005). In this case, slice cultures of rat brain
were transfected using a biolistic procedure. Both of these studies demonstrate that
correlative microscopy enables examination of identified cells within the brain at high
resolution. In one study, different regions of brains of songbirds were injected with
different fluorescent molecules. This enabled locating the regions of interest follow-
ing fixation and dissection. Selected regions were subsequently processed for electron
microscopy. One of the injected dyes, Lucifer yellow, was detected by immunofluores-
cence using an antibody to this dye. The sections were imaged by electron microscopy
and the ultrastructure of the tissue was aligned with the fluorescence images (Oberti,
Kirschmann, & Hahnloser, 2010). Correlative fluorescence and electron microscopy of
C. elegans expressing fluorescent proteins was conducted (Watanabe et al., 2011). As
noted earlier, in this study ultrathin sections of resin-embedded tissue were imaged by
super-resolution microscopy. The sections were subsequently examined by scanning
electron microscopy. Correlative light and electron microscopy has been used to moni-
tor proteasome particle-rich structures in epithelial tumors (Necchi et al., 2011).
A. Ultrathin Cryosections
The need to visualize the interior of cells and tissues has been an issue for biologist
since the earliest days of cytology and histology. Moreover, methods for preparation of
biological samples to represent the most true to life view of cells and tissues have been
a constant quest. These issues became even more acute with the advent of biological
electron microscopy. A very large number of studies from the late 1940s through to the
present day have endeavored to develop methods to peer into cells and tissues and to
image them in the most physiologically relevant state possible at high resolution. The
many approaches developed over the years have been discussed in numerous books,
review articles, and scientific papers; these are too numerous to discuss.
The desire of biologists and microscopists to examine ultrathin cryosections goes
back to earlier periods of biological electron microscopy (Bernhard & Leduc, 1967;
Christensen, 1971; Fernandez-Moran & Dahl, 1952; Fernandez-Moran & Finean,
1957; Iglesias, Berner, & Simard, 1971). An important improvement in sample prepa-
ration was the use of sucrose as an embedding material for cutting ultrathin cryosec-
tions (Tokuyasu, 1973). Sucrose-embedded tissues were then used for immunoelectron
microscopy with ferritin as the reporter system (Painter, Tokuyasu, & Singer, 1973).
These latter two papers launched an era in which immunoelectron microscopy pro-
vided a great deal of important data to increase our knowledge of cell structure and
function. The inclusion of colloidal gold as the particulate reporter was also a major
advancement in the use of ultrathin cryosections for immunoelectron microscopy
(Roth, 1996).
Ultrathin cryosections (50–100 nm) of cells and tissues have been used for over
40 years for electron microscopy. However, these sections have seldom been used for
immunofluorescence microscopy. Generally, when immunofluorescence and electron
microscopy were combined semithin (200–500 nm) cryosections sections were used
for fluorescence microscopy and an adjacent ultrathin (50–100 nm) cryosection section
was used for electron microscopy (e.g., Geuze, Slot, Strous, Hasilik, & von Figura,
1984; Van der Wel, Fluitsma, Dascher, Brenner, & Peters, 2005). We have shown that
the same ultrathin cryosection can be imaged by fluorescence and electron microscopy
(Robinson, Takizawa, Pombo, & Cook, 2001; Takizawa et al., 1998).
We have been proponents for using ultrathin cryosections of cells and tissues
for high-resolution immunofluorescence microscopy (Mori et al., 2006; Robinson,
Ackerman, Vandré, 2009; Robinson et al., 2000, 2001; Takizawa & Robinson, 2003a,
2003b; Takizawa et al., 1998; Takizawa, Anderson, & Robinson, 2005; Vandre et al.,
2007). An advantage of ultrathin cryosections in fluorescence microscopy is an appar-
ent improvement in the z-axis resolution in comparison to conventional confocal
microscopy. Using human placenta as the model tissue, we carried out 3-color fluores-
cence microscopy for the localization of CD31 and caveolin 1 (CAV-1) using antibody
methods and DNA using DAPI labeling. This was done on both 5 µm cryostat sections
and ultrathin cryosections. The sections were imaged with a conventional fluorescence
microscope and a confocal microscopy. As expected, there was significant out-of-focus
in the 5 µm sections when viewed with the conventional fluorescence microscope; this
was greatly improved with the confocal microscope (Fig. 2). Placental endothelial cells
labeled positively for both CD31 and CAV-1. In confocal optical sections of this tis-
sue, it was often difficult to distinguish areas positive for CD31 from those positive for
CAV-1. One reason for this is that there are regions of these endothelial cells where
the luminal and abluminal plasma membranes are very close together. These closely
Fig. 2 The use of ultrathin cryosections for immunofluorescence microscopy leads to an apparent improvement in z-axis resolution.
Cryostat sections (5 µm) were cut as were ultrathin cryosections. Both types of sections were double-labeled to detect CAV-1 (red) with
an antibody specific for the CAV-1α isoform (Lyden, Anderson, & Robinson, 2002) and CD31 (green). Nuclei were stained with DAPI
(blue). The 5 µm section was examined with a conventional fluorescence microscope and 3-color fluorescence images were collected
(a–c). There is considerable out-of-focus fluorescence. These same structures were examined by confocal microscopy (e–g). There is
significant reduction in out-of-focus fluorescence compared to a–c. An ultrathin cryosection was examined by both conventional and
confocal fluorescence microscopy (only confocal is shown) (i–k). There is still further improvement in image quality in the ultrathin
cryosection. In the ultrathin section, individual caveolae are resolved (i); however, they are more difficult to resolve in the confocal
image (e). Additionally, in the ultrathin cryosection both the luminal and the abluminal plasma membranes are resolved even though
they are often very closely spaced (i–k); this is clearly seen in the higher-magnification insets (i–k). The single arrow indicates the region
where the insets were from while the double arrows indicate a portion of a pericyte. Resolving the closely spaced plasma membranes is
more difficult to achieve in the confocal images of the 5 µm (e–g). The corresponding differential interference contrast (DIC) images are
shown to illustrate the tissue morphology and to provide the reference space. In all images, the lumen of a capillary is indicated (*) as is
the edge of the syncytiotrophoblast (arrowheads). The scale bars = 10 µm (a–l) and 500 nm in the insets. [Reproduced with permission
from Mori et al. (2006). For experimental details see Mori et al. (2006). (See color plate.)
spaced membranes were often difficult to distinguish in confocal optical sections,

whereas they could typically be distinguished in the ultrathin cryosections (Fig. 2).
This enabled better separation of the CD31 and CAV-1 signals. We have proposed a
model to illustrate these differences (Fig. 3).
We developed the ultrathin cryosection approach to correlative fluorescence and
electron microscopy using isolated cells (see Fig. 1) (Takizawa et al., 1998). However,
we wanted to expand this approach to be usable with tissue samples as well. For this,
we have used human placenta as the model tissue (Takizawa & Robinson, 2003a).
Fig. 3 A model illustrating the localization patterns observed in Fig. 2. The drawings of capillary sections of
different thicknesses are meant to show the cryostat section of 5 µm, a confocal optical section of ∼0.5 µm, and
an ultrathin cryosection of 100 nm. The fluorescence microscopy patterns observed with each of these sections
are indicated. As in Fig. 2, CAV-1 labeling is in red and CD31 in green. The luminal and abluminal surfaces
are indicated (LS and AS). Objects, even a small one like caveolae (∼70 nm), can “stack” in the volume of the
confocal optical section, leading to difficulties in detecting individual caveolae. In the ultrathin cryosections
(a physical section), all of the fluorescence must come from this small volume. Thus, the merging of fluores-
cence from above or below is less likely and individual objects such as caveolae are readily detected. [Repro-
duced with permission from Mori et al. (2006).] (See color plate.)
The tissue was fixed with 4% paraformaldehyde in 0.1 M cacodylate buffer, pH 7.2,
containing 5% sucrose. The fixed tissue was washed and cut into small pieces. A small
volume of tissue suspension (200 µl) was mixed with an equal volume of 20% gelatin
in cacodylate buffer (the gelatin was 300 bloom) in a microfuge tube. The tube was
immediately centrifuged to concentrate the tissue pieces and the gelatin was solidified
by chilling on ice. The gelatin was removed from the tube and the portion containing
tissue was cut in pieces (1 × 1 × 3 mm) to fit onto specimen carrier pins. The pieces were
infused with sucrose in a step-wise manner until reaching 2.3 M sucrose. The pieces
were then mounted on the carrier pins and frozen in liquid nitrogen and stored in liquid
nitrogen until used.
Ultrathin cryosections (70–100 nm) were picked up on a droplet of 0.75% gelatin-
2.0 M sucrose. The pick-up solution contained 0.05% sodium azide as a preservative
(Griffith & Posthuma, 2002). The section was then transferred onto a formvar-coated
nickel EM grid that was stabilized with evaporated carbon. “Finder” grids were used to
facilitate locating specific regions when going from the fluorescence microscope to the
electron microscope. We have used successfully several different primary antibodies
but in each case the reporter system was FNG. Initially, fluorescence microscopy was
carried out by mounting the EM grid between a glass microscope slide and a glass
coverslip in a small drop of antifading medium (Takizawa & Robinson, 2000). Images
were then collected and their locations noted with the aid of the finder grid. The tem-
porary slide preparation was then disassembled and the grid was washed in several
changes of PBS. The ultrathin cryosection was further stabilized by fixation in 2%
glutaraldehyde in PBS for 30 min. The grids were then washed in distilled water prior
to a silver-enhancement reaction to make the Nanogold more readily visible in the
electron microscope. Prior to examination, the sections were contrast enhanced to
aid in visualization of cellular membranes. This was done using the positive-contrast
enhancement method we developed (Takizawa & Robinson, 2003c). In the electron
microscope, the regions of interest were located and images were collected. A correla-
tion between the fluorescence images and the electron microscope images could then
be done (Fig. 4). A very detailed account of all of these procedures has appeared in
Takizawa and Robinson (2006).
B. Array Tomography
Another promising approach to correlative microscopy of tissues is a technique
called array tomography (Micheva & Smith, 2007). This is an extension of previous
work using ultrathin sections for immunofluorescence microscopy (Mori et al., 2006;
Takizawa et al., 1998). However, in this case, instead of using ultrathin cryosections,
resin embedded tissues were employed. Nervous tissue was fixed and subsequently
embedded in LRWhite resin. Serial sections (50–200 nm for ultrathin or 400–1000 nm
for semithin) were mounted on microscope slides, immunolabeled, and observed by flu-
orescence microscopy. This approach has the advantage that collecting serial sections
Fig. 4 Correlative fluorescence and electron microscopy using ultrathin cryosections of human placenta as the
labeling substratum and FNG as the reporter system is shown. (A) Immunofluorescence localization of CAV-1
in part of an endothelial cell is shown. Individual spot-like structures are indicated (arrow). Larger fluorescent
objects (arrowhead) are multiple spots too close to each other to be well resolved. (B) The electron micrograph
is of the same area shown in panel A at the same magnification. The arrow and arrowhead indicate the same
structures shown in panel A. A portion of a grid bar is indicated (*). The section had been subjected to the silver-
enhancement reaction. Scale bar = 5 µm. (C) A higher magnification view of the area at the arrow in panel
A. Two spots are indicated. (D) The electron micrograph is of the same exact area shown in panel C and is at
the same magnification. The silver-enhanced FNG is evident a individual small dots of electron-dense material.
This demonstrates that the spots observed by fluorescence microscopy correspond to individual caveolae. Scale
bar = 500 nm. [Reproduced with permission from Takizawa and Robinson (2003b). For experimental details see
Takizawa and Robinson (2003b, 2006).]
for 3-dimensional reconstruction is more readily achieved with resin-embedded tissue

than with frozen sections. Additionally, the labeling antibodies were eluded and the
sections were probed again with other antibodies. Finally, the sections were imaged by
scanning electron microscopy following heavy metal staining. This permits correlative
fluorescence and electron microscopy. Array tomography has now been used in several
studies (e.g., Micheva & Bruchez, 2011; Micheva, Busse, Weiler, O’Rouke, & Smith,
2010; Oberti, Kirschmann, & Hahnloser, 2011). Of course, one could image individual
resin-embedded sections for correlative fluorescence and electron microscopy without
serial sectioning.
C. Serial Block Face Scanning Electron Microscopy

An emerging methodology for acquiring large 3-dimensional datasets of cells and
tissues is focused ion beam/scanning electron microscopy (FIB/SEM) with serial block
face/scanning electron microscopy (SBF/SEM) (e.g., Heymann et al., 2006; Merchan-
Perez, Rodriguez, Alonso-Nanclares, Schertel, & DeFelipe, 2009). This approach has
also been used for correlative microscopy where fluorescence imaging was combined
with FIB/SEM and SBF/SEM (Amer et al., 2009). The authors referred to this as cor-
relative volume electron microscopy. In that study, blood vessel formation was studied
in living zebrafish embryos in which endothelial cells expressed a fluorescent protein
(EGFP). At the appropriate time after monitoring the fluorescence, the embryo was fixed
and processed for EM. The resin-embedded embryo was then examined by FIB/SEM
and SBF/SEM. In this approach, the block face was imaged by backscatter detection in
the SEM and then a set amount (e.g., 50 nm) of the block face was milled away by the
focused ion beam to reveal a new block face for imaging. This process is repeated by
successive milling, imaging, milling, imaging, etc. This methodology has tremendous
potential in that relatively large volumes of tissue can be examined and reconstructed.
There is an alternative to FIB/SEM for SBF/SEM available. In this case, an ultrami-
crotome is mounted in the SEM specimen chamber and the block face is imaged after
an ultrathin section of defined thickness is removed. This process is repeated until
the desired volume of tissue has been examined (Denk & Horstmann, 2004). As with
FIB/SEM, serial block face sectioning and SEM imaging enables the examination of
relatively large volumes of the sample. Recently, this methodology has been combined
with two-photon confocal microscopy (Briggman, Helmstaedter, & Denk, 2011).
IV. Conclusions
A number of investigators have developed procedures to combine imaging systems,

such as fluorescence and electron microscopes. These are powerful methods for doing
correlative microscopy. In our work, we have focused on using ultrathin cryosections
of cells and tissues for immunolocalization and correlative microscopy. Using FNG
as the reporter system greatly facilitated this work. A diagrammatic summary of these
methods is given in Fig. 5.
Fig. 5 A summary of the methods used for high-resolution immunofluorescence microscopy and correlative
fluorescence and electron microscopy is shown in diagrammatic form. (A) Classically, semithin cryosections
were cut for immunofluorescence microscopy and adjacent ultrathin cryosections were cut for electron micros-
copy using colloidal gold probes. (B) For high-resolution immunofluorescence microscopy, ultrathin cryosec-
tions were mounted on glass coverslips and labeled with fluorescently tagged antibodies. (C) For correlative
immunofluorescence and electron microscopy, the ultrathin cryosections were mounted on EM grids. Follow-
ing immunolabeling, the fluorescence signal from FNG was captured and then the same exact structures were
observed by electron microscopy after the silver-enhancement reaction. (D) A model of a terminal villous of the
human placenta showing the endothelium (#1), a pericyte (#2) and the syncytiotrophoblast (#3) are shown. The
right side shows the possibility for false co-localization for two different organelles of similar size (shown in
red and green). The two different types of fluorescent objects could appear colocalized if stacked over or under
each other (indicated in yellow). (See color plate.)
Fluorescence-based imaging methods are widely used in cell biological research.

While providing much valuable information, a limitation of fluorescence microscopy
is that the fluorescence signals are out of context of the cell or tissue. Providing the
context in which to place the fluorescence is often important. Correlative microscopy in
which the fluorescence signals are placed in the context of cell or tissue ultrastructure
by combining fluorescence and electron microscopy is an important means to achieve
this goal. We anticipate that correlative microscopy will be an increasingly potent tool
in biological research.
References
Ackerman, M. E., Chan, W. C., Laakkonen, P., Bhatia, S. N., & Ruoslahti, E. (2002). Nanocrystal tar-
geting in vivo. Proceedings of the National Academy of Sciences of the United States of America, 99,
12617–12621.
Agronskaia, A. V., Valentijn, J. A., van Driel, L. F., van Driel, L. F., Schneijdenberg, C. T. W. M., Humbel,
B. M., et al. (2008). Integrated fluorescence and transmission electron microscopy. Journal of Structural
Biology, 164, 183–189.
Alivisatos, A. P., Gu, W., & Larabell, C. (2005). Quantum dots as cellular probes. Annual Review of Bio-
medical Engineering, 7, 55–76.
Amer, H. E., Mariggi, G., Png, K. M. Y., Genoud, C., Monteith, A. G., Bushby, A. J., et al. (2009). Imaging
transient blood vessel fusion events in zebrafish by correlative volume electron microscopy. PLoS One,
4, e7716 doi:10.1371/journal.pone.0007716.
Ballou, B., Lagerholm, B. C., Ernst, L. A., Bruchez, M. P., & Waggoner, A. S. (2004). Non-invasive imaging
of quantum dots in mice. Bioconjugate Chemistry, 15, 79–86.
Bernhard, W., & Leduc, E. H. (1967). Ultrathin frozen sections. I. Methods and ultrastructural preservation.
The Journal of Cell Biology, 34, 757–771.
Betzig, E., Patterson, G. H., Sougrat, R., Lindwasser, O. W., Olenych, S., Bonifacino, J. S., et al. (2006).
Imaging intracellular fluorescent proteins at nanometer resolution. Science, 313, 1642–1645.
Briggman, K. L., Helmstaedter, M., & Denk, W. (2011). Wiring specificity in the direction-selectivity circuit
of the retina. Nature, 471, 183–188.
Brown, E., & Verkade, P. (2010). The use of markers for correlative light electron microscopy. Protoplasma,
244, 91–97.
Bruchez, M., Moronne, M., Gin, P., Weiss, S., & Alivisatos, A. P. (1998). Semiconductor nanocrystals as
fluorescent biological labels. Science, 281, 2013–2016.
Caplan, J., Niethammer, M., Taylor, R. M., & Czymmek, K. J. (2011). The power of correlative microscopy:
multi-modal, multi-scale, multi-dimensional. Current Opinion in Structural Biology, 21, 686–693.
Chan, W. C. W., Maxwell, D. J., Gao, X., Bailey, R. E., & Nie, S. (2002). Luminescent quantum dots for
multiplexed biological detection imaging. Current Opinion in Biotechnology, 13, 40–46.
Christensen, A. K. (1971). Frozen thin section of fresh tissue for electron microscopy, with a description of
pancreas and liver. The Journal of Cell Biology, 51, 772–804.
Cortese, K., Diaspro, A., & Tacchetti, C. (2009). Advanced correlative light/electron microscopy: current
methods and new developments using Tokuyasu cryosections. The Journal of Histochemistry and Cyto-
chemistry, 57, 1103–1112.
Danscher, G., Hacker, G. W., Hauser-Kronberger, C., & Grimelius, L. (1995). Trends in autometallographic
silver amplification of colloidal gold particles. In M. A. Hayat (Ed.), Immunogold-silver staining: Prin-
ciples, methods, and applications (pp. 11–18). Boca Raton: CRC Press.
Dantuma, N. P., Pijnenburg, M. A., Diederen, J. H., & Van der Horst, D. J. (1998). Electron microscopic
visualization of receptor-mediated endocytosis of DiI-labeled lipoproteins by diamainobenzidine photo-
conversion. The Journal of Histochemistry and Cytochemistry, 46, 1085–1090.
De Brabander, M., Geuens, G., Nuydens, R., Moeremans, M., & De May, J. (1985). Probing microtubule-
dependent intracellular motility with nanometer particle video ultramicroscopy (nanovid ultramicros-
copy). Cytobios, 43, 273–283.
Deerinck, T. J., Martone, M. E., Lev-Ram, V., Green, D. P., Tsien, R. Y., Spector, D. L., et al. (1994). Fluores-
cence photooxidation with Eosin: a method for high resolution immunolocalization and in situ hybridiza-
tion detection for light and electron microscopy. The Journal of Cell Biology, 126, 901–910.
Deerinck, T. J., Giepmans, B. N., Smarr, B. L., Martone, M. E., & Ellisman, M. H. (2007). Light and electron
microscopic localization of multiple proteins using quantum dots. Methods in Molecular Biology, 374,
43–54.
Deerinck, T. J. (2008). The application of fluorescent quantum dots to confocal, multiphoton, and electron
microscopic imaging. Toxicologic Pathology, 36, 112–116.
Denk, W., & Horstmann, H. (2004). Serial block-face scanning electron microscopy to reconstruct three-
dimensional tissue nanostructure. PLoS Biology, 2, e329.
Dukes, M. J., Peckys, D. B., & de Jonge, N. (2010). Correlative fluorescence microscopy and scanning
transmission electron microscopy of quantum-dot-labeled proteins in whole cells in liquid. ACS Nano,
4, 4110–4116.
Dulkeith, E., Ringler, M., Klar, T. A., Feldman, J., Munoz Javier, A., & Parak, W. J. (2005). Gold nanopar-
ticles quench fluorescence by phase induced radiative rate suppression. Nano Letters, 5, 585–589.
Fernandez-Moran, H., & Dahl, A. O. (1952). Electron microscopy of ultrathin frozen sections of pollen
grains. Science, 116, 465–467.
Fernandez-Moran, H., & Finean, J. B. (1957). EM and low-angle x-ray diffraction studies of the nerve
myelin sheath. The Journal of Biophysical and Biochemical Cytology, 3, 725–748.
Frens, G. (1973). Controlled nucleation for the regulation of the particle size in monodisperse gold suspen-
sion. Nature (Physical Science), 241, 20–22.
Gaietta, G. M., Deerinck, T. J., Adams, S. R., Bouwer, J., Tour, O., Laird, D. W., et al. (2002). Multicolor and
electron microscopic imaging of connexin trafficking. Science, 296, 503–507.
Geoghegan, W. D., & Ackerman, G. A. (1977). Adsorption of horse-radish peroxidase, ovomucoid and anti-
immunoglobilin to colloidal gold for the indirect detection of conconavalin A, wheat germ agglutinin and
goat anti-human immunoglobulin G on cell surfaces at the electron microscopic level: a new method,
theory and application. The Journal of Histochemistry and Cytochemistry, 25, 1187–1200.
Geuze, H. J., Slot, J. W., Strous, G. J., Hasilik, A., & von Figura, K. (1984). Ultrastructural localization of the
mannose-6-phosphatae receptor in rat liver. The Journal of Cell Biology, 98, 2047–2054.
Giepmans, B. N., Deerinck, T. J., Smarr, B. L., Jones, Y. Z., & Ellisman, M. H. (2005). Correlated light and
electron microscopic imaging of multiple endogenous proteins using quantum dots. Nature Methods, 10,
43–749.
Goodman, S. L., Park, K., & Albrecht, R. M. (1991). A correlative approach to colloidal gold labeling with
video-enhanced light microscopy, low voltage scanning electron microscopy, and high voltage electron
microscopy. In M. A. Hayat (Ed.), Colloidal gold: Principles, methods, and applications (pp. 369–409).
Boca Raton: CRC Press.
Grabenbauer, M., Geerts, W. J., Fernadez-Rodriguez, J., Hoenger, A., Koster, A. J., & Nilsson, T. (2005).
857–862.
Graham, R. C., Jr., & Karnovsky, M. J. (1966). The early stages of absorption of injected horseradish per-
oxidase in the proximal tubules of mouse kidney: ultrastructural cytochemistry by a new technique. The
Journal of Histochemistry and Cytochemistry, 14, 291–302.
Griffith, J. M., & Posthuma, G. (2002). A reliable and convenient method to store ultrathin cryosections prior
to immunolabeling. The Journal of Histochemistry and Cytochemistry, 50, 57–62.
Griffiths, G. (2001). Bringing electron microscopy back into focus for cell biology. Trends in Cell Biology,
11, 153–154.
Gustafsson, M. G. (2000). Surpassing the lateral resolution limit by a factor of two using structured illumina-
tion microscopy. Journal of Microscopy, 198, 82–87.
Hacker, G. W., Danscher, G., Grimelius, L., Hauser-Kronberger, C., Muss, W. H., Schiechl, A. , et al. (1995).
Silver staining techniques with special reference to the use of different silver salts in light and electron
microscopical immunogold-silver staining. In M. A. Hayat (Ed.), Immunogold-silver staining: Princi-
ples, methods, and applications (pp. 19–45). Boca Raton: CRC Press.
Hainfeld, J. F., & Furuya, E. R. (1992). A 1.4 nm gold cluster covalently attached to antibodies improves
immunolabeling. The Journal of Histochemistry and Cytochemistry, 40, 177–184.
Hainfeld, J. F. (1987). A small god-conjugated antibody label: improved resolution for electron microscopy.
Science, 263, 450–453.
Harata, N., Ryan, T. A., Smith, S. J., Buchanan, J., & Tsien, R. W. (2001). Visualizing recycling synaptic
vesicles in hippocampal neurons by FM 1-43 photoconversion. Proceedings of the National Academy of
Sciences of the United States of America, 98, 12748–12753.
Heymann, J. A.W., Hayles, M., Gestmann, I., Giannuzzi, L. A., Lich, B., & Subramanian, S. (2006). Site-
specific 3D imaging of cells and tissues with a dual beam microscope. Journal of Structural Biology,
155, 63–73.
Iglesias, J. R., Berner, R., & Simard, R. (1971). Ultracryotomy: a routine procedure. Journal of Ultrastruc-
ture Research, 36, 271–282.
microscopy: providing new understanding in the biomedical and plant sciences. Micron. doi:10.1016/j.
micron.2011.12.004.
Jyoti, K., Simons, J., & Simons, S. M. (2004). Potentials and pitfalls of fluorescent quantum dots for biologi-
cal imaging. Trends in Cell Biology, 14, 497–504.
Kandela, I. K., Bleher, R., & Albrecht, R. M. (2007). Multiple correlative immunolabeling for light and
electron microscopy using fluorophores and colloidal metal particles. The Journal of Histochemistry and
Cytochemistry, 55, 983–990.
Kandela, I. K., Bleher, R., & Albrecht, R. M. (2008). Immunolabeling for correlative light and electron
microscopy on ultrathin cryosections. Microscopy and Microanalysis, 14, 159–165.
Kapoor, T. M., Lampson, M. A., Hergert, P., Cameron, L., Cimini, D., Salmon, E. D., et al. (2006). Chromo-
somes can congress to the metaphase plate before biorientation. Science, 311, 388–391.
Kukulski, W., Schorb, M., Welsch, S., Picco, A., Kaksonen, M., & Briggs, J. A. G. (2011). Correlated fluo-
rescence and 3D electron microscopy with high sensitivity and spatial resolution. The Journal of Cell
Biology, 182, 111–119.
Lippincott-Schwartz, J., & Patterson, G. H. (2003). Development and use of fluorescent protein markers in
living cells. Science, 300, 87–91.
Lübke, J. (1993). Photoconversion of diaminobenzidine with different fluorescent neuronal markers into
light and electron dense reaction product. Microscopy Research and Technique, 1, 2–14.
Lučić, V., Leis, A., & Baumeister, W. (2008). Cryo-electron tomography of cells: connecting structure and
function. Histochemistry and Cell Biology, 130, 185–196.
Luo, Z., & Robinson, J. M. (1992). Co-localization of an endocytic marker and acid phosphatase in a tubular/
reticular compartment in macrophages. The Journal of Histochemistry and Cytochemistry, 40, 93–103.
Lyden, T. W., Anderson, C. L., & Robinson, J. M. (2002). The endothelium but not the syncytiotrophoblast
of human placenta expresses caveolae. Placenta, 23, 640–652.
Maranto, A. R. (1982). Neuronal mapping: a photooxidation reaction makes lucifer yellow useful for elec-
tron microscopy. Science, 217, 953–955.
Martin, O. C., & Pagano, R. E. (1994). Internalization and sorting of a fluorescent analogue of glucosyl-
ceramide to the Golgi apparatus of human skin fibroblasts: utilization of endocytic and nonendocytic
transport mechanisms. The Journal of Cell Biology, 125, 769–781.
Meisslitzer-Ruppitsch, C., Vetterlein, M., Stangl, H., Maier, S., Neumüller, J., Freissmuth, M. , et al. (2008).
Electron microscopic visualization of fluorescent signals in cellular compartments and organelles by
means of DAB-photoconversion. Histochemistry and Cell Biology, 130, 407–419.
Merchan-Perez, A., Rodriguez, J.-R., Alonso-Nanclares, L., Schertel, A., & DeFelipe, J. (2009). Counting
synapses using FIB/SEM microscopy: a true revolution for ultrastructural volume reconstruction. Fron-
tiers in Neuroanatomy, 3. doi: 3389/neuro.05.018.2009. Article 18.
Micheva, K. D., & Bruchez, M. P. (2011). The gain in brain: novel imaging techniques and multiplexed pro-
teomic imaging of brain tissue ultrastructure. Current Opinion in Neurobiology, 22, 1–7.
Micheva, K. D., Busse, B., Weiler, N. C., O’Rouke, N., & Smith, S. J. (2010). Single-synapse analysis of a
diverse synapse population: proteomic imaging methods and markers. Neuron, 68, 639–653.
Mironov, A. A., & Beznoussenko, G. V. (2009). Correlative microscopy: a potent tool for the study of rare or
unique cellular and tissue events. Journal of Microscopy, 235, 308–321.
Modla, S., & Czymmek, K. J. (2011). Correlative microscopy: a powerful tool for exploring neurological
cells and tissues. Micron, 42, 773–792.
Mori, M., Ishikawa, G., Takeshita, T., Goto, T., Robinson, J. M., & Takizawa, T. (2006). Ultrahigh-resolution
immunofluorescence microscopy using ultrathin cryosections: subcellular distribution of caveolin-1α
and CD31 in human placental endothelial cells. Journal of Electron Microscopy, 55, 107–112.
Murphy, G. E., Narayan, K., Lowekamp, B. C., Hartnell, J. A.W., Fu, J., et al. (2011). Correlative 3D imag-
ing of whole mammalian cell with light and electron microscopy. Journal of Structural Biology, 176,
268–278. Heymann.
Natera, J. E., Massad, W. A., Amat-Guerri, F., & Garcia, N. A. (2011). Elementary processes in the eosin-
sensitized photooxidation of 3-3’-diaminobenzidine for correlative fluorescence and electron micros-
copy. Journal of Photochemistry and Photobiology, 220, 25–30.
Necchi, V., Sommi, P., Vanoli, A., Manca, R., Ricci, V., & Solcia, E. (2011). Proteasome particle-rich struc-
tures are widely present in human epithelial neoplasms: correlative light, confocal, and electron micros-
copy study. PLoS One, 6, e21317. doi:10.1371/journal.pone.9921317.
Nikonenko, I., Boda, B., Alberi, S., & Muller, D. (2005). Application of photoconversion technique for
correlated confocal and ultrastructural studies in organotypic slice cultures. Microscopy Research and
Technique, 68, 90–96.
Nisman, R., Dellaire, G., Ren, Y., Li, R., & Bazett-Jones, D. P. (2004). Application of quantum dots as probes
for correlative fluorescence, conventional, and energy-filtered transmission electron microscopy. The
Oberti, D., Kirschmann, M. A., & Hahnloser, R. H. R. (2010). Correlative microscopy of densely labeled
projection neurons using neural tracers. Frontiers in Neuroanatomy, 4. Article 24.
Oberti, D., Kirschmann, M. A., & Hahnloser, R. H. R. (2011). Projection neuron circuits resolved using cor-
relative array tomography. Frontiers in Neuroscience, 5. Article 50.
Olorundare, O. E., Simmons, S. R., & Albrecht, R. M. (1992). Cytochalasin D and E: effects on fibrinogen
receptor movement and cytoskeletal reorganization in fully spread, surface-activated platelets: a correla-
tive light and electron microscopic investigation. Blood, 79, 99–109.
Painter, R. G., Tokuyasu, K. T., & Singer, S. J. (1973). Immunoferritin localization of intracellular antigens:
the use of ultracryotomy to obtain ultrathin sections suitable for direct immunoferritin staining. Proceed-
ings of the National Academy of Sciences, 70, 1649–1653.
Patla, I., Volberg, T., Elad, N., Hirschfeld-Warkenen, V., Grashoff, C., Fässler, R. , et al. (2010). Dissecting
the molecular architecture of integrin adhesion sites by cryo-electron tomography. Nature Cell Biology,
12, 909–916.
Polishchuck, R. S., Polishchuck, E. V., Marra, P., Alberti, S., Buccione, R., Luini, A. , et al. (2000). Correla-
tive light-electron microscopy reveals the tubular-saccular ultrastructure of carriers operating between
Golgi apparatus and plasma membrane. The Journal of Cell Biology, 148, 45–58.
Powell, R. D., Halsey, C. M. R., Spector, D. L., Kaurin, S. L., McCann, J., & Hainfeld, J. F. (1997). A covalent
fluorescent-gold immunoprobe: simultaneous detection of pre-mRNA splicing factor by light and elec-
tron microscopy. The Journal of Histochemistry and Cytochemistry, 45, 947–956.
Powell, R. D., Halsey, C. M. R., & Hainfeld, J. F. (1998). Combined fluorescent and gold immunoprobes:
reagents and methods for correlative light and electron microscopy. Microscopy Research and Technique,
42, 2–12.
Rieder, C. L., & Cassels, G. (1999). Correlative light and electron microscopy of mitotic cells in monolayer
cultures. Methods in Cell Biology, 61, 297–315.
Robinson, J. M., & Karnovsky, M. J. (1983). Ultrastructural localization of several phosphatases with cerium.
The Journal of Histochemistry and Cytochemistry, 31, 1197–1208.
Robinson, J. M., & Takizawa, T. (2009). Correlative fluorescence and electron microscopy in tissues: immu-
nocytochemistry. Journal of Microscopy, 235, 259–272.
Robinson, J. M., Takizawa, T., & Vandré, D. D. (2000). Applications of gold cluster compounds in immu-
nocytochemistry and correlative microscopy: comparison with colloidal gold. Journal of Microscopy,
199, 163–179.
Robinson, J. M., Takizawa, T., Pombo, A., & Cook, P. R. (2001). Correlative fluorescence and electron
microscopy on ultrathin cryosections: bridging the resolution gap. The Journal of Histochemistry and
Robinson, J. M., Ackerman, W. E., 4th, & Vandré, D. D. (2009). While dysferlin and myoferlin are co-
expressed in the human placenta, only dysferlin expression is responsive to trophoblast fusion in model
systems. Biology of Reproduction, 81, 33–39.
Roth, J. (1996). The silver anniversary of gold: 25 years of the colloidal gold marker system for immunocy-
tochemistry and histochemistry. Histochemistry and Cell Biology, 106, 1–8.
Rust, M. J., Bates, M., & Zhuang, X. (2006). Sub-diffraction-limit imaging by stochastic optical reconstruc-
tion microscopy (STORM). Nature Methods, 3, 793–795.
encoded tag for correlated light and electron microscopy of intact cell, tissues, and organisms. PLoS
Biology, 9, e1001041.
Slot, J. W., & Geuze, H. J. (1981). Sizing of protein A-colloidal gold probes for immunoelectron microscopy.
The Journal of Cell Biology, 90, 533–536.
Sutherland, A. J. (2002). Quantum dots as luminescent probes in biological systems. Current Opinion in
Solid State & Materials Science, 6, 365–370.
Takizawa, T., & Robinson, J. M. (1994). Use of 1.4 nm immunogold particles for immunocytochemistry on
ultra-thin cryosections. The Journal of Histochemistry and Cytochemistry, 42, 1615–1623.
Takizawa, T., & Robinson, J. M. (2000). Analysis of antiphotobleaching reagents for use with FluoroNano-
gold in correlative microscopy. The Journal of Histochemistry and Cytochemistry, 48, 433–436.
Takizawa, T., & Robinson, J. M. (2003a). Ultrathin cryosections: an important tool for immunofluorescence
and correlative microscopy. The Journal of Histochemistry and Cytochemistry, 51, 707–714.
Takizawa, T., & Robinson, J. M. (2003b). Correlative microscopy of ultrathin cryosections is a powerful tool
for placental research. Placenta, 24, 557–565.
Takizawa, T., & Robinson, J. M. (2003c). A new method to enhance contrast of ultrathin cryosections for
immunoelectron microscopy. The Journal of Histochemistry and Cytochemistry, 51, 31–39.
Takizawa, T., & Robinson, J. M. (2006). Correlative microscopy of ultrathin cryosections in placental
research. Methods in Molecular Medicine, 121, 351–369.
Takizawa, T., Anderson, C. L., & Robinson, J. M. (2005). A novel FcγR-defined, IgG-containing organelle in
placental endothelium. Journal of Immunology, 175, 2331–2339.
Tokuyasu, K. T. (1973). A technique for ultracryotomy of cell suspensions and tissues. The Journal of Cell
Biology, 57, 551–565.
Toni, N., Teng, E. M., Bushong, E. A., Aimore, J. B., Zhao, C., Consiglio, A. , et al. (2007). Synapse forma-
tion on neurons born in the adult hippocampus. Nature Neuroscience, 10, 727–734.
Van der Wel, N., Fluitsma, D. M., Dascher, C. C., Brenner, M. B., & Peters, P. J. (2005). Subcellular localiza-
tion of myobacteria in tissues and detection of lipid antigens in organelles using cryo-techniques for light
and electron microscopy. Current Opinion in Microbiology, 8, 323–330.
Vandre, D. D., Ackerman, W. E., 4th, Kniss, D. A., Tewari, A., Mori, M., Takizawa, T. , et al. (2007). Dys-
ferlin is expressed in human placenta but does not associate with caveolin. Biology of Reproduction, 77,
533–542.
Vicidomini, G., Gagliani, M., Cortese, K., Krieger, J., Bruescher, P., Bianchini, P. , et al. (2010). A novel
approach for correlative light electron microscopy analysis. Microscopy Research and Technique, 73,
215–224.
Watanabe, S., Punge, A., Hollopeter, G., Willig, K. I., Hobson, R. J., Davis, M. W., et al. (2011). Protein
localization in electron micrographs using fluorescence nanoscopy. Nature Methods, 8, 80–84.
Watson, A., Wu, X., & Bruchez, M. (2003). Lighting up cells with quantum dots. Biotechniques, 34,
296–300.
Westphal, V., & Hell, S. W. (2005). Nanoscale resolution in the focal plane of an optical microscope. Physi-
cal Review Letters, 94, 143903.
Xu, Y., Wang, Q., He, P., Dong, Q., Liu, F., Liu, Y. , et al. (2008). Cell nucleus penetration by quantum dots
induced by nuclear staining organic fluorophore and UV-irradiation. Advanced Materials, 20, 3468–3473.
Yuste, R. (2005). Fluorescence microscopy today. Nature Methods, 2, 902–904.
CHAPTER 4

in Parasite Research
Céline Loussert*, Claire-Lise Forestier† and Bruno M. Humbel*
* Electron Microscopy Facility, University of Lausanne, Biophore, 1015 Lausanne, Switzerland
† INSERM, Institute Pasteur, 25 Rue du Docteur Roux, 75015 Paris, France
Abstract
I. Introduction
II. Rationale
III. Methods
A. Flat Embedding Procedure with MatTek Petri Dish Containing Gridded Coverslips
B. BSA-Au Labeling of the Endocytic Pathway
IV. Experiments and Materials
V. Results and Discussion
VI. Summary
Acknowledgments
References
Abstract
The interaction of a parasite and a host cell is a complex process, which involves
several steps: (1) attachment to the plasma membrane, (2) entry inside the host cell,
and (3) hijacking of the metabolism of the host. In biochemical experiments, only an
event averaged over the whole cell population can be analyzed. The power of micros-
copy, however, is to investigate individual events in individual cells. Therefore, parasi-
tologists frequently perform experiments with fluorescence microscopy using different
dyes to label structures of the parasite or the host cell. Though the resolution of light
microscopy has greatly improved, it is not sufficient to reveal interactions at the ultra-
structural level. Furthermore, only specifically labeled structures can be seen and
related to each other. Here, we want to demonstrate the additional value of electron
microscopy in this area of research. Investigation of the different steps of parasite-host
cell interaction by electron microscopy, however, is often hampered by the fact that
there are only a few cells infected, and therefore it is difficult to find enough cells to
study. A solution is to profit from low magnification, hence large overview, and specific
60 Céline Loussert et al.
location of the players by fluorescence labels in a light microscope with the high power
resolution and structural information provided by an electron microscope, in short by
correlative light and electron microscopy.
I. Introduction
Fast developments in light microscopy and the possibility to follow the movements
of proteins in living cells have brought a revolution in the investigation of cellular
processes. With the introduction of the so-called super-resolution microscopy, the reso-
lution of light microscopy is approaching molecular levels (Betzig et al., 2006; Hell,
Dyba, & Jakobs, 2004; Rust, Bates, & Zhuang, 2006). Still, light microscopy has the
big disadvantage that only fluorescently labeled structures can be studied in relation to
each other. In electron microscopy, however, it is sufficient to label the protein of inter-
est to relate it to surrounding organelles without further specific identification. In addi-
tion, electron microscopy still profits of a 10–100 times better resolution and it will, in
cell biology, remain an important tool in future.
This fact stimulates correlating light, predominately fluorescence, microscopy with
electron microscopy. Today, there are several different approaches called correlative
light–electron microscopy. Firstly, a cellular process is followed through a fluorescent
protein, then at the time point of interest the cells are fixed, either chemically (Stöffler
et al., 1998) or by cryo-fixation (Verkade, 2008) and processed for electron microscopy.
The area imaged by light microscopy is then analyzed by electron microscopy at high
resolution. To access the labeled 3D structure, the fluorescence image is used as a guide-
line to approach the cell of interest during trimming and finally with serial sectioning
(Stöffler et al., 1998).
Secondly, fluorescence microscopy is used to locate the protein of interest in
a large area of a section (Karreman et al., 2009; Schwarz & Humbel, 2007; Schwarz
& Humbel, 2008; Takizawa & Robinson, 2006) or in a volume (Biel, et al., 2003;
Wilke et al., 2008) to ease the search in the electron microscope (Schwarz & Hum-
bel, 2007). In principle, the light micrograph serves as a map to relocate the
fluorescent spot in the electron microscope. Recently, more automated procedures were
developed, storing the coordinates obtained in the light microscope directly into the soft-
ware of the electron microscope (Gruska, et al., 2008) up to integrating a light microscope
into an electron microscope (Agronskaia et al., 2008; Jones, et al., 1982; Karreman et al.,
2009).
In this article, we want to highlight the role electron microscopy played in corre-
lation with light/fluorescence microscopy, including phase contrast and total internal
reflection fluorescence (TIRF) microscopy (Axelrod, 2001; Truskey, et al., 1992), to
dissect the different steps of parasite–host cell interaction. We will base our demon-
stration of correlative microscopy on the work published by Forestier, et al., (2011),
analyzing the entry of Leishmania donovani in bone marrow-derived macrophages
(BMMs). L. donovani, a protozoan parasite, is transmitted by a sandfly (Lutzomyia
longipalpis) vector to a mammalian host, causing visceral leishmaniasis, as disease that
4. Correlative Light and Electron Microscopy in Parasite Research 61
is often fatal (Chappuis et al., 2007). The parasite infection cycle comprises extracellu-
lar, flagellated promastigotes that proliferate in the vector and intracellular, nonmotile
amastigotes that multiply within the host cells. Parasite transmission to the host occurs
during a sandfly blood meal via injection of highly motile promastigotes, the infectious
form of the parasite. The promastigotes are capable to infect various mammalian host
cells, mainly phagocytic cells (Sacks & Perkins, 1984).
In their work, Forestier et al. (2011) investigated the dynamics of Leishmania pro-
mastigote–host cell interaction using high-resolution, temporal, and spatial microscopy
techniques to elucidate the sequential steps leading to efficient invasion.
One fact that may not be neglected in host–parasite interaction is the infection rate.
Very often we are faced with a low infection rate that makes the characterization of the
sequence of events of host–parasite interactions close to impossible. At the end of this
chapter, we propose a relatively easy method to increase the pool of parasite-infected
cells for further electron microscopy analysis, without modifying neither the parasite
infectivity nor the host cells sensitivity.
We hope, with this contribution, to prove that using a robust sample preparation
method for electron microscopy in combination with various light microscopic tech-
niques, e.g., phase contrast, fluorescence imaging makes the investigation of rare
events easier.
II. Rationale
The entry of a parasite into a host cell is characterized by numerous events (e.g.,
cell–cell contact, cell entry, and intracellular interactions), all occurring at the cellular
scale or at the subcellular scale. A large range of methods can be used to describe these
interactions, such as molecular biology or biochemistry. These methods need to pool
samples and will therefore average the analyzed processes. In our point of view, the
best way to understand a specific event is to observe individual host cell–parasite inter-
actions, combining overview (light microscopy) and detailed view (electron micros-
copy). Indeed, correlative light–electron microscopy appears as a powerful technique
to investigate the individual steps and time points of this interaction.
To make the correlative approach more efficient, in first instance the chronology of
the studied process has to be established by life cell imaging with a light microscope.
For this, a basic phase contrast light microscope is sufficient and powerful enough to
study many individual living cells and to give an overview characterizing the sequence
and time line of parasite–host cell interaction. In our study, the interaction between
L. donovani and BMMs was observed with a light microscope equipped with an
incubator to perform a long time imaging (BioStation, Nikon Instruments Inc.)
(Forestier et al., 2011). Our study dissects the dynamics of L. donovani interac-
tion with the host cells identifying four sequential phases during early infection
(Fig. 1):
• Phase I: The interaction of the host cell and the parasite is mainly initiated by
a contact of the tip of the flagellum with the host cell membrane. This contact
induces extension of host cell pseudopodes and phagocytosis of the parasite into
the parasitophorous vacuole.
• Phase II: The parasite is reorientating in the host cell by a complex mechanism and
it ends by pointing with the flagellum toward the cell membrane of the host.
• Phase III: An extensive beating of the flagellum of the parasite is observed at the
host cell membrane, which can lead to damage of the membrane.
• Phase IV: The mobility of the flagellum stops, and the parasite starts to differentiate
in the parasitophorous vacuole close to the host cell nucleus.
In the case of our study, we decide to follow the endosomes/lysosomes traffic in the
host cell during parasite infection. Using LysoTracker, this activity can easily be moni-
tored by fluorescence microscopy. The spatial resolution, however, is not sufficient to
clearly prove that interaction between the parasite, or the parasitophorous vacuole, and
these organelles occurs. Here, only electron microscopy will provide us with convinc-
ing data. In first instance, one could think about immunolabeling against lysosomes
resident proteins, but this approach has two drawbacks: the cells need to be permeabi-
lized, compromising the ultrastructure, and the dynamic aspect of the interaction is not
accessible. That is the reason why, in our experiments, we loaded, prior to infection,
the endocytic compartments with BSA-gold as described by Kleijmeer, et al., (1997),
in addition to Lysotracker labeling.
Fig. 1 Events of L. donovani entry into host cell. During phase I, the parasite establishes an interaction with
the host cell, mainly with the flagellum, which will induce phagocytosis of the parasite into a parasitophorous
vacuole. During phase II, the parasite is reoriented in the host cell ending by pointing its flagellum towards the
host cell membrane. During phase III, the flagellum of the parasite beats, generating some damages at the host
cell membrane. Finally, during phase IV, the parasite losses its motility.
III. Methods
A. Flat Embedding Procedure with MatTek Petri Dish Containing Gridded Coverslips
Cultured cells are grown in the MatTek Petri dishes (MatTek, Ashland, MA, USA)
under standard cell culture conditions. Coating of the coverslips with polylysine, fibro-
nectin, etc. can be done, but is not mandatory. Live or fixed cells can be imaged by
phase contrast, DIC, or any kind of fluorescence imaging. In case of fluorescence imag-
ing, the chemical fixation has to be gentle (e.g., 4% formaldehyde), and the fixative
solution may not contain high concentrations of glutaraldehyde that can induce autoflu-
orescence. At low magnification, the cells of interest can be identified within the coor-
dinates of the MatTek grids (Fig. 2(A,B)). The micrograph will serve as a map to find
the cells of interest during the subsequent preparation steps (Fig. 2) and finally in the
electron microscope. After imaging, the fixation is done by following a conventional
embedding protocol (Hayat, 2000), where 0.1M cacodylate buffer (pH 7.4) containing
2.5% glutaraldehyde for 2 h at room temperature was used. Before further processing
of the sample, the area of the cells of interest is also marked on the bottom of the Petri
dish with colored varnish to facilitate handling. Next, the samples were fixed in 2%
osmium tetroxide in water for 45 min at room temperature. Dehydration was done with
a graded series of ethanol: 5 min in 30%, 5 min in 50%. Then, the cells were stained with
1% uranyl acetate in 70% ethanol for 20 min. Dehydration is continued by 70% ethanol
for 1 min, 5 min in 80%, and twice 10 min in 95%, and finally thrice in 100% ethanol
for 6 min each. A drop of the Epon mixture (Luft, 1961) has to be added immediately
to avoid drying. This thin layer of resin will allow the residual ethanol to evaporate
Fig. 2 (A) In the overview, the gridded glass coverslip fused to a 35 mm Petri dish is shown. (B) At higher
magnification, the numbered fields imprinted in the coverslip are visible. One grid has a length of 600 µm.
Here, the cells or cellular events can be mapped by any kind of light microscopic imaging mode, e.g., phase
contrast, fluorescence. (C) The cells of interest are flat embedded in resin and (D) the gridded imprint again can
be followed on the bottom side of the resin block at low or (E) higher magnification. (F) Then the area of interest
is trimmed for subsequent ultrathin sectioning. Bar represents 600 µm.
during 4 h in a ventilated hood. Thereafter, Epon-filled gelatine capsules are gently put
up-side-down upon the coverslip above the spot of varnish. After about 12 h of incuba-
tion at room temperature the sample is polymerized at 60°C for 48 h (Fig. 2(C)). The
Epon block was removed (Fig. 2(D)) by heating the glass coverslip by moving a flame
of a lighter for 7 s over the bottom of the coverslip. Now the area, containing the cells,
is trimmed according to the imprinted coordinates (Fig. 2(E,F)) and sectioned. With
this flat embedding method, it is possible to collect the first sections that were in direct
contact with the coverslip. It is important to make the trapezium as small as possible to
ensure that the sectioned area is flat enough to fit into the first section.
B. BSA-Au Labeling of the Endocytic Pathway

The endocytic pathway was labeled by BSA-Au particles (Kleijmeer et al., 1997),
which can easily be identified by electron microscope. Next to a parasitophorous vacu-
ole, many vesicles containing gold particles are observed (Fig. 3(A)). According to the
Fig. 3 The micrograph shows the BSA-Au particles inside the late endosomes–lysosomes of the host cells (A).
We see a larger number of filled vesicles surrounding the parasite (arrow), but also BSA-Au inside the parasi-
tophorous vacuole (arrow head). An event of fusion of a BSA-Au containing vesicle with the parasitophorous
vacuole (B, and insert) could be caught. Bar represents 1 µm, inset 200 nm.
work of Kleijmeer et al. (1997), these are late endosomes and lysosomes. In a different
cell, we could catch the fusion of the gold labeled organelles with the parasitophorous
vacuole (Fig. 3(B)). The beauty of the technique is that we can capture a dynamic
process and it could be proven that late endosomes or lysosomes do indeed fuse with
the parasitophorous vacuole (Forestier et al., 2011).
IV. Experiments and Materials
A. L. donovani Experiments

1. Mice and Macrophages
BMMs were obtained by growing bone marrow cells from female C57BL/6 mice in
100 × 20 mm2 Petri dishes (Falcon, Becton Dickinson, Franklin Lakes, NJ, USA) at 37°C,
5% CO2 for 7 days in RPMI-1640 (Gibco Invitrogen, Carlsbad, CA, USA), supplemented
with 10% heat-inactivated fetal calf serum (FCS; Hyclone, Thermo Scientific, Canada), 1 µM
β2-mercaptoethanol, 50 U/ml penicillin, 10 mM HEPES (Sigma-Aldrich, Lyon, France),
1 mM sodium pyruvate (Gibco), and 30% L929 mouse fibroblast-conditioned medium.
2. L. donovani
Hamster-derived L. donovani promastigotes were differentiated from amastigotes
(LD1S Sudanese strain 1), isolated from spleens and livers of highly infected hamsters
and grown in vitro at 26°C in complete M199 medium (Sigma-Aldrich). Metacyclic-
enriched hamster-derived promastigotes were obtained from stationary phase culture
before passage 6.
3. Macrophages Infection Assay

BMMs were detached with 20 mM EDTA, washed and plated overnight in 35 mm
diameter culture dishes with gridded glass coverslips (MatTek) at the density of
2 × 105cells/dish in complete RPMI medium. BMMs were incubated at 37°C with meta-
cyclic-enriched hamster-derived promastigotes at the ratio of 10 parasites per cell for 2 h.
To label the late endosomes and lysosomes, a solution of BSA-gold particles (UMC,
Utrecht, The Netherlands) was dialyzed overnight against PBS at 4°C, and BMMs
were incubated with the BSA-gold particles for 1 h in RPMI-1640 at 37°C. Then after
1 h of chase, the macrophages were infected with metacyclic-enriched hamster-derived
promastigotes for 2 h in RPMI-1640 supplemented with 10% FCS.
4. Fluorescence Confocal Microscopy

Infected BBMs were incubated with LysoTracker Red DND-99 dye (Molecular
Probe, Eugene, Oregon, USA) diluted at 1/5000 in media for 1 h at 37°C and then
washed prior to adding parasites for 1 h. Samples were then fixed with 4% formal-
dehyde in PBS for at least 15 min at room temperature, blocked with 10% FCS and
mounted in Mowiol (Calbiochem, San Diego CA, USA) containing Hoechst 33342
(Sigma-Aldrich) and observed with a Leica DMI6000 confocal light microscope (Leica
Microsystems, Wetzlar, Germany).
5. TIRF Microscopy
After fixation with 4% formaldehyde, infected BBMs were permeabilized with 0.1%
Saponin (Sigma-Aldrich) in PBS for 5 min at room temperature. Immunolocalization
of LAMP-1 was done using a rat anti-mouse LAMP-1 antibody (1D4B, PharMingen)
followed by incubation with an AlexaFluor 488 conjugated goat anti-rat (Molecular
Probes) secondary antibody. Imaging was done with an Olympus IX81 microscope
(Olympus Europa Holding GmbH, Hamburg Germany) equipped with TIR-FM illumi-
nator, an Andor iXon + dv-855 EMCCD camera (Andor Technology, Belfast, Northern
Ireland), and a solid state 488 nm laser.
6. Transmission Electron Microscopy

After imaging, the infected BMMs were fixed with 2% formaldehyde, 0.2% glutaral-
dehyde in 0.2 M Sorensen buffer (pH 7.3) for 30 min followed by 2.5% glutaraldehyde
in 0.1 M cacocylate buffer (pH 7.4) for 2 h. A postfixation step was done for 45 min in
1% osmium tetroxide (Merck, Darmstadt, Germany) in water and finally incubated in
2% uranyl acetate (Merck) in 70% ethanol for 20 min. Dehydration was done in a graded
series of ethanol, and samples were embedded in Epon resin. Contrasted ultrathin sec-
tions (60 nm) were observed in a JEM 1010 electron microscope (JEOL, Tokyo, Japan).
B. Plasmodium Berghei Experiments

1. Parasites and Mosquitoes
These procedures are described in Combe et al., (2009) . Briefly, the parasites were
derivatives from P. berghei strain NK65, which expresses GFP under the eef1α pro-
moter (Janse, Ramesar, & Waters, 2006). Anopheles stephensi (Sda500 strain) mosqui-
toes were fed on infected mice 3–5 days after emergence and kept at 21°C and 70%
humidity. The percentage of fluorescent parasites was determined in midgut sporozo-
ites and in salivary gland sporozoites at various time points after infectious blood meal
by scoring at least 100 sporozoites. For infection of cell cultures, sporozoites were
isolated from infected salivary glands 18–25 days after the infectious blood meal and
kept on ice in tissue culture medium with 10% FCS.
2. HepG2 Infection
HepG2 cells were grown in DMEM + Glutamax-1 media (Gibco) supplemented
with 10% FCS (PAA Laboratories GmbH, Pasching, Austria) at 37°C in the presence of
5% CO2. Cell infection was performed at the multiplicity of 1, using 8-well Permanox
chamber slides (Lab-Tek, Nunc, Roskilde, Denmark) by adding 5 × 104 sporozoites
freshly dissected out from mosquito salivary glands to HepG2 cells. Samples were
then placed at 37°C.
3. Infected Cell Sorted by Flow Cytometry

To enrich the pool of infected cells for electron microscopy study, treated cells
were collected from the 8-well chamber slides and gently fixed in 4% formaldehyde
in Sorensen buffer pH 7.3. GFP positive cells were sorted on a BD FACSAria II cell
sorter (BD Biosciences, Franklin Lakes, USA). After sorting, the cells were fixed in
2.5% glutaraldehyde in 0.1 M cacodylate buffer for 2 h at room temperature, then for
24 h at 4°C.
4. Transmission Electron Microscopy

Due to the small number of infected cells, pellets were first embedded in 4% low melt-
ing agarose type IV (Sigma) in 0.1 M cacodylate buffer. After solidification, small cubes
of 1 mm3 were cut. After several washes in cacodylate buffer, the samples were postfixed
in 1% osmium tetroxide in water for 1 h at room temperature, and then dehydrated in a
graded series of ethanol. The samples were embedded in Epon resin and polymerized.
Contrasted ultrathin sections (60 nm) were observed in a JEM 1010 electron micro-
scope (JEOL, Tokyo, Japan)
V. Results and Discussion
Macrophages were seeded on MatTek Petri dishes that have a grid engraved to find
the coordinates of the cells back in the electron microscope. Lysosomes were stained
with LysoTracker and DNA with Hoechst 33342. For the electron microscopic stud-
ies, late endosomes and lysosomes were labeled with BSA-Au in addition to Lyso-
Tracker. In phase contrast, the desired time point of parasite–host cell interaction was
selected (Fig. 4(A)). In fluorescence, the lysosomes and the host cell nuclei can easily
be mapped (Fig. 4(B)). This fluorescence map can be overlaid with an electron micro-
graph at low magnification (Fig. 4(C)). The same cells observed in the light microscope
are found back in the electron micrograph. Now we can zoom in on each individual
parasite to be studied at magnifications where membrane interactions and organelles
can be analyzed.
Fig. 4 In the overview, we can localize a specific area of interest in phase contrast within one square of the
gridded coverslip (A). The fluorescence image of the area (B) shows numerous LysoTraker containing vesicles
(in red) in the host cells identified by Hoechst staining (in blue). The fluorescence image is overlaid with the
phase contrast image. In (C), we overlaid a low magnification electron micrograph of this specific area, with the
fluorescence image. There is a perfect match between the two images, showing the robustness of our correla-
tive approach. The parasite indicated with a white arrow is detailed in Fig. 5 and the one indicated with a white
arrow head in Fig. 6. Bar represents 20 µm. (See color plate.)
Fig. 5 In (A), we zoom in to the area indicated with a white arrow in the Fig. 4. The observed parasite is not
yet completely engulfed by the host cell. This indicates that phagocytosis is still in progress, characterized by
the extended pseudopodes of the host cell. Close to this parasite, we detect a parasitophorous vacuole containing
another parasite (B, arrow). This vacuole contains BSA-Au particles, indicating a fusion event with late endosomes/
lysosomes. The electron micrographs are a montage from a multiple images acquisition. Bar represents 5 µm.
In Fig. 4(B), the parasite indicated with an arrow shows no lysosomal stain. When
we zoom in by electron microscopy, we see a parasite showing the characteristics of
early entry into the host cell, i.e., the flagellum is pointed toward the host cell, and the
parasite is not completely engulfed by the host cell (Fig. 5(A)).
The parasitophorous vacuole containing a completely engulfed parasite at the lower
right side of Fig. 5(B) shows gold particles (arrow), indicating that the gold labeled
endocytic organelles of the host cell must have fused with the vacuole. These results
suggest that the vesicles can only fuse with the parasite containing vacuole after the
cell entry is completed.
To further analyze the interaction of L. donovani with BMMs, a parasite in the vicin-
ity of LysoTracker positive area in the fluorescence image was chosen (Fig. 4(B), arrow
head). In the electron micrograph (Fig. 6), gold particles were observed at the tip and
the bottom of the parasite contained in the parasitophorous vacuole.
With electron microscopy, we could unambiguously prove that fusion of the late
endosomes/lysosomes with the parasitophorous vacuole occurs during the early stage
of infection (Forestier et al., 2011).
Then phase III of the infection (Fig. 1) was analyzed by studying the docking of
lysosomes with the basal membrane of the host cell with TIRF microscopy. Late endo-
somes and lysosomes were marked by immunofluorescence with LAMP-1, using a per-
meabilization method. With TIRF, the cells of interest were selected and then embedded
for electron microscopy. The power of TIRF is to visualize about 100 nm deep into the
cells, thus concentrating on events occurring at the plasma membrane. Therefore, it is
of prime importance to retrieve the first one or two sections parallel to the coverslip,
because only they contain the vesicles seen in TIRF microscopy.
Fig. 6 Electron micrograph of the parasite indicated with a white arrow head in the Fig. 4. By fluorescence
imaging numerous vesicles, containing LysoTracker (Fig. 4), were detected in the vicinity of the parasite. By
electron microscopy, we could confirm that these vesicles fused with the parasitophorous vacuole by the pres-
ence of BSA-Au particles at the tip and the bottom of the parasite. Bar represents 2 µm.
In Fig. 7(A), the lysosomes are labeled with LAMP-1 after permeabilization. In
green, a homogenous distribution of the vesicles is observed inside the host cell. TIRF
microscopy allows distinguishing a large docking of vesicles under an unidentified
structure (Fig. 7A, red). In the first section parallel to the coverslip, BSA-Au particles
were found close to membranes (Fig. 7(B), arrow heads). A few sections deeper, the
presence of two parasites, in this specific area, could be confirmed (Fig. 7(C)).
With these examples, we show that many of the interesting events in cell biology are
individual and located to one up to a few cells. High resolution investigation becomes
very tedious if not unrealistic considering that a label in the electron microscope can
only be seen with a field of view of 100–400 µm2. Further, for complex mechanisms
as parasite entry, many samples have to be prepared to catch the right time point. Here,
life cell imaging with a large field of view is a necessity. With L. donovani, we have the
great advantage that a high infection rate can be achieved with a percentage of infec-
tion being around 80%. With other parasites such as P. berghei, even at high MOI only
about 20% of all liver cells get infected.
Plasmodium, genetically modified to express GFP (Janse et al., 2006), was used to
infect hepatocytes. At this low infection rate, even with correlative microscopy using
the gridded coverslips, the dissection and analysis of the different stages of infection
becomes an impossible attempt (Fig. 8) and additional tricks have to be applied.
Fig. 7 In (A), a fluorescence signal of LAMP-1, homogenously distributed in the host cell (green), is observed.
TIRF imaging identifies docking of these vesicles in a stretched area of the host cell (red). Looking at the first
section of this area by electron microscopy (B), we could observe BSA-gold particles (arrows), indicating the
presence of the filled late endosomes/lysosomes in this area. In a deeper section (C), two parasites became vis-
ible in the TIRF positive area. An overlay the TIRF image with the electron micrograph matched with the two
parasites. The electron micrographs are a montage from a multiple images acquisition. Bar represents 20 µm for
the fluorescence image and 2 µm for the electron micrographs. (See color plate.)
After infection, the cells were fixed and the infected cells, containing fluorescent par-
asites, were accumulated with the help of a cell sorter (Combe et al., 2009; Ishino et al.,
2009). Now, the extent of work became comparable to the Leishmania study, hence
realistic. Indeed (Fig. 9), the developmental stages of P. berghei could be described in
great details (Combe et al., 2009).
After enrichment by cell sorting, we could perform quantification by analyzing
more than 100 infected cells (data not shown) without the need for a multitude of
experiments. Despite the cell sorting process, the overall preservation of the cellular
ultrastructure was good. Cell sorting can be valuable to analyze similar rare events. The
sorting must be as gentle as possible that it does not change the ultrastructure of the
living cells, then it can be combined with life cell imaging.
Fig. 8 In phase contrast (A) a large number of hepatocytes can be observed on the MatTek gridded coverslip.
The fluorescence image reveals the GFP labeled parasite-infected cells (B). The overlay of the two images (C)
confirms the low rate of infection of P. berghei. Bar represents 500 µm.
Fig. 9 By increasing the pool of observed infected cells with cell sorting, it was possible to describe in details
all the developmental stages of P. berghei in hepatocytes. In (A), we observe a nonfragmented form of the
parasite inside the cytoplasm of the host cell. In (B), the parasites begin differentiation, and merozoites start to
bud from the periphery. Finally in (C), the parasite is completely differentiated and only ready to be released
merozoites are observed. Bar represents 10 µm.
VI. Summary
The great potential of microscopy is to study individual cells and single events.
When working with identical cells in cell culture, random sampling will always result
with enough material to study. In most of the cases, especially in cell biological studies
in tissue, the cell of interest is hidden like the famous needle in the haystack.
To visualize a specific cell type in a large area, e.g., an infected cell in a cell culture
or tissue, first the area of interest has to be pinpointed, e.g., by light microscopy. Differ-
ent imaging techniques, such as bright field, DIC, phase contrast, and more specifically
fluorescent labels by wide field, confocal light microscopy, or more advanced techniques
such as TIRF microscopy can be used. Thus, the area of interest can be cartographed in
2D or even 3D. With this map as a guide, the same area is approached, at higher resolu-
tion, by electron microscopy. In addition, this chapter shows how to increase the chance
of detecting a special event by using nonmicroscopic methods, e.g., FACS cell sorting.
This review shows how scarce events as developmental stages of parasite or parasite–
host interaction can be revealed by combining light and electron microscopy.
In future, we can also think of using other imaging techniques, e.g., MRI, X-ray
tomography, to screen and map from organism to tissue to cell to unravel cellular pro-
cesses involved in diseases.
Acknowledgments
The authors would like to thank Christophe Machu for the help with fluorescence
microscopy, Audrey Combe and Tomoko Ishino for the Plasmodium berghei experiments.
References
Agronskaia, A. V., Valentijn, J. A., van Driel, L. F., Schneijdenberg, C. T. W. M., Humbel, B. M., van Bergen
en Henegouwen, P. M. P., et al. (2008). Integrated fluorescence and transmission electron microscopy.
Journal of Structural Biology, 164, 183–189.
Axelrod, D. (2001). Selective imaging of surface fluorescence with very high aperture microscope objec-
tives. Journal of Biomedical Optics, 6, 6–13.
Biel, S. S., Kawaschinski, K., Wittern, K.-P., Hintze, U., & Wepf, R. (2003). From tissue to cellular ultrastruc-
ture: closing the gap between micro- and nanostructural imaging. Journal of Microscopy, 212, 91–99.
Chappuis, F., Sundar, S., Hailu, A., Ghalib, H., Rijal, S., Peeling, R. W., et al. (2007). Visceral leishmani-
asis: what are the needs for diagnosis, treatment and control? Nature Reviews Microbiology, 5, 873–882.
Combe, A., Giovannini, D., Carvalho, T. G., Spath, S., Boisson, B., Loussert, C., et al. (2009). Clonal condi-
tional mutagenesis in malaria parasites. Cellular Host Microbiology, 5, 386–396.
Forestier, C. -L., Machu, C., Loussert, C., Pescher, P., & Späth, G. F. (2011). Imaging host cell–Leishmania
interaction dynamics implicates parasite motility, lysosome recruitment, and host cell wounding in the
infection process. Cellular Host Microbiology, 9, 319–330.
Gruska, M., Medalia, O., Baumeister, W., & Leis, A. (2008). Electron tomography of vitreous sections from
cultured mammalian cells. Journal of Structural Biology, 161, 384–392.
Hayat, M. A. (2000). Principles and techniques of electron microscopy. Biological Applications. Cambridge:
Cambridge University Press.
Hell, S. W., Dyba, M., & Jakobs, S. (2004). Concepts for nanoscale resolution in fluorescence microscopy.
Current Opinion in Neurobiology, 14, 599–609.
Ishino, T., Boisson, B., Orito, Y., Lacroix, C., Bischoff, E., Loussert, C., et al. (2009). LISP1 is important
for the egress of Plasmodium berghei parasites from liver cells. Cellular Microbiology, 11, 1329–1339.
Janse, C. J., Ramesar, J., & Waters, A. P. (2006). High-efficiency transfection and drug selection of geneti-
cally transformed blood stages of the rodent malaria parasite Plasmodium berghei. Nature Protocols, 1,
346–356.
Jones, S., Chapman, S. K., Crocker, P. R., Carson, G., & Levison, D. A. (1982). Combined light and electron
microscope for routine histopathology. Journal of Clinical Pathology, 35, 425–429.
Karreman, M. A., Agronskaia, A. V., Verkleij, A. J., Cremers, F. F. M., Gerritsen, H. C., & Humbel, B. M.
(2009). Discovery of a new RNA containing nuclear structure in UVC-induced apoptotic cells by inte-
grated laser electron microscopy. Biology of the Cell, 101, 287–299.
Kleijmeer, M. J., Morkowski, S., Griffith, J. M., Rudensky, A. Y., & Geuze, H. J. (1997). Major histocom-
patibility complex Class II compartments in human and mouse B lymphoblasts represent conventional
endocytic compartments. Journal of Cell Biology, 139, 639–649.
Luft, J. H. (1961). Improvements in epoxy resin embedding methods. Journal of Biophysical and Biochemi-
cal Cytology, 9, 409–414.
Sacks, D. L., & Perkins, P. V. (1984). Identification of an infective stage of Leishmania promastigotes.
Science, 223, 1417–1419.
sections. In J. Kuo (Ed.), Electron microscopy: Methods and protocols (Vol. 369, pp. 229–256). Totowa,
NJ: Humana Press.
Schwarz, H., & Humbel, B. M. (2008). Correlative light and electron microscopy. In A. Cavalier, D. Spehner,
& B. M. Humbel (Eds.), Handbook of cryo-preparation methods for electron microscopy (pp. 527–555).
Boca Raton::CRC Press.
Stöffler, H. -E., Honnert, U., Bauer, C. A., Höfer, D., Schwarz, H., Müller, R. T., et al. (1998). Targeting of
the myosin-I myr 3 to intercellular adherens type junctions induced by dominant active Cdc42 in HeLa
cells. Journal of Cell Science, 111, 2779–2788.
Truskey, G. A., Burmeister, J. S., Grapa, E., & Reichert, W. M. (1992). Total internal reflection fluores-
cence microscopy (TIRFM). II. Topographical mapping of relative cell/substratum separation distances.
Journal of Cell Biology, 103, 491–499.
Wilke, K., Wick, K., Keil, F. J., Wittern, K.-P., Wepf, R., & Biel, S. S. (2008). A strategy for correlative
microscopy of large skin samples: towards a holistic view of axillary skin complexity. Experimental
Dermatology, 17, 73–80.
CHAPTER 5
Labeling of Ultrathin Resin Sections for

Gunar Fabig*,§, Susanne Kretschmar*, Susanne Weiche†,
Dominic Eberle*, Marius Ader* and Thomas Kurth*
* Center for Regenerative Therapies, TU Dresden, Fetscherstraße 105, D-01307 Dresden, Saxony, Germany
† Biotechnology Center, TU Dresden, Tatzberg 47-49, D-01307 Dresden, Saxony, Germany
§ Present Address: Medical Theoretical
Center, Medical Faculty Carl Gustav Carus, TU Dresden,
Fiedlerstraße 42, D-01307 Dresden, Saxony, Germany
Abstract
I. Introduction
II. Rationale
III. Methods
A. Fixation and PLT-Embedding of Samples for On-Section CLEM
B. General Labeling Protocol
C. CLEM of GFP-Labeled Rhodopsin in Retinas of Mice Using Prot A Gold and
Fluorochrome-Conjugated Antibodies
D. CLEM of α-Tubulin in the Microtubular Manchette of Mouse Spermatids
IV. Materials
V. Summary and Outlook
Acknowledgments
References
Abstract
Correlative microscopy combines the versatility of the light microscope with the
excellent spatial resolution of the electron microscope. Here, we describe fast and
simple methods for correlative immunofluorescence and immunogold labeling on
the very same ultrathin section. The protocols are demonstrated on sections of tissue
samples embedded in the methacrylate Lowicryl K4M. Ultrathin sections are mounted
on electron microscopy (EM) grids and stained simultaneously with fluorescent and
gold markers. For the detection of primary antibodies, we applied either protein A gold
or immunoglobulin G (IgG) gold in combination with secondary antibodies coupled
76 Gunar Fabig et al.
to Alexa488 or Alexa555. Alternatively, the correlative marker FluoroNanogold was

used, followed by silver enhancement. The samples have to be analyzed first at the
light microscope and then in the transmission electron microscope (TEM), because
the fluorescence is bleached by the electron beam. Labeled structures selected at the
fluorescence microscope can be identified in the TEM and analyzed at high resolution.
This way, fluorescent signals can be directly correlated to the corresponding subcellular
structures in the area of interest.
I. Introduction
Light microscopy in combination with a plethora of different fluorescent markers

for the analysis of living or fixed cells has revolutionized biology and biochemistry.
Recent developments in fluorescence microscopy (FLM) even narrowed the resolution
gap between light and electron microscopy (EM) (for recent reviews see, Hell, 2007,
2009). Although conventional FLMs, such as widefield FLM or confocal laser scanning
microscopes (CLSM), are limited by diffraction, the super-resolution microscopes, such
as stimulated emission depletion (STED) or photoactivatable localization microscopes
(PALM), have extended the diffraction barrier down to 20–60 nm (Hell, 2007, 2009).
The major drawback of these super-resolution techniques is that they are confined to
fluorescence. In other words, one can see the fluorescence emitted from labeled struc-
tures but not the unlabeled reference space. EM, on the other hand, is able to reveal
those subcellular details at high resolution but cannot be used for in vivo imaging.
Correlative light and electron microscopy (CLEM) combines the best of both worlds,
i.e., the versatility of fluorescent markers and the high spatial resolution of the EM, to
analyze dynamic subcellular architecture at high resolution (Brown et al., 2009; Gaietta
et al., 2002, 2006; Giepmans et al., 2006; Giepmans, 2008; Grabenbauer et al., 2005;
McDonald, 2009; Polishchuk et al., 2000; van Rijnsoever, et al., 2008; van Weering et al.,
2010; Verkade, 2008). However, many CLEM approaches, in particular, those combin-
ing in vivo fluorescence and transmission electron microscope (TEM)-imaging, are used
preferentially in cell culture systems (Brown et al., 2009; McDonald, 2009; Polishchuk
et al., 2000). CLEM methods are not easily applicable to “large” multicellular specimens,
such as tissue samples or whole embryos, with the exception of rather small organisms,
such as Caenorhabditis elegans and others (Kolotuev et al., 2010; Müller-Reichert et al.,
2008; Sims & Hardin, 2007). Finally, many CLEM approaches focus on single subcel-
lular events such as vesicle fusion or budding, and therefore accumulation of data that are
sufficient for statistical evaluation is laborious and time consuming.
An alternative approach for the analysis of cells and tissues is the correlative on-
section labeling of either resin or Tokuyasu cryosections using fluorochrome-coupled
antibodies and gold probes (Cortese et al., 2009; Robinson & Takizawa, 2009;
Schwarz & Humbel, 2007, 2009; Takizawa & Robinson, 2006; Vicidomini et al.,
2008, 2010). This approach has several of the following advantages: (1) Using sec-
tions through cell pellets or tissues, many areas of interest can be analyzed and statisti-
cally evaluated. This reduces the risk to lose areas of interest during EM-preparation.
5. Labeling of Ultrathin Resin Sections for Correlative Light and Electron Microscopy 77
(2) Semithin (200–400 nm) and ultrathin (50–100 nm) sections can be analyzed by
FLM. In such samples, fluorescence is emitted from a very thin optical plane, resulting in
a very precise signal that is perfectly in focus (Schwarz & Humbel, 2007, 2009). (3) The
method is relatively easy to use and widely applicable. (4) Many different antigens
may be analyzed on sections from a single sample.
Here, we describe fast and reliable protocols for the correlative staining of ultrathin
resin sections for FLM and TEM. Correlative labeling of sections is always performed
on sections from the same block/sample. However, fluorescence and gold labeling may
be done either separately on adjacent sections (Hoffmann & Schwarz, 1996; Kurth,
2003; Kurth et al., 1996, 1999, 2010; Robinson & Takizawa, 2009; Schwarz et al.,
1993a; Schwarz, 1994, 1998; Schwarz & Humbel, 2007, 2009; Takizawa & Robinson,
2006), or on the very same section mounted on a finder grid (Cortese et al., 2009;
Vicidomini et al., 2008, 2010). The choice of either method is dependent on the size
and abundance of the structures of interest. Here, we focus on the correlative label-
ing of single ultrathin sections from resin-embedded tissue blocks. The protocols
are demonstrated with two different antibodies and in two different mouse tissues:
anti-GFP staining of cells expressing a rhodopsin-GFP fusion protein in the retina, and
anti-α-tubulin staining in the spermatids of the testis.
II. Rationale
CLEM, especially the combination of in vivo fluorescence imaging and EM, is a

powerful tool to unravel subcellular processes, such as Golgi-dynamics or vesicle traf-
ficking, in hitherto unprecedented detail. However, many CLEM approaches focus on
cell culture models, and much workload has to be invested to study a single subcellular
event. To address the distribution of labeled cells and structures by CLEM in tissues
or in embryos, alternative methods are needed such as correlative on-section labeling
of semi- to ultrathin sections for fluorescence and EM. Here we provide some simple
and reliable protocols for the correlative fluorescence and gold labeling of antigens
on single ultrathin methacrylate sections (Lowicryl K4M) mounted on grids. Labeled
sections are consecutively analyzed in the FLM and the TEM. This way, many cells or
labeled subcellular structures can be analyzed in situ on sections from a single tissue
sample.
III. Methods
A. Fixation and PLT-Embedding of Samples for On-Section CLEM

Progressive lowering of temperature (PLT)-embedding into acrylic resins for immu-
nocytochemistry has been introduced decades ago to gradually decrease the tempera-
ture while the sample is dehydrated in a graded series of ethanol (Carlemalm et al.,
1982; for a recent review see Cavalier & Spehner, 2009). Compared to room tempera-
ture embedding, PLT reduces the denaturation of proteins and the extraction of lipids
d uring dehydration and infiltration. At low temperatures, the samples are infiltrated with
either Lowicryl K4M (−35°C) or Lowicryl HM20 (−35°C to −50°C) and polymerized
by UV-light. PLT-embedding is a versatile embedding method for immuno-EM and a
valuable alternative to freeze-substitution (FS) followed by low temperature embedding
(Humbel, 2009; Humbel & Schwarz, 1989; Schwarz, Hohenberg, & Humbel, 1993b) or
the Tokuyasu method (Slot & Geuze, 2007; Tokuyasu, 1980).
Fixation: For immunolabeling, samples have to be fixed mildly to preserve as
much antigenicity as possible (Griffiths, 1993). Commonly used fixatives are 2–4%
paraformaldehyde (PFA) with (or without) low concentrations of glutaraldehyde (GA,
0.05–0.5%). The retina samples presented here were fixed by perfusion with 4% PFA
in 0.1 M phosphate buffer (PB) pH 7.4 (Figs 1–3), and the testis samples were fixed
by immersion after dissection of the animal in 4% PFA in 0.1 M PB pH 7.4 (Figs 4–7).
Alternatively, samples may be cryoimmobilized through high-pressure freezing
(HPF), followed by FS and low temperature embedding, a combination of methods
considered as the “gold standard” for good preservation of both antigenicity and ultra-
structure (for further information see Schwarz & Humbel, 2009; Studer et al., 2001,
2008). However, HPF/FS is limited to samples below 200 µm thickness. Large tissue
blocks have to be dissected before HPF, for example, by using a biopsy punch (Studer
et al., 2008).
Dehydration: Samples are dehydrated in a graded series of ethanol/water mixtures
at progressively lower temperatures according to the following regime:
• 30%, 50% at 0–4°C for 45 min each, in the fridge or on ice
• 70% (and 80%, optional) at −20°C for 1 h (each), in the freezer
• 90%, 96%, 2 × 100% at −35°C, 1 h each, in the Leica AFS2 FS device.
Infiltration: Infiltration occurs in the AFS2 at −35°C
• 1 part ethanol/1 part K4M for 1 h
• 1 part ethanol/2 parts K4M for 1 h
• Pure K4M overnight
• Pure K4M 1–2 × 3 h.
Embedding and polymerization: Samples can be processed and embedded in 0.5 mL
reaction tubes, or transferred for embedding into flat embedding forms as available for
the AFS2 (Leica). Samples are polymerized by UV-irradiation for 24–48 h at −35°C
(AFS2 with mounted UV-lamp). Samples in closed 0.5 mL reaction tubes are covered
with strips of aluminum foil to trigger polymerization through indirect irradiation by
UV-light. This way, irregular polymerization can be prevented. After 12–24 h the alu-
minum strips can be removed, and the samples are irradiated for another 12–24 h. After
complete polymerization of the resin, the samples are heated to room temperature and
the K4M blocks are transferred to a fume hood to get rid of the remaining volatile and
potentially dangerous Lowicryl vapors. The samples remain in the fume hood until the
characteristic Lowicryl “smell” has vanished.
Sectioning: Blocks are mounted for ultramicrotomy, and semithin sections are pro-
duced and stained with toluidine blue/borax (Trump, Smuckler, & Benditt, 1961) to
select potentially interesting areas at the light microscope. For correlative immunola-
beling, ultrathin sections are collected on slot, mesh, or, preferably, finder grids.
Fig. 1 Correlative light and electron microscopy (CLEM) on ultrathin resin sections. (A) Schematic diagram
of the procedure. After fixation and embedding, tissue sections are mounted on EM grids and stained simultane-
ously for fluorescence light microscopy (FLM) and transmission electron microscopy (TEM). Samples are first
imaged in the FLM and then in the TEM, section through the mouse retina: outer segments of photoreceptor
cells labeled (red); asterisk, landmark that serves as a reference point. (B) Workflow for CLEM on sections using
either Prot A gold/IgG gold (top) or FluoroNanogold (bottom). (C) Fluorescence after TEM-imaging. Fluores-
cence is still intact on grid bars (arrows) but completely abolished in areas that were inspected in the TEM (white
asterisks). OS layer of the outer segments of the photoreceptors as indicated by dashed lines. (See color plate.)
B. General Labeling Protocol

Our general strategy is to label primary antibodies with protein A (Prot A) gold
or immunoglobulin G (IgG) gold (10 nm), and subsequently fill up unoccupied
binding sites with secondary antibodies conjugated to fluorochromes (Alexa488 or
Alexa555) (Fig. 1). Prot A is a bacterial protein that binds the Fc-domain of IgG
Fig. 2 CLEM of rhodopsin-GFP in resin-embedded mouse retina. (A) Schematic of the labeling approach.
The anti-GFP is a rabbit polyclonal antibody (ab 290 from Abcam). (B) Semithin section of the retina with
the characteristic layers, toluidine blue/borax staining. (C) Overview of an ultrathin section at the FLM. The
arrow indicates the outer segment displayed in D–F. The asterisks in C and D serve as landmarks and indicate
identical positions. (D) Low magnification EM micrograph of the receptor cells. (E) The boxed area indicated
in D at higher magnification. (F) The outer segment indicated by the arrowheads in C and D at high magni-
fication. The membrane discs are densely labeled. (G–H) Second example illustrating the CLEM approach.
(G) Labeled outer segments, fluorescence; boxed area indicates the region displayed in H. (H) Low magnifica-
tion EM micrograph. The square indicates an area of an outer segment displayed in I. (I) Immunogold labeling
of Rho-GFP a high magnification. (See color plate.)
molecules from different species, such as rabbit, pig, guinea pig, dog, and human.
On the other hand, it only weakly binds to antibodies from sheep, goat, donkey, rat
(with the exception of IgG2C and IgG1), or mouse (with the exception of IgG2A and
IgG2B), and it does not bind to chicken antibodies (Langone, 1982). IgG’s can be
raised in different species and can be used as specific secondary detection markers
for indirect immunolabeling. Gold particles reduce the binding capacities of both
Fig. 3 CLEM of single transplanted cells. GFP-labeled photoreceptor precursor cells were transplanted into
the subretinal space of wild-type tissue. After 3 weeks the mouse was perfusion-fixed and the retina dissected
and embedded in Lowicryl K4M. Sections were mounted on finder grids and stained with rabbit-anti-GFP
(TP401 from Torrey Pines) followed by Prot A gold and goat-anti-rabbit Alexa488. (A) FLM image, the cluster
of transplanted cells is visible at the periphery (arrows), some cells are integrated into the tissue (arrowheads);
onl, outer nuclear layer. (B) Some of the integrated labeled cells are highlighted (1–3) and a fold in the section
(*) serves as a landmark. (C) Low magnification TEM-image, the labeled outer segments are marked (1–3,
green dotted line), an unlabeled outer segment is indicated by the blue dotted line, the landmark fold by an
asterisk. (D, E) Two of the labeled cells displayed in A–C are shown at higher magnification (1, 2, green dotted
lines). The labeling is clearly visible, as well as the typical substructure of the outer segment. An unlabeled outer
segment of the host tissue is also displayed (blue dotted line); os, outer segment. (See color plate.)
Fig. 4 CLEM of ultrathin resin sections through mouse testis using Prot A gold, a bridging antibody (rabbit-
anti-mouse) and goat-anti-rabbit Alexa488. (A) Schematic of the labeling procedure. (B) Overview image
showing a section through a seminiferous tubule with α-tubulin staining predominantly in the inner part, where
the spermatids differentiate. The square indicates the region displayed in C. (C) Labeling of the spermatid man-
chettes is clearly visible, the boxed area indicates the two cells displayed in D and E. (D) High magnification
fluorescence image of labeled spermatids; nuc, nucleus. (E) Low magnification EM micrograph of the same
area. Squares in D and E indicate the region shown in F; ac, acrosome. (F) EM micrograph showing the gold-
labeled manchette (ma). (See color plate.)
Prot A and IgG. In general, labeling intensity decreases with the increasing size of
the gold particles (Humbel & Biegelmann, 1992; Humbel et al., 1998). Gold probes
of 10 nm, therefore, will probably not occupy all the potential binding sites. In addi-
tion, the fluorochrome-conjugated antibodies may bind target IgGs at different sites
than Prot A, which may leave even more binding sites for the fluorescent markers
(Griffiths, 1993).
The labeling is basically done according to the following procedure (all steps to be
carried out at room temperature):
1. 2 × 5 min 1% bovine serum albumine (BSA), fraction V in PBS (hydration of
sections and blocking of unspecific binding sites)
2. Primary antibody (rabbit-anti-GFP) in 1% BSA/PBS for 1h
3. 4–5 × 2 min PBS
4. Optional: (a) Bridging antibody, e.g., rabbit-anti-mouse IgG; necessary when the
primary antibody is a mouse monoclonal and Prot A gold serves as a colloidal
gold marker (b). FluoroNanogold-conjugated antibody; proceed with washing
(Step 5) and then go to Step 13
5. Optional: after bridging antibody or FluoroNanogold: 5 × 2 min PBS
6. Prot A gold or IgG gold in 1% BSA/PBS for 30 min to 1h (gold probe)
7. 3× short changes in PBS
Fig. 5 CLEM of ultrathin resin sections using Prot A gold, a bridging antibody (rabbit-anti-mouse) and goat-
anti-mouse Alexa488 for labeling. (A) Schematic diagram. (B) FLM image, the spermatid in the red box is
selected for EM-imaging. (C) EM micrograph of the spermatid in B; ma, manchette; nuc, nucleus. (D) Higher
magnification of the boxed area in C, immunogold labeling is clearly detectable. (See color plate.)
8. 4 × 2 min PBS

9. 1% GA in PBS for 5 min in the hood (postfixation; this step is optional and is
dependent on how much time the imaging of specific fluorescence will take;
fixes the Prot A gold; can be omitted when IgG gold is used)
10. 4–5 × 2 min PBS
11. Secondary antibody conjugated to Alexa488 or Alexa555 (fluorescence marker)
in 1% BSA/PBS for 30 min (from here on samples have to be protected from
light)
12. 4–5 × 2 min PBS
13. 5–10 min DAPI (1 µg/mL, counterstaining of nuclei)
14. 5 × 1 min water
15. Mounting of grids for FLM in 50% glycerin/water between a microscope slide
and a coverslip (Vicidomini et al., 2008)
16. Imaging at the FLM. It is mandatory to finish FLM before imaging at the TEM,
because the electron beam will destroy the fluorescence (Fig. 1(C))
18. Optional: if FluoroNanogold-conjugated antibodies were used, 5 × 1 min addi-
tional washing steps in water, followed by silver enhancement using the R-Gent
SE-Silver enhancement kit (Aurion), incubation 30–60 min
Fig. 6 Alternative approaches for CLEM on ultrathin sections using IgG gold in combination with Alexa488 conjugated antibodies.
(A) Primary antibodies are detected with goat-anti-mouse IgG gold (10 nm) and goat-anti-mouse Alexa488. (B) Primary antibodies are
detected with goat-anti-mouse IgG gold (10 nm) and donkey-anti-goat Alexa488. Arrows in (B) indicate gold particles; ma, manchette;
nuc, nucleus. The images show FLM and EM images of indicated areas. (See color plate.)
1 9. Optional: after silver enhancement: 5 × 3 min water

20. Staining with 2–4% uranyl acetate in water, 5–10 min
22. Sections are dried and imaged at the TEM.
In principle, this protocol is the basis for all experiments described here. Variables
are the use of bridging antibodies (Figs. 4 and 5(A)), the use of Prot A gold versus IgG
gold (Figs. 2–5 versus Fig. 6), and different secondary fluorochrome-coupled antibod-
ies (Figs. 2–6). Alternatively, the primary antibodies are detected with FluoroNanogold
that is applied at Step 4 and has to be silver enhanced (Step 18) to get recognizable
particles at the EM level (Fig. 7). A summary of the different labeling approaches is
displayed in Table I.
Fig. 7 CLEM of ultrathin resin sections using FluoroNanogold. (A) Schematic diagram. (B) FLM image, the
square indicates the region shown in C. (C) Staining in the microtubular manchettes of spermatids; the box
indicates the cell displayed in D and E at higher magnification (D) Close-up of the cell highlighted in C; nuc,
nucleus. (E) Low magnification EM micrograph of the very same cell; squares in D and E indicate the region
displayed in F. (F) TEM-micrograph at high magnification with silver-enhanced gold particles in the spermatid
manchette (ma). (See color plate.)
C. CLEM of GFP-Labeled Rhodopsin in Retinas of Mice Using Prot A Gold

and Fluorochrome-Conjugated Antibodies
GFP is among the most commonly used markers for cellular proteins (for a recent
review see Giepmans et al., 2006). Although designed for in vivo imaging, it can also
be used to identify certain proteins or cell types after fixation of cell and tissue samples.
It can be visualized either by its inherent fluorescence, which “survives” many fixation
protocols including PFA and moderate amounts of GA (Chalfie et al., 1994; Luby-
Phelps et al., 2003), or by immunolabeling with anti-GFP antibodies (Luby-Phelps
et al., 2003).
To visualize a translational rhodopsin-GFP (Rho-GFP) fusion protein in the retina
of transgenic Rho-GFP mice (Chan et al., 2004), the animals were perfusion-fixed (see
Section III A). The eyes were excised and retina pieces, without the removed retinal
pigment epithelium (RPE), were embedded in Lowicryl K4M. Ultrathin sections were
subsequently mounted on mesh grids and labeled with anti-GFP (rabbit-anti-GFP, ab
290 from Abcam or TP401 from Torrey Pines). Although the GFP-fluorescence can
be seen in thin Lowicryl sections and even better in Tokuyasu cryosections (data not
shown), the fluorescent signal achieved by antibody staining is much stronger, espe-
cially when sections come from older K4M blocks.
Table I
Summary of different approaches for correlative fluorescence and gold labeling on ultrathin resin sections.
Primary Ab IgG gold, bridging Ab, Prot A Prot A gold Fluorescent marker Silver Figure
gold, FluoroNanogold enhancement
(SE)
Anti-GFP (R-IgG) Prot A gold G-anti-R Alexa488 Figs 2 and 3

or Alexa555
Anti-α-tubulin Bridging Ab (R-anti-M IgG) Prot A gold G-anti-R Alexa488 Fig. 4
(M-monoclonal)
Anti-α-tubulin Bridging Ab (R-anti-M IgG) Prot A gold G-anti-M Alexa488 Fig. 5
(M-monoclonal)
Anti-α-tubulin G-anti-M 10 nm gold G-anti-M Alexa488 Fig. 6(A)
(M-monoclonal)
Anti-α-tubulin G-anti-M 10 nm gold D-anti-G Alexa488 Fig. 6(B)
(M-monoclonal)
Anti-α-tubulin G-anti-M Alexa488 SE Fig. 7
(M-monoclonal) FluoroNanogold F(a,b)
fragments
Abbreviations: Ab, antibody; D, donkey; G, goat; IgG, immunoglobulin G ; M, mouse; Prot A, protein A; R, rabbit.
Ultrathin sections were mounted on either mesh or finder grids and stained accord-
ing to the aforementioned protocol (primary antibodies from rabbit, Prot A gold, and
goat-anti-rabbit conjugated to Alexa-fluorochromes as detection markers, see Fig.
2(A)). Two different samples were analyzed (Figs 2 and 3), which are as follows:
First, retinas from Rho-GFP+/− heterozygous mice displayed strong fluorescent stain-
ing of the outer segments of photoreceptor cells (Fig. 2(C), and (G)). Selected pieces
of outer segments can be imaged in the TEM at higher magnification (Fig. 2(C)–(F)
and (G)–(I)). Strong gold labeling in the membrane stacks of the outer segments indicates
the Rho-GFP fusion protein (Fig. 2(F) and (I)). Thus, the protein of interest is highly
abundant in the sample and could be localized in the clearly defined outer segments of
the photoreceptor cells.
Second, a group of photoreceptor precursor cells expressing Rho-GFP were iso-
lated from postnatal day 4 transgenic mouse retinas and transplanted into the subretinal
space of a wild-type mouse. Most of the transplanted cells form a cluster between
retina and RPE, but some GFP-labeled cells seemingly integrate into the photore-
ceptor layer (Eberle et al., 2011). After 3 weeks, the mouse was fixed and the ret-
ina dissected. Small tissue pieces containing dispersed GFP-positive cells were
selected, embedded in K4M, sectioned, mounted on finder grids, and stained (Fig. 3).
In addition to the cell cluster on top of the outer segment layer (Fig. 3(A), arrows),
single GFP-positive cells can be identified in the photoreceptor cell layer (Fig. 3(A) and
(B), arrowheads). At the EM level, these structures were identified as outer segments
of photoreceptor cells, labeled with Prot A gold (Fig. 3(C) and (E)). Adjacent outer
segments of the host tissue were unlabeled (Fig. 3(E)). This example illustrates that the
method is quite effective in identifying a minor cell population within a tissue sample.
D. CLEM of α-Tubulin in the Microtubular Manchette of Mouse Spermatids

Spermatozoa are continuously generated from a pool of spermatogonial stem
cells. After meiosis, they differentiate into spermatids with a condensed nucleus,
a long flagellum, and the acrosome at the tip of the spermatid head. These devel-
opmental changes include the formation of a spermatid–manchette complex that
is rich in microtubules around the nucleus (Fawcett et al., 1971; Kierszenbaum &
Tres, 2004). The complex can be visualized with anti-α-tubulin antibodies (mouse
monoclonal DM1A, Sigma-Aldrich) and is presented here as a second proof-of-
principle example for correlative on-section labeling of resin sections. Mouse testis
was dissected, immersion fixed in 4% PFA/PB, and embedded in K4M. Ultrathin
sections were mounted on slot grids for further correlative labeling.
1. On-Section Labeling with Prot A Gold and Fluorochrome-Conjugated Antibodies

Protein A has only weak or no binding affinity to antibodies from sheep, goat, don-
key, rat, chicken, and mouse (Langone, 1982). If mouse monoclonal antibodies are to
be used, bridging antibodies (preferably from rabbit or guinea pig) have to be added to
the protocol (after Step 3, see Section III B). We use rabbit-anti-mouse IgGs (diluted in
1% BSA, incubation for 30 min, followed by washing in PBS, 5 × 2 min). The bridg-
ing antibodies are then again detected with Prot A 10 nm gold and goat-anti-rabbit
Alexa488 (Fig. 4(A)). The spermatid manchettes are easily detectable at the light
microscope, even at low magnifications (Fig. 4(B)). After selecting an area of interest in
the FLM (Fig. 4(B)–(D)), the very same area was identified in the TEM (Fig. 4(E)), and
the corresponding immunogold labeling visualized at high magnification (Fig. 4(F)).
In a slightly different approach, the fluorescent marker (goat-anti-mouse Alexa488)
is directed against the primary antibody (Fig. 5(A)). Although there is competition
between the bridging antibody and the fluorescent marker, the fluorescence is of suf-
ficient intensity and specificity (Fig. 5(B)) and Prot A gold labeling is apparently not
affected (Fig. 5(C) and (D)). In these experiments, we omitted the postfixation step to
avoid fixation of bridging antibodies that are not bound to Prot A gold and to enable the
binding of fluorochrome-conjugated antibodies.
2. On-Section Labeling with IgG Gold and Fluorochrome-Conjugated Antibodies

When IgG gold is used instead of Prot A gold, the bridging antibody can be omitted,
and the primary antibodies are detected with goat-anti-mouse IgG 10 nm gold and goat-
anti-mouse Alexa488 (Fig. 6(A)) or donkey-anti-goat Alexa488 (Fig. 6(B)). Correlative
labeling works in both cases. The immunogold labeling of the spermatid manchette is
weaker, which may be explained by the lack of the bridging step that also enhances the
labeling intensity (Slot & Geuze, 2007). However, in other experiments, we experienced
that Prot A gold generally revealed more intense staining than the IgG gold probes.
3. On-section Labeling with FluoroNanogold

FluoroNanogold is a correlative marker (Powell et al., 1997; Robinson et al.,
2000; Takizawa et al., 1998). It is an F(a,b)-fragment conjugated to both an Alexa-
fluorochrome and an ultrasmall gold compound (Nanogold, about 1–1.4 nm).
Using this approach, correlative labeling is a one-step reaction, after which fluo-
rescent and gold markers are both bound to the very same antigen (Fig. 7(A)).
However, Nanogold has to be silver enhanced to be detectable at convenient magni-
fications in the TEM (Lah et al., 1990; Stierhof, 2009; Stierhof et al., 1991). For the
samples displayed here, the R-Gent SE-kit from Aurion has been used according to
the manufacturers’ instructions. Again, the spermatid manchettes can be selected in
the FLM and identified in the TEM (Fig. 7(B)–(E)). At high magnification, the silver-
enhanced gold compound is clearly visible (Fig. 7(F)). The fluorescence, however, was
much weaker compared to any other approach presented here. This may be due to
quenching by the gold particles, although it has been previously reported not to be a
problem with FluoroNanogold (Powell et al., 1997).
IV. Materials
• Fixatives: For immunocytochemistry: 4% PFA in 0.1 M PB pH 7.4 (PB; 0.2 M

stock solution: 19 mL of buffer A (27.6g NaH2PO4·1H2O in 1000mL distilled
water) + 81 mL buffer B (35.6g Na2HPO4·2H2O in 1000mL distilled water)), or in
100 mM HEPES pH 7.4 or 4% PFA + 0.1% GA with the same buffer options.
• Embedding media: Lowicryl K4M kit (EMS # 14330), work always under fume
hood, because all Lowicryl formulations (K4M, HM20, K11M) are potentially
dangerous and allergenic.
• Chemicals: BSA (Merck, # 1.12018), GA (25% in water, EMS # 16220), HEPES
(EMS # 16782), PFA (granular, EMS # 19208), uranyl acetate (Polysciences #
21447), toluidine blue/borax solution: 1% toluidine blue (EMS # 22050), 0.5%
sodium tetraborate (EMS # 21130).
• Primary antibodies: Rabbit-anti-GFP (ab 290 from Abcam, Cambridge, UK;
or TP401 from Torrey Pines, San Diego, USA; both antibodies work nicely on
Tokuyasu and K4M sections); mouse monoclonal antibody against α-tubulin
(DM1A from Sigma-Aldrich, St. Louis, USA).
• Secondary antibodies and detection systems: We used several Alexa488- and
Alexa555-conjugated secondary antibodies for immunofluorescence (all from
Invitrogen/Molecular Probes), and rabbit-anti-mouse IgG (Sigma-Aldrich) as a
bridging antibody. DAPI (Invitrogen/Molecular Probes) served as a counterstain.
For EM, we use protein A conjugated to 10 nm gold particles (Aurion, Wageningen,
the Netherlands). Alternatively, goat-anti-mouse IgG 10 nm gold (Aurion) was
used. Goat-anti-mouse Alexa488 FluoroNanogold F(a,b)-fragments were pur-
chased from Invitrogen (# 24920) and the R-Gent-SE-EM silver enhancement kit
from Aurion (# 25521).
• Tools and materials: Slot grids (copper and nickel, Plano, # G 255C and # G
2500Ni), hexagonal mesh grids, 100 mesh (copper and nickel, Plano, # G 2410C
and # G 2410Ni), 200 mesh finder grids (nickel, EMS, # G 200F1-Ni), anti-
magnetic tweezers (model 5B, EMS # 78 320-5B), diamond knives (Diatome,
Biel, Switzerland).
• Instruments: Leica AFS2 FS unit and Leica EM UC6 ultramicrotome (both Leica
Microsystems, Vienna, Austria); fluorescence microscope: Keyence Biozero (BZ
8000, Keyence, Osaka, Japan), closed digital and inverted fluorescence micro-
scope; electron microscope: FEI Morgagni 268D (FEI, Eindhoven, The Nether-
lands), operated at 80 kV.
V. Summary and Outlook
Correlative labeling of ultrathin sections for FLM and TEM is a fast and versatile
method to analyze the distribution of proteins in the context of tissue architecture and
cellular ultrastructure. In the FLM, large areas of a specimen can be scored and specific
sites can be selected for EM, where labeled and unlabeled structures are analyzed simul-
taneously at high resolution. Correlative labeling can be successfully performed using
fluorescent and gold markers in different combinations (Prot A gold or IgG gold, different
fluorochrome-conjugated antibodies, FluoroNanogold). In our laboratory we prefer to use
Prot A gold as the colloidal marker, because it is a versatile probe that binds to antibod-
ies from several species (see Section III B). If mouse monoclonals are used as primary
antibodies, detection occurs with rabbit-anti-mouse bridging antibodies and Prot A gold.
By labeling serial thin sections it may even be possible to search rare cell popula-
tions or subcellular structures in larger tissue volumes. The protocol presented here can
easily be adapted to an automated approach using an immunogold labeling automat
(EM IGL, Leica Microsystems, data not shown). This way, many grids with correla-
tively labeled sections can be analyzed in the FLM, and a significantly lower number
of “interesting” samples be selected for further EM analysis.
FLM, however, is not only useful to select areas of interest. Semithin to ultrathin
sections (50–400 nm) that were prepared for EM are excellent samples for high-res-
olution light microscopy (Robinson and Takizawa, 2009; Schwarz & Humbel, 2007,
2009; Vicidomini et al., 2008). First, on-section immunolabeling does not demand the
deleterious permeabilization steps with detergents or organic solvents, routinely used
for the immunolabeling of cell culture models (Melan & Sluder, 1992; Schnell et al.,
2012). Therefore, the preservation of subcellular structures is much better than in rou-
tine tissue culture or histology samples, especially when the EM samples are prepared
by HPF, FS, and low temperature embedding (Schwarz & Humbel, 2007, 2009).
In addition, there is nearly no out-of-focus signal blurring the image, because the
section thickness is below the depth of focus of high aperture objectives resulting in
high z-resolution. Therefore, fluorescent samples can be analyzed with nonconfo-
cal widefield microscopes. However, it may be promising to combine ultrathin sec-
tions with super-resolution microscopes (e.g., 4Pi, STED, PALM, STORM) to further
improve the resolution of labeled structures at the light microscopical level. CLEM
with 4Pi-imaging of living cells and subsequent TEM-analysis has been recently
performed using a cell culture model (Perinetti et al., 2009). However, even at the
highest possible resolution using super-resolution FLM, one thing is still missing––
the subcellular reference space including the nonlabeled structures in the vicinity of
the fluorochromes or fluorescent proteins. This information can only be obtained from
the electron microscope and this is the reason why it is advantageous to combine even
super-resolution LM with EM.
Finally, a series of correlatively labeled sections can be used for 3D reconstruction
of tissue blocks. This option may be promising for the study of small populations of
labeled cells and their interactions with other nonlabeled cells (e.g., adult stem and
progenitor cells, transplanted cells (Fig. 3), or embryonic cells, such as neural crest or
primordial germ cells). With this approach, larger volumes can be reconstructed. The
z-resolution is limited to section thickness, but a thickness of 50–70 nm may be suffi-
cient for many studies. Smaller structures can be reconstructed by electron tomography
of 200 nm thick CLEM-sections (Cortese et al., 2009; Vicidomini et al., 2008, 2010).
Taken together, on-section CLEM is a rapid and versatile method for the correlative
analysis of potentially numerous antigens on sections from the very same cell or tissue
sample. Our protocols are based on standard labeling techniques and do not require
additional sophisticated instrumentation. To increase throughput, however, the use of
immunogold labeling automats, like the Leica EM IGL, or the imaging of labeled sec-
tions in an integrated fluorescence and electron microscope as described by Agronskaia
et al. (2008) might be useful.
Acknowledgments
The authors wish to thank Katrin Daniel and Attila Toth for mouse testis samples,
John Wilson for Rho-GFP mice, and the European Fund for Regional Development
(EFRE) for financial support.
References
Brown, E., Mantell, J., Carter, D. A., Tilly, G., & Verkade, P. (2009). Studying intracellular transport using
high-pressure freezing and correlative light electron microscopy. Seminars in Cell & Developmental
Biology, 20, 910–919.
Carlemalm, E., Garavito, R. M., & Villiger, W. (1982). Resin development for electron microscopy and an
analysis of embedding at low temperature. Journal of Microscopy, 126, 123–143.
Cavalier, A., & Spehner, D. (2009). Progressive lowering of temperature for immunolabeling and in situ
hybridization. In A. Cavalier, D. Spehner, & B. M. Humbel (Eds.), Hand book of cryo-preparation
methods for electron microscopy (pp. 433–465). Boca Raton, FL: CRC Press.
Chalfie, M., Tu, Y., Euskirchen, G., Ward, W. W., & Prasher, D. C. (1994). Green fluorescent protein as a
marker for gene expression. Science, 262, 802–805.
Chan, F., Bradley, A., Wensel, T. G., & Wilson, J. H. (2004). Knock-in human rhodopsin-GFP fusions as
mouse models for human disease and targets for gene therapy. PNAS, 101, 9109–9114.
methods and new developments using Tokuyasu cryosections. Journal of Histochemistry Cytochemistry,
57, 1103–1112.
Eberle, D., Schubert, S., Postel, K., Corbeil, D., & Ader, M. (2011). Increased integration of transplanted
CD73-positive photoreceptor precursors into adult mouse retina. Investigative Ophthalmology & Visual
Science, 52, 6462–6471.
Fawcett, D. W., Anderson, W. A., & Phillips, A. M. (1971). Morphogenetic factors influencing the shape of
the sperm head. Developmental Biology, 26, 220–251.
Gaietta, G. M., Deerinck, T. J., Adams, S. R., Bouwer, J., Tour, O., Laird, D. W., et al. (2002). Multicolor and
Gaietta, G. M., Giepmans, B. N. G., Deerinck, T. J., Smith, W. B., Ngan, L., Llopis, J. , et al. (2006). Golgi
twins in late mitosis revealed by genetically encoded tags for live cell imaging and correlated electron
microscopy. PNAS, 103, 17777–17782.
Giepmans, B. N., Adams, S. R., Ellisman, M. H., & Tsien, R. Y. (2006). The fluorescence toolbox for assess-
ing protein location and function. Science, 312, 217–224.
Grabenbauer, M., Geerts, W. J. C., Fernadez-Rodriguez, J., Hoenger, A., Koster, A. J., & Nilsson, T. (2005).
857–862.
Griffiths, G. (1993). Fine structure immunocytochemistry. New York: Springer.
Hell, S. W. (2007). Far-field optical nanoscopy. Science, 316, 1153–1158.
Hell, S. W. (2009). Microscopy and its focal switch. Nature Methods, 6, 24–32.
Hoffmann, W., & Schwarz, H. (1996). Ependymins: meningeal-derived extracellular matrix proteins at the
blood-brain barrier. International Review of Cytology, 165, 121–158.
Humbel, B. M., & Schwarz, H. (1989). Freeze substitution for immunochemistry. In A. J. Verklej, & J. L.M.
Leunissen (Eds.), Immuno-gold labeling in cell biology (pp. 115–134). Boca Raton, FL:: CRC Press.
Humbel, B. M., & Biegelmann, E. (1992). A preparation protocol for postembedding immunoelectron
microscopy of Dictyostelium discoideum cells with monoclonal antibodies. Scanning Microscopy, 6,
817–825.
Humbel, B. M., de Jong, M. D., Müller, W. H., & Verkleij, A. J. (1998). Pre-embedding immunolabeling for
electron microscopy: an evaluation of permeabilization methods and markers. Microscopy Research and
Humbel, B. M. (2009). Freeze substitution. In A. Cavalier, D. Spehner, & B. M. Humbel (Eds.), Hand book
of cryo-preparation methods for electron microscopy (pp. 319–341). Boca Raton, FL: CRC Press.
Kierszenbaum, A. L., & Tres, L. L. (2004). The acrosome-acroplaxome-manchette complex and the shaping
of the spermatid head. Archives of Histology and Cytology, 67, 271–284.
Kolotuev, I., Schwab, Y., & Labouesse, M. (2010). A precise and rapid mapping protocol for correlative light
and electron microscopy of small invertebrate organisms. Biology of the Cell, 102, 121–132.
Kurth, T., Schwarz, H., Schneider, S., & Hausen, P. (1996). Fine structural immunocytochemistry of catenins
in amphibian and mammalian muscle. Cell and Tissue Research, 286, 1–12.
Kurth, T., Fesenko, I. V., Schneider, S., Münchberg, F. E., Joos, T. O., Spieker, T. P., et al. (1999). Immuno-
cytochemical studies of the interactions of cadherins and catenins in the early Xenopus embryo. Devel-
opmental Dynamics, 215, 155–169.
Kurth, T. (2003). Immunocytochemistry of the amphibian embryo – from overview to ultrastructure. Inter-
national Journal of Developmental Biology, 47, 373–383.
Kurth, T., Berger, J., Wilsch-Bräuninger, M., Kretschmar, S., Cerny, R., Schwarz, H. , et al. (2010). Electron
microscopy of the amphibian model systems Xenopus laevis and Ambystoma mexicanum. Methods of
Lah, J. J., Hayes, D. M., & Burry, R. W. (1990). A neutral pH silver development method for the visualization
of 1-nanometer gold particles in pre-embedding electron microscopic immunocytochemistry. Journal of
Histochemistry and Cytochemistry, 38, 503–508.
Langone, J. J. (1982). Applications of immobilized protein A in immunochemical techniques. Journal of
Immunological Methods, 55, 277–296.
Luby-Phelps, K., Ning, G., Fogerty, J., & Besharse, J. C. (2003). Visualization of identified GFP-expressing
cells by light and electron microscopy. Journal of Histochemistry and Cytochemistry, 51, 271–274.
Melan, M. A., & Sluder, G. (1992). Redistribution and differential extraction of soluble proteins in perm-
abilized cultured cells. Implications for immunofluorescence microscopy. Journal of Cell Science, 101,
731–743.
Müller-Reichert, T., Mäntler, J., Srayko, M., & O’Toole, E. (2008). Electron microscopy of the early
Caenorhabditis elegans embryo. Journal of Microscopy, 230, 297–307.
Perinetti, G., Müller, T., Spaar, A., Polishchuk, R., Luini, A., & Egner, A. (2009). Correlation of 4Pi- and
electron microscopy to study transport through single golgi stacks in living cells with super resolution.
Traffic, 10, 379–391.
Polishchuk, R. S., Polishchuk, E. V., Marra, P., Alberti, S., Buccione, R., Luini, A. , et al. (2000). Correlative
light-electron microscopy reveals the tubular-saccular ultrastructure of carriers operating between golgi
apparatus and plasma membrane. Journal of Cell Biology, 148, 45–58.
Powell, R. D., Halsey, C. M. R., Spector, D. L., Kaurin, S. L., McCann, J., & Hainfeld, J. F. (1997). A cova-
lent fluorescent-gold immunoprobe: Simultaneous detection of pre-mRNA splicing factor by light and
electron microscopy. Journal of Histochemistry and Cytochemistry, 45, 947–956.
Robinson, J. M., Takizawa, T., & Vandre, D. D. (2000). Applications of gold cluster compounds in immu-
nocytochemistry and correlative microscopy: Comparison with colloidal gold. Journal of Microscopy,
199, 163–179.
Robinson, J. M., & Takizawa, T. (2009). Correlative fluorescence and electron microscopy in tissues: Immu-
nocytochemistry. Journal of Microscopy, 235, 259–272.
Schnell, U., Dijk, F., Sjollema, K. A., & Giepmans, B. N. G. (2012). Immunolabeling artifacts and the need
for live-cell imaging. Nature Methods, 9, 152–158.
Schwarz, H., Müller-Schmid, A., & Hoffmann, W. (1993a). Ultrastructural localization of ependymins in the
endomeninx of the brain of the rainbow trout: Possible association with collagen fibrils of the extracel-
lular matrix. Cell and Tissue Research, 273, 417–425.
Schwarz, H., Hohenberg, H., & Humbel, B. M. (1993b). Freeze substitution in virus research: A preview.
In A. D. Hyatt, & B. T. Eaton (Eds.), Immunoelectron microscopy in virus diagnosis and research
(pp. 97–118). Boca Raton, FL:: CRC Press.
Schwarz, H. (1994). Immunolabeling of ultrathin sections for fluorescence and electron microscopy. In
B. Jouffrey, & C. Colliex (Eds.), Electron microscopy 1994, ICEM 13-Paris. France: Les Editions de
Physique, Les Ulis. (Vol. 3, pp. 255–256).
Schwarz, H. (1998). Correlative immunolabeling of ultrathin sections for light and electron microscopy. In
H. A. Calderon Benavides, et al. (Ed.), Electron microscopy 1998, ICEM 14 (pp. 865–866). Bristol, PA:
Institute of Physics Publishing.
sections. Methods in Molecular Biology, 369, 229–256.
Schwarz, H., & Humbel, B. M. (2009). Correlative light and electron microscopy. In A. Cavalier, D. Speh-
ner, & B. M. Humbel (Eds.), Hand book of cryo-preparation methods for electron microscopy (pp. 537–
565). Boca Raton, FL: CRC Press.
Sims, P. A., & Hardin, J. D. (2007). Fuorescence-integrated transmission electron microscopy images: Inte-
grating fluorescence microscopy with transmission electron microscopy. Methods in Molecular Biology,
369, 291–308.
Slot, J. W., & Geuze, H. J. (2007). Cryosectioning and immunolabeling. Nature Protocol, 2, 2480–2491.
Stierhof, Y. D., Humbel, B. M., & Schwarz, H. (1991). Suitability of different silver enhancement methods
applied to 1 nm colloidal gold particles: An immunoelectron microscopic study. Journal of Electron
Microscopy Technology, 17, 336–343.
Stierhof, Y. D. (2009). Immunolabeling of ultra-thin sections with enlarged 1 nm gold or Q-dots. In
A. Cavalier, D. Spehner, & B. M. Humbel (Eds.), Hand book of cryo-preparation methods for electron
microscopy (pp. 587–616). Boca Raton, FL: CRC Press.
Studer, D., Graber, W., Al-Amoudi, A., & Eggli, P. (2001). A new approach for cryofixation by high-pressure
freezing. Journal of Microscopy, 203, 285–294.
Studer, D., Humbel, B. M., & Chiquet, M. (2008). Electron Microscopy of high pressure frozen samples:
Bridging the gap between cellular ultrastructure and atomic resolution. Histochemistry and Cell Biology,
130, 877–889.
thin cryosections. Proof of principle. Journal of Histochemistry and Cytochemistry, 46, 1097–1102.
Tokuyasu, K. T. (1980). Immunocytochemistry of ultrathin frozen sections. Histochemical Journal, 12,
381–403.
Trump, B. F., Smuckler, E. A., & Benditt, E. P. (1961). A method for staining epoxy sections for light micros-
copy. Journal of Ultrastructure Research, 5, 343–348.
van Rijnsoever, C., Oorschot, V. M., & Klumperman, J. (2008). Correlative light-electron microscopy
(CLEM) combining live-cell imaging and immunolabeling of ultrathin cryosections. Nature Methods,
5, 973–980.
van Weering, J. R.T., Brown, E., Sharp, T. H., Mantell, J., Cullen, P. J., & Verkade, P. (2010). Intracellular
membrane trafficking at high resolution. Methods of Cell Biology, 96, 619–648.
Verkade, P. (2008). Moving EM: The rapid transfer system as a new tool for correlative light and electron
microscopy and high throughput for high-pressure freezing. Journal Microscopy, 230, 317–328.
Vicidomini, G., Gagliani, M. C., Canfora, M., Cortese, K., Frosi, F., Santangelo, C., et al. (2008). High data
Vicidomini, G., Gagliani, M. C., Cortese, K., Krieger, J., Buescher, P., Bianchini, P., et al. (2010). A novel
approach for correlative light electron microscopy analysis. Microscopy Research and Technique, 73,
215–224.
CHAPTER 6
3D HDO-CLEM: Cellular Compartment

Analysis by Correlative Light-Electron
Microscopy on Cryosection
Katia Cortese*,1, Giuseppe Vicidomini§,1, Maria Cristina
Gagliani*, Patrizia Boccacci*, Alberto Diaspro*,†,‡,§ and Carlo
Tacchetti*,†║
* MicroScoBio Research Center, Department of Experimental Medicine (DIMES), Department of Physics
(DIFI) and Department of Informatics, Bioengineering, Robotics and Information Science (DIBRIS),
University of Genoa, Genoa, Italy
† IFOM, Fondazione Istituto FIRC di Oncologia Molecolare, Milan, Italy
‡ LAMBS-IFOM, MicroScoBio, Department of Physics, University of Genoa, Italy
§ Nanophysics, Istituto Italiano di Tecnologia, Genoa, Italy
║Experimental Imaging Center, Scientific Institute San Raffaele, Milan, Italy
1 These authors have equally contributed.
Abstract
I. Introduction
II. Materials
A. Reagents
B. Antibodies
C. Sectioning and Imaging Equipments
D. Software
III. Methods
A. Preparing Reagents
B. Preparation of Ultrathin Cryosections for HDO-CLEM
C. Immunofluorescence and Immunogold Labeling for HDO-CLEM
D. HDO-CLEM
E. Hybrid Morphometry: A Systematic Error-Based Approach
IV. Notes
Acknowledgments
References

96 Katia Cortese et al.
Abstract
Fundamental to obtaining a depth-understanding of the function and structure of

cells is the ability to study and correlate their molecular topography with the ultrastruc-
tural morphology, for example, to visualize the position of a given protein relative to
a given cell compartment and its morphology. Standard fluorescence light microscopy
(FLM) relies on simple sample preparations, and localizes proteins in living or fixed
cells with a resolution in the range of few hundred nanometers, allowing large field of
view. However, FLM is unable to visualize the unlabeled cellular context. On the other
hand, electron microscopy (EM) techniques reveal protein topology with the resolu-
tion in a range of a few tens of nanometer, retains the cellular context, but can only be
applied on a limited field of view. Therefore, both approaches present shortcomings,
in terms of field of view, statistical output, resolution, sample preparation, and context
analysis, that can likely complement each other. To bridge the gap between FLM imag-
ing and EM, several laboratories have developed methods for correlative light-electron
microscopy (CLEM). In a nutshell, CLEM enables one to investigate the same exact
region of interest utilizing the two microscope platforms, and thereby virtually combine
their capabilities.
In this chapter, we describe a protocol based on immunolabeling of Tokuyasu cryo-
sections that allows correlation of LM and EM images with excellent preservation of
cellular ultrastructure. We will refer to this method as high-data-output CLEM (HDO-
CLEM).
The major benefits of HDO-CLEM are the possibility to (1) correlate several hun-
dreds of events at the same time, (2) perform three-dimensional (3D) correlation, (3)
immunolabel both endogenous and recombinantly tagged proteins at the same time,
and (4) combine the high data analysis capability of FLM with the high precision of
transmission EM in a CLEM hybrid morphometric analysis.
We have identified and optimized critical steps in sample preparation, defined rou-
tines for sample analysis and retracing of regions of interest, developed software for
semi/fully automatic 3D FLM reconstruction and set the basis for a hybrid light/EM
morphometry approach.
I. Introduction
In modern cell biology, imaging techniques such as fluorescent light microscopy

(FLM), electron microscopy (EM), and electron tomography (ET) provide the most
powerful tools to study cellular structure and function. All approaches hold positive
and negative aspects. FLM has the advantage to be performed on both living and fixed
cells, and can provide results on a large set of data. However, the relatively poor resolu-
tion of standard far-field fluorescence microscopes reduces the possibility to precisely
identify the nature of a given fluorescent signal. For example, a fluorescent spot might
represent an entire organelle, a subdomain of an organelle, or even aggregates of pro-
teins. In addition, the relationship between the fluorescent spot and the surroundings
6. 3D HDO-CLEM: Cellular Compartment Analysis by Correlative Light-Electron Microscopy on Cryosection 97
cellular unlabeled structures is missing. On the other hand, immuno-electron micros-

copy (immuno-EM) allows the localization of proteins with a resolution of tens of
nanometers and an exquisite depiction of subcellular structures of all compartments,
labeled and unlabeled. However, EM methods are static in nature and cannot track
highly dynamic processes such as membrane trafficking events. Moreover, the local-
ization of certain rare cellular structures within the enormous volume of a cell, or group
of cells, may be difficult by EM and may limit the analysis to a small set of data.
Although comparatively new FLM techniques, usually referred as super resolution
light microscopy approaches (Hell, 2007; Watanabe et al., 2011), have almost closed
the resolution gap between FLM and EM, the correlation of the information obtained by
EM and FLM can answer many important open questions in cell biology, mostly linked
to the definition of the relationships between labeled and unlabeled compartments.
Thus, ideally it would be very useful to combine the strengths of both FLM and EM,
while observing the very same sample by both techniques. This combined approach is
referred to as the correlative light-electron microscopy (CLEM). Many EM laborato-
ries have spent considerable efforts in the development of easy, robust, and quantitative
methods to perform CLEM. Each different approach focuses on different aspects of
CLEM: the retracing of the regions of interest (ROIs), identified first by FLM and suc-
cessively retraced by EM (Kukulski et al., 2011; Sartori et al., 2007); the development
of new correlative probes, which can be visualized by both FLM and EM (Ellinger,
Messlitzer-Ruppitsch, Rohrl, Neumuller, & Pavelka, 2009; Gaietta et al., 2002; Gra-
benbauer et al., 2005; Kukulski et al., 2011; Takizawa, Suzuki, & Robinson, 1998);
improvement of the live-cell imaging, or of fixation conditions (Brown & Verkade,
2010; van Rijnsoever, Oorschot, & Klumperman, 2008; Kukulski et al., 2011; Shu
et al., 2011). The choice of a specific CLEM method ultimately depends on the given
scientific question to answer.
The HDO-CLEM method (Vicidomini et al., 2008, 2010) has been designed with
the aims such as (1) maximizing labeling efficiency, (2) performing analysis on large
data sets, (3) obtaining good preservation of morphology, and (4) allowing 2D and 3D
analyses. The approach is based on the use of ultra/semithin Tokuyasu cryosections
(Tokuyasu, 1980), and integrates immunogold and immunofluorescence labeling.
The Tokuyasu method is based on sucrose embedding of aldehyde-fixed samples
and it allows a reasonable balance between morphology and antigenicity retention. The
use of sucrose prevents ice crystal formation during the freezing procedure of alde-
hyde-fixed samples (Tokuyasu, 1980). In noncrystalline ice (vitreous ice), the freez-
ing procedure has minimal impact on the ultrastructure and allows ultrathin sectioning
of the sample. With the introduction of the Tokuyasu cryosectioning method and the
subsequent improvements (Slot & Geuze, 2007; van Donselaar, Posthuma, Zeuschner,
Humbel, & slot, 2007), immuno-EM has become widely accessible to many EM labo-
ratories and has set the standard for antigen localization at the nanometer resolution.
The labeling procedures adopted in the HDO-CLEM method follows a stepwise
approach. The antigens of interest are first labeled with a specific primary antibody
and then by a specific secondary fluorescent antibody and protein A-gold (PAG). The
procedure is suitable for both single and double labeling approaches. The HDO-CLEM
protocol has been optimized to correlate the highest amount of different cells struc-
tures, or events, in a single experiment. The possibility to apply the method to semithin
cryosections, or ribbons of serial cryosections, makes the method suitable for 3D stud-
ies, by both EM tomography and FLM, respectively. FLM 3D relies on a free soft-
ware, MicroScoBioJ (Vicidomini et al., 2008, 2010), and a collection of ImageJ-based
plugins (Abramoff, Magalhaes, & Ram, 2004), able to create 3D multicolor volume
rendering from serial 2D images.
A high content-data analysis procedure is also available, to provide large scale
FLM morphometric data sets, with an accuracy close to the EM. The method involves
the correction of the FLM data by a systematic error obtained from the FLM/EM
correlation of a significant cohort of samples.
Here, we report a step-by-step description of the HDO-CLEM method we have
developed, as we have applied it to correlate FLM and EM analyses of Russell bodies
(Mattioli et al., 2006; Russell, 1890).
Russell bodies are dilated endoplasmic reticulum (ER) cisternae, generally sepa-
rated from the normal ER network, containing condensed aberrant immunoglobulins
(Ig) originally identified in plasmocitoma cells. We used HeLa cells stably transfected
with a mouse immunoglobulin µ-chain, lacking the first constant domain (µ-ΔCH1),
together with a wild-type immunoglobulin λ-light chain. In this case, immunoglobu-
lin cannot reach the Golgi, and the cells display spherical dilated rough ER cister-
nae resembling the Russell bodies identified in plasmocitoma cells (rough Russell
bodies, RRBs). RRBs are about 1 µm in diameter and contain protein aggregates of
mutant immunoglobulin M. Conversely, when µ-ΔCH1 is transfected in the absence
of λ, it aggregates in the smooth ERGIC tubular compartments (smooth Russell bod-
ies, SRBs), as shown by the association with the ERGIC marker ERGIC-53. SRBs are
composed of narrow (100 nm wide, n-SRB) or dilated (200 nm wide, d-SRB), curled
tubular cisternae (Mattioli et al., 2006).
II. Materials
A. Reagents
1. 0.1M phosphate buffered saline (PBS) pH 7.4
2. 4% Paraformaldehyde
3. 70% Glutaraldehyde
4. 4% Paraformaldehyde/0.4% glutaraldehyde
5. 0.15% Glycine
6. 2.3M sucrose
7. 2% Methylcellulose (25 cP)
8. 12% Gelatin
9. 10% Sodium azide solution
10. 10% Bovine serum albumin (BSA) stock solution
11. 4% Uranyl acetate
1 2. 0.3M oxalic acid

13. 2% Uranyl acetate/0.15M oxalic acid
14. 25% Ammonium hydroxide
15. 0.4% Uranyl acetate/1.8% 25 ctp methylcellulose
16. 50% Glycerol in water
17. Formvar powder
18. 40,6-Diamidino-2-phenylindole (DAPI) (1 mg/ml stock solution, use 100 ng/ml
working solution)
B. Antibodies
1. Rabbit anti-IgG (µ-chain) (Zymed) primary antibodies (diluted in 1:200 in 1%
BSA in PBS)
2. Rabbit anti-calreticulin (Stressgen) primary antibodies (diluted in 1:200 in 1%
BSA in PBS)
3. Cy3-conjugated donkey anti-rabbit (Zymed) secondary antibody (diluted in 1:200
in 1% BSA in PBS)
4. Cy2-conjugated donkey anti-rabbit (Zymed) secondary antibody (diluted in 1:200
in 1% BSA in PBS)
5. 10 and15 nm PAG (Utrecht University, Utrecht, Netherlands)
C. Sectioning and Imaging Equipments

1. T5293, DUMONT biology tweezers nr. 5 stainless.
2. Aluminum specimen pins for ultra-cryo-microtomy (local manufacturer).
3. Diamond knife, cryo, dry, 3 mm – 45°, 1 mm/sec cutting speed (Drukker or
Diatome).
4. Leica EM FCS Ultracut cryo-ultramicrotome (Leica Microsystems, Wetzlar,
Germany).
5. Finder gold EM grids (Agar).
6. Microscope square coverslips (No. 1, 0.13–0.16 mm thick and 20 mm × 20 mm
size) (see Note 1).
7. Microscope slides (1.0–1.2 mm thick, 25.4 × 76.2 mm size).
8. A spectral laser scanning microscope equipped with an HeNe laser (1mW, 543 nm),
a multiline Ar laser (100 mW, 457-476-488-514 nm) and a near ultraviolet laser
diode (405 nm), such as a Olympus FV1000, with a transmitted light detector for
recording brightfield images. A PlanApo 60×/1.42 oil objective (Olympus).
9. Transmission electron microscope capable of recording tomographic tilt series,
such as a Tecnai 12-G2 Bio-Twin EM (FEI Company). This microscope is
equipped with a LaB6 emitter for high brightness, a CompuStage for accurate
movement of the specimen on five axes (X, Y, Z, α, β), a SIS Mega View III CCD
camera (Olympus-SIS) for digital image acquisition, and Kodak film 4489 for
recording some EM micrographs.
D. Software
1. ImageJ (http://rsbweb.nih.gov/ij/) free software package (Abramoff et al., 2004)
(see Note 2)
2. MicroScoBioJ (http://imagejdocu.tudor.lu/doku.php?id=plugin:stacks:microscob
ioj:start) free software package (Vicidomini, et al., 2008 , 2010) (see Note 3)
3. TurboReg (http://bigwww.epfl.ch/thevenaz/turboreg/) free software package (see
Note 4)
4. MosaicJ (http://bigwww.epfl.ch/thevenaz/mosaicj/) free software package (see
Note 5)
5. IMOD (http://bio3d.colorado.edu/imod/) free software package (Kremer, Mastro-
narde, & McIntosh, 1996) (see Note 6)
6. iTEM platform (Olympus-SIS) equipped with the iTEM Solution Tomography
(Olympus-SIS) extension
III. Methods
Outline of the methods are (1) preparation of reagents, (2) preparation of semi/
ultrathin cryosections of cell samples by Tokuyasu technique, (3) immunofluorescence
and immunogold labeling of semi/ultrathin cryosections, (4) image acquisition and
processing (ImageJ and MicroScoBioJ software), and (5) hybrid morphometry.
A. Preparing Reagents
1. 0.1M PBS: Weigh 80g NaCl, 2g KCl, 14.4g Na2HPO4˙2H2O, and 2.3g
NaH2PO4˙H2O, and add distilled water (Milli Q) up to 1000 ml. Dilute 1:10
(v:v) with distilled water prior to use.
2. 4% Paraformaldehyde: Dissolve 4g of granular paraformaldehyde in 100 ml of
PBS. This solution must be prepared under the chemical fume hood. Preheat the
PBS to 60°C to allow the paraformaldehyde to completely dissolve, and then add
50 µl of 1N NaOH while stirring.
3. 4% Paraformaldehyde/0.4% glutaraldehyde: Mix 70% glutaraldehyde and 4%
paraformaldehyde in a 1:175 (v:v) ratio.
4. 0.15% Glycine: To 30 mg of glycine, add PBS up to a 20 ml final volume.
5. 2.3M sucrose: Dissolve 78.73g of sucrose in 50 ml of PBS and stir until the
sucrose is completely dissolved. Add PBS to a 100 ml final volume and store
50 ml aliquots at 4°C.
6. 2% Methylcellulose: Add 4g of methylcellulose to 196 ml of distilled water while
stirring at 90°C. Rapidly cool to 10°C on ice while stirring, and then add water
to a final volume of 200 ml. Seal with parafilm and stir overnight at 4°C. Store
methylcellulose at 4°C in the dark, for up to 3 months.
7. 12% Gelatin: Dissolve 12g of gelatin powder (local food brand is satisfactory) in
75 ml of PBS by first stirring the solution for 10 min at room temperature and
then for 4–6 h at 60°C. When the gelatin is completely dissolved, cool the solu-
tion to 37°C, add 200 µl of a 10% sodium azide solution, and finally add PBS to
a 100 ml final volume. Aliquot and store at 4°C, up to several months.
8. 10% BSA stock solution: Add 10g of serum albumin in 100 ml of distilled water
(Milli Q) and gently stir overnight, to prevent foaming, at 4°C. Set the pH to
7.4 with 1N NaOH, and then add sodium azide to a 0.02% final concentration.
Centrifuge the solution for 1 h at 100,000 × g, retrieve the supernatant, and store
in small aliquots at 4°C.
9. 4% Uranyl acetate: Dissolve 4g of uranyl acetate in 100 ml of distilled water by
stirring at room temperature.
10. 0.3M oxalic acid: Dissolve 3.78g of oxalic acid in 100 ml of distilled water by
stirring at room temperature.
11. 2% Uranyl acetate/0.15M oxalic acid: Mix 4% uranyl acetate and 0.3M oxalic
acid in a 1:1 (v:v) ratio. Adjust the pH to 7–8 by adding drops of 25% ammonium
hydroxide while stirring.
12. 0.4% Uranyl acetate/1.8% 25 ctp methylcellulose: Gently mix 1 ml of 4% uranyl
acetate to 9 ml of 2% methylcellulose solution. Store the solution at 4°C in the
dark, for up to 3 months.
13. 50% Glycerol in water: Add 100 ml of glycerol to 100 ml of deionized water
while stirring. Aliquot to 100 ml bottles and autoclave.
14. Formvar: Weigh 1.1g of formvar and place in a 100 ml volumetric flask. Add
chloroform to a 75 ml volume, and stir until dissolved. Subsequently, add more
chloroform to a 100 ml final volume and let rest overnight at room temperature
(see Note 7). Store formvar solution in a clean bottle for up to 3 months.
B. Preparation of Ultrathin Cryosections for HDO-CLEM

HeLa cells, treated to induce Russell bodies formation (Mattioli et al., 2006), are
plated on tissue culture dishes until they reach 80% confluence. Unless otherwise indi-
cated, all steps are performed at room temperature.
1. Fix cells by adding a mixture of 4% paraformaldehyde/0.4% glutaraldehyde to an
equal volume of culture medium for 2 h at room temperature (see Note 8).
2. Quench free aldehyde groups by incubating the cells for 10 min in 0.15%
glycine.
3. Add a small volume of 1% gelatin in PBS (see receipt 12% gelatin). Scrape off the
cells from the tissue culture dishes with a cell scraper.
4. Transfer the cells to a 1.5 ml microcentrifuge tube, pellet the cells for 1 min at
13,000 rpm and remove the supernatant.
5. Resuspend the cells with 1 ml of 12% gelatin for 10 min at 37°C.
6. Pellet the cells for 1 min at 13,000 rpm and carefully remove the supernatant
(gelatin 12%).
7. Resuspend the pellet in 1 ml of 12% gelatin.
8. Repeat the centrifugation step and remove the supernatant.
9. Transfer the pellet to 4°C for 15 min.
10. Carefully cut the microcentrifuge tube, at the level of the pellet, using a razor
blade and remove the solidified gelatin/cell pellet from the microcentrifuge tube.
11. Submerge the small blocks in 2.3M sucrose solution, and slowly rotate the tubes
overnight at 4°C (see Note 9).
12. Carefully retrieve the specimen blocks from the 2.3M sucrose solution, using
stainless biology tweezers nr. 5, and transfer to the top of the aluminum pins.
13. Position the specimen in the best orientation for sectioning (see Note 10).
14. Rapidly freeze pins by dipping in liquid nitrogen and store in liquid nitrogen (see
Note 11).
15. Cut semithin (200–300 nm) or ultrathin cryosections (60–65 nm) with a diamond
knife using the cryo-ultramicrotome. The temperature of sample, knife, and
chamber are set at −100°C for semithin sections, and −120°C for ultrathin cryo-
sections, respectively. The cutting speed is set at 1–2.5 mm/s (see Note 12).
16. Collect ribbons of flat cryosections on a droplet of a 1:1 mixture of 2.5M sucrose
and 2% methylcellulose and transfer to formvar film-coated, carbon stabilized,
gold finder EM grids.
17. Store the cryosections at 4°C, for up to 1 month (see Note 13).
C. Immunofluorescence and Immunogold Labeling for HDO-CLEM (see Note 14)

Every step of the labeling procedure is performed by floating the section-side grid
on drops of the different solutions placed on a strip of parafilm. Unless otherwise indi-
cated, all steps are performed at room temperature.
1. Immunolabeling (see Note 15)

1. Sequentially incubate grids with 2% gelatin in PBS for 20 min at 37°C, to remove
the methylcellulose coat, with 1% BSA in PBS for 10 min, to quench nonspecific
protein binding sites, and then with rabbit anti-calreticulin antibody for 30 min.
2. Rinse samples in 1% BSA in PBS.
3. Incubate for 45 min with the Cy3-conjugated donkey anti-rabbit antibody in the
presence of DAPI.
4. Rinse samples in 1% BSA in PBS.
5. Incubate for 20 min with PAG, in 1% BSA in PBS.
2. Double Labeling (see Fig. 1)

1. For double calreticulin/IgG µ-chain labeling, after the calreticulin labeling in the
previous section is complete (in this case do not use mixture with DAPI), fix the
sample for 5 min with 1% glutaraldehyde in PBS in order to prevent any cross-
reactivity of antibodies.
2. Incubate for 30 min with rabbit anti-IgG µ-chain.
3. Rinse the sample with 1% BSA in PBS.
4. Incubate with the Cy2-conjugated donkey anti-rabbit secondary antibody in the
presence of DAPI for 45 min.
5. Rinse the sample with 1% BSA in PBS.
Fig. 1 Double immunolabeling on ultrathin (60 nm) cryosections. (A) Immunofluorescence of calreticulin (Cy3,
red) and µ-ΔCH1 (Cy2, green). Nuclei labeled with DAPI (blue). To better appreciate colocalization, insets repre-
sent the two single channel corresponding to calreticulin (left) and µ-ΔCH1 (right) labeling in a HeLa cell express-
ing RRBs. (B) Low-magnification (2,200×) TEM image of the same ROI observed at confocal microscope.
(C) EM high-magnification (6,600×) image of the region pointed by the black arrow in (B). Immunogold labeling
with two different size gold nanoparticles confirms the colocalization of calreticulin (PAG-10 nm) and µ-ΔCH1
(PAG-15 nm) in RRBs (C). Scale bar: 2 µm. Figure adapted from Vicidomini et al. (2010). (See color plate.)
6. Incubate with PAG-15 nm for 20 min.

7. Fix with 1% glutaraldehyde in PBS for 2 min.
8. Lastly, wash in water.
3. Preparing Grids for FLM Observations

1. Coat formvar film-coated, carbon stabilized, gold finder EM grids with a 200 nm
thin layer of methylcellulose (see Note 16).
2. Place the gold finder grid on a 200 µl drop of 50% glycerol in water present on a
microscope coverslip (see Note 17).
3. Sandwich the grid between the coverslip and the microscope slide.
4. Sonicate the assembled microscope slide for 10 min in distilled water to remove
potential air bubbles in the glycerol medium (see Note 18).
4. Retrieval of Grids and EM Preparation

Once FLM imaging is completed, samples are processed for EM. The described
sample preparation is fully compatible with ET.
1. Carefully remove the grids from the glass slide.
2. Extensively wash grids in water to remove glycerol and methylcellulose.
3. To increase contrast for EM analysis, incubate cryosections with 2% uranyl
acetate/0.15M oxalic acid (200 nm semithin cryosections for 5 min, or 60 nm
ultrathin cryosections for 10 min), followed by incubation of all cryosections with
0.4% uranyl acetate/1.8% 25 ctp methylcellulose for 5 min.
D. HDO-CLEM
Cryosections of pelleted cells permit high sensitivity labeling and allow multimodal
(FLM/EM) imaging of an elevated number of investigated structures, cells, or events.
Moreover, different consecutive physical sections of the same structure or cell can be
retraced in the long ribbon and used for 3D analysis. A key point for retracing the ROIs
observed at FLM is to produce large field of view maps of the whole ribbon of cryosec-
tions at both the FLM and the EM levels.
1. FLM Imaging
1. Set the CLSM spectral imaging parameters as follows: excite DAPI using the
405 nm laser diode, and collect fluorescence in the 425–475 nm spectral region;
excite Cy3 using the 543 nm HeNe laser, and collect fluorescence in the 555–655 nm
spectral region; excite the Cy2 dye using the 488 nm Ar laser line, and collect
fluorescence in the 500–540 nm spectral region. Use the 543 nm HeNe laser to
acquire a bright field image. Set sequential individual channel scans to prevent
spectral crosstalk.
2. Collect multichannel (DAPI/Cy3/brightfield or DAPI/Cy2/Cy3/brightfield)
partially overlapping images of the whole ribbon of serial cryosections at the
maximum field of view of the FLM system.
3. Use the set of images to produce a mosaic of the entire ribbon (see Fig. 2(A)). In
particular, use the MosaicJ plugin (Thevenaz & Unser, 2007) to obtain a high-
quality mosaic. The plugin requires a set of overlapping images of the entire rib-
bon, along with their user-provided coarse mosaic. The solution then refines this
initial mosaic in a fully automatic fashion.
4. Identify in the mosaic the ROI to successively retrace for the EM imaging. The
ROI can be a single cell, structure, or event (see Fig. 2(A)), or can be a consecutive
series of physical sections of the same cell or structure (serial ROI) (see Fig. 2(A)).
The patterns of the nuclei help to identify the physical sections of the same cell,
and thereby the same ROI in different cryosections.
5. The ROIs previously identified in the mosaic may be further imaged using differ-
ent settings. If the pixel sampling used during the imaging for the ROI identifica-
tion does not satisfy the Nyquist criterion (see Note 19), the image obtained does
not preserve the finer details of the structures. As a consequence, any process (e.g.,
registration and segmentation) performed on this image will be compromised.
With the help of the ribbon mosaic, retrace all the ROIs and acquire them using a
pixel size of 50–25 nm (see Note 19) (see Fig. 2(B–D)).
2. FLM Image Processing: Registering Serial ROIs

Use the TurboReg plugin to align and register the series of multichannel (DAPI/Cy3
or DAPI/Cy2/Cy3) images representing the consecutive physical sections of a cell or
structure.
1. Registration is performed using a consecutive pair of images (a reference image
and a source image): register the second image (source) to the first image (refer-
ence) and use the result as the new reference image for the third image, and so on
(see Fig. 2(A–C)).
2. Use the nuclei patterns to specify the landmarks for the registration of the respective
channel and save the landmarks.
Fig. 2 FLM imaging. Immunofluorescence labeling for µ-ΔCH1 (Cy3) on SRB-expressing HeLa cells. (A) High-resolution confo-
cal map of three consecutive serial cryosections. Automatic MosaicJ montage of 16 multichannel images (DAPI-Cy3-Bright field).
Bright field channel the grid mesh bars and letters. Images acquired with the maximum image size allowed by the CLSM system
(2048 × 2048 pixels, 100 nm pixel size). Insets (ROI1, ROI2 Sect. 1, ROI2 Sect. 2, and ROI2 Sect. 3) are direct digital enlargements
of the corresponding areas boxed in (A). ROI2 Sect. 1–3 show the same ROI identified on three consecutive serial sections. (B–D)
High-quality image acquisition of ROI2 Sect. 1–3 (50 nm pixel size). Scale bars: 100 µm (A) and 5 µm (B). Figure adapted from
Vicidomini et al. (2008.) (See color plate.)
3. Successively, load the landmarks and apply the same transformation that is applied
to the nuclei channel to the remaining channels. Use the “rigid body rotation” or
“affine” registration type (Thevenaz, Ruttimann, & Unser, 1998) and the auto-
matic mode of TurboReg. Occasional dramatic differences in the nuclei pattern
occur between consecutive cryosections. In this case, disengage the automatic
refinement procedure and use the manual mode of registration.
3. FLM Image Processing: 3D Surface Rendering

Surface rendering is a method used to display a 2D projection of an isosurface rep-
resenting a 3D data set. The first step for surface rendering is the extraction of the
surface (isosurface) from the 3D data set at the boundary between the background and
the structure to display. The isosurface is usually approximated by a set of polygo-
nal meshes (in our case triangles). The Mesh Maker plugin implements the marching
cube algorithm, one of the most common algorithms used to extract isosurfaces from a
3D data set (William & Harvey, 1987). The algorithm requires a value (isovalue) that
specifies if a pixel in the image belongs to the background or to the structure to display
(see Fig. 3(D and E)).
Fig. 3 FLM image processing: registering serial ROIs. Immunofluorescence labeling for µ-ΔCH1 (Cy3) SRB-expressing HeLa cells;
nuclei labeled with DAPI (blue). (A,B) ROIs identified on consecutive 200 nm serial cryosections. Images are selected from Fig. 1.
Warped images obtained by affine transformation using defined landmarks. (C) Registration of (A, target image) on (B, source image)
using chromosomes (blue channel, DAPI), followed by registration of SRBs (red channel). (D) Surface rendering of the final registered
3D stack from the three consecutive serial sections displayed in Fig. 1 by TurboReg iteration of the process used to obtain (C). (E)Higher
magnification of details boxed in (D) The white dot in (E) locate an area in the 3D rendering, previously identified in (C). Scale bar:
5 µm. Figure adapted from Vicidomini et al. (2008). (See color plate.)
1. Choose the isovalue for the channel representing the structure to visualize (e.g.,
nuclei or Russell bodies). The user-friendly interface of the Mesh Maker plugin
facilitates the choice of the value.
2. Introduce the sampling interval (voxel size) of the 3D data set (see Note 20).
3. Save the file containing the isosurface of the structure to visualize (file format
[.off]).
4. Repeat the same procedure for the other channels of the 3D data set.
5. Use the Mesh Viewer plugin to display the isosurfaces. It can display up to four
different isosurfaces that can be viewed simultaneously. The plugin provides dif-
ferent navigation functions such as zooming, translation, and rotation.
4. EM Imaging
Once FLM imaging is completed, samples are processed for EM (see Section 3.3.4).
1. Use the “multiple image alignment” (MIA) function of the iTEM software to gen-
erate the EM high-resolution mosaic map of the whole cryosection ribbon. A more
time-consuming manual acquisition can also be performed.
2. Use the EM and FLM mosaic to manually track the ROIs within the cryosection.
Number and letter landmarks on the grid mesh, and the nuclei pattern, allow an
easy recognition of the ROIs. The number and letter landmarks present within the
grid mesh are visualized in the brightfield channel of the FLM mosaic. EM digital
image or EM micrographs of the ROIs are recorded (see Fig. 4(A and B)).
3. Record the tomographic tilt series of the selected ROIs.
5. EM image Processing: Electron Tomography

1. Generate tomograms of the selected ROIs using the IMOD software package
(Kremer et al., 1996) or the iTEM solution tomography. Use PAGs (see Sections
3.3.1 and 3.3.2) as fiducial markers for the alignment of the tomographic tilt series
(see Fig. 4(D)).
6. CLEM image Processing

A CLEM overlay image is obtained by rescaling and registering the FLM image to
the EM images (see Fig. 4(C)).
1. Rescale the FLM image using basic function of ImageJ (Image → Scale). The scale
factor is given by the ratio of EM image pixel size and FLM image pixel size.
2. Register the FLM image and the EM image by using manual mode and “rigid-
body rotation” as a function of the TurboReg plugin. FLM and EM images are set
as source and reference, respectively.
3. Superimpose the grayscale EM image and the color-coded FLM image using the
basic function of ImageJ (Plugins → Color Functions → Color Composite Merge).
E. Hybrid Morphometry: A Systematic Error-Based Approach

EM provides detailed information about cellular ultrastructure, membrane shapes,
and organelle architecture. However, certain cellular structures are rare and, there-
Fig. 4 EM imaging. (A) Immunofluorescence labeling for µ-ΔCH1 (Cy3) on d-SRB-expressing HeLa cells (same as in Figs 1 and 2);
nuclei labeled with DAPI (blue). Solid line insets display higher magnification/resolution views of areas in dashed boxes. (B) EM images
of (A). Black line inset shows higher magnification of areas encircled in the white line boxes and highlight portion of a d-SRB. (C) CLEM
overlay of the confocal image on the corresponding EM images. (D) TE analysis of the area displayed in the insets in (B). Virtual sec-
tions of the tomogram is set as background image. ETM obtained from a −45° to a +45° tilt series, with 1° increments. Voxel sizes
22.7 × 22.7 × 7 nm (240 × 159 × 28 voxels) (original tomograph was interpolated to obtain a better 3D model). Scale bars: 50 µm (A),
2.5 µm (B and C). Figure adapted from Vicidomini et al. (2008). (See color plate.)
fore, difficult to locate by EM within the enormous volume of a cell or group of cells.
Even if such structures can be located occasionally, a meaningful quantitative analysis
requires observation of multiple copies. FLM is ideal for the localization of rare and
dispersed structures within the cell, but it suffers from poor spatial resolution and,
therefore, it lacks fine structural information. These complementary advantages and
disadvantages set the motivation for hybrid morphometry CLEM approaches combin-
ing large data sets (FLM) and detailed morphometry information (EM). In summary,
statistical morphometric features are inferred from a large FLM data set and, subse-
quently, a small subset of the data is correlated with EM data to evaluate the quality
of the FLM analysis. Because of the higher spatial resolution of EM with respect to
FLM, when a morphometric property of the very same object is measured by the two
modalities, the EM value can be considered as the “real” one, as it is essentially unaf-
fected by error introduced by poor spatial resolution. Broadly speaking, comparing the
FLM and the EM morphometric values measured on a cohort of correlated structures
allows an estimate of the “systematic error” introduced by the FLM. Such a system-
atic error can be used to better evaluate and weigh the results obtained by the FLM
morphometric analysis performed on a statistical relevant (large) data set. Here, we
describe the hybrid morphometric analysis step-by-step to study the surface area of
smooth and RRBs.
1. FLM Image Analysis: Surface-Area Estimation

FLM imaging misses the unlabeled surrounding cellular structures. In most
cases, this is an undesirable property, but becomes an important characteristic used
for automatic surface-area estimation. The high background to structure contrast
of the FLM image helps segmentation algorithms to separate the labeled structures
from the surrounding background. In summary, the segmentation algorithm returns
a binary image, in which “zero” represent the background and “one,” the object or
structure. Many different segmentation algorithms have been applied and designed
for FLM imaging (Dima et al., 2011). Here we use a hybrid algorithm that combines
a classical single-threshold approach and a dynamic-threshold approach (see Note 21).
We called this method weight-adaptive-threshold (WAT) and implemented it in the
MicroScoBioJ software package. However, there is no universally applicable seg-
mentation method that will work for all images, and no segmentation technique is
perfect (Dima et al., 2011). Thereby, before starting the high-content analysis, we
highly recommend to test all the available segmentation algorithms provided by
ImageJ.
1. Apply a Gaussian blur filter (Process → Filter → Gaussian Blur) on the raw FLM
images in order to reduce the noise (see Note 22).
2. Apply the segmentation algorithm.
3. Use the binary images (1: structure, 0: background) obtained from the segmenta-
tion algorithm to measure the surface area (Analyze → Analyze Particles) for each
structure of interest (SOI). At the end of these steps, every structure of interest
{ SOIi : i ∈ U } is characterized by a value representing an estimation of the surface
area, SAFLM(SOIi).
2. EM Image Analysis: Surface-Area Estimation

1. Estimate for a small subset of the structures of interest { SOIj : j ∈ S ⊂ U } the sur-
face area using the correlated EM images. Use manual edge-detection (membrane
contour) on EM images to measure the surface area of the structures of interest,
SAEM(SOIj). Manual segmentation is a time-consuming procedure and, therefore,
unrealistic when dealing with thousands of structures.
3. Calculating Systematic Error

1. Average the FLM measurements, SA
FLM 
 FLM
SA (SOIi ) / N , where N = #U
denotes the number of structures analyzed. The result represents an estimate of
the surface area of the investigated structure.
2. Even if the estimation of the surface-area system is obtained on a large data
set one has to keep in mind that FLM has poor spatial resolution compared to
EM. To understand the accuracy of the surface-area estimate SA compute
the average of the relative errors assuming the EM measures the true values:
E FLM EM EM
((SA (SOI j )  SA (SOI j )) / SA (SOI j )  M )), where M = #S denotes
the number of structures correlated. Generally, the value E can be considered as an
index of the systematic error introduced using FLM for the morphometric analysis.
IV. Notes
1. Each objective lens is designed and assembled to achieve optimum performance

with specific microscope coverslips. Using a coverslip with wrong thickness can
introduce spherical aberrations during imaging and thereby degrade the resolu-
tion. Consult the manufacturer of your objective lens about which coverslip to use.
2. ImageJ is a public domain Java image processing program. It can display, edit, ana-
lyze, process, save, and print many different types of microscopy images. ImageJ
was designed with an open architecture that provides extensibility via Java plugins.
3. MicroScoBioJ is a collection of three plugins designed to work in the ImageJ soft-
ware package: (1) Mesh Maker computes triangle meshes from a 3D fluorescence
microscope data set; (2) Mesh Viewer is a 3D viewer able to visualize the triangle
meshes created by the Mesh Maker; and (3) WAT, performs a segmentation algo-
rithm on a fluorescence microscope image. The Mesh Viewer plugin requires the
installation of the Java3D library, release 1.4.0_01 (http://java3d.java.net/binary-
builds-old.html).
4. TurboReg is written as a plugin for ImageJ. The purpose of this plugin is to reg-
ister, that is, to align or to match two images, one of them being called the source
image and the other the target image. Three alignment modes are available: man-
ual, automatic, and batch. In all three cases, the user is given the opportunity
to interactively specify some landmarks, which establishes the initial correspon-
dence between the images. In the automatic and batch modes, the landmarks of
the source image are then automatically refined to better match those of the target
image. In the manual mode, this automatic refinement procedure is disengaged.
5. MosaicJ is a plugin designed to work in the ImageJ software package. The

purpose of this plugin is to facilitate the assembly of a mosaic of overlapping
individual images. It provides a semiautomated solution where the initial rough
positioning of the images must be performed by the user, and where the final
delicate adjustments are performed by the plugin. The MosaicJ plugin requires
that the plugin TurboReg is installed.
6. IMOD is a set of image processing, modeling, and display programs used for
tomographic reconstruction and for 3D reconstruction of EM serial sections and
optical sections. The package contains tools for assembling and aligning data
within multiple types and sizes of image stacks, viewing 3D data from any ori-
entation, and modeling and display of the image files.
7. Chloroform is highly volatile, and, therefore, the formvar solution must always
be kept within sealed containers.
8. Aldehyde-based fixative can modify the antigen configuration and reduce the
antibody binding capabilities. Therefore, labeling efficiency in immuno-EM
can vary depending on the fixation protocol used relative to the specific anti-
gen studied and antibody used. Thus, optimization of fixation conditions for
a given antibody in a non-CLEM setup is highly recommended before start-
ing CLEM studies. An overview of fixatives suitable for cryo-immunolabel-
ing has been described (Slot & Geuze, 2007). In our experience, for CLEM
studies, it is advisable to use low concentrations of glutaraldehyde in the
fixative (from 0.025% to 0.2%). These concentrations often result in opti-
mal ultrastructural preservation and limited autofluorescence, allowing easier
confocal imaging of flat ribbons of semithin cryosections, and good antigen
preservation.
9. Each block should be small enough to be accommodated on the tip of an alumi-
num pin for cryo-ultramicrotomy (about 1–2 mm3).
10. Ice crystal formation during freezing of the specimens will damage the ultra-
structure of the cell and make sectioning difficult, since good cryosections can
be cut only from vitreous frozen blocks.
11. The mounting procedure should be as rapid as possible. Water evaporating from
the specimen during air exposure results in an increase in sucrose concentration
so that sucrose crystals may form. It is advisable to perform this transfer step
with a dissection microscope in a cold room. If a suitable cold room is unavail-
able, use the dissection microscope at room temperature and mount the blocks
rapidly using a box filled with ice.
12. A properly trimmed block face is absolutely crucial to obtain long ribbons of
cryosections. Use automatic cutting cycle of the ultramicrotome to produce rela-
tively thick (~100 nm) sections and allow the machine to trim into the block
until a desired facet is produced (usually 100 cycle). Repeat the process for all
the facets in order to trim the sample to a square or rectangle, with all four sides
being straight and even.
13. Use cryovials for storage, with the caps punched to make a hole to prevent high
internal pressure. Cryovials are then stored in cryoboxes in liquid nitrogen.
14. HDO-CLEM analysis can also be applied to cell expressing green fluorescent
protein (GFP-tagged proteins). In this immunofluorescence labeling is not nec-
essary (Fig. 5)
15. Use of the rabbit anti-calreticulin antibody labels the ER, and use of the rabbit
anti-IgG (µ-chain) labels the two forms of Russell bodies (i.e., RRB and SRB).
16. The methylcellulose coat plays different roles. It protects the section during
FLM imaging, ensuring a high rate of success for the subsequent EM imaging
session, and protects the formvar film coating and cryosections during the disas-
sembling of the grid from the temporary FLM slide.
17. Gold grids are preferred to other materials to avoid changes in the spectral prop-
erties of DAPI.
18. During FLM imaging, the section side of the EM grid is facing toward the objec-
tive. The glycerol/water layer between the cryosections and the coverslip ensures
optimal imaging conditions, avoiding potential spherical aberration induced by a
mismatch of refractive index.
Fig. 5 HDO-CLEM approach applied to recombinantly expressed GFP-Rab5. 200 nm cryosections of HeLa cells transiently trans-
fected with Rab5-GFP (kindly provided by Giorgio Scita, IFOM, Milan). (A–C) Rab5-GFP localizes in discrete compartments (green)
of a transfected cell. Images refer to the same ROI in three consecutive serial sections. Nuclei are stained by DAPI (blue). (D) 3D
surface rendering obtained. (E) Low-magnification EM map of the mesh area containing the ROI displayed in (A). (F) CLEM overlay
of the area boxed in (E), and the CLSM image shown in (A). (G) Enlarged EM image of the area boxed in (F). Immunogold labeling,
using rabbit antibodies to GFP (Molecular Probe) and PAG (10 nm), identifies Rab5-GFP associated to the surrounding membranes.
Inset represents a high-magnification TEM image of the boxed ROI. The GFP-positive compartment is morphologically identifiable as
an endosome. (H) 3D model of the boxed area in (G). TE was obtained from a −45° to +45° tilt series, with 1° increments. Tilt series
were used to obtain tomograms of 1024 × 1024 × 31 voxels with 1.5 × 1.5 × 1.5 nm voxel size. Computer-assisted tracing of the membrane
contours on the tomogram was used to generate the 3D model of the endosome in (G). Moreover, to validate the possibility of EM to
identify also unlabeled structures, we generated also the 3D model of a mitochondria (blue), located close to the endosome. Scale bars:
5 µm (A–C), 50 µm (E), 5 µm (F), 1 µm (G). Figure adapted from Vicidomini et al. (2008). (See color plate.)
19. Pixel size (or sampling distance) for FLM imaging has to be chosen according
to the Nyquist sampling criterion. This criterion determines the minimal sam-
pling distance needed to capture all the spatial information from the microscope
into the image. When the sampling distance is larger than the Nyquist distance,
spatial information about the image is lost. As a rule of thumb, the Nyquist
distance equals half of the spatial resolution of the system. For our confocal
system configuration a spatial resolution of ~200 nm is expected, thereby a
pixel size of 100–80 nm is recommended in the normal case. Using a smaller
pixel size, no additional spatial information is afforded, and the image is said
to have been oversampled. The extra pixels do not theoretically contribute to
the spatial resolution, but can often help improve the accuracy of feature mea-
surements taken from the image. To ensure adequate sampling for successive
image processing, a pixel size equal to a third of the expected spatial resolution
is suggested.
20. The voxel size of the FLM 3D data set is given laterally by the pixel size of the
FLM image and axially by the physical section of the cryosection.
21. The algorithm we used is based on the concept of threshold. A threshold value is
computed for each pixel of an object image. If the pixel intensity is higher than
the threshold, the pixel belongs to the object. The IsoData (Ridler & Calvard,
1978) algorithm identifies a global threshold for all pixels of an image, using
statistical properties of the image histogram. Unfortunately, this algorithm is
not able to accommodate changes in the distribution of different fluorescence
concentrations in a structure. Chow and Kaneko (Chow & Kaneko, 1972) solved
this problem but dynamically changing the threshold over the image: each local
pixel threshold is obtained by statistical investigation of the intensity values
of the local neighborhoods of such pixel. The statistic that is most appropriate
depends largely on the input image: the more complex is the statistic, the better
the results. For this work we combined the IsoData threshold (ThIsoD), with the
dynamic threshold (Thd), to find a weight-adaptive-threshold Thwa(x,y) = wIsoD-
ThIsoD + wIsoDThd(x,y) + wbaseThbase with the sum of the weights w equals to one.
Thbase can be used to increase or decrease the final threshold on the base of prior
information. Moreover, further information, such as background, can be directly
added to the algorithm.
22. The sigma value for the Gaussian filter has to be chosen according to the imaging
conditions. In particular, the sigma value has to be in the order of the resolution
of the FLM system. In such a regime, only the frequencies dominated by noise
are eliminated in the image, while frequencies containing information about
the structure are preserved. Assuming a pixel size of 50 nm and a resolution of
~200 nm, a sigma of 4 pixel is suggested. Alternatively, it is possible to use
a deconvolution algorithm (Bertero, Boccacci, Desidera, & Vicidomini, 2009)
instead of the Gaussian filter. These algorithms simultaneously reduce noise,
blurring and enhancing the contrast of the images. It has been demonstrated that
segmentation and edge-detection algorithms applied on deconvolved images
leads to superior results (Vicidomini, Boccacci, Diaspro, & Bertero, 2009).
Acknowledgments
The authors thank all members of Centro di Ricerca MicroScoBio and of Italian
Institute of Technology, Department of Nanophysics, for support and discussions.
References
Abramoff, M. D., Magalhaes, P. J., & Ram, S. J. (2004). Image processing with ImageJ. Biophotonics Inter-
national, 11(7), 36–42.
Bertero, M., Boccacci, P., Desidera, G., & Vicidomini, G. (2009). Image deblurring with Poisson data: From
cells to galaxies. Inverse Problems, 25(12).
244(1–4), 91–97.
Chow, C. K., & Kaneko, T. (1972). Automatic boundary detection of the left ventricle from cineangiograms.
Computers and Biomedical Research, 5(4), 388–410.
Dima, A. A., Elliott, J. T., Filliben, J. J., Halter, M., Peskin, A., Bernal, J., et al. (2011). Comparison of seg-
mentation algorithms for fluorescence microscopy images of cells. Cytometry A, 79(7), 545–559.
Ellinger, A., Meisslitzer-Ruppitsch, C., Rohrl, C., Neumuller, J., & Pavelka, M. (2009). Photooxidation tech-
nology for correlated light and electron microscopy. Journal of Microscopy Oxford, 235(3), 322–335.
Gaietta, G., Deerinck, T. J., Adams, S. R., Bouwer, J., Tour, O., Laird, D. W., et al. (2002). Multicolor and
electron microscopic imaging of connexin trafficking. Science, 296(5567), 503–507.
Correlative microscopy and electron tomography of GFP through photooxidation. Nature Methods,
2(11), 857–862.
Hell, S. W. (2007). Far-field optical nanoscopy. Science, 316(5828), 1153–1158.
image data using IMOD. Journal of Structural Biology, 116(1), 71–76.
Kukulski, W., Schorb, M., Welsch, S., Picco, A., Kaksonen, M., & Briggs, J. A. (2011). Correlated fluores-
cence and 3D electron microscopy with high sensitivity and spatial precision. Journal of Cell Biology,
192(1), 111–119.
Mattioli, L., Anelli, T., Fagioli, C., Tacchetti, C., Sitia, R., & Valetti, C. (2006). ER storage diseases: A role
for ERGIC-53 in controlling the formation and shape of Russell bodies. Journal of Cell Science, 119
(Pt 12), 2532–2541.
Ridler, T., & Calvard, S. (1978). Picture thresholding using an iterative selection method. IEEE Transaction
on Systems Man and Cybernetics, 8, 629–632.
Russell, W. (1890). An address on a characteristic organism of cancer. British Medical Journal, 2(1563),
1356–1360.
Sartori, A., Gatz, R., Beck, F., Rigort, A., Baumeister, W., & Plitzko, J. M. (2007). Correlative microscopy:
Bridging the gap between fluorescence light microscopy and cryo-electron tomography. Journal of Struc-
tural Biology, 160(2), 135–145.
encoded tag for correlated light and electron microscopy of intact cells, tissues, and organisms. PLoS
Biology, 9(4), e1001041.
Slot, J. W., & Geuze, H. J. (2007). Cryosectioning and immunolabeling. Nature Protocols, 2(10), 2480–2491.
thin cryosections. Proof of principle. Journal of Histochemistry and Cytochemistry, 46(10), 1097–1102.
Thevenaz, P., & Unser, M. (2007). User-friendly semiautomated assembly of accurate image mosaics in
microscopy. Microscopy Research and Technique, 70(2), 135–146.
Thevenaz, P., Ruttimann, U. E., & Unser, M. (1998). A pyramid approach to subpixel registration based on
intensity. IEEE Transaction on Image Processing, 7(1), 27–41.
Tokuyasu, K. T. (1980). Immunochemistry on ultrathin frozen sections. The Histochemical Journal, 12(4),
381–403.
van Donselaar, E., Posthuma, G., Zeuschner, D., Humbel, B. M., & Slot, J. W. (2007). Immunogold labeling
of cryosections from high-pressure frozen cells. Traffic, 8(5), 471–485.
combining live-cell imaging and immunolabeling of ultrathin cryosections. Nature Methods, 5(11),
973–980.
output and automated 3D correlative light-electron microscopy method. Traffic, 9(11), 1828–1838.
Vicidomini, G., Boccacci, P., Diaspro, A., & Bertero, M. (2009). Application of the split-gradient method to
3D image deconvolution in fluorescence microscopy. Journal of Microscopy, 234(1), 47–61.
Vicidomini, G., Gagliani, M. C., Cortese, K., Krieger, J., Buescher, P., Bianchini, P., et al. (2010). A novel
approach for correlative light electron microscopy analysis. Microscopy Research and Technique, 73(3),
215–224.
localization in electron micrographs using fluorescence nanoscopy. Nature Methods, 8(1), 80–84.
William, E. L., & Harvey, E. C. (1987). Marching cubes: A high resolution 3D surface construction algorithm.
Computer Graphics, 21(4), 163–168.
CHAPTER 7

of GFP
Markus Grabenbauer
Department of Systems Cell Biology, Max-Planck-Institute for Molecular Physiology, Otto-Hahn-Str. 11,
D-44227 Dortmund, North Rhine-Westphalia, Germany
Abstract
I. Introduction
II. Rationale
III. Methods
A. Day 1––Cell Culture
B. Day 2––Fixation, Bleaching, and EM Preparation
C. Day 3––Electron Microscopy
D. Electron Tomography, 3D Reconstruction, and Modeling
IV. Materials
A. Cell Culture
B. Fixation, Bleaching, and EM Preparation
C. Electron Microscopy and Electron Tomography
V. Discussion
Acknowledgments
References
Abstract
The correlation of light and electron microscopy (EM) is a powerful tool as it combines
the investigation of dynamic processes in vivo with the resolution power of the electron
microscope. The green fluorescent proteins (GFPs) and its derivatives revolutionized
live-cell light microscopy. Hence, this review outlines correlative microscopy of GFP
through photo-oxidation, a method that allows for the direct ultrastructural visualization
of fluorophores upon illumination. Oxygen radicals generated during the GFP bleach-
ing process photo-oxidize diaminobenzidine (DAB) into an electron dense precipitate
that can be visualized both by routine EM of thin sections and by electron tomography
for 3D analysis. There are different levels of correlative microscopy, i.e. the correlation
of certain areas, cells, or organelles from light to EM, where photo-oxidation of DAB
through GFP allows the highest possible degree––the correlation of specific molecules.
118 Markus Grabenbauer
I. Introduction
Photo-oxidation is a powerful tool in correlating fluorescent signals as detected by

light microscopy, embedded in their ultrastructural environment as detected by elec-
tron microscopy (EM) and electron tomography. Free oxygen radicals that form upon
illumination, precipitate 3,3’-diaminobenzidine (DAB) into an electron-dense reaction
product (Maranto, 1982). Most organic fluorophores generate reactive oxygen species
in their exited state, such as singlet oxygen radicals (1O2). These interact very fast in
their immediate vicinity on the nanometer scale with DAB monomers, which form
radical intermediates (Fig. 1). Upon oxidative polymerization and oxidative cycliza-
tion, the insoluble DAB polymer is formed. The osmiophilic DAB precipitate can then
be visualized by EM to reveal detailed spatial information. The main purpose of cor-
relative microscopy is to combine the advantages of both imaging techniques. In bio-
logical studies this is achieved by analyzing the dynamics of live cells or organisms in
low magnification by light microscopy, and subsequently by observing the same struc-
tures at higher resolution by EM. Cells or tissues can be fixed and processed through
a photo-oxidation procedure (Fig. 2), so that finally a correlative analysis of the very
same structure is possible (Fig. 3).
Fig. 1 Reaction scheme of the DAB photo-oxidation through GFP. By strong illumination of GFP, electrons
are excited and generate singlet oxygen radicals. These react with DAB monomers in their direct vicinity and
start the process of oxidative polymerization and oxidative cyclization. The insoluble deposition of DAB poly-
mer can be stained by osmium tetroxide. Image provided by Katja van Eickels (MPI Dortmund, Germany).
7. Correlative Light and Electron Microscopy of GFP 119
The photo-oxidation technique was originally described for neuronal mapping, using
micro-injected Lucifer yellow as fluorescent marker (Maranto, 1982), and in this inter-
relation, it is still in use (Nikonenko, Boda, Alberi, & Muller, 2005). In the mean time,
dozens of organic fluorophores have been reported to photo-oxidize DAB in correlative
microscopy studies. These include carbocyanine dyes such as 1,1’-dioctadecyl-3,3,3’,3’-
tetramethylindocarbocyanine perchlorate (DiI) (von Bartheld, Cunningham, & Rubel,
1990), boron dipyrromethene difluoride (BODIPY) (Pagano, Sepanski, & Martin, 1989;
Fig. 2 Bleaching process in higher (A–D) and lower magnification (E–H), and the same area after osmification
and Epon embedding (I–L). (A) Fluorescence signal of GFP-coupled Golgi-marker in HeLa cells. (B) Same
cells in bright field microscopy appear transparent. Bleaching process for 3 min (C) and 5 min––which would
be slightly overdeveloped for further EM analysis (D). Fluorescence of the same cells in lower magnification
before (E) and after bleaching (F) shows the extinction of the signal in a defined area (white box). Differential
interference contrast of the same area before (G) and after bleaching (H) shows appearance of DAB precipitate
in the bleached spot. After osmification and Epon embedding, the bleached area is heavily stained by osmium,
clearly visible in lower (I) and medium magnification (J). Higher magnification shows cells partly stained with
a clearly defined precipitation border (K–L). Scale bar: 10 µm.
Pagano, Martin, Kang, & Haugland, 1991), eosin (Deerinck et al., 1994), and many others.
Most reports on staining synaptic vesicles use styryl dyes such as FM1-43 (Branco,
Marra, & Staras, 2010; Denker et al., 2011; Harata, Ryan, Smith, Buchanan, & Tsien,
2001; Hoopmann, Rizzoli, & Betz, 2012). To get a comprehensive overview, the
reader is referred to the following reviews: Lubke, 1993; Meisslitzer-Ruppitsch, Rohrl,
Neumuller, Pavelka, & Ellinger, 2009; Modla & Czymmek, 2011; Sosinsky, Giepmans,
Deerinck, Gaietta, & Ellisman, 2007. In contrast to label complete cells by micro-injecting
fluorophores unspecifically into their cytosol, BODIPY as ceramide analogue was
already used in an improved way to target the correlative fluorescent tag to specific
intracellular membranes (Pagano et al., 1991). The staining of specific lipids to study
their synthesis, subcellular localization, and degradation was an important application
of early photo-oxidation experiments. A detailed overview on lipid staining gives the
excellent review of Meisslitzner-Ruppitsch et al. (2009) gives an excellent review.
There are now approaches available to use membrane-permeable biarsenical fluorescein
derivatives (FlAsH and ReAsH) to specifically label protein chimeras bearing a short
tetracysteine polypeptide sequence (Adams et al., 2002). This allowed for multicolored
fluorescence analysis of connexin 43 dynamics and subsequent correlated EM of DAB
photo-oxidized by ReAsH (Gaietta et al., 2002). Until now, a successful photo-oxidation
of FlAsH has not yet been reported. The combination of green fluorescent protein (GFP)
live cell imaging and subsequent photo-oxidation of ReAsH bound to a tetracysteine tag
in the same chimeric molecule allowed for new insights into Golgi apparatus behavior
during and after mitosis (Gaietta et al., 2006).
In the literature, there is currently some confusion about the terms “photo-oxidation” and
“photoconversion.” Although the original report already determined “photo-oxidation”
in its title as the “name” for this correlative microscopic technique (Maranto, 1982),
there are numerous publications wrongly reporting about “photoconversion,” even the
very recent ones. Apart from being just wrong, it is also misleading, as photoconversion
is an occupied scientific term for the biophysical process of inducing structural changes
to fluorophores and changing their properties. Photoconversion is mainly mediated by
irradiation, for example, on photoreceptors or on specifically designed fluorophores,
such as the green-to-red conversion of the fluorescent protein Kaede (Mizuno et al.,
2003). A comprehensive overview on photoswitchable proteins was recently provided
Fig. 3 Correlative light and electron microscopy of the same Golgi apparatus. HeLa cells expressing the cyan
mutant ECFP fused to GalNAc-T2 are shown by fluorescence light microscopy (A) and fluorescence combined
with the bright field channel (B). During the bleaching process in presence of DAB, fluorescence fades away
and background staining appears: 0 min (C), 4 min (D), and 6 min (E). After 8 min (F), the reaction was stopped.
At this time-point, the fluorescence was completely erased. One of the cells (highlighted by arrows) is identified
in the electron microscope (G). A juxtanuclear Golgi stack (arrow) becomes visible at higher magnifications
(H, I). The Golgi stack shows specific and very precisely localized DAB precipitation in the lumen of two
cisternae (J). This represents DAB photo-oxidized by ECFP bleaching and correlates directly to the observed
fluorescence (arrow in A). Other dark structures like mitochondria, autophagic vacuoles, background precipita-
tion, as well as osmium stained membranes were distinguished from the specific DAB staining by comparing
with the control cells without GFP fusion proteins. Adapted from Grabenbauer et al. (2005). Scale bars: 25 µm
(A–F), 10 µm (G), 5 µm (H), 2 µm (I), and 500 nm (J).
by the Lukyanov brothers (Lukyanov, Chudakov, Lukyanov, & Verkhusha, 2005).

The necessary distinction between the terms “photo-oxidation” versus “photoconver-
sion” is clearly described in a recent review on photo-oxidation (Meisslitzer-Ruppitsch
et al., 2009).
One of the most frequently used fluorophores in cell biology studies is based on
the GFP from Aequorea victoria. Discovered in the 1960s (Shimomura, Johnson,
& Saiga, 1962), cloned in the early 1990s (Prasher, Eckenrode, Ward, Prendergast,
& Cormier, 1992), and later successfully expressed in model organisms (Chalfie,
Euskirchen, Ward, & Prasher, 1994), GFP and its mutants revolutionized cell biology.
Along with spectral mutants and fluorescent proteins from other Cnidarian sources,
GFP fusion proteins have provided important insights into both temporal and spa-
tial organization of molecules inside living cells. The use of GFP and their mutants
for cell and developmental biology is summarized in excellent reviews (Lippincott-
Schwartz & Patterson, 2003; Meisslitzer-Ruppitsch et al., 2009; Shimomura, 2006;
Tsien, 1998).
At the level of EM, the detection of GFP proved to be challenging. Most studies rely
on the use of primary antibodies against GFP, analyzed by immuno-EM (Schwarz &
Humbel, 2007) (Spiegelhalter et al., 2010). The first attempt to detect GFP through
photo-oxidation was done on peroxisomal-targeted GFP in yeast (Monosov, Wenzel,
Luers, Heyman, & Subramani, 1996). In this study, wild-type GFP was used, dis-
playing suboptimal behavior in photo-oxidation. Due to the presence of catalase in
yeast peroxisomes, it is likely that DAB polymerization was driven by the peroxidatic
activity of catalase, which is still enzymatically active in cells after their fixation
(Fahimi, 2009). That protocol has not been used in additional studies.
About 10 years later, the first successful photo-oxidation of DAB through GFP
was reported in mammalian cells (Grabenbauer et al., 2005)(Fig. 4). Due to the
low capacity of GFP to release singlet oxygen during illumination, existing photo-
oxidation protocols have been modified to gain more efficient preparation, allow-
ing for the ultrastructural detection of the fine DAB precipitate. Rather than single
modifications, it was the combination of diverse steps to enhance sensitivity over a
certain threshold. These steps included reduction of background DAB precipitation
by blocking endogenous enzyme activities and degrading autofluorescence without
affecting the GFP fluorescence. Most momentous was probably the addition of pure
oxygen gas during bleaching, intensifying the production of reactive oxygen species,
and the abdication of any postcontrasting metals or procedures. After photo-oxidation
through GFP, the membrane contrast was mainly induced by the “reduced osmium
method” using potassium ferrocyanide (Karnovsky, 1971). In correlative experiments
conducted on cells and organelles of different fluorescence intensity, a linear relation-
ship between initial fluorescence and final DAB precipitation was shown (Fig. 5).
This linear correlation of signals in light and electron microscopy proves photo-
oxidation through GFP as quantitative technique, at least in certain experimental
conditions. The highly sensitive photo-oxidation protocol which is outlined in the fol-
lowing sections were applied and improved in further studies (Meisslitzer-Ruppitsch
et al., 2008) (Asp et al., 2009).
Fig. 4 Correlative microscopy of GFP through photo-oxidation. HeLa cells expressing the Golgi-resident
enzyme GalNAc-T2 fused to EGFP. High-resolution light microscopy shows fluorescence exclusively localized
to the Golgi apparatus, nuclear counterstain by DAPI (A). After photo-oxidation and Epon embedding, the DAB
staining can be identified by electron microscopy, resembling the GFP localization at the juxtanuclear Golgi stack
(B). At higher magnification, the DAB polymerization sites can be precisely localized to the cisternal lumina with
a gradient-like distribution across the Golgi stack (C). The GFP is tagged to the luminal side of the transmembrane
protein anchor of GalNAc-T2. Control cells without GFP expression do not show any DAB precipitation at the
Golgi apparatus after illumination (D). Slight poststaining using uranyl-acetate results in stronger membrane con-
trast, but the DAB signal is less prominent (E). GalNAc-T2 coupled to the ECFP mutant resembles the identical
DAB distribution across the Golgi stack (F). A–D adapted from Grabenbauer et al. (2005). Scale bars: 5 µm (A),
500 nm (B), 100 nm (C), 200 nm (D–E), 100 nm (F). (See color plate.)
Fig. 5 Quantitative analysis of the signal-to-noise ratio by correlative microscopy. Part of GalNAc-T2ECFP
expressing cells (A) were prebleached in the absence of DAB substrate (dashed line in B) and completely
bleached in the presence of DAB. The same cells identified by bright field light microscopy (dashed box in C) are
retraced in a thin section by electron microscopy (D). By enhancing the magnification (D–G) individual Golgi
stacks are revealed (G), which are directly correlated to fluorescent Golgi apparatus (arrow in A). The dashed
squares (C–F) outline the proximate higher magnified images (D–G). Transversal cut Golgi stacks of the
same ultra thin section are imaged in the same magnification and correlated to individual Golgi apparatus fluores-
cence signals after prebleaching (B). The signal-to-noise ratios of specific ECFP Golgi fluorescence against cyto-
sol and specific DAB precipitation against cytosolic background is plotted in a graph and shows a linear regression
(H). Strongly prebleached cells do not show any DAB precipitation in their cisternal lumen (I). Note the absence
of background staining in mitochondria due to blocking endogenous oxidases. Adapted from Grabenbauer et al.
(2005). Scale bars: 25 µm (A–C), 10 µm (D), 5 µm (E), 2 µm (F), 200 nm (G, I). (See color plate.)
II. Rationale
GFP and related variants are very commonly used in light microscopy studies. The
rationale here is to describe the conditions to perform correlative light and electron
microscopy using free radicals generated by GFP to photo-oxidize DAB. The goal is
to establish sensitive conditions that produce a DAB precipitate of sufficient quality to

enable electron microscopic investigations in both 2D and 3D. The method presented
here is optimized for correlative visualization of GFP-tagged proteins residing in
secretory pathway of mammalian cells. Specific adaptations of this technique are most
likely necessary for the analysis of different organelles, cellular compartments, or other
model systems including tissues. In the following paragraphs, a detailed step-by-step
description of the method is given.
III. Methods
An established Golgi marker, human Golgi-resident enzyme N-acetylgalactos-

aminyltransferase 2, was fused to the enhanced green or cyan GFP-mutants, EGFP
(GalNAc-T2GFP) and ECFP (GalNAc-T2CFP), respectively. Both markers are stably
expressed in HeLa cells and have been extensively characterized both at the light and
at the ultrastructural levels (Storrie et al., 1998).
A typical photo-oxidation procedure can usually be performed in 3 days (excluding 3D
tomographic analysis). Briefly, day 1 is mainly to prepare for the actual bleaching experi-
ment and seeding of cells into glass bottom culture dishes (e.g., MatTek). Eventually,
perturbations of the cells or certain treatments have to be performed before proceeding
further. During day 2, the central bleaching experiment takes place including fluores-
cence microscopy, fixation of samples, preincubation steps, the actual GFP bleaching,
postfixation, dehydration, and finally embedding in resin. Day 3 includes the removal
of the bottom glass from samples, ultrathin sectioning, EM, and the data evaluation.
It is not necessary, however, to accomplish the preparation procedure within the three
consecutive days. Moreover, it is recommended to provide at least 24 h in between each
step for the proper attachment of cells to their support (after day 1) and sufficient resin
polymerization (after experimental day 2). If the intense procedure of experimental day 2
is utterly exhausting, it is favorable to leave the samples in buffer after the osmium post-
fixation step and finalize the dehydration and resin embedding steps the next morning.
For the experimenter, some basic experience in fluorescence microscopy, cytochem-
ical sample preparation, and EM is required. To avoid false positive or false negative
ultrastructural localization of GFP, a detailed fluorescence microscopic evaluation of
the GFP localization should be done before examining the sample at the ultrastruc-
tural level. Control experiments without illumination and mock experiments without
GFP expression should always be performed in parallel, as well as positive controls.
This will minimize the risk of misinterpretations. The chemical fixation procedure was
optimized to achieve a morphology of the Golgi apparatus and the secretory pathway
that resembles the high-resolution results obtained by cryo-EM. Adaptation to other
compartments, fluorescent proteins, cell types, or organisms may require further opti-
mization of the procedure.
It is important to note here that some chemicals used in this protocol are highly
poisonous and therefore dangerous! Precautions should be taken at all steps of the pro-
cedure. These include the following: safe storages, use of a fume hood, wearing safety
gloves, as well as disposing the poisonous waste in an appropriate way.
A. Day 1––Cell Culture

Cells are seeded on uncoated glass bottom Petri dishes 24 h to 48 h before use. For
correlative microscopy, the “CELLocate” or “grid” glass bottom dishes are recom-
mended (MatTek). In general, the bleached spot can be detected after osmification and
resin embedding using a binocular microscope. Alternatively, individual “landmarks”
can be scratched on the glass bottom using a diamond pen, followed by UV-light
sterilization before seeding the cells.
For correlating samples after embedding based on morphological features, the cells
should be grown in loose groups and not as a confluent monolayer, so that each cell
could be later identified by its individual shape. Alternatively, for confluent mono-
layers (e.g., polarized cells) or tissues, more sophisticated correlative identification
techniques can be used.
B. Day 2––Fixation, Bleaching, and EM Preparation

All incubation steps including fixation, bleaching, dehydration, and resin polym-
erization are carried out in the glass bottom tissue culture dishes, prepared on day 1.
(1) Take cells from incubator and wash immediately with prewarmed (37°C)
calcium- and magnesium-free phosphate-buffered saline (PBS) pH 7.4 for 30 s
or less.
(2) Fix cells with prewarmed (37°C) fixative containing 2% glutaraldehyde and 2%
sucrose in PBS for 20–30 min.
(3) Wash samples for 3 × 5 min with PBS.
(4) Block endogenous enzyme activity with 50–100 mM potassium cyanide and
100 mM glycine in PBS for at least 1 h.
(5) Block endogenous autofluorescence by 100 mM ammonium chloride in PBS
for 40 min.
(6) Block endogenous autofluorescence by 10 mg/ml sodium borohydrate freshly
dissolved in PBS for 40 min. Potassium cyanide eliminates background DAB
oxidation that occurs as a consequence of mitochondrial respiration. If back-
ground staining remains persistent in the mitochondrial matrix, the potassium
cyanide incubation time has to be prolonged. Alternatively, potassium cyanide
can be added to the final DAB solution (see Step 8). Glycine, ammonium chlo-
ride, and sodium borohydrate reduce glutaraldehyde-induced autofluorescence
and consequently, background DAB precipitation.
(7) Wash samples for 3 × 5 min in Tris-buffered saline (TBS) at pH 7.4.
(8) Photo-oxidation: Samples are incubated in a solution of 1–2 mg/ml DAB in
TBS. The DAB solution should be freshly prepared, kept on ice in the dark, and
only be used on the very same day. Immediately before applying to the samples,
the DAB solution should be oxygen enriched by letting pure oxygen gas bub-
bling through. Only a well visible GFP fluorescence at this stage will result in
effective photo-oxidation evaluable by EM. To bleach GFP for photo-oxidation,
samples are illuminated with the appropriate filter settings for EGFP (excitation
filter BP 470/40) or ECFP (excitation filter BP 436/20) using a bright fluores-

cence lamp on an inverted microscope and a high-numerical-aperture objec-
tive (63× Plan-Apochromat 1.4 NA). If available, the climate chamber of the
microscope should be set to the lowest temperature (below 10°C) for effective
DAB precipitation. Alternatively, a specimen cooling-chamber can be used. In
general, the lower the temperature, the higher the oxygen saturation, and there-
fore photo-oxidation efficiency! But freezing and crystallization of the sample
has to be avoided in any circumstances. The development of the DAB reaction
should be monitored very carefully and immediately stopped if strong cytosolic
background staining occurs, typically after 20–30 min using EGFP or after 5–10
min using ECFP.
(9) Wash samples 3 × 5 min in distilled water.
(10) Postfixation of samples in 1% osmium tetroxide reduced by 1.5% potassium
ferrocyanide for 30 min on ice.
(11) Wash samples for 5 min in distilled water.
(12) Dehydration by incubation in graded ethanol series of 50%, 70%, 80%, 90%,
95%, 3 × 100% ethanol, 5 min each. The cells should not be allowed to dry out
during this procedure.
(13) Apply Epon on the cells for 1 h at room temperature.
(14) Remove Epon by dripping off. Take care not to damage or dry the samples.
Apply fresh Epon. Put a small Eppendorf tube without lid and filled with Epon
on top of the cells. Once polymerized, it will serve elegantly as block for mount-
ing on ultramicrotome.
(15) Put samples in an oven to polymerize at 60°C overnight.
C. Day 3––Electron Microscopy

After polymerization, the culture dish glass bottom is removed by immersion in
concentrated hydrofluoric acid (30% HF) (Please act very carefully! Do not use glass
containers for this!). Alternatively, the glass bottom can be disrupted by harsh tempera-
ture changes (by dipping in liquid nitrogen at −196°C). Under a binocular microscope,
it should be checked that the glass has been removed completely, as any remnants will
damage the diamond knife during cutting. Bleached spots containing the cells of inter-
est should be visible using either the ultramicrotome or a more sophisticated binocular
microscope. If the glass had correlation marks before embedding (CELLocate or self-
scratched by a diamond pen), this will be visible now as elevation on the surface of the
Epon block. At this point, the resin block can be further processed with an EM trimmer
at an angle of 30° to get a pyramid containing the bleached cells.
Ultrathin sections should be cut parallel to the surface of the former coverslip on
an ultramicrotome. This ensures an appropriate orientation of the cells in the sections,
resembling the axis of imaging by light microscopy. Approaching the diamond knife
edge to the sample should be done with great care and as parallel as possible, to get
the first sections already from the complete block. As flat embedded cells are about
5–20 µm in thickness, there is only limited material for smoothing the block surface.
Ultrathin sections of 40–70 nm thickness are collected on the surface of the diamond
knife water bath, and mounted on Formvar-coated EM grids. After drying for 1 h, the
grids are ready for electron microscopic analysis. If necessary, sections can be counter-
stained with 1% aqueous uranyl acetate and lead citrate. However, as these salts seem
to have a higher affinity to osmicated compounds rather than to DAB precipitates, the
DAB signal-to-noise ratio will be reduced. Therefore, this is not recommended.
EM should be carried out using small objective and condenser apertures to enhance
the contrast. Imaging with sensitive digital cameras will be advantageous. Semithin
sections (100–300 nm) can also be analyzed by electron tomography, if the neces-
sary environment (electron microscope capable of recording well-aligned tilt series, 3D
reconstruction software) is available.
D. Electron Tomography, 3D Reconstruction, and Modeling

Epon sections of deep purple interference color—corresponding to 250 nm thickness—
can be decorated on both sides with 10 nm sized colloidal gold as fiducial markers. For
electron tomography, dual axis tilt series are automatically recorded around two orthogo-
nal axes using either the Open Tomo program (Ziese et al., 2002) or the SerialEM software
(Mastronarde, 2005). Tomograms of the two tilt series are generated and combined using
the IMOD program package (Kremer, Mastronarde, & McIntosh, 1996). Golgi stacks
and vesicles are modeled by computer-assisted tracing of membrane contours based on
gray value thresh-holding. The presence or absence of DAB precipitation in cisternae
and vesicles is decided upon local intensity differences (thresh-holding) within the tomo-
gram and the modeled objects are annotated to visualize localization of DAB precipitate
representing GFP fluorescence.
IV. Materials
A. Cell Culture
HeLa cells (No. CCL 185; American Type Culture Collection, Rockville, MD, USA)
were grown in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal calf
serum (FCS) and 2 mM L-glutamine (GIBCO BRL, Life Technologies) at 37°C and
5% CO2 on uncoated glass bottom (No. 1.5) Petri dishes (MatTek, Ashland, MA, USA).
For correlative microscopy, the “CELLocate” or “grid” glass bottom dishes are rec-
ommended (MatTek). Stable transfectants of HeLa cells were generated as described
previously (Storrie et al., 1998). Briefly, plasmids encoding GalNAc-T2GFP or Gal-
NAc-T2CFP were transfected into HeLa cells cultured in 10 cm tissue culture dishes in
the presence of 5% FCS using the calcium phosphate protocol (Paabo, Weber, Nilsson,
Schaffner, & Peterson, 1986). Selection was for ∼3 weeks in the culture medium sup-
plemented with geneticin (G-418 sulfate, 300 µg/ml)(Sigma) or puromycin (1 µg/ml)
(Sigma), respectively. After significant cell death occurred, positive clones were recul-
tured and checked for appropriate fluorescence.
B. Fixation, Bleaching, and EM Preparation

If not stated otherwise, all chemicals were purchased by Sigma-Aldrich.
Chemicals: PBS (calcium and magnesium free), TBS, glutaraldehyde (25% stock solu-
tion, Merck, Darmstadt, Germany), sucrose (USBCorporation, Cleveland, Ohio, USA),
potassium cyanide, glycine, sodium borohydrate, 3,3’-DAB hexahydrate (Polysciences,
Eppelheim, Germany), osmium tetroxide (Serva, Eppelheim, Germany), potassium fer-
rocyanide, Epon 812 (Serva, Eppelheim, Germany), hydrofluoric acid.
Microscopes: Different inverted fluorescence microscopes with high-numerical-aper-
ture objectives (63× Plan-Apochromat 1.4 NA) have been used. Best results were obtained
with the following instruments: Zeiss Axiovert 200 and Zeiss Axio Observer.Z1 (both
by Carl Zeiss, Oberkochen, Germany). Chosen were appropriate filter settings for EGFP
(38 HE excitation filter BP 470/40) or ECFP (47 HE excitation filter BP 436/25). Samples
were bleached using a 100 W mercury lamp (AttoArc) or a 120 W HXP metal-halogenide
lamp (both by Carl Zeiss, Oberkochen, Germany). Sample cooling was achieved by speci-
men cooling chamber “Temperable Insert P” (Carl Zeiss, Oberkochen, Germany). For
finding slightly bleached spots on Epon samples, a sophisticated stereomicroscope gives
advantages: Stereo-Lumar.V12 (Carl Zeiss, Oberkochen, Germany).
C. Electron Microscopy and Electron Tomography

Preparative instrumentation: Resin block trimmer (Leica EM Trim, Leica Microsys-
tems, Bensheim, Germany), ultramicrotome (Leica Ultracut S, Leica Microsystems,
Bensheim, Germany).
Microscopes: Samples were examined at 100 kV accelerating voltage on a Philips
BioTwin CM 120 equipped with a top mount Gatan TEM camera or at 200 kV acceler-
ating voltage on a Philips CM 200 FEG field emission gun microscope, equipped with
a Tietz slow scan CCD camera (both microscopes by Philips Eindhoven, Netherlands).
For electron tomography, dual axis tilt series were acquired at 200 kV on an FEI Tecnai
20 LaB6 transmission electron microscope using a TemCam F214 setup with 14 µm
sized pixels (TVIPS GmbH, Gauting, Germany) or at 120 kV on a JEOL JEM-1400
LaB6 transmission electron microscope (JEOL Germany, Eching, Germany) using either
2K × 2K or 4K × 4K CCD cameras (F-214FS or F-416; TVIPS, Gauting, Germany).
V. Discussion
Photo-oxidation remains a versatile tool for correlating light microscopy and EM,
since its introduction more than 30 years ago (Maranto, 1982). The preparation is techni-
cally demanding, needs fine handling and experience in cytochemistry, light microscopy
and EM, and is––due to the numerous controls, which have to be processed in parallel––
also time consuming. This might be reflected in the relatively large number of scientists
discontinuing their application of the technique, including the original founder Anthony
R. Maranto himself. Nevertheless, once established in a laboratory, photo-oxidation is
one of the most powerful techniques to correlate in vivo fluorescence signals with the
superior resolution of EM. For both, fluorescence microscopy and EM, rather “stan-
dard” instruments can be used. A regular epifluorescence microscope with little addi-
tional equipment such as cooling device will provide sufficient bleaching, as long as the
fluorescence lamp is powerful and the numerical aperture of the objective lens is good
enough (around 1.2 NA or higher). Bleaching either by latest confocal laser scanning
microscopes or by directly coupled lasers of a total internal reflection fluorescence (TIRF)
microscope, our samples displayed nondetectable DAB precipitation, or appeared heav-
ily overstained. Using shutters in the beam path of epifluorescence microscope, elabo-
rated bleaching patterns can already be generated manually on cells and even on defined
parts of the cell, which is shown in Fig. 2. Also, the electron microscopic analysis does
not need necessarily high-end instrumentation. Cells and even organelles can be corre-
lated based on morphological features using bright field or differential interference con-
trast micrographs of the corresponding preparation (Fig. 3). Enhancing contrast on the
light microscope by phase contrast is not recommended, at least not for the bleaching
process, as the phase ring in the objective beam path reduces its light transmittance.
The main challenge in using low emitters of reactive oxygen radicals, such as GFP
and its mutants, for photo-oxidation is the improvement of the preparation procedure
combined with acceptable ultrastructural preservation. This already starts with the fixa-
tion of the sample. As high-pressure freeze fixation/freeze substitution is not compati-
ble with the photo-oxidation procedure, the chemical fixative was optimized to achieve
a morphology of the Golgi apparatus (Fig. 4) and secretory pathway closest resem-
bling to high-resolution results obtained by cryo-electron microscopic preparations
(Bouchet-Marquis, Starkuviene, & Grabenbauer, 2008). Concerning the background
DAB precipitation by sample autofluorescence, one could use less autofluorescence-
inducing fixatives such as paraformaldehyde, but the compromise will be an artificially
modified ultrastructure. A more gentle way is reducing the glutaraldehyde-induced
autofluorescence by chemical blocking using glycine, ammonium chloride, and sodium
borohydrate (Baschong, Suetterlin, & Laeng, 2001; Harata et al., 2001). Further, there
have been numerous adjustments to existing photo-oxidation protocols, as outlined in
Chapter III. Oxygenation of the DAB solution had very significant influence. Already
in 1988, Sandell and Masland argued that reactive oxygen species generated during
illumination are the driving force for photo-oxidation of DAB (Sandell & Masland,
1988). Later, it was shown that replacing oxygen in the working solution with carbon
dioxide (Lubke, 1993) or argon (Deerinck et al., 1994) inhibited the reaction, whereas
an oxygen-enriched reaction chamber increased the photo-oxidation efficiency (Kacza,
Hartig, & Seeger, 1997). To detect the fine DAB precipitation by EM, it is mandatory
to omit contrast enhancement by uranyl acetate and/or lead citrate, as these heavy metal
salts have higher affinity to osmium in membranes than to DAB polymers, lowering the
signal-to-noise ratio of DAB detection dramatically. In this case, the use of “reduced
osmium” for membrane contrast enhancement is recommended (Karnovsky, 1971).
Finally, the sum of these protocol enhancement steps resulted in the first successful
photo-oxidation using GFP mutants (Grabenbauer et al., 2005) (Fig. 4). However, there
have been numerous adjustments and applications of this protocol (Asp et al., 2009;
Meisslitzer-Ruppitsch et al., 2008). A very important finding was the direct and linear
correlation of the initial fluorescence signal in organelles to the final DAB precipi-
tate (Fig. 5). This makes photo-oxidation through GFP a quantitative method opening
numerous new possibilities to analyze cells and cellular networks in conjunction with
quantitative systems biology approaches.
In contrast to immuno-EM, where 3D data evaluation is very sparse (Zeuschner
et al., 2006), electron tomography analysis of photo-oxidation preparations is not
hampered by the inhomogeneous distribution of the labeling density throughout the
sample, which can be obtained by postembedding techniques using antibodies or other
substances with reduced penetration into the sections. Three-dimensional imaging is
necessary to extract spatial information when using correlative markers in highly con-
voluted membrane structures such as those found in the secretory pathway (Marsh,
Mastronarde, Buttle, Howell, & McIntosh, 2001). Early photo-oxidation approaches on
Golgi apparatus already took advantage of studying lipid traffic between the trans-Golgi
cisternae and the trans-Golgi network in 3D (Ladinsky, Kremer, Furcinitti, McIntosh,
& Howell, 1994). By electron tomography of DAB photo-oxidized through GFP, it was
unambiguously shown that Golgi-resident enzymes are not excluded from peri-Golgi
vesicles (see Figs. 6 and 7), which was in accord with previous in vivo (Martinez-
Menarguez et al., 2001) and in vitro findings (Malsam, Satoh, Pelletier, & Warren,
2005), as well as with recent quantitative proteomics data (Gilchrist et al., 2006).
The “labeling resolution” defines the radius of the marker around his putative target.
For immuno-EM, in conjunction with colloidal gold labeling, the radius of the gold
particle (2.5–7.5 nm) has to be extended by protein bridges and the length of primary
antibody, which can be in the range of 10–20 nm in total (Roth, 1996). The radius using
quantum dots as correlative markers can be in the same range. Quantum dots are inor-
ganic and fluorescent nanocrystals, which can be identified by EM according to their
distinct shapes (Giepmans, Deerinck, Smarr, Jones, & Ellisman, 2005; Sosinsky et al.,
2007). Recently, the activation states of kinases have been determined by electron
tomography of quantum dots bound to Eph receptor tyrosine kinases, using their dis-
tance to a reference membrane to differentiate active from inactive forms (Janes et al.,
2009). Such exact measurements of 4–8 nm distances require specialized experimental
settings. However, the labeling resolution of DAB photo-oxidation through GFP is in
the range of 5–10 nm, due to the faint DAB deposition (Grabenbauer et al., 2005), so
far one of the best labeling resolutions achieved for EM markers.
Photo-oxidation in general and photo-oxidation through GFP in particular has also
limitations. Analyzing GFP-tagged proteins occurring in low copy number, or working
with samples suffering from high initial autofluorescence might be very cumbersome.
Even if fine DAB signals might be detectable, such results will be less convincing
than prominent DAB precipitations emerging from densely packed membrane proteins.
Furthermore, if the analyzed compartment is initially electron dense such as autopha-
gic vacuoles, the differentiation between “real” signal and dark background might be
impossible. The same ambiguity might occur if pure cytosolic proteins are observed
without fluorescent aggregations. Such signals will result in evenly distributed cytosolic
DAB precipitation, which has very low signal-to-noise ratio by electron microscopic
Fig. 6 Electron tomographic 3D analysis of a Golgi stack containing GalNAc-T2GFP. In a virtual 5 nm sec-
tion through the tomogram (A), the DAB precipitation representing GalNAc-T2GFP is clearly visible in two
Golgi cisternae, a peri-Golgi vesicle (arrow) and a COP-coated bud (arrowheads). In the 3D model extracted
from the tomogram, the GalNAc-T2GFP containing cisternae are colored in green, GalNAc-T2GFP-containing
vesicles in red, other vesicles in white, nonlabeled cisternae in blue, and ER in deep purple (only visible in the
colored online version) (B). A virtual section of the tomogram is set as background image. A GalNAc-T2GFP
containing vesicle (arrow in A) was reconstructed by gray level thresh-holding as an alternative representation
method to manual tracing. Shown in perspective views with and without virtual sections (C–E). Adapted from
Grabenbauer et al. (2005). Scale bar: 100 nm. (See color plate.)
evaluation. Moreover, compartments with a known potential of generating high back-

ground, such as mitochondrial matrix, need an enormous number of control experi-
ments to get convincing results. On the other hand, if mitochondria are not the primary
target of a particular experiment, one could omit blocking their oxidases to use their
DAB precipitation for easier relocation and correlating specific subcellular areas (see
Fig. 8). Nevertheless, compared to enzyme-driven DAB localization techniques such
as horseradish peroxidase (HRP) coupled markers, photo-oxidation through GFP has
a far better resolution of labeling, because the DAB reaction is not saturated and DAB
polymers do not diffuse far away from their target (maximum about 5–10 nm). Fur-
thermore, HRP enzyme cytochemistry driven morphological artifacts, such as severely
swollen Golgi cisternae, can be minimized (see Fig. 9), probably influenced by lower
temperatures during DAB photo-oxidation.
Fig. 7 Three-dimensional analysis of peri-Golgi vesicles containing GalNAc-T2GFP. In virtual sections (A),
the vesicle has to be localized and can be analyzed by viewing at different angles as presented in perpendicular
views showing x-, y-, and z-axis projections (B–D). Only if it is a rounded profile in all three dimensions, it can
be determined as true vesicle and distinguished from buds or tubular extensions emanating from Golgi cister-
nae. B–C shows DAB precipitate containing peri-Golgi vesicles. For comparison, the vesicle in D is devoid
of DAB. The microscopic resolution is not homogeneous throughout the three axes. Note that the vesicle in B
represents the vesicle highlighted in Fig. 6, but is seen from other perspectives. B–D adapted from Grabenbauer
et al. (2005).
Actually, three GFP variants are reported to work in correlative microscopy through
photo-oxidation, namely enhanced green fluorescent protein (EGFP), enhanced cyan
fluorescent protein (ECFP), and yellow fluorescent protein (Grabenbauer et al., 2005,
Meisslitzer-Ruppitsch et al., 2008). In future, more GFP variants are expected to be
Fig. 8 Photo-oxidation background in mitochondria. HeLa cells expressing fluorescent Golgi marker photo-
oxidized without prior blocking of mitochondrial oxidases (A). Before bleaching, the cells appear transparent
(B). After few minutes irradiation, mitochondria get already stained black in bright field light microscopy (C).
By electron microscopy, the highly specific background can be detected in mitochondrial matrix, while the
intercristae space remains devoid of DAB precipitation. Such background staining can be useful in finding back
certain cells or organelles for correlative microscopy, if the examined GFP-tag resides outside mitochondria.
Scale bars: 10 µm (A–C), 200 nm (D).
successfully adapted to photo-oxidation and could perhaps be modified or optimized for

generating higher amounts of singlet oxygen. Preferably, these GFP versions produce high
amounts of reactive oxygen species after fixation, but not in vivo. Probably, a photocon-
vertible mutant can be generated to be switchable to a more active form after sample fixa-
tion. Recently, a novel genetically encoded tag for correlative microscopy was developed,
based on phototropin 2 of Arabidopsis thaliana (Shu et al., 2011), (Sosinsky et al. 2012,
this MCB volume). This tag generates very high amounts of singlet oxygen, and was named
small singlet oxygen generators (“miniSOGs”), being very potent in photo-oxidation.
Unfortunately, this new protein marker tool needs cofactors to fluoresce, such as flavin
mononucleotide (FMN), interfering with the cellular homeostasis of their host FMN
levels. Much more problematic seems miniSOGs’ generous production of singlet oxygen
species during illumination, rendering their use in live cell imaging questionable. Never-
theless, double-tagged with GFP mutants using different fluorescence excitation maxima,
miniSOG chimeras might be larger, but still suited for correlative studies in live cells.
The photo-oxidation preparation of GFP-tagged proteins with highly sensitive
protocols––as described in this review––is already shown to work for organelles and
Fig. 9 Photo-oxidation in comparison to horseradish peroxidase-mediated DAB precipitation. HeLa cells

expressing the same Golgi-resident enzyme, either coupled to EGFP (A) or to horseradish peroxidase (B). The
ultrathin section in (B) was counterstained with uranyl acetate and lead citrate. By this enhancement of mem-
brane contrast, the strong DAB precipitate remains visible. Note that cisternae containing DAB precipitate are
extremely swollen after peroxidase reaction. Such artificial compartment deformations make any conclusion
on morphology of cisternae and peri-Golgi vesicles highly doubtable. Scale bars: 500 nm (A), 200 nm (B).
compartments such as plasma membrane, endosomes, nuclear envelope, endoplas-

mic reticulum, and the secretory pathway with its main organelle, the Golgi apparatus
(Grabenbauer et al., 2005; Meisslitzer-Ruppitsch et al., 2008). As photo-oxidation in
general is not limited to these organelles, more studies with different compartments
will come up. The future exercise is to adjust this challenging technique from cul-
tured mammalian cells to tissues and to different model organisms like bacteria, yeast,
worms, flies, and plants. In the slipstream of the GFP revolution on light microscopic
level, the GFP detection at ultrastructure might also revolutionize EM and establish
correlative microscopy of GFP as widely applied experimental approach. To sensitize
and optimize the protocol further might finally enable to specifically detect single mol-
ecules at ultrastructural level, whose dynamics have been observed before in live cells.
This would be an enormous break through for getting quantitative biology approaches
to the electron microscopic level.
Acknowledgments
The excellent technical support of Sabine Dongard, Katja van Eickels, and Dr. Oliver
Hofnagel is gratefully acknowledged. Some of the work presented in this review was
supported by an EMBO fellowship (to M.G.) and by the German Ministry of Education
and Research (BMBF Grant NanoCombine).
References
Adams, S. R., Campbell, R. E., Gross, L. A., Martin, B. R., Walkup, G. K., Yao, Y., et al. (2002). New biar-
senical ligands and tetracysteine motifs for protein labeling in vitro and in vivo: synthesis and biological
applications. Journal of the American Chemical Society, 124, 6063–6076.
Asp, L., Kartberg, F., Fernandez-Rodriguez, J., Smedh, M., Elsner, M., Laporte, F., et al. (2009). Early stages
of Golgi-vesicle and -tubule formation require diacylglycerol. Molecular Biology of the Cell. 20, 780–790.
Baschong, W., Suetterlin, R., & Laeng, R. H. (2001). Control of autofluorescence of archival formaldehyde-
fixed, paraffin-embedded tissue in confocal laser scanning microscopy (CLSM). The Journal of Histo-
chemistry and Cytochemistry, 49, 1565–1572.
Bouchet-Marquis, C., Starkuviene, V., & Grabenbauer, M. (2008). Golgi apparatus studied in vitreous sec-
tions. Journal of Microscopy, 230, 308–316.
Branco, T., Marra, V., & Staras, K. (2010). Examining size-strength relationships at hippocampal synapses
using an ultrastructural measurement of synaptic release probability. Journal of Structural Biology, 172,
203–210.
cence photooxidation with eosin: a method for high resolution immunolocalization and in situ hybridiza-
Denker, A., Bethani, I., Krohnert, K., Korber, C., Horstmann, H., Wilhelm, B. G., et al. (2011). A small pool
of vesicles maintains synaptic activity in vivo. Proceedings of the National Academy of Sciences of the
United States of America, 108, 17177–17182.
Fahimi, D. H. (2009). Peroxisomes: 40 years of histochemical staining, personal reminiscences. Histochem-
istry and Cell Biology, 131, 437–440.
Gaietta, G. M., Giepmans, B. N., Deerinck, T. J., Smith, W. B., Ngan, L., Llopis, J., et al. (2006). Golgi
microscopy. Proceedings of the National Academy of Sciences of the United States of America, 103,
17777–17782.
electron microscopic imaging of multiple endogenous proteins using Quantum dots. Nature Methods, 2,
743–749.
Gilchrist, A., Au, C. E., Hiding, J., Bell, A. W., Fernandez-Rodriguez, J., Lesimple, S., et al. (2006). Quan
titative proteomics analysis of the secretory pathway. Cell, 127, 1265–1281.
Grabenbauer, M., Geerts, W. J., Fernadez-Rodriguez, J., Hoenger, A., Koster, A. J., et al. (2005). Correla-
tive microscopy and electron tomography of GFP through photooxidation. Nature Methods, 2, 857–862.
Harata, N., Ryan, T. A., Smith, S. J., Buchanan, J., & Tsien, R. W. (2001). Visualizing recycling synaptic
vesicles in hippocampal neurons by FM 1-43 photoconversion. Proceedings of the National Academy of
Hoopmann, P., Rizzoli, S. O., & Betz, W. J. (2012). FM dye photoconversion for visualizing synaptic vesicles
by electron microscopy. Cold Spring Harbor Protocols, 84–86.
Janes, P. W., Wimmer-Kleikamp, S. H., Frangakis, A. S., Treble, K., Griesshaber, B., Sabet, O., et al.
(2009). Cytoplasmic relaxation of active Eph controls ephrin shedding by ADAM10. PLoS Biology,
7, e1000215.
Kacza, J., Hartig, W., & Seeger, J. (1997). Oxygen-enriched photoconversion of fluorescent dyes by means
of a closed conversion chamber. Journal of Neuroscience Methods, 71, 225–232.
Karnovsky, M. (1971). Use of ferrocyanide reduced osmium tetroxide in electron microscopy. The Journal
of Cell Biology, 51, 146A.
Ladinsky, M. S., Kremer, J. R., Furcinitti, P. S., McIntosh, J. R., & Howell, K. E. (1994). HVEM tomography
of the trans-Golgi network: structural insights and identification of a lace-like vesicle coat. The Journal
of Cell Biology, 127, 29–38.
Lubke, J. (1993). Photoconversion of diaminobenzidine with different fluorescent neuronal markers into a
light and electron microscopic dense reaction product. Microscopy Research and Technique, 24, 2–14.
Lukyanov, K. A., Chudakov, D. M., Lukyanov, S., & Verkhusha, V. V. (2005). Innovation: photoactivatable
fluorescent proteins. Nature Reviews, 6, 885–891.
Malsam, J., Satoh, A., Pelletier, L., & Warren, G. (2005). Golgin tethers define subpopulations of COPI
vesicles. Science, 307, 1095–1098.
Maranto, A. R. (1982). Neuronal mapping: a photooxidation reaction makes Lucifer yellow useful for elec-
tron microscopy. Science, 217, 953–955.
Marsh, B. J., Mastronarde, D. N., Buttle, K. F., Howell, K. E., & McIntosh, J. R. (2001). Organellar relationships
in the Golgi region of the pancreatic beta cell line, HIT-T15, visualized by high resolution electron tomog-
raphy. Proceedings of the National Academy of Sciences of the United States of America, 98, 2399–2406.
Martinez-Menarguez, J. A., Prekeris, R., Oorschot, V. M., Scheller, R., Slot, J. W., Geuze, H. J., et al. (2001).
Peri-Golgi vesicles contain retrograde but not anterograde proteins consistent with the cisternal progres-
sion model of intra-Golgi transport. The Journal of Cell Biology, 155, 1213–1224.
Mastronarde, D. N. (2005). Automated electron microscope tomography using robust prediction of specimen
movements. Journal of Structural Biology, 152, 36–51.
Meisslitzer-Ruppitsch, C., Rohrl, C., Neumuller, J., Pavelka, M., & Ellinger, A. (2009). Photooxidation
technology for correlated light and electron microscopy. Journal of Microscopy, 235, 322–335.
Meisslitzer-Ruppitsch, C., Vetterlein, M., Stangl, H., Maier, S., Neumuller, J., Freissmuth, M., et al. (2008).
Electron microscopic visualization of fluorescent signals in cellular compartments and organelles by
means of DAB-photoconversion. Histochemistry and Cell Biology, 130, 407–419.
Mizuno, H., Mal, T. K., Tong, K. I., Ando, R., Furuta, T., Ikura, M., et al. (2003). Photo-induced peptide
cleavage in the green-to-red conversion of a fluorescent protein. Molecular Cell, 12, 1051–1058.
Monosov, E. Z., Wenzel, T. J., Luers, G. H., Heyman, J. A., & Subramani, S. (1996). Labeling of per-
oxisomes with green fluorescent protein in living P. pastoris cells. The Journal of Histochemistry and
Nikonenko, I., Boda, B., Alberi, S., & Muller, D. (2005). Application of photoconversion technique for
correlated confocal and ultrastructural studies in organotypic slice cultures. Microscopy Research and
Paabo, S., Weber, F., Nilsson, T., Schaffner, W., & Peterson, P. A. (1986). Structural and functional dissection
of an MHC class I antigen-binding adenovirus glycoprotein. The EMBO Journal, 5, 1921–1927.
Pagano, R. E., Martin, O. C., Kang, H. C., & Haugland, R. P. (1991). A novel fluorescent ceramide analogue
for studying membrane traffic in animal cells: accumulation at the Golgi apparatus results in altered spec-
tral properties of the sphingolipid precursor. The Journal of Cell Biology, 113, 1267–1279.
Pagano, R. E., Sepanski, M. A., & Martin, O. C. (1989). Molecular trapping of a fluorescent ceramide ana-
logue at the Golgi apparatus of fixed cells: interaction with endogenous lipids provides a trans-Golgi
marker for both light and electron microscopy. The Journal of Cell Biology, 109, 2067–2079.
Prasher, D. C., Eckenrode, V. K., Ward, W. W., Prendergast, F. G., & Cormier, M. J. (1992). Primary structure
of the Aequorea victoria green-fluorescent protein. Gene, 111, 229–233.
Roth, J. (1996). The silver anniversary of gold: 25 years of the colloidal gold marker system for immunocy-
tochemistry and histochemistry. Histochemistry and Cell Biology, 106, 1–8.
Sandell, J. H., & Masland, R. H. (1988). Photoconversion of some fluorescent markers to a diaminobenzidine
product. The Journal of Histochemistry and Cytochemistry, 36, 555–559.
Shimomura, O. (2006). Discovery of green fluorescent protein. Methods of Biochemical Analysis, 47, 1–13.
Shimomura, O., Johnson, F. H., & Saiga, Y. (1962). Extraction, purification and properties of aequorin, a
bioluminescent protein from the luminous hydromedusan, Aequorea. Journal of Cellular and Compara-
tive Physiology, 59, 223–239.
Biology, 9, e1001041.
Sosinsky, G. E., Giepmans, B. N., Deerinck, T. J., Gaietta, G. M., & Ellisman, M. H. (2007). Markers for
correlated light and electron microscopy. Methods in Cell Biology, 79, 575–591.
Spiegelhalter, C., Tosch, V., Hentsch, D., Koch, M., Kessler, P., Schwab, Y., et al. (2010). From dynamic live
cell imaging to 3D ultrastructure: novel integrated methods for high pressure freezing and correlative
light-electron microscopy. PloS One, 5, e9014.
Storrie, B., White, J., Rottger, S., Stelzer, E. H., Suganuma, T., Nilsson, T., et al. (1998). Recycling of golgi-
resident glycosyltransferases through the ER reveals a novel pathway and provides an explanation for
nocodazole-induced Golgi scattering. The Journal of Cell Biology, 143, 1505–1521.
Tsien, R. Y. (1998). The green fluorescent protein. Annual Review of Biochemistry, 67, 509–544.
von Bartheld, C. S., Cunningham, D. E., & Rubel, E. W. (1990). Neuronal tracing with DiI: decalcification,
cryosectioning, and photoconversion for light and electron microscopic analysis. The Journal of Histo-
chemistry and Cytochemistry, 38, 725–733.
Zeuschner, D., Geerts, W. J., van Donselaar, E., Humbel, B. M., Slot, J. W., Koster, A. J., et al. (2006). Immuno-
electron tomography of ER exit sites reveals the existence of free COPII-coated transport carriers. Nature
Ziese, U., Janssen, A. H., Murk, J. L., Geerts, W. J., Van der Krift, T., Verkleij, A. J., et al. (2002). Automated
high-throughput electron tomography by pre-calibration of image shifts. Journal of Microscopy, 205,
187–200.
CHAPTER 8
Picking Faces out of a Crowd: Genetic

Labels for Identification of Proteins in
Correlated Light and Electron Microscopy
Imaging
Mark H. Ellisman*,†, Thomas J. Deerinck*, Xiaokun Shu‡ and
Gina E. Sosinsky*,†
* National
Center for Microscopy and Imaging Research, University of California, San Diego, La Jolla, CA
92093-0608, USA
† Department of Neurosciences, University of California, San Diego, La Jolla, CA 92093-0608, USA
‡ Department of Pharmaceutical Chemistry, University of California, San Francisco, CA 92093-0608, USA
Abstract
I. Introduction
II. The Crowded Cell and Spatiotemporal Proteomics
III. What EM Has to Offer
IV. Immunomarkers
V. Genetically Appended or Inserted Protein Tags
VI. Types of Genetic Tags Currently Available
A. GFP
B. Tetracysteine Domain/ReAsH Ligand
C. Other Genetic Tags That Covalently Bind Exogenous Fluorescent Ligands
D. MiniSOG: A Genetic Tag That Tightly Binds Endogenous Flavin Mononucleotide
E. Metal Ligand-Based Labels (Metallothionein and Ferritin)
F. Genetically Appended Peroxidase-Based Labels
VII. Future Directions and Challenges
Acknowledgments
References

Copyright 2012, Elsevier Inc. All rights reserved. 139 http://dx.doi.org/10.1016/B978-0-12-416026-2.00008-X
140 Mark H. Ellisman et al.
Abstract
Correlated light and electron microscopic (CLEM) imaging is a powerful method

for dissecting cell and tissue function at high resolution. Each imaging mode provides
unique information, and the combination of the two can contribute to a better under-
standing of the spatiotemporal patterns of protein expression, trafficking, and function.
Critical to these methods is the use of genetically appended tags that highlight specific
proteins of interest in order to be able to pick them out of their complex cellular envi-
ronment. Here we review and discuss the current generation of genetic labels for direct
protein identification by CLEM, addressing their relative strengths and weaknesses and
in what experiments they would be most useful.
I. Introduction
In 2008, the Nobel Prize for Chemistry was awarded to Osamu Shimomura, Martin
Chalfie, and Roger Y. Tsien for the discovery and development of the jellyfish protein
green fluorescent protein (GFP) as an imaging tool for light microscopy. The devel-
opment of GFP, similarly constructed 11-stranded β barrel proteins such as the coral
protein DsRed-derived monomeric RFP and their spectral derivatives, enabled a revo-
lution in live-cell light microscopic imaging as well as fluorescence microscopy of
fixed preparations. Using these genetically appended tags, individual proteins could be
identified and tracked within a cell as well as expressed as soluble fluorescent proteins
(FPs) that can be used to delineate a particular subset of cells. Particularly in neuro-
nal tissues, soluble FPs enable researchers to delineate and track fine neuronal pro-
cesses in an otherwise highly complex milieu. Soluble FPs can also be used to denote
the transfection event of a second tagged protein within the same cell. In addition,
the development of photo-activated fluorescent proteins (FP) has made possible the
development of a subset of super-resolution light microscopy (LM) techniques such as
PALM (Henriques, Griffiths, Hesper Rego, & Mhlanga, 2011; Leung & Chou, 2011).
There has been much interest within the imaging field to build or extend these kinds of
genetically appended light microscopy (LM) labels to protein detection in the electron
microscope (EM) with the advantage of superior resolution and a different modality of
imaging based on mass density.
II. The Crowded Cell and Spatiotemporal Proteomics
The human genome contains 23,000 genes that give rise to larger numbers of active
proteins. Adding to this number are different isoforms of these proteins with alterna-
tive splicing and posttranslational modifications (Andersen & Mann, 2006). Organelles
such as the nuclear pore complex contain many proteins (Rout et al., 2000) while the
cytoplasm is a highly complex and organized matrix. The cytosol contains components
mostly in a nonequilibrium system due to the multiple levels of organization of protein
8. Genetic Tags for Proteins Imaged by Correlated LM and EM 141
in it (Fulton, 1982). Viscosity measurements of the cytosol are similar to pure water;
however, the diffusion of small molecules is about fourfold slower than pure water solu-
tions (Verkman, 2002). In the axon, the protein content has been estimated to be ∼ 2%
of the weight of the axoplasm, while in the oocytes the cytoplasm is between 30% and
40% protein by weight, excluding the protein of the yolk. Muscle cells contain approxi-
mately 23% protein by weight; red blood cells contain about 35% protein by weight;
and in general, actively growing cells contain between 17% and 26% protein by weight
(Fulton, 1982). These measurements are comparable to the protein content of crystals
used for structure determination (Vergara, Lorber, Sauter, Giege, & Zagari, 2005).
The field of “spatial cell biology” (Hurtley, 2009) has been defined as the study of the
spatiotemporal distribution of cellular components. In particular, the location of a cell
within an organism and the location within the cell of its constituent parts affect the func-
tions cells, tissues, and organisms perform, molecular signaling partners, growth, and
division. The establishment and maintenance of cellular and organismal order dictates that
cell components such as proteins “know their place.” In particular, the regulatory machin-
ery of the cell must be organized in a highly sophisticated manner, both spatially as well
as temporally, ensuring that, for example, signaling enzymes are able to correctly encoun-
ter their intracellular substrates. These processes involve the coordinated movement of
molecules and complexes through the crowded cytoplasm to their correct positions. In
order to observe this elegant molecular ballet, the development of highly specific imaging
probes is crucial to identifying them from among the molecular and organellar “crowd.”
III. What EM Has to Offer
Microscopy has the ability to examine single cells among many and their inter-
nal molecular organization. Fluorescence imaging has proved to be very powerful
because probes such as FPs, fluorescent antibodies, and small probes (e.g., phalloidin
staining for actin or DAPI for DNA) provide a signal from only specific proteins or
nucleic acids. FPs, in particular, have been of tremendous value for live-cell imaging
of dynamic events. However, interacting components and organelles must be stained
as well in order to visualize multiple interactions. While the resolution of LM imaging
has increased due to the development of super-resolution techniques, LM can still only
provide information related to a labeled molecule or set of molecules. Even with super-
resolution techniques, electron microscopy still has a minimum ∼5-fold improvement
in practical resolution, and transmission electron microscopy has the benefit of being
able to view other cellular components when there is sufficient mass density. However,
because of the limitation of putting biological samples into a vacuum, EM imaging is
currently limited to biological specimens fixed or frozen at a specific time.
Techniques have been developed to combine light and electron microscopy on the
same specimens in order to exploit the advantages of both imaging methods. This has
been referred to as CLEM, for correlated light and electron microscopy (the same area
is examined by both imaging techniques) or correlative light and electron microscopy
(the same specimen, but not the same area, is imaged by LM and EM). The latter is
typically less time-consuming because less searching is required. The combination

of LM and EM has proved to be very powerful since it combines the specificity and
dynamics of the fluorescent imaging probes using LM with the increased resolution
and cellular context made available by EM techniques. Specialized probes, mostly for
proteins, have been developed for CLEM and have the characteristics of exhibiting
fluorescence for LM observations, and either have or can be used to generate mass
density contrast during the EM specimen preparation process.
IV. Immunomarkers
Antibody labeling techniques remain an extremely important tool for the subcellular
localization of proteins. These probes have the advantage of high specificity and signal
strength when used in combination with secondary antibodies with fluorescent tags, col-
loidal gold markers, immunoenzymatic methods, or eosin-based fluorescence photo-
oxidation (Sosinsky, Giepmans, Deerinck, Gaietta, & Ellisman, 2007). The primary
disadvantage to preembedding procedures for immunolabeling is that the milder fixation
conditions and detergent permeabilization incubation necessary to allow antibodies to
enter a cell or tissue result in a greatly compromised cellular ultrastructure. In particular,
for proteins embedded in interior membrane compartments or soluble within the cyto-
plasm, structures are often compromised or soluble proteins lost during incubation steps.
Some of this damage can be partially mitigated by the addition of small amounts of glu-
taraldehyde (0.01–1%) in a standard 2–4% paraformaldehyde fixation. However, there
is typically a trade-off in the accessibility and conformation of epitopes to these immu-
noprobes, often resulting in a reduced LM and EM signal. CLEM using the genetic tags
described below has the potential for significantly improved preservation of ultrastructure.
Since no permeabilization step is necessary and because the label is intrinsically incorpo-
rated into the sample, strong fixatives such as 1–2% glutaraldehyde can be used. However,
it should be noted that in certain cases, such as when examining proteins that adopt unique
conformations during part of their life cycle, conformation-dependent antibodies are nec-
essary. For example, phospho-specific antibodies are still the only way to discriminate
between different phospho-forms of proteins within the cell (Sosinsky, Solan, et al., 2007).
For CLEM imaging, secondary antibodies must be fluorescent and contain an elec-
tron-dense label. Two types of CLEM secondary antibodies are currently commercially
available. These are quantum dot–conjugated secondary antibodies (Nisman, Dellaire,
Ren, Li, & Bazett-Jones, 2004) and FluoroNanogold™, a secondary antibody that has
both a fluorescent moiety and a 1.8 nm gold cluster. Quantum dots have strong and
stable fluorescence and unique geometries for multi-labeling in electron microscopy
(Giepmans, Deerinck, Smarr, Jones, & Ellisman, 2005) (Fig. 1(A–C)); however, quan-
tum dot–conjugated secondary antibodies are about the same size as small colloidal
gold beads so that epitope accessibility can still be problematic. In addition, quantum
dots are less dense than gold spheres and thus, can be difficult to find within a dense
cytoplasm. FluoroNanogold™ has the advantages of being fluorescent and having a
smaller gold particle for better penetration (Takizawa & Robinson, 2000; Takizawa,
Fig. 1 Comparison of antibody particle labels versus antibody fluorescent photo-oxidation labeling.
(A) Schematic of indirect (secondary) antibody labeling for correlative light and electron microscopy where
the secondary antibody is conjugated to a gold bead or quantum dot. (B) Electron micrograph of microtubules
(arrows) labeled with immuno-quantum dots. The primary antibody is an α-tubulin monoclonal antibody.
The fluorescence image from the quantum dots is shown in the inset. (C) Higher magnification of microtu-
bule. (D) Schematic of photo-oxidized eosin-immunolabeling. The gray ellipse represents the layer of DAB
precipitate after photo-oxidation of the eosin (green sun) on the secondary antibody. Eosin is a brominated
version of the common fluorophore fluorescein, and the chemical formula of the isothiocyanate used for con-
jugation is shown in the upper right hand corner. (E) Electron micrograph of microtubules (arrows) labeled
with anti-α tubulin primary antibodies, eosin secondary antibodies and then photo-oxidized in the presence
of DAB and oxygen. An eosin fluorescence image is shown in the inset. (F) Higher magnification of a single
microtubule. The difference in ultrastructure between B and E is primary due to the less stringent fixation
methods used so the larger quantum dot secondary antibodies could penetrate into the cell.
Suzuki, & Robinson, 1998); however, it requires a gold or silver enhancement step
to grow the gold cluster immunolabeling in order to easily detect the particle in EM
images. Fluorescence photo-oxidation (see Section V ) can also be used as a detection
method using eosin conjugated to secondary antibodies or small ligands (Capani et al.,
2001; Deerinck et al., 1994) (Fig. 1(D–F)).
V. Genetically Appended or Inserted Protein Tags
The key to using genetic tags for the EM portion of a CLEM label is the incorpo-
ration of diaminobenzidine (DAB) as part of the labeling protocol. DAB has a long
history as a reagent for correlated light and electron microscopy specimen preparation
methods. DAB is used as a substrate for enzymatic-based polymerization using peroxi-

dase- or singlet oxygen–based polymerization during photo-oxidation (Deerinck et al.,
1994; Maranto, 1982). When oxidized, DAB forms a highly insoluble reaction product
that can be made electron-dense through treatment with osmium tetroxide. Although
peroxidase-based systems generally have limited resolution due to diffusion of reaction
product as compared to photo-oxidation, enzymatic systems have greater sensitivity.
DAB-based reaction products are excellent for filling cells for correlated microscopy,
either through enzymatic reactions or through photo-oxidation (Li, Wang, Chiu, &
Cline, 2010; Maranto, 1982). For molecular localization using photo-oxidation, we
have shown that fluorescence photo-oxidation of DAB provides superb resolution, and
diffusion of the reaction product is minimized by extensive chemical cross-linking by
glutaraldehyde prior to the generation of the reaction product. Unlike particulate mark-
ers such as protein-conjugated colloidal gold, photo-oxidation-based reaction products
do not just decorate the targeted protein but also act almost as a negative stain, in some
cases, providing exquisite structural detail (Shu et al., 2011). Staining occurs in close
proximity to the molecule rather than being separated by many nanometers, as in anti-
body methods where the secondary antibody is covalently linked to a detection agent
such as a colloidal gold bead, quantum dot, or eosin and is separated from the target by
length of the primary antibody.
To be useful as a CLEM label, there are certain requirements when using genetic
tags. Genetic tags should be nontoxic when expressed in a cellular environment,
have good sensitivity in the physiological range of the experiment and be active in
the required physiological environment (including pH, temperature, aqueous solution,
enzymatic environment, and ionic concentrations), exhibit fluorescence, and either
independently or with the addition of additional ligands or cofactors create an electron-
dense label. Genetic tags have the additional advantage that they are stoichiometric to
their target protein. This quality can be important for low copy number proteins within
cells. Table I summarizes the properties of several CLEM genetic labels. As with any
exogenously expressed protein, it is critical that independent control experiments be
done to ensure that recombinant proteins be expressed and trafficked in the same man-
ner as endogenously expressed proteins. Often, these controls mean a comparison to
immunolabeled native systems.
VI. Types of Genetic Tags Currently Available
A. GFP
Historically, GFP was one of the first protein tags to be used for CLEM (Mono-
sov, Wenzel, Luers, Heyman, & Subramani, 1996). While the shielding of its
chromophore helps to make GFP extremely bright, photostable, and nontoxic,
it also prevents GFP from producing enough singlet oxygen necessary for good
DAB polymerization. GFP has been used as a CLEM tag in certain cases where
there is high concentration of GFP-tagged expressed protein (Grabenbauer et al.,
2005) that in special cases provides sufficient singlet oxygen generation, even
Table I
8. Genetic Tags for Proteins Imaged by Correlated LM and EM

Comparison of genetic tags for CLEM.
Genetic tag MW (kDa) FQY 1O

2 QY Advantages Disadvantages References
GFP 26.9 0.6 < 0.004 Bright fluorescence, well- Poor photo-oxidizer (Grabenbauer et al.,
established technology Very low reactive oxygen yield 2005)
Tetracysteine/ 2.22 0.42–0.471 0.024 Small size, versatility of ligands Low reactive oxygen yield (Gaietta et al., 2002,
ReAsH especially for pulse-chase Limited diffusion and nonspecific 2006; Griffin et al.,
(present 4C or labeling at LM and EM levels background of ligands 1998; Martin et al.,
4N generation). 2005)
miniSOG/ FMN 15.4 0.37 0.47 Medium size between GFP an Weak fluorescence (Shu et al., 2011)
tetracysteine Bleaches rapidly
Very good singlet oxygen
generator; FMN is an
endogenous ligand. Strong
EM signal after photo-
oxidation.
HRP 34 NA NA Enzymatic reaction. Needs to be combined with (Li et al., 2010;
Labeling limited to expressing fluorescent protein (GFP) for LM Schikorski, 2010;
cells. No photo-oxidation Does not work in oxidized Schikorski et al.,
necessary. High sensitivity. environments 2007)
Limited so far to plasma membrane
or ER targeting
Metallothionein 6 NA NA When gold atoms bind, delivers Needs to be combined with (Diestra, Cayrol, et al.,
(MTH) high contrast. High spatial fluorescent protein (GFP) for LM 2009; Diestra,
resolution. Relatively small Difficult to get gold atoms into cells Fontana, et al.,
tag size. Toxic 2009; Mercogliano
& DeRosier, 2006,
2007; Nishino et al.,
2007)
Ferritin 19.42 NA NA Iron less toxic. Strong signal Needs to be combined with (Wang et al., 2011)
from iron cores. High spatial fluorescent protein (GFP) for LM
resolution. Large size of label when
oligomerized
Need to get exogenous iron atoms
into cells for labeling
FQY = fluorescence quantum yield; 1O2 QY = singlet oxygen quantum yield; NA = not applicable
1 Measurements done using FRET-based emission of GFP-TC proteins.
2 monomer size.
145
though the level of GFP singlet oxygen quantum yield is close to zero (quantum yield
<0.004) (Jimenez-Banzo, Nonell, Hofkens, & Flors, 2008).
B. Tetracysteine Domain/ReAsH Ligand

In 1998, Albert Griffin, Stephen Adams, and Roger Tsien demonstrated that recom-
binant proteins in intact cells can be imaged by genetically appending or inserting
a small tetracysteine motif, -Cys-Cys-Xaa-Xaa-Cys-Cys- and then exposing the cells
to a membrane-permeant nonfluorescent biarsenical derivative of fluorescein, termed
FlAsH (Griffin, Adams, & Tsien, 1998; Griffin, Adams, Jones, & Tsien, 2000). The
uniquely designed stereochemistry of the tetracysteine domain binds small mole-
cules containing two arsenic atoms with a specific distance that coordinate with the
SH groups of the Cys residues. FlAsH binds with high affinity and specificity to the
tetracysteine motif and becomes strongly green fluorescent. Toxicity from binding of
the trivalent arsenic atoms to endogenous thiols is mitigated by simultaneous admin-
istration of micromolar concentrations of arsenic antidotes such as 1,2-ethanedithiol
(EDT) or British anti-Lewisite (BAL; dimercaprol; 2,3-dimercaptopropanol). Upon
removal of excess FlAsH, the fluorescent complexes survive for days in the absence
of excessive (mM) concentrations of competing EDT. We engineered a recombinant
connexin43 (Cx43) by genetically fusing a 17 amino-acid-long tetracysteine receptor
domain (EAAAREACCRECCARA) to the COOH terminus of Cx43 (Gaietta et al.,
2002) and showed localization to gap junctions and internal trafficking structures
including vesicles, lysosomes, as well as portions of the ER and the Golgi apparatus.
One of the advantages of the tetracysteine tag/biarsenical ligand labeling approach
is that one can employ multiple ligands with specialized properties for different imag-
ing purposes, and since its introduction, multiple ligands have been successfully syn-
thesized. One such ligand, termed ReAsH, a red fluorescent biarsenical derivative of
resorufin, was found to be a modest singlet oxygen generator suitable for fluorescence
photo-oxidation of DAB (Adams et al., 2002). Using a directed evolution approach,
the tetracysteine domain sequence was optimized (Martin, Giepmans, Adams, & Tsien,
2005) to greatly increase the ReAsH affinity over the original peptide. This peptide has
the sequence FLNCCPGCCMEP and can be appended to either the N or C terminus
(referred to as a 4N or 4C tag, respectively). While an improvement, the absolute con-
trast value of the bound ReAsH fluorescence was still lower than that of GFP. How-
ever, this enhanced tetracysteine domain has the advantages of its small size (typically
between 14 and 16 amino acids)––the ability to use ligands with different spectral
properties to achieve an optical pulse chase labeling of new and old proteins and its
improved EM contrast due to the higher signal-to-noise ratio.
Because of their small size, tetracysteine tags can be incorporated more easily into
proteins. Typically, tags are appended to the protein either at the N or C terminus (see
actin example in Fig. 2); however, in many instances, tetracysteine domains have been
incorporated into internal sites within a protein without any disruption of function or
trafficking (Boassa et al., 2010; Hoffmann et al., 2005; Lanman et al., 2008; Ziegler,
Batz, Zabel, Lohse, & Hoffmann, 2011). In particular, viral function and infectivity can
Fig. 2 Genetic tetracysteine domain and ReAsH provide good labeling and ultrastructure of specific proteins. (A) Schematic of fluo-
rescence photo-oxidation for correlative LM and EM (Reprinted from Sosinsky et al. (2003)). (B) Confocal fluorescence image (inset)
and thin section electron microscopy of the same two adjacent cells, one transfected with an N-terminal appended tetracysteine domain.
Labeling is obvious in the transfected cell (top cell) as opposed to the untransfected cell (bottom cell). Arrows denote labeled stress fiber
containing actin filaments, versus an unlabeled bundle (arrowhead). (C) Slice from an EM tomogram of stress bundles containing tetra-
cysteine/ReAsH β actin. Arrows point to individual actin filaments. (D) Tracing of actin filaments in the tomogram. The outlines of the
actin bundles are shown in brown. Cyan filaments are those actin filaments separated from the actin bundle while the yellow filaments
are those that were easily traced within the bundle. (See color plate.)
be maintained when small tetracysteine tags are incorporated into internal positions
(Lanman et al., 2008; Rudner et al., 2005; Whitt & Mire, 2011).
Later improvements to the tetracysteine tagging system included incorporating it
into GFP. The tetracysteine tag can be appended to either the N or C terminus of GFP
for subsequent fusion of the complex to either the N or C terminus of the target pro-
tein. Fusing this newer tetracysteine to GFP and exciting ReAsH indirectly via fluo-
rescence resonance energy transfer from the fluorescent protein (Gaietta et al., 2006)
can increase the contrast of the 4C/ReAsH to within a factor of two of GFP. This novel
combinatorial tag has several advantages for dynamic imaging, in that it combines the
properties of GFP (brightness, high specificity, and ease of use) with small tetracysteine
tags (pulse-chase, photo-oxidation, excellent preservation for EM). It is worth noting
that both GFP and tetracysteines do not fluoresce well in oxidative compartments such
as the Golgi or ER, but this effect can be ameliorated by carrying out the biarsenical
labeling step during a short incubation that includes triethylphosphine (TEP), a mem-
brane permeant phosphine (Gaietta et al., 2006).
C. Other Genetic Tags That Covalently Bind Exogenous Fluorescent Ligands

A strategy similar to tetracysteine/biarsenical labeling whereby an exogenous ligand
covalently binds to a genetically engineered protein domain is the SNAP-tag system.
Covalent labeling with a small molecule is achieved through the mammalian O6-
alkylguanine-DNA-alkyltransferase (hAGT) that irreversibly transfers the alkyl group
from its substrate, O6-alkylguanine-DNA, to one of its cysteine residues. These ligands
have guanine or chloropyrimidine leaving groups via a benzyl linker that bind to hAGT
(Keppler et al., 2003). Dyes conjugated to benzyl linkers facilitate fluorescence label-
ing. A mutant version of hAGT that binds benzocysteine provides an orthogonal label-
ing system. The commercial versions are marketed as SNAP and CLIP (New England
Biolabs). The size of the SNAP and CLIP protein tags are ∼20 kDa. The commercially
available HALOtag™ is 33 kDa enzyme derivitized from a modified prokaryotic deha-
logenase (DhaA), which makes stable bonds that can be readily made with synthetic
molecules appended to a chloroalkane linker (Los et al., 2008) (Promega Corporation).
While there are no reports of ligands currently available that would allow EM obser-
vation, there is a potential for having an orthogonal set of genetic labels with these
reagents.
D. MiniSOG: A Genetic Tag That Tightly Binds Endogenous Flavin Mononucleotide

Addition of an exogenous ligand is a requirement for fluorescence imaging and
DAB photo-oxidation with the tetracysteine/biarsenical labeling system. While this is
typically not a problem in tissue culture cells (Griffin, Adams, & Tsien, 1998; Gaietta
et al., 2002) or transfected primary cells in culture (Ju et al., 2004), the labeling with
exogenous ligands in intact tissues can be highly problematic. In particular, diffusion
of biarsenicals across the blood–brain barrier or through large expanses of tissue to
their correct targets presents a challenge. In addition, tetracysteine/biarsenical labeling
requires the co-administration of an antidote such as ethanedithiol to prevent cellular

toxicity and needs careful precautions to mitigate nonspecific background labeling.
Our efforts to create other new and novel genetic tags were in part driven by the need
to develop protein tags that did not require the application of an endogenous fluorescent
ligand. Thus, when the tagged protein is expressed within the cellular environment, the
recombinant protein would become fluorescent and capable of generating a reaction
product upon light stimulation without any further treatments.
Using a completely new approach, Shu et al. (2011) developed a new genetically
encoded tag that has the advantages of a tag size significantly smaller than conven-
tional fluorescent proteins, exhibits intrinsic fluorescence without the need for an
exogenously applied fluorescent ligand, has low cellular toxicity, and exhibits a sub-
stantially higher singlet oxygen quantum yield than ReAsH (Shu et al., 2011) (Fig. 3).
MiniSOG (for mini Singlet Oxygen Generator) was engineered from the LOV2 (light,
oxygen, and voltage) domain of the protein phototropin 2, a blue light plant photore-
ceptor from Arabidopsis thaliana that binds flavin mononucleotide (FMN), a highly
efficient singlet oxygen photosensitizer. FMN is ubiquitous in cells and is involved
in biological functions such as mitochondrial electron transport, fatty acid oxida-
tion, and vitamin metabolism. In phototropin 2, the excited state energy of FMN is
consumed to form a covalent bond with a cysteine. To divert this energy into fluores-
cence and singlet oxygen generation, saturation mutagenesis of regions surrounding
the chromophore of the LOV2 domain of phototropin 2 was performed. The current
version of miniSOG, at 10.6 kDa, is less than half the size of GFP and has two emis-
sion peaks at 500 and 528 nm and a singlet oxygen quantum yield of 0.47, an almost
20× improvement over ReAsH. For comparison, eosin has a singlet oxygen quantum
yield of 0.57; however, its fluorescence yield is only about 0.05 (10–20× less than
fluorescein). Furthermore, in the absence of light, miniSOG causes no discernable
cellular toxicity.
The fluorescence from miniSOG can be used to successfully localize a wide variety
of proteins and organelles in cultured mammalian cells in much the same way as can
be done with GFP. Its green fluorescence, while modest compared to GFP (fluores-
cence quantum yield of 0.37 vs. 0.6), revealed that the fusion proteins appeared to
have correct localizations. These targets included the ER and Golgi apparatus (using
signal sequence localizations), Rab5a, zyxin, tubulin, β-actin, and α-actinin fusions
as examples of proteins tagged in cytosolic compartments (see α-actinin examples in
Fig. 3). Mitochondrial targeting and nuclear histone 2B-fusions show that miniSOG
expresses within those organelles. In primary neurons, SynCAM–miniSOG fusions
localized to the synapse. Using the fluorescence and photo-generated singlet oxygen
from miniSOG for fluorescence photo-oxidation of DAB, correlated confocal and EM
imaging could be performed with many miniSOG fusion proteins with high sensitiv-
ity. Shu et al. (2011) also showed that miniSOG was functional in transgenic animals,
showing EM localization of labeled mitochondria in the muscle walls of C. elegans
transfected with miniSOG appended to a cytochrome C targeting motif as well as at
synapses labeled with SynCAM-1-miniSOG or SynCAM-2-miniSOG in mice brain
slices where the mice were transfected by in utero electroporation.
Fig. 3 Fluorescence photo-oxidation of mini-SOG labeled α-actinin. (A) Richardson diagram of miniSOG
with FMN (left). The photo-oxidation process is shown in the right-hand schematic. (B–E) miniSOG was
genetically appended to the C-terminus of α-actinin. (B) Confocal microscope image of miniSOG fluorescence
highlight actin bundles (C) Transmitted light micrograph of the same area after photo-oxidation. (D) Thin sec-
tion EM of the same area. (E) Higher magnification of the actin bundle shows DAB precipitate surrounding
labeled α-actinin. (Figure reproduced from Shu et al. (2011)).
E. Metal Ligand-Based Labels (Metallothionein and Ferritin)

Theoretically, the binding of metal clusters to proteins could act as markers for
EM observations because these metal clusters contain elements that sufficiently scat-
ter electrons. In order for these tags to be useful for CLEM, these protein tags could
be appended to FPs. Two protein tags that have been shown to be useful for metal
deposition for EM imaging are concatenated metallothionein (MTH) (Mercogliano &

DeRosier, 2006, 2007) and bacterial ferritin. Both use the principle that additions of
exogenous metal atoms when bound to MTH or ferritin cluster enough atoms for EM
detection.
MTH facilitates direct gold labeling of the fusion protein via a reaction with auro-
thiomalate to its MTH moiety. Mercogliano and DeRosier (2006) first demonstrated
that purified MTH proteins can form gold clusters that can be imaged by EM. MTH-
based labels were originally developed as an electron-dense label to be used in identifi-
cation of protein domains in single particle reconstruction of isolated macromolecular
complexes (Mercogliano & DeRosier, 2007). MTH binds ∼12–20 gold atoms per copy,
approximately as much as is found in the EM reagent, Nanogold® (Nanoprobes Inc.,
Yaphank, NY) that has a 1.4 nm diameter. One benefit is that at ∼6 kDa, the present
concatenated MTH is a relatively small tag and can be used as a tandem dimer to
increase signal, still leaving the tag at about the same size as miniSOG (Mercogliano
& DeRosier, 2007). There are a few reports of MTH as molecular labels in cells for
electron microscopic imaging using either gold or cadmium ions to grow clusters to
chimeric proteins expressed in Escherichia coli or Cos7 mammalian cells (Diestra,
Cayrol, Arluison, & Risco, 2009; Diestra, Fontana, Guichard, Marco, & Risco, 2009;
Fukunaga et al., 2007; Nishino, Yasunaga, & Miyazawa, 2007). The major drawback
to this tagging system is that exogenous metals must be added to and taken up by cells
and these metal solutions can be toxic to the cell. As a result, most cellular studies have
used bacteria that are more tolerant of higher concentrations of heavy metals. Zinc
clusters, which are better tolerated, can also create an electron-dense label (Bouchet-
Marquis, Pagratis, Kirmse, & Hoenger, 2011).
An additional and recently developed heavy metal binding protein probe is based
on the bacterial iron binding ferritin complex that is assembled from E. coli FtnA pro-
tein (19.4 kDa) (Wang, Mercogliano, & Lowe, 2011). Twenty-four FtnA monomers
form a shell with a 7.5 nm central cavity. The center of the ferritin complex becomes
loaded with iron when cells grow under iron-rich conditions and the advantage is its
high imaging contrast and low toxicity. However, this is a relatively large structure of
∼12 nm diameter and because the FtnA needs to oligomerize, a high concentration of
chimeric protein is necessary. Test cases of the bacterial chemoreceptor sensor, CheY
protein, and the septal ring protein ZapA genetically fused to GFP and FtnA showed
detectable metal clusters in the correct cellular locations.
F. Genetically Appended Peroxidase-Based Labels

Horseradish peroxidase has been used extensively in light and electron microscopy
as a reagent for immunocytochemistry (Porstmann & Kiessig, 1992), and there is much
interest in adapting it as a genetic tag because of its excellent enzymatic activity in
creating an electron-dense label. In order for HRP to be functional, it must be glycosyl-
ated (Veitch, 2004). Since HRP is a plant glycoprotein, the secretory signal sequence
of the human growth hormone on HRP was appended onto the HRP cDNA (ssHRP)
to ensure proper trafficking in mammalian cells (Connolly, Futter, Gibson, Hopkins,
& Cutler, 1994). ssHRP was then targeted to ER by adding the KDEL-retention signal
(Norcott, Solari, & Cutler, 1996). Schikorski and coworkers subsequently showed that
this ER marker could be made neuron-specific by modifying the construct so that it was
controlled by the synapsin promoter (Schikorski, Young, & Hu, 2007) and can be used
with transmitted light imaging as a LM probe (Schikorski, 2010). Examples of specific
HRP protein labeling for EM include the HRP-Wingless chimera used to show directed
Wingless trafficking in Drosophila embryos (Dubois, Lecourtois, Alexandre, Hirst, &
Vincent, 2001) and a fusion protein of HRP and type I transmembrane protein, CD2,
that labeled specific populations of gamma neurons in Drosophila (Watts, Schuldiner,
Perrino, Larsen, & Luo, 2004) and HRP-synaptophysin-GFP that highlighted populations
of synaptic vesicles (Ruthazer, Li, & Cline, 2006).
Recombinant HRP has been proposed as a potentially good target for EM cell filling
that would act analogous to soluble cytosolic FPs for LM. The major application of
this technique would be to identify specific cells, such as neurons, and aid segmenting
their processes in three-dimensional reconstructions obtained either through EM serial
section, EM tomography, or serial block face scanning electron microscopy (SBEM).
However, the major drawback to using the current HRP as a probe for cell filling is that
it does not have enzymatic activity in the reduced environment of the cytosol (Li et al.,
2010). A membrane-targeted HRP has recently been developed that allows identifica-
tion of neurons because of the intense staining of their plasma membranes. This protein
contains a plasma membrane signal sequence and transmembrane domain fused to the
C-terminus of HRP. In combination with a bicistronic vector, soluble GFP was also
expressed serving as a fluorescent, cytosolic label, thus facilitating easy identification
and segmentation in 3D EM volumes.
VII. Future Directions and Challenges
Two challenges represent the frontiers for further development of CLEM genetic
tags. The first is the development of specific nucleic acid CLEM probes analogous
to the protein probes described in this review. Markers for specific DNA or RNA
sequences appended to FPs are still lacking. Notably, a recent study developing spec-
tral imaging RNA probes have generated a fluorescent toolkit similar to FPs (Paige,
Wu, & Jaffrey, 2011) and thus, there is a potential that such imaging tools could be
modified for CLEM. The second challenge is expression of EM genetic probes in ani-
mals. Several of the genetic tags discussed above require the administration of exog-
enous ligands for detection of the tagged protein. While ligand-based imaging tags can
be very versatile, getting these exogenous reagents requires diffusion through tissues
and especially across the blood–brain barrier. MiniSOG and peroxidase-based probes
would circumvent this and in combination with strong fluorescing FPs proved an ave-
nue for CLEM imaging of transgenic animals (Dubois et al., 2001; Shu et al., 2011).
Finally, future CLEM improvements will also need to include combining these genetic
labeling protocols with methods that optimally preserve the ultrastructure of cells and
tissues, such as cryo-fixation methods. High pressure freezing and freeze substitution
using hybrid chemical/cryo-fixation protocols show great promise for achieving this
goal (Sosinsky et al., 2008).
Acknowledgments
We thank Dr. Roger Tsien and members of the Tsien laboratory for their inspiration.
We also thank Dr. Ben Giepmans for preparing the tetracysteine-actin constructs shown
in Fig. 2. National Institutes of Health Grants P41RR004050 (MHE), P41GM103412
(MHE), GM086197 (Roger Y. Tsien, PI), GM065937, and GM072881 (GES) provided
funding for this research.
References
Adams, S. R., Campbell, R. E., Gross, L. A., Martin, B. R., Walkup, G. K., Yao, Y., et al. (2002). New biar-
senical ligands and tetracysteine motifs for protein labeling in vitro and in vivo: synthesis and biological
applications. Journal of the American Chemical Society, 124, 6063–6076.
Andersen, J. S., & Mann, M. (2006). Organellar proteomics: turning inventories into insights. EMBO
Reports, 7, 874–879.
Boassa, D., Solan, J. L., Papas, A., Thornton, P., Lampe, P. D., & Sosinsky, G. E. (2010). Trafficking and
recycling of the connexin43 gap junction protein during mitosis. Traffic, 11, 1471–1486.
Bouchet-Marquis, C., Pagratis, M., Kirmse, R., & Hoenger, A. (2012). Metallothionein as a clonable high-
density marker for cryo-electron microscopy. Journal of Structural Biology, 177, 119–127.
Capani, F., Deerinck, T. J., Ellisman, M. H., Bushong, E., Bobik, M., & Martone, M. E. (2001). Phalloidin-
eosin followed by photo-oxidation: a novel method for localizing F-actin at the light and electron micro-
scopic levels. Journal of Histochemistry and Cytochemistry, 49, 1351–1361.
Connolly, C. N., Futter, C. E., Gibson, A., Hopkins, C. R., & Cutler, D. F. (1994). Transport into and out of
the Golgi complex studied by transfecting cells with cDNAs encoding horseradish peroxidase. Journal
tion detection for light and electron microscopy. Journal of Cell Biology, 126, 901–910.
Diestra, E., Cayrol, B., Arluison, V., & Risco, C. (2009). Cellular electron microscopy imaging reveals the
localization of the Hfq protein close to the bacterial membrane. PLoS One, 4, e8301.
Diestra, E., Fontana, J., Guichard, P., Marco, S., & Risco, C. (2009). Visualization of proteins in intact cells
with a clonable tag for electron microscopy. Journal of Structural Biology, 165, 157–168.
Dubois, L., Lecourtois, M., Alexandre, C., Hirst, E., & Vincent, J. P. (2001). Regulated endocytic routing
modulates wingless signaling in Drosophila embryos. Cell, 105, 613–624.
Fukunaga, Y., Hirase, A., Kim, H., Wada, N., Nishino, Y., & Miyazawa, A. (2007). Electron microscopic
analysis of a fusion protein of postsynaptic density-95 and metallothionein in cultured hippocampal neu-
rons. Journal of Electron Microscopy (Tokyo), 56, 119–129.
Fulton, A. B. (1982). How crowded is the cytoplasm? Cell, 30, 345–347.
Gaietta, G. M., Giepmans, B. N., Deerinck, T. J., Smith, W. B., Ngan, L., Llopis, J., et al. (2006). Golgi
microscopy. Proceedings of the National Academy of Sciences of the United States of America, 103,
17777–17782.
electron microscopic imaging of multiple endogenous proteins using Quantum dots. Nature Methods, 2,
743–749.
857–862.
Griffin, B. A., Adams, S. R., & Tsien, R. Y. (1998). Specific covalent labeling of recombinant protein mol-
ecules inside live cells. Science, 281, 269–272.
Griffin, B. A., Adams, S. R., Jones, J., & Tsien, R. Y. (2000). Fluorescent labeling of recombinant proteins in
living cells with FlAsH. Methods in Enzymology, 327, 565–578.
Henriques, R., Griffiths, C., Hesper Rego, E., & Mhlanga, M. M. (2011). PALM and STORM: unlocking
live-cell super-resolution. Biopolymers, 95, 322–331.
Hoffmann, C., Gaietta, G., Bunemann, M., Adams, S. R., Oberdorff-Maass, S., Behr, B., et al. (2005).
A FlAsH-based FRET approach to determine G protein-coupled receptor activation in living cells. Nature
Methods, 2, 171–176.
Hurtley, S. (2009). Spatial cell biology. Location, location, location. Introduction. Science, 326, 1205.
Jimenez-Banzo, A., Nonell, S., Hofkens, J., & Flors, C. (2008). Singlet oxygen photosensitization by EGFP
and its chromophore HBDI. Biophysical Journal, 94, 168–172.
Ju, W., Morishita, W., Tsui, J., Gaietta, G., Deerinck, T. J., Adams, S. R., et al. (2004). Activity-dependent
regulation of dendritic synthesis and trafficking of AMPA receptors. Nature Neuroscience, 7, 244–253.
Keppler, A., Gendreizig, S., Gronemeyer, T., Pick, H., Vogel, H., & Johnsson, K. (2003). A general method
for the covalent labeling of fusion proteins with small molecules in vivo. Nature Biotechnology, 21,
86–89.
Lanman, J., Crum, J., Deerinck, T. J., Gaietta, G. M., Schneemann, A., Sosinsky, G. E., et al. (2008). Visual-
izing flock house virus infection in Drosophila cells with correlated fluorescence and electron micros-
copy. Journal of Structural Biology, 161, 439–446.
Leung, B. O., & Chou, K. C. (2011). Review of super-resolution fluorescence microscopy for biology. Applied
Spectroscopy, 65, 967–980.
Li, J., Wang, Y., Chiu, S. L., & Cline, H. T. (2010). Membrane targeted horseradish peroxidase as a marker
for correlative fluorescence and electron microscopy studies. Front Neural Circuits, 4, 6.
Los, G. V., Encell, L. P., McDougall, M. G., Hartzell, D. D., Karassina, N., Zimprich, C., et al. (2008).
HaloTag: a novel protein labeling technology for cell imaging and protein analysis. ACS Chemical Biol-
ogy, 3, 373–382.
Maranto, A. (1982). Neuronal mapping: a photooxidation reaction makes Lucifer Yellow useful for electron
microscopy. Science, 217, 953–955.
Martin, B. R., Giepmans, B. N., Adams, S. R., & Tsien, R. Y. (2005). Mammalian cell-based optimization of
the biarsenical-binding tetracysteine motif for improved fluorescence and affinity. Nature Biotechnology,
23, 1308–1314.
Mercogliano, C. P., & DeRosier, D. J. (2006). Gold nanocluster formation using metallothionein: mass spec-
trometry and electron microscopy. Journal of Molecular Biology, 355, 211–223.
Mercogliano, C. P., & DeRosier, D. J. (2007). Concatenated metallothionein as a clonable gold label for
electron microscopy. Journal of Structural Biology, 160, 70–82.
Monosov, E. Z., Wenzel, T. J., Luers, G. H., Heyman, J. A., & Subramani, S. (1996). Labeling of per-
oxisomes with green fluorescent protein in living P. pastoris cells. The Journal of Histochemistry and
Nishino, Y., Yasunaga, T., & Miyazawa, A. (2007). A genetically encoded metallothionein tag enabling
efficient protein detection by electron microscopy. Journal of Electron Microscopy (Tokyo), 56, 93–101.
Norcott, J. P., Solari, R., & Cutler, D. F. (1996). Targeting of P-selectin to two regulated secretory organelles
in PC12 cells. Journal of Cell Biology, 134, 1229–1240.
Paige, J. S., Wu, K. Y., & Jaffrey, S. R. (2011). RNA mimics of green fluorescent protein. Science, 333,
642–646.
Porstmann, T., & Kiessig, S. T. (1992). Enzyme immunoassay techniques. An overview. Journal of Immu-
nological Methods, 150, 5–21.
Rout, M. P., Aitchison, J. D., Suprapto, A., Hjertaas, K., Zhao, Y., & Chait, B. T. (2000). The yeast nuclear
pore complex: composition, architecture, and transport mechanism. Journal of Cell Biology, 148,
635–651.
Rudner, L., Nydegger, S., Coren, L. V., Nagashima, K., Thali, M., & Ott, D. E. (2005). Dynamic fluorescent
imaging of human immunodeficiency virus type 1 gag in live cells by biarsenical labeling. Journal of
Virology, 79, 4055–4065.
Ruthazer, E. S., Li, J., & Cline, H. T. (2006). Stabilization of axon branch dynamics by synaptic maturation.
Journal of Neuroscience, 26, 3594–3603.
Schikorski, T., Young, S. M., Jr., & Hu, Y. (2007). Horseradish peroxidase cDNA as a marker for electron
microscopy in neurons. Journal of Neuroscience Methods, 165, 210–215.
Schikorski, T. (2010). Horseradish peroxidase as a reporter gene and as a cell-organelle-specific marker in
correlative light-electron microscopy. Methods in Molecular Biology, 657, 315–327.
Biology, 9, e1001041.
Sosinsky, G. E., Gaietta, G. M., Hand, G., Deerinck, T. J., Han, A., Mackey, M., et al. (2003). Tetracysteine
genetic tags complexed with biarsenical ligands as a tool for investigating gap junction structure and
dynamics. Cell Communication & Adhesion, 10, 181–186.
Sosinsky, G. E., Solan, J. L., Gaietta, G. M., Ngan, L., Lee, G. J., Mackey, M. R., et al. (2007). The C-terminus
of connexin43 adopts different conformations in the Golgi and gap junction as detected with structure-
specific antibodies. Biochemical Journal, 408, 375–385.
Sosinsky, G. E., Crum, J., Jones, Y. Z., Lanman, J., Smarr, B., Terada, M., et al. (2008). The combination of
chemical fixation procedures with high pressure freezing and freeze substitution preserves highly labile
tissue ultrastructure for electron tomography applications. Journal of Structural Biology, 161, 359–371.
Takizawa, T., & Robinson, J. M. (2000). FluoroNanogold is a bifunctional immunoprobe for correlative
fluorescence and electron microscopy. The Journal of Histochemistry and Cytochemistry, 48, 481–486.
Veitch, N. C. (2004). Horseradish peroxidase: a modern view of a classic enzyme. Phytochemistry, 65,
249–259.
Vergara, A., Lorber, B., Sauter, C., Giege, R., & Zagari, A. (2005). Lessons from crystals grown in the
Advanced Protein Crystallisation Facility for conventional crystallisation applied to structural biology.
Biophysical Chemistry, 118, 102–112.
Verkman, A. S. (2002). Solute and macromolecule diffusion in cellular aqueous compartments. Trends in
Biochemical Sciences, 27, 27–33.
Wang, Q., Mercogliano, C. P., & Lowe, J. (2011). A ferritin-based label for cellular electron cryotomogra-
phy. Structure, 19, 147–154.
Watts, R. J., Schuldiner, O., Perrino, J., Larsen, C., & Luo, L. (2004). Glia engulf degenerating axons during
developmental axon pruning. Current Biology, 14, 678–684.
Whitt, M. A., & Mire, C. E. (2011). Utilization of fluorescently-labeled tetracysteine-tagged proteins to study
virus entry by live cell microscopy. Methods, 55, 127–136.
Ziegler, N., Batz, J., Zabel, U., Lohse, M. J., & Hoffmann, C. (2011). FRET-based sensors for the human
M1-, M3-, and M5-acetylcholine receptors. Bioorganic & Medicinal Chemistry, 19, 1048–1054.
CHAPTER 9
Correlated Light Microscopy and Electron

Microscopy
Klaas A. Sjollema, Ulrike Schnell, Jeroen Kuipers, Ruby
Kalicharan and Ben N.G. Giepmans
Department of Cell Biology, University Medical Center Groningen (UMCG), University of Groningen,
A. Deusinglaan 1, Bldg 3215, room 749, 9713 AV Groningen, The Netherlands
Abstract
Abbreviations
I. Introduction
II. Correlated Light and Electron Microscopy: Using the Best of Two Worlds
A. Conventional Use with Transmission LM
B. Fluorescence LM and CLEM: Exploration and Exploitation
III. ROIs: Search & Find Tools
A. Large-Scale LM and EM
B. Marks
C. Photoconversion and CLEM
D. Immunolabeling and Combinatorial Probes and CLEM
E. Live-Cell Imaging and CLEM
F. Super Resolution Microscopy and CLEM
G. Correlation of LM and EM Using LM Image of Block Face
H. Stage Holders for CLEM and Integrated Microscopes
IV. Our Approach:Virtual Reality Overlay during Preparation
A. Aim
B. Requirements (See also Fig. 1)
C. Principle
D. Workflow (See also Figs. 2 and 3)
V. Discussion and Conclusion
VI. Future Perspective
Acknowledgments
References
Abstract
Understanding where, when, and how biomolecules (inter)act is crucial to uncover

fundamental mechanisms in cell biology. Recent developments in fluorescence light
microscopy (FLM) allow protein imaging in living cells and at the near molecular level.
158 Klaas A. Sjollema et al.
However, fluorescence microscopy only reveals selected biomolecules or organelles but

not the (ultra)structural context, as can be examined by electron microscopy (EM). LM and
EM of the same cells, so-called correlative (or correlated) light and electron microscopy
(CLEM), allow examining rare or dynamic events first by LM, and subsequently by EM.
Here, we review progress in CLEM, with focus on matching the areas between different
microscopic modalities. Moreover, we introduce a method that includes a virtual overlay
and automated large-scale imaging, allowing to switch between most microscopes. Ongo-
ing developments will revolutionize and standardize CLEM in the near future, which thus
holds great promise to become a routine technique in cell biology.
Abbreviations
CLEM correlated light and electron microscopy

DAB diaminobenzidine
EM electron microscopy
FLM fluorescence LM
FP fluorescent protein
FRAP fluorescent recovery after photobleaching
GFP green FP
LM light microscopy
ROI region of interest
SEM scanning EM
STED stimulated emission depletion
TEM transmission EM
PALM photoactivated localization microscopy
I. Introduction
Fundamental to understanding how biomolecules regulate life is the ability to
study them at the near-molecular level. Microscopic techniques are highly valuable
to determine protein location in space and time. Ongoing development of geneti-
cally encoded fluorescent proteins (FPs), such as the green fluorescent protein (GFP),
allows imaging in living cells. In parallel, fluorescent microscopic techniques are
being developed that allow near-molecular imaging of biomolecules (∼10–50 nm
resolution). Although these developments have revolutionized light microscopy, it
can still not provide information about the (ultrastructural) cellular context, as can
be determined with electron microscopy (EM). However, protein identification with
EM can only be performed in fixed dead cells, and techniques that are used for 3D-
protein determination usually depend on destructive procedures. Can experiments be
designed that give both spatiotemporal information in living cells using (fluorescence)
light microscopy (LM), and subsequently reveal the ultrastructural details and cellular
context using EM? In general, the combination of multiple experimental approaches is
9. Correlated Light Microscopy and Electron Microscopy 159
a basic ingredient in scientific explorations. An excellent example is microscopy: When

either LM or EM reaches limitations, the other modality is being used in parallel. While
these side-by-side approaches usually are straightforward, there is an additional option
to use LM and EM techniques in a combined approach on the same sample, which may
be challenging, but can have added benefits.
II. Correlated Light and Electron Microscopy: Using the Best

of Two Worlds
A. Conventional Use with Transmission LM

Transmission LM is routinely used to identify specific cells and/or regions in tissues
to be processed for EM. Usually, cells or tissue samples are subjected to a color stain,
which allows recognition of tissue structures, certain cell types, or macromolecules.
Once the selection has been made, samples are further processed and examined with
EM. An overlay of the data of the two modalities is typically not made, since LM is
only used as guidance for the region of interest (ROI) for EM.
B. Fluorescence LM and CLEM: Exploration and Exploitation

Fluorescence microscopy combined with EM is more troublesome and will be the
focus of this chapter. Importantly, these combinatorial approaches should only be
implemented when there is an added benefit over using a single modality, since cor-
related light and electron microscopy (CLEM) procedures have limitations, including
compromising optimal sample preparation for both LM and EM. Severe ultrastructural
changes occur during standard (immuno)fluorescence sample preparation as shown by
EM (reviewed in Schnell, Dijk, Sjollema, & Giepmans, 2012). Furthermore, sample
preparation for EM requires fixation, cutting of the sample, or even permeabilization of
cells or tissue, which may ruin the 3D-ultrastructural context.
A prerequisite to correlate FLM and EM images in CLEM is to find back the area
imaged with LM at cm (lateral) and µm (axial) scale, with EM, typically at lateral mm
scale during sample preparation and µm–nm scale for acquisition. Several research groups
encountered problems in correlating CLEM images. Since the 1980s (Maranto, 1982;
Roth, Bendayan, & Orci, 1980), CLEM has been explored and exploited (reviewed in
Cortese, Diaspro, & Tacchetti, 2009; Giepmans, 2008; Kolotuev, Schwab, & Labouesse,
2009; McDonald, 2009; Sosinsky, Giepmans, Deerinck, Gaietta, & Ellisman, 2007; van
Driel, Valentijn, Valentijn, Koning, & Koster, 2009; van Weering et al., 2008). A number
of inventive procedures and probes have now been developed to match LM and EM data,
especially to examine organelles or molecules. In this chapter, the most recent develop-
ments in CLEM and the available methods of correlating the areas are presented, includ-
ing the introduction of a widely applicable technique we use in our laboratory. Most
methods result from the revolution of FPs, which play a central role in advanced FLM
techniques, such as super resolution microscopy and dynamic imaging of molecules.
1. High Resolution LM Markers

Biomolecules can be studied with fluorescence microscopy down to the near-
molecular level with 10–50 nm precision, so-called “super resolution microscopy”
(see for reviews Leung & Chou, 2011; Patterson, Davidson, Manley, & Lippincott-
Schwartz, 2010; Schermelleh, Heintzmann, & Leonhardt, 2010; Walter, Huang,
Manzo, & Sobhy, 2008). CLEM provides the unique possibility to correlate super
resolution FLM data with the ultrastructural EM information, thereby showing single
molecules in the context of macromolecules, membranes, organelles, cells and tissue.
The importance of CLEM in combination with super resolution techniques was shown
using photo activated localization microscopy (PALM), locating the fluorescent signal
of a mitochondrial protein within the mitochondria following cryosectioning (Betzig
et al., 2006). More recently, both PALM and stimulated emission depletion (STED)
have been used in combination with EM, where fluorescent super resolution data
could be localized to mitochondria and nuclei (Watanabe et al., 2011). In the latter
study, EM procedures have been optimized to allow super resolution FLM in fixed
and sliced sections for EM (see below). Although these developments hold promise
for broader application, super resolution CLEM applications are in a pioneering stage
and therefore still limited.
2. CLEM of Sparse Events and Dynamic Processes

Changes in levels, interactions, activity and subcellular localization of molecules
determine cell behavior. Key to understanding cellular processes is to know when and
where molecules are present with high accuracy. FP-techniques allow spatiotemporal
imaging of proteins during short-lived events, such as cell division, and detection of
rare cells or events in large samples. Using FLM markers and subsequent EM, dif-
ferent states of cells or organelles, e.g. the state of mitochondria during apoptosis,
can be complemented with ultrastructural examination (Sun et al., 2007). Moreover,
CLEM allows matching ultrastructure where dynamic information has been acquired
with live-cell FRAP-imaging, followed by photoconversion of newly formed signals
(Darcy, Staras, Collinson, & Goda, 2006). These new possibilities of studying protein
dynamics in the ultrastructural context of cells can uncover details that would have
remained ‘in the dark’ using FLM only.
3. Combinatorial Probes
Development of probes suitable for LM and EM started in 1980 with affinity
labeling using protA labeled with both FITC and immunogold (Roth et al., 1980;
reviewed in Robinson & Takizawa, 2009). In this chapter, we focus on recognition of
proteins. However, also markers for complete cells or membranes, e.g., DAPI, Luci-
fer yellow (LY; Maranto, 1982), or labeled lipids (Dantuma, Pijnenburg, Diederen, &
Van der Horst, 1998) have been shown great tools for CLEM. Besides affinity-based
targeting, genetically encoded proteins or combinatorial probes, e.g., FluoroNano-
gold or Quantum dots, are available that exhibit fluorescence and simultaneously
can be visualized by EM due to electron-dense precipitates or particles (see Cortese
et al., 2009; Sims et al., 2006; Sosinsky et al., 2007). Such probes ensure specific
targeting of the protein of interest and can be broadly applied to study protein activ-
ity and localization (reviewed in Giepmans, 2008; Giepmans, Adams, Ellisman, &
Tsien, 2006). Recently, a genetically encoded protein has been developed that does
not need additional labeling (mini-SOG). Because of its high singlet-oxygen quan-
tum yield, it forms a great target for photoconversion (Shu et al., 2011). Although
combinatorial probes for CLEM hold great promise, some approaches using the
super resolution techniques might make a subset of probes that deposit an EM mark
obsolete (see Watanabe et al., 2011).
While the sample preparation using either affinity-based or genetically encoded
probes might seem straightforward, it is not easy to overlay the FLM data with EM.
Currently, the bottleneck in most CLEM-approaches that include live-cell imaging is
to find back your ROI imaged by LM for EM. Below we review the current “search-
&-find protocols,” some of which rely on LM imaging of EM-prepared samples, others
that require a new type of microscope that integrates LM and EM. As an important step
toward routine CLEM, we present a more generic method that allows combining data
from almost any LM with SEM or TEM.
III. ROIs: Search & Find Tools
A. Large-Scale LM and EM
Traditionally, large-scale high resolution imaging has been performed by manual
acquisition followed by gluing pictures together. Automation and computer capacity to
handle GB datasets have been the basis of the development of automated imaging of
large areas at high magnification, which is a recent development (Chow et al., 2006).
Currently, this large volume imaging is a standard feature on most advanced fluores-
cence microscopes. Similar progress has been made on electron microscopes, allowing
acquisition of high resolution datasets of large areas (typically >1 GB). The orientation
between LM and EM datasets is relatively easy: hallmarks in the sample (especially in
tissue) or the borders of the culture dish or slide can be used for guidance. In principle,
LM or EM datasets are reformatted and rotated until they match, and the CLEM is done
entirely post image acquisition, since all ROIs are present in both datasets.
Major benefits of large-scale imaging include the chance of capturing rare spatial
events and the ability of high-throughput analysis, allowing quantitative studies. How-
ever, acquisition of a single point in time during imaging of the mosaic is relatively
time consuming, which limits the possibilities for temporal imaging. The resulting
large datasets contain a wealth of information, which should preferably be analyzed
in an automated manner. This usually requires dedicated software and compilation of
macros or plugins for image analysis programs such as ImageJ. While automated LM
acquisition and stitching have been standardized, automated acquisition and reconstitu-
tion of EM data is still in a pioneering phase, although very promising (Anderson et al.,
2009; Briggman, Helmstaedter, & Denk, 2011; Cardona et al., 2010; Kaynig, Fischer,
Muller, & Buhmann, 2010; Sims & Hardin, 2007).
B. Marks
Essential for CLEM is to find back the ROI selected by LM (cm scale) during EM
sample preparation (mm scale) and acquisition (µm–nm scale). Grids or marks included
in cell culture dishes are of great help for orientation. To aid in relocating ROIs, glass
coverslips with etched grids can be used, which are reproduced on the resin block after
embedding (Perkins, Sun, & Frey, 2009). Another way of highlighting an area of interest
is to use a diamond pen to induce a scratch, which will also be visible in the resin. The
marked area can then be selected either by eye or by hand, or for instance by using an
objective holder with a diamond (or other marker) instead of a lens to encircle the area
of interest, which is further processed for EM (Merdes & De Mey, 2011). Gridded cov-
erslips with positional markers are also used to combine live-cell imaging and immu-
nolabeling (Hanson, Reilly, Lee, Janssen, & Phillips, 2010; van Rijnsoever, Oorschot,
& Klumperman, 2008). Because the grid is not visible during imaging at high magni-
fication, a separate image at lower magnification is needed to see the grid pattern and
coordinates. Another disadvantage of growing cells on a finder grid is that the sample
may be partially obscured (Sartori et al., 2007). Alternatively, cells can be grown on
Formvar that is coated with a finder mask of a fine layer of gold or carbon. During the
coating, a “masking device” can be used that prevents gold deposition on the Formvar
in certain positions, thereby creating a “negative” grid pattern (Auinger & Small, 2008).
A seemingly straightforward way of pinpointing areas of interest is to extract the
ROI during sample preparation, which is routinely performed when selecting certain
areas of tissues for further analysis. This method can also be applied when (pre-)select-
ing cells of interest in a culture dish. Using a pipet tip, the cells surrounding the ROI
can be carefully removed.
C. Photoconversion and CLEM

Several methods for correlated microscopy depend on polymerization of osmiophilic
diaminobenzidine (DAB; Meisslitzer-Ruppitsch, Rohrl, Neumuller, Pavelka, & Ellinger,
2009; Sosinsky et al., 2007). When fluorophores are used to generate singlet oxygen to
polymerize DAB by intense illumination using an objective, the illuminated area will
be darker than its surrounding, and is thereby intrinsically labeled. This straightforward
method to define ROIs imaged at LM level has been used by several researchers (pioneered
by Deerinck et al., 1994; Maranto, 1982). A highly informative cartoon with the steps of
this procedure was recently presented (Meisslitzer-Ruppitsch et al., 2009), and several
examples of CLEM using photoconversion of genetically encoded probes are shown by
Shu and coworkers (2011). While photoconversion provides a fast, straightforward orien-
tation at the mm scale, obviously, the match has to be made at the cellular scale as well.
D. Immunolabeling and Combinatorial Probes and CLEM

The most developed way of CLEM is labeling thin sections with immunoreagents
that can be detected at both LM and EM levels. This is a generic way of labeling,
which is being explored for CLEM since the 1980s, prior to the GFP revolution. The
Tokayasu(-like) labeling (Bos et al., 2011; Tokuyasu, 1973) of ultrathin sections no
longer only includes immunogold, but combinatorial probes combining gold and fluo-
rescent dyes. Following labeling, the sections are first analyzed by LM and subse-
quently by EM. Special finder grids for FluoroNanogold labeling of thin sections have
been explored by Takizawa (Robinson, Takizawa, Pombo, & Cook, 2001; Takizawa,
Suzuki, & Robinson, 1998). Combinatorial probes have been applied as markers not
only to identify proteins or structures, but also to orient the sample at the multi-nm
scale: Protein A-gold in combination with fluorophores can be used as fiducial markers
to overlay LM and EM data (Vicidomini et al., 2008, 2010). Obviously, this approach
is highly suitable for serial sections followed by 3D reconstruction. To detect numerous
antigens with high resolution, a procedure has been developed to repeat immunolabel-
ing and imaging several times, applying different antibodies. Implementing this on
many thin sections and reconstructing the image is presented as “Array Tomography,”
which also can be followed by EM analysis of the specimen (Micheva & Smith, 2007).
An added bonus of this technique is the improved axial resolution by using thin sec-
tions at the FLM level. The axial resolution is the physical section thickness, typically
∼60–80 nm. As opposed to many other CLEM setups, the use of combinatorial probes
is true-correlated CLEM at the molecular scale: the same antigen is detected at both
modalities. This might even be further optimized using intrinsic properties of the probe,
e.g., quantum dots for postembedding EM (Nisman, Dellaire, Ren, Li, & Bazett-Jones,
2004), for both LM and EM. The Tokayasu approach focuses on immunolabeling. In
addition, for affinity dyes, such as nuclear stains or phalloidin, the procedure also has
been nicely illustrated by Jahn et al. (2009).
E. Live-Cell Imaging and CLEM

A disadvantage of immunolabeling is the common incapability of live-cell stud-
ies. One straightforward solution is the use of FPs for live-cell imaging followed by
cryoimmunogold labeling for EM (van Rijnsoever et al., 2008). Here, a mark on the
coverslip or gridded coverslips is usually needed for orientation. For vitrified samples,
the approach of whole-grid imaging may also be used, including imaging of living cells
using certain probes. LM imaging of the grid requires a special cryoholder as described
by Sartori. Also here, finder grids will aid in relocating the region of interest (Briegel
et al., 2010; Sartori et al., 2007). Additionally, the use of software-directed guidance to
find back coordinates of cells grown on EM grids may be useful (Sartori et al., 2007).
A recent report described fabrication of special holders for cryo-LM imaging of sam-
ples at low costs (Carlson & Evans, 2011). Nowadays, genetically encoded proteins can
also directly be detected in thin sections that have a well-preserved ultrastructure: mul-
ticolor live-cell imaging can be more easily combined with tomography when the EM-
samples also still have their fluorescence using cryopreparations, as well demonstrated
by Briggs and coworkers (Kukulski et al., 2011; see Watanabe et al., 2011 for chemical
fixation). This can be even extended to intravital CLEM, shown already for zebrafish
embryos (Nixon et al., 2009). Besides cryo-EM, CLEM is also compatible with high-
pressure freezing (HPF). In this approach, cells are typically grown on special sapphire
discs or aclar (Jimenez et al., 2010). As for grid-labeling, the area of the discs restricts
the area of interest at the (live) FLM level. Thus, the HPF method is suitable for high
resolution EM analysis, but not for large-scale CLEM (Lucic et al., 2007). To facili-
tate a fast transition from the FLM imaging toward the vitrified state, a rapid transfer
system has been developed that allows to process living cells, in which dynamics have
been analyzed with fluorescent probes, within 5 s for EM-analysis (Verkade, 2008).
F. Super Resolution Microscopy and CLEM

The most recent development in correlating both the FLM signal and the ultrastruc-
ture in thin sections is with the aid of super resolution microscopy. While this has been
performed on cryomaterial in the PALM-studies mentioned earlier (Betzig et al., 2006),
here the ultrastructural preservation was hampered. Recently, CLEM for genetically
encoded probes has been optimized. Crucial ingredient to balancing ultrastructural
and fluorescence preservation was a cocktail of 0.1% potassium permanganate com-
bined with 0.001% osmium tetroxide. To retain the fluorescence capability, samples
were embedded in glycol methacrylate. This protocol led to sufficient fluorescence to
perform PALM and STED imaging on semithin sections, and simultaneously allowed
good preservation of the ultrastructure (Watanabe et al., 2011). 3D reconstitution of
large volumes using LM markers has been pioneered by Smith, using array-tomography
(Micheva & Smith, 2007) as mentioned earlier.
G. Correlation of LM and EM Using LM Image of Block Face

Alternative to imaging ultrathin sections, imaging of the complete block face might
be an option. This allows relative easy orientation of the sample, and all LM lateral
information will correspond to the EM sample, since both images can be overlaid
completely. We showed an example previously, where chemically fixed samples had
been labeled preembedding using quantum dots. Postembedding, excluding Osmium
to avoid loss of fluorescence, the block face was imaged first with FLM, followed by
ultrathin-sectioning and TEM analysis (Giepmans, Deerinck, Smarr, Jones, & Ellisman,
2005). While this procedure leads to x/y CLEM, the axial information taken with
the LM will extend several ultrathin-sections. Alternative to serial sectioning EM as
described earlier (see also Anderson et al., 2009; Cardona et al., 2010), the whole
sample can be embedded after LM and imaged as a block in a SEM using the backscat-
ter detector. While no selection of the area is needed, the disadvantage of this method
is the lower resolution.
H. Stage Holders for CLEM and Integrated Microscopes

Custom stage holders or rapid transfer systems have been built to facilitate live-cell
imaging followed by cryo-EM analysis as described earlier. Currently, devices are con-
structed that are designed to find back the area of interest. A commercially available
device is “Shuttle & Find” from Zeiss. This system is based on positioning a mark on
the holder, and with software intervention it allows to automatically allocate the posi-
tion which is imaged with a dedicated inverted LM on a dedicated SEM. Currently, this
device is not suitable for TEM because it is incompatible with ultrathin sectioning of
the samples. Further developments are needed to make such a system more versatile.
When samples are prepared for LM and EM analysis, using fluorescence imaging
integrated in an electron microscope is a straightforward way to identify ROIs. Some
systems have now been developed that allow FLM in the EM, suited to study rare
events/features in tissue but incompatible with live-cell imaging. Systems are built by
several manufacturers, including a TEM-based system with an integrated LM objective
by FEI (Agronskaia et al., 2008), as well as a system that has the optical and electron
detecting devices at the opposite sides of the sample (JEOL Clairscope; Nishiyama
et al., 2010), and an LM positioned in situ in the SEM underneath the sample (Delmic
Secom; J. Hoogenboom personal communication). The latter integrated microscope
may benefit from improving wet EM, i.e., imaging specimens with both LM and EM
in liquid, pioneering the work being undertaken by de Jonge and coworkers (Dukes,
Peckys, & de Jonge, 2010; de Jonge & Ross, 2011).
Depending on the experimental question, a certain microscopy modality is selected
for CLEM. All methods mentioned earlier have advantages, but also restrictions, such
as lack of freedom of switching between any LM and/or EM of choice.
IV. Our Approach: Virtual Reality Overlay during Preparation
While mosaic microscopy was still challenging a decade ago, it is nowadays avail-
able on many microscopes. Utilizing automated large-scale imaging, our method allows
a relative straightforward correlation of LM and EM images taken on almost any sys-
tem and of almost any type of sample (both cultured cells and tissue). Subsequent
analysis of the underlying ultrastructure with EM does not need expensive equipment.
Orientation and correlation of LM and EM images is facilitated by an intrinsic hallmark
(tissue, dust, etc.) or an induced mark (e.g., scratch) in the sample.
A. Aim
Facilitate finding back objects imaged in any LM and EM.
B. Requirements (See also Fig. 1)

• Fluorescence microscope, preferably capable of making automated mosaics
• A binocular equipped with a drawing tube
• Computer and monitor
• Image analysis software (we use Adobe Photoshop, ImageJ, TissueFaxs viewer
from TissueGnostics and ICE from Microsoft)
• Standard embedding and sectioning facility
• Electron microscope, preferably capable of making automated mosaics
Fig. 1 Setup for searching and finding areas of interest. Microscope setup to facilitate tracking ROIs: a Zeiss
Stemi SV11 binocular equipped with continuous zoom and a drawing tube. The drawing tube is pointed toward
a computer monitor, which is rotated by 90° to match the computer screen image in the microscope.
C. Principle
In our approach, a complete LM mosaic is being overlaid with the embedded
sample using a drawing tube (virtual overlay). ROIs can be marked in the image as
well as on the embedded sample, enabling processing exactly the same area for EM
analysis.
D. Workflow (See also Figs. 2 and 3)

1. Grow cells on a glass bottom Petri dish or chamber slide. Preferably, glass
should contain a mark like a finder grid or diamond pen scratch. Later orienta-
tion is easier in a chamber than in a Petri dish.
2. Preferably use a nonconfluent layer of cells providing hallmarks.
3. If needed, follow cells of interest with live-cell imaging or fix cells.
4. Make a mosaic image of a large area of the chamber at sufficient magnification
(fluorescent channels and bright field/DIC), preferably including one or more
corners of the chamber.
Fig. 2 CLEM using a virtual overlay. Example of tracking down cells of interest from low magnification FLM to high magnification
EM. (A) Stitched mosaic (12 × 29 images; TissueFaxs) of living cells grown on a chamber slide; merge of bright field and red (mCherry)
and green (GFP) fluorescence images. This overview will be projected on the embedded cells. Yellow and green rectangles: silhouettes
of two eventually trimmed pyramids prior to ultrathin sectioning. (B) 1 µm thick, toluene blue stained section of the upper yellow pyra-
mid indicated in (A). Arrows indicate landmarks also visible in the ultrathin section (C) of the same pyramid. (D) Higher magnification
of (A) (see inset), which is overlaid with the low magnification TEM image (E) to facilitate the localization of cells of interest on TEM
scale. (F) Overlay of (D) with a higher magnification of the TEM overview (G), enabling the tracking of cells of interest in the TEM for
high magnification imaging. The arrow indicates a green fluorescent cell that is selected for higher magnification imaging. (H) Enlarged
fluorescent image of the cell of interest as indicated in (F), which can be overlaid with a mosaic reconstruction of ten TEM images of
the same region (I, J). The rectangle indicates the final high resolution TEM image of this cell (K). (See color plate.)
Fig. 3 Schematic flow chart of mosaic based CLEM back track. (1) A chamber slide on which cells are grown, only one cell is fluorescent
(green). Aim is to select this particular cell for TEM. A mark in the chamber, e.g., a scratch in the glass (diamond pen), facilitates later
matching of the fixed cells with the fluorescent images. The indicated grid in the chamber represents positions on which LM images are
made at 20× or 40× magnification (bright field/DIC and fluorescence), using an automated stage. Typically, a hundred or a few hundred
images are needed. (2) Cells are fixed, dehydrated, stained, and embedded in Epon. After polymerization, the glass bottom of the chamber
is removed. (3) Images taken in step 1 are stitched into one large image (TissueFaxs Viewer), with fluorescence and bright field as separate
files. Images are converted to separate layers in one Photoshop file, facilitating easy switching between bright field, fluorescence and the
merged image. In this large-scale image, ROIs can be marked by drawing a circle in a new layer. (4) Viewing the embedded sample from
step 2 in a binocular (bottom-up), the sample is overlaid with the image constructed in step 3, using a drawing tube which is pointing toward
a computer monitor. Marks in chamber and image can be used for positioning of the sample. Note that orientation and magnification of
sample and Photoshop image have to match (see Fig. 1). (5) Areas of interest selected in step 3 are marked with scratches using a razor-
blade, which are also drawn into the Photoshop image. (6) Areas of interest are sawed out and trimmed for ultra-thin sectioning. The draw-
ing tube and Photoshop image are used to control the exact position. (7) Low magnification TEM image or mosaic of the complete section is
acquired and overlaid with the Photoshop image as a new layer. (8) High magnification images of relocated area of interest can be acquired.
5. Fix and stain the cells according to a TEM protocol (cells have to be visible in a
light microscope; therefore, include Osmium or another clearly visible dye).
6. Repeat imaging as in step 4 if necessary.
7. Dehydrate and embed the cells in a suitable resin such as Epon (carefully, there
should not be any disruption of the cell layer).
8. Polymerize the resin.
9. Remove the glass bottom of the dish, e.g., with hydrofluoric acid (>4 h).
10. Make mosaic reconstructions of the images acquired in step 4, fluorescence and
bright field separated on aligned layers in one Photoshop file.
11. If necessary, improve the quality of the bright field reconstruction by contrast
enhancement (e.g., by using the shadow function of ImageJ) and background
subtraction to remove the halo.
12. Match the improved bright field mosaic image with the embedded cells in the cham-
ber (bottom-up) using the binocular and drawing tube. The tube is pointed forward
toward the computer monitor, which, in our case, is rotated by 90° to match the view
in the microscope. The mosaic image has to be flipped horizontally. The magnifica-
tion of the microscope has to match the size of the image on the monitor. A zoom
function on the microscope is obligate. If Photoshop is used, the image can easily be
zoomed in or out and can be panned on the monitor (keep all layers linked).
13. When positions of the cells in resin and on the monitor match, switch to the
fluorescent layer in Photoshop. The fluorescent signal is now projected on the
embedded cells.
14. Locate ROIs and mark them with some scratches in the resin. Looking through
the drawing tube, mark exactly the same areas in the Photoshop document.
15. Use a jigsaw to isolate small blocks (<25 mm2) that contain the selected areas.
16. Insert the block in a microtome sample holder and place it back in the drawing
tube microscope.
17. Overlay the areas of interest and match it with the cells visible in the block face.
18. Mark the area with a razorblade to facilitate the location and trimming of the area
of interest.
19. Make a pyramid to be able to section the area of interest in an ultramicrotome.
Draw the exact outline of the pyramid in the Photoshop document.
20. Make ultrathin sections of the block face, and place them on large mesh grids,
preferably single hole. Stain them if necessary.
21. Acquire a low magnification TEM image or a mosaic so that the complete sec-
tion is in one image. Overlay and correlate this image with the Photoshop image
as a new layer in the Photoshop document.
22. Locate the area of interest in the TEM and acquire high magnification images.
V. Discussion and Conclusion
Correlative microscopy is being explored since the 1980s and has been revolution-
ized since the introduction of FPs. Still, no routine method is available, and procedures,
probes, microscopes, and add-ons are under full development. The lack of a standard
procedure is partially caused by (1) compromising sample preparation or imaging when
needing to use both the modalities and (2) the difficulty of finding back areas of interest
as discussed here. The latter problem has been approached by several laboratories and
researchers, resulting in custom-made solutions, which have their specific benefits, but
also are of limited usability.
Considerations when using CLEM are first and foremost what kind of data is needed
to solve the research question. For example, we advocate automated mosaic imaging,
but the acquisition of 1000 images of a large area is much more time consuming than
imaging a single area. This precludes high temporal imaging of the individual areas
which might be needed for studying fast processes in living cells. Tokayasu-like label-
ing on ultrathin sections is well developed, but may be incompatible with live-cell imag-
ing when solely immunoreagents are used. Alternatively, GFP can be immunolabeled
on ultrathin sections following live-cell imaging. Another limitation of Tokayasu-like
labeling is the restriction on ultrathin sections, precluding 3D analysis, which can be
solved by array tomography. At the molecular level, labeling ultrathin sections gives a
genuine, molecular correlation, which is typically not seen when combining preembed-
ding (live) images with EM because of the limited resolution at LM level.
A recurrent theme of this chapter is the problem of correlating images from differ-
ent microscopes. This is not easy when selecting an ∼25 mm2 area of interest, but even
more difficult when making the selection at the cellular or macromolecular scale. In
most approaches, including ours, marks in the sample are important to match images.
These marks can either be intrinsic (stained nuclei, cell shapes, or groups of cells) or
introduced (finder grids, scratches, etc.).
Regarding the choice of microscopes, commercial systems are available for CLEM,
which are, however, quite costly. Preferably, microscopes are used that are easily
accessible and most helpful to solve the particular research question. When limited to
certain instruments, several options are available to correlate LM and EM data. Our
approach has the advantages that it can be done at low costs, and that it provides a
high degree of freedom regarding the choice of microscopes. Given all the different
approaches being undertaken to allow a more routine CLEM, what would these develop-
ments bring us next?
VI. Future Perspective
A recent prediction of the direction of CLEM was the increase of super resolution LM
and EM and software assistance in overlaying the LM and EM images (Cortese et al.,
2009). Indeed, recent studies discussed here have shown that these predictions were
right. Ongoing challenges, including orientation of samples, will soon be overcome by
(1) further development of integrated LM/EM microscopes and (2) large-scale LM and
EM imaging of large areas of interest at high magnification. We predict that these develop-
ments will help to establish CLEM as a routine technique available at institutional and
academic microscopy centers within the next 5 years.
Acknowledgments
Part of this work was supported by the Groningen University Graduate School
of Medical Sciences; a Marie Curie International Reintegration Grant within the 7th
European Community Framework Program to B.N.G.G. and was performed at the
University Medical Center Groningen Microscopy and Imaging Center, which is spon-
sored by Netherlands Organization for Scientific Research grants 40-00506-98-9021
and 175-010-2009-023.
References
Anderson, J. R., Jones, B. W., Yang, J. H., Shaw, M. V., Watt, C. B., Koshevoy, P., et al. (2009). A computa-
tional framework for ultrastructural mapping of neural circuitry. PLoS Biology, 7, e1000074.
Auinger, S., & Small, J. V. (2008). Correlated light and electron microscopy of the cytoskeleton. Methods in
Imaging intracellular fluorescent proteins at near-molecular resolution. Science, 313, 1642–1645.
Bos, E., SantAnna, C., Gnaegi, H., Pinto, R. F., Ravelli, R. B., Koster, A. J., et al. (2011). A new approach to
improve the quality of ultrathin cryo-sections; its use for immunogold EM and correlative electron cryo-
tomography. Journal of Structural Biology, 175, 62–72.
Briegel, A., Chen, S., Koster, A. J., Plitzko, J. M., Schwartz, C. L., & Jensen, G. J. (2010). Correlated light
and electron cryo-microscopy. Methods Enzymology, 481, 317–341.
Briggman, K. L., Helmstaedter, M., & Denk, W. (2011). Wiring specificity in the direction-selectivity circuit
of the retina. Nature, 471, 183–188.
Cardona, A., Saalfeld, S., Preibisch, S., Schmid, B., Cheng, A., Pulokas, J., et al. (2010). An integrated
micro- and macroarchitectural analysis of the Drosophila brain by computer-assisted serial section elec-
tron microscopy. PLoS Biology, 8, e1000502.
Carlson, D. B., & Evans, J. E. (2011). Low-cost cryo-light microscopy stage fabrication for correlated light/
electron microscopy. Journal of Visualized Experiments (52), doi:10.3791/2909. pii: 2909.
Chow, S. K., Hakozaki, H., Price, D. L., MacLean, N. A., Deerinck, T. J., Bouwer, J. C., et al. (2006). Automated
microscopy system for mosaic acquisition and processing. Journal of Microscopy, 222, 76–84.
methods and new developments using Tokuyasu cryosections. The Journal of Histochemistry and Cyto-
chemistry, 57, 1103–1112.
Dantuma, N. P., Pijnenburg, M. A., Diederen, J. H., & Van der Horst, D. J. (1998). Electron microscopic
visualization of receptor-mediated endocytosis of Dil-labeled lipoproteins by diaminobenzidine photo-
conversion. The Journal of Histochemistry and Cytochemistry, 46, 1085–1089.
Darcy, K. J., Staras, K., Collinson, L. M., & Goda, Y. (2006). An ultrastructural readout of fluorescence
recovery after photobleaching using correlative light and electron microscopy. Nature Protocols, 1,
988–994.
de Jonge, N., & Ross, F. M. (2011). Electron microscopy of specimens in liquid. Nature Nanotechnology,
6, 695–704.
Dukes, M. J., Peckys, D. B., & de Jonge, N. (2010). Correlative fluorescence microscopy and scanning
transmission electron microscopy of quantum-dot-labeled proteins in whole cells in liquid. ACS Nano,
4, 4110–4116.
Giepmans, B. N., Deerinck, T. J., Smarr, B. L., Jones, Y. Z., & Ellisman, M. H. (2005). Correlated light and elec-
tron microscopic imaging of multiple endogenous proteins using Quantum dots. Nature Methods, 2, 743–749.
Hanson, H. H., Reilly, J. E., Lee, R., Janssen, W. G., & Phillips, G. R. (2010). Streamlined embedding of
cell monolayers on gridded glass-bottom imaging dishes for correlative light and electron microscopy.
Jahn, K. A., Barton, D. A., Su, Y., Riches, J., Kable, E. P., Soon, L. L., et al. (2009). Correlative fluorescence
and transmission electron microscopy: an elegant tool to study the actin cytoskeleton of whole-mount
(breast) cancer cells. Journal of Microscopy, 235, 282–292.
Jimenez, N., Van Donselaar, E. G., De Winter, D. A., Vocking, K., Verkleij, A. J., & Post, J. A. (2010). Gridded
Aclar: preparation methods and use for correlative light and electron microscopy of cell monolayers, by
TEM and FIB-SEM. Journal of Microscopy, 237, 208–220.
Kaynig, V., Fischer, B., Muller, E., & Buhmann, J. M. (2010). Fully automatic stitching and distortion correc-
tion of transmission electron microscope images. Journal of Structural Biology, 171, 163–173.
Kukulski, W., Schorb, M., Welsch, S., Picco, A., Kaksonen, M., & Briggs, J. A. (2011). Correlated fluo-
rescence and 3D electron microscopy with high sensitivity and spatial precision. The Journal of Cell
Biology, 192, 111–119.
Leung, B. O., & Chou, K. C. (2011). Review of super-resolution fluorescence microscopy for biology. Applied
Spectroscopy, 65, 967–980.
Lucic, V., Kossel, A. H., Yang, T., Bonhoeffer, T., Baumeister, W., & Sartori, A. (2007). Multiscale imaging
of neurons grown in culture: from light microscopy to cryo-electron tomography. Journal of Structural
Biology, 160, 146–156.
Maranto, A. R. (1982). Neuronal mapping: a photooxidation reaction makes Lucifer yellow useful for electron
microscopy. Science, 217, 953–955.
Meisslitzer-Ruppitsch, C., Rohrl, C., Neumuller, J., Pavelka, M., & Ellinger, A. (2009). Photooxidation
technology for correlated light and electron microscopy. Journal of Microscopy, 235, 322–335.
Merdes, A., & De Mey, J. (2011). Kinetochore microtubules, analyzed by correlated light and immunoelectron
microscopy. Methods in Molecular Biology, 777, 209–221.
Nishiyama, H., Suga, M., Ogura, T., Maruyama, Y., Koizumi, M., Mio, K., et al. (2010). Atmospheric scan-
ning electron microscope observes cells and tissues in open medium through silicon nitride film. Journal
Nixon, S. J., Webb, R. I., Floetenmeyer, M., Schieber, N., Lo, H. P., & Parton, R. G. (2009). A single method
for cryofixation and correlative light, electron microscopy and tomography of zebrafish embryos. Traffic,
10, 131–136.
Patterson, G., Davidson, M., Manley, S., & Lippincott-Schwartz, J. (2010). Superresolution imaging using
single-molecule localization. Annual Review of Physical Chemistry, 61, 345–367.
Perkins, G. A., Sun, M. G., & Frey, T. G. (2009). Chapter 2 correlated light and electron microscopy/electron
tomography of mitochondria in situ. Methods Enzymology, 456, 29–52.
Robinson, J. M., Takizawa, T., Pombo, A., & Cook, P. R. (2001). Correlative fluorescence and electron
microscopy on ultrathin cryosections: bridging the resolution gap. The Journal of Histochemistry and
Robinson, J. M., & Takizawa, T. (2009). Correlative fluorescence and electron microscopy in tissues: immu-
nocytochemistry. Journal of Microscopy, 235(Sep (3)):259–272.
Roth, J., Bendayan, M., & Orci, L. (1980). FITC-protein A-gold complex for light and electron microscopic
immunocytochemistry. The Journal of Histochemistry and Cytochemistry, 28, 55–57.
bridging the gap between fluorescence light microscopy and cryo-electron tomography. Journal of Struc-
tural Biology, 160, 135–145.
Schnell, U., Dijk, F., Sjollema, K. A., & Giepmans, B. N. (2012). Immunolabeling artifacts and the need for
live-cell imaging. Nature Methods, 9, 152–158.
Biology, 9, e1001041.
Sims, P. A., & Hardin, J. D. (2007). Fluorescence-integrated transmission electron microscopy images: inte-
369, 291–308.
Sims, P., Albrecht, R., Pawley, J. B., Centonze, V., Deerink, T., & Hardin, J. (2006). When light microscope
resolution is not enough: correlational light microscope and electron microscope. In J. Pawley (Ed.),
Handbook of biological confocal microscopy (pp. 846–860). New York: Springer.
Sun, M. G., Williams, J., Munoz-Pinedo, C., Perkins, G. A., Brown, J. M., Ellisman, M. H., et al. (2007).
Correlated three-dimensional light and electron microscopy reveals transformation of mitochondria dur-
ing apoptosis. Nature Cell Biology, 9, 1057–1065.
Takizawa, T., Suzuki, K., & Robinson, J. M. (1998). Correlative microscopy using FluoroNanogold on ultrathin
cryosections. Proof of principle. The Journal of Histochemistry and Cytochemistry, 46, 1097–1102.
Tokuyasu, K. T. (1973). A technique for ultracryotomy of cell suspensions and tissues. The Journal of Cell
Biology, 57, 551–565.
van Driel, L. F., Valentijn, J. A., Valentijn, K. M., Koning, R. I., & Koster, A. J. (2009). Tools for correlative
cryo-fluorescence microscopy and cryo-electron tomography applied to whole mitochondria in human
endothelial cells. European Journal of Cell Biology, 88, 669–684.
combining live-cell imaging and immunolabeling of ultrathin cryosections. Nature Methods, 5, 973–980.
van Weering, J. R., Wijntjes, R., de Wit, H., Wortel, J., Cornelisse, L. N., Veldkamp, W. J., et al. (2008).
Automated analysis of secretory vesicle distribution at the ultrastructural level. Journal of Neuroscience
Methods, 173, 83–90.
Verkade, P. (2008). Moving EM: the Rapid Transfer System as a new tool for correlative light and electron
Vicidomini, G., Gagliani, M. C., Cortese, K., Krieger, J., Buescher, P., Bianchini, P., et al. (2010). A novel approach
for correlative light electron microscopy analysis. Microscopy Research and Technique, 73, 215–224.
Walter, N. G., Huang, C. Y., Manzo, A. J., & Sobhy, M. A. (2008). Do-it-yourself guide: how to use the
modern single-molecule toolkit. Nature Methods, 5, 475–489.
CHAPTER 10
Capturing Endocytic Segregation Events

with HPF-CLEM
Edward Brown*, Jan van Weering*, Thom Sharp*, Judith
Mantell*,†, and Paul Verkade*,†,‡
* Department
of Biochemistry, School of Medical Sciences, University of Bristol, University Walk, Bristol,
BS8 1TD, UK
† Wolfson Bioimaging Facility, School of Medical Sciences, University Walk, Bristol, BS8 1TD, UK
‡ Department of Physiology and Pharmacology, School of Medical Sciences, University Walk, Bristol, BS8
1TD, UK
Abstract
I. Introduction
A. Intracellular Trafficking
B. Fixation for Electron Microscopy
C. Fixation of Membrane Tubules
D. The Use of HPF-CLEM for Studying Intracellular Trafficking
II. Methods
A. CLEM Markers
B. Specimen Carriers
C. Preparation of Live Cell Carriers
D. Development of a Modified CLEM Carrier
E. Preparation of a Modified CLEM Carrier
F. Modification to Rapid Loader
G. Improvement of Image Acquisition
H. Live Cell Imaging and HPF-CLEM
I. Sample Processing for EM

J. Electron Microscopy and Data Correlation
K. Tomography and Modeling
III. Outlook
Acknowledgments
References

176 Edward Brown et al.
Abstract
We have advocated the use of high-pressure freezing (HPF) in specific types of

Correlative Light Electron Microscopy (CLEM) experiments because the intracellular
components such as the cytoskeleton and membrane tubules can only be adequately
preserved via cryofixation. To allow fast transfer from the light microscope into a cryo-
fixation device, we have developed the Rapid Transfer System (RTS) for the EMPACT2
high-pressure freezer. In this chapter, we will describe how to prepare and perform a
CLEM experiment using this device and will highlight the latest changes made to the
original system to optimize the workflow.
I. Introduction
Our lab is interested in intracellular trafficking and sorting mechanisms inside cells.
One of the ways cells achieve this sorting is by the formation of membrane tubules.
When tagged with fluorescent molecules and introduced into cells, both specific types
of cargo and structural components can be seen in the light microscope to form elon-
gated structures (tubules) that move around in the cell (see, e.g., Brown & Verkade,
2010). However, their numbers and life span are limited compared to the more spheri-
cal structures and at any given time point only a handful of tubules may be visible or
present in a cell.
In order to gain a better understanding of those tubules, we need to study them at
higher resolution than possible with light microscopy (LM); hence, we turn to electron
microscopy (EM). Although this technique can provide high-resolution data, it only
provides static images, and with a limited field of view finding, a membrane tubule may
be like finding the proverbial needle in the haystack.
Correlative light electron microscopy (CLEM) aims at combining advantages of
both light and EM. One way to employ CLEM is to first search for rare or dynamic
events live and in overview in the light microscope. Then, after fixation and process-
ing, the location of the event can be retraced and studied both in high resolution in the
electron microscope and also in relation to other structures inside the cell, providing a
reference space.
We have advocated a strategy for CLEM whereby after live-cell imaging, the sam-
ple is cryofixed by high-pressure freezing (HPF) rather than standard chemical fixation.
Some cellular structures such as parts of the cytoskeleton and membrane transport
tubules are susceptible to breakdown following chemical fixation and for our case, it is
exactly those membrane tubules that we want to study.
In order to perform CLEM experiments in combination with HPF, we have previ-
ously developed the rapid transfer system (RTS) attachment to the EMPACT2 high-
pressure freezer and described some of the tools necessary in the process (Brown,
Mantell, Carter, Tilly, & Verkade, 2009; Verkade, 2008). In this chapter, we will discuss
the different steps involved in our CLEM process and highlight the most recent modi-
fications we have made to the system.
10. Capturing Endocytic Segregation Events with HPF-CLEM 177
A. Intracellular Trafficking
Intracellular transport, via membrane-bound carriers, is essential for proper func-
tioning of a cell. It, for instance, delivers newly synthesized proteins via the endoplas-
mic reticulum (ER) and the Golgi complex to the location where they can perform
their function. It is also responsible for maintaining lipid bilayers, storage of signal
molecules, recycling or degrading of used proteins, and providing membrane expan-
sion in cell division. Although intracellular trafficking is not the main focus of this
paper, we feel a short introduction into the subject is useful to better comprehend our
CLEM strategy.
The cell contains many trafficking routes with different start and end stations
(Fig. 1). This requires that most of the steps in the trafficking processes are highly
regulated to obtain specificity. In general, there are two main routes: the biosynthetic
pathway and the degradative route. The biosynthetic pathway (Fig. 1, blue arrows)
starts on the ribosomes of the ER with the translation of mRNA into proteins. The
proteins are transported via vesicles to the Golgi complex. After posttranslational
modification in the Golgi, the proteins are sorted and packaged in the trans-Golgi
network (TGN). They are then transported to different locations in the cell where
each protein performs its function, for instance, in the plasma membrane or in the
endosomal system.
A similar route with different sorting stations is present for material taken up from
outside the cell that starts with the internalization or endocytosis of material from the
plasma membrane via small vesicular profiles. These endocytosed vesicles fuse with
an early endosome where the cargo is sorted in a tubular endosomal network. Most of
this material will take the degradative route (Fig. 1, orange arrows), but a number of
proteins are recycled back to the plasma membrane or the TGN (Fig. 1, gray arrows).
There are many points of cross talk between these major transport pathways but again
it is beyond the scope of this review to discuss these in detail.
Our main research focus lies in the study of two membrane-bound receptors: the
Epidermal Growth Factor receptor (EGFr, Gorden, Carpentier, Cohen, & Orci, 1978;
Haigler, McKanna, & Cohen, 1979) and the Transferrin receptor (Tfr). When bound to
their respective ligands, they are internalized via the same early pathway, but later on
Tfr is sorted away from these endosomes and recycled back to the plasma membrane
via membrane tubules; whereas, EGFr is going further down the degradative pathway
(Stocheck and Carpenter, 1984) (Fig. 1). It is this point of sorting that we are trying to
elucidate with our CLEM approach (Fig. 2). The need to apply CLEM to these studies
can be illustrated by the fact that there still remains confusion over issues, such as the
colocalization of Tfr and EGFr throughout the endocytic pathway (Leonard et al., 2008).
B. Fixation for Electron Microscopy

Although EM can be a very powerful tool to study intracellular trafficking (Van
Weering, Brown, Mantell, Sharp, & Verkade, 2010), an intrinsic part of any EM
experiment is that the sample will need to be fixed before it can be analyzed inside the
EM. It is well known that this fixation (and the proceeding processing) can introduce
EE
ation
tur
ma TEN
RE
maturation
TGN
LE/MVB
G
ER
L
ERGIC
TfR
EGFR
Coat
Fig. 1 Trafficking routes in cells. Model depicting the major membrane trafficking routes in cells. Orange
arrows mark the endocytic and degradative pathway. Ligands and receptors in this pathway, such as the trans-
ferrin receptor (TfR) and the epidermal growth factor receptor (EGFR) located at the plasma membrane are
internalized and initially trafficked to early endosomes (EE). From here the EGFR is further following the
degradative pathway via late endosomes/multivesicular bodies (LE/MVB) to be degraded in the lysosomes (L).
The TfR however is recycled back to the plasma membrane (gray arrows), either directly or via recycling endo-
somes (RE). Proteins synthesized in the endoplasmic reticulum (ER) follow the biosynthetic pathway (blue
arrows) via the Golgi apparatus (G) to its destination, for example the plasma membrane. ERGIC, ER–Golgi
intermediate compartment; TGN, trans-Golgi network; TEN, tubular endosomal network. (See color plate.)
Fig. 2 The principle of correlative light electron microscopy. In a CLEM experiment for studying
intracellular transport, a trafficking event is first studied (live, not shown) in the fluorescence light
microscope (A). This provides an overview of the location and movement of marked proteins and allows
one to focus on an event of interest. This event is captured (fixed) and then the exact structure of interest
is studied at higher resolution in the electron microscope (B). This not only provides a more detailed
location of the marker proteins but also additional information about the ultrastructural context and its
reference space. (See color plate.)
major artifacts (Penttila et al., 1974, Kellenberger et al., 1992). There are two types of
fixation widely used in EM: chemical and cryo. Chemical fixation relies on the diffu-
sion of aldehydes through the specimen that cross-links proteins, immobilizing them
in their local environment. As a result, it is not instantaneous as the rate of diffusion
governs the speed of fixation (Rosenmund and Stevens, 1997). Areas of the cells fur-
thest from the surface will receive fixative last and may have altered since the original
application. What’s more! Aldehydes do not cross-link all cellular components; lip-
ids, for example, remain unfixed until later exposure to an additional fixative such as
osmium tetroxide (Small et al., 1981).
Cryofixation is preferred to chemical fixation when preservation of ultrastructure is
of utmost importance. In contrast to fixation by chemical means, cryofixation physi-
cally immobilizes all cellular contents within milliseconds (Dahl & Staehelin, 1989).
The goal in cryofixation is to convert all the water in the sample to vitreous ice. Vitre-
ous ice occupies the same volume as water, whereas crystalline and cubic ice, on the
other hand, occupies a larger volume. In practice, the production of small ice crys-
tals is unavoidable and its effects are largely unnoticed. Large ice crystal formation,
on the other hand, can cause rupturing of membranes and aggregation of intracellular
proteins in a biological sample. To avoid large ice crystal formation and to achieve
well-frozen specimens, rapid removal of thermal energy is required somewhere in the
region of 10,000°C/s (Dahl & Staehelin, 1989). The distance of the specimen from the
cooled surface largely determines the rate of cooling. This places limitations on the
thickness of the specimen; for plunge freezing, the maximum specimen thickness is
around 20 µm. Because the physical properties of water are altered when pressurized,
this thickness can be increased if pressure is applied to the specimen. Its freezing point
is lowered and its propensity to form ice crystals is reduced, and thus the critical cool-
ing rate required can be lowered up to 100 fold (Moor, 1987). It is these characteristics
of water that led to the development of HPF.
C. Fixation of Membrane Tubules

Chemical fixation had been shown to be particularly poor at preserving early endo-
cytic compartments (Murk et al., 2003). Moreover, breakdown of tubules into series of
vesicular structures has been shown in the electron microscope as a direct consequence
of fixation artifacts (Hyde, Lancelle, Kepler, & Hardham, 1991; Mersey and McCully
1978; Rosenbluth 1963; Sandborn & Bendayan 1996; Tormey, 1964; Wilson, Canny,
McCuly, & Lefkovitch, 1990). As tubular extensions emanating from early endosomes
are our objects of study, we wanted to investigate whether chemical fixatives would not
only cause ultrastructural changes as seen in the electron microscope but would also cause
visible alteration of tubules in the light microscope. Cells were incubated with a fluores-
cent marker (Tfn-Alexa488) that after internalization recycles back to the plasma mem-
brane and is thus included in these membrane tubules. Tfn-Alexa488 was internalized
5 min prior to live cell imaging (Brown & Verkade 2010). At a moment during live cell
imaging where numerous tubules were visible, 2.5% glutaraldehyde (warmed to 37°C)
was added to the cells. This addition caused a loss of focus and meant an adjustment
to the focal plane was required to find the same tubules. Glutaraldehyde addition also
caused a decrease in intensity from the Alexa dye and an increase in autofluorescence.
In order that this reduction in signal intensity does not affect tubule measurements,
brightness was increased to the prefixation level. Measurements of the same tubules in
pre- and postfixation images (Fig. 3) were made along the length of the tubule from tip
to tip. We found that there is a neglectable reduction in the length of the fluorescence
signal of the tubules (3% on average). However, we did observe that tubules before fixa-
tion had an even distribution of fluorescence. This had changed to a pattern where the
fluorescence was more punctate and occasionally showed gaps in the “tubule” (Fig. 3).
It could be that the fluorescent molecules have become more concentrated, possibly as a
consequence of the breakdown of the tubule in smaller segregated vesicles. It is highly
likely however that the breakdown of tubules occurring during chemical fixation and
observed in the electron microscope is below the resolution of the light microscope and
can only be observed occasionally. Currently, we cannot think of an experiment which
would conclusively settle this matter.
D. The Use of HPF-CLEM for Studying Intracellular Trafficking

As the term indicates, intracellular trafficking is a highly dynamic process. It is thus
very amenable for live cell imaging studies using fluorescent proteins. These studies
will tell us about the history of the structures but they do not always provide enough
detail. Thus, we combine live LM together with high-resolution EM in a CLEM experi-
ment. As the retention of the tubules during fixation is a critical step in the switch from
LM to EM, we have to use cryofixation. To reliably fix not only the outer edges of cells,
we have resorted to HPF. To this end, we have developed the RTS attachment to the
EMPACT2 high-pressure freezer together with Leica Microsystems (Verkade, 2008).
This high-pressure freezer is mobile, so it can be placed next to any light microscope,
and with the RTS, it is possible to transfer a sample from the light microscope stage into
the high-pressure freezer, allowing for a time resolution between the last LM image
and point at which the sample is frozen of about 4 s (Verkade, 2008). In the remainder
of this chapter, we will describe procedures of how we have recently improved on this
process and what other solutions we have tried but were less successful.
II. Methods
A. CLEM Markers
One of the initial and most important decisions to make when performing a CLEM
experiment is the choice of the marker or probe. A CLEM probe must be fluorescent
in the LM and electron-dense in the EM. Different experimental questions also require
different probes. There is a wealth of publications on CLEM probes (Giepmans, Adams,
Ellisman, & Tsien, 2006, but also see a number of other chapters in this book), but
thus far, no genetically encoded CLEM probe, like FlAsh/ReAsh or miniSog (Gaietta
Fig. 3 The influence of fixation on fluorescently labeled membrane tubules. Transferrin labeled with Alexa488
was internalized into HeLa cells and followed by live-light microscopy imaging. (A) shows the last frame of the
live imaging sequence before the addition of fixative. (B) shows a frame of the same cell after the addition of the
fixative. Not much difference can be observed, but in some tubules there appear to be breaks in the fluorescent
signal, possible indications of the breakdown of these membrane tubules.
Fig. 4 Schematic model of the CLEM markers. Schematic representation of the endocytic CLEM markers
used in our studies. On the left, the structure of transferrin is depicted with the red fluorophore Alexa594 and
a 5 nm gold particle is attached. On the right, EGF is tagged with the green Alexa488 and 10 nm gold particle
is attached. (See color plate.)
et al., 2002, Shu et al., 2011), is available that is compatible with HPF and freeze
substitution (requiring processing in a nonwatery milieu). Therefore for our research
question, we can only use CLEM markers that are internalized through endocytosis
(Brown & Verkade, 2010; Verkade, 2008). We can make use of probes such as quantum
dots and fluoro (nano) gold. However, we observed that transferrin and other tracers,
coupled to quantum dots can be miss-sorted (Brown & Verkade, 2010 and unpublished
results, respectively) as has been shown by others (Jaiswal & Simon, 2004). To avoid
the need to enhance ultrasmall 1 nm gold during the freeze substitution (Morphew, He,
Bjorkman, & McIntosh, 2008), we chose gold particles of 5 and 10 nm that are directly
visible in the electron microscope as our EM part of the CLEM marker. To be able to
discriminate them on the light microscopical level as well we also tagged Tf and EGF
with a red and green dye, respectively. This resulted in the following probes: EGF-
Alexa488-10 nm gold and Tf-Alexa594-5 nm gold (Fig. 4) that have been described in
Brown and Verkade (2010).
B. Specimen Carriers
If postfixation labeling is not required and the preservation of ultrastructure is the
primary concern as in our case, then LM must be followed by HPF. CLEM with HPF
can be performed without the need of a specialized transfer mechanism. For an excel-
lent review on the use of HPF in CLEM, see McDonald (2009). A specimen can be
imaged at the light microscope until an event of interest occurs. It can then be manipu-
lated using a stereo microscope into an appropriate carrier before the specimen is fro-
zen. This process could take from about 30 s up to several minutes; this is obviously too
long if you wish to capture events occurring in the order of seconds. To achieve this, the
sample must already be in a carrier while imaging so that when the event is observed
the sample can be loaded directly into the HPF. The need to achieve high-pressure fro-
zen samples in as short a time as possible has lead to the development of the RTS on
the EM PACT2 (Verkade, 2008).
For controlled pressurization of a sample, it must be contained within a closed sys-

tem. In HPF this is achieved by placing the sample in a sealable cavity known as a
specimen carrier. The presence of a carrier, however, must not significantly reduce the
rate of heat transfer from the sample. To that end carriers are typically made from met-
als such as copper, gold, and aluminum due to their high thermal conductivity (Table I).
Two main carrier shapes have evolved to accommodate the freezing of different sample
types. Tubes are used for samples suspended in solution and cups that can accommo-
date a range of specimens such as tissue, embryos, and cell monolayers. In the case of
cell monolayers, cells must first be grown on an inert substrate which can be contained
within a carrier and still allow efficient heat transfer from the cells. Glass cannot be
used in HPF as it acts as an insulator and is fragile. Sapphire however has much better
thermal conductivity than glass (Table I) and can also support cell growth without the
need of additional biological coatings. They are widely used in CLEM studies as they
are transparent, allowing transmission of light for LM. Another suitable substrate for
culturing cells for HPF is Aclar (Jiménez, Humbel, Van Donselaar, Verkleij, & Burger,
2006); while also being transparent and therefore useful for CLEM, we were surprised
to find that according to the data found on various web sources it has inferior heat-
transfer properties compared to sapphire and even glass (Table I), but then again Aclar
is of course a plastic material not renowned for its heat-conductive properties.
C. Preparation of Live Cell Carriers

After a sample has been frozen and processed for EM, the same region of interest
as viewed in LM must be relocated. The finder grid system employed in the live cell
clamp (Verkade, 2008) is one way to achieve this and the process for loading samples
this way is described below. The finder grid is clamped into the carrier so that the posi-
tion of cells can be determined using the surrounding numbers as a reference. There
are several issues with this finder grid system, however. Firstly, construction of live
cell clamps is not straightforward and requires considerable patience. The presence of
Table I
Thermal conductivity values of materials commonly used in HPF experiments.
Material at RT Thermal conductivity: Watt/meter × Kelvin
Aclar 0.19–0.22
Water 0.57–0.61
Glass 0.8–1.4
Ice 0°C 1.6–2.2
Sapphire 23–35
Aluminum 200–250
Gold 310–318
Copper 385–401
Diamond 800–2200
Sources: www.engineeringtoolbox.com, www.emsdiasum.com, www.matweb.com,

www.valleydesign.com, and Wikipedia.
a finder grid between objective and cells while imaging can cause interference if the
cell of interest is near a grid bar. The ability to relocate cells depends on the finder grid
remaining in place throughout HPF, freeze substitution, and embedding. However, if
it falls out of the carrier at some point or even moves significantly, it is not possible
to relocate the cell. As it is currently the only commercial system available for the
EMPACT2, we will describe the loading process in more detail in the following sec-
tions.
The live cell clamp kit is a commercially available product from Leica Microsys-
tems.
(1) Cells are grown on a 1.4 mm sapphire disk to a confluency of 60–70%.
(2) A 3 cm Petri dish is filled with CO2-independent medium (Gibco).
(3) A sapphire disk with cells is placed in the medium.
(4) A live-cell carrier and Nickel finder grid are submerged in the medium and all air
bubbles are removed using the pointed end of a tweezer (Fig. 5(A and B).
(5) The sapphire disk is lifted with a pair of fine tweezers and placed in the live-cell
carrier with the cells facing upward (Fig. 5(C)).
(6) The finder grid is lifted with a pair of nonmagnetic tweezers (the finder grid will
stick otherwise!) and placed on top of the sapphire disk (Fig. 5(D)). The finder grid
is slightly bigger than the opening of the carrier, so it sits only on one side on the
sapphire disk.
(7) The finder grid is held in place by one pair of tweezers. Using the second pair, the
finder grid is lightly pushed down and “bends” into the carrier opening (Fig. 5(E)).
The elasticity of the metal creates tension on the wall of the carrier holding the
finder grid and sapphire disk in place (see also Fig. 5(K–N)).
(8) To check whether the finder grid is secured tightly in the carrier, the carrier is
flipped over (Fig. 5(F)). Both finder grid and sapphire disk should stay in the car-
rier and can now be used for the CLEM experiment.
D. Development of a Modified CLEM Carrier

An alternative solution is to grow cells on a substrate with an incorporated grid
pattern that is transferable to the embedding resin (van Weering et al., 2010). Mem-
brane carriers (McDonald, Morphew, Verkade, & Muller-Reichert, 2007) are named
as such due to the thin, flexible membrane that makes the well of the carrier. Pressur-
ized hydraulic fluid transmits pressure to the sample through this membrane rather
than being applied directly; this avoids disrupting the sample and potentially damaging
it (McDonald et al., 2007). Membrane carriers are designed to freeze small-volume
specimens such as cell suspensions, embryos, and biopsies. Live cell carriers (Leica
Microsystems, Verkade, 2008) differ from membrane carriers in a number of ways. As
they are designed for imaging on a light microscope, they have a 1 mm opening in the
bottom of the well of the carrier to allow the transmission of light. The depth of the
well in the live cell carrier is 200 µm, compared to the 100 µm of the membrane carrier.
The inner diameter of the well is also slightly bigger at 1.51 mm instead of 1.5 mm.
This increased the well depth and the diameter of the well is necessary to accommodate
the finder grid that is used in the live cell clamp system (Verkade, 2008). An adverse
consequence of the well depth is that there is potentially a distance of 150 µm created
between the cells and the objective lens, the repercussions of which are discussed later.
In order to overcome this imaging issue, a new CLEM carrier was designed using
the thinner membrane carriers. The distance between cells and objective when using
membrane carriers is no more than 50 µm (see also Fig. 5(L)). Crucially, this increased
proximity allows the use of higher resolution objectives.
E. Preparation of a Modified CLEM Carrier

As the development of an improved CLEM carrier is still ongoing, we cannot pro-
vide a detailed protocol. What we will do in the following section is describe what
directions we have taken and what has worked and what has not. To adapt membrane
carriers (Fig. 5(G)) for CLEM, a hole is needed to be made in the bottom of the well.
Initial attempts to do this using a drill with a 1 mm drill bit were largely unsuccessful
as drilling through the center of the carrier proved problematic. The torque of the drill
bit touching the carrier caused it to jump, moving it off the center, and resulting in a
hole through the wall of the well rather than the bottom. Even after construction of a
jig to hold carriers in place while drilling, the numbers of unusable carriers generated
was unacceptably low. This was largely due to the burr generated on the inside of the
well, which meant that sapphire disks would no longer sit flat in the base. As an alter-
native solution, holes were made by piercing the membrane with a tapered histology
needle under a stereo microscope. By placing the carrier with the well facing upward
on a block of parafilm, the needle was passed through the center of the well until the
diameter of the hole was between 0.5 and –1 mm (Fig. 5(H)). Importantly, this did not
generate a burr on the inside of the well, although it did on the underside of the carrier.
This was quite easily removed, however, by simply running a razor blade across the
underside.
As the membrane carriers cannot accommodate a finder grid, another method of
securing the sapphire disk was needed. Several adhesives were trialed in order to
find one that was nontoxic to cells, capable of withstanding the pressure generated
by the HPF, nonsoluble in water or acetone, and easy to apply. The adhesives trialed
included EPON, dental wax, poly-L-lysine, super glue, araldite, and Loctite 350 UV
Fig. 5 The CLEM carriers. A to F show the steps in the loading process of the commercial “old” CLEM carrier. A Petri dish is filled with
CO2-independent medium and a CLEM carrier, finder grid, and the sapphire disk with cells grown on are placed in the dish (A and B).
The sapphire disk is loaded into the carrier (C). The finder grid is placed on top of the sapphire disk (D). The finder grid is wedged in
using the nonmagnetic tweezers (E). To check if the finder grid is secured, the complete carrier system is flipped over (in the medium).
The finder grid and sapphire disk should stay in place (F). To improve on the imaging quality, the thinner membrane carrier was modi-
fied as the “new” CLEM carrier (G–J). A membrane carrier is cut out of the carrier plate (see McDonald et al. 2007 and Brown et al.,
2010)(G). A hole is drilled in the middle (H). The sapphire disk is glued in place and overlaid with the finder grid (I). This is placed in
a carbon evaporator system and the finder pattern is carbon coated onto the sapphire disk (and carrier) (J). K shows a schematic repre-
sentation of the “old” CLEM carrier where due to the presence of the finder grid there is considerable distance between the cells and the
objective (see also M). In the “new” CLEM carrier there is no finder grid (L) and the distance of the cells to the objective is considerably
less (N) allowing the use of a higher NA 100× objective.
curing glue. Of all these, araldite and Loctite 350 performed the best. They caused
no noticeable toxic effects to cells; their viscosity meant that they could be easily
painted in the carrier and the curing methods allowed time to manipulate the sap-
phire disk into position. Sapphire disks always remained attached during cell cul-
ture and live cell imaging; however, detachment after HPF and freeze substitution
was unavoidable. When securing the sapphire disk in the well with the adhesive
(Fig. 5(H)), care was taken to ensure no air was trapped underneath the disk. If air
is present in an enclosed system, pressurization of the fluid within that system will
be diminished as the air is more easily compressed. The result in the case of HPF is
insufficient pressure on the sample generating inadequate freezing rates for optimal
preservation.
The finder grid in the live-cell clamp setup not only retained the sapphire disk but
also acted as the means of relocation of the cell of interest. Cell relocation with the
membrane carriers had to be achieved by another method, carbon coating of a finder
pattern directly on the sapphire disk. A similar technique is used by Vic Small’s lab
(Auinger & Small, 2008) although gold is used instead of carbon, and cells are chemi-
cally fixed rather than high-pressure frozen. Throughout sapphire disk attachment, car-
riers remained in the sheet as purchased from the manufacturer. The regular spacing
of carriers in this sheet meant multiple carriers could be coated simultaneously by the
use of a batch stencil. This was constructed by punching 1 mm holes of the same spac-
ing in filter paper and fixing finder grids over these holes. The stencil was then held in
place over the sheet of carriers with masking tape ensuring that each finder grid was
positioned directly over the center of the sapphire disk (Fig. 5(I)). This was then placed
under a carbon source in an Auto 306 evaporation system (Edwards) and carbon got
deposited for 10 s (Fig. 5(J)). The resulting carbon coat was found to flake off from
sapphire disks after 24 h in cell culture medium. As a result of personal communication
with Kent McDonald, the carbon finder pattern was baked on to the sapphire disks at
120°C overnight. A comparison of the new CLEM membrane carrier and the Live-cell
clamp is shown in Fig. 5(K–N).
When the finished CLEM membrane carriers were submerged in cell culture
medium, it consistently resulted in the incorporation of air bubbles in and around the
carrier. To avoid the aforementioned issues of insufficient pressure build up these were
removed. However, when working in a tissue culture hood without the use of a stereo
microscope this was not a trivial task. A sterile pipette tip and a 200 µl Gilson medium
were pulsed around the carrier while being held with fine forceps in order to dislodge
any trapped air.
Although the glues used gave no sign of toxicity from cells cultured in the pres-
ence of the polymerized form, the puncturing of membrane carriers potentially
exposes copper in the immediate vicinity of the cells. Leeching of copper ions into
solution has been shown to be toxic to epithelial cells (Hanagata et al., 2011); all
Leica carriers are made of copper due to its thermal conductivity but are coated with
gold to prevent toxic effects on cells. HeLa cells could be cultured on the punctured
membrane carriers for up to 5 days without showing visible signs of toxicity such as
rounding up.
F. Modification to Rapid Loader

The standard rapid loader as sold with the RTS for the EMPACT2 package (Leica
Microsystems) is designed to hold the live cell carriers. For this purpose, it has a recess
in the fork of the loader of 0.4 mm. This loader can still be used when freezing samples
in membrane carriers that have a thickness of 0.15 mm as the carrier falls to the base of
the recess. When using these carriers for CLEM, however, the top of the carrier must be
flushed with the top edge of the fork so that the cells are as close to the objective lens
as possible. To achieve this, a membrane carrier was carefully positioned at the desired
height in the fork and Araldite was injected into the recess created behind the carrier.
The Araldite was then hardened and the carrier cut away leaving a platform to easily
load membrane carriers at the correct height (Fig. 6(C)). Imaging with the standard
rapid loader was carried out on a coverslip (Fig. 6(A)). The drawback with this method
is that only a droplet of medium can be used while imaging as too large a volume would
overflow. In addition, the capillary forces draw fluid away from the coverslip onto the
stage insert. Imaging can therefore only be performed for short periods of time to guard
against the medium and sample drying out. We tried to overcome these issues by taking
out part of the straight neck of the rapid loader and enlarging the imaging hole of the
stage insert to fit a glass-bottomed imaging dish (Brown et al., 2009). A better-designed
rapid loader dedicated for CLEM experiments was presented around the same time
(McDonald, 2009) and given to us as a kind gift from Paul de George from Marine Reef
International. This new loader has a bridged neck (Fig. 6(B)) rather than being flat;
this permits the use of glass-bottomed imaging dishes that can be filled with sufficient
medium to allow long periods of imaging. This modified rapid loader is now commer-
cially available through Marine Reef International (www.marinereef.com).
G. Improvement of Image Acquisition

The above-mentioned modifications were made with the idea that they would
improve imaging properties, most importantly, the image quality. We therefore com-
pared the standard live-cell clamps and the new CLEM membrane carriers in live cell
imaging experiments. Cells were allowed to internalize EGF-Alexa488-10 nm gold and
Tfn-Alexa594-5 nm gold for 5 min before washing with PBS and imaging on a spinning
disk confocal microscope (dedicated for live cell imaging)(Fig. 7).
The focal plane of the cells in live cell clamps cannot be achieved with a 100× oil-
immersion lens, instead a 63× glycerol-immersion lens must be used. The increased
proximity of cells to the objective with the CLEM membrane carriers allows the use of
a 100× oil-immersion lens. The longer exposure times needed to detect adequate signal
from the probes with the live cell clamps resulted in a higher signal-to-noise ratio than
the CLEM membrane carrier. With the 100× lens, it is possible to resolve Tfn positive
tubule forming from early endocytic compartments (boxed region in Fig. 7(A)). With
the 63× lens, however, it is much harder to resolve individual punctate, and tubules are
below the limit of resolution (Fig. 7(B)). As generation and pinching off of tubules are
the main mechanism for sorting in early endosomes, the inability to visualize them in
the LM means sorting events cannot be captured with HPF-CLEM.
Fig. 6 The CLEM stage inserts. (A) shows the original setup as it is currently commercially available. The
standard rapid loader holding the CLEM carrier rests on a glass coverslip and a drop of medium is added to this
“open” system. In the new setup (B), the modified dedicated CLEM rapid loader bridges into a glass-bottomed
imaging dish that is filled with imaging medium. We have adapted the CLEM rapid loader by filling the fork
with a bit of Epon (brown area in C) to make sure that the new CLEM carrier is as close to the coverslip as
possible (C).
Fig. 7 Imaging properties of the new CLEM carrier. Using both the “new” and “old” CLEM carrier system,
live cell imaging data of the internalization of EGF and Transferrin were analyzed. Whereas we could not visu-
alize tubules using the “old” system (B), we could observe them using the “new” one (A).
H. Live Cell Imaging and HPF-CLEM

Having established that the modifications made were indeed improvements over
the existing system, we decided to apply these modifications in a CLEM experiment.
Before an experiment can be started, it needs a few logistical preparations, which are
listed as follows:
(1) Few days before
(a) Prepare the modified CLEM carrier.
(b) Sterilize the CLEM carrier. We use the UV lamp of the AFS2 for 5 h.
(2) One day before
(a) Seed the cells on the CLEM carrier.
(3) One hour before the experiment
(a) Fill the EMPACT2 with liquid nitrogen and the pressurizing fluid methyl-
cyclo-hexane.
(b) Move the EMPACT2 next to the light microscope, a spinning disk confocal in
our case (Fig. 8(C)), and insert the modified CLEM into the microscope stage.
(c) Warm the incubation chamber of the light microscope to 37°C.
(d) Prewarm the 20% BSA in CO2-independent medium containing 10% serum
to 37°C.
(4) The experiment
(a) The CLEM carrier was submerged in a 10 µl drop of EGF-Alexa488-10 nm
gold (100 ng/ml) and Tfn-Alexa594-5 nm gold (20 µg/ml) and incubated for
5 min at 37°C (Brown & Verkade, 2010).
(b) Wash cells briefly with PBS warmed to 37°C.
(c) Mount the CLEM carrier in the new CLEM rapid loader (Fig. 8(A)). As only
a small volume of liquid can be supported on the carrier, it was important that
the carrier was mounted quickly to prevent cells drying out.
(d) Submerge the end of CLEM rapid loader in an eppendorf tube containing 20%
BSA in CO2-independent medium containing 10% serum and transport to the
spinning disk confocal microscope.
(e) Place the loader on the modified stage (Fig. 8(B)) with the carrier in a glass-
bottomed dish containing 20% BSA in CO2-independent medium containing
10% serum.
(f) A field of view to image was selected. The criteria for selection of the cell or
cells were based on a normal morphology (i.e., not rounding up, blebbing,
etc.) and a detectable fluorescence signal from both probes.
(g) First, bright field images of the cells and their position on the carbon finder
grid were taken (see Fig. 10(A)).
(h) Acquire live fluorescence images of the green and red channels while moving
the focal plane through the volume of the cell. This allowed the identification
of an endosome containing both probes (Fig. 9(C) arrow) which would be the
main focus of the CLEM experiment. Images from green and red channels were
acquired continuously in the focal plane of the endosome with approximately 0.7
sec between each acquisition. One minute into the live cell imaging, the overlap-
ping fluorescent signals within the endosome begin to separate slightly (Fig. 9).
Fig. 8 Setup of the EMPACT2 + RTS next to the light microscope. After clamping the “new” CLEM
carrier in the fork of the modified (Brown et al., 2010) or dedicated rapid loader (A), the loader is placed
on the modified stage insert that fits a glass-bottomed imaging dish (B). The EMPACT2 + RTS is placed
right next to the light microscope that is going to be used for the live cell imaging to obtain the highest
time resolution between the last image from the light microscope and the point at which the sample is
frozen.
Fig. 9 Live cell imaging of endocytic segregation. Cells have internalized both the EGF-Alexa488-10 nm gold
(top row A) and the Tf-Alexa594-5 nm gold (middle row B) for 5 min before live cell imaging is started, the
final frames from this part of a HPF-CLEM experiment are shown (time until the sample is frozen is shown on
the bottom row C). The bottom row shows the merged image and at time point −4 the last frame is captured and
several red, green and yellow fluorescent structures are visible. (See color plate.)
(i) The CLEM rapid loader was quickly taken from the stage and entered into the
RTS of the EM PACT2 (Leica Microsystems) that was positioned close-by
(Fig. 8(C)). Sealing the specimen pod, firing the sample into the HPF cham-
ber, and freezing takes approximately 2.5 s. Combining this time with the
roughly 1.5 s it takes to transfer the rapid loader from the stage to the HPF, the
time from the last image acquired to the specimen being frozen is 4 s.
( j) The sample was stored in a 0.5 ml eppendorf tubes in a liquid nitrogen storage
dewar until further processing.
I. Sample Processing for EM

Frozen samples can be processed in a number of ways. Ideally, the sample would
be sectioned frozen-hydrated without any processing and imaged in a cryo-electron
microscope (CEMOVIS, Al-Amoudi et al., 2004). However, retracing a specific event
in this way is at the moment technically impossible. So, we process our samples via
freeze substitution to be embedded in a resin. This can be for instance Lowicryl HM20
that would allow for subsequent immunolabeling (van Weering et al., 2010). For most
experiments, we however embed our samples in Epon for which we have developed a
reasonably short protocol (Verkade, 2008)
(1) Fill a freeze substitution device (e.g., AFS2, Leica Microsystems) for about 1/3
with liquid nitrogen and set the temperature to −90°C.
(2) Thaw a Cryogenic vial (NUNC) containing a freeze substitution cocktail of 1%
(w/v) Osmium tetroxide and 0.1% (w/v) Uranyl acetate in acetone in a total
volume of 0.5 mL (pre-prepared and stored in a liquid nitrogen storage dewar).
(3) Pour the frozen sample from the storage eppendorf into a metal well or on a pre-
cooled metal plate in the AFS2.
(4) Carefully place the carrier in the bottom of the vial using fine forceps and orien-
tate, so that the cells are facing upward.
(5) Tighten the cap of the vial to prevent the evaporation of the freeze substitution
cocktail.
(6) The temperature inside the AFS is maintained at −90°C for 5 h before being raised
to 0°C at a constant rate over 18 h.
(7) At 0°C remove the cryovials containing the samples from the AFS and transfer to
a fume hood.
(8) Immediately pipette off the freeze substitution cocktail and replace with pure
acetone. Adding and removing liquid to and from the vial is done with care to
ensure the sapphire disk remains in the carrier and that the cells are not disturbed.
(9) After 1 min repeat leave for 5–10 min, and then repeat one more time.
(10) After the final acetone wash is removed, 25% (v/v) Epon (TAAB 182) in acetone
is added to the sample and incubated for 1 h at room temperature on a rocker.
(11) After 1 h this was replaced by 50% (v/v) Epon in acetone.
(12) After 1 h this was replaced by 75% (v/v) Epon in acetone.
(13) After 1 h this was replaced by 100% (v/v) Epon.
(14) Following the infiltration of Epon for 1–2 h, the carrier is transferred to a flow-
through ring and cup (Leica Microsystems), ensuring that the cells remained
facing up at all times (Fig. 10(A)). After a further hour of incubation on the
rocker at room temperature, the flow-through ring was placed in an oven at 60°C
and the resin was allowed to harden for 24 h.
After the embedding and hardening, the samples need to be processed and sectioned.
The plastic cup around the flow-through ring was cut away using a pair of cutting pli-
ers. Using the pointed edge of one of the hands of the pliers, the samples were punched
out of the flow-through ring. Excess resin was trimmed from around the carrier with
a razor blade and a line was scored in the resin just below the carrier. The carrier was
submerged in liquid nitrogen for 10 s before the blade of a detachment tool (Leica
Microsystems) heated to 60°C was used to separate the carrier from the resin (see Fig.
10(C) for the workflow). It must be noted that the detachment tool was developed for
the thicker carriers and not for the membrane carriers. This makes precise adjustment
of the resin block containing the modified CLEM carrier on the knife-edge necessary.
Removal of the carrier left the sapphire disk in the resin. Resin was trimmed from
around the sapphire disk that was then carefully lifted off the stub with a razor blade.
The carbon-coated finder grid was transferred from the sapphire onto the resin. The
position of the cell on the finder grid determined from bright field images (Fig. 10(A))
was used to trim the block to the area shown in Fig. 10(B). The block was then serial
sectioned with a thickness between 70 and 300 nm and collected on Pioloform-coated
EM slot grids. Care was taken while aligning the block face so that the entire face was
cut within the first few sections. This would facilitate correlation of the LM and EM
data. No additional staining was performed on the sections.
Fig. 10 Retracing the area of interest. From the carbon pattern deposited on the sapphire disk, it can be
observed that the cell and the event of interest were located on the right side of the number 6 (A). The same
number 6 can be found back as an imprint on the Epon block (B) and the block is then trimmed to a pyramid
containing the cell of interest. In order to obtain this smooth Epon surface with the cells exposed, the carrier and
sapphire disk have to be removed. This procedure is shown schematically in (C). After embedding in the flow-
through ring the CLEM carrier will have some resin on all sides (1), the resin from the top and sides is carefully
removed with a razor blade (2). The carrier (3) and subsequently the resin between the carrier and the sapphire
disk (4) are carefully removed. As a final step, the sapphire disk itself is lifted off the resin block (5), directly
exposing the bottom side of the embedded cells.
J. Electron Microscopy and Data Correlation

The positions of cells relative to one another in the bright field images (Fig. 11(A))
were used to identify the corresponding cells in a low-magnification EM micrograph
(Fig. 11(B)). Once the cell selected for live cell imaging had been identified, the sec-
tion/sections corresponding to the focal plane of live cell imaging were found. This was
achieved by scanning through serial sections of the same cell and making note of the
cell shape and distribution of endocytic compartments. This was then compared with
the images acquired from moving through the focal plane in the light microscope. When
the correct section was located, a higher magnification EM micrograph of the cell was
superimposed with the last frame of fluorescence imaging (Fig. 11(C)). Fluorescence over-
lays were created using Adobe Photoshop CS3 in a similar method to the one described
in Keene, Tufa, Lunstrum, Holden, & Horton, 2008. The LM image was first resized to
the same scale at the EM image. With the LM above the EM image, the opacity of the
LM image was reduced to 60%. By overexposing the fluorescence signal, using the levels
function, the boundary of the cell was more easily identified. The boundary of the cell in
the LM image was then aligned with that in the EM image by rotating and resizing. Loca-
tion of endocytic compartments was also used to confirm accurate alignment. Uncon-
strained resizing of the LM image was necessary, as the proportions of the cell seen in
EM had changed slightly to what was seen in LM due to compression during sectioning.
Further examination of the areas of the cell highlighted by the fluorescence superim-
position reveals endocytic compartments containing gold particles (Fig. 11(D)). These
compartments appear to be mainly early endocytic in nature, based on the number of intra-
luminal vesicles within (Mari et al., 2008). Also given that the time from ligand addition
to fixation was approximately 7–8 min in this example, it would be expected that ligands
would still be present in early endosomes. From these images, it can however already be
appreciated that the larger 10 nm gold particles tagged with EGF are mainly on one side
of the endocytic structures, whereas the smaller 5 nm gold particles bound with transferrin
are on the opposite site, closer to the plasma membrane. This localization corresponds well
with the fluorescent pattern of the last image from the live cell imaging data (see Fig. 11).
As a result of the point-spread function from the fluorescent signal, the axial resolu-
tion (by full width at half maximum) achievable by a spinning disk confocal is approxi-
mately 800 nm (Egner, Andresen, & Hell, 2002). That means potentially eleven 70 nm
sections could be contributing to the fluorescence seen from the focal plane of live cell
imaging. Looking at consecutive serial sections of the same endosome that appeared
to show separation in the LM, showed that this structure spanned at least 7 sections
(Fig. 12) making it at least 490 nm in the z-axis. There were two sizes of gold particle
associated with the membrane of this structure: the 10 nm gold from EGF-Alexa488-10
nm gold and the 5 nm gold from Tfn-Alexa594-5 nm gold. From the sections, it was
difficult to picture the 3D structure and the distribution of gold particles within the
endosome. It appeared to consist of a number of vesicular compartments; some of
which were obviously connected whereas other connections were indistinct. The only
way to fully appreciate and understand the complexity of the structure was to construct
a 3D model using electron tomography (Frank, 2006; Frey, Perkins & Ellisman, 2006;
McIntosh, Nicastro, & Mastronarde, 2005).
Fig. 11 Retracing the cell and event of interest. After the block has been trimmed and sectioned, a bright field
image (A) is taken and the pattern of cells is used to trace back the cell of interest in the EM (B). Identical cells
are numbered on both images. Our main interest was in cell number 6. When the fluorescent pattern of this cell
was overlaid on the EM image (C), it allows us to zoom into the different endocytic structures present in the
cells. Our particular interest was in the second structure on the bottom row. (See color plate.)
Fig. 12 Retracing a structure in the third dimension. As this particular sample was sectioned at 70 nm thick-
ness, the structure of interest can be observed in different shapes in a number of serial sections. The shape of
the structure changes dramatically in this Z-series, whereas it is captured as 1 fluorescent “blob” in the approxi-
mately 500–800 nm thick confocal section (see also 11C).
K. Tomography and Modeling

Tomography is normally performed on sections with a thickness between 200 and
300 nm. In our example, we have made use of 70 nm thick sections. The acquisition of
several tilt series and joining of the resulting tomograms were therefore required. Tilt
series of the serial sections were acquired on a Tecnai20, 200 kV Transmission Electron
Microscope (FEI company), with image acquisition for every 1° tilt angle between
±65°. A tilt series of the four uppermost sections (Fig. 12, bottom panel) of the endo-
some was acquired. A fold in the next consecutive section (Fig. 12, top right) prevented
acquisition at most tilt angles and was not further processed.
Although electron tomography is an essential part of our CLEM workflow, it is a
research field on its own, with its own technology developments and improvements
(Gan & Jensen, 2012). Here, we will only briefly outline the packages and strategy we
use for our applications. After acquisition of the tilt series on an FEI Tecnai20 using
the FEI tomography acquisition software, they were aligned using the FEI Explore 3D
software. These aligned series were then imported into the IMOD software (Kremer,
Mastronarde, & McIntosh, 1996) and four tomograms were constructed and subse-
quently joined together. Finally, the full tomogram was read into the Amira ResolveRT
software. The endosome membranes on each slice were traced (Fig. 13) and different
colors were used to give a better sense of perspective to the final model.
When all membrane and gold particles had been traced, the model was triangulated
and simplified (Fig. 13). The model reveals that the endosome is in fact far more inter-
connected than was first thought or visible. It is now evident that compartments that
appeared quite distinct from one another in a single serial section are actually part of the
same single structure. When looking at the endosome in the context of its surroundings
Fig. 13 Electron Tomography of the structure of interest. After reconstruction of the tomogram, the structure
of interest was modeled by hand. (A) shows part of the model placed on top of a slice of the tomogram. This
view highlights the vicinity of the endosome to the plasma membrane (white arrows). The endosome appears
to be directly connected to the plasma membrane via a membranous structure (in pink). (B) and (C) show indi-
vidual slices of the tomogram in which the membrane of the connection is visible (arrows). (See color plate.)
(Fig. 13), its proximity to the plasma membrane can be appreciated. In fact, there even
appears to be a tubular structure connecting the endosome to the plasma membrane
(Fig. 13). We must emphasize that the manual tracing and modeling within the tomog-
raphy process are highly subjective and that since this is only one example, the presence
of a direct connection between an endosome and the plasma membrane may be intrigu-
ing but for the moment is very speculative.
III. Outlook
What these studies have shown us is that it is the combination of a number of

microscopy techniques that is indeed more than just the sum of each individual tech-
nique. Live LM showed us how dynamic the system is and allowed us to choose a
specific event in time to study at higher resolution. The electron microscope allowed
us to analyze that the fluorescent “blob” is actually a structure consisting of a number
of sub compartments. It also allowed us to see that the labeled structure is very close
to the plasma membrane. But it was only by using electron tomography that we could
determine that the subcompartments visible in one section are actually part of one large
interconnected structure that may even be directly connected to the plasma membrane.
It is clear that this is a very promising approach that we will apply to different questions
within the intracellular trafficking field.
In order to extract even more data from these methods, there is however still a lot of
scope for improvements
(1) There is a need for a CLEM probe that is biosynthetic and compatible with
cryofixation. This will allow us to study the adaptor proteins located on the
cytoplasmic side of the membrane which are involved in the membrane tubu-
lation process.
(2) By using automated tracking and analysis, we should be able to analyze more
than one event from a sample. Rather than analyzing a phenomenon, quan-
titation is a key factor nowadays. Although by just looking for numbers we
may miss the special cases.
(3) A more automated (objective) approach to model electron tomography data
has been a wish within the tomography community for a long time. Espe-
cially with the growing need for analyzing several data sets, it is not doable
to perform this all by hand.
Acknowledgments
The authors would like to thank Debbie Carter and Gini Tilly of the Electron
Microscopy Unit of the Wolfson Bioimaging Facility and Kati Jepson, Alan Leard, and
Mark Jepson of the Light Microscopy Unit of the Wolfson Bioimaging Facility for their
excellent support throughout these studies
References
Al-Amoudi, A., Chang, J. J., Leforestier, A., McDowall, A., Salamin, L. M., Norlen, L. P., et al. (2004). Cryo-
electron microscopy of vitreous sections. The EMBO Journal, 23, 3583–3588.
Auinger, S., & Small, J. V. (2008). Correlated light and electron microscopy of the cytoskeleton. Methods in
244, 91–97.
high-pressure freezing and correlative light electron microscopy. Seminars in Developmental Biology,
20, 910–919.
theory and practice. Journal of Electron Microscopy Technique, 13, 165–174.
Egner, A., Andresen, V., & Hell, S. W. (2002). Comparison of the axial resolution of practical Nipkow-disk
confocal fluorescence microscopy with that of multifocal multiphoton microscopy: theory and experi-
ment. Journal of Microscopy, 206, 24–32.
Frank, J. (2006). Electron tomography: Methods for three-dimensional visualization of structures in the cell.
Berlin Heidelberg: Springer-Verlag.
Frey, T. G., Perkins, G. A., & Ellisman, M. H. (2006). Electron tomography of membrane-bound cellular
organelles. Annual Review of Biophysics and Biomolecular Structure, 35, 199–224.
Gan, L., & Jensen, G. J. (2012). Electron tomography of cells. Quarterly Reviews of Biophysics, 45,
27–56.
Gorden, P., Carpentier, J. L., Cohen, S., & Orci, L. (1978). Epidermal growth factor: morphological demon-
stration of binding, internalization, and lysosomal association in human fibroblasts. Proceedings of the
Haigler, H. T., McKanna, J. A., & Cohen, S. (1979). Direct visualization of the binding and internalization
of a ferritin conjugate of epidermal growth factor in human carcinoma cells A-431. The Journal of Cell
Biology, 81, 382–395.
Hanagata, N., Zhuang, F., Connolly, S., Li, J., Ogawa, N., & Xu, M. (2011). Molecular responses of human
lung epithelial cells to the toxicity of copper oxide nanoparticles inferred from whole genome expression
analysis. ACS Nano, 5, 9326–9338.
Hyde, G. K., Lancelle, S., Kepler, P. K., & Hardham, A. R. (1991). Freeze substitution reveals a new model
for sporangial cleavage in phytophtora, a result with implications for cytokinesis in other eukaryotes.
Journal of Cell Science, 100, 735–746.
Jaiswal, J. K., & Simon, S. M. (2004). Potentials and pitfalls of fluorescent quantum dots for biological imaging.
Trends in Cell Biology, 14, 497–504.
Jiménez, N., Humbel, B. M., Van Donselaar, E., Verkleij, A. J., & Burger, K. N.J. (2006). Aclar discs: a ver-
satile substrate for routine high-pressure freezing of mammalian cell monolayers. Journal of Microscopy,
221, 216–223.
Keene, D. R., Tufa, S. F., Lunstrum, G. P., Holden, P., & Horton, W. A. (2008). Confocal/TEM overlay
microscopy: a simple method for correlating confocal and electron microscopy of cells expressing GFP/
YFP fusion proteins. Microscopy and Microanalysis, 14, 342–348.
Kellenberger, E., Johansen, R., Maeder, M., Bohrmann, B., Stauffer, E., & Villiger, W. (1992). Artefacts and
morphological changes during chemical fixation. Journal of Microscopy, 168, 181–201.
Leonard, D., Hayakawa, A., Lawe, D., Lambright, D., Bellve, K. D., Standley, C., et al. (2008). Sorting of
EGF and transferrin at the plasma membrane and by cargo-specific signaling to EEA1-enriched endo-
somes. Journal of Cell Science, 121, 3445–3458.
Mari, M., Bujny, M. V., Zeuschner, D., Geerts, W. J., Griffith, J., Petersen, C. M., et al. (2008). SNX1
defines an early endosomal recycling exit for sortilin and mannose 6-phosphate receptors. Traffic, 9,
380–393.
McDonald, K., Morphew, M., Verkade, P., & Muller-Reichert, T. (2007). Recent advances in high-pressure
freezing: equipment- and specimen-loading methods. Methods in Molecular Biology, 369, 143–173.
McIntosh, R., Nicastro, D., & Mastronarde, D. (2005). New views of cells in 3D: an introduction to electron
tomography. Trends in Cell Biology, 15, 43–51.
Mersey, B. G., & McCully, M. E. (1978). Monitoring of the course of fixation in plant cells. Journal of
Microscopy, 114, 49–76.
Moor, H. (1987). In R. A. Steinbrecht, & K. Zierold (Eds.), Cryo-techniques in biological electron micros-
copy (pp. 175–191). Berlin Heidelberg: Springer-Verlag.
Morphew, M., He, W., Bjorkman, P. J., & McIntosh, P. R. (2008). Silver enhancement of nanogold particles
during freeze substitution for electron microscopy. Journal of Microscopy, 230, 263–267.
Murk, J. L., Posthuma, G., Koster, A. J., Geuze, H. J., Verkleij, A. J., Kleijmeer, M. J., et al. (2003). Influence
of aldehyde fixation on the morphology of endosomes and lysosomes: quantitative analysis and electron
tomography. Journal of Microscopy, 212, 81–90.
Penttila, A., Kalimo, H., & Trump, B. F. (1974). Influence of glutaraldehyde and-or osmium tetroxide on cell
volume, ion content, mechanical stability, and membrane permeability of Ehrlich ascites tumor cells. The
Perkins, G. A., Sun, M. G., & Frey, T. G. (2009). Chapter 2 correlated light and electron microscopy/electron
tomography of mitochondria in situ. Methods in Enzymology, 456, 29–52.
Rosenbluth, J. (1963). Contrast between osmium-fixed and permanganate-fixed toad spinal ganglia. The
Rosenmund, C., & Stevens, C. F. (1997). The rate of aldehyde fixation of the exocytotic machinery in cul-
tured hippocampal synapses. Journal of Neuroscience Methods, 76, 1–5.
Sandborn, E. B., & Bendayan, M. (1996). The existence of a circulatory system in animal cells. Cell Vision,
3, 180–190.
Biology, 94, e1001041. doi:10.1371/journal.pbio.1001041.
Small, J. V. (1981). Organization of actin in the leading edge of cultured cells: influence of osmium tetroxide
and dehydration on the ultrastructure of actin meshworks. The Journal of Cell Biology, 91, 695–705.
Stoscheck, C. M., & Carpenter, G. (1984). Down regulation of epidermal growth factor receptors: direct
demonstration of receptor degradation in human fibroblasts. The Journal of Cell Biology, 98, 1048–1053.
Tormey JMcD. (1964). Differences in membrane configuration between osmium tetroxide-fixed and glutar-
aldehyde-fixed ciliary epithelium. The Journal of Cell Biology, 23, 658–664.
Van Weering, J., Brown, E., Mantell, J., Sharp, T., & Verkade, P. (2010). Studying membrane traffic in high
resolution. Methods in Cell Biology, 96, 619–648. Electron Microscopy of Model Systems.
Wilson, T. P., Canny, M. J., McCuly, M. E., & Lefkovitch, L. P. (1990). Breakdown of cytoplasmic vacuoles.
A model of endoplasmic membrane rearrangement. Protoplasma, 155, 144–152.
CHAPTER 11
Targeted Ultramicrotomy: A Valuable

Tool for Correlated Light and Electron
Microscopy of Small Model Organisms
Irina Kolotuev*, Daniel J. Bumbarger†, Michel Labouesse‡
and Yannick Schwab‡
* Institut
de Génétique et Développement de Rennes, UMR 6290 CNRS, Université Rennes 1, Faculté de
Medecine/1. Fédération de Recherche BIOSIT, UMS 3480 Université de Rennes 1, Campus santé. 2 avenue
du Pr Leon Bernard CS34317, 35043 Rennes Cedex, France
† Department of Evolutionary Biology, Max Planck Institute for Developmental Biology, Spemannstrasse
37/IV, Tübingen 72076, Germany
‡ Institut de Génétique et de Biologie Cellulaire et Moléculaire, 1 rue Laurent Fries, 67404 Illkirch Cedex,
France
Abstract
I. Introduction
II. Rationale
III. Methods
A. Preparation of the Specimen for LM
B. High-Pressure Freezing
C. Freeze Substitution and Flat Embedding
D. Targeted Ultramicrotomy
E. Sectioning
F. Electron Microscopy
A. Organizing the Bench for LMP Agarose Embedding
B. Confocal Imaging
C. High-Pressure Freezing and Freeze Substitution
D. Flat Embedding
E. Laser Carving
F. Serial Sectioning
G. Transmission Electron Microscopy
V. Discussion
Acknowledgments
References

Copyright 2012, Elsevier Inc. All rights reserved. 203 http://dx.doi.org/10.1016/B978-0-12-416026-2.00011-X
204 Irina Kolotuev et al.
Abstract
Correlative light and electron microscopy (CLEM) is used when one needs to com-
bine both imaging modalities on the same sample. When working on living small model
organisms, such as Caenorhabditis elegans, specific CLEM protocols are required to
acquire high-resolution light microscopic images of a region of interest and thereafter
to relocate and study the same object at the ultrastructural level using a transmission
electron microscope. In this chapter, we describe how to process living specimens from
the confocal microscope to the transmission electron microscopy (TEM), focusing on
an improved ultramicrotomy technique that allows a precise and reliable targeting of
the object of interest. This improvement significantly reduces the time consuming and
frequently frustrating search for the region of interest. Our targeted ultramicrotomy
protocol is versatile enough to be applied on a variety of bulk specimens, such as fly
and fish embryos, or mouse tissues.
I. Introduction
Correlative light and electron microscopy (CLEM) is a “family” of techniques for

combining on the same object, data obtained from a light microscope with data acquired
with an electron microscope. Both imaging techniques have multiple variations. There-
fore, the precise definition of CLEM differs between studies, models, and labs. In
this chapter, we describe the approach for image acquisition of living Caenorhabditis
elegans larvae and embryos by light microscopy (LM) followed by observation of
the same regions using ultrastructural analysis by transmission electron microscopy
(TEM).
CLEM is an imaging tool that combines the powers of LM with those of electron
microscopy (EM). LM offers access to a wide range of sample sizes varying from
hundreds of nanometers to tens of millimeters. Thanks to a large diversity of record-
ing chambers that allow controlling the local environment, imaging living samples is
also feasible. It is therefore possible to obtain functional data from any living sample,
from the simplest unicellular model organism to the more complex vertebrates. Elec-
tron microscopy, and especially TEM, offers increased spatial resolution, reaching the
nanometer range when working on biological specimens. It is a unique technique in
cell biology especially when it is combined with the use of affinity markers, to localize
molecules.
The concept of CLEM is not new and was introduced in the 1960s and 1970s for
research on cultured cells (Price, 1974) and the nervous system (for a review see Modla
& Czymmek, 2011). The use of LM to identify regions for further TEM analysis is a
routine procedure in any EM lab. Inspecting a semithin section stained with toluidine
blue is a useful screening procedure prior to any thin sectioning with an ultramicrotome.
This is, in itself, a kind of correlative approach.
The breakthroughs provided by antibody-based immunocytochemistry and engi-
neering of fluorescent proteins have given a wide range of options for correlative
11. Targeted Ultramicrotomy 205
microscopy. One approach is to combine immunofluorescence with immunogold label-

ing, either on consecutive sections, or simultaneously on the same section using bifunc-
tional probes (Albrecht, Seulberger, Schwarz, & Risau, 1990; Micheva & Smith, 2007;
Powell, Halsey, & Hainfeld, 1998). Another strategy is to use GFP fluorescence which
is then correlated with TEM after immunodetection or photoconversion (Grabenbauer
et al., 2005; Knott, Holtmaat, Trachtenberg, Svoboda, & Welker, 2009). Interestingly,
some recent studies have shown that the fluorescence of several probes can even be
retained during TEM sample preparation, and can still be detected in cryosections
(Sartori et al., 2007; Schwartz, Sarbash, Ataullakhanov, McIntosh, & Nicastro, 2007)
or in plastic sections (Kukulski et al., 2011; Nixon et al., 2009; Watanabe et al., 2010).
One of the major challenges to be able to target the TEM study on a specific region
is the addition of reference points to the sample. When working on adherent cultured
cells, external reference coordinates are often used to track the precise position of the
cell of interest (McDonald, 2009; Polishchuk et al., 2000; Sartori et al., 2007; Spiegel-
halter et al., 2010; van Rijnsoever, Oorschot, & Klumperman, 2008; Verkade, 2008).
However, because of their size, similar approaches cannot be efficiently applied to
multicellular organisms such as C. elegans, Drosophila, zebrafish, or rodents. Specific
strategies have therefore been developed, relying mainly on the specimen’s anatomy to
relocalize the targeted region.
A number of correlative studies have been carried out on C. elegans embryos
(Muller-Reichert, Srayko, Hyman, O’Toole, & McDonald, 2007; Sims & Hardin,
2007; Watanabe et al., 2010). Specifically, the strategy described here has been applied
for the study of C. elegans excretory canal defects and serves as an example of recent
technical developments (Kolotuev, Schwab, & Labouesse, 2009). We describe the steps
required for confocal microscopy imaging of living samples, followed by cryofixation
by high-pressure freezing (HPF), freeze substitution (FS), and flat embedding. We give
a detailed explanation on the technique of laser-etched landmarks that we use to pre-
cisely and rapidly find the region of interest by ultramicrotomy, a process that can be
named “targeted ultramicrotomy.”
II. Rationale
To carry out a correlative approach on a small model organism, the following criti-
cal points have to be addressed in the workflow: (a) finding the appropriate conditions
for LM imaging on living specimen, (b) selecting the most efficient way to fix the
specimen for further processing in EM, (c) being able to image the embedded blocks,
and (d) finally to concentrate the ultramicrotomy effort on the region that contains the
object of interest. The techniques should be adaptable to various sample types, includ-
ing the different developmental stages of C. elegans (i.e., embryos, larvae, or adults),
but also to flies, fish embryos, or mouse tissues. One of the key steps in our approach
is the immobilization of the specimen in low-melting point agarose, which allows live
imaging followed by freezing using HPF. Another crucial point is flat embedding in
resin, which facilitates postfixation analysis and measurement of the animal’s position.
The addition of precise landmarks on the surface of the block allows precise identifica-
tion of the regions of interest (ROI) after ultramicrotomy. As a result, the time spent
for localizing the ROI is significantly improved, from several hours to less than 1 h.
III. Methods
A. Preparation of the Specimen for LM

Working with small model organisms, such as C. elegans, often involves imag-
ing single living specimens. Live recordings then allow the dynamic and functional
exploration of samples. The organisms can be selected according to their phenotype,
their developmental stage, or even specific treatments, such as single-cell ablation or
acute injuries. The first step in a correlative LM/EM analysis is to optimize the imag-
ing by LM. For living specimens, the specimen must be sufficiently immobile to allow
the use of high magnification objectives. The strategy used here is similar to that used
in labs working on zebrafish embryos. It consists in immobilization of the organisms
by anesthesia, followed by embedding in low melting point (LMP) agarose. The main
improvement presented here is to make these embedded specimens compatible with
further processing for freezing prior to EM. This entails embedding in agarose to a
thickness compatible with high-pressure freezing (i.e., thickness <200 µm and diame-
ter of 1.5 mm), a technique of choice for the cryofixation of nematodes (Hall, Hartwieg,
& Nguyen, 2012; Muller-Reichert, Hohenberg, O’Toole, & McDonald, 2003; Muller-
Reichert, Mantler, Srayko, & O’Toole, 2008; Weimer, 2006).
1. LMP Agarose Embedding

This embedding is performed between a bottom sheet of 200 µm and a top sheet of
50 µm Aclar film. A spacer of 200 µm Aclar is used to prevent undesirable compres-
sion of the material (Fig. 1(A)). Worms (typically 10–20) are deposited on the bottom
sheet in a small drop (approximately 2 µl) of buffer (egg buffer for embryos; M9 buf-
fer supplemented with anesthetics for larvae and adults; see Section IV for buffers
Fig. 1 LMP agarose flat embedding. (A) A glass slide is topped with a piece of Aclar and an Aclar spacer with
a rectangular hole in the middle. Animals are transferred into a 2-µl drop to the bottom Aclar. A 30 µl drop of
LMP agarose is applied on the slide. After mixing both drops, the top Aclar and a glass slide are placed on top
of the agarose to spread the drop. (B) After agarose gelling, the layout is gradually dismounted. (C) Round
pieces containing the sample are made with a biopsy punch precisely fitting the HPF carrier. From Kolotuev
et al.(2009).
composition). Anesthesia prevents worms escaping from the agarose pad, even after
embedding. A 30 µl drop of warm 4% LMP agarose is pipetted using a prewarmed tip
and deposited close to the worms. The worm suspension is then rapidly mixed with
the agarose and the sandwich closed by adding the top Aclar film. This step leads to
a spreading of the samples across the surface of the bottom sheet. After 10–20 s, the
agarose solidifies and the top sheet can be removed (Fig. 1(B)). A biopsy punch is
then used to isolate a round slab of agarose containing embedded worms (Fig. 1(C)).
The inner diameter of the slab (1.5 mm) matches quite precisely with the size of HPF
carriers that will be used after the imaging using LM. If manipulated gently and kept
wet, these discs can preserve the specimen for a prolonged period. For example, adult
worms have been kept embedded for more than 60 min, after which they were still
able to move, feed, and lay eggs. Note that when stored for a long time in buffer, the
agarose discs may have a tendency to swell and therefore could exceed the dimensions
of the HPF carriers. Therefore, if prolonged observation is required, it may be neces-
sary to make thinner slabs with reduced diameters. The transfer from the dissecting
microscope to the confocal microscope is performed by sliding the discs on the tip of a
scalpel blade with an eyelash (see below).
2. Confocal Acquisition
The confocal imaging is performed on an inverted microscope. To allow the use of
an oil immersion objective, the samples are mounted on a glass slide and covered by
a 50 µm thick Aclar sheet. Another 50 µm thick Aclar sheet is used as a spacer to cre-
ate a recording chamber (1 × 1 cm) (Fig. 2(A)). Aclar is flexible and facilitates sample
retrieval after imaging. However, if glass coverslips are used, the samples can undergo
shearing and compression. It is important that the thickness of the spacer is smaller than
the actual thickness of the agarose round piece, as this prevents movement during the
imaging. Nevertheless, special care should be taken to avoid compressing the sample.
Before closing the recording chamber, a drop of M9 buffer is added on top of the round
piece. The montage is then flipped upside down and transferred to the stage of the
inverted microscope for confocal analysis (Fig. 2(B)).
B. High-Pressure Freezing
We have used the EMPACT2 from Leica Microsystems to freeze worms using this
method. We next describe the transfer from LM to the freezing machine. It is important
that all of the equipment is installed close to each other to minimize the time between
each step.
1. Transfer Confocal––EMPACT2
The LM recording chamber is removed from the stage of the confocal, and the top
layer of Aclar is then dismounted carefully. As the agarose disc can either stick to the
top or to the bottom Aclar sheet, it is important to check its position while dismounting
the sandwich. With the help of an eyelash, the sample is gently pushed to the edge of a
scalpel blade (Fig. 2(C)) and transferred to the freezing carrier (Fig. 2(D)).
Fig. 2 Recording chamber and manipulation of the agarose embedded samples. (A) The round pieces of aga-
rose are positioned inside a recording chamber made of a glass slide, topped with a 50 µm thick Aclar spacer.
(B) The chamber is closed with another sheet of Aclar and tilted upside down for recording on an inverted
confocal microscope. (C) The agarose pieces are manipulated with an eyelash and carried on a razor blade. The
round pieces of agarose fit into the HPF carriers. Modified from Kolotuev et al. (2009).
2. Freezing
Prior to the freezing, the carriers are pretreated with 1% phosphatidylcholine
(diluted in chloroform). This coating will improve the success rate of sample removal
during the FS. An empty HPF carrier is installed onto the rapid loader before the LM
imaging and prefilled with hexadecene, an external cryo-protectant commonly used in
HPF. The sample is then fitted inside the carrier and frozen. With training, the whole
transfer, from the light microscope to the freezing, can take less than 30 s. For better
time-resolved experiments, further modifications should be considered, as mentioned
in the discussion.
C. Freeze Substitution and Flat Embedding

1. Freeze Substitution
A large variety of protocols have been described for the FS of high-pressure frozen
nematodes. We have selected two, one adapted for morphological analysis and the
other for immunocytochemistry.
For morphological analysis, the FS mix contains 2% osmium tetroxide, 0.25% ura-
nyl acetate, and 0.25% glutaraldehyde in acetone. As described previously (Buser &
Walther, 2008), a small percentage of water is added to enhance the contrast of mem-
branes. From the start the samples are placed in microtubes that can be closed with a
hermetic cap to prevent evaporation. The tubes are prefilled with 0.5 ml of FS mix and
cooled to liquid nitrogen temperature. The HPF samples are then transferred on top of
Table I
Comparison of embedding protocol using either Epoxy resin or HM20.
Dehydration Embedding in epoxy resin Embedding in HM20
−90°C 48 h −90°C 48 h
Fixation Ramp 5°C/h Ramp 5°C/h

−30°C 12 h −50°C 12 h
Rinses −30°C Acetone; 2 h −50°C Ethanol; 2 h
Infiltration −30°C Resin 25%; 2 h −50°C Resin 25%; 3 h
−30°C Resin 50%; 2 h −50°C Resin 50%; 3 h
−30°C Resin 75%; 3 h −50°C Resin 75% ; 16 h
−50°C Resin 100% ; 3 h
Ramp 5°C/h
Room temperature resin 100%; 3 h −50°C Resin 100%; 3 h
Room temperature resin 100%; 3 h −50°C Resin 100%; 3 h
Polymerization 60°C 48 h −50°C; UV light 48 h
Ramp 5°C/h
Room temp. 24 h
the frozen mix and placed in the chamber of the FS unit that is precooled to −90°C.
After 10–15 min, the mix melts and the samples sink to the bottom of the tube. It is
important here to check for their proper orientation and to make sure the samples are
positioned upside-up so they are exposed to the FS mix.
For immunocytochemical analysis, the chemical fixation after the dehydration step
is performed at −50°C, a temperature at which the acetone does not evaporate. Open
molds can therefore be used to process up to 10 different carriers simultaneously. The
FS cocktail consists of 0.1% of uranyl acetate diluted in acetone. The mold is prefilled
with 1 ml of the cocktail and cooled to liquid nitrogen temperature. The samples are
then distributed to each slot of the mold and transferred to the FS unit. When the tem-
perature of the mix reaches −90°C, the acetone melts and the samples fall to the bot-
tom. The volume of mix is then topped to 3 ml. The detailed protocols for the FS are
shown in Table I.
2. Removing the Specimen from the HPF Carrier and Flat Embedding
Prior to the final steps of the embedding, the samples are removed from the HPF
carriers. This is performed to allow a further LM examination of the embedded worms
after the resin polymerization. For the morphology protocol, the samples are removed
from the carriers at room temperature in the cavity of a porcelain plate under a dissect-
ing microscope. The carriers are manipulated with a pair of fine forceps, and the sample
is pushed out by applying a slight pressure or rotary motion on its side. Provided the
carriers have been precoated with phosphatidylcholine, the samples generally come out
easily. For the immunocytochemistry protocol, removal of the sample is difficult as it
must be performed at low temperature. This is done either inside the chamber of the FS
Fig. 3 Flat embedding after high-pressure freezing and FS. (A) For embedding in Lowicryl, a modified ver-
sion of the flat embedding molds (Leica) is used to reduce the block thickness to 500 µm. (B) When embedded
in epoxy resins, a sandwich made of two Aclar sheets and of one space Aclar is used. The resulting blocks are
200 µm thick. From Kolotuev et al. (2009).
unit, using appropriate tools such as long forceps and eyelashes mounted on long wood
sticks, or inside the chamber of a cryo-ultramicrotome.
When the infiltration steps in resin are completed, the samples are prepared for flat
embedding (Fig. 3). For Lowicryl embedding, a modified version of the flat embed-
ding molds (Leica Microsystems) is used. This consists of a cup vial and a thinned
inner piece (Fig. 3(A)). The inner piece is supplied at a thickness of 3 mm which leads
to thick resin blocks in which the samples are not clearly visible in a standard light
microscope (due to light diffraction in the transmitted mode). Therefore, the embed-
ding molds are filed to a thickness of 0.5 mm with a milling machine. To obtain a
smooth block surface, two round pieces of Aclar are placed at the bottom and on top of
the mold (Fig. 3(A)). For Epon embedding, the samples are placed between two sheets
of Aclar separated by a spacer also made of Aclar. All these sheets are 200 µm thick
(Fig. 3(B)). The resins are polymerized either by heat (60°C) for epoxy resins or by UV
light illumination for Lowicryls.
D. Targeted Ultramicrotomy
1. ROI Marking by Laser Etching
The flat blocks are removed from their molds and placed on a glass slide. For epoxy
embedded samples, the removal of the Aclar sheets should be performed gently and not
too long after the end of the polymerization step. This precaution avoids any fractures
extending through sample region, where the resin can be less dense and more fragile.
A laser microdissecting device is then used to etch landmarks on the block surface.
We have used the LMD6000 from Leica, which consists of an upright microscope
equipped with a pulsed UV laser. It is important at this step to create a box that is
aligned to the main axis of the worm or embryo, which will later be used as a guide
to set the alignment of the diamond knife. If several samples were embedded together,
Fig. 4 Sample localization and orientation for targeted ultramicrotomy. (A) The surface of the block is carved
with a pulsed UV laser to create landmarks aligned to the samples. (B and C) The use of a grid in the eyepiece
of the ultramicrotome allows aligning the trimming diamond to the landmarks.
as shown in Fig. 4, several boxes can each be defined with alignments appropriate for
each specimen of interest.
2. Measurements and Positioning

The marked blocks are then inspected under a light microscope and pictures of the
region of interest are taken in order to include the landmarks in the frame. Measure-
ments are performed on the images, using a software such as ImageJ (Abramoff, Mag-
alhaes, & Ram, 2004) to estimate the distance between the landmark and the sample
(Fig. 4(A)). Although the laser-carved line is rather thick (around 10 µm) and not
present in the same focus plane as the sample, it can still be used to estimate the posi-
tion of the ROI. Subsequently, a more accurate approach is utilized, as described in
Section D.3.
To start the trimming, the block is mounted in an ultramicrotome, in the sample
holder specifically designed for flat specimens (Fig. 4(C)). The block is set horizontal,
with the carved surface facing up to allow the observation of the landmarks. By tilting
the sample holder, the carved face is brought perpendicular to the optical path of the
binocular. A gridded eyepiece mounted on the binocular of the ultramicrotome is then
aligned to the marks (Fig. 4(B)) and serves to align a trimming diamond parallel to
the desired cutting plan (Fig. 4(C)). The sample is tilted back to its horizontal position
before starting the trimming of its face. The 200 µm thick block is extremely fragile
when trimmed horizontally. For this reason, the trimming of the face is performed in
two steps (Fig. 5). First, a reduced width (<500 µm) of the block is trimmed to create a
surface which is parallel to the desired cutting orientation (Fig. 5(A)). In doing so, the
amount of resin removed at each turn is not too important. After rotating the sample
by 90° the previously trimmed face is used to align the block vertically (Fig. 5(B)).
Trimming in this orientation allows rapid trimming because the resistance of the block
to the cutting action is high. When using a grid in the optics of the ultramicrotome, it
is possible to evaluate the amount of resin to trim to reach the carved landmarks. The
block is then rotated back to its horizontal position (Fig. 5(C)) to trim its two sides. All
the subsequent ultramicrotomy steps are done with the block in its vertical position
(Fig. 6(A)).
Fig. 5 Trimming of the flat embedded blocks. The trimming of the block face is processed in two steps.
(A) First, a reduced surface (yellow) is trimmed when the block is set horizontal. (B) Second, the block is
rotated 90° and realigned to bring the previously trimmed surface to a vertical position. The trimming of the
face is then finished to reach the front of the carved landmark. (C) The block is then rotated 90° to its horizontal
position to trim its sides, using the beveled edges of the trimming diamond. (See color plate.)
Fig. 6 Determination of the sample origin. (A) The block is set vertical to cut serial thick sections that are
stained with toluidine blue and inspected with a light microscope. (B) The first section that shows the sample
defines the origin used for further determination of the cutting depth. Modified from Kolotuev et al. (2009).
3. Precise Trimming and Determination of an Internal Landmark

As mentioned previously, the laser etching produces thick lines that can vary from
5 to 10 µm. It is therefore difficult to estimate with precision when the trimming reaches
the front landmark. To improve the precision in reaching a specific ROI, an origin of
measurement, intrinsic to the sample, should be found. For C. elegans larvae or adults,
this can be the anterior tip of the animal. For embryos, this can be any position along
the eggshell. The rough trimming of the block is therefore stopped at least 5 µm before
the estimated position of the origin, which is then sought by collecting and inspecting
thick sections (Fig. 6(A)). Making serial thick sections through a 5–20 µm depth is not
Fig. 7 Ultramicrotome setup for improved serial sectioning. (A) Microscope arms are used to bring an ionizer
and an eyelash in the working space of the ultramicrotome. (B) Closer view of the ionizer. (C) The eyelash is
mounted on a micromanipulator to precisely reach the ribbon of sections.
a tedious task and can be performed rapidly. The thick sections are collected on glass
slides, stained with toluidine blue, and inspected under a light microscope. In case of
whole worms, the first section that shows the worm’s tip represents the origin of mea-
surements for further position determination of the ROIs (Fig. 6(A)). The feed of the
ultramicrotome is then used to keep track of the progression of the sectioning, down
to the targeted region. Note that the precision of measurements is determined by the
thickness set to find the origin. If 500 nm thick sections were used, the precision should
then be 500 nm.
E. Sectioning
1. Serial Sectioning
To obtain large serial section datasets (Bumbarger, Crum, Ellisman, & Baldwin,
2006; Bumbarger, Crum, Ellisman, & Baldwin, 2007; Bumbarger et al., 2009), care
must be taken to minimize the loss of sections. An efficient workstation should be set
up to reduce movements during grid handling (Fig. 7(A)). The microtome should be in
a room where air currents, including those from heating and cooling systems, will not
disturb section ribbons. If airflow is a problem, a Plexiglas cabinet can be built around
the microtome.
A special consideration for larger series of sections is that it is desirable to fit as
many sections as possible on each grid. This reduces the amount of grid handling,
and thus opportunities for section loss, and will later speed up image acquisition. The
height of the block face is usually trimmed very short, about 100 µm, and the leading
and trailing edges are trimmed parallel to each other at approximately a 60° angle with
respect to the block face. The resulting sections should take up most of the width of the
grid slot. This will make ribbons easier to handle and will help stabilize the grid film.
The sides are trimmed close to perpendicular to the block face in order to minimize the
number of times the face will have to be recut.
Two strategies can help sections adhere to each other in a ribbon. A small amount
of cyanoacrylate adhesive can be applied to the leading and trailing edges of the block
face. Alternatively, when forming the block face, the final cuts can be done with a
slightly, but not overly, dull razor blade. The resulting rough surface causes sections to
adhere to each other more effectively.
An ionizer, such as the Diatome Static Line II, is mounted a few centimeters
away from the edge of the diamond knife to prevent the buildup of charges that can
induce wrinkles and other problems (Fig. 7(B)). Consideration should be given to
section thickness. Thinner sections will give more z-resolution, but will take longer
to image for a given volume, and the likelihood of section loss is higher. Thicker
sections will be more robust, but the consequence of a single missing section can
be more problematic. A thickness of 50–70 nm seems to be a good compromise for
many projects.
For large serial section datasets, it is useful to keep a record of each section and its
interference color as it comes off the knife. The purpose is to identify the location and
size of gaps in the dataset, as well as to identify sections if the ribbon breaks prior to
being fixed on the film. The information on the lost sections is precisely recorded in
order to have a complete track of sectioning information.
When the ultramicrotome is started, the first cut section will typically be thinner
than subsequent sections. Therefore, longer ribbons are cut if possible, to avoid the
possible loss, or thickness variations, of sections between each cutting session. The
number of sections in the ribbon is determined primarily by the size of the block face
as well as by the skill of the microtomist. Ribbons are separated into segments that
are shorter than the length of the slot in the grid (usually 2 mm). This can be per-
formed by gently pressing an eyelash probe onto the gap between the two sections.
If it does not break after a few sessions of light pressure, move the eyelash to a dif-
ferent spot of the same gap and try again. Pig eyelashes are more rigid than human
eyelashes, and thus make ideal probes. They can either be purchased premounted to
a handle or can be obtained in large numbers from a butcher, washed and glued to a
handle. Delicate and precise movement of the eyelash can be achieved by attaching
the probe to a micromanipulator mounted on a microscope arm that can be swung
in and out of position (Fig. 7(C)). With such a setup, sections are seldom lost while
breaking the ribbon.
2. Support Film on Metallic Slides and Sections Pickup

When working with small specimen, such as C. elegans, it is often important to
avoid using mesh grids, as the probability to get a bar on the ROI is high. This is why
slot grids are very often used for such samples. To collect the thin sections, the slot
grids are usually covered by a support film, such as pioloform. Another strategy has
been developed, where the support film is prepared on a metallic stage (Abad, 1988).
We have used a modified version of this stage by preparing a metallic slide perforated
by holes of 4 mm in diameter (Fig. 8). On the same day of sectioning, empty wide-slot
grids are glow-discharged to ensure they are not hydrophobic and a small amount of
adhesive is applied to ensure a proper adherence to the film. This is done either with a
Coat-Quick “G” grid coating pen or with adhesive made by dissolving Scotch® tape in
acetone. The sections are then picked up from the knife using an empty slot grid as a
loop. They stay trapped on a drop of water inside the slot, and can be deposited on the
support film (Fig. 8(A)). The water is then evaporated in an oven (set to 50°C), allowing
the grid and the sections to attach to the film (Fig. 8(B)).
This technique has the advantage to first improve the preparation procedures of
the pioloform coating, but also to decrease the occurrence of folds during the pickup
procedure. This can be extremely important when dealing with serial sections, were no
section loss is tolerated.
The grids are removed gently from the pioloform-coated metallic slide by poking
the film at their edge. Sections are then stained with uranyl acetate and lead citrate.
When studying serial sections to model structures in 3D, deposition of a carbon layer
will improve the stability of the sections under the beam of the electron microscope.
F. Electron Microscopy
1. ROI Localization
The use of the targeted ultramicrotomy protocol is efficient for finding rare or dis-
crete structures in a complex sample, such as C. elegans. When structures can be seen
with a light microscope, the estimation of the ROI position, relative to anterior tip of
Fig. 8 Pickup procedure for serial sections. (A) The sections are picked up on bare slot grids and put on top
of a pioloform film mounted on a homemade metallic slide. (B) Evaporation of the water (at 50°C in an oven)
causes the grids and the sections to adhere to the support film.
Fig. 9 Calculations of the distances and angles. ImageJ and Adobe Photoshop software are used for evaluating
the ROI position within the sample. When a fluorescent signal is the target, the major calculations for the ROI
finding are based on the localization of the fluorescent signal in live animals, and transposed to the carved resin
block. From Kolotuev et al. (2009).
the animal, serves as a guide for the ultramicrotomy. Previously, our approach has been
applied to study the formation of cysts in the excretory canal of mutant C. elegans
embryos or larvae (Kolotuev et al., 2009). In this example, the abnormalities were
clearly visible using a transmitted light microscope equipped with phase contrast. The
use of fluorescent lines is also of great interest for highlighting specimen anatomy.
For example, the pore forming cells of the excretory canal are clearly detectable in the
che-14::GFP strain (Fig. 9). A three-dimensional map of the worm is then built from
confocal Z-stacking, and serves to relocate the target structure.
There are numerous applications where this procedure for ROI localization can be
used to focus an EM study on a specific region of a given sample (i.e., specific cell type,
mosaic expression of proteins, etc.).
2. TEM Analysis
When the region of interest is collected on sections for TEM, many kinds of analy-
ses can be performed, ranging from immunogold experiments to electron tomography.
In the example shown in Fig. 10, the sections containing the pore forming cells of the
excretory canals were treated for immunogold staining of VHA5, a subunit of the Vo
ATP-ase expressed in C. elegans epithelial cells. A standard protocol for immunostain-
ing was used, consisting of a blocking step, incubation with the primary antibody, and
a detection with a secondary antibody coupled to colloidal gold (for a detailed protocol
see Kolotuev et al., 2009).
A. Organizing the Bench for LMP Agarose Embedding

Instrumentation: Dissecting microscope with a bottom and a top illumination. Two
heater blocs, one set to 70°C to melt the LMP and the other to 37°C to keep the melted
agarose liquid and to prewarm the pipette tips used for the agarose deposit on the
Fig. 10 Precise localization of the excretory duct cell and immunogold labeling with the anti-VHA-5 anti-
body. (A) Confocal micrograph of an L4 stage larva expressing che-14::gfp. The arrow indicates the excretory
duct as visualized by che-14::gfp expression. Scale bar, 10 µm. (B) TEM micrograph of a transverse section
after staining with anti-VHA-5 antibody and gold labeling, showing the excretory duct cell and excretory canal
of the animal embedded in Lowicryl resin as shown in (A). VHA-5 localizes to the excretory canal lumen and
canaliculi (scale bar, 1 µm). (C) Inset of the region boxed in (B). Arrows point to the precise localization of the
VHA-5 antibody to membrane stacks of the duct cell. Scale bar, 0.2 µm. Modified from Kolotuev et al. (2009).
mounting slide. For this, we cut the first 2–3 mm of 1 ml pipettes tips to account for the
viscosity of the LMP.
Material: Aclar sheets 7.8 mil (198.12 µm EMS catalogue number #50425-25) and
2 mil (50.8 µm EMS catalogue number #50426-25––note that at present, EMS is no
longer distributing the 50 µm Aclar, but other companies such as Science Services
GmbH are proposing alternatives).
The Aclar sheets for embedding are prepared in advance: bottom part 7.8 mil 7.6 ×
2.6 cm––spacer 7.8 mil 7.6 × 2.6 cm with a window cut at 1 × 1 cm––top part 2 mil
7.6 × 2.6 cm.
Scalpel blades (type 10). Platinum-pick for worm manipulation. Biopsy punch of
1.5 mm diameter (Miltex––ref 33-31A).
Reagents: Egg buffer (NaCl, 118 mM; KCl, 48 mM; CaCl2, 2 mM; MgCl2, 2 mM;
Hepes 25 mM; pH 7.3); M9 buffer (KH2PO4, 22 mM; Na2HPO4 2H2O, 34 mM; NaCl,
86 mM; MgSO4 1 mM) (Mc-Carter et al., 1997); LMP agarose (Cat # 16520100, Invi-
trogen); anesthetic (0.1% tricaine, 0.01% tetramisole in M9 buffer).
B. Confocal Imaging
Instrumentation: Confocal microscope––Either a Leica SP2 AOBS or a Leica SP5
AOBS with a 10× PL APO 0.4 objective, or an oil immersion 63× PL APO 1.4 objec-
tive is used.
C. High-Pressure Freezing and Free Substitution

Instrumentation: HPF (Leica EMPACT2 equipped with the RTS) and FS unit (Leica
EM AFS).
Material: HPF carriers––flat carriers (Leica Microsystems ref 16706898) or mem-
brane carriers (Leica Microsystems ref 16707899), precoated with phosphatidylcho-
line; microtubes with cap (Sarstedt ref 72.694); disposable plastic pipettes. Embedding
molds made from a reagent bath (Leica Microsystems ref 16707154) and from a milled
flat embedding insert (Leica Microsystems ref 16707155).
Reagents: 1% Phosphatidylcholine (Sigma ref 61755) in chloroform. Hexadecene-1
92% (Sigma-Aldrich, ref H-7009). FS mix for morphological analysis consisting
of 2% osmium tetroxide (EMS), 0.25% uranyl acetate (Laurylab ref 4885113),
0.25% glutaraldehyde (EMS), and 1% distilled water in acetone. FS mix for immu-
nogold analysis: 0.1% uranyl acetate in acetone. Note that the uranyl acetate is
prepared from a 10% stock solution in methanol that is stored at −20°C. Epoxy
embedding medium: 45% Epon 812 substitute (Fluka ref 45345), 30% DDSA
(Fluka 45346), 23% MNA(Fluka 45347), and 2% DMP30 (Fluka 93331). Lowic-
ryl HM20 monostep (EMS ref 14345).
D. Flat Embedding
Instrumentation: Oven set to 60°C. UV lamp for the AFS (Leica EMUV ref 702727).
Material: Aclar® sheets for Epon embedding: top and bottom part 7.8 mil 7.6 × 2.6
cm, spacer 7.8 mil 7.6 × 2.6 cm with an inner window of 6 × 1 cm. Glass slides. Aclar
round pieces for Lowicryl embedding (Leica Microsystems ref 16707156).
E. Laser Carving
Instrumentation: Laser microdissection microscope Leica LMD6000 equipped with
UV pulsed laser, 50 µJ, 80 MHz. The 10× (Plan fluotar 0.25) and the 20× (N Plan 0.4)
objective lenses were used to prepare the landmarks.
Material: Glass slide
F. Serial Sectioning
Instrumentation: Ultramicrotome Leica UCT, Diatome Static Line II, Märzhäuser
MM33 micromanipulator, and Edwards coating unit EM12E4.
Material: Cryotrim (Diatome), diamond knife ultra35 (Diatome), pioloform-coated
copper rhodium slot grids (EMS CR2010), fine forceps, and pig eyelashes mounted on
a handle.
G. Transmission Electron Microscopy

Instrumentation: Transmission electron miscoscope CM12 (FEI), operated at 80 kV.
Orius 1000 CCD camera from Gatan, controlled by the acquisition software Digital
Micrograph.
Reagents: For section staining, 5% uranyl acetate in water, lead citrate, and double-
distilled water.
V. Discussion
In the example described in this chapter, the CLEM approach was used to find
precise locations along the body axis of an adult nematode, that is, the cells forming
the pore of the excretory canal in C. elegans (Kolotuev et al., 2009). The LM was per-
formed on a living specimen with a setup that allows high-resolution imaging using a
confocal microscope. Using the targeted ultramicrotomy protocol, relying on the mark-
ing of flat-embedded worms, the precision for reaching the ROI was better than 1 µm.
Furthermore, the work effort was considerably reduced, compared to other commonly
used and time-consuming serial-sectioning strategies.
We have applied this protocol to other invertebrate specimens, such as Drosophila
embryos (Kolotuev et al., 2009), wings or wing imaginal discs (unpublished), as well
as vertebrate tissue, such as zebrafish embryos (unpublished) and slices of mouse brain
(Rezai et al., 2012). While the flat-embedding and the production of landmarks on the
block face were applied the same way for all these sample types, the fixation of the
specimens had to be adapted, as not all specimens are compatible with HPF. HPF is
the technique of choice for C. elegans samples (Favre et al., 1995; Hall et al., 2012;
Muller-Reichert et al., 2003; Muller-Reichert, Mancuso, Lich, & McDonald, 2010), but
is more difficult for fly, zebrafish, or mouse tissues due to their biochemical composi-
tion or size. Conventional chemical fixation protocols have therefore been applied in
the CLEM procedure.
When performing a CLEM experiment on living organisms, the goal is often to
capture dynamic events. In the present work, a delay close to 30 s between the last
LM image and the fixation of the sample via HPF has been achieved. For now, the
best time-resolved CLEM trial performed on bulk specimens has been described for
C. elegans embryos with a delay of 5 s (Muller-Reichert et al., 2007). This represents
a technical limitation at the moment, as the available HPF machines cannot process
the samples faster. A delay in sample transfer in the same order has been reported for
cultured cells (Spiegelhalter et al., 2010; Verkade, 2008). In the future, new freezing
machines with faster transfer systems have then to be developed to be able to capture
events occurring in the millisecond range.
Currently, the precision of the ROI localization in CLEM experiments has reached
the sub-micrometric scale (Bishop et al., 2011; Kolotuev et al., 2009), which is a break-
through when looking for specific structures in the volume of a bulk specimen. But this
targeting is difficult, if the ROI has no characteristic shape or appearance. This was not
the case in our study on the pore of the excretory canal (Kolotuev et al., 2009), neither
in works focused on dendritic spines (Bishop et al., 2011; Knott et al., 2009). Recent
work has shown that the fluorescence of some reporter proteins can be preserved in
resins after embedding (Kukulski et al., 2011; Nixon et al., 2009; Watanabe et al.,
2010). The inspection of the resin sections with LM can then be used to focus the EM
study on the fluorescent spots with high precision, either by using bifunctional fiducial
markers (Kukulski et al., 2011) or by tracking the fluorescence with super-resolution
microscopes (Watanabe et al., 2010). Similar strategies should now be adaptable to 3D
volumes to achieve the same precision directly on the embedded specimen, and not
only on sections.
The targeted ultramicrotomy strategy that is described in this chapter relies on mark-
ing the resins blocks with a microdissection apparatus. The use of lasers to add land-
marks on a sample for TEM has also been described when working with adherent
cultured cells (Colombelli et al., 2008; Spiegelhalter et al., 2010). More recently, the
same principle has been used on fixed tissues (Bishop et al., 2011). In this study, a near-
IR laser was applied to create landmarks in the volume of the specimen. Combined
with a photo-conversion step, the electron-dense landmarks were used to localize the
targeted region in the depth of the resin block while inspecting serial thick and thin sec-
tions. The power of this technique is that it reliably allows a precision of about 1 µm in
ROI search and opens up to the possibility to register the subcellular volumes acquired
with and electron microscope to the 3D confocal data.
The current protocol, together with these technical improvements for ROI reloca-
tion, helps to address the growing need for correlative microscopy in intact cells and
tissues. The automation of serial EM acquisitions provided by recently developed
microscopes (see Hall et al., 2012 for their application to C. elegans) offers a great
potential. It can be expected that in a near future, the yield and the reliability of the
volume acquisition in the EM will be greatly improved and will lead to efficient cor-
relations in 3D, in complex multicellular specimen.
Acknowledgments
We would like to thank Manuela d’Alessandro, Vincent Hyenne, and John Lucocq
for critical reading of the manuscript, the members of the IGBMC Imaging Center for
support in LM and EM experiments, and Serge Taubert from the IGMBC workshop for
help in designing the tools for sample preparation.
References
Abad, A. (1988). A study of section wrinkling on single-hole coated grids using TEM and SEM. Journal of
Electron Microscopy Technique, 8(Feb (2)):217–222. PubMed PMID: 3246609.
Abramoff, M. D., Magalhaes, P. J., & Ram, S. J. (2004). Image processing with Image. Journal of Biophoton-
ics International, 11, 36–42.
Albrecht, U., Seulberger, H., Schwarz, H., & Risau, W. (1990). Correlation of blood-brain barrier function
and HT7 protein distribution in chick brain circumventricular organs. Brain Research, 535, 49–61.
Bishop, D., Nikic, I., Brinkoetter, M., Knecht, S., Potz, S., Kerschensteiner, M., et al. (2011). Near-infrared
branding efficiently correlates light and electron microscopy. Nature Methods, 8, 568–570.
Bumbarger, D. J., Crum, J., Ellisman, M. H., & Baldwin, J. G. (2006). Three-dimensional reconstruction of
the nose epidermal cells in the microbial feeding nematode, Acrobeles complexus (Nematoda: Rhabdi-
tida). Journal of Morphologyl, 267, 1257–1272.
Bumbarger, D. J., Crum, J., Ellisman, M. H., & Baldwin, J. G. (2007). Three-dimensional fine structural
reconstruction of the nose sensory structures of Acrobeles complexus compared to Caenorhabditis
elegans (Nematoda: Rhabditida). Journal of Morphology, 268, 649–663.
Bumbarger, D. J., Wijeratne, S., Carter, C., Crum, J., Ellisman, M. H., & Baldwin, J. G. (2009). Three-
dimensional reconstruction of the amphid sensilla in the microbial feeding nematode, Acrobeles com-
plexus (Nematoda: Rhabditida). The Journal of Comparative Neurology, 512, 271–281.
Buser, C., & Walther, P. (2008). Freeze-substitution: the addition of water to polar solvents enhances the reten-
tion of structure and acts at temperatures around −60 degrees C. Journal of Microscopy, 230, 268–277.
Colombelli, J., Tangemo, C., Haselman, U., Antony, C., Stelzer, E. H., Pepperkok, R., et al. (2008). A cor-
relative light and electron microscopy method based on laser micropatterning and etching. Methods in
Molecular Biology, 457, 203–213.
Favre, R., Hermann, R., Cermola, M., Hohenberg, H., Muller, M., & Bazzicalupo, P. (1995). Immuno-
gold-labelling of CUT-1, CUT-2 and cuticlin epitopes in Caenorhabditis elegans and Heterorhabditis sp.
processed by high pressure freezing and freeze-substitution. Journal of Submicroscopic Cytology and
Pathology, 27, 341–347.
Grabenbauer, M., Geerts, W. J., Fernadez-Rodriguez, J., Hoenger, A., Koster, A. J., & Nilsson, T. (2005). Cor-
relative microscopy and electron tomography of GFP through photooxidation. Nature Methods, 2, 857–862.
Hall, D. H., Hartwieg, E., & Nguyen, K. C. (2012). Modern electron microscopy methods for C. elegans.
Methods of Cell Biology, 107, 93–149.
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 micro-
scopic analysis. Nature Protocols, 4, 1145–1156.
and electron microscopy of small invertebrate organisms. Biology of Cell, 102, 121–132.
Kukulski, W., Schorb, M., Welsch, S., Picco, A., Kaksonen, M., & Briggs, J. A. (2011). Correlated fluores-
cence and 3D electron microscopy with high sensitivity and spatial precision. Journal of Cell Biology,
192, 111–119.
McCarter, J., Bartlett, B., Dang, T., & Schedl, T. (1997). Soma-germ cell interactions in Caenorhabditis
elegans: multiple events of hermaphrodite germline development require the somatic sheath and sperma-
thecal lineages. Developmental Biology, 181(Jan 15 (2)):121–143. PubMed PMID: 9013925.
Muller-Reichert, T., Hohenberg, H., O’Toole, E. T., & McDonald, K. (2003). Cryoimmobilization and three-
dimensional visualization of C. elegans ultrastructure. Journal of Microscopy, 212, 71–80.
Muller-Reichert, T., Srayko, M., Hyman, A., O’Toole, E. T., & McDonald, K. (2007). Correlative light and
101–119.
Muller-Reichert, T., Mantler, J., Srayko, M., & O’Toole, E. (2008). Electron microscopy of the early
Caenorhabditis elegans embryo. Journal of Microscopy, 230, 297–307.
Muller-Reichert, T., Mancuso, J., Lich, B., & McDonald, K. (2010). Three-dimensional reconstruction
methods for Caenorhabditis elegans ultrastructure. Methods in Cell Biology, 96, 331–361.
10, 131–136.
Polishchuk, R. S., Polishchuk, E. V., Marra, P., Alberti, S., Buccione, R., Luini, A., et al. (2000). Correlative
apparatus and plasma membrane. Journal of Cell Biology, 148, 45–58.
Powell, R. D., Halsey, C. M., & Hainfeld, J. F. (1998). Combined fluorescent and gold immunoprobes:
reagents and methods for correlative light and electron microscopy. Microscopy Research and Technique,
42, 2–12.
Price, Z. H. (1974). Correlative light and electron microscopy of single cultured cells. In M. A. Hayat (Ed.),
Principles and techniques of electron microscopy (Vol. 4, pp. 45–63). New York: Van Nostrand Reinhold.
Rezai, X., Faget, L., Bednarek, E., Schwab, Y., Kieffer, B. L., & Massotte, D. (2012). Mouse delta opioid
receptors are located on presynaptic afferents to hippocampal pyramidal cells. Cellular Molecular Neu-
robiology, 32(4), 509–516.
Schwartz, C. L., Sarbash, V. I., Ataullakhanov, F. I., McIntosh, J. R., & Nicastro, D. (2007). Cryo-fluorescence
microscopy facilitates correlations between light and cryo-electron microscopy and reduces the rate of
photobleaching. Journal of Microscopy, 227, 98–109.
369, 291–308.
Spiegelhalter, C., Tosch, V., Hentsch, D., Koch, M., Kessler, P., Schwab, Y., et al. (2010). From dynamic live
cell imaging to 3D ultrastructure: novel integrated methods for high pressure freezing and correlative
light-electron microscopy. PLoS One, 5(2), e9014.
Weimer, R. M. (2006). Preservation of C. elegans tissue via high-pressure freezing and freeze-substitution
for ultrastructural analysis and immunocytochemistry. Methods of Molecular Biology, 351, 203–221.
CHAPTER 12

of Intermediate Stages of Meiotic Spindle
Assembly in the Early Caenorhabditis elegans
Embryo
Ina Woog*, Silke White*, Mandy Büchner*, Martin Srayko†
and Thomas Müller-Reichert*
* Medical Theoretical Center (MTZ), Medical Faculty Carl Gustav Carus, University of Technology Dresden,
Fiedlerstraße 42, 01307 Dresden, Germany
† Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada
Abstract
I. Introduction
II. Methods
A. Staging of Isolated Meiotic Embryos by Light Microscopy
B. High-Pressure Freezing and Freeze-Substitution
C. Thin-Layer Embedding, Serial Sectioning, and Prescreening of Samples
D. Electron Tomography
III. Instrumentation and Material
A. Staging of Early Embryos
B. High-Pressure Freezing and Cryoprocessing of Staged Embryos
IV. Discussion
A. Live-Cell Imaging
B. Buffer Solution for Live-Cell Imaging and Freezing
C. Duration of Freeze-Substitution
D. Software for Electron Tomography
Acknowledgments
References

224 Ina Woog et al.
Abstract
This chapter is an update of the previously published book chapter “Correlative

Light and Electron Microscopy of Early C. elegans Embryos in Mitosis” (Müller-
Reichert, Srayko, Hyman, O’Toole, & McDonald, 2007). Here, we have adapted and
improved the protocol for the isolated meiotic embryos, which was necessary to meet
the specific challenges a researcher faces while investigating the development of very
early Caenorhabditis elegans embryos ex-utero. Due to the incompleteness of the egg-
shell assembly, the meiotic embryo is very fragile and much more susceptible to changes
in the environmental conditions than the mitotic ones. To avoid phototoxicity associated
with wide-field UV illumination, we stage the meiotic embryos primarily using trans-
mitted visible light. Throughout the staging and high-pressure freezing, we incubate
samples in an isotonic embryo buffer. The ex-utero approach allows precise tracking of
the developmental events in isolated meiotic embryos, thus facilitating the comparison
of structural features between wild-type and mutant or RNAi-treated samples.
I. Introduction
The early embryo of the roundworm Caenorhabditis elegans is perfectly suitable

for investigating early developmental events (Sonnichsen et al., 2005). In particular,
meiotic spindles assemble in a highly stereotypical manner, allowing the use of mutants
and RNA-mediated interference (RNAi) to detect subtle defects at various stages of the
process (Müller-Reichert, Greenan, O’Toole, & Srayko, 2010). In order to form a com-
plete understanding of the mechanism of spindle assembly, high-resolution imaging
and 3-D reconstructions are required for all stages of the assembly process. Compared
to the mitotic spindle assembly, little is known about the structure of the female meiotic
spindles in C. elegans and the proteins involved in spindle formation and chromosome
segregation. In part, this is due to the unavailability of a routine method of prepara-
tion for parallel light microscopy (LM)/ electron microscopy (EM) analysis of meiotic
embryos outside of the hermaphrodite uterus. When exposed to osmotic changes or
phototoxic light levels, the meiotic C. elegans embryo will abruptly halt in develop-
ment and this is the main reason why meiotic embryos in a previous study were ana-
lyzed by preparing whole worms for electron microscopy (Srayko, O’Toole, Hyman,
& Müller-Reichert, 2006). Using this approach, however, one has to section through
whole worms to get to the relevant embryo.
In meiosis, haploid gametes are produced from a diploid progenitor cell. However,
the process of genome haploidization differs between oogenesis and spermatogene-
sis. Using electron microscopy, acentrosomal anastral spindle assembly in C. elegans
female meiosis was first analyzed by Albertson and Thomson (1993). As the chromo-
somes condense, an array of microtubules first forms around the chromatin and then
becomes arranged in an elongated (“pointed”) bipolar spindle that is about 13 µm long
at early metaphase. This spindle then shortens to roughly 3–4 µm and adopts a barrel-
like shape, and this shortening is accompanied by a rotation of the spindle relative
12. Correlative Light and Electron Microscopy of Intermediate Stages of Meiotic Spindle assembly 225
to the surface of the embryo. The microtubule severing protein katanin is required
for shortening of microtubules during spindle formation and increasing the number
of short microtubules (Srayko et al., 2006). At anaphase, chromatin is separated to a
maximum of 3–4 µm with microtubules arranged between the chromosomes. Interest-
ingly, the holocentric kinetochores of the female meiotic chromatin seem to help ori-
ent meiotic chromosomes, but are not essential for chromosome segregation (Dumont,
Oegema, & Desai, 2010). The second meiotic division then follows a similar pattern
(Albertson & Thomson, 1993; McNally, Audhya, Oegema, & McNally, 2006).
Similar to other species including Drosophila, Xenopus, mice, and humans, the
female meiotic spindle in C. elegans forms without the participation of the centrosomes
(Manandhar, Schatten, & Sutovsky, 2005). Unlike male meiosis, only one quarter of the
genetic material of the oocyte is retained for karyogamy prior to mitosis, the rest is dis-
carded as polar bodies. In meiosis I, homologous chromosomes separate and one half
are extruded out of the embryo into the first polar body. The remaining chromosomes
undergo meiosis II, during which sister chromatids separate. After anaphase II, half of
Fig. 1 Schematic drawing of female and male meiosis in C. elegans. (A) Morphology and orientation of the
bipolar spindle in C. elegans female meiosis. Genome haploidization involves spindle rotation and shortening,
and “transformation” of the bipolar spindle into an “inside-out” one. Meiosis I and II basically follow the same
scheme. (B) Process of male meiosis in C. elegans. Genome haploidization gives rise to four equivalent haploid
sperm cells.
226 Ina Woog et al.
the chromatids are extruded into a second, slightly smaller polar body, leaving behind
a single copy of chromatids in the cytoplasm, which will later replicate and migrate
toward their sperm-derived counterparts in preparation for the first mitotic division
(Fig. 1(A); Fig. 2). Relatively little is known about the male meiosis in C. elegans.
Unlike the female spindle, the male meiotic spindle resembles that of a mitotic spindle,
with centrosomes at the poles. Separation of homologous chromosomes in meiosis I
and of sister chromatids in meiosis II gives rise to four equivalent haploid sperm cells,
each associated with a pair of centrioles (Fig. 1(B)).
Light microscopy of fluorescent protein-tagged C. elegans strains combined with
silencing of genes by RNAi has given new insights into the dynamics of meiotic as
well as mitotic spindle formation, kinetochore assembly, and chromosome separation
(Müller-Reichert, Müller-Reichert, et al., 2010). The limited resolution, however, is
the “weakness” of conventional LM. Electron microscopy (EM) on the other hand can
provide a high-resolution image of the structure of interest, but cannot deliver infor-
mation about the dynamic nature of a given subcellular structure. The combination
of both microscopic techniques, i.e. the application of correlative light and electron
microscopy (CLEM), is used to (i) combine contextual information from LM with the
resolution of EM, (ii) catch a dynamic process at a known point, (iii) locate a rare event
and/or structure, or (iv) increase EM sample size and throughput (McDonald, 2009).
CLEM is a perfect tool to better understand many biological processes in different
systems and it is remarkably powerful for embryonic studies in C. elegans as shown
previously (Kirkham, Müller-Reichert, Oegema, Grill, & Hyman, 2003; Pelletier,
O’Toole, Schwager, Hyman, & Müller-Reichert, 2006; Olsen et al., in preparation).
Here, we will concentrate on the special requirements for sample preparation related
to the analysis of early meiotic C. elegans embryos ex-utero and present an improved
protocol for staging and cryoimmobilization of isolated embryos.
II. Methods
A. Staging of Isolated Meiotic Embryos by Light Microscopy

For staging and high-pressure freezing, a “capillary tube” was used to collect meiotic
embryos and transfer them into the membrane carriers used for cryoimmobilization.
Cellulose microcapillaries were cut into pieces of about 2 cm length, placed into the
pointed end of a micropipette tip, and sealed with nail polish (Hohenberg, Mannweiler,
& Müller, 1994; Müller-Reichert et al., 2007). The tip holding the capillary was then
mounted on a Pipetman. The high transparency of the capillary tube allows observing
of cellular processes by LM prior to high-pressure freezing.
For staging and freezing, young adult worms of C. elegans strain expressed a fluo-
rescently tagged protein. Worms were picked into a 25 µl drop of minimal Edgar’s
growth medium (MEGM) on a high precision (0.17 mm) coverslip. MEGM is a spe-
cial embryonic cell medium, suitable for short-term culture of permeabilized embryos
(Edgar, 1995). Using a stereomicroscope, embryos were released by cutting a worm in
the gonad region with two hypodermic needles. A meiotic embryo was sucked into the
home-made “capillary device.” Unfortunately, while the pronuclei of mitotic one-cell
embryos are good indicators of specific mitotic stages, for meiotic embryos there are
no such clear “landmarks.” However, the shape and size of the embryo, the appearance
of the cytoplasm, the presence of an eggshell, and the presence or absence of a polar
body can give useful hints on the stages of the embryo. While this helps presorting the
embryos even under a transmitted light stereoscope, fluorescently tagged worm strains
were used in order to precisely stage the embryo with the fluorescence microscope prior
to freezing. After capturing a suitable embryo into the tube, the capillary was released
by cutting near the pipette tip. The ∼1 cm long capillary containing the embryo was
then cut to a final length of about 1 mm, thus being small enough to fit into the Leica
EMPACT2+RTS membrane carrier for subsequent freezing. In order to prevent the
embryo from floating out, the capillary was cut with the scalpel’s backside. This closed
the ends of the tube and sealed the capillary with the embryo inside. This method did not
always result in a properly closed capillary; however, with some luck the embryo can be
successfully transferred to the membrane carrier within an open capillary. Depending
on the time constraints of the experiment, it is often advantageous to attempt an open
capillary transfer for an embryo that is at the desired stage, rather than repeatedly trying
to seal the capillary by recutting.
Prior to transfer to the membrane carrier, the coverslip with the capillary tube in
media was mounted on the stage of an inverted light microscope equipped with an LED
light source and the meiotic development was observed using fluorescence. Shortly
before the desired stage was reached, a last image was taken before the sample was
rapidly frozen. This transfer takes about 10–30 s. The LM imaging setup was opti-
mized to avoid any kind of refractive index mismatch and thus ensure high quality
imaging. However, it has to be noted that unlike “normal” ex-utero imaging, where
the embryo sits freely in the medium on the coverslip (as for Fig. 2), the capillary in
which the embryo is embedded for the subsequent freezing distorts the light such that
the resulting image will be of slightly lower quality.
Fig. 2 Morphology and orientation of the bipolar spindles in C. elegans female meiosis as observed by fluo-
rescence light microscopy. The upper and lower panel shows spindles in female meiosis I and female meiosis II,
respectively. A C. elegans strain (MAS91) expressing GFP::β-tubulin (red) and mCherry::Histone (turquoise)
was used. Scale bar: 5 µm. (See color plate.)
228 Ina Woog et al.
Fig. 3 Sample preparation for correlative light microscopy and electron tomography. (A) Sucking of a pre-
selected embryo into a capillary tube. (B) Staging of the selected embryo with an inverted fluorescence light
microscope. (C) High-pressure freezing of the embryo. (D) Thin-layer embedding. (E) Preparation of serial
sections of a staged embryo on a coated slot grid (modified from Müller-Reichert et al., 2007).
B. High-Pressure Freezing and Freeze-Substitution

After reaching the desired developmental stage, the meiotic embryo is rapidly cry-
oimmobilized using a high-pressure freezer (Fig. 3). This physical fixation method
preserves the fine structural details of the embryo in its native state (McDonald, Mor-
phew, Verkade, & Müller-Reichert, 2007). As described in detail previously, we used
the rapid transfer system (RTS) of the Leica EMPACT2 high-pressure freezer. One of
the big advantages of this freezing machine is its portability, and can be easily moved
directly next to the light microscope (Verkade, 2008). This allows for minimal delay
times between staging and cryoimmobilization of the embryo. This method allows cor-
relative LM/EM experiments of C. elegans embryos at defined stages in early develop-
ment (McDonald et al., 2007).
Preparation of the meiotic embryo for EM closely resembles the already published
technique for the mitotic embryo (Müller-Reichert et al., 2007). Nevertheless, the key
steps are described in brief. First, the pod was connected to the bayonet and was placed
into the machine. The membrane carrier was premounted into the rapid loader (for
further details see McDonald et al., 2007; Verkade, 2008). Shortly before freezing,
MEGM with 20% BSA was filled into the membrane carrier and the tube contain-
ing the meiotic embryo in “pure” MEGM was transferred from the microscope to the
membrane carrier. MEGM containing 20% BSA is used as a filler and cryoprotectant
to avoid the formation of ice crystals during the freezing process. Care needs to be
taken not to let the BSA sit in the membrane carrier too long before loading the embryo
because it will harden. If there is BSA on the lip of the membrane carrier, it will harden
quickly and prevent a good seal of the carrier during freezing, resulting in a pressure
error. Pushing the rapid loader with the membrane carrier into the preloaded RTS initi-
ated freezing of the embryo. The carrier was then released from the pod and collected
in the LN2 bath.
For the freeze-substitution step, the membrane carriers including the samples were
put in precooled cryovials containing 1% OsO4 and 0.1% uranyl acetate in anhydrous
acetone (McDonald & Müller-Reichert, 2002). The cryovials were placed in a freeze
substitution machine (Leica) and the samples were left for 8–48 h at −90°C. The freeze-
substitution samples were allowed to warm up 5°C/h for 12 h. At −30°C, the samples
were maintained for further 5 h and after that the temperature was increased again
from 5°C/ h to 0°C. Following three washing steps with fresh anhydrous acetone, the
samples were gradually infiltrated with Epon/Araldite (3:1 for 1 h, 1:1 for 2 h, 1:3 for
2 h, 100% resin for 1 h, overnight and another hour).

Cleaned glass slides were dipped into a Teflon solution. After drying, the slides were
polished for better transparency. Two folded sheets of Parafilm were used as spacers
at each end of a glass slide. A drop resin was given onto the slide and the membrane
carrier containing the capillary was placed into it. The tube was then taken out of the
membrane carrier using needles and a stereomicroscope. A second coated and cleaned
glass slide was placed on the first one and the resin was polymerized for 48 h at 60°C.
After polymerization, the slides were separated using a razor blade. In order to relocate
the embryo prior to sectioning, the location of the embryo in the thin layer of resin was
marked using the diamond tip of an objective marker. At this point, it is recommended
to sketch the position of the embryo in the resin.
Using a scalpel, small cubes of 1 × 1 mm2 resin containing the capillary tube were
cut out under a stereomicroscope and pasted onto the dummy block (Müller-Reichert
et al., 2007). The glue was left to polymerize at 60°C for 30 min and the mounted sam-
ple was then trimmed to a trapezoid block for serial sectioning. For subsequent tomog-
raphy, ribbons of semithick (300 nm) sections were collected on FORMVAR-coated
copper slot grids. At this step, it is absolutely crucial to note the order of the ribbons on
the EM grids (for a detailed protocol for serial sectioning see Müller-Reichert, Man-
cuso, Lich, & McDonald, 2010). The post-staining of the sections was routinely carried
out using 2% uranyl acetate in 70% methanol followed by Reynold’s lead citrate. For
subsequent electron tomography, the stained grids were dipped into a droplet of 15 nm
gold solution for 1 min, left to air dry for a minute, and then shortly washed in ddH2O.
To identify and document regions of interest for subsequent electron tomography, the
stained sections with attached fiducial gold markers were screened using an electron
microscope operated at 80 kV.
The regions of interest were recorded at different tilts to compute a tomogram
and generate a 3-D model. Samples were placed in an intermediate voltage (300 kV)
electron microscope equipped with a eucentric tilting stage. Serial, tilted views were
collected over ±60° range at 1° increments. To collect a larger area, 2 × 2 montages
were acquired (Marsh, Mastronarde, Buttle, Howell, & McIntosh, 2001; O’Toole et al.,
2003). After the first tilt series was obtained, the grid was rotated by 90° and a sec-
ond tilt series was acquired over the same range. Calculations of the tomograms were
done using the IMOD software package (eTomo) (Mastronarde, 1997). The software
packages IMOD and AMIRA were used for displaying and modeling of meiotic fea-
tures (Fig. 4), such as microtubules and chromatin (Kremer, Mastronarde, & McIntosh,
1996; Weber et al., 2011).
230 Ina Woog et al.
Fig. 4 Electron tomography and 3-D modeling of a meiotic spindle in an isolated C. elegans embryo. (A)
Semithick (300 nm) section showing a meiotic spindle in early anaphase during the first meiotic division at
low magnification. (B) Corresponding 3-D model illustrating microtubules (pink lines), the surface of the chro-
mosomes (blue), and microtubule ends (closed, green spheres; open, white spheres). Scale bars: 1 µm (A) and
0.5 µm (B). (See color plate.)
III. Instrumentation and Material
A. Staging of Early Embryos

Instrumentation: Stereomicroscope with white light source, light microscope
equipped with epifluorescence (e.g., a Zeiss Observer Z1 with a 40×/1.2 W C-Apochromat
objective, Axiocam Mrm 16-bit digital camera controlled by Axiovision Software,
time-lapse intervals of 15–20 s), 470 nm and 590 nm LED as fluorescence light source.
Materials: C. elegans strain XA3501 expressing GFP::β-tubulin and GFP::Histone
to visualize microtubules and chromatin. C. elegans strain MAS91 expressing GFP::
β-tubulin and mCherry::Histone to distinguish microtubules and chromatin, dialysis tub-
ing with an inner diameter of 200 µm (Leica Microsystems), a micropipettor (0.5–10 µl
size), 25 gauge hypodermic needles, high precision coverslips (0.17 mm thickness,
24 × 60 mm), No. 23 scalpel tip, pipette tips, nail polish, fine tipped tweezers.
Reagents: Worm medium MEGM (Edgar, 1995): 0.125 mM polyvinylpyrrolidone
(Sigma P0930); 1 mM inulin (Sigma I3754); 58.8 mM NaCl; 25.2 mM KCl; 25 mM
HEPES (pH 7.4); 0.55 mM galactose; 0.96 mM l-glutamine; 0.022 µl/ml pyruvic acid
(Sigma 107360); 10 µl/ml lactate syrup (Sigma L4263); 10 µl/ml penicillin–streptomycin
(Sigma P3539); 400 µl/ml FBS.
B. High-Pressure Freezing and Cryoprocessing of Staged Embryos

Instrumentation: Stereomicroscope with white light source, high-pressure freezer
(EM PACT2+RTS, Leica Microsystems), freeze-substitution machine (Leica EM AFS,
Leica Microsystems)
Materials: Membrane carriers (100 µm deep)
Reagents: MEGM + 20 % BSA, anhydrous acetone (EM grade), osmium tetroxide,
uranyl acetate, Ep on-Araldite

Instrumentation: Objective marker with a diamond tip (Leica), ultramicrotome
(Leica UCT, Leica Microsystems), electron microscope operated at 80 kV (LEO 906,
Zeiss) equipped with a compustage.
Materials: Glass slides, Parafilm®, “dummy” blocks.
Reagents: Teflon® solution (MS-143V TFE Release Agent, Miller-Stephenson
Chemical Co., Inc., Danbury, CT, USA), Epon/Araldite, fast glue, Formvar, copper slot
grids, methanol, uranyl acetate, lead citrate, 15 nm colloidal gold (BBInternational).
Instrumentation: Intermediate-voltage electron microscope operated at 300 kV (we
use a TECNAI TF30 FEG, FEI, Eindhoven, the Netherlands), high-tilt rotating stage
(Gatan model 650, Pleasanton, CA), 2k × 2k CCD camera (Gatan), image capture soft-
ware package (SerialEM), and 3-D reconstruction software (IMOD), modeling soft-
ware (Amira & IMOD).
Materials: Fine tipped tweezers.
IV. Discussion
The following steps of the described correlative LM/EM approach for the isolated
C. elegans embryos were discussed in depth previously: (1) the advantages of using
capillary tubes, (ii) the use of BSA as a filler and cryoprotectant, (iii) high-pressure
freezing with the EMPACT2+RTS, (iv) serial sectioning for large-scale reconstruc-
tions, and (v) electron tomography for 3-D-modeling of the microtubule apparatus
(Müller-Reichert et al., 2007; Müller-Reichert, Mancuso, et al., 2010). Because most of
the steps of sample preparation, i.e. high-pressure freezing, thin-layer embedding and
serial sectioning, are almost identical for both mitotic and meiotic embryos, we will
briefly discuss mainly the challenges of live-cell imaging of early developing embryos
with incomplete eggshell assembly.
A. Live-Cell Imaging
Live-cell microscopy of meiotic C. elegans embryos differs from imaging mitotic
cells. It is difficult to define the meiotic stage via bright field contrasting techniques,
such as DIC, as it is routinely done for mitotic embryos (Müller-Reichert et al., 2007).
The meiotic spindle can usually be observed as a “clearing” in the yolk granule-filled
cytoplasm. It is difficult, however, to distinguish the various stages of spindle assembly
reliably using this method alone. Only with the help of fluorescence microcopy can
the meiotic embryo be staged precisely. This not only demands worm lines that stably
express the desired fluorescent markers, it also poses a threat to the viability of the early
embryo. The early meiotic C. elegans embryo is very fragile due to the incompleteness
232 Ina Woog et al.
of the eggshell assembly, which makes it susceptible to both osmotic changes and phys-
ical disruption. In addition, imaging fluorophores adds a risk of photo toxicity resulting
from the generation of free radicals in the cytoplasm. To handle the osmotic sensitivity
of the early meiotic embryo, an embryonic cell medium developed by Edgar was used
for imaging blastomere cultures ex-utero. This minimal embryonic growth medium
(MEGM) has been used previously for short-term culture of early embryos by provid-
ing sufficient osmotic conditions (Edgar, 1995). To minimize the disruption of embry-
onic development due to UV irradiation, we use light emitting diodes (LEDs) as a light
source, which emit discrete wavelengths and thus use less total energy to achieve the
desired excitation of the fluorophore. Moreover, we empirically determine the lowest
excitation energy (exposure time and LED power) that will allow reasonable imaging
results without disrupting the phenotype or cell cycle progression of control embryos.
Another practical consideration for live-cell imaging is to prevent evaporation of
the medium in which the embryo is imaged. In addition to using a sufficient amount
of MEGM, one can also cover the sample. For this, a second, small coverslip is care-
fully lowered onto the sample. A spacer such as a line of high-vacuum grease should be
used in between the two glasses to prevent the embryo from getting crushed. However,
it can prove more difficult to extract the capillary tube for freezing if a cover is used.
Also, removal of the upper coverglass will slightly increase the time between staging
and freezing.
B. Buffer Solution for Live-Cell Imaging and Freezing

In contrast to the previously published chapter on the preparation of mitotic embryos
for CLEM (Müller-Reichert et al., 2007), we used only MEGM for specimen prepara-
tion. Importantly, the membrane carrier was prefilled with MEGM + 20% BSA shortly
before transfer of the embryo-containing tube. Specimen preparation in buffer contain-
ing 20% BSA has the disadvantage that the medium becomes more viscous over time.
Also, BSA in the capillary tube increases the darkness of the whole specimen due to
the osmium tetroxide in the freeze-substitution “cocktail.” Nevertheless, BSA is a good
choice as a filler and cryoprotectant when used to completely fill the membrane carrier
prior to freezing.
C. Duration of Freeze-Substitution
There is a wide range of published freeze-substitution “cocktails” and protocols for
various cell types and different organisms (for a review, see Hippe-Sanwald, 1993). For
the freeze-substitution of the meiotic C. elegans embryos, we used a previously pub-
lished cocktail, containing 1% osmium tetroxide and 0.1% uranyl acetate in anhydrous
acetone (Müller-Reichert, Hohenberg, O’Toole, & McDonald, 2003). Freeze-substi-
tution with acetone was carried out at −90°C for 8–48 h. Afterward the sample was
warmed up 5°C/h to −30°C. Good fixation results are obtained if the sample is left at
−30°C for further 5 h. After this step of “contrast enhancement,” the sample is warmed
up at 5°C/h until a final temperature of 0°C. Recently, it has been reported for whole
C. elegans worms and other specimens that the duration of freeze-substitution can be
drastically reduced from 48 to 3 h (McDonald & Webb, 2011). Using basic laboratory
tools, such as a platform shaker, liquid nitrogen, a metal block with holes for cryo-
tubes, and an insulated container such as an ice bucket, the temperature increase from
−80°C to 0°C was significantly reduced to about 8 h. This quick freeze-substitution
protocol gave excellent ultrastructural results for whole worms and it is expected that
good structural preservation can also be obtained for both isolated meiotic and mitotic
embryos.
D. Software for Electron Tomography

The raw tilt series were converted into dual-axis tomograms using the software
package, eTomo (Mastronarde, 1997). The subsequent 3-D segmentation of cellular
features, such as microtubules and chromosomes, is the most time-consuming part of
3-D modeling. Currently, we use two well-developed software packages for tracing
of microtubules. The first one is IMOD (Kremer et al., 1996), where microtubules are
traced manually by placing points (i.e., contours) along the length of the microtubules.
Manual segmentation of microtubules, however, is very time consuming, especially if
one keeps the high number of microtubules in a semithick section of a meiotic spindle
in mind. We estimate that around 3000 microtubules built up the meiotic spindle. The
second software package is AMIRA (Visage Imaging GmbH). This software package
provides semiautomated tracing of microtubules (Weber et al., 2011). AMIRA runs
a template through the tomogram, which has the same hollow cylindrical shape as
a microtubule. After the automated screening for matching microtubule geometry, it
is recommended to check the microtubules for accurate tracing. To achieve this, an
“editor” has been specifically designed and implemented for correcting the automatic
tracing. The automatic tracing together with a manual correction has been reported to
significantly reduce the amount of manual labor for tracing microtubule centerlines.
This semiautomatic handling of 3-D data from electron tomograms is certainly an
important step toward large-scale analysis of cellular microtubule assemblies.
Acknowledgments
This work was supported by a grant from the Deutsche Forschungsgemeinschaft
(MU 1423/3–1) to T. Müller-Reichert.
References
Albertson, D. G., & Thomson, J. N. (1993). Segregation of holocentric chromosomes at meiosis in the nema-
tode, Caenorhabditis elegans. Chromosome Research, 1, 15–26.
Dumont, J., Oegema, K., & Desai, A. (2010). A kinetochore-independent mechanism drives anaphase chro-
mosome separation during acentrosomal meiosis. Nature Cell Biology, 12, 894–901.
Edgar, L. G. (1995). Blastomere culture and analysis. Methods of Cell Biology, 48, 303–321.
Hippe-Sanwald, S. (1993). Impact of freeze substitution on biological electron microscopy. Microscopy
Research and Techniques, 24, 400–422.
234 Ina Woog et al.
Hohenberg, H., Mannweiler, K., & Muller, M. (1994). High-pressure freezing of cell suspensions in cel-
lulose capillary tubes. Journal of Microscopy, 175, 34–43.
Kirkham, M., Muller-Reichert, T., Oegema, K., Grill, S., & Hyman, A. A. (2003). SAS-4 is a C. elegans
centriolar protein that controls centrosome size. Cell, 112, 575–587.
Manandhar, G., Schatten, H., & Sutovsky, P. (2005). Centrosome reduction during gametogenesis and its
significance. Biology of Reproduction, 72, 2–13.
Marsh, B. J., Mastronarde, D. N., Buttle, K. F., Howell, K. E., & McIntosh, J. R. (2001). Organellar relation-
ships in the Golgi region of the pancreatic beta cell line, HIT-T15, visualized by high resolution electron
tomography. Proceedings of the National Academy of Sciences of the United States of America, 98,
2399–2406.
Mastronarde, D. N. (1997). Dual-axis tomography: an approach with alignment methods that preserve reso-
lution. Journal of Structural Biology, 120, 343–352.
McDonald, K., & Müller-Reichert, T. (2002). Cryomethods for thin section electron microscopy. Methods
in Enzymology, 351, 96–123.
McDonald, K. L., Morphew, M., Verkade, P., & Müller-Reichert, T. (2007). Recent advances in high-pressure
McDonald, K. L., & Webb, R. I. (2011). Freeze substitution in 3 hours or less. Journal of Microscopy, 243,
227–233.
McNally, K., Audhya, A., Oegema, K., & McNally, F. J. (2006). Katanin controls mitotic and meiotic spindle
length. Journal of Cell Biology, 175, 881–891.
Müller-Reichert, T., Greenan, G., O’Toole, E., & Srayko, M. (2010). The elegans of spindle assembly. Cel-
lular and Molecular Life Science, 67, 2195–2213.
Müller-Reichert, T., Hohenberg, H., O’Toole, E. T., & McDonald, K. (2003). Cryoimmobilization and three-
dimensional visualization of C. elegans ultrastructure. Journal of Microscopy, 212, 71–80.
Müller-Reichert, T., Mancuso, J., Lich, B., & McDonald, K. (2010b). Three-dimensional reconstruction
101–119.
Olson, S. K., Greenan, G., Desai, A., Müller-Reichert, T., & Oegema, K. (2012). Distinct steps direct
hierarchical assembly of the eggshell and permeability barrier following fertilization in C. elegans. in
preparation.
O’Toole, E. T., McDonald, K. L., Mantler, J., McIntosh, J. R., Hyman, A. A., & Müller-Reichert, T. (2003).
Morphologically distinct microtubule ends in the mitotic centrosome of Caenorhabditis elegans. Journal
Pelletier, L., O’Toole, E., Schwager, A., Hyman, A. A., & Müller-Reichert, T. (2006). Centriole assembly in
Sonnichsen, B., Koski, L. B., Walsh, A., Marschall, P., Neumann, B., Brehm, M., et al. (2005). Full-genome
RNAi profiling of early embryogenesis in Caenorhabditis elegans. Nature, 434, 462–469.
Srayko, M., O’Toole, E. T., Hyman, A. A., & Müller-Reichert, T. (2006). Katanin disrupts the microtubule
lattice and increases polymer number in C. elegans meiosis. Current Biology, 16, 1944–1949.
Weber, B., Greenan, G., Prohaska, S., Baum, D., Hege, H. -C., Müller-Reichert, T., Hyman, A. A., Verbavatz,
J. -M. (2011). Automated tracing of microtubules in electron tomograms of plastic embedded samples of
Caenorhabditis elegans embryos. Journal of Structural Biology, 178, 129–138.
CHAPTER 13
Precise, Correlated Fluorescence

Microscopy and Electron Tomography
of Lowicryl Sections Using Fluorescent
Fiducial Markers
Wanda Kukulski*, †, Martin Schorb*, Sonja Welsch*,
Andrea Picco†, Marko Kaksonen†, *and John A.G. Briggs*, †
* Structural
and Computational Biology Unit, European Molecular Biology Laboratory, Meyerhofstr. 1,
D-69117 Heidelberg, Germany
† CellBiology and Biophysics Unit European Molecular Biology Laboratory, Meyerhofstr. 1, D-69117
Heidelberg, Germany
Abstract
I. Introduction
II. Rationale
III. Methods
A. High-Pressure Freezing
B. Freeze-Substitution and Embedding
C. Sectioning, Pick-Up and Application of Fluorescent Fiducial Markers
D. Fluorescence Microscopy
E. Electron Tomography
F. Fluorescent Fiducial-Based Correlation
G. Compensating for Possible Fluorescence Image Shifts
H. Application to Thin Section 2D Electron Microscopy
A. High-Pressure Freezing of Y
east Cells
B. High-Pressure Freezing of Mammalian Cells
C. Freeze-Substitution/Lowicryl Embedding
D. Ultramicrotomy, EM Grids, FluoSpheres
E. Fluorescence Microscopy
F. Electron Tomography
G. Fluorescent Fiducial-Based Correlation

236 Wanda Kukulski et al.
V. Discussion
A. Flexibility in the Sample Preparation Protocol
B. The Choice of the EM Grids
C. The Choice of FluoSpheres as Fluorescent Fiducial Markers
D. Estimation of the Correlation Accuracy
E. Photobleaching of EGFP and mCherry in Lowicryl Sections
F. General Applicability of the Method
References
Abstract
The application of fluorescence and electron microscopy to the same specimen

allows the study of dynamic and rare cellular events at ultrastructural detail. Here, we
present a correlative microscopy approach, which combines high accuracy of correla-
tion, high sensitivity for detecting faint fluorescent signals, as well as robustness and
reproducibility to permit large dataset collections. We provide a step-by-step protocol
that allows direct mapping of fluorescent protein signals into electron tomograms.
A localization precision of <100 nm is achieved by using fluorescent fiducial markers
which are visible both in fluorescence images and in electron tomograms. We explain
the critical details of the procedure, give background information on the individual
steps, present results from test experiments carried out during establishment of the
method, as well as information about possible modifications to the protocol, such as its
application to 2D electron micrographs. This simple, robust, and flexible method can be
applied to a large variety of cellular systems, such as yeast cell pellets and mammalian
cell monolayers, to answer a broad spectrum of structure–function related questions.
I. Introduction
Many cell biological questions that aim at the functional characterization of a cel-
lular process also involve questions on ultrastructure. Such studies often rely both on
dynamic information obtained from fluorescence microscopy (FM) and on high-resolution
data from electron microscopy (EM) or electron tomography (ET).
FM, in particular the use of green fluorescent protein (GFP) and its variants to genet-
ically label cellular proteins, offers unique capabilities to observe processes in living
cells over time (Lippincott-Schwartz & Patterson, 2003). It provides the cellular local-
ization and distribution pattern of the fluorescent protein (FP)-labeled proteins, and it
can give dynamic information on the behavior of the proteins, such as patch lifetimes,
movement, and protein turnover. The large field of view (usually over many cells) and
the fact that only what is fluorescently labeled is visible, make it possible to search for
rare cellular events marked by FP-labeled proteins. By labeling different proteins with
different color-variants of FPs (Shaner, Steinbach, & Tsien, 2005), it is possible to dis-
tinguish multiple proteins within a single sample. Together, these capabilities allow a
13. Precise, Correlated Fluorescence Microscopy and Electron Tomography of Lowicryl Sections 237
functional state of a cellular structure, an intermediate stage in a cellular process or the

sequence of events during a process to be defined based upon the presence or absence
of proteins labeled with different color variants, or observations of the time points at
which the proteins arrive and depart.
The major limitations of FM are the low attainable resolution as well as the lack of
information regarding the cellular context, since most cellular components are unla-
beled and therefore invisible. It is a dream of cell biologists to visualize the ultrastruc-
ture hidden under the fluorescent spot whose motion they observe and to know what
its surrounding environment looks like. The repertoire of recently developed super-
resolution techniques can partially fill the resolution gap (Patterson, Davidson, Manley,
& Lippincott-Schwartz, 2010), but remain blind to the cellular context; still only what
is fluorescently labeled can be seen.
In contrast, cellular EM and particularly cellular ET are unbeatable tools for provid-
ing three-dimensional structural data at nanometer-resolution within cells (Baumeister,
2002; Hoenger & McIntosh, 2009). The power of imaging by EM lies in its ability to
display the full cellular landscape in the field of view, and to show the ultrastructure of
interest within the crowdedness of the cell at high resolution. The drawbacks, on the
other hand, are equally important: The complexity of the cellular landscape makes it
sometimes very difficult to identify the ultrastructure of interest. If the structure is rare,
searching for it becomes searching for a needle in a haystack, and screening large areas
is tedious if not impossible. Further, the electron micrograph is a static snapshot of the
system; dynamic information gets lost due to the necessity of fixing the sample.
Clearly, the complementarity of FM and EM makes both equally indispensable tools
for cell biology. The many efforts being made in recent years to combine both tech-
niques on the very same specimen attest to the potential of correlative microscopy
methods in cell biology.
Many and varied correlative microscopy procedures are available these days, as
discussed in the other chapters of this volume. The first family of methods consists
of those where FM is applied to the system prior to fixation. These methods have
proven to be very powerful in identifying cells expressing a certain fluorescent pattern
that marks a defined stage in the cell cycle, or the developmental stage of an organ-
ism (Guizetti et al., 2011; Kolotuev, Schwab, & Labouesse, 2010; Müller-Reichert,
Srayko, Hyman, O’Toole, & McDonald, 2007; Pelletier, O’Toole, Schwager, Hyman,
& Müller-Reichert, 2006; Verkade, 2008). The second family of methods identifies
fluorescent cells or even organelles within the EM specimen after preparation. This
can be achieved either directly by FM or indirectly by immunogold-labeling or chemi-
cal reactions which transform fluorescent signals into electron contrast (Sartori et al.,
2007; Schwartz, Sarbash, Ataullakhanov, McIntosh, & Nicastro, 2007; Shu et al., 2011;
van Driel, Valentijn, Valentijn, Koning, & Koster, 2009; van Rijnsoever, Oorschot, &
Klumperman, 2008; Watanabe et al., 2011).
To answer questions concerning suborganelle-sized ultrastructures associated with
highly dynamic, complex cellular processes, two further hurdles must be overcome.
Firstly, high precision of correlation must be obtained. In most methods, the position-
ing accuracy is sufficient to localize cells within a tissue or organelles within a cell, but
does not permit to unambiguously identify and localize features with a precision below
100 nm. Secondly, high sensitivity is required to detect faint fluorescent signals. Most
approaches abrogate the fluorescent signal during sample preparation, or do not allow
the use of high numerical aperture and oil-immersion objectives, which require short
working distances. The achieved sensitivity is therefore far behind the state-of-the-art
light microscopy setups during live cell imaging. Approaches which perform live cell
imaging before fixation can potentially produce high-quality FM images, but the tem-
poral delay between the FM and the EM image is at least a few seconds, which does
not permit the study of fast processes or mobile features.
These limitations mean that a number of cell biological problems are challenging
to address. Such questions include the following: What is the conformational state of
a flexible cellular component when it binds to a specific auxiliary protein? Is there a
defined 3D ultrastructure underlying the diffraction-limited fluorescent spot which rep-
resents the cellular function of my interest? Can we unambiguously identify and assign
an unknown structure to our fluorescently labeled protein of interest? How do ultra-
structural intermediates of a very dynamic cellular process look like in 3D at a precisely
defined time point? These types of questions require a correlative microscopy approach
by which the specimen is imaged by FM and EM at the same time point (i.e., after
preparation) in a manner compatible with high-sensitivity FM as well as with state-of-
the-art ET. The method must be sufficiently robust and reliable to permit large datasets
to be recorded and must allow correlation of FM and EM data with high precision.
We have recently presented a correlative procedure that fulfills these conditions
(Kukulski et al., 2011). This approach is based on the observation that the fluorescent
signal of GFP, expressed in cells which have been high-pressure frozen, can be retained
in Lowicryl resin sections (Nixon et al., 2009). The freeze-substitution and embedding
protocols are optimized such that the ultrastructural preservation is as good as in current
state-of-the-art ET studies. At the same time, faint fluorescence signals are sufficiently
well preserved that a detection sensitivity similar to live cell imaging can be achieved.
RFPs are preserved as well as GFPs. FM is performed after the sections have been
placed on EM grids, and the imaging setup is designed to permit the use of high numeri-
cal aperture oil-immersion lenses. To achieve the required accuracy of correlation, a
fluorescent fiducial marker system is introduced. As fluorescent fiducial markers, we
use fluorescent microspheres, which are added onto the section surface and are visible
by FM in a fluorescent channel distinct from GFP or RFP. By ET, they become visible
on the section surface after tomogram reconstruction. The set of coordinates provided
by the positions of the beads in FM and in ET allows the precise position of the FP spot
of interest within the electron tomogram to be calculated. In addition, the fluorescent
fiducial system provides means for estimating the accuracy of the correlation.
The potential of this method has been illustrated by application to different types
of cell biological questions. We have demonstrated its power to find rare events such
as virus–cell interactions during virus entry, by pinpointing fluorescent HIV particles
on the surface of MDCK cells. We could determine the tip conformation of grow-
ing microtubules in fission yeast, marked by an RFP-labeled TIP-binding protein.
Finally, we have used the method to describe intermediate membrane states within a
10-seconds-time window during endocytosis in budding yeast (Kukulski et al., 2011).

The diversity of these examples demonstrates that the spectrum of possible applica-
tions of the method is large.
The method is simple, robust, and does not require any specially designed equip-
ment. It can therefore be established and used in any cell biology lab where there is
access to both FM and EM including high-pressure freezing technology.
II. Rationale
This chapter is intended as a description of our recently developed correlated FM

and ET method (Kukulski et al., 2011) providing comprehensive and detailed informa-
tion on the protocol, focusing on critical points. An overview of the complete workflow
and time line of the procedure is shown in Fig. 1. Within the chapter we will also
present technical findings and conclusions drawn while establishing the method. We
describe variations on the method, including application of the correlative procedure to
thin sections for 2D EM.
III. Methods
The primary specimens are cells that express the proteins of interest tagged with
fluorescent proteins. The first stage of the method is cryo-immobilization of the sample
by high-pressure freezing. For high-pressure freezing of yeast cells, we follow the pro-
tocols which were described in detail earlier in this book series (Höög & Antony, 2007;
McDonald, 2007). In brief, yeast cells are pelleted using a vacuum filtration device
onto a nitrocellulose filter, which is then placed onto an agar plate. The yeast paste is
loaded into membrane carriers with a toothpick, and high-pressure frozen using the
Rapid Transfer System of the EMPACT2. For high-pressure freezing of MDCK cells,
we also use an established protocol described in detail in an earlier volume of this series
(Walther, Wang, Liessem, & Frascaroli, 2010). To give a brief description, cells are
grown on precleaned, carbon-coated sapphire discs. For high-pressure freezing with
the BAL-TEC HPM-010, sapphire discs are clamped between aluminum planchettes
(Sapphire disc placed on the flat side of a Wohlwend platelet type 242, covered with
the 50 µm deep space of a Wohlwend platelet type 390), and the residual space in the
planchette cup is filled with 1-hexadecene. The planchette sandwich is disassembled
under liquid nitrogen prior to freeze-substitution.
B. Freeze-Substitution and Embedding

The embedding procedure we employ is based on the protocol published by Nixon
et al. (2009), which reports GFP signal retention in Lowicryl resin. For this, high-pressure
frozen samples are processed by freeze-substitution and embedding in Lowicryl HM20.
TIME LINE WORKFLOW
flexible time
cell culture
fixed time
possible
break cryo fixation (HPF)
choice of freeze
freeze substitution
substitution solution
3 - 7 days
Lowicryl embedding choice of embedding resin
2 days
UV box
0 - 2 weeks hardening of blocks in dark
trimming and sectioning of blocks
section pickup onto grids choice of grids

within
1- 2 days
addition of fluospheres distribution control by EM
sandwich assembly washing fluospheres
choice of fluospheres color

fluorescence microscopy
addition of gold fiducials

1 hour
post-staining
1 day (5-10
fluorescent electron tomography 2D electron microscopy
spots)
tomogram reconstruction
1 -2 days extraction of digital slices for correlation
correlation procedure
Fig. 1 Workflow and time schedule of the complete correlative microscopy protocol. In the central column,
the individual work steps are listed. In the left side column, the estimated time required for different parts of
the process is given. For steps marked by a continuous line, time is critical and should be kept at a minimum,
whereas for steps marked by a dashed line the duration is flexible. The right side column shows optional steps
including preparation or controls to be done. Grey boxes indicate steps which are ideally combined into one
wet-lab session.
We use a temperature-controlling AFS2 (Leica) with an FPS robot, and apply the fol-
lowing protocol to both yeast and MDCK cells. Freeze-substitution occurs at −90°C for
48–58 h with 0.1% (w/v) uranyl acetate in glass distilled acetone. The temperature is
then raised to −45°C (5°C/h), samples are washed 3 times with acetone and infiltrated
with increasing concentrations (10%, 25%, 50%, 75%, 4 h each) of Lowicryl HM20 in
acetone while the temperature raises further to −25°C. 100% Lowicryl is exchanged 3
times in 10 h steps, and UV-polymerized at −25°C for 48 h after which the temperature
is raised to 20°C (5°C/h) and UV polymerization continued for 48 h. After polymeriza-
tion, the samples should be stored protected from light, and should be processed further
within a few weeks. We noticed that the fluorescent signals in samples stored for sev-
eral months tend to be weak or even completely bleached.
C. Sectioning, Pick-Up and Application of Fluorescent Fiducial Markers

Sections of 300 nm are cut with a diamond knife in a microtome and picked up on
carbon-coated 200 mesh copper grids. As fluorescent fiducial markers, FluoSpheres
(Molecular Probes) are adhered to the section (see Discussion for information on
choice of fluorescent fiducials). We use 0.02 µm Blue FluoSpheres (excitation 365 nm/
emission 415 nm), because they do not fluoresce strongly in the GFP and RFP chan-
nel, but a small amount of signal bleeds through into these channels. This effect can be
used for drift-correction during the correlation procedure (see below). To reduce their
intensity and improve their adhesion properties, FluoSpheres are diluted 20× in a 0.1%
Tween-20 in PBS solution, incubated for 10 min at room temperature, washed twice by
30 min ultracentrifugation at 100,000 g, 4°C, resuspended in PBS to a concentration
approximately 50× diluted compared to the initial solution, and sonicated for 5 min.
Before applying on sections, it is advisable to check the monodispersity and concentra-
tion of the FluoSpheres preparation by EM. Large or fuzzy bead aggregates instead of
monodisperse spheres indicate a too harsh treatment with Tween-20. A pretreated Fluo-
Spheres preparation can be stored at 4°C for several weeks. To adsorb FluoSpheres to
sections, pretreated FluoSpheres are sonicated for 5 min, then grids are placed section-
face down onto a 15 µl drop of FluoSpheres for 10 min, blotted with filter paper and
washed with 3 drops of water with blotting between the washing steps.
D. Fluorescence Microscopy
To achieve fluorescence image quality and signal sensitivity equivalent to that
obtained in high-end live cell imaging, it is desirable to use high numerical aperture
oil-immersion lenses. This requires that there is no air layer between the sections and
the objective. To realize this, sections can be dried onto coverslips, which gives excel-
lent fluorescence imaging conditions (Watanabe et al., 2011), but makes recovery for
transmission EM impossible. We therefore decided to image sections adhered to an
EM grid, which is immersed in a thin water layer on a coverslip. To minimize the
working distance, the section-side of the EM grid has to face toward the objective
during imaging. The water layer between coverslip and sections ensures good imaging
conditions and allows the use of oil immersion objectives. To prevent drying, the grid
in water can be sandwiched between the coverslip and a glass-slide, and sealed with
vacuum grease. In practice, we found it more convenient to sandwich the grid between
two circular coverslips held together by a ring holder (custom-made or purchased
“Attofluor cell chamber” from Invitrogen) (Fig. 2). For this, drops of 15 µl of water
are placed onto two round coverslips, of which one has a layer of vacuum grease at the
rim (Fig. 2(A and B)). The grid is placed on one of the drops, and the sandwich is very
carefully closed, avoiding the formation of an air layer or bubbles between sections
and coverslip, as well as preventing the grid from sliding into the rim of vacuum grease
(Fig. 2(C and D)).
The assembled sandwich can then be imaged in a fluorescence microscope of choice.
The choice of fluorescence microscope depends upon the sample of interest, but in all
cases it should have sufficient sensitivity to detect the signals of interest, and appro-
priate filters for detecting all FPs as well as the fluorescent fiducials. In general, the
microscopy setup used for characterization of the signals by live cell or conventional
A B
C D
Fig. 2 Assembly of the sandwich setup for FM of sections on EM grids. (A) The ring holder consists of two
parts which can be screwed together to hold the coverslips in between. An O-ring ensures tight fixation of the
coverslips while avoiding breakage. The sandwich is assembled from two coverslips each carrying a drop of
water. The arrow indicates a rim of vacuum grease on one of the coverslips. (B) The grid is placed section-face
down onto one drop of water, and the sandwich is carefully closed with the second coverslip, avoiding that the
grid moves toward the rim of the coverslips. (C) The closed sandwich is placed in the lower part of the ring
holder. (D) The two parts are screwed together, just as tight as necessary to keep the sandwich fixed. Too tight
screwing will press the water out of the coverslip sandwich and cause the sections to stick to the coverslip.
FM will also be appropriate for imaging sections. We found that a wide-field imaging
set-up with a 100× oil objective was ideal for our needs (see Materials).
During FM, the positions on the grid of the imaged areas of interest should be
recorded so that they can be found back easily by EM (Fig. 3). We recommend the use
of grids with landmarks in the grid center (see Materials), which aid in locating the
region of interest to a defined gridsquare. In addition, 200 mesh grids have a gridsquare
size of about 90 µm, which roughly matches the entire field of view by FM in our setup,
making orientation relative to grid bars easy. It is thus best to mark the gridsquare
imaged by FM on a grid scheme (Fig. 3(A)).
The section is usually not perfectly flat; therefore, several focal planes will be
needed to obtain in-focus images of the full area of interest. This means that for each of
the three channels (GFP, RFP, and blue fluorescent fiducials), a number of 2–5 images
need to be taken in different focal planes. It is important to plan the recording scheme
to avoid any unnecessary exposure that will cause bleaching of the FP signal. Since
photobleaching of the bright FluoSpheres is insignificant, they should be imaged after
all the more critical GFP and/or RFP images have been collected on the given area. It is
recommended to start imaging in the bright field mode by finding the lowest (or high-
est) focal plane, record both GFP and RFP images, change the focus, and then record
both GFP and RFP again. This can be repeated until the highest (or lowest) focus is
reached. The number of focal planes that can be taken may be limited to about 3 or
4 planes due to the bleaching of the GFP signal, depending on its brightness (see “Pho-
tobleaching of EGFP and mCherry in Lowicryl sections”). After collecting all GFP and
RPF images, the focal stack of blue fluorescent fiducial images is recorded by work-
ing the way back from the highest to the lowest focal plane. Collecting a sufficiently
large series of blue fluorescent fiducial images ensures coverage of the whole area of
A B C
Fig. 3 Visual orientation on the EM grid from the fluorescence to the electron microscope. (A) The gridsquare
imaged by FM is marked on a grid scheme (white dashed box). This helps to find this gridsquare back in the
electron microscope. (B) An overlay of the whole GFP and RFP image frames, which correspond to approxi-
mately the size of one gridsquare. The gridbars are indicated by grey shading (top right and bottom left corner).
Their positions are identified in bright field mode (not shown). The approximate area recorded in the low
magnification tomogram is marked by a white dashed square. This mark greatly helps to re-find the area during
the fluorescent fiducial-based correlation procedure. (C) An enlarged view of the area marked in (B). Here the
dashed square represents the dimensions of the high magnification tomogram. Scale bars are 10 µm (B) and 2
µm (C). (See color plate.)
interest by in-focus images of FluoSpheres. Because the flatness of the sections varies
from area to area, the spacing between focal planes may need to be adjusted to obtain
a maximum area of coverage with a minimum of exposure. For this reason, we did not
find it helpful to automatically collect focal stacks for all areas.
Fluorescence imaging of EM grids should be done within approximately 1 day after
sectioning, as we noticed loss of fluorescence signal in sections on EM grids after longer
time periods. Fluorescent signals in sectioned material appear to deteriorate significantly
faster than in the resin blocks, which can be stored in the dark for at least a few weeks.
We recommend to make composite images from the three channels and to have
these at hand when sitting at the electron microscope during the subsequent ET ses-
sion (Fig. 3(B)). They can be used to quickly find the cell of interest in the electron
microscope based on the pattern of cells, and to note the approximate area where
tomograms are taken. This mark will be helpful in the later correlation procedure
to restrict the area in which the fluorescent signals of fluorescent fiducial beads are
found (see Fig. 4).
E. Electron Tomography
To enhance the contrast, grids with yeast sections are poststained after FM imaging
with Reynolds lead citrate for 12 min, whereas we found the contrast in MDCK cell
sections sufficient without further staining. As tomographic fiducial markers, 15 nm
protein A–coupled gold beads are adsorbed on both sides of the grids. Grids are placed
in a high-tilt holder, and digital images are collected as dual-axis tilt series over a −60°
to 60° tilt range (1° increment). Typical pixel sizes we use are 1.18 nm, 1.52 nm, or
1.97 nm at the specimen level.
With a tomogram frame size of 2048 pixels (we acquire binned images on a 4k × 4k
CCD camera), a pixel size of 1.18 nm results in a tomogram field of view of 2.4 µm.
This corresponds to a frame size of 37 pixels in FM (considering a pixel size of 64.5
nm as we have on our camera). The fluorescent signal from a point source such as one
FluoSphere (or any diffraction-limited spot of interest) typically spreads over 5–10
pixels, depending on its brightness. It is thus likely that in such a small area, not enough
FluoSpheres will be resolved by fluorescence imaging (see also “Fluorescent fiducial-
based correlation”) to perform the correlation procedure. Therefore, if the structures of
interest require the collection of high magnification tomograms, it is recommended to
additionally record lower magnification tilt series where the field of view will contain
more FluoSpheres for correlation. Low magnification tomograms are typically col-
lected as single-axis tilt series at 2.53 nm or 5.07 nm pixel size (corresponding to 5.2 or
10.4 µm frame size) and at 3° or 2° increment, respectively.
For all tomogram reconstructions, we use the IMOD software package (Kremer,
Mastronarde, & McIntosh, 1996). Lower magnification tomograms, which are only
used for fluorescent fiducial-based correlation and not for showing the ultrastructure
of interest at high resolution, can be reconstructed using the patch-tracking algorithm
available in IMOD version 4.1.4, which is much faster than the gold fiducial alignment
procedure.
input for fiducial-based correlation procedure
fluorescence microscopy electron microsocopy
1. assigning fiducial markers in FM and EM

A fiducial fluorescence B C
2. assigning FP spot of interest

calculate optimal
D RFP fluorescence E centroid of FP spot transformation between
of interest fiducial and EM image
based on fiducial positions
apply transformation to
fluorescent spot coordinates
G
3. shift between fluorescent images
F identify fiducial fluorescent
signal bleedthrough in RFP image
calculate shift
and apply
to fluorescent
spot coordinates
Result: transformed FP spot

coordinates: x=1106 y=1210
Fig. 4 Fluorescent fiducial-based correlation procedure. The example area shown here corresponds to the one
shown in Fig. 3. Yellow frames indicate input images (fiducial fluorescence image (A), average EM image (C),
and fluorescence image containing the RFP spot of interest (D)). The central column contains enlargements
of the FM area of interest (B and E). In the first panel, the manual assignment of fluorescent fiducials in FM
and EM is presented (yellow circles). The second panel illustrates the input of the RFP spot of interest (yel-
low circle), and the Gauss-fitting-based centroid determination is indicated as a red cross (E). In the third row,
an overlay of the fluorescent fiducial and RFP fluorescence images is shown (F), and the fluorescent fiducials
which bleed-through to the RFP channel are marked by yellow circles. These are used to calculate the average
shift between the fiducial fluorescence image and the RFP image. This shift is then applied to the RFP spot
coordinates shown in (E). The new coordinates of the RFP spot of interest are then converted into EM coordi-
nates (red mark in G) by applying the calculated optimal transformation between fluorescence and EM fiducial
coordinates. Scale bars are 10 µm (A, D, F) and 1 µm (C, G). (See color plate.)
F. Fluorescent Fiducial-Based Correlation

The concentration of fluorescent fiducial markers needs to be adjusted to achieve
the appropriate distribution density. The density must be low enough to clearly resolve
the beads from each other in the FM, but high enough that about 6–10 fluorescent
fiducials fall within the region reconstructed in the electron tomogram. We found that
approximately 50-fold dilution after pretreatment of FluoSpheres usually gives the
right distribution. To aid in visualizing all fluorescent fiducials included in the EM
field of view, those tomographic slices which contain the fluorescent fiducials adhered
to the section surface were averaged. To do so, we use the “Stack” functions of the
ImageJ software.
An overview of the correlation procedure is presented in Fig. 4. The correlation
process is performed with a simple in-house-written MATLAB procedure. As inputs
it requires the averaged EM image as well as those fluorescence images containing
fluorescent fiducials, GFP and RFP in-focus signals from the area of interest. After
entering the input data, the fluorescence micrographs are shown on screen next to the
average EM image (it is possible to adjust the brightness and contrast). Next, using the
MATLAB Control Point Selection Tool, the signals of FluoSpheres in the fluorescent
image are identified and assigned to the corresponding FluoSpheres visible in the aver-
aged EM images by clicking with the mouse on 6–10 FluoSpheres in each image. The
precise coordinates of the fluorescent signals of the fluorescent fiducials are determined
using a centroid fit with subpixel accuracy, which is carried out on a fluorescence image
which is high-pass-filtered with a smooth cut-off of 70 pixels. The filter excludes gradi-
ent effects from cellular background, which would otherwise affect the centroid fitting.
The centroid coordinate of the GFP or mCherry labeled signal of interest is determined
in the same manner, after manually marking its position in the fluorescence image.
The coordinates of the fluorescent fiducial pairs are used as landmarks for gener-
ating linear conformal transformations which relate the fluorescence image with the
EM image. Transformations are calculated using all possible combinations of subsets
of fluorescent fiducials containing a minimum of three fluorescent fiducial pairs. The
transformations are then scored by comparing the sum of squared residuals of the pre-
dicted fluorescent fiducial positions from the coordinates of the centroids of fluorescent
fiducials in the EM image, and the best-scoring transform is selected. This transforma-
tion is then applied to the fluorescence coordinates of the spot of interest to calculate its
coordinates in the electron tomogram.
To verify the success of the correlation, the predicted positions of the transformed
fluorescent fiducials are presented to the user to visually assess their deviation from the
beads’ actual positions in the EM image. In this way, any incorrectly picked fluorescent
fiducials can be identified. Although the above procedure excludes most outliers, in
rare cases they are included in the calculation. Thus, if the correlation is inaccurate due
to wrongly assigned fluorescent fiducials or outliers which are positioned on locally
deteriorated section areas, the fluorescent fiducial assignment can be repeated at this
stage after they are discarded.
Once the bead positions are accepted by the user, the final transformed coordinates
are saved in a log file. In case a further visual check of the correlation is desired,
transformed fluorescent images as well as images showing the positions of fluorescent

fiducials and the transformed spot in the EM image are created and can be overlaid.
If a higher magnification tomogram was recorded, a further transformation of the
coordinates of the fluorescent spot from the low magnification to the high magnifica-
tion tomogram is necessary (Fig. 5). This is done in a second correlation step using a
MATLAB Control Point Selection Tool-based procedure, which is equivalent to the
first, but which uses the tomographic gold fiducials to calculate the transform. Averages
of high and low magnification tomogram slices showing the tomographic gold fiducials
are calculated in ImageJ, and serve as input images displayed in MATLAB to assign
pairs of tomographic gold fiducials. Since the EM image showing the FluoSpheres,
which was used in the first round of correlation, also shows the section surface, it may
contain enough tomographic gold fiducials to be readily used in this step as well. The
tomographic gold fiducials are generally distributed more densely than the FluoSpheres
and are therefore better suited for correlating small fields of view.
G. Compensating for Possible Fluorescence Image Shifts

The desired accuracy of the correlation is below 100 nm, which is in the range of
the FM pixel size. This means that chromatic aberration or small sub-pixel to pixel-
sized shifts between the fluorescence image containing the fluorescent fiducials and the
image containing the FP signal of interest will cause localization errors of the FP coor-
dinates in the electron tomograms in the 50–100 nm range. It is therefore absolutely
critical to either avoid any shift between the fluorescence images or to correct for any
shift or aberration which has occurred during imaging.
input for 2nd correlation step if higher magnification tomogram required

assigning gold fiducials in low mag and high mag
A B C

coordinates in high mag
tomogram: x=1034 y=964
Fig. 5 Correlation of FP signal with high magnification tomograms. If a higher magnification tomogram has
been recorded, the FP spot coordinates are transformed onto this high magnification tomogram based on tomo-
graphic gold fiducials. Average images of the low magnification (A) and high magnification tomogram (B) are
used to assign gold fiducials. A detail from a tomographic slice of the high magnification tomogram is shown
in (C); the transformed FP spot coordinates reveal the structure of interest. The inner and outer circles mark the
50% and 80% accuracy, respectively. Scale bars are 1 µm (A), 200 nm (B), and 100 nm (C).
Such shifts might typically be caused by slight drift of the microscope stage between
collection of the FP image and collection of the fluorescent fiducial image. We decided
to correct for the shift after imaging, which could be comfortably done during the cor-
relation procedure.
In order to correct for any shift between the fiducial fluorescence image and the fluo-
rescence image containing the FP spot of interest, it is necessary to measure the amount
and direction of the shift. To do this, we look at an overlay of these two images and
click the positions of bright FluoSpheres signals which bleed-through from the fiducial
fluorescence channel into the fluorescence image containing the FP spot of interest.
We then measure the differences between their positions in the two images. From these
measurements, the average shift in x and y directions is calculated, and used to correct
the coordinates of the fluorescent spot of interest.
H. Application to Thin Section 2D Electron Microscopy

For some ultrastructural questions, 3D information is not required, and it is suffi-
cient to collect conventional 2D EM images instead of tomograms. We therefore tested
the applicability of our correlative procedure to 2D images. We found that FluoSpheres
were visible on poststained thin sections and that it is therefore very straightforward to
apply the method for 50 nm sections (Fig. 6). The protocol is generally the same as for
300 nm sections. Instead of tomograms, we collected EM images of the spot of interest
at various magnifications and applied the fluorescent fiducial-based procedure to cor-
relate these images to the corresponding GFP and RFP images. If higher magnifications
are desired, gold fiducials can be adsorbed to the sections in addition to FluoSpheres
and used to correlate low and high magnification images, analogously to the procedure
described above using low and high magnification tomograms.

A. High-Pressure Freezing of Yeast Cells
Instrumentation: Vacuum pump, filtration apparatus (McDonald, 2007). As high-
pressure freezer, we use the Leica EMPACT2 equipped with the Rapid Transfer System.
Materials: 0.45 µm pore size nitrocellulose filters, toothpicks, YPD agar plate,
0.2 mm membrane carriers (Leica Microsystems)
B. High-Pressure Freezing of Mammalian Cells

Instrumentation: For mammalian cells grown on sapphire discs, we use the
BAL-TEC HPM-010.
Materials: Carbon-coated sapphire discs and aluminum planchettes (M. Wohlwend,
Sennwald-CH) (Walther et al., 2010)
Reagents: 1-hexadecene
A C
Fig. 6 The correlation procedure applied to 2D images of 50 nm Lowicryl sections. (A) Overlay of EGFP
and mCherry images of a 50 nm section of yeast cells expressing proteins labeled with both FPs. A spot where
EGFP and mCherry colocalize has been targeted, and the dashed rectangle marks the area that corresponds to
the electron micrograph shown in (B). In (C) a part of the fiducial fluorescence image is shown, and the selected
fluorescent fiducial signals assigned to beads seen in (D) are illustrated by yellow circles. The green cross rep-
resents the transformed EGFP spot coordinates. (E) A higher magnification micrograph of the region of interest.
Scale bars are 10 µm (A), 2 µm (C), and 250 nm (E). (See color plate.)
C. Freeze-Substitution/Lowicryl Embedding
Instrumentation: Leica AFS2 with FPS robot for automated reagent handling
Materials: AFS2 consumables (reagent bottles, flow-through rings, reagent baths,
dispenser syringes)
Reagents: Glass-distilled acetone, 20% uranyl acetate in dried methanol, Lowicryl
HM20 (Polysciences, Inc.)
D. Ultramicrotomy, EM Grids, FluoSpheres

Instrumentation: Microtome (Leica Ultracut UCT), diamond knife (Drukker), ultra-
centrifuge (Tabletop Beckmann Instruments Inc., with rotor TLA 120.1)
Materials: Centrifuge tubes (Beckman Instruments, Inc.), 200 mesh copper grids
with carbon support film (Plano, product number S160)
Reagents: Blue FluoSpheres (365/415), 0.02 µm in size (Invitrogen, catalog num-
ber F8781, a large selection of colors and sizes is available at Invitrogen), Tween 20
(Sigma-Aldrich), phosphate buffered saline (PBS)
E. Fluorescence Microscopy
Instrumentation: We use an Olympus IX81 microscope equipped with a 100×
NA1.45 objective, Orca-ER CCD camera (Hamamatsu), electronic shutters, and filter
wheels (Sutter Instruments). An X-Cite 120 PC lamp (EXFO) is used for fluorescence
excitation with the following filters: 470/22 nm for GFP, 556/20 nm for mCherry and
377/50 nm for Blue FluoSpheres. For GFP and Blue FluoSpheres we use 520/35 nm,
and for mCherry 624/40 nm emission filters. We use Metamorph software (Universal
Imaging) to control the CCD camera, filter wheels, and shutters.
Materials: Round coverslips (Menzel-Gläser, diameter 25 mm, number 1), custom-
made ring-holder or Attofluor cell chamber (Invitrogen)
Reagents: Beckman vacuum grease silicone (Beckman Instruments Inc.)
F. Electron Tomography
Instrumentation: High-tilt tomography holder (Model 2020; Fischione Instruments)
or a DualAxis tomography holder (Model 2040, Fischione Instruments). The electron
microscope we use for data collection is a FEI Tecnai TF30 operated at 300 kV. We
record digital images on a FEI 4k Eagle camera as dual-axis tilt series.
Software: SerialEM for automated tilt series acquisition (Mastronarde, 2005),
IMOD software package for tomogram reconstruction (Kremer et al., 1996)
Reagents: 15 nm Protein-A covered gold beads, Reynolds lead citrate for poststain-
ing of yeast sections.
G. Fluorescent Fiducial-Based Correlation

Software: Matlab 7.4 (The MathWorks, Inc.) with the Image Processing Toolbox
installed. ImageJ 1.44 (National Institutes of Health, USA)
V. Discussion
In this chapter, we have described a correlative microscopy method for the study
of unknown, rare, transient, and dynamic cellular ultrastructures on the 100 nm scale,
which are identified by tagging proteins of interest with FPs. We aimed to provide the
reader with detailed information on how to perform correlative FM and ET based on

preserved fluorescent protein signals in resin sections of embedded cells, and how to
achieve an accuracy of correlation of better than 100 nm. The method is robust and
allows acquisition of large correlative datasets, and thus facilitates quantitative analysis
of the system under study.
A. Flexibility in the Sample Preparation Protocol

Several points in the protocol must be stringently followed, others can be adapted
to the needs of the actual project. Stringent requirements of the method include
cryo-immobilization of the specimen, the limited use of chemicals in the freeze-substitution
solution, and fast processing of the samples after UV-polymerization of Lowicryl through
fluorescence imaging, in particular after sectioning. Aspects of the protocol which can be
tuned for the cellular system or biological question of interest include, for example, the
temperature program of the freeze-substitution and embedding. The freeze-substitution
solution also tolerates some changes and small amounts of fixative, if necessary.
We tested freeze-substitution solutions with 0, 0.1, and 0.2% uranyl acetate and found
that with 0.2% uranyl acetate, the fluorescence preservation was decreased, and that
with 0% uranyl acetate the ultrastructural preservation was impaired. We also tested the
addition of 1–3% water, which is in some cases desired to improve membrane contrast
(Walther & Ziegler, 2002), and found it had no effect on fluorescence retention, and
that it did not significantly improve membrane contrast in our samples. The addition of
0.1% glutaraldehyde in the freeze-substitution solution was tested as well: we observed
no significant improvement in ultrastructure preservation, but found that weak fluores-
cent signals were no longer preserved. Signals originating from a large number of FPs,
such as MA-EGFP-labeled HIV particles, could, however, still be visualized. Since the
ultrastructural preservation was equally satisfying without chemical fixatives, we did
not further investigate the effects of higher glutaraldehyde concentrations.
Although we did not explore this ourselves, various resins have been reported to
retain the fluorescent signal from FPs. These are LR White for YFP fluorescence, par-
ticularly when the sample is not fully dehydrated (Micheva & Smith, 2007), or GMA
resin to retain Citrine, tdEOS and Dendra fluorescence (Watanabe et al., 2011). These
reports suggest that the embedding medium does not necessarily need to be Lowicryl
HM20. In addition, while our proof of principle experiments focused on GFP and RFP
variants (Kukulski et al., 2011), the above-mentioned studies showed the retention of
other FPs, indicating that it is likely to be possible to retain the fluorescence of a wide
range of FPs.
B. The Choice of the EM Grids

While exploring procedures to image resin sections by FM, we tested various grid
types as support for the fluorescent Lowicryl sections. We observed that Formvar and
Pioloform, both used as support for carbon film on EM grids or as exclusive support
film on slot grids, show significant autofluorescence in both GFP and RFP channels,
creating a speckle-type background which obscures faint fluorescent spots during FM.
We therefore recommend the use of grids coated only with carbon (see Materials).
Alternatively, for example, if the use of slot grids is important, using a confocal micro-
scope for fluorescence imaging can permit the resin section plane to be focally sepa-
rated from the plastic support film.
C. The Choice of FluoSpheres as Fluorescent Fiducial Markers

When selecting the type of beads suitable to form the fluorescent fiducial system for
correlating fluorescent images to electron tomograms, the requirements were defined as
follows: (i) a bright fluorescent signal with a spectrum distinct from the FPs of choice;
(ii) a good visibility by EM on cellular sections; (iii) monodispersity as well as good
adsorptive properties on cellular sections; and (iv) commercial availability. We tested
different variants of FluoSpheres, as well as fluorescent gold beads and quantum dots
as fluorescent fiducial markers.
We tested Rhodamine-labeled 8 nm gold beads by adsorbing them to resin sections
on EM grids and imaged them both by FM and by EM. We were able to assign fluores-
cent signals to the corresponding gold beads, which in most cases were in fact clusters
of 3–10 gold beads with the fluorescence intensity correlating with the number of gold
beads in the cluster. Single gold beads were rarely detected by FM. Although their high
contrast in electron micrographs, even on strongly stained cellular background, was an
advantage, we found several disadvantages in using fluorescently labeled gold. Firstly,
it is necessary to discriminate them from tomographic gold fiducials used for tomo-
gram reconstruction. The same fiducials cannot be used for both correlation and tomo-
gram reconstruction, because the latter must be present on the section at a much higher
density than can be used for correlation. Secondly, single beads give a signal that is too
weak to be detected, and most reliable signals originate from bead clusters, which hin-
der precise localization of the signal. Thirdly, the beads are not commercially available,
and need to be prepared in several centrifugation cycles with sequential coating steps.
We also tested imaging of quantum dots Q655 adsorbed to carbon-coated EM grids by
FM and EM. The main disadvantage of quantum dots is their “blinking” behavior in
FM. This requires imaging over a long time period in order to collect signals from all
fiducials in the imaged areas. In addition, the small size, low contrast, and nonspherical
shape of the quantum dots make it difficult to detect them on stained sections of cells.
In contrast to this, commercially available FluoSpheres give a bright, constant fluores-
cent signal, can be easily distinguished from tomographic gold fiducials, and their con-
centration and monodispersity on the section surface can be well controlled. Their color
can be chosen according to the color of the FP signals. Although FluoSpheres appear less
electron dense than gold beads or quantum dots, which makes them difficult to see in pro-
jection images of 300 nm sections of stained resin sections, they can easily be detected
after tomographic reconstructions and on thin 50 nm sections. The diameters we observe
for FluoSpheres with a nominal size of 20 nm vary between approximately 20 and 200
nm, and correlate rather well with their fluorescence intensity, which can be helpful when
assigning fluorescent signals to the fluorescent fiducials seen in the EM image.
D. Estimation of the Correlation Accuracy

For interpretation of the data, it is important to determine the accuracy and confidence
with which a specific FP signal is localized within the electron tomogram. This accuracy
will be sample-dependent: the FP signal itself, the cellular system under study, the density
of the fluorescent fiducials and the degree of distortion of the sample during section-
ing may all have an impact on how precisely the correlation will work. If a reasonably
large dataset has been collected, it is possible to use the data to estimate the accuracy
with which the position of the fluorescent signal can be determined in the tomogram.
In the correlation procedure described above, the coordinates of fluorescent fiducials
within fluorescence and EM images serve as a basis to calculate the transform that maps
the FP signal of interest onto EM images. To estimate the accuracy of the correlation, one
fluorescent fiducial bead is excluded from calculating the transformation, but is instead
treated like the FP spot of interest: this means that its position in the EM image is predicted
based on the transformation calculated using the coordinate pairs of the other fluorescent
fiducials. The predicted fluorescent fiducial position can then be compared to its actual
position. This procedure is repeated automatically for all the fluorescent fiducials in the
dataset, resulting in a set of deviations of the predicted positions from the true positions.
From this set, the fraction of predicted positions which fall within a certain radius from the
actual position can be derived. For example, in the case of the RFP-labeled microtubule-
tip binding protein mal3p localized in fission yeast cells, we used this method to estimate
that the accuracy of RFP-mal3p localization was 52 nm for 50% of the correlations, and
121 nm for 80% of the correlations (see dashed circles in Fig. 5(C)).
This calculation does not take into account any error in the determination of the shift
between the fiducial fluorescent image and the fluorescent image of the FP of interest
(see Section III.G). This error is typically on the order of 30 nm and can be removed
completely if the correlation is carried out using only fluorescent fiducials that bleed
through into the fluorescent channel of the object of interest, and if their coordinates
are measured in that channel.
E. Photobleaching of EGFP and mCherry in Lowicryl Sections

When collecting several fluorescence images to cover the entire focal depth of a sec-
tion (see Section III.D), we noticed that GFP signals bleach rather fast, while RFP sig-
nals remain almost constant. We compared the fluorescence photobleaching in sections
of budding yeast co-expressing Lsp1-mCherry and Pil1-EGFP, which are the major
components of eisosomes, and are present at eisosomes in roughly equal, high copy
numbers of 2000 to 5000 molecules (Brach, Specht, & Kaksonen, 2011; Walther et al.,
2006). The used excitation intensities gave similar signal-to-noise ratios with 1 s expo-
sure times for both channels, but after 60 frames of 1 s exposure time each, Pil1-EGFP
is almost completely bleached, whereas Lsp1-mCherry still shows significant fluores-
cent signals (Fig. 7). These observations indicate that the photostability of mCherry is
greater than that of EGFP when embedded in resin sections. We have not quantified this
effect, but have observed it repeatedly in various samples. It should therefore be con-
sidered when selecting the FP tag for a set of experiments. This choice will of course
A Lsp1-mCherry B Pil1-EGFP
300 frame 1 300 frame 1

290 frame 30 290 frame 30
frame 60 frame 60
280 280
fluorescence intensity
fluorescence intensity
270 270
260 260
250 250
240 240
230 230
220 220
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
pixel along the profile pixel along profile
Fig. 7 Comparison of GFP and RFP photobleaching in Lowicryl sections. The bleaching of Lsp1-mCherry is
shown in (A) and of Pil1-EGFP in (B). The gallery shows a selection of frames from a series of 60 frames,
1 s exposure each. The numbering of frames corresponds to the cumulative exposure when the given frame was
recorded. Below, the fluorescence intensity along a line (positioned as shown in the inset frame) is plotted for the
frames corresponding to a total exposure of 1 s, 30 s, and 60 s. The plot shows that spot and background intensi-
ties (peaks and baseline) of mCherry and EGFP are similar after 1 s of exposure. After a total exposure of 60 s,
the mCherry signal over background remains clearly visible, whereas EGFP spots are almost undetectable after
the same amount of exposure.
also depend on the same factors considered for live cell imaging experiments such as
the cytosolic background, fluorescence intensity, protein maturation rate, etc.
F. General Applicability of the Method

It is important to emphasize that this correlative microscopy method can be applied to
a large variety of cell biological questions in various cell types or organisms, provided that
the following requirements of the specimen are fulfilled: First, cryo-immobilization (i.e.,
high-pressure freezing) protocols for the given organisms must be available and result in
good ultrastructural preservation. For many model systems, high-pressure freezing proto-
cols can be found in various chapters of Volume 96 of Methods in Cell Biology (Electron
Microscopy of Model Systems). Second, signals from the FP-tagged proteins of interest
must be unambiguously identifiable and well characterized by conventional FM or live
cell imaging. Information should be available or obtainable on the cellular distribution of
the signal, its brightness, and its shape (i.e., whether the signal originates from structures
which are seen as resolution-limited spots in FM). The lower detection limit of the signal
will depend on a combination of copy number, brightness of the selected FP, and the cyto-
solic background of the tagged protein: these are the same considerations as for live cell
imaging. We have found that in general, even at low copy number, if a signal is clearly
detectable by conventional FM or live cell imaging, it will also be detectable in resin sec-
tions (Kukulski et al., 2011). If a transient or dynamic structure is to be described, dynamic
information such as signal lifetimes and movement are also essential for a comprehensive
interpretation of the “static” correlative microscopy data derived from plastic sections.
A simple variant of the method described here is its application on 2D electron micro-
graphs, instead of electron tomograms. This approach is faster because tomographic
reconstruction is avoided, and it does not require the availability of a high-end electron
microscope equipped for automated tomography. In some cases, visualization or iden-
tification of the ultrastructure of interest may not require 3D information. However,
acquiring electron tomographic data instead of 2D micrographs in order to correlate
cellular ultrastructures to fluorescence data has great advantages: for example, when
studying endocytic invaginations in yeast, only in 3D does their tubular shape become
obvious and can be unambiguously distinguished from sections through furrow-like
invaginations of the plasma membrane (Kukulski et al., 2011; Stradalova et al., 2009).
Subtle conformational variants such as the different structures of microtubule tips can
only be reliably determined in electron tomograms (Höög et al., 2007, 2010). A 2D
electron micrograph is a projection through the section volume (with a minimal thick-
ness of approximately 50 nm), which means that subtle fine structural features such as
microtubule protofilaments will be largely obscured by the dense cytoplasm above and
below it. In contrast to that, virtual tomographic slices represent volumes of few nano-
meters thickness, giving crisp and clear images. On the other hand, many problems do
not require 3D information, and these can be addressed by the 2D correlative approach.
For instance, projection images are adequate for questions on the subcellular localiza-
tion of the protein of interest, or the compartment with which the labeled protein is asso-
ciated. The 2D approach can in some cases serve as an alternative to immuno-EM if,
for example, immunogold labeling is impractical due to the lack of suitable antibodies.
Correlative light and electron microscopy methods have enormous potential to answer
fundamental cell biological questions. The correlated FM and ET method described in
this chapter allows 3D spatial resolution to be combined with dynamic information from
FM on the ultrastructural scale. The procedure is simple, robust, and provides the flexi-
bility to be applied to many different cell biological questions in various cellular systems.
References
Baumeister, W. (2002). Electron tomography: towards visualizing the molecular organization of the cyto-
plasm. Current Opinion in Structural Biology, 12, 679–684.
Brach, T., Specht, T., & Kaksonen, M. (2011). Reassessment of the role of plasma membrane domains in the
regulation of vesicular traffic in yeast. Journal of Cell Science, 124, 328–337.
Guizetti, J., Schermelleh, L., Mäntler, J., Maar, S., Poser, I., Leonhardt, H., et al. (2011). Cortical constriction
during abscission involves helices of ESCRT-III-dependent filaments. Science, 331, 1616–1620.
Hoenger, A., & McIntosh, J. R. (2009). Probing the macromolecular organization of cells by electron tomog-
raphy. Current Opinion in Cell Biology, 21, 89–96.
Höög, J. L., & Antony, C. (2007). Whole-cell investigation of microtubule cytoskeleton architecture by
electron tomography. Methods in Cell Biology, 79, 145–167.
Höög, J. L., Schwartz, C., Noon, A. T., O’Toole, E. T., Mastronarde, D. N., McIntosh, J. R., et al. (2007).
Organization of interphase microtubules in fission yeast analyzed by electron tomography. Developmental
Cell, 12, 349–361.
Höög, J. L., Huisman, S. M., Sebö-Lemke, Z., Sandblad, L., McIntosh, J. R., Antony, C., et al. (2010). Electron
tomography reveals growing MT ends to have flared morphology. Journal of Cell Science, 124, 693–698.
Kukulski, W., Schorb, M., Welsch, S., Picco, A., Kaksonen, M., & Briggs, J. A. (2011). Correlated fluo-
Biology, 192, 111–119.
Mastronarde, D. N. (2005). Automated electron microscope tomography using robust prediction of specimen
movements. Journal of Structural Biology, 152, 36–51.
McDonald, K. (2007). Cryopreparation methods for electron microscopy of selected model systems. Meth-
ods in Cell Biology, 79, 23–56.
Micheva, K. D., & Smith, S. J. (2007). Array Tomography: a new tool for imaging the molecular architecture
Müller-Reichert, T., Srayko, M., Hyman, A., O’Toole, E. T., & McDonald, K. (2007). Correlative light and elec-
tron microscopy of early Caenorhabditis elegans embryos in mitosis. Methods in Cell Biology, 79, 101–119.
10, 131–136.
Pelletier, L., O’Toole, E., Schwager, A., Hyman, A. A., & Müller-Reichert, T. (2006). Centriole assembly in
Shaner, N. C., Steinbach, P. A., & Tsien, R. Y. (2005). A guide to choosing fluorescent proteins. Nature
Methods, 2, 905–909.
Biology, 4, e1001041.
Stradalova, V., Stahlschmidt, W., Grossman, G., Blazikova, M., Rachel, R., Tanner, W., et al. (2009). Furrow-
like invaginations of the yeast plasma membrane correspond to membrane compartment of Can1. Jour-
nal of Cell Science, 122, 2887–2894.
Walther, P., & Ziegler, A. (2002). Freeze substitution of high-pressure frozen samples: the visibility of bio-
logical membranes is improved when the substitution medium contains water. Journal of Microscopy,
208, 3–10.
Walther, T. C., Brickner, J. H., Aguilar, P. S., Bernales, S., Pantoja, C., & Walter, P. (2006). Eisosomes mark
static sites of endocytosis. Nature, 439, 998–1003.
Walther, P., Wang, L., Liessem, S., & Frascaroli, G. (2010). Viral infection of cells in culture – approaches
for electron microscopy. Methods in Cell Biology, 96, 603–618.
CHAPTER 14
Integrative Approaches for Cellular

Cryo-electron Tomography: Correlative
Imaging and Focused Ion Beam
Micromachining
Alexander Rigort, Elizabeth Villa, Felix J.B. Bäuerlein,
Benjamin D. Engel and Jürgen M. Plitzko
Department of Molecular Structural Biology, Max Planck Institute of Biochemistry, Am Klopferspitz 18,
D-82152 Martinsried, Germany
Abstract
I. Introduction
II. Rationale
III. Methods
A. Cryo-correlative Microscopy
B. Cryo-fluorescence Instrumentation
C. Cryo-focused Ion Beam Milling
D. Cryo-focused Ion Beam Instrumentation
E. Experimental Workflow for Preparing Thin Cellular Specimens
F. Cryo-electron Microscopy and Tomography
G. Quantitative Analysis of Data from Cellular Tomograms
A. Preparation and Vitrification of Cells
B. Cryo-fluorescence Correlative Microscopy
D. Cryo-electron Microscopy and Tomography
V. Discussion and Outlook
Acknowledgments
References
Abstract
The application of cryo-electron tomography to cells and tissues is commonly

referred to as “cellular tomography,” and enables visualization of the supramolecular
260 Alexander Rigort et al.
architecture of cells in a near-native state. However, in order to access structural fea-

tures hidden deep inside cellular volumes, it is necessary to use hybrid techniques to
identify and localize features of interest and prepare such regions for subsequent analy-
sis by transmission electron microscopy. We present a workflow that integrates dif-
ferent approaches: (1) correlative cryo-fluorescence microscopy to localize features
within frozen-hydrated cells, (2) focused ion beam milling to thin these specimens in
a targeted manner, and (3) cryo-electron tomography to provide detailed information
about the cellular ultrastructure of thinned samples. We describe the combined use of
these techniques and the instrumentation required to enable cryo-electron tomography
for a vast range of cellular samples.
I. Introduction
Cells and tissues are poorly accessible to transmission electron microscopy (TEM)
since their thickness requires preparation techniques to render them thin enough
to be transparent to the electron beam. Conventional methods achieve this by cutting
thin sections of resin-embedded samples with an ultramicrotome, at the expense of
preparation artifacts caused by chemical fixation, dehydration, and embedding steps.
These artifacts can be avoided through the use of cryogenic preparation methods
(Dubochet & Sartori Blanc, 2001), which preserve native structures by maintaining the
fully hydrated state of biological specimens. Subsequently, cryo-electron tomography
can be employed to investigate the three-dimensional (3D) architecture and organiza-
tion of frozen-hydrated cells and tissues (Baumeister, 2005; Leis, Rockel, Andrees, &
Baumeister, 2009; Lucic, Forster, & Baumeister, 2005). However, such a cryo-electron
microscopy approach requires dedicated high-end instrumentation and sophisticated
computational methods for the extraction of information from tomographic data sets.
Depending on the dimensions of the sample, there are two main approaches to
immobilize biological material by vitrification (solidification while avoiding the for-
mation of crystalline ice; Fig. 1):
(i) Thin film vitrification by plunge freezing in a secondary cryogen, such as eth-
ane (Dubochet & McDowall, 1981; Frederik, de Haas, & Storms, 2009). This
method is limited to sample thicknesses of <5–10 µm and can be applied to
entire cells grown directly on EM grids.
(ii) High-pressure freezing (Moor, 1987). This approach is suitable for vitrifying
bulk biological samples with thicknesses of up to 200 µm (Studer, Graber,
Al-Amoudi, & Eggli, 2001).
As depicted in the scheme shown in Fig. 1, only a few cellular specimens can be
examined directly by cryo-TEM. Besides isolated “single particles,” this includes some
prokaryotic cells (Kurner, Frangakis, & Baumeister, 2005) as well as the peripheral
regions or “appendages” of whole eukaryotic cells that are sufficiently thin and there-
fore transparent to the electron beam (Medalia et al., 2002; Nicastro et al., 2006). How-
ever, the majority of the eukaryotic cell volume is impenetrable, and thus requires
additional preparation techniques prior to imaging.
14. Integrative Approaches for Cellular Cryo-electron Tomography 261
Fig. 1 Schematic diagram outlining the general workflow for specimen preparation required by cryo-electron
tomography. “Hybrid approaches” involve the correlative use of light microscopy, cryo-ultramicrotomy, and
focused ion beam (FIB) instrumentation.
Cutting ultrathin sections from high-pressure frozen biological material is one way
of dealing with the sample thickness constraint, and considerable effort has been made
to establish cryo-ultramicrotomy as a routine preparation method (Al-Amoudi, Diez,
Betts, & Frangakis, 2007; Hsieh, Leith, Mannella, Frank, & Marko, 2006; McDowall
et al., 1983). However, sections of vitrified cells inevitably suffer from distortions and
deformations caused by the mechanical cutting process. The most significant artifact
is unavoidable sample compression of up to 30–50% occurring in the cutting direction
(Al-Amoudi, Studer, & Dubochet, 2005; Richter, 1994).
A novel alternative to cryo-ultramicrotomy for site-specific thinning of frozen-
hydrated biological specimens uses focused ion beam (FIB) instrumentation, originally
developed for material science applications. In contrast to mechanical sectioning, thin-
ning of the specimen occurs via sputtering with focused ions, typically gallium. Pilot
studies showed that cryo-FIB milling can indeed be applied to frozen-hydrated mate-
rial, resulting in samples that contained bacterial cells transparent enough for TEM
(Marko, Hsieh, Schalek, Frank, & Mannella, 2007; Rigort et al., 2010).
Identifying and localizing regions of interest within ice-embedded specimens poses
another important challenge to cellular electron tomography (Fig. 1). At high magnification
of the electron microscope, it becomes difficult to target specific subcellular features,

especially those present at low copy numbers within the vast expanse of the cellular land-
scape. Correlative cryo-fluorescence microscopy can help bridge the gap in resolution
between light and electron microscopy (Sartori et al., 2007; Schwartz, Sarbash, Ataul-
lakhanov, McIntosh, & Nicastro, 2007; van Driel, Valentijn, Valentijn, Koning, & Koster,
2009). Under cryogenic conditions, features of interest are localized by their fluorescent
signal before zooming in on the structures of these features by cryo-TEM (Fig. 2).
This chapter provides an introduction to correlative and hybrid approaches, and
details the methods that enable a reliable and reproducible workflow for cellular cryo-
tomography. First, we focus on light microscopy (LM) approaches used to identify and
localize features of interest. Second, we give a detailed description of the FIB mill-
ing technique, followed by a discussion of the workflow for sample preparation and
electron tomography. Third, we show an approach for the automated segmentation of
cellular features from tomograms.
II. Rationale
The primary goal of any cellular tomography study is to obtain 3D images at molec-
ular resolution without compromising structural preservation. However, in order to
attain the necessary resolution, the sample of interest must be thin enough for inves-
tigation by TEM. Accomplishing this goal requires a reliable preparation workflow,
including methods for processing frozen-hydrated samples and computational tools for
statistical analysis.
Fig. 2 Correlative LM/EM of mammalian culture cells. A cellular region is shown from the scale captured by the light microscope to
the high magnification and relatively small field of view of the electron microscope. The arrows indicate areas enlarged in subsequent
panels. (A) Live-cell phase contrast image and (B) cryo-fluorescence image of rat hippocampal neurons grown on EM “Finder” grids
and labeled with FM1-43 vital dye. (C) Corresponding cryo-EM image of the region, exhibiting various neuronal processes surround-
ing the area where the tomogram was recorded. (D) Tomographic slice showing two neuronal processes and an extracellular vesicle
connected to both processes. (E) Surface rendering showing the extracellular vesicle (blue), two neighboring neuronal structures (gray),
connections between them (yellow and orange), and vesicle-bound molecular complexes (external, green; internal, red). The vesicle is
shown in a cut-away view to expose the complexes. Reproduced from Plitzko et al. (2009).
III. Methods
A. Cryo-correlative Microscopy
The field of view in cryo-electron tomography covers at most a few square microm-
eters. Localizing specific areas of interest on an EM grid, three orders of magnitude
larger, often exceeds the time required to record a tomographic tilt series. This “screen-
ing” process is frequently aggravated by the low signal-to-noise ratios (SNRs) of single
TEM projection images, which must be recorded under low-dose conditions to avoid
radiation-induced structural damage. However, the degree of damage caused by the
cumulative electron dose becomes evident only after data collection. Due to these
restrictions, structures of interest can be difficult to identify within low-dose electron
micrographs, which are dominated by noise that obscures finer structural details. Fur-
thermore, without a priori information, it is almost impossible to assess the functional
state of a structure, since electron micrographs represent only a series of static snap-
shots from which the sequence of events must be inferred. Here, fluorescence micros-
copy can offer a solution, permitting independent and unambiguous confirmation about
the functional state of a feature of interest, albeit at low resolution and devoid of the
structural context in which the events occur. The use of optical microscopy to screen the
cellular landscape and identify features of interest before magnifying these features via
cryo-EM provides a powerful hybrid approach for studying cellular processes (Fig. 2).
In general, the biological sample can be immobilized by vitrification techniques
either before or after imaging with the light microscope. Imaging at physiological
conditions has the advantage that oil-immersion objectives with high numerical aper-
tures (NA) can be used. This, in turn, allows investigations at higher spatial resolution,
which can be further increased through the application of super-resolution techniques
such as PALM/STORM (Betzig et al., 2006; Rust, Bates, & Zhuang, 2006; for overview
see Patterson, Davidson, Manley, and Lippincott-Schwartz 2010). However, molecu-
lar reorganization occurs during the time required to transfer the cells to a vitrification
device, complicating the accurate correlation of LM and EM observations. For example,
remodeling of the cellular actin network occurs within milliseconds in fast migrating
cells, such as neutrophils and Dictyostelium discoideum (Diez, Gerisch, Anderson,
Muller-Taubenberger, & Bretschneider, 2005). In practice, transfer to a vitrification
device can take at least 30 seconds for thin film vitrification and several minutes for
high-pressure freezing. The latter method is notably prolonged since medium or buffer
must be supplemented with a high molecular weight cryoprotectant to ensure uniform
vitrification throughout the entire sample. The vitrification process arrests cellular events
for subsequent cryo-electron tomography within milliseconds (Dubochet, 2007). Thin
film vitrification occurs in about 5 milliseconds (Berriman & Unwin, 1994), while high-
pressure freezing can be accomplished within 15–20 milliseconds. However, it remains
a challenge to gain temporal control over the freezing process (Frederik & Sommerdijk,
2005), to capture cellular events at specific stages by time-resolved vitrification.
In order to image the exact same structure in the electron microscope as was previ-
ously observed by LM, it is necessary to perform fluorescence microscopy after vit-
rification. Such a cryo-fluorescence microscopy approach allows direct correlation
of the frozen-hydrated sample between the two imaging modalities (Fig. 2). Unlike
approaches that image the specimen before vitrification, this method is not affected
by structural changes that occur during the time between fluorescence observation and
cryoimmobilization. However, the specimen has to be observed at temperatures below
the devitrification point (<−135°C), which excludes the use of standard immersion
objectives (although an experimental liquid-propane immersion lens is in development
(Le Gros, McDermott, Uchida, Knoechel, & Larabell, 2009). Consequently, the attain-
able resolution in cryo-fluorescence microscopy is limited by the NA of air objective
lenses. So far, the best experimentally measured resolution using fluorescent beads
for calibration is around 0.4 µm (van Driel et al., 2009). However, optical aberrations
(e.g., scattering of light in the frozen medium), subtle changes in sample quality, and
the locally varying ice thickness on a grid can affect the attainable resolution and com-
plicate accurate colocalization. A detailed description of the influencing factors and the
requirements for instrumentation in cryo-fluorescence microcopy is given in Section B.
Fluorescence labeling techniques are of fundamental importance to successful
cryo-correlative studies (Lippincott-Schwartz & Patterson, 2003; Sosinsky, Giepmans,
Deerinck, Gaietta, Ellisman, 2007). The labeling of intracellular epitopes in living cells
before freezing, either by electron-dense markers or fluorescent antibodies or compounds,
is not only invasive but also notoriously difficult. Immunocytochemistry labeling cannot
be performed under cryogenic conditions since antibodies do not have access to target sites
embedded in the ice and unbound labels cannot be removed via washing steps. Currently,
the best labels for cryo-correlative microscopy are clonable fusion proteins, such as green
fluorescent protein (GFP). These fluorophores are expressed within the cell, enabling non-
invasive screening for cellular phenotypes and localization of low-copy-number structures
on an EM grid. However, currently available fluorescent proteins do not provide contrast to
EM micrographs. Clearly, there is demand for a genetically engineered electron-dense con-
trasting agent suitable for both LM and EM. The first attempts to generate such a label were
based on fusing a small metal-binding protein, metallothionein, to a target protein. This
approach tried to leverage the capacity of metallothionein to initiate gold-cluster forma-
tion (Mercogliano & DeRosier, 2006). However, this approach has several shortcomings,
including nonspecific intracellular aggregation and adverse physiological effects.
B. Cryo-fluorescence Instrumentation
Fluorescence microscopy of vitrified biological specimens directly on EM grids
requires dedicated instrumentation that ideally can be adapted to various existing LM
platforms. For this purpose, different nitrogen-cooled cryostage systems have been
developed that can be mounted to either inverted (Sartori et al., 2007) or upright light
microscopes (Schwartz et al., 2007; van Driel et al., 2009). All systems must comply
with the basic requirements for a cryogenic workflow:
(a) Temperature stability: The vitrified specimen must be kept at temperatures below
the devitrification point of approximately −135°C at all times.
(b) Protection from contamination: Contamination by ice crystals or frost particles on
the EM grid surface must be avoided by proper isolation and shielding of the sample.
(c) Appropriate specimen holder: The system must be adapted to hold standard EM
grids.
These technical demands impose major constraints on the resolution and signal
detection of LM. Due to the low temperature of the cryostage, oil- or water-immersion
objective lenses with high NAs cannot be used. Instead, the system is limited to lower
NA dry objectives with long working distances (to account for the additional space
required for sample isolation and cooling). To date, only objectives with NAs up to
0.75 have been used on these systems.
The current cryostage implemented in our laboratory is shown in Fig. 3A. This
second-generation cryoholder, the “Cryostage2” (MPI of Biochemistry, Martinsried,
Germany and FEI Company, Eindhoven, the Netherlands), is adapted to the motor-
ized stage of an inverted epifluorescence microscope (Rigort et al., 2010; Sartori et al.,
2007). It comprises two main parts: a central cooling block connected to an automated
liquid-nitrogen (LN2) supply system (Norhof LN2 Microdosing system, Series 900,
Fig. 3 Schematic illustration of the cryo-correlative stage (Cryostage2) for an inverted fluorescence microscope. (A) Cut-away perspec-
tive view of the Cryostage2 holder with a 63× long working distance objective (NA 0.75), showing housing and insulation requirements.
(B) The Cryostage2, with an opening for the condenser in the lid, attached to an inverted light microscope (Zeiss Axiovert 200 M). (C)
Schematic drawing and magnified view of the central part of the Cryostage2, depicting the imaging position and the specimen slider,
with four loading positions for vitrified EM grids. Frozen specimens that are not docked in the imaging position are stored and protected
from ice contamination beneath the main cooling block. The free working distance using the 63× objective is in the range of 2.4–1.8 mm.
Scale bars: (A) 80 mm, (B) 20 mm. Inset (B) not drawn to scale. Adapted from Rigort et al. (2010).
Maarsen, the Netherlands) and a stable supporting frame with an insulation box that
separates the interior from the ambient atmosphere. A lid made of a synthetic polyure-
thane-based material, with an opening for the microscope condenser, covers the top
of the cryostage (Fig. 3B). Vitrified samples can be loaded directly within the cooled
nitrogen atmosphere of the stage, rendering an additional transfer station unnecessary.
To avoid potential cytotoxic effects caused by copper ions, eukaryotic cells are typi-
cally cultivated on gold EM grids. These gold grids are particularly malleable and can be
easily deformed by the gripping and clamping tools used throughout the entire workflow.
This obstacle can be circumvented by stabilizing the grid after vitrification in a rigid
specimen support. “Autogrids” (FEI Company, Eindhoven, the Netherlands), also known
as “C-clip rings,” consist of a rigid reinforcement ring that provides steadier specimen
support during the multiple transfer and handling steps (Rigort et al., 2012). The speci-
men slider of this cryostage (Cryostage2) enables direct loading and unloading of mul-
tiple autogrids (Fig. 3C), avoiding complicated manipulation steps. This is of special
importance if the grid will be subjected to further manipulation steps, as discussed below.

With the advent of cryo-preparation systems for FIB microscopes, a fundamentally
different method for thinning frozen-hydrated samples became feasible. FIB instru-
ments were originally developed for material science applications. These instruments
are commonly used to directly modify semiconductor devices, fabricate optoelectronic
components, and perform failure analysis (Volkert & Minor, 2007). The FIB technique
for TEM specimen preparation was introduced more than 20 years ago and is now a
standard preparation method in the material sciences because of its unsurpassed site-
specific preparation abilities (Kirk, Williams, & Ahmed, 1989). However, applying this
technique to frozen-hydrated biological specimens requires hardware and protocols
adapted to cryogenic needs.
During the milling procedure, the frozen-hydrated sample is kept in high vacuum and
is exposed to a beam of focused ions (see Fig. 4). Heavy ions, such as gallium (Ga+), are
typically used. The controlled bombardment of the sample with these ions results in the
removal of atoms from the specimen surface through sputtering. Notably, FIB milling
avoids the mechanical sectioning artifacts intrinsic to cryo-ultramicrotomy (Al-Amoudi
et al., 2005). Nonetheless, subtle ion-induced structural alterations, due to local heating
or the formation of a damaged layer on the milled surface (as a consequence of the ion
impacts), can be expected and are the subject of ongoing investigations. Moreover, the
surfaces of the milled areas are not completely planar and “streak-like” surface irregu-
larities along the milling direction can sometimes be observed. Such parallel striations
are related to compositional changes within the ice-embedded specimen, which result
in differential sputtering rates and thus preferential milling, referred to as “curtaining”
(Heymann et al., 2006).
To date, it has been shown that FIB milling of vitreous ice with an ion current of
10 pA (at a nominal incidence angle of 75° from the surface normal) does not induce
heating to the extent that devitrification occurs (Marko, Hsieh, Moberlychan, Mannella,
Fig. 4 Schematic illustration of the cryogenic setup for a FIB microscope. In an external loading station, vitrified EM “autogrids”
are mounted under liquid-nitrogen into a cryo-shuttle. The shuttle is picked up by a shuttered transfer device, which is evacuated by a
roughing pump before it is docked to a turbo-pumped cryo transfer unit attached to the high-vacuum chamber of the dual-beam micro-
scope. After pressure equalization, the shuttle is transferred to a nitrogen gas-cooled cryostage within the FIB instrument. An external
liquid-nitrogen dewar is used for cooling the cryostage with cold nitrogen gas to approximately −170°C. The focused ion beam is used
for micromachining the frozen-hydrated specimen.
& Frank, 2006). Moreover, simulations indicate that the Ga+ implantation zone should
be restricted to a relatively tolerable superficial layer of 10–20 nm when oblique or
“grazing” angles of incidence are chosen (Ziegler, Ziegler, & Biersack, 2010).
The cryo-FIB thinning technique can be applied to vitrified biological mate-
rial obtained by either plunge freezing or high-pressure freezing. So far, the successful
combination of FIB technology and electron tomography has been shown for only a few
selected applications using plunge frozen specimens on EM grids (Marko et al., 2007;
Rigort et al., 2010, 2012). Only a single study has demonstrated the preparation of thin
lamellae by FIB milling from high-pressure frozen samples (Hayles et al., 2010).
Cross-beam (ZEISS, Oberkochen, Germany) or dual-beam (FEI Company, Eind-
hoven, the Netherlands) microscopes combine the FIB with a scanning electron
microscope (SEM; Fig. 4). The combination with the SEM allows another “mode” of
tomography, where 3D information is obtained by imaging sequentially milled block-
faces using secondary electron or backscattered electron detection. Although the reso-
lution provided by this technique is currently modest compared to TEM, substantially
larger areas can be explored. In the so-called “slice and view” applications, the layer-
by-layer removal of material is accomplished by FIB milling (Heymann et al., 2006;
Knott, Marchman, Wall, & Lich, 2008). Serial block-face SEM (Denk & Horstmann,
2004), a variant of this technique for resin-embedded samples, combines imaging with
mechanical slicing by a built-in ultramicrotome. Tomograms obtained in this way
can be informative for bulk, stained specimens (e.g., brain tissues). However, these
approaches are currently not compatible with frozen-hydrated specimens, since they
impose very high demands on sample and beam stability (Muller-Reichert, Mancuso,
Lich, & McDonald, 2010).
D. Cryo-focused Ion Beam Instrumentation

Successful application of the cryo-FIB technique to the frozen-hydrated biological
material is directly tied to robust hardware implementation and reliable operation pro-
tocols. The following requirements must be taken into consideration:
(a) Temperature stability: The specimen must remain below the devitrification point
of approximately −135°C at all times. Significant heating during ion milling
(beyond the interaction layer) has to be avoided.
(b)C ontamination protection: The specimen must remain free from frost or other
potential contaminants during all cryogenic transfer steps.
(c) Sample and beam stability: Ion-induced damage should be limited to a surface
layer of not more than ∼10–20 nm. Processing times should be minimized to
avoid sample alterations and adverse drift effects.
(d) Sample dimensions and geometry: The thinned area of the specimen must be suit-
ably oriented for cryo-electron tomography (e.g., with respect to the orthogonal
tilting directions).
One of the major challenges in routinely applying FIB milling to cellular samples is
the multiple transfers of EM grids, during which the samples can be contaminated or
lost. This is especially an issue for projects that require cryo-correlative microscopy,
which adds further manipulation steps to the cryogenic workflow. For this reason, vari-
ous engineering solutions have been implemented to facilitate sample mounting and
transfer for LM, FIB, and cryo-TEM instruments (Fig. 5).
Several different column-mounted cryo-SEM preparation systems are commercially
available (e.g., Quorum, Gatan, Leica, Hummingbird). These systems enable the trans-
fer of specimens onto a stable SEM cold stage for further observation or manipulation.
Both the turbomolecular-pumped cryo-preparation chamber and the SEM chamber
contain cold traps (anticontaminators) to prevent frost contamination of the sample.
Depending on the system used, minor or major modifications must be made to account
for the requirements of cryo-electron tomography. Here, we focus on adaptations that
have been made to the Quorum “Polarprep” system.
After loading EM grids into rigid autogrids at cryogenic temperatures (Fig. 5A),
they are mounted into a transfer “shuttle” (Fig. 5B) in a cryo-FIB loading station
(Fig. 5C). The cryo-shuttle accommodates two autogrids, which are held in position
by a clamping device (Fig. 5B). A shutter covers both autogrid slots to protect the
samples from potential ice contamination during transfers. This shutter can be opened
or closed using the manipulation rod within the Quorum preparation chamber. The
pretilted (45°) design of the cryo-shuttle accounts for the tilting geometry permitted by
the FIB microscope stage and allows precise orientation of the autogrids with respect
to the incident ion beam (Fig. 5B, E). In order to enable oblique or “grazing” ion beam
Fig. 5 Essential components required for specimen preparation in cryo-EM. Perspective drawings of the (A) modified “autogrid”
frame, (B) cryo-FIB shuttle, (C) cryo-FIB transfer station, and (D) custom-made “Polara” TEM cartridge. The annotations highlight the
key features of the devices. The slot-like modification on the “autogrid” frame allows milling of the frozen specimen at oblique or “graz-
ing” angles with the focused ion beam. (E) A cryo-SEM micrograph displaying the cut-out region of an “autogrid” mounted in the cryo-
shuttle. (F) The same EM grid as shown in (E) viewed at parallel beam incident angle. From this orientation, parallel milling approaches
can be performed on the frozen-hydrated EM grid. Scale bars: (A) 1 mm, (B) 10 mm, (C) 40 mm, (D) 3 mm, (E) 1 mm, (F) 400 µm.
incidence angles for milling, a slot-like modification was made to the autogrid sup-
port (Fig. 5A, D–F). The loading station is comprised of a solid insulated housing that
accommodates a LN2 reservoir (Fig. 5C). Within this reservoir, a central metal insert
designed for the placement of standard EM grid boxes ensures the smooth cryogenic
transfer of frozen-hydrated grids into a small autogrid loading device (not shown) and
then into the cryo-shuttle. The cryo-shuttle is then loaded onto a support, which can be
tilted to permit uptake of the cryo-shuttle with the Quorum “Polarprep” transfer rod and
subsequent transfer into the FIB instrument (see Fig. 4).
Inside the high-vacuum chamber of the SEM/FIB microscope, a thermally isolated,
nitrogen gas-cooled cryostage is attached to the SEM/FIB stage. The cryostage and
anticontaminator are cooled by two separate cold gas circuits. The nitrogen gas-cooling
dewar is remotely positioned (an “off-column” dewar system). For milling experi-
ments, the cryostage is maintained at approximately −170°C. The details of the milling
procedure are described in the next section. Once the milling experiment is completed,
the cryo-shuttle is retrieved with the transfer rod and returned to the loading station.
Fig. 6 Illustration of possible FIB-milling strategies for vitrified cellular samples. (A) The frozen cell is attached to the carbon support
film of an EM grid and embedded in a vitreous ice layer. A thin layer (delineated by the dashed line) represents the specimen thickness
appropriate for cryo-electron tomography (<500 nm). (B) Parallel milling: the incident angle of the ion beam is parallel to the EM grid
surface. This approach usually involves halving a frozen grid and is demanding, as subsequent transfers with the weakened half-grid are
necessary. (C) Wedge-shaped milling: The ion beam impinges upon the frozen specimen at oblique or “grazing” angles. This approach
is the most feasible demonstrated thus far, as it can be performed without physically cutting the EM grid. (D) Cryo-lamella preparation:
the frozen specimen is milled vertically to expose a thin lamella, thereby preserving cellular features along the z-axis but necessitating
physical removal and reorientation of the lamella for TEM. At present, a lift-out option for cryogenic lamella preparations is not avail-
able. However, there has been recent success in milling self-supported lamellae, which do not require lift-out for cryo-TEM imaging
(Rigort et al., 2012). Reproduced from Rigort et al. (2010).
In LN2, the autogrids are recovered and either stored or directly transferred to a TEM
sample holder (Fig. 5D) for tomography experiments.
E. Experimental Workflow for Preparing Thin Cellular Specimens

The preparation of thin samples for electron tomography should provide sufficiently
large, electron-transparent regions covering a maximum cellular volume. Different mill-
ing geometries can be leveraged to open windows into the cell (Fig. 6). Currently, the
most suitable procedures rely on milling either parallel (Fig. 6B) or oblique to the sam-
ple support, the latter resulting in a wedge-shaped area (Fig. 6C). Milling free-standing
lamellae, as commonly performed in the material sciences (Fig. 6D), is currently not
feasible, primarily because of the unresolved technical challenge of transferring lamel-
lae at cryogenic temperatures. Current progress has been made in the milling of self-
supported lamellae (not shown), which are anchored on their sides to the surrounding
material and do not require lift-out for cryo-TEM imaging (Rigort et al., 2012).
Setting up a cryo-FIB experiment requires establishing parameters for ion beam
milling of vitrified samples (Fig. 7, IV). These parameters include the time, current,
patterns, and geometries for milling. Currently, the following procedure is recom-
mended for a cryo-FIB experiment using EM grids:
Fig. 7 Schematic diagram illustrating the sequence of experimental steps required to produce a thin cellular
specimen for cryo-electron tomography. Cells are grown on EM grids and vitrified by plunge freezing. Cryo-
genic transfer steps harbor the risk of damaging the sample due to ice contamination and physical deformation
with tools, such as fine tweezers. During steps III and V, the autogrid is mechanically manipulated by either
mounting to or dismounting from the appropriate FIB or TEM holder devices.
(1) An appropriate ion beam milling approach must be selected (see Fig. 6),
depending on the biological question being asked and the sample preparation
method used.
(2) Once the frozen-hydrated specimen is mounted onto the gas-cooled cryo-
FIB stage, it must be properly oriented with respect to the ion beam inci-
dence angle. This requires precise adjustments of height, rotation, and tilt on
the FIB microscope compustage.
(3) The areas of interest on the vitrified EM grid must be identified before the milling
process begins. This can be assisted by correlative LM techniques (see Fig. 8).
(4) Patterns for ion milling must be defined across the target areas (a process termed
“patterning”). A pattern defines the area that is rastered by the ion beam. The
sample is milled by moving the beam along a scan path in a series of passes
through the pattern. It is important to define the pattern size, pattern loop time
(the period of time before the beam repeats a scan of the milling pattern), dwell
time (the time the ion beam dwells at each pixel), and beam overlap (the ratio
between the pixel spacing and the beam diameter). Dwell times of 0.1–1 µs and
a beam overlap of 50% are suitable for milling vitrified biological specimens.
Fig. 8 Navigating and targeting areas of interest on a frozen-hydrated EM grid. (A and B) Merged cryo-phase
contrast and cryo-fluorescence image of prion-infected Saccharomyces cerevisiae cells (GFP). The white circles
(see arrowheads in B) indicate the sites selected for FIB milling. (C and D) Cryo-scanning electron micrograph
of the area shown in A and B, where the milled areas (arrowheads in D) can be easily recognized. Scale bars:
(A and C) 100 µm, (B and D) 40 µm. Adapted from Rigort et al. (2010).
(5) The main operation parameters of the FIB are the accelerating voltage and
the milling currents. In order to avoid surface irregularities (implantation of
gallium ions and “curtaining”) and to minimize the risk of thermal stress,
low ion currents must be selected. For coarse milling of bulk ice material,
ion currents of up to 0.1 nA can be used, while fine milling should be per-
formed at ion currents of 10–30 pA.
(6) The milling time depends on the aforementioned parameters (points 4 and 5)
as well as the specific size of the pattern defined around the feature of inter-
est. Going to lower ion beam currents will increase the time needed for
Fig. 9 Comparison between cryo-FIB milling and vitreous cryosectioning. (A) Scanning electron micrograph of a 50 nm thin cryosec
tion from a M. smegmatis sample obtained by cryo-ultramicrotomy. The cryosection is not well attached at its sides to the EM grid and
shows some “waviness.” (A’) Corresponding TEM projection micrograph. Compression in the cutting direction (white arrow) can be
clearly detected as parallel distortions. (B) Cryo-scanning electron micrograph of vitrified M. smegmatis cells on an EM grid. (B’) TEM
projection image from FIB-milled M. smegmatis cells. (C) Cryo-scanning electron micrograph of vitrified D. discoideum cells on an EM
grid. (C’) TEM projection image of a FIB-thinned D. discoideum cell. Dashed lines in B’ and C’ depict the milling edges. The milling
direction is indicated by the white arrows (in B’–C’). Scale bars: (A) 100 µm, (A’) 500 nm, (B) 5 µm, (B’) 500 nm, (C) 30 µm, (C’) 200
nm. Adapted from Rigort et al. (2010).
illing. The volume of ice that must be removed depends on the vitrification
m
method used (plunge freezing or high-pressure freezing), the nature of the
frozen-hydrated specimen (e.g., cells in suspension or adherent cells), and
the density of the cells on the grid.
F. Cryo-electron Microscopy and Tomography

Ion beam micromachining is a promising technique that allows the preparation of elec-
tron-transparent “windows” containing structures that reside deep inside the cells. The FIB
technique avoids mechanical knife-induced compression artifacts and is versatile enough
to target individual cells on grids under visual control, guided by correlative fluorescence
microscopy or cryo-SEM (Fig. 8). Fig. 9 shows SEM images and TEM projections of vitri-
fied Mycobacterium smegmatis and D. discoideum cells, obtained by cryo-ultramicrotomy
and FIB milling. Attachment of the cryosection ribbon to the carbon support film
of an EM grid (Fig. 9A) is often problematic, resulting in instabilities during tomographic

acquisition and difficulties in the subsequent alignment process. Compression in the
cutting direction is clearly visible within the 50 nm thin cryosection of M. smegmatis
(Fig. 9A’). In contrast to cryo-ultramicrotomy, the FIB-thinned bacteria sample shows
no signs of compression artifacts (Fig. 9A’). The corresponding SEM image of the frozen
mycobacterial cell suspension reveals the hydrophobic nature of their cell walls, as por-
tions of the cells protrude from the vitreous ice layer (Fig. 9B). Fig. 9C displays an SEM
micrograph of frozen-hydrated D. discoideum cells attached to the carbon support film
of an EM grid. Individual cells can be selected for thinning by FIB micromachining. The
resulting TEM projection image in Fig. 9C’ shows a thinned, electron transparent region
exposing a large portion of the eukaryotic cytoplasm. In both cases, ion beam milling was
performed with 30 keV gallium ions and a relatively low ion current of 10–30 pA. In the
tomographic reconstruction (Fig. 10) of the region shown in Fig. 9C’, a part of the cell’s
cytoplasm is exposed, delimited by the cell membrane. The tomogram encompasses a vol-
ume of approximately 0.12 µm3 of cellular space. Numerous macromolecular complexes
can be discerned and interspersed between a mitochondrion, a vacuolar compartment, and
part of the rough endoplasmic reticulum (Fig. 10A). The corresponding 3D model shows
ribosomes and actin filaments that were extracted using computational methods (Fig. 10B)
(Forster, Medalia, Zauberman, Baumeister, & Fass, 2005; Rigort et al., 2012).
G. Quantitative Analysis of Data from Cellular Tomograms

Due to the sensitivity of frozen-hydrated biological specimens to ionizing radiation,
the cumulative electron dose must be minimized and distributed among the images of
Fig. 10 Cryo-electron tomography of a FIB-thinned cellular specimen. (A) Slice from a tomographic recon-
struction of a D. discoideum cytoplasmic region (as shown in Fig. 9C’), revealing the cell membrane (cm) with
underlying cortical actin filaments (arrowheads), a mitochondrion (mt), parts of the endoplasmic reticulum (er),
and a vacuolar compartment (vc). The cytoplasm is crowded with a great abundance of unidentified macromo-
lecular complexes. (B) Corresponding surface rendering, color-coded for actin filaments (red) and putative ribo-
somes (yellow). The cell membrane and the membranous structures of mitochondrion, endoplasmic reticulum,
and vacuolar compartment are shown in gray. Scale bar: (A) 200 nm.
a tomographic tilt series. As a consequence of electron dose fractionation, low con-

trast of ice-embedded specimens, and artifacts arising from incomplete data sampling
(the “missing wedge”), the SNR in 3D reconstructions is very low. This makes image
analysis difficult and complicates structure determination and interpretation. In order
to enhance the SNR in tomograms, a number of denoising approaches have been devel-
oped for cryo-electron tomography (for an overview see Narasimha et al., 2008).
To allow for reproducible and objective analysis of cellular features contained within
cryo-tomograms, it is necessary to “separate” them from the background in a process
referred to as segmentation. For most features, such as intracellular membranes, this
is performed manually by assigning a label to voxels sharing the same visual char-
acteristics. From the corresponding contours of these labels, a 3D shape can then be
computed. However, such a manual approach relies on subjective user-dependent
selection and tracing steps, which are very time consuming and difficult to reproduce
by others. Furthermore, with the exception of some large-scale structures, it is virtually
impossible to perform a complete segmentation based solely on visual inspection of a
tomogram. Therefore, statistically sound analysis of cellular tomograms must leverage
automated segmentation techniques, allowing higher throughput and unbiased com-
parisons of results from numerous experiments (Rigort et al., 2012; Weber et al., 2012).
An example of such an automated segmentation approach for actin networks, which
combines template matching with a tracing algorithm, is given in Fig. 11. Polymeriza-
tion of monomeric actin into filaments drives the formation of different types of protru-
sions at the cell periphery and underlies cytoplasmic organization and cellular motility
(Pollard & Borisy, 2003; Ridley et al., 2003). Cryo-electron tomography has the unique
potential to provide 3D insights into the structural arrangement of unperturbed cellular
actin networks. Tomograms were acquired from different regions of D. discoideum
and fibroblast cells (REF-52 cell line). A sheet-like membrane protrusion (Fig. 11A)
containing a dense, branched network of actin filaments can be discerned, as well as
an actin stress fiber (Fig. 11B) consisting of densely packed bundles of actin filaments.
Another tomogram shows a finger-like filopodium (Fig. 11C) characterized by par-
allel bundles of actin filaments. Automated segmentation with reproducible statistics
is essential for understanding the functional organization of such complex filament
networks in the cellular context. This computational approach yields geometric data
on individual actin filaments, which can be used to characterize the actin networks in
terms of filament length, orientation, density, and stiffness.
A. Preparation and Vitrification of Cells

Instrumentation: Inverted light microscope for quick inspection of living cells (e.g.,
Zeiss Primo Vert). Vitrification device (Plunge-freezer; either custom-built or commer-
cially available system).
Materials: Dictyostelium discoideum strain AX2-214 (wild type) and polystyrene
cell culture dishes. Holey carbon-coated 200 mesh gold EM grids (Quantifoil, Jena,
Fig. 11 Automated segmentation of actin networks from cryo-electron tomograms, with actin filaments
c olored according to their angular orientation (color coding according to scale given in A). (A) Protrusive mem-
brane region from a D. discoideum cell exhibiting a dense network of actin filaments in various orientations.
(B) Part of a stress fiber within a migrating REF-52 fibroblast cell. Individual actin filaments are densely bun-
dled to form a higher order structure. (C) Filopodium from a D. discoideum cell emerging from a region of the
cell membrane where most of the actin filaments are oriented parallel to the membrane. (A’-C’) Tomographic
reconstruction slices from the corresponding tomograms. Scale bars: (A–C) 400 nm. Reproduced from
Rigort et al. (2012).
Germany and Protochips, Raleigh, NC, USA). Blotting filter paper (e.g., Whatman
No.1, Fairfield, CT, USA) and fine-tipped tweezers. Primary cryogen: liquid nitrogen;
secondary cryogen: liquid ethane.
Reagents: Nutrient medium (per liter: maltose, 20 g; yeast extract, 5 g; proteose
peptone, 5 g; thiotone E peptone, 5 g; Na2HPO4 · 7H2O, 0.67 g; KH2PO4, 0.34 g;
dihydrostreptomycin-sulfate, 0.05 g; final pH 6.4 to 6.6). 17 mM K/Na-phosphate buf-
fer, pH 6.0 (PB).
B. Cryo-fluorescence Correlative Microscopy

Instrumentation: Inverted light microscope (e.g., Zeiss Axiovert 200M inverted
uorescence microscope equipped with LD Plan-Neofluar 63×/0.75 Corr Ph2 objective,
fl
AxioCam MRm CCD camera). Cryo-correlative stage (either custom-built or
commercial-available system; e.g. Cryostage2, FEI, Hillsboro, OR, USA).
Materials: AutoGrid™ (FEI, Hillsboro, OR, USA) sample carriers.
Reagents: None.

Instrumentation: FIB microscope (e.g., dual-beam (FIB/SEM) instrument Quanta
3D FEG, FEI, Hillsboro, OR, USA), equipped with a cryo-SEM preparation system
(e.g., PP2000T, Quorum, East Sussex, UK).
Materials: Fine-tipped tweezers. Modified AutoGrid™ (FEI, Hillsboro, OR, USA)
sample carriers to permit accessibility for shallow incident ion beams in the FIB micro-
scope. Shuttered sample holder (e.g., cryo-shuttle) (Rigort et al., 2010) for transfer of
the specimen carriers.
Reagents: None.
D. Cryo-electron Microscopy and Tomography

Instrumentation: Intermediate-voltage electron microscope operated at 300 kV (e.g.,
Tecnai G2 Polara microscope (FEI, Eindhoven, the Netherlands) equipped with a GIF
2002 postcolumn energy filter and slow-scan CCD camera with 2048 × 2048 pixels
(Gatan, Pleasanton, CA, USA). Software for alignment (using fiducial markers and/
or cross-correlation procedures) and 3D reconstruction (e.g., TOM software package,
MPI Martinsried, Germany; Nickell et al., 2005; Korinek et al., 2011). Software for
data analysis and visualization (Amira, Visage Imaging, Berlin, Germany). Details
about actin segmentation (Rigort et al., 2012) can be found at http://www.zib.de/en/vi
sual/software/CryoEM.html.
Materials: Fine-tipped tweezers.
Reagents: Colloidal gold beads as electron-dense reference objects (fiducial mark-
ers; e.g., Ted Pella, Redding, CA, USA).
V. Discussion and Outlook
One of the ultimate goals of structural biology is the in situ visualization of the
concerted action of macromolecular assemblies within the cellular context. While
cryo-electron tomography is ideally suited to unveil cellular structures at molecular
resolution, it is challenged by the following paradox: while the information content of
a small volume increases (as one zooms further and further into the cell), sampling sta-
tistics decrease (as the volume imaged represents only a small proportion of the cell).
Therefore, approaches are needed that can be applied to a wide range of sample sizes,
bridging several orders of magnitude in spatial resolution.
The need to study biological systems on different scales––from “organisms
to atoms”––will make integrative methods indispensable. A unified methodol-
ogy is required that is capable of navigating cellular landscapes (e.g., correlative
microscopy), targeting features of interest (e.g., micromachining by FIB milling),
acquiring structural information (e.g., electron tomography), and analyzing complex
cellular samples (computational methods). To date, the successful combination of
various techniques into a robust and reliable workflow remains a major challenge.
Handling and transfer steps have to be precisely controlled throughout the process to
avoid adverse contamination and structural damage of the specimen. Suitable soft-
ware tools must be developed to provide a streamlined and transparent framework.
In addition, the automation of repetitive tasks will greatly increase the throughput of
hybrid imaging approaches.
Super-resolution microscopy techniques are a promising new avenue for correla-
tive microscopy (Toomre & Bewersdorf, 2010). Compared to conventional diffrac-
tion-limited LM, methods based on single molecule localization (PALM and STORM)
currently improve the resolution by roughly tenfold (∼20 nm). There may soon be a
convergence between the scales of fluorescence microscopy and cryo-EM, which will
yield a powerful combination of identifying fluorescently labeled molecules within the
cellular structural context provided by EM. However, all efforts will be hampered by
the cryogenic restrictions imposed by investigating frozen-hydrated samples.
FIB micromachining of frozen-hydrated cells promises to deliver samples devoid of
sectioning artifacts for tomography. FIB tools are routinely used in the semiconductor
industry and similarly, it is hoped that cryo-FIB will mature to a stage where it becomes
a common laboratory research tool. However, cryo-FIB technology is still at an early
stage, and improved devices and methodologies need to be developed to optimize its
application.
While cryo-electron tomography has already provided extraordinary insights into
the supramolecular organization of whole cells, the resolution that is routinely achieved
is far from the physical limit (∼2–3 nm). However, the near future should see consider-
able improvements that will greatly expand the possibilities of cryo-EM. For example,
direct electron detectors show great promise in terms of improved sensitivity and res-
olution performance, which will result in optimal signal collection in low-dose EM
(McMullan, Clark, Turchetta, & Faruqi, 2009; McMullan, Faruqi, et al., 2009; Milazzo
et al., 2011; Bammes et al., 2005). Moreover, the implementation of phase contrast
methods, namely phase plates (as routinely used in LM), will create opportunities for
in-focus phase contrast imaging at unprecedented resolutions. Recent applications of
Zernike-type phase contrast in cryo-EM of single particles and cryo-electron tomog-
raphy promise to launch a new era in EM (Danev & Nagayama, 2008; Murata et al.,
2010). The large increase in contrast demonstrated in these images suggests that it
may be possible to study structures that were previously considered “too small”
(<200 kDa) to be imaged by standard defocus-based phase contrast methods. However,
the development of phase plate devices is currently far from maturity, and several tech-
nical hurdles remain that prevent routine application of the technique.
In summary, the development and application of innovative technologies will eventu-
ally facilitate the structural characterization of macromolecular complexes in their func-
tional cellular context, bridging the gap between molecular and structural cell biology.
Acknowledgments
The authors would like to thank Tim Laugks and the members of the fine mechanic
workshop of the department for structural biology for their engineering. Work at the Max
Planck Institute of Biochemistry in Martinsried was supported by the European Com-

mission’s 7th Framework Programme (grant agreement HEALTH-F4-2008-201648/
PROSPECTS and a Marie Curie fellowship to E.V.), from the Federal Ministry of
Education and Research (BMBF) and from an inter-institutional research initiative of
the Max Planck Society.
References
Al-Amoudi, A., Studer, D., & Dubochet, J. (2005). Cutting artefacts and cutting process in vitreous sections
for cryo-electron microscopy. Journal of Structural Biology, 150, 109–121.
Al-Amoudi, A., Diez, D. C., Betts, M. J., & Frangakis, A. S. (2007). The molecular architecture of cadherins
in native epidermal desmosomes. Nature, 450, 832–837.
Bammes, B. E., Rochat, R. H., Jakana, J., Chen, D. H., & Chiu, W. (2012). Direct electron detection yields
cryo-EM reconstructions at resolutions beyond 3/4 Nyquist frequency. Journal of Structural Biology,
177, 589–601.
Baumeister, W. (2005). From proteomic inventory to architecture. FEBS Letters, 579, 933–937.
Berriman, J., & Unwin, N. (1994). Analysis of transient structures by cryomicroscopy combined with rapid
mixing of spray droplets. Ultramicroscopy, 56, 241–252.
Danev, R., & Nagayama, K. (2008). Single particle analysis based on Zernike phase contrast transmission
electron microscopy. Journal of Structural Biology, 161, 211–218.
dimensional tissue nanostructure. Plos Biology, 2, 1900–1909.
Diez, S., Gerisch, G., Anderson, K., Muller-Taubenberger, A., & Bretschneider, T. (2005). Subsecond reor-
ganization of the actin network in cell motility and chemotaxis. Proceedings of the National Academy of
Dubochet, J., & McDowall, A. W. (1981). Vitrification of pure water for electron-microscopy. Journal of
Microscopy-Oxford, 124, Rp3–Rp4.
Dubochet, J., & Sartori Blanc, N. (2001). The cell in absence of aggregation artifacts. Micron, 32, 91–99.
Dubochet, J. (2007). The physics of rapid cooling and its implications for cryoimmobilization of cells.
Methods in Cell Biology, 79, 7–21.
Forster, F., Medalia, O., Zauberman, N., Baumeister, W., & Fass, D. (2005). Retrovirus envelope protein
complex structure in situ studied by cryo-electron tomography. Proceedings of the National Academy of
Frederik, P. M., & Sommerdijk, N. (2005). Spatial and temporal resolution in cryo-electron microscopy: a
scope for nano-chemistry. Current Opinion in Colloid & Interface Science, 10, 245–249.
Frederik, P. M., de Haas, F., & Storms, M. M. H. (2009). Controlled vitrification. In A. Cavalier, et al. (Ed.),
Handbook of cryo-preparation methods for electron microscopy (pp. 71–102)). Boca Raton, FL: CRC
Press, Taylor & Francis Group.
Hayles, M. F., de Winter, D. A. M., Schneijdenberg, C. T. W. M., Meeldijk, J. D., Luecken, U., Persoon, H. ,
et al. (2010). The making of frozen-hydrated, vitreous lamellas from cells for cryo-electron microscopy.
Journal of Structural Biology, 172, 180–190.
Heymann, J. A.W., Hayles, M., Gestmann, I., Giannuzzi, L. A., Lich, B., & Subramaniam, S. (2006). Site-
155, 63–73.
Hsieh, C. E., Leith, A., Mannella, C. A., Frank, J., & Marko, M. (2006). Towards high-resolution three-
dimensional imaging of native mammalian tissue: electron tomography of frozen-hydrated rat liver sec-
tions. Journal of Structural Biology, 153, 1–13.
Kirk, E. C. G., Williams, D. A., & Ahmed, H. (1989). Cross-sectional transmission electron-microscopy
of precisely selected regions from semiconductor-devices. Institute of Physics Conference Series, 100,
501–506.
Knott, G., Marchman, H., Wall, D., & Lich, B. (2008). Serial section scanning electron microscopy of adult
brain tissue using focused ion beam milling. Journal of Neuroscience, 28, 2959–2964.
Korinek, A., Beck, F., Baumeister, W., Nickell, S., & Plitzko, J. M. (2011). Computer controlled cryo-elec-
tron microscopy–TOM(2) a software package for high-throughput applications. Journal of Structural
Biology, 175, 394–405.
Kurner, J., Frangakis, A. S., & Baumeister, W. (2005). Cryo-electron tomography reveals the cytoskeletal
structure of Spiroplasma melliferum. Science, 307, 436–438.
Le Gros, M. A., McDermott, G., Uchida, M., Knoechel, C. G., & Larabell, C. A. (2009). High-aperture cryo-
genic light microscopy. Journal of Microscopy-Oxford, 235, 1–8.
Leis, A., Rockel, B., Andrees, L., & Baumeister, W. (2009). Visualizing cells at the nanoscale. Trends in
Biochemical Sciences, 34, 60–70.
Lucic, V., Forster, F., & Baumeister, W. (2005). Structural studies by electron tomography: from cells to
molecules. Annual Review of Biochemistry, 74, 833–865.
Marko, M., Hsieh, C., Moberlychan, W., Mannella, C. A., & Frank, J. (2006). Focused ion beam mill-
ing of vitreous water: prospects for an alternative to cryo-ultramicrotomy of frozen-hydrated biological
samples. Journal of Microscopy-Oxford, 222, 42–47.
Marko, M., Hsieh, C., Schalek, R., Frank, J., & Mannella, C. (2007). Focused-ion-beam thinning of frozen-
hydrated biological specimens for cryo-electron microscopy. Nature Methods, 4, 215–217.
McDowall, A. W., Chang, J. J., Freeman, R., Lepault, J., Walter, C. A., & Dubochet, J. (1983). Electron
microscopy of frozen hydrated sections of vitreous ice and vitrified biological samples. Journal of
Microscopy, 131, 1–9.
McMullan, G., Clark, A. T., Turchetta, R., & Faruqi, A. R. (2009). Enhanced imaging in low dose electron
microscopy using electron counting. Ultramicroscopy, 109, 1411–1416.
McMullan, G., Faruqi, A. R., Henderson, R., Guerrini, N., Turchetta, R., Jacobs, A., et al. (2009). Experi-
mental observation of the improvement in MTF from backthinning a CMOS direct electron detec-
tor. Ultramicroscopy, 109, 1144–1147.
Medalia, O., Weber, I., Frangakis, A. S., Nicastro, D., Gerisch, G., & Baumeister, W. (2002). Macromo-
lecular architecture in eukaryotic cells visualized by cryoelectron tomography. Science, 298, 1209–1213.
Mercogliano, C. P., & DeRosier, D. J. (2006). Gold nanocluster formation using metallothionein: mass spec-
trometry and electron microscopy. Journal of Molecular Biology, 355, 211–223.
Milazzo, A. C., Cheng, A., Moeller, A., Lyumkis, D., Jacovetty, E., Polukas, J., et al. (2011). Initial evaluation
of a direct detection device detector for single particle cryo-electron microscopy. Journal of Structural
Biology, 176, 404–408.
Moor, H. (1987). Theory and practice of high pressure freezing. In R. A. Steinbrecht, & K. Zierold (Eds.),
Cryotechniques in biological electron microscopy (pp. 175–191). Berlin: Springer-Verlag.
methods for caenorhabditis elegans ultrastructure. Electron Microscopy of Model Systems, 96, 331–361.
Murata, K., Liu, X. A., Danev, R., Jakana, J., Schmid, M. F., King, J., et al. (2010). Zernike phase contrast
cryo-electron microscopy and tomography for structure determination at nanometer and subnanometer
resolutions. Structure, 18, 903–912.
Narasimha, R., Aganj, I., Bennett, A. E., Borgnia, M. J., Zabransky, D., Sapiro, G., et al. (2008). Evaluation
of denoising algorithms for biological electron tomography. Journal of Structural Biology, 164, 7–17.
Nicastro, D., Schwartz, C., Pierson, J., Gaudette, R., Porter, M. E., & McIntosh, J. R. (2006). The molecular
architecture of axonemes revealed by cryoelectron tomography. Science, 313, 944–948.
Nickell, S., Forster, F., Linaroudis, A., Net, W. D., Beck, F., Hegerl, R., Baumeister, W., & Plitzko, J. M.
(2005). TOM software toolbox: acquisition and analysis for electron tomography. Journal of Structural
Biology, 149, 227–234.
Plitzko, J. M., Rigort, A., & Leis, A. (2009). Correlative cryo-light microscopy and cryo-electron tomog-
raphy: from cellular territories to molecular landscapes. Current Opinion in Biotechnology, 20, 83–89.
Pollard, T. D., & Borisy, G. G. (2003). Cellular motility driven by assembly and disassembly of actin fila-
ments. Cell, 112, 453–465.
Richter, K. (1994). Cutting artifacts on ultrathin cryosections of biological bulk specimens. Micron, 25,
297–308.
Ridley, A. J., Schwartz, M. A., Burridge, K., Firtel, R. A., Ginsberg, M. H., Borisy, G., et al. (2003). Cell
migration: integrating signals from front to back. Science, 302, 1704–1709.
Rigort, A., Bauerlein, F. J. B., Leis, A., Gruska, M., Hoffmann, C., Laugks, T., et al. (2010). Micromachining
tools and correlative approaches for cellular cryo-electron tomography. Journal of Structural Biology,
172, 169–179.
Rigort, A., Gunther, D., Hegerl, R., Baum, D., Weber, B., Prohaska, S., et al. (2012). Automated seg-
mentation of electron tomograms for a quantitative description of actin filament networks. Journal of
Structural Biology, 177, 135–144.
Rigort, A., Bauerlein, F. J., Villa, E., Eibauer, M., Laugks, T., Baumeister, W., et al. (2012). Focused ion beam
micromachining of eukaryotic cells for cryo-electron tomography. Proceedings of the National Academy
of Sciences of the United States of America, 109, 4449–4454.
Sosinsky, G. E., Giepmans, B. N. G., Deerinck, T. J., Gaietta, G. M., & Ellisman, M. H. (2007). Markers for
Studer, D., Graber, W., Al-Amoudi, A., & Eggli, P. (2001). A new approach for cryofixation by high-pressure
freezing. Journal of Microscopy-Oxford, 203, 285–294.
Toomre, D., & Bewersdorf, J. (2010). A new wave of cellular imaging. Annual Review of Cell and Devel-
opmental Biology, 26, 285–314.
Volkert, C. A., & Minor, A. M. (2007). Focused ion beam microscopy and micromachining. MRS Bulletin,
32, 389–395.
Weber, B., Greenan, G., Prohaska, S., Baum, D., Hege, H. C., Muller-Reichert, T., et al. (2012). Auto-
mated tracing of microtubules in electron tomograms of plastic embedded samples of Caenorhabditis
elegans embryos. Journal of Structural Biology, 178, 129–138.
Ziegler, J. F., Ziegler, M. D., & Biersack, J. P. (2010). SRIM: the stopping and range of ions in matter.
Nuclear Instruments Methods B, 268, 1818–1823.
CHAPTER 15
Visualizing Proteins in Electron

Micrographs at Nanometer Resolution
Shigeki Watanabe and Erik M. Jorgensen
Howard Hughes Medical Institute and Department of Biology, University of Utah, Salt Lake City, UT
84112-0840
Abstract
I. Introduction
II. Rationale
III. Methods
B. Freeze-Substitution
C. Plastic Embedding
D. Sectioning
E. Fluorescence Imaging
F. SEM Imaging
G. Image Alignment
IV. Instrumentations and Materials
D. Coverslip Cleaning
E. Sectioning
F. PALM Imaging
G. Staining
H. SEM Imaging
I. Image alignment
V. Discussion
A. Fixation
B. Plastic
C. Super-Resolution Fluorescence Imaging
D. Electron Microscopy Imaging
E. Alignment
F. Quantification

284 Shigeki Watanabe et al.
VI. Perspective
Acknowledgments
References
Abstract
To understand protein function, we need a detailed description of the molecular

topography of the cell. The subcellular localization of proteins can be revealed using
genetically encoded fluorescent proteins or immunofluorescence. However, the pre-
cise localization of proteins cannot be resolved due to the diffraction limit of light.
Recently, the diffraction barrier has been overcome by employing several microscopy
techniques. Using super-resolution fluorescence microscopy, one can pinpoint the loca-
tion of proteins at a resolution of 20 nm or even less. However, the cellular context is
often absent in these images. Recently, we developed a method for visualizing the sub-
cellular structures in super-resolution images. Here we describe the method with two
technical improvements. First, we optimize the method to preserve more fluorescence
without compromising the morphology. Second, we implement ground-state depletion
and single-molecule return (GSDIM) imaging, which does not rely on photoactivatable
fluorescent proteins. These improvements extend the utility of nano-resolution fluores-
cence electron microscopy (nano-fEM).
I. Introduction
First described by Robert Hooke in 1665 (Hooke, 1665), a cell is the fundamental
unit of life. Since then, light-based microscopes have been used extensively to identify
structures such as the nucleus, mitochondria, and Golgi apparatus within a cell. Further
technological improvements, such as dark field, phase contrast, differential interfer-
ence contrast (DIC), and fluorescence imaging (Murphy, 2001), have expanded our
ability to probe the structure of a cell. However, due to the diffraction limit of light,
structures cannot be pinpointed in a cell (McCutchen, 1967). Thus, although Golgi
staining was used to reveal the basic anatomy of neurons (Cajal, 1894, 1899, 1903),
neither the synaptic connections between neurons nor subcellular structures, such as
synaptic vesicles, have yet been directly observed by light microscopy.
To overcome the resolution limit of light microscopy, electron microscopy was
developed in 1931 (Knoll & Ruska, 1932). With the shorter wavelength of electrons,
resolution was improved to less than 1 nm, allowing for a complete depiction of sub-
cellular structures. However, proteins important for cellular functions cannot be eas-
ily identified in electron micrographs. An immunocytochemical approach on plastic
sections has been the best method available to identify the location of a protein at the
ultrastructural level. However, this approach is difficult. The proteins lose antigenic-
ity or are inaccessible due the fixation and plastic embedding. Thus, although electron
microscopy has tremendous resolution advantages over light microscopy, its use has
been limited in biology.
15. Visualizing Proteins in Electron Micrographs at Nanometer Resolution 285
Despite the resolution limit imposed by the diffraction of light, fluorescence micros-
copy has become the most popular imaging technique in cell biology since the turn of
the twenty-first century. There are two types of fundamental innovations that continue
to extend the utility of fluorescence microscopy in cell biology: first, the discovery and
synthesis of new fluorescent probes and second, developments in microscopy tech-
niques. The most important innovation was the discovery and adaptation of genetically
encoded fluorophores, the first of which was green fluorescent protein (GFP) (Chalfie,
Tu, Euskirchen, Ward, & Prasher, 1994; Shimomura, 2009). Subsequently, fluorophores
with many different excitation and emission spectra have been developed (Tsien, 1998;
Zhang, Campbell, Ting, & Tsien, 2002). Using different sets of fluorescent probes,
the spatial relationships of multiple proteins in a cell have been studied. Furthermore,
microscopy techniques that allow three-dimensional visualization of proteins have been
developed. The first step in three-dimensional fluorescence imaging was the inven-
tion of confocal microscopy. In confocal microscopy, specimens are scanned in x–y
dimensions while out-of-focus fluorescence is blocked by an aperture. Only in-focus
fluorescence is detected for the reconstruction of an image. However, these techniques
are light based and hence the resolution is still limited to ~200 nm. Recently, ground-
breaking improvements in microscopy techniques have extended resolution beyond the
diffraction limit of light.
Super-resolution fluorescence microscopy overcomes the diffraction barrier by
isolating fluorescence from each molecule spatially and temporally (Hell, 2007). For
example, stimulated emission depletion (STED) microscopy allows the emission of
fluorescence from an area smaller than the diffraction limit (Hell & Wichmann, 1994).
STED is similar to confocal microscopy in that all fluorescent proteins in a diffraction-
limited spot are excited by a scanning laser. However, a doughnut-shaped de-excitation
beam that immediately follows the excitation beam brings the excited molecules back
to the ground state without allowing the emission of fluorescence. Thus, fluorescence
is only collected from the center of the doughnut, which is smaller than the diffrac-
tion limit. This subdiffraction spot is scanned across a specimen to generate an image.
A resolution of ~30 nm can be achieved using this technique (Harke et al., 2008).
Another super-resolution fluorescence technique is photoactivated localization
microscopy, PALM (Betzig et al., 2006), and its related techniques: fluorescence photo-
activation localization microscopy, fPALM (Hess, Girirajan, & Mason, 2006); ground-
state depletion with single-molecule return, GSDIM (Fölling et al., 2008); stochastic
optical reconstruction microscopy, STORM (Rust, Bates, & Zhuang, 2006); and direct
stochastic optical reconstruction microscopy, dSTORM (Heilemann et al., 2008). In
these techniques, each molecule is present one at a time by changing the fluorescence
state of individual fluorophores. For example, in PALM and STORM, fluorophores
can be photoactivated from a nonresponsive state to a responsive state one at a time
(Fig. 7(A); Gurskaya et al., 2006; Patterson & Lippincott-Schwartz, 2002; Wiedenmann
et al., 2004). Alternatively in GSDIM, fluorophores can be driven into a dark state, from
which they return to the ground state in a stochastic manner (Fig. 8(A); Fölling et al.,
2008; Heilemann et al., 2008). By taking advantage of such responses, each molecule is
isolated stochastically in space and time. By acquiring multiple images and calculating
the centroid of each fluorescent spot, all the molecules in the field are mapped
to produce the final image. Using these techniques, resolution can be improved to
~20 nm (Betzig et al., 2006; Hess et al., 2006; Rust et al., 2006; Shtengel et al.,
2009).
Irrespective of the technique used, all super-resolution fluorescence imaging has
the disadvantage that cellular context is missing from the images. To overcome this
problem, attempts have been made to combine fluorescence microscopy with electron
microscopy.
Methods for correlating protein localization in light and electron micrographs can
be separated into three approaches: preembedding fluorescence microscopy, preem-
bedding diaminobenzidine (DAB) imaging, and postembedding fluorescence electron
microscopy (fEM). Preembedding fluorescence microscopy is the simplest: the whole
cell is imaged using normal fluorescence techniques first to localize the protein and the
specimen is then processed for electron microscopy (Müller‐Reichert, Srayko, Hyman,
O’Toole, & McDonald, 2007; Oberti, Kirschmann, & Hahnloser, 2010; Polishchuk
et al., 2000; Verkade, 2008). This approach makes fluorescence imaging simple; how-
ever, precisely locating the same region of interest in the electron microscope is difficult
and it does not localize proteins with high-definition. The DAB precipitation method
uses photooxidation (Grabenbauer et al., 2005; Shu et al., 2011; Sosinsky, Giepmans,
Deerinck, Gaietta, & Ellisman, 2007) or enzymatic oxidation (Schikorski, 2010) of
DAB, which forms precipitation that can be observed via both light and electron
microscopy. This method can be useful when the proteins of interest are not abundant
in the cell. However, the electron-dense precipitation of DAB molecules may obscure
the fine details of electron micrographs, especially when the proteins are abundant.
Moreover, this technique is only compatible with conventional chemical fixation, and
thus fixation artifacts caused by dehydration (McDonald, 2007) are inevitable. Cor-
relative fEM preserves fluorescence in plastic sections and images fluorescence and
electron-dense structures from the same sections (Kukulski et al., 2011; Micheva &
Smith, 2007; Nixon et al., 2009; Schwarz & Humbel, 2007; Sims & Hardin, 2007;).
Fluorescence signals can be overlaid on electron micrographs using this approach, but
the disparity between diffraction-limited fluorescence imaging and ultrastructure means
that the proteins cannot be precisely localized, and the preservation of fluorescence
and morphology is often compromised. Recently, we have developed a method, called
nano-fEM, that preserves 60–70% fluorescence while retaining good ultrastructure
(Watanabe et al., 2011). By applying super-resolution imaging techniques, proteins can
be localized to subcellular structures with ~20 nm resolution. Here we will describe
our improved methods for protein localization in electron micrographs and discuss the
current issues and future directions of this approach.
II. Rationale
To fully characterize the functions of a protein, the location of the protein relative
to the subcellular structures must be revealed. A fluorescence-tagging approach is
a very robust method to localize proteins at a subcellular level. If super-resolution

fluorescence imaging is applied, the resolution of fluorescence images can be com-
parable to that of electron micrographs (Hell, 2007). However, the lack of cellular
context in fluorescence images limits the ability to localize proteins to subcellular
structures. On the other hand, immuno-electron microscopy has high resolution, but
it is limited by the poor quality of fixation and availability of antibodies (Morphew,
2007; Rostaing, Weimer, Jorgensen, Triller, & Bessereau, 2004; Roth, Bendayan,
Carlemalm, Villiger, & Garavito, 1981). A combination of these two techniques can
merge the robustness of fluorescence imaging with the ultrastructures depicted by
electron microscopy and potentially reveal protein localization at its maximum reso-
lution. Therefore, we sought fixation conditions that preserve both fluorescence and
morphology in the plastic sections.
We recently published a paper describing methods for correlating STED or PALM
images with electron micrographs (Watanabe et al., 2011). With 60–70% preserva-
tion of fluorescence after fixation and plastic embedding, signals appeared very bright
using STED microscopy. This result was partly due to the brightness and the superior
quantum efficiency of Citrine (Heikal, Hess, Baird, Tsien, & Webb, 2000) used for
STED. However, the resolution of STED was ~50–70 nm (Watanabe et al., 2011). On
the other hand, the average localization precision of PALM images, that is, a standard
error of mean of a localization, was ~12 nm (Watanabe et al., 2011). However, pho-
toconvertible fluorescent proteins appeared more sensitive to the sample preparation.
Moreover, PALM imaging is often restricted to a single color. Therefore, we sought
to improve fEM technique in two ways. First, we have been refining the protocol to
preserve more photoconvertible fluorescent proteins without compromising the mor-
phology. In this chapter, we will present a new protocol that preserves ~90% of fluores-
cence with the same quality of ultrastructure. Second, since Citrine is very bright and
has shown to be compatible for GSDIM (Lalkens, Testa, Willig, & Hell, 2011), we also
tested GSDIM imaging using Citrine. Fluorescent proteins compatible for multicolor
PALM imaging are rare; thus, if this approach is successful, GSDIM imaging will
allow the use of specimens that have already been tagged with fluorescent proteins for
confocal microscopy.
III. Methods
To perform postembedding fluorescence imaging, fluorescent proteins must be func-

tional in plastic sections. The retention of fluorescence through the fixation and plas-
tic embedding processes is extremely difficult because of the oxidizing, acidic, and
dehydrated conditions required for sample preparation. Fluorescent proteins require a
hydrated environment with alkaline pH for their best performance (Tsien, 1998). Thus,
we optimized the protocol to balance the preservation of morphology and fluorescence.
Protein localization using correlative nano-fEM consists of the following seven steps:
high-pressure freezing, freeze-substitution, plastic embedding, sectioning, super-resolution
imaging, electron microscopy imaging, and image alignment (Fig. 1).
Experimental procedures for nano-fEM
A Generating transgenics
(TOM-20::flur)
F Plastic embedding
B Mounting specimens C High-pressure freezing D Transfer specimens into

E Freeze-substitution G Trimming to ROI
for freezing (bacteria as freeze-substitution media with
cryoprotectants) and fixation in AFS
fixatives under liquid Nitrogen
K Image alignment J SEM imaging after post- I Super-resolution H Sectioning

staining with uranyl acetate fluorescence imaging
and carbon-coating
Fig. 1 Schematic diagram showing the procedure for nano-fEM. Transgenic animals (A) are subject to high-pressure freezing (B, C).
The cellular water is substituted with acetone, while tissues are fixed with 0.1% potassium permanganate and 0.001% osmium tetroxide
(D, E). The specimens are embedded into GMA plastic (F). The plastic block is trimmed to the region of interest (G), and thin slices
(~80 nm thick) of tissues are cut using a diamond knife (H). Sections are imaged first with super-resolution fluorescence microscopy
(I) and second with SEM (J). The correlative image is then aligned based on the fiducial markers (K).
High-pressure freezing is the first step required for preserving both fluorescence and
morphology. The high-pressure freezing and freeze-substitution techniques allow near-
instantaneous immobilization of specimens, prevent fixation and dehydration of arti-
facts such as protein aggregations (Payne, 1973), and also prevent membrane shrinkage
(McDonald, 2007) often observed using the conventional method. The preparation can
be carried out in the way recommended for the specimens of choice (McDonald et al.,
2010; Müller-Reichert, 2010). For Caenorhabditis elegans, transgenic animals express-
ing photoconvertible fluorescent proteins are raised in an opaque box to minimize their
exposure to ultraviolet light. Raising them in darkness prevents spontaneous photocon-
version of fluorescent proteins. Once animals are raised to the desired developmental
stage, typically to adulthood, they are subjected to freezing. For Baltec HPM 10, two
types of specimen carriers are used: type A and type B. Type A has two chambers. One is
A Ethanol substituted B Acetone substituted
500 nm
500
Fig. 2 Preservation of morphology by different freeze-substitution media. Transgenic animals expressing TOM-20–tdEos were freeze-
substituted in ethanol (A) or acetone (B). The ventral nerve cord in an adult C. elegans is shown in each case. Note that ethanol washed
out the tissues. 0.1% potassium permanganate and 0.001% osmium tetroxide was used as the fixative in both cases. The membrane
contrast was enhanced by both en bloc staining and section staining using uranyl acetate.
100 µm deep and the other is 200 µm deep. Type B has a 300 µm-deep well on one side
and a flat surface on the other side. Hexadecane should be applied to the 100 µm-deep
well of type A and the flat side of type B. Hexadecane should be removed from the wells
by blotting the carriers against a piece of filter paper. Since the diameter of an adult C.
elegans is about ~70 µm, we mount animals into the 100 µm-deep side of a type A speci-
men carrier. We add animals and bacteria to the specimen carrier by scooping animals
from the lawn of bacteria (Escherichia coli) with a paintbrush. The bacteria serve as
cryoprotectants. This configuration allows us to separate the animals easily from the bac-
teria during the plastic embedding step. Alternatively, 20% BSA in M9 solution can be
used (McDonald, Morphew, Verkade, & Müller-Reichert, 2007). However, it is harder to
dissociate the animals from the BSA solution. The chamber should be slightly overfilled
to avoid trapping air bubbles in the chamber when it is capped with the flat side of the
type B specimen carrier. We then freeze the specimens as instructed by the manufacturer.
Following high-pressure freezing, the vitrified water is replaced with organic sol-
vents such as acetone so that the water does not recrystallize when the specimens are
brought back to higher temperatures for plastic embedding. In addition, fixation of
tissues is carried out simultaneously to stabilize the subcellular structures and enhance
the contrast of tissues.
Prior to the freeze-substitution, cryovials containing 0.1% potassium permanganate
and 0.001% osmium tetroxide in 95% acetone (a mix of anhydrous acetone with milliQ
water) are frozen in liquid nitrogen. The use of acetone as a freeze-substitution medium
is critical for preserving ultrastructure because acetone cross-links with membrane
(Weibull & Christiansson, 1986; Fig. 2). The automated freeze-substitution (AFS) unit
should also be prepared as instructed by the manufacturer. The program should be set
as summarized in Table I. The duration at –90°C can be shortened or lengthened to
accommodate one’s schedule.
Table I
Freeze-substitution programs for AFS 1 and 2.
AFS 1 AFS 2
–90°C for 5–36 h –90°C for 30–36 h

5°C/h to –50°C 5°C/h to –50°C
–50°C for 2 h Pause
5°C/h to –30°C Unpause after acetone wash
–30°C for 72 h 5°C/h to –30°C
Stop the program and initiate another program after plastic Pause
embedding
–30°C for 0 h Unpause after plastic embedding
10°C/h to –20°C 10°C/h to –20°C
–20°C for 24 h –20°C for 24 h
The frozen samples are transferred into the cryovials containing the fixative
under liquid nitrogen. Only the specimen carrier containing specimens (typically
type A) is transferred into the cryovials. The cryovials are capped and placed in
the AFS, which is held at –90°C. To ensure the homogeneity of the fixative and
to accelerate the diffusion of chemicals, the cryovials are shaken periodically (at
least twice a day). When the temperature reaches –60°C, a glass vial containing
95% acetone is placed in the AFS chamber. As soon as the temperature reaches –50°C,
the fixative is replaced with 95% acetone. To ensure the complete removal of the
fixative, this step is repeated five times with 20 min intervals. In the meantime,
0.1% uranyl acetate in 95% acetone is prepared and precooled to –50°C. At the
last washing step, 95% acetone is replaced with the uranyl acetate solution. Uranyl
acetate stains phospholipids and nucleic acids and does not affect fluorescence
(Kukulski et al., 2011). Uranyl acetate is often included in the fixative; however,
we have observed improved morphology when uranyl acetate is applied after fixa-
tion. It is possible that the uranyl acetate interferes with osmium cross-linking the
membranes. The AFS program should be resumed if paused. The vials containing
95% ethanol (anhydrous ethanol + 5% milliQ water) are placed in the chamber so
that the solution is precooled. When the program reaches the –30°C step, the uranyl
acetate solution is replaced with 95% ethanol. Again, this step is repeated five
times over a period of 2 h. Although ethanol can extract lipids (Weibull & Christiansson,
1986), switching from acetone to ethanol is necessary to make the specimen com-
petent for acrylic resin embedding, since acetone interferes with plastic polym-
erization by scavenging free radicals (Newman & Hobot, 1993). About 90% of
fluorescence can be preserved using this protocol (Fig. 3). If significant reduction in
fluorescence level is observed using this protocol, one may need to alter the duration of
uranyl acetate application because some batches of uranyl acetate is highly acidic
and quenches fluorescence. One hour of incubation is sufficient for enhancing the
contrast of tissues.
TOM-20-tdEos TOM-20-tdEos
A No fixatives B With fixatives
50 µm
C Quantification of fluorescence intensity from

animals treated with or without fixatives
Fluorescence intensity
(arbitary unit) 200
100
0
Control (no fixatives, Fixed (0.1% KMnO4 + 0.001% OsO4,
embedded in GMA) en bloc UA staining, embedded in GMA)
Fig. 3 Preservation of fluorescence after fixation and plastic infiltration. (A) Fluorescent mitochondria in an
unfixed adult worm. Transgenic animals expressing the mitochondrial marker TOM-20–tdEos were freeze-
substituted with 95% acetone in the absence of fixatives. Mitochondria in the muscles of an adult worm are
visible. 100% of fluorescence is preserved when compared to unfixed animals (not shown). (B) Fluorescent
mitochondria in a fixed adult worm. Transgenic animals expressing TOM-20–tdEos were freeze-substituted
with 95% acetone in the presence of fixatives (0.1% KMnO4 + 0.001% OsO4) and en bloc stained with uranyl
acetate. (C) Fluorescence intensity of TOM-20–tdEos was measured from the control and the fixed animals and
quantified. A 10% reduction in the fluorescence level was observed when animals were treated with fixatives
and staining solution. (See color plate.)
After freeze-substitution, specimens are infiltrated with glycol methacrylate (GMA).
The infiltration is carried out in a stepwise fashion (30%, 70%, and 100%) at –30°C.
The stock solution of GMA can be prepared by mixing 22.7 ml of GMA, 10 ml of butyl
methacrylate, 1 ml of milliQ water, and 200 mg of benzoyl peroxide in a 50 ml conical
tube. Dissolving benzoyl peroxide may require a few minutes, and thus it is best to
leave the tube on the nutator in a cold room. Once the benzoyl peroxide is fully dissolved,
the stock solution is transferred into 20 ml scintillation vials and stored at –20°C until use.
To prepare the infiltration media, we mix the GMA stock solution media with 95% ethanol.
The infiltration media is then placed in the AFS chamber. This step should be carried out
concurrently with the ethanol-washing step of freeze-substitution so that the medium is
cooled down to –30°C before use. Each step of the infiltration process can be spaced to
accommodate one’s schedule. Specimens are typically incubated with the 30% solution for
2–4 h, the 70% solution for 3–5 h, and the 100% solution overnight.
The next day, the specimens are transferred into the cap from a polypropylene
BEEM capsule. An Aclar disk, prepared by a 3/8× punch, should be placed at the
bottom of the cap. The disk makes the surface of the plastic smooth which makes trim-
ming of the block easier. An Aclar disk must be placed on the top surface of the resin
as well if you are using LR White, since oxygen will prevent the polymerization of
the acrylic resin. Specimens are placed on the top of the Aclar disk, and thus the Aclar
disk blocks oxygen at the base of the cap. The media is exchanged twice more, with a
2 h interval. Animals are separated from the bacteria so that individual nematodes are
exposed to plastic without being surrounded by bacteria or other animals. Acrylic resin
does not cross-link with tissues; thus, when sectioned, specimens can dissociate from
the surrounding plastic, resulting in distortion of the plastic and discontinuity of the rib-
bon of sections (Fig. 4(A)). If individual animals that are completely surrounded by the
plastic are sectioned one at a time, the distortion of the plastic will be minimal, improv-
ing the sectioning result (Fig. 4(B)). Therefore, the removal of the cryoprotectants is
critical. At the final step, a catalyst, n,n-dimethy-p-toluidine, is added to the GMA stock
solution at a concentration of 1.5 µl catalyst per 1 ml GMA stock. The reaction is highly
exothermic and can be complete within 30 min of the addition of the catalyst at room
temperature, so the catalyst should be added while the GMA stock solution is held in
the AFS chamber. The specimens should be left in the chamber overnight to ensure
completion of the process. Specimens are stored in nitrogen-gas-filled vacuum bags
(Ziploc) at –20°C to reduce oxidation of the fluorescent protein.
D. Sectioning
Once the plastic has polymerized, the specimens can be manipulated at room tem-
perature. A room with bright sunlight or even bright room light should be avoided at
all times. To find a region of interest before sectioning, the specimens can be exposed
briefly (~1 s) to blue light. Once the area of interest has been located under the stereo-
microscope, a knife mark is left near this area, and the block is trimmed to the knife
mark using a razor blade. Sectioning of GMA plastic can be challenging because of the
hydrophilic nature of the plastic. If further trimming of the block with a glass knife is
necessary, water in the boat should be underfilled. If water gets on the block face, the
section submerges rather than floats on the surface and cannot be retrieved.
Ultrathin sectioning can be carried out in a manner similar to that used for epoxy
resins. However, sectioning quality of GMA is inconsistent. Cutting even the same
block on a different day may result in poor appearance of morphology (Fig. 4(C and
D)). Occasionally, folding of a tissue due to sectioning can also be observed (Fig. 4(E)).
In addition, structures such as lipid droplet in gut can also dissociate while sectioning
(Fig. 4(F)). Sectioning thickness and cutting speed can be adjusted to minimize the sec-
tioning artifacts––we often cut 80 nm thick sections with the cutting speed of 1.6 mm/s.
If sections are imaged on a TIRF microscope, the sections must be mounted directly
on a coverglass; otherwise, sections can be mounted on a transmission electron micros-
copy (TEM) grid. For PALM imaging, autofluorescence from dust particles can also
be detected because of the high sensitivity of the EMCCD camera. Therefore, cover-
glasses should be cleaned using piranha solution. The piranha solution can be prepared
by mixing three parts of sulfuric acid and one part hydrogen peroxide. The coverglasses
are left in the piranha solution for 1 h, washed thoroughly with milliQ water six times,
sonicated for half an hour, and washed again with milliQ water six times. This treatment
Fig. 4 Problems associated with GMA sectioning. (A, B) Low magnification electron micrographs, showing the dissociation of worms
from the surrounding media. The section is more distorted in a specimen that is surrounded by the cryoprotecting bacteria (B) than
when the specimen is surrounded by plastic (A). The bacterial cryoprotectant in the gallette should be dissected away from the fixed
sample before plastic embedding. (C, D) Electron micrographs of neurons from the same specimen, sectioned on different dates. The
preservation of tissues is superb on one day (C), but such morphology is obscured by the inconsistent sectioning quality (D). (E) Elec-
tron micrograph of nerve ring, showing folding of sections due to the incomplete polymerization of plastic (black arrows). (F) Electron
micrograph of intestine, showing dropouts of tissues (black circles).
A Picking up sections using a loop B Picking up sections directly onto a coverslip
Fig. 5 Schematic diagrams showing two methods for mounting sections onto a coverslip. Sections can be picked up using a loop (A)
or by simply immersing a coverslip in the boat of a diamond knife (B). A longer ribbon of sections can be mounted by immersing a
coverslip. (See color plate.)
leaves the surface of the coverglasses hydrophilic, which is not ideal for picking up
sections. The addition of a few drop of hexamethyldisilazane (HMDS) in a sealed jar
containing the coverglasses creates the hydrophobic surface.
There are two ways to mount sections on a coverglass (Fig. 5). First, sections can be
picked up using a perfect loop (Electron Microscopy Scieneces) or something similar and
placed on a coverglass (Fig. 5(A)). A problem with this approach is that sections may dis-
sociate from the ribbon in the process. The loop should be cleaned using ethanol prior to
use so that dust particles are not transferred with the sections. The use of a Kimwipe or
filter paper to remove water from the loop also leaves dust on the sections, so the drop of
water containing sections should be blown off to a coverglass from the loop. Once the water
evaporates, sections will adhere to the surface of the coverglass. The second method is to
use a diamond knife with a large boat so that a coverglass can be submerged in the water to
pick up sections, just as they are on a TEM grid (Fig. 5(B)). Mounting sections in the middle
of the coverglass is technically more challenging because sections become invisible in the
concave water surface created from submerging the coverglass in water. A longer ribbon
can be mounted on a coverglass using this method, whereas the dimension of the ribbon is
restricted by the size of the loop in the former case. Ideally, sectioning should be carried out
in a darkened room. Once sections from the region of interest are collected, the coverglass
should be covered with aluminum foil and stored at –20°C until further processing.
E. Fluorescence Imaging
Prior to fluorescence imaging, a solution containing gold nanoparticles should be
applied to the sections for 1 min. The sections are then soaked in milliQ water for 2 min.
The 100 nm gold nanoparticles emit red fluorescence when excited by green light due to
surface plasmon resonance (Link & El-Sayed, 1999) and reflect electrons when imaged
under an electron microscope. Therefore, they can mark both fluorescence images
and electron micrographs and serve as alignment markers for the image correlation.
The solution containing gold nanoparticles with a concentration at 1.2 × 1010 particles/
ml is applied to the specimen for ~30 s to 1 min and then removed by blowing the solu-
tion off to the side of the coverslip and blotting it using a Kimwipe or filter paper. Next,
a drop of water is applied to the sections. The water restores the function of fluorescent
proteins. After fixation and plastic embedding, some of the fluorescent proteins are
preserved in a nonfluorescent state (Watanabe et al., 2011). We found that the addition
of water prior to imaging increases the fluorescence intensity 30-fold (Watanabe et al.,
2011). However, autofluorescence in GMA resin also increases after hydration. This
increase is detected by an EM-CCD camera, with which the PALM scope is equipped.
Therefore, PALM imaging cannot be carried out while immersing the samples in water.
Water is only applied to specimens for 2 min and is removed just prior to imaging. The
use of a Kimwipe or filter paper to draw the solution off directly from the specimen would
introduce dust particles into the sections and induce autofluorescence, so the solution must
be blown to the edge of the coverslip and then absorbed onto a Kimwipe or filter paper.
For PALM imaging, the microscope should be set up as recommended by the manu-
facturer. The frame rate should be set at 20 Hz or less. Bright field illumination should
be used to locate specimens. A low-intensity 488 nm laser can be employed to localize
proteins in the region of interest if necessary but should be avoided if possible. Once
the region of interest is set, 561 nm laser light at the highest intensity should be applied
for ~2 min to the specimen to bleach the autofluorescence. To collect images, the acti-
vation laser and readout laser should be set at their lowest and maximum intensities,
respectively. We usually collect ~ 6000 frames per section; this is sufficient to sample
the distribution of fluorophores. However, if every molecule in the region must be
localized, the frame number should be increased accordingly. If the activated molecules
in a given space are sparse, the activation energy can be gradually increased. It is impor-
tant to note that the computer calculates the centroid of each fluorescent spot; thus, if two
molecules emit fluorescence in a given diffraction-limited spot at the same time, only
one dot would be placed in the middle of these two molecules, resulting in a false nega-
tive. PALM images should be acquired from each section on the coverslip. The sections
become dehydrated during a long imaging session, so they are rehydrated with a drop of
milliQ water approximately every half-hour. Again, the milliQ water should be blown to
the side before absorbing it onto a Kimwipe or filter paper to avoid contamination.
For GSDIM imaging, fluorescent proteins must be driven into the dark state in the
absence of oxygen (Fölling et al., 2008). Plastic seems to serve as a diffusion barrier
for oxygen; thus, a solution containing oxygen scavengers or a chamber filled with
nitrogen gas may not be necessary. Following the gold particle application and hydra-
tion described in the previous section, 488 nm laser at 5 mW/mm2 intensity is applied
to induce the state shift of fluorescent proteins from the ground to the dark state (Fig.
8(B and C)). About 5000 images are collected at a frame rate of 20/s while applying a
continuous 488 nm laser at 1 mW/mm2.
After their acquisition, images should be processed for PALM (Fig. 7(B and C) or
GSDIM (Fig. 8(D and E). Since some background fluorescence is inevitable when
imaging plastic sections, it should be filtered out in the final fluorescence image. For
background subtraction, we impose two thresholds: fluorescence lifetime and bright-
ness. The emission from tdEos molecules typically lasts for ~500 ms or less before they
become bleached when 5 mW/mm2 laser is applied. A threshold of 500 ms is set so that
any molecules that fluoresce for longer than 500 ms do not appear in the final image.
Additionally, low-intensity background autofluorescence can be subtracted by visual-
izing the molecules with high-localization precision. Localization precision depends
on the number of photons collected at each spot. Fluorescence from tdEos or Citrine
molecules often appears brighter than background fluorescence; hence, by imposing a
threshold for localization precision, low-intensity noise can be filtered.
F. SEM Imaging
To observe the ultrastructure of biological specimens under a scanning electron
microscopy (SEM), tissues must first be post-stained with uranyl acetate, and then
coated with carbon. Although uranyl acetate is applied to the specimens during the
cryosubstitution process, the contrast is insufficient for SEM imaging. To post-stain the
sections, a solution containing 2.5% uranyl acetate in milliQ water should be prepared.
Filtering the solution using a 0.2 µm syringe filter is highly recommended. A drop of
the solution is applied directly to the top of the sections, and after 4 min, it is washed
off by gently running filtered milliQ water thoroughly over the specimens. The sections
should be air dried rather than blotted with filter paper.
Plastic sections and coverglasses are not electron-conductive materials. Thus, the
surfaces of such specimens will be charged when placed under the electron beam.
To prevent charging, specimens must be coated with carbon. Using a carbon sput-
ter, a layer of carbon is applied to the specimen so that the glass slides appear dark
brown. The surface of the specimen is then grounded by applying one end of a piece
of carbon tape to the edge of the glass and attaching the other end to the base of the
specimen stub. The electrons that accumulate on the glass surface will now run to the
ground.
After preprocessing and the routine alignment of the microscope, the samples can
be imaged. Backscattered electrons should be collected. The FEI Nova Nano has
two features that make it more sensitive to backscattered electrons. First, a magnetic
field can be applied between the pole piece and the specimen so that backscattered
electrons do not leave the beam path. The stage can also be negatively biased so
that backscattered electrons can be accelerated toward the detector. A combination
of the magnetic field and stage bias thus enhances detection efficiency. Typically,
the specimen current is set at 0.11 nA, the accelerating voltage at 5 keV, and the
landing energy at 3 keV. “Immersion mode” should be turned on. Typically, two sets
of images are generated: low magnification and high magnification. A fluorescence
image is aligned to the low magnification image. The higher magnification image is
then used to observe the distribution of signals relative to the subcellular structures.
The grayscale of the image should be inverted to resemble a transmission electron
micrograph.
G. Image Alignment
A fluorescence image and an electron micrograph are aligned based on gold fiducial
markers. This is possible because the 100 nm gold nanoparticles emit fluorescence in
red and reflect electrons, marking both fluorescence and electron micrographs (Fig. 6(A
and B). To align the images, a new window of 5000 × 5000 pixels with a resolution at
300 pixels/in. should be opened in Photoshop. The low magnification electron micro-
graph is then copied and pasted to the window and transformed (“command” + “t” for
Mac) so that the image occupies about 80% of the space. The aspect ratio of the image
should not be changed when the image is transformed (holding down the “shift” key
while scaling the image will keep the ratio). The sum TIRF image is copied to a new
layer and transformed (scaling, translation, and rotation) so that the fluorescence from
the gold particles lies on top of particles visualized in the electron micrographs (Fig.
6(B)). For alignment, although some gold particles are visible in a PALM image, a sum
TIRF image should be used because fluorescence from gold particles is more distinc-
tive from the background noise in the sum TIRF image. Nonlinear transformations
may be necessary to align the images perfectly because electron beams can distort the
specimens (Newman & Hobot, 1993). The fluorescence image is then copied to another
layer and aligned on the top of the sum TIRF image using the same transformational
value (Fig. 6(C)).
For presentation purposes, a gradient transparency in the fluorescence image may be
applied so that only the background pixels become transparent. First, the transformed
fluorescence image should be duplicated so that the original file is not manipulated.
Using “color range” under the “select” dropdown menu, select the background “black
pixel” and invert the selection (the “invert” option should be clicked). The fuzziness
should be adjusted so the black pixels surrounding the signals are not selected. Only
now the signals should be selected. The signals are then cut and pasted to a new layer.
This action leaves blanks in the locations where the signals were in the manipulated
Fig.6 Gold nanoparticles are used as fiducial markers to align fluorescence and electron micrographs. Image is a cross section of a
C. elegans adult expressing TOM-20–tdEos. (A) A low magnification SEM image is used to orient and place the high magnification
SEM image. The low magnification image (5000×) is opened in photoshop first. A higher magnification electron micrograph (50,000×) is
aligned on top of the low magnification image. High magnification image is outlined with a white line. Black dots, indicated by arrows,
are the fiducial marks from 100 nm gold nanoparticles applied prior to PALM imaging. (B) A sum TIRF image is aligned onto an electron
micrograph based on fluorescence from the gold fiducials. A “sum TIRF” image represents all of the fluorescence obtained during the
length of the imaging session. (C) A PALM image is then rotated and translated based on the values obtained in (B). (See color plate.)
Fig. 7 Nano-fEM using photoactivation localization microscopy (PALM). (A) Schematic diagram showing the concept of PALM.
Molecules are activated stochastically, imaged, and photobleached. The centroid of each fluorescence spot is then calculated and recon-
structed in a final image. (B) Sum TIRF image of TOM-20–tdEos acquired from a thin section (80 nm) of a GMA-embedded sample.
White arrow indicates fluorescence emitted from gold particles. (C) Corresponding PALM image of TOM-20–tdEos. (D) Electron
micrograph of a body muscle acquired from the same section. (E) Nano-fEM of TOM-20–tdEos. The fluorescent signals are localized
to the outer membrane of the mitochondria. (See color plate.)
layer—this layer thus contains only the background pixels. The transparency should
be set to 10% in the background layer. Now the signals are distinctive from the back-
ground, and can be overlaid on the electron micrograph. The representative images
for correlative microscopy using PALM (Fig. 7(B–E)) or GSDIM (Fig. 8(D–G)) from
the transgenic animals expressing TOM-20::tdEos or TOM-20::Citrine, respectively,
showed similar distribution of proteins on the outer membrane of mitochondria.
IV. Instrumentations and Materials
1. High-pressure freezer (HPM010, Baltec)
2. Specimen carriers type A (#241-200, TechnoTrade)
3. Specimen carriers type B (#242-200, TechnoTrade)
4. Hexadecene (#H2131-100 ML, Sigma-Aldrich)
5. Albumin from bovine serum (A2153-10G, Sigma-Aldrich)
6. Worm buffer M9 (22 mM KH2PO4, 19 mM NH4Cl, 48 mM Na2HPO4, and 9 mM
NaCl)
7. Forceps (insulated; #16LZ01873KN, Leica Microsystem)
Fig. 8 Nano-fEM using ground-state depletion (GSDIM). (A) Schematic diagram showing the concept of GSDIM. Fluorescent pro-
teins, Citrine, are excited and driven into a dark state with intense laser in the presence of an oxygen scavenger. Fluorophores return to
the ground state stochastically and are imaged. The centroid of each fluorescent spot is calculated and localized in a similar manner as
PALM to reconstruct a final image. (B) TIRF image of TOM-20–Citrine acquired from a thin section (80 nm). Fluorescence signals are
diffraction limited because all molecules fluoresce at the same time. (C) TIRF image of the same region imaged in (B). All the molecules
are driven into a dark state (the triplet state) by intense 488 nm laser, and thus no signals are observed. (D) Sum TIRF image recon-
structed from all the fluorescence spots detected by the camera. Background fluorescence is typically higher using GSDIM because
sections cannot be prebleached using the readout laser. (E) Corresponding GSDIM image of TOM-20–Citrine. (F) Electron micrograph
of a body wall muscle acquired from the same section. (G) Correlative GSDIM and electron microscopy of TOM-20–Citrine. The
fluorescent signals are localized to the outer membrane of the mitochondria. (See color plate.)
8. Paintbrush #00 (7/16 inch)

9. Whatman filter paper (#09-810D, Fisher)
10. Isopropanol (#19030-4L, Sigma-Aldrich)
11. Liquid nitrogen
1. Automated freeze-substitution unit (AFS 2, Leica microsystem)
2. Cyrovials (#D9912, Nalgene)
3. 50 ml screw cap conical tubes (#62.547.004, Starstedt)
4. 15 ml screw cap conical tubes (#62.554.002 PP, Starstedt)
5. Acetone (RT10016, EMS)
6. Ethanol (#459844-1L, Sigma-Aldrich)
7. Osmium tetroxide (crystals, 1/10g; RT19134, EMS)
8. Potassium permanganate (RT20200, EMS)
9. Uranyl acetate (#21447-252, Polysciences)
10. Disposable transfer pipette (#14670-201, VWR)
11. Disposable pasteur pipette (Borosilicate glass; #13-678-20A, Fisher)
12. Pipetman and tips (P1000, P200, and P20)
1. GMA kit (low acid and TEM grade; #02630-AA, SPI)
2. N,N-Dimethyl-p-toluidine (#D9912, Sigma-Aldrich)
3. Scintillation vials (20 ml; #72632, EMS)
4. Aclar film (#50425-10, EMS)
5. BEEM capsule (polypropylene; #TC, EBSciences)
6. 3/8” disk punches (#54741, Ted Pella)
7. 25 ml serological pipette (#13-678-11, Fisher)
8. 10 ml serological pipette (#13-678-11 E, Fisher)
9. Petri dish (35 × 10 mm; #351008, Fisher)
D. Coverslip Cleaning
1. Coverslip (#1.5, round; #40300, Warner Instruments)
2. Sulfuric acid (#320501-2.5L, Sigma-Aldrich)
3. Hydrogen peroxide (#216763-500ML, Sigma-Aldrich)
4. Coverslips rack (Teflon; C-14784, Invitrogen)
5. Sonicator
E. Sectioning
1. Ultramicrotome (UC6, Leica microsystem)
2. Diamond knife (Ultra jumbo, 45°, 4.0 mm; DiATOME)
3. Glass strips (#8030, Ted Pella)
4. Glass knife boats (#123-3, Ted Pella)
5. Nail polish (clear)
6. Perfect loop (#70944, EMS)
7. Hair tool for manipulation of plastic sections
8. Razor blade (Double edge; #72000, EMS)
9. High profile microtome blades (#818, Leica microsystem)
F. PALM Imaging
1. Zeiss PAL-M (ELYRA P.1, Zeiss)
2. Gold nanoparticles (call for 2× concentrated solution; #790122-010, micro-
spheres-nanospheres.com)
3. Canned air
4. Attofluor cell chamber for microscopy (#A7816, Invitrogen)
G. Staining
1. Uranyl acetate (#21447-252, Polysciences)
2. Syringe (10 ml)
3. Syringe filter (0.22 µm; SLGP033RB, Millipore)
H. SEM Imaging
1. FEI Nova Nano
2. Backscattered electron detector (vCD, FEI)
3. Carbon fiber cord (#4539, EMS)
4. SEM pin stub (#16111, Ted Pella)
5. Carbon conductive tape (double coated; #16084-7, Ted Pella)
I. Image alignment
1. Photoshop (CS5, Adobe)
V. Discussion
Super-resolution fluorescence microscopy allows one to precisely pinpoint the loca-

tion of a single molecule in a cell (Betzig et al., 2006; Hell & Wichmann, 1994; Hess
et al., 2006; Rust et al., 2006). However, precise localization in a field of black is not
useful. Where is that protein in the context of cellular structure? Where are previously
described organelles in relationship to this protein? Here, we demonstrate that fluores-
cently tagged proteins can be correlated with ultrastructure in electron micrographs
to identify the cellular location of a protein. Five issues in particular bear discussion:
fixation, plastic, imaging, alignment, and quantification.
A. Fixation
Here we use a combination of potassium permanganate and osmium tetroxide
followed by en bloc uranyl acetate staining to balance the fluorescence and morphol-
ogy preservation at synapses. Neuronal tissues are the most sensitive to the fixation
condition (Watanabe et al., 2011), and thus the use of lipid cross-linkers, such as
osmium tetroxide and potassium tetroxide, is essential. One drawback is that these
cross-linkers are also oxidizing agents; consequently, fluorescent proteins may not
be functional after fixation. These chemicals are less reactive at lower temperatures,
so the aforementioned cryomethods are required for preserving fluorescence during
sample preparation. On the other hand, fixation may not be necessary if the tissue
of interest has a naturally high contrast in electron microscopy and thereby requires
no additional fixation and staining with uranyl acetate. In this case, cryosectioning
(Betzig et al., 2006) can be employed. However, such high-contrast tissues or sub-
cellular structures are rare, and thus reaction with electron dense fixatives is required
to elucidate the subcellular details in tissues. Recently, the use of uranyl acetate
alone in the cryosubstitution medium has shown to enhance the contrast of mem-
branes while preserving almost 100% of the fluorescence level in yeast cells and
cell cultures (Kukulski et al., 2011). Therefore, fixation conditions can be altered
according to need.
B. Plastic
The choice of resin media is critical for the functionality of fluorescent proteins and
the quality of sectioning. Fluorescent proteins require a hydrated environment that is
neutral or alkaline in pH. There are two types of resins often used in electron micros-
copy: epoxy and acrylic. Epoxy-based resins have a superior sectioning quality because
they can cross link with tissues; however, they are very hydrophobic, making them
incompatible with fluorescent proteins. Moreover, they require a temperature of 60°C
for polymerization. Fluorescent proteins are denatured at such a temperature. Acrylic
resins, on the other hand, are hydrophilic, fulfilling the requirement for a hydrated envi-
ronment, and can be polymerized at low temperatures by either UV light or chemicals
(Newman & Hobot, 1993). However, most acrylic resins have a low pH––fluorescent
proteins are quenched in such a condition. Here we chose GMA because it fulfilled the
hydrated and alkaline requirements of fluorescent proteins. However, since GMA does
not cross link with tissues, dropouts of tissues are often observed. Ultimately, a plastic
with good features of both types of resins should be developed.
C. Super-Resolution Fluorescence Imaging

To understand the functions of proteins, the relationship of one protein to another
and to subcellular structures must be revealed. Photoactivated localization microscopy
can pinpoint the location of proteins at nanometer resolution, but the applications have
been somewhat limited due to the lack of fluorescent proteins compatible for multicolor
imaging. Here we show that fEM can provide the structural information in the PALM
image. The method described here also introduces the possibility for multiple colors to
be imaged by using GSDIM. Potentially, a large number of preexisting transgenic lines,
expressing GFP, for example, can be imaged at the single-molecule level using GSDIM
instead of requiring the construction of strains using photoactivatable tags such as Eos
for PALM imaging. Although Citrine is unique, and other fluorescent proteins such
as GFP are not compatible with GSDIM imaging, other super-resolution fluorescence
imaging modalities, such as structured illumination microscopy, might be employed to
localize fluorescence in sections before electron microscopy. Moreover, depending on
the application, diffraction-limited fluorescence spots may provide enough information
so that simple correlative fEM is sufficient for one’s imaging needs. For example, the
subtype of neurons in a neuropil can be identified simply by overlaying the diffraction-
limited fluorescence images on the electron micrographs (Micheva & Smith, 2007).
Thus, the choice of fluorescent proteins should be made based on the application. One
must also remember that fEM without super-resolution imaging still produces subdif-
fraction fluorescence images because the axial resolution is determined by the thick-
ness of the section, which is typically about 50 nm, rather than the actual point-spread
function of the fluorescence, which is over 700 nm (Micheva and Smith, 2007).
D. Electron Microscopy Imaging

For fEM, we have been using a SEM extensively, but TEM can also be used.
However, two factors should be considered before using TEM. First, sections must
be ultrathin for TEM imaging. Typically, 50- to 70 nm-thick sections are imaged by
TEM. If specimens are very darkly stained with osmium tetroxide, contrast can still
be obtained using a typical accelerating voltage of 100 keV. However, because the
primary fixative used here is potassium permanganate, the tissue contrast is low with
such a high accelerating voltage. A lower voltage can be used to enhance the contrast,
but in this case, the sections must be even thinner, but obtaining such thin sections
with GMA-embedded specimens can be challenging. Moderately high-resolution
images (~5 nm resolution) can be achieved from the surface of the specimens in SEM
with a low accelerating voltage (~5 keV), and thus the thickness of sections is not an
issue. Second, a grid with lateral dimensions of 2 mm × 1 mm must be used for TEM
imaging. Therefore, one must manipulate many ribbons of sections without losing
sections or losing the order of the sections, whereas with SEM, long ribbons of sec-
tions can be mounted on the coverglass. Therefore, unless resolution better than 5 nm
is necessary to capture cellular details, SEM offers advantages over TEM.
E. Alignment
A fluorescence image and an electron micrograph are aligned based on gold fiducial
markers, but the alignment may not be precise for two reasons. First, gold particles emit
fluorescence constantly. Because each particle is not stochastically activated, the fluo-
rescence signals from a few neighboring gold particles appear as one fluorescent spot
due to the diffraction limit of light. On the other hand, electron microscopy depicts the
actual dimensions of the particles. The disparity in the resolution results in some ambi-
guity in the alignment. Second, electron beams can distort the specimen even though
the sections are mounted on a solid glass surface. The distortion is usually more severe
in acrylic resins such as GMA. Aligning a portion of the image or applying nonlinear
transformations could help overcome this distortion problem.
F. Quantification
Counting the number of fluorescent molecules in PALM images is very tempting, but
one must consider three factors that affect the number of molecules present in each spot
of an image. First, fluorescence is lost during sample preparation. We reported above that
most of the fluorescence is preserved when measured in our specimens at the beginning
and end of the fixation and embedding procedure, but we may be overestimating fluores-
cence because of mechanical damage to the fluorophores during sectioning and oxidation
after sectioning has not been quantified. Second, the stochastic nature of photoactivatable
fluorescent proteins affects quantification. Each molecule must be activated separately
in space and time and must be permanently bleached after emitting fluorescence for
calculating the precise number of molecules. However, fluorophores can be easily under-
counted or overcounted. The fluorophores can emit too little light and fail to be counted.
Alternatively, they may return from the triplet dark state instead of becoming bleached
and be counted twice. This uncertainty is compounded in GSDIM, in which overcount-
ing is routine and thus this method is not quantitative. Third, tagged proteins are typically
not expressed at the endogenous level. Therefore, quantification cannot be precise.
VI. Perspective
Nano-resolution fEM (nano-fEM) using super-resolution fluorescence microscopy

can reveal the precise location of proteins within a cell. Unlike immuno-EM, protein
localization is limited only by the ability to tag the protein of interest with fluoro-
phores. This method can thus reveal the localization of proteins more quantitatively
and at very high resolution. In the future, a few improvements could greatly extend
the utility of this method. These improvements include the development of hydro-
philic plastics that cross link with tissues, fluorophores that are osmium resistant,
photoactivatable fluorescent proteins that can be imaged simultaneously in two color
spectra, and a method to combine three-dimensional PALM imaging with electron
tomography.
Acknowledgments
We thank Harald Hess and Eric Betzig for access to the PALM microscope for
proof-of-principle experiments; Rick Fetter for sharing fixation protocols, reagents,
and encouragement; Michael Davidson, Geraldine Seydoux, Stefan Eimer, Rudolf
Leube, Keith Nehrke, Christian Frøkjær-Jensen, Aude Ada-Nguema, and Marc Ham-
marlund for DNA constructs; Gunther Hollopeter for generating transgenic lines
used in these experiments; Jackson Richards for a critical reading of the manuscript;
and Carl Zeiss Inc. for providing access to the beta version of the Elyra PALM
microscope.
References
Cajal, S.R.Y. (1894). Die Retina der Wirbelthiere: Untersuchungen mit der Golgi-Cajal’schen Chromsilber-
methode und der Ehrlich’schen Methylenblaufärbung (Wisbaden, Bergmann).
Cajal, S.R.Y. (1899). Comparative study of the sensory areas of the human cortex. Worcester: Clark Uni-
versity.
Cajal, S.R.Y. (1903). Studien über die Hirnrinde des Menschen (Johann Ambrosius Barth).
Fölling, J., Bossi, M., Bock, H., Medda, R., Wurm, C. A., Hein, B., et al. (2008). Fluorescence nanoscopy by
ground-state depletion and single-molecule return. Nature Methods, 5, 943–945.
857–862.
Gurskaya, N. G., Verkhusha, V. V., Shcheglov, A. S., Staroverov, D. B., Chepurnykh, T. V., Fradkov, A. F.,
et al. (2006). Engineering of a monomeric green-to-red photoactivatable fluorescent protein induced by
blue light. Nature Biotechnology, 24, 461–465.
Harke, B., Keller, J., Ullal, C. K., Westphal, V., Schönle, A., & Hell, S. W. (2008). Resolution scaling in
STED microscopy. Optics Express, 16, 4154–4162.
Heikal, A. A., Hess, S. T., Baird, G. S., Tsien, R. Y., & Webb, W. W. (2000). Molecular spectroscopy and
dynamics of intrinsically fluorescent proteins: coral red (dsRed) and yellow (Citrine). Proceedings of the
Heilemann, M., van de Linde, S., Schüttpelz, M., Kasper, R., Seefeldt, B., Mukherjee, A., et al. (2008).
Subdiffraction‐resolution fluorescence imaging with conventional fluorescent probes. Angewandte
Chemie International Edition, 47, 6172–6176.
Hell, S. W., & Wichmann, J. (1994). Breaking the diffraction resolution limit by stimulated emission: stim-
ulated-emission-depletion fluorescence microscopy. Optics Letters, 19, 780–782.
Hell, S. W. (2007). Far-field optical nanoscopy. Science, 316, 1153–1158.
Hess, S. T., Girirajan, T. P. K., & Mason, M. D. (2006). Ultra-high resolution imaging by fluorescence pho-
toactivation localization microscopy. Biophysical Journal, 91, 4258–4272.
Hooke, R. (1665). Micrographia: Or some physiological descriptions of minute bodies made by magnify-
ing glasses with observations and inquiries thereupon Jo. Matryn and Ja. Allestry. London: The Royal
Society.
Knoll, M., & Ruska, E. (1932). Elektronenmikroskop. Zeitschrift Physikalische, 78, 318–339.
Kukulski, W., Schorb, M., Welsch, S., Picco, A., Kaksonen, M., & Briggs, J. A. G. (2011). Correlated fluo-
Biology, 192, 111–119.
Lalkens, B., Testa, I., Willig, K. I., & Hell, S. W. (2012). MRT letter: nanoscopy of protein colocalization in
living cells by STED and GSDIM. Microscopy Research and Technique, 75, 1–6.
Link, S., & El-Sayed, M. A. (1999). Spectral properties and relaxation dynamics of surface plasmon elec-
tronic oscillations in gold and silver nanodots and nanorods. The Journal of Physical Chemistry B, 103,
8410–8426.
McCutchen, C. W. (1967). Superresolution in microscopy and the Abbe resolution limit. Journal of the Opti-
cal Society of America, 57, 1190–1192.
McDonald, K. L., Morphew, M., Verkade, P., & Müller-Reichert, T. (2007). Recent advances in high-pressure
McDonald, K., Schwarz, H., Müller-Reichert, T., Webb, R., Buser, C., & Morphew, M. (2010). “Tips and
tricks” for high-pressure freezing of model systems. Methods in Cell Biology, 96, 671–693.
McDonald, K. (2007). Cryopreparation methods for electron microscopy of selected model systems. Meth-
ods in Cell Biology, 79, 23–56.
Micheva, K., & Smith, S. (2007). Array tomography: a new tool for imaging the molecular architecture and
ultrastructure of neural circuits. Neuron, 55, 25–36.
Morphew, M. K. (2007). 3D immunolocalization with plastic sections. Methods in Cell Biology, 79, 493–513.
electron microscopy of early Caenorhabditis elegans embryos in mitosis. In cellular electron microscopy.
Oxford, Elsvier. pp. 101–119).
Müller-Reichert, T. (2010). (1st ed.). Electron microscopy of model systems. (Vol. 96). New York: Academic
Press.
Murphy, D. B. (2001). Fundamentals of light microscopy and electronic imaging. Hoboken, NJ: John Wiley
and Sons.
Newman, G. R., & Hobot, J. A. (1993). Resin microscopy and on-section immuno-cytochemistry. Berlin,
Germany: Springer-Verlag.
10, 131–136.
Oberti, D., Kirschmann, M. A., & Hahnloser, R. H.R. (2010). Correlative microscopy of densely labeled
projection neurons using neural tracers. Frontiers in Neuroanatomy, 4, 24.
Patterson, G. H., & Lippincott-Schwartz, J. (2002). A photoactivatable GFP for selective photolabeling of
proteins and cells. Science, 297, 1873–1877.
Payne, J. W. (1973). Polymerization of proteins with glutaraldehyde. Soluble molecular-weight markers.
Journal of Biochemistry, 135, 867–873.
Polishchuk, R. S., Polishchuk, E. V., Marra, P., Alberti, S., Buccione, R., Luini, A., et al. (2000). Correlative
apparatus and plasma membrane. The Journal of Cell Biology, 148, 45–58.
Rostaing, P., Weimer, R. M., Jorgensen, E. M., Triller, A., & Bessereau, J.-L. (2004). Preservation of immu-
noreactivity and fine structure of adult C. elegans tissues using high-pressure freezing. The Journal of
Histochemistry and Cytochemistry, 52, 1–12.
Roth, J., Bendayan, M., Carlemalm, E., Villiger, W., & Garavito, M. (1981). Enhancement of structural pres-
ervation and immunocytochemical staining in low temperature embedded pancreatic tissue. The Journal
of Histochemistry and Cytochemistry, 29, 663–671.
Schikorski, T. (2010). Horseradish peroxidase as a reporter gene and as a cell-organelle-specific marker
in correlative light-electron microscopy. In S. D. Schwartzbach, & T. Osafune (Eds.), Immunoelectron
microscopy (pp. 315–327). Totowa, NJ:: Humana Press.
Shimomura, O. (2009). Discovery of green fluorescent protein (GFP) (Nobel Lecture). Angewandte Chemie
International Edition in English, 48, 5590–5602.
Shtengel, G., Galbraith, J. A., Galbraith, C. G., Lippincott-Schwartz, J., Gillette, J. M., Manley, S., et al.
(2009). Interferometric fluorescent super-resolution microscopy resolves 3D cellular ultrastructure. Pro-
ceedings of the National Academy of Sciences, 106, 3125–3130.
Biology, 9, e1001041.
369, 291–308.
Sosinsky, G. E., Giepmans, B. N.G., Deerinck, T. J., Gaietta, G. M., & Ellisman, M. H. (2007). Markers for
Tsien, R. (1998). The green fluorescent protein. Annual Review of Biochemistry, 67, 509–544.
microscopy and high throughput for high‐pressure freezing. Journal of Microscopy, 230, 317–328.
Weibull, C., & Christiansson, A. (1986). Extraction of proteins and membrane lipids during low temperature
embedding of biological material for electron microscopy. Journal of Microscopy, 142, 79–86.
Wiedenmann, J., Ivanchenko, S., Oswald, F., Schmitt, F., Röcker, C., Salih, A., et al. (2004). EosFP, a flu-
orescent marker protein with UV-inducible green-to-red fluorescence conversion. Proceedings of the
Zhang, J., Campbell, R. E., Ting, A. Y., & Tsien, R. Y. (2002). Creating new fluorescent probes for cell biol-
ogy. Nature Reviews. Molecular Cell Biology, 3, 906–918.
CHAPTER 16
Atmospheric Scanning Electron

Microscope for Correlative Microscopy
Ian E.G. Morrison*, Clare L. Dennison*, Hidetoshi Nishiyama†,
Mitsuo Suga†, Chikara Sato‡, Andrew Yarwood§
and Peter J. O’Toole*,¶
* Technology Facility, Biology Department, University of York,York YO10 5DD, UK
† JEOL Ltd., Advanced Technology Division, Akishima, Tokyo 196-8558, Japan
‡ Biomedical Research Institute, National Institute of Advanced Industrial Science and Technology, Tsukuba,
Ibaraki 305-8568, Japan
§ JEOL UK Ltd., Jeol House, Watchmead, Welwyn Garden City, Herts AL7 1LT, UK
¶ Corresponding author
Abstract
I. Introduction: Atmospheric Scanning Electron Microscopy
II. Design of the JEOL JASM-6200
III. Sample Preparation
A. ASEM Dish Preparation for Culture and Stabilizing of Cells and Bacteria
B. Staining of Samples for the ClairScope
C. Immunogold Labeling
D. Heavy Metal Staining
IV. Application Studies
A. Nonbiological: Studies Using Heated or Electrochemical ClairScope Dishes
B. Biological Studies of Cells in Culture or Tissue Sections
C. Cell Surface Receptors and Lipid Rafts
D. Examination of Mycoplasma in Solution
E. Detection of Cathodoluminescence from Phosphor Particles as Correlative Markers
V. Discussion
Acknowledgments
References
Abstract
The JEOL ClairScope is the first truly correlative scanning electron and optical
microscope. An inverted scanning electron microscope (SEM) column allows electron
308 Ian E.G. Morrison et al.
images of wet samples to be obtained in ambient conditions in a biological culture dish,

via a silicon nitride film window in the base. A standard inverted optical microscope
positioned above the dish holder can be used to take reflected light and epifluorescence
images of the same sample, under atmospheric conditions that permit biochemical
modifications. For SEM, the open dish allows successive staining operations to be per-
formed without moving the holder.
The standard optical color camera used for fluorescence imaging can be exchanged
for a high-sensitivity monochrome camera to detect low-intensity fluorescence signals,
and also cathodoluminescence emission from nanophosphor particles. If these particles
are applied to the sample at a suitable density, they can greatly assist the task of perfect-
ing the correlation between the optical and electron images.
I. Introduction: Atmospheric Scanning Electron Microscopy
Electron microscopes (EM) and optical microscopes (OM) have one great practi-
cal difference when working with biological materials. An EM sample will be under a
high vacuum to prevent dispersion of the electron beam, and hence must be either (a)
fixed and dried or (b) frozen and sectioned, and metal-coated. Conversely, visible light
microscopy is nearly always performed in atmospheric conditions where the biological
sample is fully hydrated and in many cases still living, in the sense that biochemical
turnover is still occurring. This fundamental difference has been a major difficulty in
developing truly correlative microscopy techniques.
Environmental EM has been investigated along two different approaches. One uti-
lizes a differential pressure sample chamber, so that electrons can pass through a series
of orifices between the electron beam chamber and a sample container at saturated
water pressure to prevent drying (Danilatos, 1988). The second technique is to enclose
the biological material in a capsule behind a membrane that is strong enough to support
a high vacuum, yet thin enough to be electron transparent. Such membranes can be
supported polymeric films (Thiberge, Nechustan, et al., 2004; Thiberge, Zik, & Moses,
2004) or low atomic mass compounds such as silicon nitride (Creemer et al., 2008).
In both these cases, there is no access to the container while in the microscope,
so that experiments on the sample cannot be carried out in situ. In capsules, the thin
membrane is susceptible to damage by changes in the sample volume, e.g., by emission
of gas or thermal expansion. To address these points and also to enable true correla-
tive imaging, in 2010 JEOL Japan introduced the JASM-6200, a combination of OM
and EM (Nishiyama et al., 2010). This uses a modified cell culture dish as the sample
holder. In the center of the dish is a silicon chip (Fig. 1(A)) with a layer of silicon
nitride on the inside, and a 250 µm square is thinned externally to expose the silicon
nitride film window, typically 100 nm thick. The underside of the dish is seated against
a mechanical support and a vacuum seal. Optical microscopy is performed from above
and outside the EM vacuum chamber and thus at atmospheric pressure, usually with
a water-dipping objective lens, while the electron imaging is from below, using an
inverted scanning electron microscope (SEM) column (Fig. 1(B)).
16. Atmospheric Scanning Electron Microscope for Correlative Microscopy 309
Fig. 1 (A) The ClairScope culture dish. The silicon chip in the middle is 4 mm square, and at the center of the chip can be seen the
250 µm square window of SiNx, 100 nm thick. (B) Diagram of the ClairScope. The electron beam from the inverted SEM passes through the
window of the dish shown in (A), and backscattered electrons are detected by the BEI detector. The optical microscope above the dish has a
water-dipping 40× lens and images are recorded by the camera above. (C) “Brightfield” and FITC fluorescence images of cells growing on
a ClairScope dish. The edges of the left-hand image show blue in reflected light from the silicon, and the green center is from the SiNx window.
A549 lung carcinoma cells were stained with a lipid raft marker (cholera toxin b subunit, linked to Alexa Fluor-488) and appear as dots.
(See color plate.)
Since the dish contents are under ambient conditions, sample processing can be
performed without removing the dish from the microscope; it is also possible to control
temperature and atmospheric composition, and electric fields can be applied using spe-
cialized electrode dishes. This new system is called an atmospheric scanning electron
microscope (ASEM); the hybrid system of the OM and the ASEM is called ClairScope
JASM-6200 (JEOL Ltd.).
II. Design of the JEOL JASM-6200
A 100 nm silicon nitride film will support 1 atmospheric pressure differential, while
being thin enough for minimal scattering of 10–30 keV electrons. The electron beam
passes through a flight tube, which can be removed for cleaning. Backscattered electrons
are registered in a detector at the top of the column under the dish. To prevent con-
tamination in the event of a film failure, the ASEM has a countermeasure, the column
protection system. This consists of a shutter, an inner pipe, and an air-leak valve. The
shutter has a 4 ml capacity to retain the sample, and is located between the ASEM dish
and the ASEM column. The inner pipe with an orifice is located in the ASEM column
to prevent sample material from entering the gun. The air-leak valve function is to
return the space under the ASEM dish to atmospheric pressure immediately in the event
of a failure of the silicon nitride film.
Electron imaging can be performed at magnifications between 100× and 100,000×,
and the optical image corresponds to 430×. The electron gun operates at 10, 20, or 30
kV, and is capable of a resolution of 8 nm (Nishiyama et al., 2010), but the practical
resolution when imaging through the 100 nm SiNx film depends on the sample and the
depth of the target in the hydrated layer (Fig. 2). Images are recorded at 20–160 s per
frame, but lower resolution videos can be recorded at speeds up to 13 frames s−1.
The inverted optical microscope system is a modified Olympus BX head, with a
dichromatic filter selector controlled through software. Illumination is from a standard
100 W mercury lamp, through an optic fiber, so that any suitable excitation source
could be coupled in; intensity can be varied by a software-controlled neutral density
filter set. Nonfluorescence images are obtained in reflection mode (Fig. 1(C)), and cap-
tured by a QI Retiga 2000R color camera having 1600 × 1200 pixels. When using the
standard Olympus LUMPlanFL 40× NA0.80 water immersion lens, the camera relay
lens provides 0.185 µm per pixel, a little less than required by the Nyquist criterion
(0.162). The practical resolution limit with this camera is found to be about 300 nm.
The Kodak KAI-2020 sensor in the camera has 37% quantum efficiency (QE) at
540 nm and 31% at 620 nm (www.kodak.com/go/imagers). This is sufficient for many
applications; some higher efficiency cameras can be fitted in its place (e.g., Hamamatsu
Fig. 2 (A) SEM image of part of a U251 glioblastoma cell at magnification 7500×, stained with platinum
blue. (B) Cross sections through parallel fibers were summed for 7 pixels along the length, and fitted by three
Gaussian peaks constrained to have the same width, which was found to be 30 nm.
ORCA) but currently need to be operated in the standalone mode on a separate com-
puter, as the JEOL software is only suitable for the Retiga. The image output can be up
to 18 frames s−1, but fluorescence exposure times are often in the order of seconds. The
ClairScope package controls the two imaging systems through very similar graphical
user interfaces (GUIs), but true simultaneous imaging is not readily possible so as to
protect the electron detector.
There are some restrictions on use. The dish window is susceptible to damage
by strong acids or alkalis, but has been found to withstand molar acid for several
minutes with no effect. As with all environmental electron imaging, the penetration
depth of the electrons into the aqueous medium is restricted to a few microns and
depends on the accelerating voltage. This in itself can be a useful factor in restrict-
ing background signal, and thus improving signal-to-noise ratio when studying fine
detailed structures.
III. Sample Preparation
A. ASEM Dish Preparation for Culture and Stabilizing of Cells and Bacteria
The 35 mm polystyrene culture dish can be used for nearly all tissue culture opera-
tions; standard coating protocols can be applied, such as poly-l-lysine, PEI, and col-
lagen. The dish holds a maximum of 8 ml of culture medium, but volumes can be
reduced to <500 µl by a Teflon ring around the silicon chip, so that high-cost media
such as cell stimulants can be used more economically. Specialized high temperature
dishes, made of titanium with a ceramic heater, have been used for observations of
crystal growth for materials science studies, and cells with electrodes can be used, e.g.,
for electro-stimulation of neuronal cells.
Protocol 1: Culturing Cells in ClairScope Dishes
Standard culture methods are used, but since the observation area is 250 µm square, it may be neces-
sary to seed the cells at a 2-fold higher density. Adherent cells can be grown in the dish without pretreat-
ment, but to improve adhesion the surface may be coated by normal methods such as poly-l-lysine, or
polyethylimine, which can also assist transfection (Vancha et al., 2004):
Poly-l-lysine (Sigma, P8920) is diluted 10-fold, and 200 µl is applied as a drop at the center of
the dish, or 2 ml to the entire dish. Alternatively, polyethylimine (Fluka, P3143) is diluted in 150 mM
NaCl solution, and applied to the dish at 25 µg ml−1. After incubation at room temperature for 10–20
min, the dish is washed in ultrapure water three times.
These adhesion treatments may also be used to stabilize nonadherent cells, organelles, or bacteria,
on the dish window. Other typical coatings are albumin (Yamazoe, Uemura, & Toshizumi, 2008),
fetuin, collagen, and fibronectin (Sato et al., 2012).
B. Staining of Samples for the ClairScope

To obtain good contrast in optical and electron images of biomaterials, and to identify
the target (organelle, receptor complex, protein, etc.), it is generally necessary to stain
the material. In OM, fluorescent staining by targeted antibodies is often the method of
choice, and the sensitivity of detection is now at the single molecule level in some situ-
ations (Sako, 2006). The standard camera and 40× objective lens of the ClairScope is
not able to image with such extreme sensitivity, but the camera can be replaced, e.g., by
an electron multiplying (EMCCD) detector, in combination with a higher powered lens
such as an Olympus LUMFLN 60× water immersion having NA1.1 (Sato et al., 2012).
C. Immunogold Labeling
Similar targeted staining can be attained in EM using antibody- or Fab-fragment
tagged nanoparticles such as gold; nanogold (1.4 nm diameter: Nanoprobes, NY 11980,
USA) gives the highest resolution but cannot be detected by standard SEM except by
enhancing by silver or gold. The enhancement procedure (GoldEnhance™, Nanoprobes,
NY 11980, USA) increases the diameter of individual particles above the resolution
limit of the ClairScope SEM. The protocols are detailed in the manufacturer’s instruc-
tions, but will need optimization for the precise cell line and cellular target required.
Nonspecific enhancement can occur, and can be minimized by selective prewashing, e.g.,
in EDTA to remove extraneous metal ions.
When applied to nanogold inside cells, the enhancement procedure can result in
nonuniform size and intensities as shown in Fig. 3(A). The B104 neuroblastoma cells
were transfected with a pH sensitive version of GFP, and after fixing and permeabiliz-
ing, the GFP was labeled with a monoclonal antibody, with a 1.4 nm nanogold Fab, and
finally gold enhanced. The size and intensity of the gold particles were found by fitting
two-dimensional Gaussians on sloping backgrounds to the spots; they have a wide
range of diameters and intensities, shown by histograms 3b and 3c where the widths
have a coefficient of variation (CV) of 33%.
Alternatively, for precise molecular counting inside cells, gold nanoparticles larger than
the resolution limit of the SEM may be bound to an epitope by using a secondary antibody,
e.g., anti-GFP with 20 nm gold beads. These have a very narrow size distribution, and
when imaged on a dry silicon nitride film they have the size distribution as shown in Fig.
3(G) with CV = 10%, and the intensity spread shown in Fig. 3(H). The distributions found
in cells are shown in Fig. 3(I and J), in a small area of a B104 cell identified by fluorescence
in Fig. 3(D) and enlarged in Fig. 3(E). The larger range of apparent widths (CV = 19%)
is probably due to the increased distance of the particles from the film window; the higher
intensity spots will result from particles in the same position but separated in the z-direction.
D. Heavy Metal Staining

More often, staining is accomplished by heavy metals, but some of them are non-
specific, and often require severe chemical treatments (e.g., osmium staining has to
Protocol 2 Immunolabeling Fluorescent Protein with 20 nm Gold Particles
Cells transfected with the GFP gene are cultured on the ClairScope dish at a suitable density: 104
cells per dish for B104 neuroblastoma cells. Cells can be differentiated on the dish while under cul-
ture; to reduce sample volumes while preventing evaporation, the volume of medium can be reduced
by surrounding the silicon chip by a hollow Teflon disc (1 mm thick, 15 mm inner diameter, sealed
to the dish with a thin layer of silicone grease).
They are then fixed, e.g., by paraformaldehyde (30 min in 4% PFA and 4% sucrose in water, or 2%
PFA in phosphate buffered saline PBS), and washed three times in a blocking solution of 1% bovine
serum albumin in PBS.
The cells are permeabilized in 0.1% Triton X-100, 1% BSA in PBS for 20 min and again washed
three times in blocking solution. Primary labeling is then performed by incubating with anti-GFP
(e.g., Invitrogen A11120, mouse IgG2a, monoclonal 3E6) in 1% BSA/PBS at a concentration of 200
ng ml−1. After incubating for 2 h in the dark, in a humid atmosphere, the cells are washed three times
in 1%BSA/PBS.
The secondary gold label (e.g., Abcam 27242, Goat anti-mouse IgG, 20 nm gold conjugate) is
diluted one part in five of 1% BSA/PBS containing 0.1% Triton X-100, and applied carefully as a
20 µl drop on the window to reduce reagent volume, and incubated for 1 h in the dark and in humid
conditions. The cells are then washed three times in 1% BSA/PBS before finally filling the dish with
dextrose-500 (10 mg ml−1 in PBS).
Protocol 3 Staining Biomaterials with Platinum Blue
Platinum blue (commercial name TI Blue, Nisshin EM, Japan, 6% solution) is diluted 10-fold, and
applied to the dish for a suitable period; 10 min may be enough for light staining of cells, or longer
to increase the contrast level. After this period, excess stain can be removed and the sample washed
several times with ultrapure water. The open dish can remain in the instrument while this is done,
and thus restaining can be performed. If the staining protocol requires toxic materials, e.g., osmium,
the dish holder can be removed to a fume hood and replaced with a fairly high degree of accuracy.
For bacteria such as E. coli, the preferred dilution of platinum blue is 3-fold.
be carried out in a hood) but can be performed on fully hydrated materials. An alter-
native material, platinum blue (Inaga et al., 2007), stains cells in a different pattern
to uranyl acetate and can give good contrast in a few minutes at ambient conditions;
although it is toxic, with careful handling it can be applied in situ in the ClairScope.
The size of features identified by this stain can be determined as shown in Fig. 2,
where Fig. 2(A) is an image of a U251 glioblastoma cell labeled by the above protocol
using 0.6% stain solution for 30 min. The thin features can be found to have a width of
Fig. 3 (A) Part of a B104 neuroblastoma cell expressing synapto-pHluorin (Miesenböck, De Angelis,
& Rothman, 1998); labeled with anti-GFP and Fab-linked nanogold, and enhanced with GoldEnhance.
The cell shows a large number of gold particles, whose widths and intensities were quantified by fitting a
two-dimensional Gaussian function to the local pixels around the estimated centroid, with a local sloping
background. The widths and intensities above background are collected into histograms (B) and (C). (D)
Fluorescence images of B104 cells: overlay of DAPI (blue) and synapto-pHluorin (green). The box at lower
right is enlarged in (E). The cells were then labeled with anti-GFP and Fab-linked nanogold, and imaged by
SEM (F); the exact position of the area in the SEM image cannot be identified, but is estimated to be in the
box shown in (D). (G) and (H) Data for 20 nm gold particles dispersed on a polylysine-coated ClairScope
dish, and imaged and analyzed as in (B) and (C). Widths and intensity histograms show highly uniform
widths. (I) and (J) Data for 20 nm gold particles in (F) analyzed as mentioned earlier. The broader distribu-
tion of widths is probably due to the increased distance from the film when inside the cell, but there may be
a number of unresolved pairs of particles. (The intensity scales are arbitrary as the intensity depends on the
contrast and brightness.)
30 nm (Fig. 2(B)); this value was obtained for fibers in the body of the cell (i.e., away
from the film window) and using spot size 40; this value could be reduced but at the
cost of decreased signal-to-noise ratio.
Prolonged imaging by SEM of wet materials can lead to damage by generation of
radicals. To reduce this effect, the medium should be supplemented by 10 mg ml−1 of a
radical trapping agent such as ascorbate, glucose, or dextran-500.
IV. Application Studies
A. Nonbiological: Studies Using Heated or Electrochemical ClairScope Dishes

The ASEM can be used for investigations of chemical and physical processes in
the liquid state at sub-100 nm resolution (Suga et al., 2011). The instrument has been
employed to study crystallization processes, and melting and solidification of low melt-
ing point alloys such as solder in a dish heated up to 250°C. An electrode dish has
been used to observe electrochemical changes. Typically, growth of disordered metal
structures has been observed in electrolytes, in real time and with quasi-synchronous
observation at better than 50 nm resolution by ASEM and mesoscopic imaging by the
optical microscope. This combination of the two methodologies will also permit, e.g.,
observation of evolution of gas bubbles in these types of experiment, which is not possible
with conventional SEM.
Additionally, the ordering of particles such as protein aggregates or silica spheres
into close-packed structures has been seen to be induced by electron beam irradiation
(Suga et al., 2011). The clustering and dispersion of such particles depends critically
on the electron dose density, so that magnification or probe current changes can switch
the particle ordering behavior on and off. Salt concentration in the liquid medium
is also a factor as it changes the depth of the ionic layer at the surface of the particles
(L. G. Gardiner & J. Yuan, personal communication).
B. Biological Studies of Cells in Culture or Tissue Sections

The first publication featuring the ClairScope showed a wide range of biological
applications (Nishiyama et al., 2010). The first version of the silicon nitride window
dish used 30 nm thick film, which resulted in slightly better resolution than the pro-
duction 100 nm thickness; the observed width of gold beads on the 30 nm thick film
approached the actual 15 nm value, while with film thicknesses greater than 100 nm,
the resolution was impaired and the noise and apparent diameter increased. On the thin
edges of COS7 cells labeled with wheat germ agglutinin-colloidal gold conjugates,
neighboring 15 nm beads could be resolved.
PC12 cells can be cultured to develop neuronal-type synapses (Nishiyama et al.,
2010). These synapses were visualized by staining with platinum blue after Triton X100
permeabilization, and show a fine network of intercellular connections. A similar stain-
ing approach in COS7 cells was used to image ultrastructural details of the endoplasmic
reticulum (ER), which was identified by using the fluorescence microscope to locate a
specific immunolabel, protein disulfide isomerase antibody with an Alexa Fluor-488
secondary. For cells where the ER was very close to the lower cell membrane, and
hence close to the film window, fine detail could be observed with structures much
narrower than 100 nm.
Platinum blue was found to preferably stain chromosomes (Nishiyama et al., 2010).
In COS7 cells that had been fixed, permeabilized and stained as discussed earlier,
nuclei were visible as a lighter shade above the surrounding stained cytoplasm, with
nucleoli especially bright. Mitotic cells showed much brighter with the chromosomes
clearly separated, though the resolution is reduced due to the mitotic nucleus being
located deeper inside the more rounded cell (Fig. 4).
Tissue sections could be stained by platinum blue, and the images of these were
compared with similar sections stained using heavy metal protocols (uranyl acetate
or osmium tetroxide). The section may not lie completely flat on the film window but
applying a cover slip to the top of the tissue can improve contact. Platinum blue was
found to be much better for location of nuclei than the conventional protocols, and a
combination of stains can be used to show up the desired features of the tissue.
C. Cell Surface Receptors and Lipid Rafts

Lipid rafts are suitable objects to be studied by the ASEM. These objects are domains
of the cell plasma membrane that are enriched in cholesterol and sphingolipids (Simons
& Ikonen, 1997), and are believed to have sizes ranging from 10 to 200 nm. The rafts
are believed to act as a focus point for membrane proteins, where they can interact
and thus form the start of a signaling cascade. They can be labeled by the β-subunit of
cholera toxin (CTxβ), which binds preferentially to the ganglioside GM1, but their size
is below the resolution limit imposed by optical methods, and there is always the ques-
tion whether binding of the marker is affecting the temporal and spatial dimensions of
Fig. 4 Chromosomes in COS7 cells, fixed with glutaraldehyde and stained with platinum blue. (A) The mitotic
cells are much brighter than other nuclei at magnification 450×, and (B) at 4000×, the chromosomes are vis-
ibly extended. The distance of the chromosomes from the lower cell face on the SiNx film limits the resolution.
the raft. A typical image of A549 lung carcinoma cells labeled with Alexa Fluor-488
CTxβ is shown in Fig. 1(C).
A study of the ultrastructural characteristics of the cell surface receptor CD44
employed the ASEM among several techniques, to elucidate the effect of cholesterol
on the localization and clustering of the adhesion molecule in lipid rafts (Murai et al.,
2010). CD44 acts as a receptor for the extracellular matrix moiety hyaluronan, and
cleavage and release of the external domain of the membrane protein (shedding) is a
major factor in the migration of tumor cells. In the glioblastoma cell line U251, the CD44
molecules are concentrated in rafts; the study examined the distribution of CD44 in the
cell membranes, as cholesterol was depleted by the action of methyl β-cyclodextrin
(MβCD). Receptors were labeled by a monoclonal antibody (Hermes-3), and then by
fluorescent (Alexa Fluor-488) or nanogold conjugated goat anti-mouse secondary. The
1.4 nm gold particles were then enhanced using GoldEnhance EM (Nanoprobes) to
enlarge the gold particles to a size suitable for the ASEM, and then imaged in 10 mg
ml−1 ascorbic acid buffer.
Some typical ASEM images are shown in Fig. 5. The images revealed that the gold
particles formed aggregates in the size ranging between 100 and 1000 nm in normal
U251 cells, but were widely dispersed and nonclustered in cells treated with MβCD.
These images supported results from several different techniques that cholesterol
depletion affects the clustering and localization of CD44 in the plasma membrane.
When the depletion is modulated by a statin agent, CD44-dependent cell mobility was
suppressed; this could be an effective treatment for reducing the progression of malig-
nant tumors (Murai, 2012).
D. Examination of Mycoplasma in Solution

Mycoplasma is the name given to a group of bacteria lacking a cell wall that are the
smallest known free-living organisms (Razin, Yogev, & Naot, 1998). They are enclosed
in a three-layer lipid membrane typically with diameters <300 nm, so that individual
cells are difficult to identify by light microscopy, and subcellular structures are very
difficult to image. The complete thickness of the cell is clearly within the 3 µm electron
penetration depth of the ASEM, and hence is very suitable for study by the ClairScope
(Sato et al., 2012).
The mobility of these bacteria is believed to depend on flexible projections from a
protusion (Miyata, 2010); this method is very different from the cytoskeletal or motor
protein type of mobility found in other organisms. The end of each projection can bind
to sialic acid moieties on the host organism cells, and is complicit in the infection pro-
cess by which Mycoplasma pneumoniae can be found in many asthma and respiratory
tract illnesses.
The small size of the organism makes diagnosis of infection difficult and slow, and
so there is a need for a super-resolution microscopy technique to identify pathogens
at an early stage of infection, when their distinctive morphology could be recognized.
The ClairScope was used to identify the cells, which can be attached by coating the
silicon nitride film with fetuin, by heavy metal staining using a sequence of platinum
Fig. 5 ASEM observation of CD44 distribution on the cell surface: (A–E) control cells or (F–J) treated with 5
mM MβCD for 1 h and fixed with 4% paraformaldehyde. CD44 proteins on the cell surface were labeled with
anti-CD44 mAb Hermes-3 and then with Alexa Fluor 488- and Nanogold (1.4 nm)-conjugated goat anti-mouse
IgG followed by gold enhancement. (A and F) Fluorescence images, (B and G) SEM images at 1500× in ascorbic
acid solution, (C and H) at 3000×, (D and I) at 10,000×, and (E and J) at 30,000× magnification. (See color plate.)
blue, tannic acid, osmium tetroxide, uranyl acetate, and lead citrate. Alternatively, they
could be labeled by immuno-fluorogold and gold enhancement (Sato et al., 2012).
The cells of Mycoplasma mobile are bottle-shaped (Fig. 6) with DNA at the larger
end, and with a protusion that identifies the direction of motion. Nanogold labeling of
MvspI protein (Fig. 6(D)) helps to identify the preponderance of this protein at the bul-
bous and head regions of the cell; such localization is beyond the resolution of simple
fluorescence imaging (Fig. 6(B)). Similarly, the protein identified as the mobility factor,
Fig. 6 (A) ASEM images of Mycoplasma mobile cells in water on fetuin-coated dishes, stained with platinum
blue, tannic acid, uranyl acetate, osmium tetroxide, and lead citrate. DNA is located in the bulbous end (arrows),
a ring-like structure can be seen at the distal end (white arrowheads) and in between are more lightly stained
features (black arrowheads), which may be the features involved in mobility. Fluorescence from DAPI (blue)
and Alexa Fluor-594 (red) are overlaid in (B) for anti-MvspI and (F) for anti-Gli349 labeling, with maps of cell
positions (C) and (G), estimated from SEM images. The label also carried nanogold which is shown (d and H)
by ASEM images after gold enhancing, and finally (E) and (I) after additional staining with tannic acid, uranyl
acetate, and lead citrate to increase the contrast of the cell outline. The different patterns of gold labeling can be
seen represented by grey shading in the maps (C) and (G). (See color plate.)
Gli349, is shown to localize preferably between these two regions (Fig. 6(H)). Heavy
metal staining served to identify the outlines of the cells (Fig. 6(E and I)) and the
nanogold labeling can be seen to exhibit different patterns identified by the outlines in
Fig. 6(C and G).
These results suggest that specific identification of Mycoplasma spp. is pos-
sible using the ClairScope to achieve super-resolution images (relative to optical
microscopy) of fully hydrated samples. The heavy metal staining protocol can be
simplified to reduce the processing time. The ability to correlate the structures and
sizes observed by ASEM with the specificity of fluorescent antibody staining, in the
same hydrated sample, should enable the instrument to be used as a high-throughput
monitoring device.
E. Detection of Cathodoluminescence from Phosphor Particles as Correlative Markers

Phosphors are materials that emit visible light upon interaction with electrons. The
color of the emission depends on the phosphor dopant, typically a rare-earth atom.
Many phosphor nanoparticles are biologically inert and can be microinjected into cells,
or can be conjugated to antibodies to create probes. In some cases, the phosphors are
also fluorescent under UV or multiphoton excitation (PhosphorDotsTM Sun Innova-
tions); such probes can thus be used as markers in correlative microscopy.
The color camera of the ClairScope is not sufficiently sensitive to detect single
nanoparticles, but it can be replaced by a high-QE (quantum efficiency) CCD for sin-
gle-color imaging (e.g., Hamamatsu Orca with a back-illuminated CCD detector, 70%
QE, or an Andor Ixon electron multiplying CCD, 90% QE). Separating the camera
software from the ClairScope also allows camera exposure while electron imaging.
We have applied nanoparticles of gadolinium oxysulfide doped with terbium
(Gd2O2S:Tb) to cells growing on the ClairScope dish window. Aggregates can be seen
in the brightfield image Fig. 7(A); when illuminated by UV light (ClairScope “DAPI”
filter), the nanocrystals are identified (Fig. 7(B)), but smaller particles are difficult to
detect in this mode due to high background. If the camera is allowed to record while
the SEM is scanning, then the cathodoluminescence emissions are detected from the
area scanned; typical images are shown in Fig. 7(C), and the background is much
lower. By scanning smaller areas and enlarging the SEM spot size to increase the
electron dose, the smallest particles can be detected (Fig. 7(D). In addition, the nano-
crystals can be imaged by the SEM, which shows that single particles of ∼50 nm length
can easily be detected in the CL image and thus distinguished from any other objects
of the same size.
The CL intensity depends on the probe current (set by the spot size), voltage, and
pixel dwell time, which is a combination of beam scan speed and magnification. Full
details of this behavior will be presented elsewhere (Morrison & O’Toole, manuscript in
preparation), but the images in Fig. 7(D and E) show that nanocrystals can be detected
both by fluorescence and in CL by electron scanning at magnifications around 5000×,
thus permitting their use as correlation markers as well as the potential to be used
directly as single probes for super-resolution imaging of individual labeled particles.
Fig. 7 Images of phosphor particles distributed on U251 cells grown on an ASEM dish, and imaged using the Hamamatsu Orca ER
camera in place of the standard QI Retiga. (A) “Brightfield” reflectance image; the edge of the dish window appears at the bottom. The
dark patches are aggregates of phosphor particles; single crystals are too small to be visible in this mode. (B) DAPI false-color fluores-
cence image, showing autofluorescence in cells, diffraction-limited punctate dots from small clusters, and fluorescent patches from larger
aggregates. (C) Cathodoluminescence image showing emission from phosphor particles under 30 kV electron excitation at magnification
1200×, in false color (color guide on the right). Note that the phosphors on the cells at right, which are out of focus in the DAPI
image, are clearly visible on the CL image due to the very large depth of focus of the electron beam. Also, phosphor particles that are
below the edge of the film window show emission, from secondary electron excitation. (D) Small area of (C), recorded with 10kV elec-
trons at magnification 5000×, to increase the electron excitation and thus CL signal. More phosphor particles are visible than in (C). (E)
SEM image of the arrowed phosphor particle, at 40,000×, shows that it is a single crystal with length less than 100 nm. (See color plate.)
V. Discussion
The ClairScope is a novel instrument, and applications of ASEM are still being
developed by experience. The principal limitation is the small window size in the dish,
which in turn leads to restricted experimental electron beam time, but this will hope-
fully be addressed by the development of a multiwindow dish. This should help to
bring the technique into use as a high-throughput imaging method for microbiological
analysis.
The great advantage of this technology is the ability to access and handle wet
samples that are being imaged by SEM. Conventional SEM preparation of biological
samples involves fairly aggressive chemical techniques, and once the material has been
electron imaged it is generally not easy to re-stain and re-examine the specimen. In the
ClairScope, it is a routine matter to stain a sample, take a quick SEM image, and then
if necessary repeat to obtain a quality image with the minimum chemical intervention.
This is in addition to the correlative ability conferred by the optical microscope.
Again, the open ASEM dish allows the user to observe the specimen by fluorescence
methods, and then to stain the sample for SEM in situ. In general, the heavy metal
stain techniques result in greatly reduced or totally quenched fluorescence; in the
ClairScope, there is no need to move the sample between the two imaging instru-
ments, so it is easier to correlate the two images. The ability to image prestaining
also enables time-lapse microscopy with fixations possible at the critical time point,
although the fixation process itself may not be instant and thus not suitable for all
applications.
The ability to examine the specimen under a variety of staining conditions is well
illustrated by the images of mycoplasma by Sato et al. (2012). Widefield fluores-
cence of such small objects can only indicate the approximate position and orienta-
tion of the cells, but the SEM images of heavy metal–stained organisms allow much
more detailed conclusions to be drawn about their ultrastructure, and the resolv-
ing power of the instrument when using nanogold as a selective probe for specific
details of structure and function, which can best be investigated while the organism
is fully hydrated. It will prove of great use in the understanding of the ultrastructure
of microorganisms.
The technique compliments many of the emerging “super-resolution” optical micros-
copy techniques (Huang, Bates, & Zhuang, 2009). As with these methods, this technique
can resolve fine detailed structures in fixed, hydrated samples. The staining protocols
have added appeal as they are widely applicable with small 20 nm gold particles replac-
ing the fluorescent tags on the antibodies or need for genetic engineering with fluores-
cent proteins. The technique has the added benefit of then being able to further stain the
sample with metal stains to gain greater structural information. Although the process
is not suited to “live” cell imaging for anything more than a snapshot due to potential
damage from prolonged exposures to electrons, the ability to undertake classic light
microscopy before committing to EM makes the potential for correlative super-resolu-
tion techniques and ASEM very appealing.
There are many possible applications of this novel instrument for research in biophys-
ics and biomaterials. Mineral growth is a field of considerable interest (Weiner, 2008), and
providing the restricted depth of electron interaction (<3 µm) is adequate, there should be
a number of research areas that can use the ClairScope given its easy-to-use atmospheric
imaging capability.
In most correlative microscopy, there is still a problem in comparing the optical
images, which are detected simultaneously at all pixels and generally with subsecond
exposures, hence suffering no distortions, with the high-resolution images from a slow-
scanning electron beam when distortion and drift can occur. The use of simultaneous
cathodoluminescence detection using a nanophosphor may assist in this process; by
use of single particle imaging techniques, the position of the phosphor particle may
be defined to better than 10 nm precision in the CL image, and about 20 nm by flu-
orescence. Thus, the phosphors are used as markers for precise correlation between
fluorescence and SEM images. The ability to compare the CL data with fluorescence
images, at different focus positions and with deconvolution, should give some three-
dimensional information about the SEM images.
The ideal correlative instrument would replace the widefield optical head by a
confocal microscope; a laser scanning system would be bulky and expensive, but a
Nipkow disc confocal could be integrated without major modifications to the ASEM
instrument. It must be hoped that a development of this kind can be achieved; com-
bining high resolution, high-sensitivity three-dimensional fluorescence imaging with
the ultrastructural imaging ability of the electron beam would be a major advantage.
Meanwhile, if higher resolution fluorescence images are needed, it is relatively simple
to acquire images on a separate confocal microscope (upright type with water-dipping
lens, or inverted microscope with long working distance air objective), before replacing
the dish on the ClairScope.
Acknowledgments
We thank Dr. Jack Franssen (UMC St Radboud, Nijmegen, Netherlands) for the
U251 glioblastoma cells, and Dr. Gareth Evans (Biology Department, University of
York, York, UK) for the B104 neuroblastoma cells, and both for helpful advice. Prof.
Jack Silver and Dr. George Fern (Wolfson Centre for Materials Processing, Brunel
University, Uxbridge, UK) provided the phosphor sample.
CLD gratefully acknowledges the award of a CASE scholarship by Jeol, UK and the
BBSRC, and IEGM thanks the Biology Department, University of York for support.
References
Creemer, J. F., Helveg, S., Hoveling, G. H., Ullmann, S., Molenbroek, A. M., Sarro, P. M., et al. (2008).
Atomic-scale electron microscopy at ambient pressure. Ultramicroscopy, 108, 993.
Danilatos, G. D. (1988). Foundations of environmental scanning electron microscopy. In P. W. Hawkes
(Ed.), Advances in electronics and electron physics (Vol. 71, pp. 109–250). San Diego: Academic Press.
Huang, B., Bates, M., & Zhuang, X. (2009). Super-resolution fluorescence microscopy. Annual Review of
Biochemistry, 78, 993–1016.
Inaga, S., Katsumoto, T., Tanaka, K., Kameie, T., Nakane, H., & Naguro, T. (2007). Platinum blue as an
alternative to uranyl acetate for staining in transmission electron microscopy. Archives of Histology and
Cytology, 70, 43.
Miesenböck, G., De Angelis, D. A., & Rothman, J. E. (1998). Visualizing secretion and synaptic transmission
with pH-sensitive green fluorescent proteins. Nature, 394(6689), 192–195.
Miyata, M. (2010). Unique centipede mechanism of mycoplasma gliding. Annual Review of Microbiology,
64, 519–537.
Morrison, I. E. G. & O’Toole, P. J. manuscript in preparation.
Murai, T. (2012). The role of lipid rafts in cancer cell adhesion and migration. International Journal of Cell
Biology. Article No.763283.
Murai, T., Maruyama, Y., Mio, K., Nishiyama, H., Suga, M., & Sato, C. (2010). Low cholesterol triggers
membrane microdomain-dependent CD44 shedding and suppresses tumour cell migration. Journal of
Biological Chemistry, 286, 1999–2007.
Nishiyama, H., Suga, M., Ogura, T., Maruyama, Y., Koizumi, M., Mio, K., et al. (2010). Atmospheric scanning
electron microscope observes cells and tissues in open medium through silicon nitride film. Journal of
Structural Biology, 169, 438–449.
PhosphorDotsTM Sun Innovations, Fremont CA, USA 94539. http://www.nanomaterialstore.com/nano-

phosphor.php
Razin, S., Yogev, D., & Naot, Y. (1998). Molecular biology and pathogenicity of mycoplasmas. Microbiology
and Molecular Biology Reviews, 62, 1094–1156.
Sako, Y. (2006). Imaging single molecules in living cells for systems biology. Molecular Systems Biology,
2, 56.
Sato, C., Manaka, S., Nakane, D., Nishiyama, H., Suga, M., Nishizaka, T., et al. (2012). Rapid imaging of
mycoplasma in solution using atmospheric scanning electron microscopy (ASEM). Biochemical and
Biophysical Research Communications, 417, 1213–1218.
Simons, K., & Ikonen, E. (1997). Functional rafts in cell membranes. Nature, 387, 569–572.
Suga, M., Nishiyama, H., Konyuba, Y., Iwamatsu, S., Watanabe, Y., Yoshiura, C., et al. (2011). The atmo-
spheric scanning electron microscope with open sample space observes dynamic phenomena in liquid or
gas. Ultramicroscopy, 111, 1650–1658.
Thiberge, S., Nechustan, A., Sprinzak, D., Gileadi, O., Behar, V., Zik, O., et al. (2004). Scanning electron
microscopy of cells and tissues under fully hydrated conditions. Proceedings of the National Academy of
Sciences of the United States of America, 101(10), 3346–3351.
Thiberge, S., Zik, O., & Moses, E. (2004). An apparatus for imaging liquids, cells, and other wet samples in
the scanning electron microscope. Review of Scientific Instruments, 75, 2280.
Vancha, A. R., Govindaraju, S., Parsa, K. V. L., Jasti, M., González-García, M., & Ballestero, R. P. (2004).
Use of polyethyleneimine polymer in cell culture as attachment factor and lipofection enhancer. BMC
Biotechnology, 4, 23–34.
Weiner, S. (2008). Biomineralization: a structural perspective. Journal of Structural Biology, 163, 229–234.
Yamazoe, H., Uemura, T., & Toshizumi, T. (2008). Facile cell patterning on an albumin-coated surface.
Langmuir, 24, 8402–8404.
CHAPTER 17
Bridging Microscopes: 3D Correlative

Light and Scanning Electron Microscopy
of Complex Biological Structures
Miriam S. Lucas, Maja Günthert, Philippe Gasser, Falk Lucas
and Roger Wepf
Electron Microscopy ETH Zurich – EMEZ, ETH Zurich, Switzerland
Abstract
I. Introduction
A. Why High-Pressure Freezing?
B. 3D Correlative Light and Electron Microscopy
C. Sample Preparation for 3D CLEM
II. Rationale
III. Methods
B. Freeze-Substitution for CLEM
C. 3D Correlation of En-bloc CLSM and FIB-SEM
D. Serial-Section SEM and Correlative Array Tomography
IV. Materials and Instrumentation
D. Serial-Section SEM and Array Tomography
V. Discussion
A. Fluorescence Staining During Freeze-Substitution is a Comprehensive Tool
for CLEM
B. 3D CLEM: FIB-SEM or Correlative Array Tomography?
VI. Summary
Acknowledgments
References

326 Miriam S. Lucas et al.
Abstract
The rationale of correlative light and electron microscopy (CLEM) is to collect data on
different information levels––ideally from an identical area on the same sample––with the
aim of combining datasets at different levels of resolution to achieve a more holistic view
of the hierarchical structural organization of cells and tissues. Modern three-dimensional
(3D) imaging techniques in light and electron microscopy opened up new possibilities to
expand morphological studies into the third dimension at the nanometer scale and over
various volume dimensions. Here, we present two alternative approaches to correlate 3D
light microscopy (LM) data with scanning electron microscopy (SEM) volume data. An
adapted sample preparation method based on high-pressure freezing for structure preser-
vation, followed by freeze-substitution for multimodal en-bloc imaging or serial-section
imaging is described. The advantages and potential applications are exemplarily shown
on various biological samples, such as cells, individual organisms, human tissue, as well
as plant tissue. The two CLEM approaches presented here are per se not mutually exclu-
sive, but have their distinct advantages. Confocal laser scanning microscopy (CLSM) and
focused ion beam-SEM (FIB-SEM) is most suitable for targeted 3D correlation of small
volumes, whereas serial-section LM and SEM imaging has its strength in large-area or
-volume screening and correlation. The second method can be combined with immuno-
cytochemical methods. Both methods, however, have the potential to extract statistically
relevant data of structural details for systems biology.
I. Introduction
Correlative light and electron microscopy (CLEM) has become a powerful tool in life
science (Jahn et al., 2012). The rationale is to collect data on different information levels
from an identical area of one sample and combine these to achieve a holistic understand-
ing of the ultrastructure of living systems. Three-dimensional (3D) imaging techniques
in light and electron microscopy (EM) have opened up new possibilities to expand mor-
phological context description into the third dimension at nanometer scale. This chapter
focuses on two methods for correlative 3D imaging, combining confocal laser scanning
microscopy (CLSM) with focused ion beam-scanning electron microscopy (FIB-SEM),
or wide-field light microscopy (LM) and SEM using serial sections of embedded samples.
The two approaches are schematically outlined in Fig. 1. A specially adapted sample prep-
aration method based on high-pressure freezing (HPF) for optimum ultrastructural preser-
vation is presented as connecting link between the afore-mentioned imaging techniques.
A. Why High-Pressure Freezing?

Since its introduction in the 1960s (Moor, 1973; Riehle, 1968; Riehle & Hoechli, 1973),
high-pressure freezing has proven in numerous applications and among various kinds of
samples from the kingdoms of life (Dahl & Staehelin, 1989; Hess, 2007; Hohenberg, Man-
nweiler, & Muller, 1994; Hohenberg, Tobler, & Muller, 1996; McDonald & Morphew,
17. 3D CLEM on Biological Structures 327
Fig. 1 Workflow scheme for 3D CLEM: Specimens are high-pressure frozen immediately after sampling, followed by freeze-substitu-
tion and embedding in a resin. Such resin blocks can be directly imaged in a CLSM. Subsequently, the same ROI can be investigated by
FIB-SEM. If needed, ultrathin lamellae could be extracted for TEM tomography work afterwards. Alternatively, ultrathin serial sections
can be collected on ITO-coated coverglass supports and investigated by LM in brightfield, darkfield, or fluorescence mode (BF/DF,
FLM). Once imaged by LM, the corresponding areas can be further investigated in a low voltage SEM (LVSEM). This latter approach
is called “correlative array tomography” (CAT).
1993; Müller, 1992; Studer, Hennecke, & Muller, 1992; Studer, Michel, & Muller, 1989;
Studer, Michel, Wohlwend, Hunziker, & Buschmann, 1995; Szczesny, Walther, & Muller,,
1996) to preserve biological tissue better and closer to its lifelike state than any other sample
preparation method for EM (Murk et al., 2003; Schwarz & Humbel, 2007; Tiedemann,
Hohenberg, & Kollmann, 1998). The pinpoint precision of preserving the functional locus
of biomolecules has been demonstrated for human skin and other tissues (Kirschning, Rut-
ter, & Hohenberg, 1998; Pfeiffer et al., 2000; Ripper, Schwarz, & Stierhof, 2008; Schwarz,
Hohenberg, & Humbel, 1993; Vielhaber, Brade, et al., 2001; Vielhaber, Pfeiffer, et al., 2001)
and is summarized in Fig. 2. Human skin biopsies were prepared according to different fixa-
tion protocols and immediately immobilized by HPF to preserve the ultrastructural state of
the tissue. These frozen samples were then freeze-substituted and embedded in parallel and
with identical preparation steps as reported in Pfeiffer et al. (2000) and Biel, Kawaschinski,
Wittern, Hintze, and Wepf (2003), and compared to the method of progressive lowering of
temperature, or PLT embedding (Robertson, Monaghan, Clarke, & Atherton, 1992).
One sample (Fig. 2(A)) was chemically fixed with Karnovsky’s fixative (1.25%
glutaraldehyde and 0.2% formaldehyde in cacodylate buffer) prior to HPF.
Fig. 2 Localization and distribution of glucosylceramide on semithin sections (200 nm) of human epidermis in
dependence of sample fixation and dehydration applied to preserve human skin biopsies for imaging. (A) Prior
to HPF, this skin sample was chemically fixed in Karnovsky’s fixative. After freezing the sample was dehydrated
using a standard freeze-substitution protocol for skin. (B) This sample was chemically fixed with Karnovsky’s
fixative, then dehydrated and emdedded using the PLT method (C) A fresh skin biopsy immediately high-pressure
frozen after withdrawal from the donor was freeze-substituted with the same protocol as in sample (A). All
samples have been finally embedded in HM20 and polymerized at −50°C. Glucosylceramide can be found all
over the epidermis as well as in the dermis in (A) and (B), which does not match to the natural appearance as
vesicles (lamellar bodies) in the upper stratum spinosum (Ssp), stratum granulosum (SG), and the lower part of
the stratum corneum (SC) as clearly shown in Fig. 1c after HPF. Glucosylceramide should also not be found in
the upper SC, since these lipids are deglucosylated by enzyme activities in the SC. The redistribution of this small
biomolecule into nonfunctional areas can only be explained as an artifact of chemical fixation, which can even be
enhanced by PLT dehydration. Scale bar: 20 µm.
araformaldehyde or formaldehyde is considered a faster but weaker penetrating fixa-

P
tive compared to glutaraldehyde (Elias & Friend, 1975; Karnovsky, 1965). The second
sample (Fig. 2(B)) was chemically fixed as mentioned above at 4°C, but was then
dehydrated and embedded in HM20 using the PLT method. The third sample (Fig.
2(C)) was high-pressure frozen immediately (i.e., less than 30 sec) after extraction
of the punch biopsy and freeze-substituted as reported previously (Biel et al., 2003;
Pfeiffer et al., 2000). Ultrathin sections of the embedded tissue were labeled against
glucosylated precursor lipids (mouse Glucosylceramide antiserum; Brade, Vielhaber,
Heinz, & Brade, 2000) which are very specific to the last viable cell layers of the epi-
dermis and finally are exported into the intercellular space, where they are deglucosyl-
ated by beta-glucocerebrosidase (Holleran et al., 1994) and not recognized anymore by
the mouse-anti glucosylceramide antibody. Therefore, the localization of this antigen
is functionally defined in lamellar bodies, which are transported from the trans-Golgi
network in the upper stratum spinosum and stratum granulosum toward the skin sur-
face. The antigens are then secreted into the intercellular space at the “life–death tran-
sition interface” between stratum granulosum and the first dead layers of the stratum
corneum.
Fig 2 shows clearly that this lipid precursor is delocalized and found in different skin
layers even down to the dermis (Fig. 2 (A and B)) after chemical fixation, independent
of the follow-up dehydration protocol. The structure–function related localization of

such small biomolecules (<1000 MW) is only preserved after HPF and gentle, optimized
freeze-substitution (FS) as reported here. Our methods take full advantage of localization
techniques, such as immunolocalization for LM and EM or even super-resolution imaging
in the near future, and thereby enable bridging between these imaging techniques.
B. 3D Correlative Light and Electron Microscopy

Morphological investigations in life science rely on the description of lipid mem-
branes delineating intracellular compartments, i.e. organelles or vesicles, as well as cell
borders and intercellular connections. These membranes are only 4.5–5 nm thick and
thus invisible to LM, but resolvable by EM. FIB-SEM has proven to be a valuable
application for ultrastructural investigations in life science (Young, Dingle, Robinson,
& Pugh, 1993). It offers the combination of precise and site-specific milling of mate-
rial from a bulk sample, using a focused ion beam, with high-resolution SEM imaging.
Using the FIB as a serial-sectioning tool to produce stacks of consecutive image planes
allows automated recording of large volumes (Knott, Marchman,Wall, & Lich, 2008;
Lucas et al., 2008). Routine overnight FIB-SEM cross-section imaging can extend up to
50 × 40 µm2. With a slice thickness of 5–10 nm, these stacks can expand up to several
10 µm in the x-axis. This makes FIB-SEM a valid alternative to standard serial-section
transmission electron microscopy (TEM) for targeted volumes, as the preparative effort
is reduced and a better z-resolution compared to serial sections can be achieved.
Modern field emission SEMs allow even at low acceleration voltages of 1 kVto resolve struc-
tural details of contrasted or unstained biological samples in the range of 1–2 nm, using either-
back-scattered electron (BSE) or secondary electron (SE) imaging. Therefore, on-section
SEM imaging has the lateral resolution power to image lipid membranes in cells and tis-
sues. Necessary adaptations for sample preparation and imaging parameter are reported
in this chapter. Hence, SEM imaging of either ultrathin sections or block-faces may be an
alternative approach for ultrastructure investigations, which are to date dominated by TEM.
C. Sample Preparation for 3D CLEM

Dyes and specific labels for LM often are incompatible or only partially compatible
with heavy metal salts used for EM. Therefore, specimen preparation for 3D CLEM
requires a compromise, which ideally results in a sample suitable for all light and elec-
tron microscopic approaches described above. En-bloc imaging in CLSM and FIB-
SEM poses some challenges: (1) the region of interest (ROI) has to be accessible for
imaging as well as for FIB-milling, i.e. near the surface, (2) the material should facili-
tate stable and artifact-free milling conditions in FIB-SEM, and most important (3) flu-
orescence imaging should be enabled, while optimum signal-to-noise ratio is provided
for BSE imaging. Resin embedding meets these requirements inasmuch as first, the
ROI can easily be uncovered by trimming or ultramicrotomy as is routinely done for
TEM samples; and second, the material is sufficiently homogenous in terms of hard-
ness and density to enable artifact-free FIB-milling (Knott et al., 2008). Fluorescence
labeling during FS followed by embedding in HM20 proved a powerful tool for cor-
relative microscopy (Biel et al., 2003). It enables en-bloc CLSM investigation of the
sample prior to preparation of ultrathin sections for TEM.
Obtaining a good signal-to-noise ratio in FIB-SEM, especially with respect to the
visibility of lipid membranes, is more challenging, as the samples need to be inves-
tigated en-bloc and conventional post-staining of ultrathin sections is not possible.
Several protocols have been described, aiming to optimize preembedding membrane
staining after chemical fixation, based on potassium ferrocyanide-mediated oxidation
of osmium tetroxide, followed by uranyl acetate (De Winter et al., 2009; Knott et al.,
2008), or tannic acid (Armer et al., 2009).
For ultrastructural investigations of very fast cellular processes, e.g. virus entry into
host cells, HPF is preferable to chemical fixation, as it is faster and avoids fixative-induced
artifacts such as shrinkage or membrane perturbation (Droste et al., 2005; Griffiths, 1993;
Murk et al., 2003). Enhancing membrane staining during FS after HPF for TEM applica-
tions has been discussed intensely. Pfeiffer et al. (2000) and Humbel and Schwarz (1989)
showed that FS in acetone with uranyl acetate preserves the membrane details and lipids
well. FS in acetone with uranyl acetate and glutaraldehyde was also described as the
method of choice by Giddings (2003), when compared to osmium tetroxide and tannic
acid in acetone. Hess (2003) on the other hand, concluded that osmium tetroxide in ace-
tone, with optional addition of uranyl acetate, was the best choice for morphological inves-
tigations. Tannic acid-mediated osmium impregnation at room temperature following FS
in osmium in acetone provided good membrane contrast even in semithin sections used
for TEM tomography, and rendered post-staining of the sections unnecessary (Jimenez
et al., 2009). Furthermore, addition of water to the FS-medium has been described to
enhance membrane visibility (Buser & Walther, 2008; Walther & Ziegler, 2002), which
also significantly improved the mordant-mediated osmium staining described by Jimenez
et al. (2009). Nevertheless, most of these methods either involve additional post-staining
of ultrathin sections or are incompatible with fluorescence labeling.
Array tomography, on the other hand, has different requirements in terms of speci-
men preparation. As this method employs serial sections of resin-embedded material,
labels do not necessarily have to be introduced before embedding. This leaves a margin
for adaptations according to the requirements of the respective experiment, and opens
up more possibilities for specific labeling either by histological dyes, fluorescent dyes
(e.g., DAPI), or by immunocytochemical techniques to identify an ROI. If intense appli-
cation of heavy metal salts interferes with fluorescence labeling, the sections can be
post-stained at a later stage for SEM investigation. The protocol described in this chap-
ter provides a compromise, which allows characterization of the embedded specimen by
CLSM prior to either FIB-SEM investigation or serial sectioning for array tomography.
II. Rationale
The purpose of this chapter is to present and discuss volume imaging techniques for
correlative light and scanning electron microscopy, including a protocol for FS adapted
for multimodal en-bloc imaging. Fluorescent labeling during FS, prior to embedding,
provides histological context information of the ROI and can be employed to relo-
cate the ROI in SEM and FIB-SEM. As an alternative to the en-bloc CLEM approach,
serial-section SEM and array tomography are described, and assets and drawbacks of
both methods are discussed.
III. Methods
The described protocol for FS and the different CLEM approaches are applicable to a
wide variety of samples, such as cultured cells, small organisms, tissue samples, or plant
specimens. To demonstrate this, we used monolayers of MDCK and HeLa cells, nema-
todes (C. elegans), biopsies of human skin, and mung bean (Vigna radiata) root nod-
ules. The specimen preparation for these samples is described in the following sections.
Pretreatment or culture conditions are omitted. These may vary significantly according
to sample type and are not relevant for the description of the general concept of the
described methods. The reader is referred to previous publications (Mercer & Helenius,
2008; Pfeiffer, Vielhaber, Pflucker, Wepf, & Hintze, 1997; Rothen-Rutishauser, Kramer,
Braun, Gunthert, & Wunderli-Allenspach, 1998; Studer et al., 1992).
Practical aspects of HPF, including some “tips and tricks” have been described in
detail previously (McDonald et al., 2010; Monaghan, Cook, Hawes, Simpson, Tomley,
2003; Reipert, Fischer, & Wiche, 2004; Studer et al., 1992; Tiedemann et al., 1998).
Therefore, we will focus on specific adaptations for 3D CLEM.
Cells were cultivated directly on sapphire discs to facilitate in situ fixation by
HPF (Eppenberger-Eberhardt et al., 1991; Hawes, Netherton, Mueller, Wileman, &
Monaghan, 2007; Schwarb, 1990). Before adding the sapphire discs to cell culture
dishes, they were washed in sulfuric acid, followed by three washes each in tap water,
distilled water, and ethanol. Subsequently, the sapphire discs were evaporation coated
with a layer of carbon (15–18 nm). Scratching an asymmetrical character into this car-
bon layer will help to recognize the side with the attached cells. Cells were maintained
at culture conditions until HPF.
To facilitate handling during FS without losing the nematodes, worms were loaded
into cellulose capillaries with an inner diameter of 200 µm as described by Hohenberg
et al. (1994). The capillaries were attached to gel-loading pipette tips, and the worms
were then picked from the agar and loaded directly into the cellulose tubes. Filled capil-
laries were cut into 2 mm pieces and sealed by crimping the ends.
Mung bean root nodules were abscised and cut into 200 μm, thick slices using a
hand microtome. These slices were degassed under prevacuum in 1-hexadecene until
“bubbling” ceased (Studer et al., 1989). Biopsies of human skin were taken as fresh as
possible before freezing using standard 2 mm biopsy punches. The subcutis was care-
fully removed prior to freezing.
Tissue samples and cellulose capillaries were placed in standard aluminum platelets
for HPF, filled with incompressible 1-hexadecene to improve energy transfer and pro-
tect the specimen. For cell culture samples, two sapphire coverslips were sandwiched
using a 100 µm aluminum spacer ring as a separator. The cavity was filled with stan-
dard cell culture medium. After freezing, the specimens were stored in liquid nitrogen
until further processing.

The following protocol yielded excellent results for all tested samples. However,
concentrations and times for sample processing may have to be adapted for different
specimens. The list of tested fluorophores is not intended to be exhaustive, and further
investigation is encouraged.
1. Freeze-Substitution and Embedding Procedure

The HPF sandwiches were opened under liquid nitrogen and transferred into the FS
medium, precooled to −90°C. FS baskets made of pipette tips were used to facilitate
transfer and subsequent medium exchange (Hohenberg et al., 1994). We used anhy-
drous acetone supplemented with either 2% uranyl acetate or 1% osmium tetroxide,
or with a combination of both. Fluorescent dyes were added to the medium before
cooling. All dyes were applied in such concentrations that the solutions were saturated,
forming a sediment in the cooled solutions (−90°C). FS was started at −90°C and kept
for 8 h for cell culture samples, or 24 h for larger specimens, such as skin biopsies, root
nodules, or nematodes. Subsequent steps were adjusted according to the embedding
protocol for the respective resin. For embedding in Epon, the samples were held at −60°C,
−30°C, and 0°C for 8 h each, followed by washing in anhydrous acetone and room
temperature embedding in Epon. For embedding in Lowicryl HM20, FS was continued
for 8 h each at −70°C and −50°C, before washing in anhydrous acetone. Resin infiltra-
tion with a graded series of HM20 in acetone was performed at −50°C (1 h each step
for cell culture samples, 2–4 h for the larger specimens). UV polymerization in fresh
HM20 was performed at −35°C.
After polymerization, samples were trimmed with either a razor blade or a dia-
mond knife to uncover the specimens for en-bloc imaging or serial sectioning.
Sapphire coverslips were removed from the cell culture samples by briefly plung-
ing into liquid nitrogen. The resulting block-face thus represented the plane of the
cell substrate with the cells embedded upside down just beneath the surface in the
resin block.
2. Fluorescent Dyes for Preembedding Labeling

Several fluorescent dyes were tested regarding the suitability for labeling of differ-
ent sample types, the quality of labeling with respect to the application of heavy metal
stains for EM, and the possibility of embedding in different types of resins. Samples
included cultured cells, tissue samples, plant material, and nematodes. A full list of
tested dyes and specimens is given in Table I.
Table I
Fluorescent dyes suitable for application in FS, and labeling results for resin-embedded specimens.
Dye Tested with Highlighting Image quality
1.8 ANS Skin (human) Collagen and/or elastic fibers +

Acridine orange Skin (human) Nuclei, cytoplasm, stratum corneum ++
Liver (mouse) Nuclei, cytoplasm +
Lymph node (mouse) Nuclei, cytoplasm ++
HeLa cells Cytoplasm, nuclei +
Mung bean roots Bacteria population of root nodules ++
Bodipy 560 Skin (human) Nuclei, cell membranes, stratum corneum, whole +
dermis
Dapi MDCK cells Entire cell, unspecific -
DCVJ Skin (human) Nuclei, collagen, and/or elastic fibers +
DiD Skin (human) Lipophilic domains, cell membranes +
Liver (mouse) Lipophilic domains ++
HeLa cells Cell membranes ++
MDCK cells Lipophilic domains, cell membranes (also ++
intracellular)
DiIC18 Skin (human) Lipophilic domains, cell membranes ++
Liver (mouse) Lipophilic domains, cell membranes ++
HeLa cells Cell membranes ++
MDCK cells Cell membranes +
Nematodes (C. elegans) Lipophilic domains -
DiOC6 Skin Cytoplasm (lipophilic domains), stratum corneum +
Nile blue sulfate Skin (human) Nuclei, cytoplasm, stratum corneum, collagen, ++
elastic fibers
Liver (mouse) Nuclei, cytoplasm +
HeLa cells Nuclei, cytoplasm +
MDCK cells Nuclei, cytoskeleton ++
Nematodes (C. elegans) Cytoplasm, membranes +
Nile red Skin (human) Nuclei, cytoplasm, stratum corneum -
Liver (mouse) No labeling --
HeLa cells Nuclei, cytoplasm +
MDCK cells Nuclei, cytoplasm ++
Oregon Green Skin (human) Nuclei, collagen, elastic fibers +
PhalloidinAlexa 488 HeLa cells Cytoskeleton ++
Safranin O Skin (human) Nuclei, cytoplasm, stratum corneum, collagen, ++
elastic fibers
Liver (mouse) Cytoplasm -
HeLa cells Nuclei, cytoplasm ++
Nematodes (C. elegans) Nuclei, cytoplasm, cell membranes ++
Mung bean roots Entire tissue, unspecific +
Sudan III Skin (human) Nuclei, whole dermis +
Syto 16 MDCK cells No labeling --
Syto 24 MDCK cells Entire cell, unspecific -
Syto 83 MDCK cells Entire cell, unspecific +
Sytox Green MDCK cells Entire cell including membranes, unspecific +
Tannin Skin (human) Nuclei, stratum corneum -
++, good; +, fair;-, poor; --, no labeling.

Osmium tetroxide is described to interfere with fluorescence. To evaluate the effect

of osmium tetroxide, we compared the fluorescence of samples freeze-substituted with
uranyl acetate alone, a combination of uranyl acetate and osmium tetroxide, or osmium
tetroxide alone. The tested dyes showed the best signal-to-noise ratio and most detailed
structures after FS with uranyl acetate alone. The presence of osmium tetroxide in
the FS-medium yielded rather weak and diffuse labeling. This effect was more pro-
nounced in the range of green and yellow fluorescence, while far red fluorescence was
not impaired (Fig. 3(A-C)). Additionally, we compared two standard types of embed-
ding resins: HM20 and Epon. Epon showed a slight autofluorescence in the range of
500–570 nm, interfering with detection of fluorescence in this wavelength and thus
making detection of faint signals difficult. Fluorescence in the far red was not impeded
(Fig. 3(D)). In contrast, HM20 did not show any notable autofluorescence.
We found pronounced differences in the labeling characteristics depending on the
sample type. For example, DiIC18 showed detailed labeling of membranes in HeLa and
MDCK cells. In contrast, Nile blue sulfate labeled cytoplasm and nuclei in HeLa cells,
but was limited to the nuclei in MDCK cells (Fig. 3(D)). The same dyes yielded strong
fluorescence signals even with the addition of osmium tetroxide when applied to skin
tissue (Fig. 3(G and H)), but rather weak fluorescence in nematodes (Fig. 3(K and L)).
However, signal quality was sufficient for a histological description of the sample or
recognition of the ROI. Other stains such as Safranin O or Nile red produced rather
unspecific staining results for both MDCK and HeLa cells (Fig. 3(E and F)), while
Safranin O gave a very detailed description of nematodes (Fig. 3(M)).

FIB-SEM provides a valid alternative to serial-section TEM (Knott et al., 2008;
Lucas et al., 2008). State-of-the-art SEM platforms allow imaging with an x/y-resolution
Fig. 3 Preembedding fluorescence staining during FS. All samples were high-pressure frozen, and freeze-
substituted in acetone containing either uranyl acetate alone or a combination of uranyl acetate and osmium
tetroxide, as well as fluorescent dyes. Large images of cell cultures represent samples labeled with uranyl ace-
tate only, insets show the respective fluorescence when osmium tetroxide was applied additionally. Human skin
and nematodes were freeze-substituted with uranyl acetate and osmium tetroxide. All samples were embedded
in HM20, unless otherwise stated. (a–c) HeLa cells after infection with Vaccinia virus stained with DiIC18 to
label cell membranes (A), and Nile blue for staining of the nucleus and cytosol (B), (C) merge of both signals.
Only the fluorescence of DiIC18 is slightly impaired by the application of osmium tetroxide (insets), while the
Nile blue signal is not affected. (D) MDCK cells stained with Nile blue and embedded in Epon. Note the dif-
ference in labeling, with MDCK cells the Nile blue signal is restricted to the nucleus, while it stains the nucleus
and cytoplasm uniformly in HeLa cells. The fluorescence signal was not changed by addition of osmium
tetroxide (inset). (E) HeLa cells (infected with Vaccinia virus) labeled with Safranin O or Nile red (F). Both
dyes stain the entire cell rather unspecifically. (G–J) Biopsy of human skin. DiIC18 labels cell membranes and
stratum corneum (G), while Nile blue highlights the cytoplasm, collagen, and elastic fibers (H). The merged
signals allow a detailed histological description of the sample. (K) DiIC18 staining in C. elegans is rather weak
and can only be roughly attributed to lipophilic domains, while Nile blue labels the cytoplasm and membranes
equally strong (l, ROI corresponding to k). (M) Safranin O provides a very detailed description, highlighting
the nuclei and membranes, while cytoplasmic labeling is weaker. Scale bar: 25 µm. (See color plate.)
of 1–2 nm, which is sufficient for ultrastructure research and subcellular structural biol-
ogy projects. An advantage of FIB-SEM is the automation: volumes of up to 50 × 30 ×
30 µm3 can be recorded overnight, with a voxel size of 5–10 nm, which is not achiev-
able both in time and z-resolution by standard serial sectioning. Combining FIB-SEM
with CLSM preinvestigation, this en-bloc CLEM technique allows a comprehensive
study of a sample at the light and electron microscopic level in 3D.
1. Block-Face Imaging with CLSM

One prerequisite for en-bloc imaging of resin-embedded samples by CLSM is a
clean, smooth, and even sample surface. Pronounced scratches, rough uneven surfaces,
or indentations, as well as a sloping surface impede with imaging quality and should be
avoided. Excess resin covering the sample or ROI also needs to be removed. Therefore,
the samples are trimmed in the same way as for TEM sectioning. For the creation of
the imaging surface, a diamond-trimming knife is highly recommended, as this will
yield best results. We have designed a simple CSLM-sample holder to accommodate
the specimen holder of the ultramicrotome. This enables easy switching back and forth
between the CLSM and the ultramicrotome without unclamping the resin block, if the
ROI is not yet uncovered and further trimming is required.
The signal penetration depth for efficient signal detection from resin blocks depends
on the sample type and embedding resin. For HM20-embedded samples, the penetra-
tion depth of 50 µm can easily be achieved. However, the closer to the surface the
better is the signal detection efficiency. In most cases, the fluorescence from the embed-
ded samples is too weak to be detected with standard air objective lenses. Immersion
media, such as water, oil, or glycerine, may be used to overcome this. When working
with immersion oil or glycerine, the specimen surface should be cleaned thoroughly
immediately after imaging, to avoid transferring traces of these media either to the
diamond knife (in case additional trimming is required), or into the FIB-SEM. Further-
more, immersion oil may affect the resin.
2. Focused Ion Beam-Scanning Electron Microscopy

For FIB-SEM investigation, the resin blocks were cut to approximately 5 mm in height
to fit the movement range of the FIB-SEM stage, mounted on SEM specimen stubs using
conductive silver paste, and sputter coated with a 5–7 nm layer of gold. In the FIB-SEM,
an additional carbon protection layer was deposited on the ROI to protect the specimen
surface from ion beam damage. To obtain an initial imaging field, a trench was milled
perpendicular to the sample surface using 30 kV acceleration voltage and a current of
6.5 nA, and the milled cross-section was finally polished using 30 kV and 1.5 nA. Before
starting image acquisition, a settling time of at least one hour after inserting the sample
into the FIB-SEM system was allowed, to prevent stage drift and resulting focus changes.
SEM images of cross-sections were recorded at 2 kV, using an EsB detector (EsB grid
1250 V). Three-dimensional volumes were automatically acquired by sequential milling
(30 kV, 300 pA) and SEM imaging. Volumes were recorded with isotropic voxels, i.e. the
nominal x/y-resolution for imaging was chosen to match the average slice thickness (=
z-resolution). This enables detailed inspection and analyses of the recorded volume from
any direction, using arbitrary imaging planes. The actual voxel size was chosen for each
sample according to the respective aim of the study, usually ranging from 5 to 20 nm.
These milling and imaging conditions may be used as guidelines, but it is strongly
recommended to optimize the parameters for different sample and resin types, as well
as other FIB-SEM systems. Especially, the imaging conditions may alter with different
sample preparations.
3. Optimized Membrane Staining for 3D CLEM

The optimum FS protocol for 3D CLEM involves strong membrane staining for
FIB-SEM, while not impairing fluorescence detection. The best results for mem-
brane visualization have been achieved with osmium tetroxide at room temperatures
(Jimenez et al., 2009; Knott et al., 2008). However, osmium quenches fluorescence by
competitively binding to carbon double bonds. Aditionally, when applied above −30°C
renders the samples dark brown or black, which significantly reduces the fluorescence
yield. To identify the optimum compromise between heavy metal staining and main-
taining fluorescence, we used cell culture samples and compared the staining results
obtained after the use of either uranyl acetate, osmium tetroxide, or both reagents in the
FS medium. Additionally, we tested two of the most commonly used embedding resins,
Epon and Lowicryl HM20, with respect to signal-to-noise ratio and milling character-
istics in FIB-SEM. To assess the quality of membrane staining in HPF-FS samples, our
methods were compared to potassium ferrocyanide-enhanced osmium staining after
chemical fixation (Knott et al., 2008). The results are summarized in Fig. 4.
Membrane visibility was best in samples prepared with the benchmark protocol
(Fig. 4(C and F)). But this preparation is not compatible with fluorescence staining due
to osmium-mediated fluorescence quenching and blackening of the sample. To mini-
mize or prevent this effect, FS media containing osmium tetroxide were exchanged
for anhydrous acetone at 0°C for embedding in Epon, or already at −50°C for HM20
embedding. The use of uranyl acetate alone resulted in a reasonable staining of the
cytoplasm, the nucleus, and the membranes of some larger vesicles, but the outer cell
membrane and nuclear envelope could not be detected (Fig. 4(A and D)). When only
osmium tetroxide was applied, the samples showed an extremely weak imaging con-
trast, which was not suitable for FIB-SEM imaging. However, the combination of ura-
nyl acetate and osmium tetroxide produced a detailed staining of the cytoplasm, the
nucleus, and the membranes (Fig. 4(B and E)). The comparison of Epon and HM20
showed no detectable difference with respect to milling properties. Concerning image
quality, HM20 is preferable to Epon due to a better signal-to-noise ratio.
4. Retrieving the Region of Interest

Having defined the ROI in CLSM, relocating it in the FIB-SEM can present a chal-
lenge. While CLSM records fluorescence signals from beneath the surface of the resin
block, SEM is primarily a technique for surface imaging, and signal detection from
within a bulk sample is limited. Increasing the acceleration voltage yields a larger inter-
action volume of the electrons with the sample, and thus allows limited visualization of
subsurface structures. However, the penetration depth is limited to a few micrometers
Fig. 4 Optimization of membrane staining for FIB-SEM. Cells are embedded upside down. After resin curing
the sapphire substrate is removed, leaving the cells directly beneath the surface of the resin block. For FIB-SEM,
the resin block is sputter coated with 5–7 nm gold (indicated by blue, dashed line) for conductivity. To protect
the sample and prevent curtaining effects during milling, the ROI is additionally coated with carbon (asterisk).
The images were acquired with identical settings for brightness and contrast to enable direct comparison of
staining quality. (A, B) MDCK cells grown on sapphire discs were freeze-substituted after HPF in acetone and
embedded in HM20. (A) Imaging contrast achieved with uranyl acetate and (B) a combination of uranyl acetate
and osmium tetroxide in the FS medium. Membrane visibility is improved by addition of osmium tetroxide.
(C) Benchmark preparation according to a protocol for chemical fixation optimized for membrane detection,
using uranyl acetate, pure and reduced osmium tetroxide followed by embedding in Epon (Knott et al., 2008).
The membranes show strong staining while cytoplasm and other details of the cell are less pronounced. (D–F)
Magnified display corresponding to areas marked by dotted lines in images (A–C). Scale bar: 1 µm.
maximum, and more importantly, extended high-voltage imaging will damage the resin
surface. Therefore, this method is only recommended as a tool for fine-tuning rather
than large-scale navigation while searching for the ROI.
Reference points are inevitably necessary to be able to navigate to the ROI. These
reference points may be random surface features, markings specifically scratched into
the surface, or structural features in the sample itself. State-of-the-art CLSMs allow the
recording of a reflection signal, which can be used to display the surface topography
of the block-face.
For cell culture monolayers grown on carbon-coated sapphire discs and processed
in situ, scratches in the carbon layer provide a starting point. The carbon layer per-
sists on the surface of the polymerized resin block after removal of the sapphire disc.
The character scratched into the carbon to help identify the face holding the cells
remains visible, and provides simple reference points for relocalization (Fig. 5).
After identifying the ROI using the fluorescence of the cells (Fig. 5(A)), a reflection
image of the same region is recorded (Fig. 5(B)). The scratch is easily recognizable in
both CLSM and FIB-SEM (Fig. 5(C)), and thus facilitates the retrieval of the ROI for
FIB-SEM milling (Fig. 5(D)).
Larger tissue samples or small specimens enclosed in cellulose capillaries provide
navigational aid via structural features exposed on the block-face by trimming. Dis-
tinctive features, such as the outlines of individual cells or the cellulose capillaries, are
clearly identifiable in the CLSM image as well as in the top view of the block-face in
the FIB-SEM (Fig. 5–7).
Fig. 5 CLEM for cell culture applications. Monolayers of cells are grown on a carbon-coated sapphire disc,
high-pressure frozen and freeze-substituted in the presence of uranyl acetate and Safranin O, then embedded in
HM20. Investigation of the resin-embedded sample by CLSM facilitates selection and 3D characterization of an
ROI (A). Surface features of the resin block can be recognized in the reflection mode of the CLSM (B) as well as
in SEM (C) and can be used to navigate to the ROI for FIB-SEM investigation (D). (E) Gallery of single images
from an FIB-SEM volume of a HeLa cell infected with Vaccinia virus. (F) 3D model of a 4 µm thick wedge of
the cell, showing the cell surface with protruding filopodia and virus particles (red). The virus infection induces
large membrane blebs (asterisk). (G) Virus particles are observed in association with endocytic invaginations
next to retracting blebs (arrowheads). Virions seem to enter the host cell via macropinocytic uptake as part of the
bleb retraction process (Mercer & Helenius, 2009). Scale bars: (A–D) 50 µm, (F, G) 500 nm. (See color plate.)
5. 3D Correlation of CLSM and FIB-SEM and Its Practical Relevance

The datasets acquired by CLSM and FIB-SEM are of very different modalities, which
are reflected mainly in data type, voxel size, and imaging orientation. In fluorescence
microscopy, a structure is represented by the presence or absence of a fluorescent signal.
Segmentation (manual or automatic) of this kind of data is relatively easy. However,
the information contained in electron microscopic data is much more complex. Merg-
ing these two types of datasets automatically is currently impossible without manual
interactions. Software tools, such as Amira (Visage Imaging GmbH, Germany), enable
the simultaneous visualization of the two datasets, and offer algorithms for automatic
alignment of the two datasets after manual prealignment. The original imaging plane of
the FIB-SEM volume is perpendicular to that of the CLSM volume, and merging these
is only meaningful for later analyses, if at least one of the two datasets was recorded
with isotropic voxels. This and any preprocessing of the datasets, such as image reg-
istration of one of the image stacks, may result in mismatching during merging. Reg-
istration of image stacks may become necessary due to drift during volume recording.
Freeware tools, such as Fiji or Image J, allow easy and reliable algorithms for registra-
tion, stitching, or filtering (Schindelin, 2008; Walter et al., 2010).
For some applications, a full 3D correlation of the two datasets may not be necessary.
When working with monolayers of cells, the fluorescent imaging may function as a map
for navigation to locate a specific ROI or cell in the FIB-SEM. HeLa cells infected with
Vaccinia virus (western reserve strain) react with extrusion of large, transient mem-
brane blebs, which upon retraction cause endocytic internalization of the virus (Mercer
& Helenius, 2008). Studying this virus entry mechanism inevitably requires investiga-
tion of membrane interactions and membrane structures at electron microscopic resolu-
tion. As the cell surface with protruding filopodia, membrane blebs, and bound virus
particles constitutes a rather complex 3D object (Fig. 5(F)), effective analysis of the
ultrastructure implicates 3D imaging. After HPF, FS in the presence of Safranin O
and embedding in HM20, cells showing extensive blebbing were chosen by CLSM
and FIB-SEM volumes were recorded from the corresponding ROI (Fig. 5(A-D)).
Three-dimensional models of the FIB-SEM volumes revealed virus particles (red) in
close association with endocytic invaginations next to retracting blebs (Fig. 5(G)).
These data suggest that the virus particles enter the cell adjacent to retracting blebs
and that macropinocytic uptake of virions occurs as part of the bleb retraction process,
rather than through opening of the bleb itself (Mercer & Helenius, 2009). This could
only be discovered by 3D EM, as the exact position of a single virion with respect to the
intra- or extracellular sides of the cell membrane cannot be determined by LM.
Applications involving larger volumes require a complete correlation of CLSM
and FIB-SEM data in order to comprehend the 3D organization of the sample, e.g. to
understand the colonization process of legume root nodules by nitrogen-fixing bacteria
(Fig. 6). This is a symbiosis providing the host plant with nitrogen, which in return pro-
vides nutrients for the bacteria (Bradyrhizobium japonicum). Sections of mung bean root
nodules were stained with Acridine orange during FS. An ROI was chosen by CLSM
Fig. 6 3D CLEM for plant samples. A root nodule from mung bean (Vigna radiata) colonized with nitrogen-fixing
bacteria (B. japonicum) was extracted fresh from the plant. 200 µm thick sections were high-pressure frozen after
degassing in 1-hexadecene, followed by FS and embedding in HM20. The FS medium contained uranyl acetate,
osmium tetroxide, and Acridine orange for fluorescent labeling of the bacteria. (A) In the CLSM image obtained
from the resin block-face cells with (P) and without (*) bacteria population can be distinguished, as well as vacu-
oles (V) and weakly stained nuclei (N). The fluorescence of the cell walls is rather attributed to autofluorescence
of the plant sample. The corresponding area imaged in SEM is shown in (B). Four cells are outlined (yellow dot-
ted line) to illustrate the correspondence. The white rectangle marks the position where an FIB-SEM volume was
recorded. The CLSM and FIB-SEM volumes were merged using the Amira software package. A single plane of
the merged volumes demonstrates the correlation of the two datasets (C). Parts of four cells were included in the
FIB-SEM volume, three of which showed intensive fluorescent signal attributed to bacteria, while the fourth cell
seemed empty. FIB-SEM imaging confirmed that the fluorescent signal coincides with the presence of bacteria.
(D) A 3D representation of both volumes shows the spatial relation. The imaging plane of FIB-SEM is perpendicu-
lar to that of CLSM. (E) A detailed 3D model of the FIB-SEM volume illustrates the organization of the bacteria
inside the root nodule cells. Several bacteria (green) are surrounded by a peribacteroid membrane (whitish) forming
a so-called symbiosome. Internal structures such as the nucleoid (blue) and the polyhydroxybutyrate inclusions
(green bubbles) can also be distinguished in FIB-SEM images. Scale bars: (A–C) 10 µm and (E) 500 nm.Volumes
[x-y-z]: (D) CSLM 57.2 × 57.2 × 9.0 µm3, FIB-SEM 21.8 × 6.2 × 15.9 µm3. (See color plate.)
(Fig. 6(A)) including cells colonized with bacteria as well as noncolonized cells. Cells
devoid of bacteria showed no fluorescence or only a few “spots,” as they usually contain
large vacuoles. The corresponding volume recorded by FIB-SEM proved the direct cor-
relation of the fluorescence signal with the occurrence of bacteria or large vesicles in the
cells (Fig. 6(A-D)). The fluorescence from the cell walls is rather attributed to autofluo-
rescence of the plant material. A 3D reconstruction of the volume (Fig. 6(E)) illustrates
the intracellular organization of the bacteria (green) in so-called symbiosomes (Oldroyd
& Downie, 2008) surrounded by a peribacteroid membrane (whitish), which is not vis-
ible by CLSM. Within the cytoplasm of the bacteria, substructures can be distinguished,
showing electron-dense domains with different contrast, which can be attributed to the
nucleoid (blue) and polyhydroxybutyrate inclusions (green bubbles; Fig. 6(E)).
D. Serial-Section SEM and Correlative Array Tomography

As an alternative to the “destructive” 3D CLEM approach based on FIB-SEM,
serial-section SEM and correlative array tomography (CAT) enable even larger ROIs to
be investigated. Serial sections mounted on glass support can be labeled for light and
electron microscopy. With the introduction of special surface coatings, such as chro-
mium or indium tin oxide (ITO), glass supports become highly conductive for SEM
investigation, while transparency for LM applications is maintained. We recommend
the ITO-coverslips from Optics Balzers (Liechtenstein), because of their minimized
surface roughness. Relocation of the ROI after LM investigation can be facilitated by
“Shuttle & Find” (Carl Zeiss Microscopy, Germany), a software- and hardware-based
interface to connect LM imaging with SEM.
1. Serial Sectioning and Labeling for Light Microscopy

The preparation of serial sections for LM and SEM, known also as array tomo
graphy (Micheva & Smith, 2007) is less demanding than for TEM. Individual sections
and section ribbons can be much larger, and are easier to handle and pick up onto glass
coverslips without being masked by a supporting grid. Additionally, the section thick-
ness can be increased up to 150–200 nm if needed, which also simplifies the handling.
As the preparation of serial sections has been described elaborately elsewhere (Blumer,
Gahleitner, Narzt, Handl, & Ruthensteiner, 2002; Micheva, O’Rourke, Busse, & Smith,
2010; Micheva & Smith, 2007), we will focus here on the most important aspects
concerning array tomography. While cutting serial sections, best possible trimming of
the cutting face to trapezoid or rectangular shape with a perfectly parallel cutting edge
is essential for obtaining consistent ribbons; otherwise, the ribbons form arcs or tear
apart. However, imperfect ribbons can still be analyzed without loss of information,
as the substrate onto which the ribbons are mounted can be much larger (e.g., 22 ×
22 mm) than the standard 3 mm TEM grid. In addition, the field of view can easily be
adapted accordingly in both LM and SEM. If sections do not form ribbons, applying a
thin film of glue to the cutting face of the resin block or adding a few droplets of ethanol
to the water in the boat may help holding the sections together (Blumer et al., 2002).
To facilitate adherence of the sections, it is advisable to discharge glass coverslips and
especially ITO-coated glass supports before use. A positive charge has proven conve-
nient for ITO-coated glass. We usually let the sections dry on a hotplate at 40°C for
30 min, to ensure that the sections attach well to substrate and are not washed off dur-
ing post-staining or labeling. Folds and ripples should be avoided, as they are potential
sources for charging artifacts during SEM imaging.
As fluorescent dyes are introduced already during FS, this labeling may be sufficient to
characterize a sample and choose an ROI by LM. This inherent fluorescence is detectable
by wide-field fluorescence microscopy even in 70 nm sections (Fig. 7(F) and Fig. 8(A)).
For samples without preembedding labeling, histological staining methods, such as Hema-
toxylin and Eosin, or Toluidine blue can be applied (Fig. 8(A)). Fluorescent immunolabel-
ing can be used to specifically identify ROIs for high-resolution SEM investigations.
Histological staining or antibody-labeling was done according to standard protocols
as described, e.g. by Trump, Benditt, and Smuckler (1961) for Toluidine blue, or for
immunolabeling by Schwarz and Humbel (2007), or Vielhaber, et al. (2001), respec-
tively. However, when immunolabeling for SEM applications, serum-containing buf-
fers should be omitted, because the serum residues may smear the surface of the section
and interfere with imaging.
2. Large Field of View SEM on Resin Sections

Samples freeze-substituted with uranyl acetate and osmium tetroxide, and embed-
ded in HM20 as described above proved well suited for SEM imaging without any
further staining (Fig. 7(G) and Fig. 8). However, should the inherent staining provide
insufficient contrast, sections can be post-stained in the same way as for TEM applica-
tions (Ellis, 2007). Staining with aqueous uranyl acetate for 5 min, followed by 30–60 s
lead citrate (Reynolds, 1963), works fine with HM20 sections.
The “Shuttle & Find” application, linking LM and SEM by means of a 3-point calibra-
tion on a dedicated sample holder, simplifies the task of relocating an ROI in the SEM.
The calibration creates a set of relative coordinates for every image acquired by LM that
is saved in the header of each image. After repeating the calibration in the SEM and thus
setting up the same relative coordinate system, the LM images can be used for navigation.
Options for scanning large fields of view are available for most modern SEM platforms.
We employ the ATLAS system, an addendum for Zeiss SEM systems, consisting of an
additional scan generator and control software allowing automated acquisition of fields
of view of up to 32k square pixel. A mosaic function facilitates flexible definition of very
large fields of view. Having relocated the ROI using the “Shuttle & Find” tool, the area
to be recorded by SEM and the imaging conditions are defined. Multiple positions can be
saved into one ATLAS experiment for automated image acquisition. Imaging conditions,
such as type of signal detection, acceleration voltage, focus, brightness and contrast, astig-
matism corrections, and working distance are set using the standard SEM controls and are
then saved into the ATLAS control software for each position. Autofunctions for focus,
astigmatism, brightness and contrast are available, but are usually not required.
As for FIB-SEM imaging, optimum imaging conditions have to be determined for
each sample and preparation method, but the settings used for the skin sample presented
in Fig. 8 may serve as a starting point. These images were recorded using the in-lens SE
Fig. 7 Multiple use of one sample for different imaging modes. High-pressure frozen samples of C. elegans in cellulose capillar-
ies stained during FS with uranyl acetate and osmium tetroxide, and embedded in HM20 can be used for multiple light and electron
microscopic investigations. The sample shown in (A–E) was additionally stained with Nile blue sulfate during FS, and Safranin O was
used for the sample depicted in (F) and (G). (A) The CSLM volume shows a segment of one nematode (u-shaped structure). The wall
of the enclosing cellulose capillary has a strong autofluorescence (asterisk). The blue box marks the position of the FIB-volume shown
in (B). (C) Prior to FIB-SEM investigation, 75 nm sections of this sample were investigated by TEM without post-staining. This image
corresponds to the area just above the blue box indicating the position of the FIB-SEM volume. (D) A single frame taken from the stack
acquired by FIB-SEM demonstrates the imaging quality which is comparable to TEM. The protective carbon layer on the surface of the
resin block is marked by a dashed line. (E) Resin sections can also be investigated by SEM. For this, 100 nm sections were mounted
on ITO-coated coverslips and imaged at 2 kV using an in-lens SE detector without post-staining. ITO-coated glass support is not only
highly conductive for SEM application, but also remains transparent for LM. The fluorescence signal of this resin section (Safranin O
was applied during FS) is well detectable (F). This section was taken after FIB-SEM investigation of the sample; the asterisk marks a
hole in the section resulting from the FIB-trench. (G) “Shuttle & Find” was used to navigate and find the same area (as indicated by
orange frame in (F)) for SEM imaging. Fissures in the resin section reveal the ITO substrate (arrowheads).Volumes [x-y-z]: (A) 90 × 90
× 43 µm3, (B) 15.7 × 18.8 × 13.6 µm3. Scale bars: (C–E) 5 µm, (F) 20 µm, and (G) 5 µm. (See color plate.)
Fig. 8 Array tomography. 100 nm thick serial sections of a human skin biopsy, high-pressure frozen, freeze-substituted with uranyl
acetate, osmium tetroxide, and Safranin O, and embedded in HM20 were mounted on ITO-coated glass coverslips. (A) The first 15 sec-
tions were imaged by fluorescence LM without further staining, using the Safranin O signal introduced during FS. Further 19 sections
were stained with Toluidine blue for LM imaging, and the third part of the series (17 sections) was stained with Dapi. “Shuttle & Find”
was employed to navigate to the ROI for every section, and all 51 sections were then imaged by SEM at 2 kV using in-lens SE detection.
Combination of all images yields a volume that comprises multiple levels of information, ideal for in-depth characterization of complex
biological structures. (B) The entire ROI was imaged by SEM with a pixel size of 5 nm, yielding a large field of view at TEM-like imaging
resolution. Scanning was performed using the ATLAS system in tile scan mode. To cover the complete ROI, 4 × 3 tiles were recorded, and
thus the overlapping regions of the tiles were scanned twice. The resulting charging effect changes the contrast in the image, in this case
these areas appear darker. Enlargements are showing nuclei, cell membranes, and desmosomes (C), melanosome clusters in the basal cell
layer (D), and collagen fibers in the dermis (E). The contrast of these insets is reversed to demonstrate the TEM-like quality and resolution
of the SEM micrographs.Volumes [x-y-z]: (A) complete LM stack: 321.0 × 221.5 × 5.1 µm3, SEM stack: 226.4 × 170.0 × 5.1 µm3. Scale
bars: (B) 20 µm and (C–E) 1 µm. (See color plate.)
detector with an acceleration voltage of 1.8 kV and a pixel size of 5 nm. According to
sample quality, preparation method, and thickness of the resin sections, the acceleration
voltage may be varied between 0.8 and 3 kV. While setting up experiments with the
ATLAS system, one should keep in mind that a single 32k × 32k pixel dataset amounts to
1 GB, which poses a considerable challenge in terms of image processing, even by mod-
ern standards. As a compromise we chose to record overlapping tiles of 16k × 16k pixel
images instead, accepting the time loss and beam damage caused by multiple scanning
of the overlap areas.
The most common sources for artifacts while imaging resin sections by SEM are
beam damage, contamination, and charging effects. Despite the conductive ITO-
support, ripples or poor contact between sections and supporting layer may result in
charge deposition and hence charging artifacts. The heavy metal staining in the sec-
tions acts as local charge distributor. In order to achieve stable imaging conditions,
the acceleration voltage has to be adapted according to the section thickness, so that
a fraction of the electron interaction volume includes the ITO support. Beam damage,
on the other hand, depends on the “hardness” of the resin; sections of “softer” resin
are more likely to show beam damage. Long dwell times while scanning, although
improving image quality, induce break down and evaporation of parts of the resin.
This changes the imaging properties, which becomes visible while scanning an area
twice, e.g. for mosaics (dark vertical bands in Fig. 8(B)). However, this beam dam-
age is minimized in modern SEMs capable of low-dose, fast image acquisition with
shortened dwell times.
3. Correlative Array Tomography

The imaging tools described above facilitate tracking of an ROI over long ribbons
of serial sections. Various methods for labeling can be employed to identify ROIs by
LM and relocate them for high-resolution SEM. The identification of an ROI can be
based either on a descriptive histological overview or specific antibody labeling. As an
example, the presence of a sweat gland and its association to a hair shaft in sections of
human skin can readily be identified. This morphological description can be provided
e.g. by the Safranin O labeling introduced during FS, as well as Toluidine blue or Dapi
post-staining (Fig. 8(A)). Antibody labeling can be applied to recognize the ROI if
morphological features are insufficient, or to identify and localize biomolecules (e.g.,
glucosylceramide, a skin-barrier lipid precursor; Fig. 9).
Image stacks can be recorded in both light and electron microscopy tracking the ROI
in different dimensions. The resulting 3D datasets can be combined to a multimodal
volume, comprising a multilevel description of the sample. Modern image processing
algorithms allow stable, automated image alignment to combine the images of serial
sections to volume datasets. TrakEM2 (Cardona et al., 2010), a plug-in for Fiji, pro-
vides the necessary tools for stitching of mosaic tiles and registration of image stacks
from light and electron microscopy. This can be done individually or by combining the
different volumes. As for the combination of CLSM and FIB-SEM data, 3D modeling
and visualization can be done in any software package enabling simultaneous process-
ing of more than one dataset at a time (e.g., Amira).
Fig. 9 Immunolabeling as a tool to identify the ROI for high-resolution SEM. Sections of human skin were
labeled for glucosylceramide, a sphingolipid involved in mitogenesis and final epidermal differentiation. The
sections were mounted on ITO-coated coverslips for correlative imaging in LM and SEM. (A) Immunofluores-
cence shows the localization of glucosylceramide in lamellar bodies in the stratum granulosum and upper stratum
spinosum, accounting for the typical bead chain-like pattern once exported into the intercellular space at the inter-
face to the stratum corneum (represented in orange; Vielhaber et al., 2001). Nuclei were stained with Dapi (blue).
(B) The ROI was relocated in the SEM using “Shuttle & Find.” (C) The overlay image of fluorescence signal
and SEM demonstrates the correlation. The positions of the cutouts are marked by white frames in the fluo-
rescence (A) and SEM image (B), respectively. (D) The difference in imaging resolution becomes apparent in
the detail of the overlay image (position indicated by white frame in C). The fluorescence signal shows rather
large ovoid shapes of up to 500 nm in diameter (delineated by white dotted lines for better visibility in the grey-
value representation). But the SEM image reveals clusters of up to five vesicles or lamellar bodies no larger
than 100–200 nm where only one fluorescence spot is detected, or expose vesicles that have no corresponding
fluorescence spot at all. Scale bars: (A, B) 25 µm, (C) 2 µm, and (D) 500 nm. (See color plate.)
4. Immunolocalization of Biomolecules
As mentioned in the beginning, HPF combined with FS is the only method to
preserve the exact position of small biomolecules, such as lipids for high-reso-
lution imaging (Humbel & Schwarz, 1989; Verkleij, Humbel, Studer, & Muller,
1985). Localizing these molecules can be performed by the use of primary antibod-
ies against the target antigen (Fig. 9: rabbit antiserum against glucosylceramide)
combined with secondary fluorescence-labeled antibodies directed against the pri-
mary antibodies (Fig. 9(A)). This allows relatively fast to capture the distribution
of the antigen on a larger scale by LM. Correlative microscopy then enables the
recording of the identical area at higher resolution in the SEM (Fig. 9(B) shows the
corresponding area to 9(A)). Fig. 9(C) and (D) also highlights the advantages and
limitations of immunofluorescence, i.e. a fast and easy localization of the binding
sites, but on the other hand, losing the context in the tissue since glucosylceramide
is limited to lamellar bodies in the first layers of the stratum corneum. Hence, the
cellular context is neither stained nor visible without additional staining. The reso-
lution of the highlighted “fluorescence spots” does not allow to distinguish if the
label is localized in the intracellular or intercellular space, nor if there are differ-
ent structures related to these binding sites (Fig. 9(C)). With the underlying high-
resolution information gained by SEM imaging of the same section (Fig. 9(D)), one
can easily distinguish both. As shown in Fig. 9(D), SEM reveals up to 5 vesicular
entities covered by only one “fluorescence spot”, while other vesicular structures
show no corresponding fluorescence label at all. Therefore, these vesicles cannot
be lamellar bodies but must be other vesicular structures present in these kerati-
nocytes. The functional distribution can be easily and directly extracted from the
distribution of the fluorescent labels in the LM images, whereas the correspond-
ing subcellular structures can be allocated by a simple superimposition of the two
image modes even without immunolabeling for EM. The ideal probe for CAT would
involve a fluorescent and a colloid metal label simultaneously linked to the second-
ary antibody (e.g., Fluoro-Gold or Fluoronanogold), to enable optimum localization
precision at the SEM level.
IV. Materials and Instrumentation
Instrumentation: We use two models of high-pressure freezers: the Bal-Tec HPM 010
(Bal-Tec AG, Principality of Liechtenstein) and the Leica EM HPM 100 (Leica Micro-
systems, Austria). Additional instrumentation used for the described experiments include
a carbon coater for coating of sapphire discs (BAE 120, Bal-Tec AG, Principality of
Liechtenstein), a hand microtome (Windaus Labortechnik GmbH, Germany), and a pre-
vacuum chamber for degassing of samples before freezing, if needed.
Materials: Standard aluminum planchettes (3 mm diameter, 200 µm indentation)
and aluminum spacer rings (3 mm diameter, 100 µm thick) were used for freezing
of tissue samples in both instruments (Leica Microsystems, Austria). Round sapphire
discs (3 mm diameter, 100 µm thick; Rudolf Brügger SA, Minusio, Switzerland),

evaporation coated with carbon. Cellulose capillaries with an inner diameter of 200
µm (Leica Microsystems, Austria), and gel-loading pipette tips (Vaudaux-Eppendorf
AG, Switzerland). Sterile punch biopsies with a diameter of 2 mm (Stiefel GmbH,
Germany).
Reagents: 1-Hexadecene (Fluka, Switzerland) served as filling medium of alumi-
num planchettes for all types of samples, except for cell culture samples, which were
frozen in the respective culture medium.

Instrumentation: Automated FS device (self-build) consisting of a cooling chamber
holding the sample containers with ethanol as cooling agent and a temperature control-
ler (Tecon AG, Switzerland) for automatic temperature shifting. A UV lamp can be
placed on top of the cooling unit for low-temperature embedding and polymerization.
Alternatively, an oven for heat curing (60°C) of resins after room temperature embed-
ding is needed.
Materials: Eppendorf tubes. FS baskets made from Eppendorf pipette tips, cut to 5
mm length, with a solvent-resistant ethylene tetrafluoroethylene netting with a mesh
size of 210 µm (Bückmann GmbH & Co. KG, Mönchengladbach, Germany) glued to
the bottom of the truncated pipette tip by carefully melting the plastic on a hot plate.
Reagents: Acetone (ACS grade; Sigma, Switzerland) dried over molecular sieve,
supplemented with 2% uranyl acetate (Fluka, Switzerland), and/or 1% osmium tetrox-
ide (EMS, USA). Fluorescent dyes were added to the FS cocktail. A list of tested dyes,
together with the type of samples they were tested on can be found in Table I. Samples
were embedded in either Lowicryl HM20 (Polysciences Europe GmbH, Germany) or
Epoxy resin (Epoxy embedding medium kit; Fluka, Switzerland).

Instrumentation: Ultramicrotome Leica EM UC6 (Leica Microsystems, Austria).
Confocal Laser Scanning Microscope Zeiss LSM-510 Meta (Carl Zeiss Microscopy,
Germany). FIB-SEM CrossBeam workstation NVision 40 (Carl Zeiss Microscopy,
Germany), equipped with a gas injection system (GIS) for local deposition of carbon
or platinum, and SE in-lens and EsB (BSE) detectors. Sputter coater (gold or platinum)
to render resin blocks conductive for SEM (SCD 050, Bal-Tec AG, Principality of
Liechtenstein).
Materials: Ultramicrotomy sample holders (Leica Microsystems, Austria), diamond
trimming knife (Cryotrim 20°; Diatome AG, Switzerland), and SEM sample stubs
(Plano GmbH, Germany). Specially designed sample holders for CSLM to investigate
resin blocks while clamped into the ultramicrotomy holder, or mounted on SEM stubs
(self-built).
Reagents: Immersion oil, conductive silver paste (Demetron Leitsilber 200, Hanau,
Germany).
Software: Fiji (freeware; www.fiji.sc). Amira (Visage Imaging GmbH, Germany).
D. Serial-Section SEM and Array Tomography

Instrumentation: Ultramicrotome Leica EM UC6 (Leica Microsystems, Austria),
Glow discharging system (Emitech Plasma Cleaner K100X; Quorum Technologies
Ltd., United Kingdom). Hotplate for drying and staining sections on coverglass support.
Wide field light microscope Zeiss Axio Imager M1 (Carl Zeiss Microscopy, Germany)
and scanning electron microscope Zeiss Gemini 1530 FEG (Carl Zeiss Microscopy,
Germany) equipped with SE in-lens, SE2 and BSE detector systems. Both LM and
SEM are operated with the AxioVision software to facilitate the “Shuttle & Find” con-
nectivity (Carl Zeiss Microscopy, Germany). Additional equipment on the SEM for
scanning of large fields of view: ATLAS system (release 3.5.2; Carl Zeiss Microscopy,
Germany) comprising an additional scan generator and control software.
Materials: Razor blades for manual trimming. Diamond knife with large boat spe-
cially designed for serial sectioning and pick-up of section ribbons onto glass slides
(Ultra Jumbo 35°, Diatome AG, Switzerland). Glass coverslips coated with indium tin
oxide (Optics Balzers AG, Principality of Liechtenstein). Tweezers and eyelash glued
to a wooden pick for handling and pickup of serial sections onto glass slides. “Shuttle
& Find” sample holders with calibration marks suitable for both LM and SEM.
Reagents: Acrylic glue (e.g., Pattex compact by Henkel). Labeling solutions for
resin sections: Rabbit antiserum against glucosylceramide (GlycoTech Produktions-
und Handelsgesellschaft mbH, Germany), secondary antibody: Cy2 AffiniPure Don-
key Anti-Rabbit IgG (Jackson ImmunoResearch Europe Ldt., United Kingdom), Dapi
(4´,6-diamidino-2-phenylindole, dihydrochloride; Invitrogen); Toluidine blue (Merck,
Switzerland) in sodium tetraborate decahydrate buffer.
Software: Fiji (freeware; www.fiji.sc), Amira (Visage Imaging GmbH, Germany).
V. Discussion
The two 3D-CLEM approaches presented here are not per se mutually exclusive but
have their distinct advantages. CLSM combined with FIB-SEM is most suitable for
targeted 3D correlation of small volumes, whereas serial-section LM and SEM (i.e.,
CAT) imaging has its strength in the large field of view and volume screening, and cor-
relation and can be combined with immunocytochemical methods. Both methods have
the potential to extract statistically relevant data of structural details, if the structures of
interest can be selected and highlighted.
A. Fluorescence Staining During Freeze-Substitution is a Comprehensive Tool for CLEM

Application of fluorescent labels during FS presents a form of preembedding labeling,
but without the drawback of interfering with the preservation of ultrastructure. Com-
mon preembedding labeling techniques involve permeabilization of cell membranes
to facilitate the introduction of specific labels or extensive treatment with reagents to

trigger site-specific reactions (Giepmans, Deerinck, Smarr, Jones, & Ellisman, 2005;
Grabenbauer et al., 2005). Although these methods allow specific labeling for CLEM,
the artifacts induced by the labeling process may result in changes of ultrastructure and
redistribution of intracellular and extracellular components (Griffiths, 1993). Labeling
during FS after HPF avoids these artifacts. However, this kind of preservation of the
ultrastructure comes at the price of unspecific preembedding labeling: the fluorophores
introduced during FS offer only semispecific quality of labeling. Lipophilic domains,
such as lipid membranes, can be labeled reliably using DiIC18, DiD, or DiO6, while
most common nuclear stains do not produce any specific labeling. Some general his-
tological dyes as Acridine orange, Safranin O, or Nile blue sulfate can be used to stain
nuclei, if needed. On the other hand, FS after HPF and embedding in HM20 has been
proven to retain antigenicity (Pfeiffer et al., 2000; Robertson et al., 1992; Schwarz &
Humbel, 1989, 2007), allowing specific labeling at a later stage of sample preparation.
Dyes have to be chosen according to sample type and aim of the project. For cell culture
samples, the fluorophores may serve as a navigational aid for subsequent FIB-SEM inves-
tigation. In this case, specific labeling is unnecessary, as long as a cell of interest can be
identified by its shape and/or position. For complex tissue samples, the introduced fluores-
cent dyes provide a histological description, enabling the identification of an ROI based on
morphological features. In both cases, the process of elaborate searching for an ROI in EM
is simplified. As the staining results may vary even between different types of cell cultures,
dyes must be evaluated for each new application and sample type. However, the combina-
tion of the lipophilic dye DiIC18 and the histological stain Nile blue sulfate produced suit-
able results for most of the tested sample types, and can be recommended for initial testing.
Combining simultaneous fluorescence labeling with heavy metal staining for EM
is possible. However, intense membrane staining for EM, especially using osmium
tetroxide, means a trade-off in fluorescence. The combination of uranyl acetate and flu-
orescent dyes is a good starting point, especially if the same sample is going to be used
for CAT or immunocytochemical localization in conjunction with FIB-SEM inves-
tigations. Most preparation methods aiming at optimum membrane labeling for EM
techniques utilize either osmium tetroxide alone or in combination with either potas-
sium ferrocyanide or tannic acid (De Winter et al., 2009; Giddings, 2003; Hess, 2003;
Jimenez et al., 2009; Knott et al., 2008). Irrespective of the fixation and embedding
method, this intense membrane labeling is incompatible with fluorescence labeling. On
the other hand, osmium tetroxide is indispensable for a clear detection of membranes
when employing en-bloc imaging techniques as FIB-SEM.
The FS protocol including fluorophores was initially developed for combined appli-
cation of CLSM and TEM using post-stained sections (Biel et al., 2003). We now
adapted this protocol to facilitate en-bloc FIB-SEM imaging to include 1% osmium
tetroxide in the FS medium. An important modification in this protocol, compared to
those discussed above, is to keep the samples below −50°C during osmium penetration.
This prevents the blackening of the sample which interferes with fluorescence detec-
tion, and also reduces osmium-mediated fluorescence quenching. Any excess osmium
is removed at −50°C and the substitution medium is replaced with pure organic solvent
to finish the FS experiment. The resulting membrane staining is not as pronounced as,
e.g., after potassium ferrocyanide-mediated osmium treatment (Fig. 4), but is sufficient
to detect and trace the membranes for 3D analysis and modeling.
B. 3D CLEM: FIB-SEM or Correlative Array Tomography?

Correlative microscopy covers a lot of different imaging techniques. Here, we have
presented two approaches for correlating LM and SEM imaging to acquire 3D datas-
ets. Combined with the sample preparation methods discussed above, both FIB-SEM
and array tomography have the potential to replace ultrastructural investigations that are
nowadays predominantly performed by TEM. Cutting-edge SEM platforms allow imag-
ing with an x/y-resolution of 1–2 nm, which is sufficient to image lipid membranes.
Which of the two techniques might be better suited for a specific project has to be
determined according to the size and volume of the sample of interest, the desired
z-resolution, and the type of labeling essential for gaining the desired information.
In general, FIB-SEM has the advantage of automation. Volumes of up to 50 × 30
× 30 µm3 can be recorded overnight with a voxel size of 5–10 nm. This is not achiev-
able both in time and z-resolution by standard serial-section TEM, making FIB-SEM a
valid alternative (Knott et al., 2008; Lucas et al., 2008). However, specific fluorescent
labeling during FS is limited, and thus the identification of an ROI is achieved mainly
by comparison of recognizable features in the sample or on the sample surface.
In contrast, serial-section SEM or array tomography can provide a much larger field
of view. Additional equipment is available for most modern SEM platforms for auto-
mated recording of up to 32k square pixel and for relocation of the ROI. The introduc-
tion of special glass supports promoted this correlative imaging approach. Chromium
or ITO surface coatings render glass supports ideal for correlative imaging, as the cov-
erslips are highly conductive, while remaining transparent for LM. As a drawback, the
z-resolution of array tomography is limited by the sectioning process. It is challenging
to routinely cut serial sections of a thickness of less than 30–50 nm. Contrary to FIB-
SEM, which is a “destructive” method, serial sections can be stored and reinvestigated
ad libitum. In addition, such sections can be post-stained or labeled in order to facilitate
retrieval of the ROI for high-resolution SEM investigation. Histological staining or
immunolabeling is feasible (Biel et al., 2003; Vielhaber, Pfeiffer, et al. (2001)), and
if the inherent contrast is insufficient for SEM imaging, additional post-staining with
uranyl acetate and lead citrate can be applied.
VI. Summary
Fluorescent labeling during FS not only facilitates the task of relocating an ROI, but
additionally provides multimodal information of the same ROI in 3D. CLSM investiga-
tion prior to EM supplies valuable information on the structural context of the ROI. Such
data can be combined with high-resolution imaging of the identical ROI by either FIB-
SEM or serial-section SEM. Both approaches, CLSM combined with FIB-SEM, as well
as CAT combined with prior in-vivo LM investigation, have the potential to bridge the
gap between systems biology and high-resolution EM. The method of choice, however,
has to be determined de novo for each project. By merging data from different imaging
modalities, we are able to better understand the complexity of living systems in 3D.
Acknowledgments
The authors would like to thank all collaborators for providing samples and fruit-
ful discussions: Prof. Hans-Martin Fischer (Institute of Microbiology, ETH Zurich)
provided the mung bean root nodules. Dr. Stefanie Krämer (Institute of Pharmaceutical
Sciences, ETH Zurich) contributed the MDCK cells and the glucosylceramide anti-
serum. HeLa cells were obtained from Dr. Jason Mercer (Institute of Biochemistry,
ETH Zurich). The nematodes were kindly provided by Dr. Christel Genoud (Friedrich
Miescher Institute for Biomedical Research). Realizing automatic CAT SEM technol-
ogy would not have been possible without massive support and development by Carl
Zeiss Microscopy, Germany. In particular, we would like to thank the project leaders
Dr. Thomas Albrecht, Dr. Christian Boeker, and Dr. Martin Edelmann for continuous
and great support. We are also very grateful for the development of the special ITO-
coated coverglass for CAT by Alex Vogt (Optics Balzers, Lichtenstein).
References
Armer, H. E. J., Mariggi, G., Png, K. M.Y., Genoud, C., Monteith, A. G., Bushby, A. J., et al. (2009). Imag-
ing transient blood vessel fusion events in zebrafish by correlative volume electron microscopy. PLoS
ONE, 4, e7716.
Biel, S. S., Kawaschinski, K., Wittern, K. P., Hintze, U., & Wepf, R. (2003). From tissue to cellular ultrastruc-
ture: closing the gap between micro- and nanostructural imaging. Journal of Microscopy, 212, 91–99.
Blumer, M. J. F., Gahleitner, P., Narzt, T., Handl, C., & Ruthensteiner, B. (2002). Ribbons of semithin sec-
tions: an advanced method with a new type of diamond knife. Journal of Neuroscience and Methods,
120, 11–16.
Brade, L., Vielhaber, G., Heinz, E., & Brade, H. (2000). In vitro characterization of anti-glucosylceramide
rabbit antisera. Glycobiology, 10, 629–636.
Buser, C., & Walther, P. (2008). Freeze-substitution: the addition of water to polar solvents enhances the
retention of structure and acts at temperatures around −60 degrees C. Journal of Microscopy, 230,
268–277.
Cardona, A., Saalfeld, S., Preibisch, S., Schmid, B., Cheng, A., Pulokas, J., et al. (2010). An integrated
micro- and macroarchitectural analysis of the Drosophila brain by computer-assisted serial section elec-
tron microscopy. PLoS Biology, 8, e1000502.
theory and practice. Journal of Electron Microscopy Technique, 13, 165–174.
Droste, M. S., Biel, S. S., Terstegen, L., Wittern, K. P., Wenck, H., Wepf, R. (2005). Noninvasive
measurement of cell volume changes by negative staining. Journal of Biomedical Optics. 10, 064017.
Elias, P. M., & Friend, D. S. (1975). Permeability barrier in mammalian epidermis. Journal of Cell Biology,
65, 180–191.
Ellis, E. A. (2007). Poststaining grids for transmission electron microscopy: conventional and alternative
protocols. Methods in Molecular Biology, 369, 97–106.
Eppenberger-Eberhardt, M., Riesinger, I., Messerli, M., Schwarb, P., Muller, M., Eppenberger, H. M., et al.
(1991). Adult-rat cardiomyocytes cultured in creatine-deficient medium display large mitochondria with
paracrystalline inclusions, enriched for creatine-kinase. Journal of Cell Biology, 113, 289–302.
Giddings, T. H. (2003). Freeze-substitution protocols for improved visualization of membranes in high-

pressure frozen samples. Journal of Microscopy, 212, 53–61.
Giepmans, B. N., Deerinck, T. J., Smarr, B. L., Jones, Y. Z., & Ellisman, M. H. (2005). Correlated light and elec-
tron microscopic imaging of multiple endogenous proteins using Quantum dots. Nature Methods, 2, 743–749.
857–862.
Griffiths, G. (1993). Fine structure immunocytochemistry (1st ed.). Heidelberg, Germany: Springer-Verlag.
Hawes, P., Netherton, C. L., Mueller, M., Wileman, T., & Monaghan, P. (2007). Rapid freeze-substitution
preserves membranes in high-pressure frozen tissue culture cells. Journal of Microscopy, 226, 182–189.
Hess, M. W. (2003). Of plants and other pets: practical aspects of freeze-substitution and resin embedding.
Journal of Microscopy, 212, 44–52.
Hess, M. W. (2007). Cryopreparation methodology for plant cell biology. In J. R. McIntosh (Ed.), Cellular
electron microscopy (Vol. 79, pp. 57–100). San Diego, CA: Academic Press.
Hohenberg, H., Mannweiler, K., & Muller, M. (1994). High-pressure freezing of cell-suspensions in cel-
lulose capillary tubes. Journal of Microscopy, 175, 34–43.
Hohenberg, H., Tobler, M., & Muller, M. (1996). High-pressure freezing of tissue obtained by fine-needle
biopsy. Journal of Microscopy, 183, 133–139.
Holleran, W. M., Takagi, Y., Menon, G. K., Jackson, S. M., Lee, J. M., Feingold, K. R., et al. (1994). Per-
meability barrier requirements regulate epidermal beta-glucocerebrosidase. Journal of Lipid Research, 35,
905–912.
Humbel, B. M., & Schwarz, H. (1989). Freeze-substitution for immunochemistry. In A. J. Verkleij, & J. L. M.
Leunissen (Eds.), Immuno-gold labeling in cell biology (pp. 115–134)). Boca Raton, FL:: CRC Press, Inc.
microscopy: providing new understanding in the biomedical and plant sciences. Micron, 43, 565–582.
Jimenez, N., Vocking, K., van Donselaar, E. G., Humbel, B. M., Post, J. A., et al. (2009). Tannic acid-mediated
osmium impregnation after freeze-substitution: a strategy to enhance membrane contrast for electron tomog-
raphy. Journal of Structural Biology, 166, 103–106.
Karnovsky, M. J. (1965). A formaldehyde-glutaraldehyde fixative of high osmolality for use in electron
microscopy. Journal of Cell Biology, 27. A137-A138.
Kirschning, E., Rutter, G., & Hohenberg, H. (1998). High-pressure freezing and freeze-substitution of native
rat brain: suitability for preservation and immunoelectron microscopic localization of myelin glycolipids.
Journal of Neuroscience Research, 53, 465–474.
Lucas, M. S., Gasser, P., Günthert, M., Mercer, J., Helenius, A., & Wepf, R. (2008). Correlative 3D micros-
copy: CLSM and FIB/SEM tomography used to study cellular entry of vaccinia virus. Presented at EMC
2008, Aachen, Germany.
McDonald, K., & Morphew, M. K. (1993). Improved preservation of ultrastructure in difficult-to-fix organ-
isms by high pressure freezing and freeze substitution: I. Drosophila melanogaster and Strongylocentro-
tus purpuratus embryos. Microscopy Research and Techniques, 24, 465–473.
McDonald, K., Schwarz, H., Müller-Reichert, T., Webb, R., Buser, C., & Morphew, M. (2010). Tips and
tricks for high-pressure freezing of model systems. In M. -R. Thomas (Ed.), Methods of cell biology
(Vol. 96, pp. 671–693). San Diego: Elsevier Academic Press Inc.
Mercer, J., & Helenius, A. (2008). Vaccinia virus uses macropinocytosis and apoptotic mimicry to enter host
cells: supporting online material. Science, 320(5875), 531–535.
Mercer, J., & Helenius, A. (2009). Virus entry by macropinocytosis. Nature of Cell Biology, 11, 510–520.
Micheva, K. D., O’Rourke, N., Busse, B., & Smith, S. J. (2010). Array tomography: production of arrays.
Cold Spring Harbor Protocols. 2010, pdb, prot5524.
Monaghan, P., Cook, H., Hawes, P., Simpson, J., & Tomley, F. (2003). High-pressure freezing in the study
of animal pathogens. Journal of Microscopy, 212, 62–70.
Moor, H. (1973). Cryotechnology for the structural analysis of biological material. Freeze-etching Tech-
niques and Applications, 11–19.
Müller, M. (1992). The integrative power of cryofixation-based electron microscopy in biology. Acta Micro-
scopica, 1, 37–44.
Murk, J. L., Posthuma, G., Koster, A. J., Geuze, H. J., Verkleij, A. J., Kleijmeer, M. J., et al. (2003). Influence
of aldehyde fixation on the morphology of endosomes and lysosomes: quantitative analysis and electron
tomography. Journal of Microscopy, 212, 81–90.
Oldroyd, G. E., & Downie, J. M. (2008). Coordinating nodule morphogenesis with rhizobial infection in
legumes. Annual Review of Plant Biology, 59, 519–546.
Pfeiffer, S., Vielhaber, G., Pflucker, F., Wepf, R., & Hintze, U. (1997). Structural conservation of human
skin: a comparison of chemical fixation with cryoimmobilization and different cryopreservation meth-
ods. European Journal of Cell Biology, 74. 101–101.
Pfeiffer, S., Vielhaber, G., Vietzke, J. P., Wittern, K. P., Hintze, U., & Wepf, R. (2000). High-pressure
freezing provides new information on human epidermis: simultaneous protein antigen and lamellar lipid
structure preservation. Study on human epidermis by cryoimmobilization. Journal of Investigative Der-
matology, 114, 1030–1038.
Reipert, S., Fischer, I., & Wiche, G. (2004). High-pressure freezing of epithelial cells on sapphire coverslips.
Reynolds, E. S. (1963). Use of lead citrate at high pH as an electron-opaque stain in electron microscopy.
Riehle, U., & Hoechli, M. (1973). The theory and technique of high pressure freezing. In E. L. Benedetti,
& P. Favard (Eds.), Freeze-etching techniques and applications (pp. 11–19). Paris: Société Française de
Microscopie Electronique.
Riehle, U. (1968). Über die Vitrifizierung verdünnter wässriger Lösungen Dissertation, Eidgenössische
Technische Hochschule ETH Zürich. ETH No. 4271.
Ripper, D., Schwarz, H., & Stierhof, Y. D. (2008). Cryo-section immunolabelling of difficult to preserve
specimens: advantages of cryofixation, freeze-substitution and rehydration. Biology of the Cell, 100,
109–123.
Robertson, D., Monaghan, P., Clarke, C., & Atherton, A. J. (1992). An appraisal of low-temperature embed-
ding by progressive lowering of temperature into Lowicryl-Hm20 for immunocytochemical studies.
Rothen-Rutishauser, B., Kramer, S. D., Braun, A., Gunthert, M., & Wunderli-Allenspach, H. (1998). MDCK
cell cultures as an epithelial in vitro model: cytoskeleton and tight junctions as indicators for the defini-
tion of age-related stages by confocal microscopy. Pharmaceutical Research, 15, 964–971.
Schindelin, J. (2008). Fiji is just ImageJ – Batteries included. Presented at ImageJ User and Developer
Conference, Luxembourg.
Schwarb, P. (1990). Morphologische Grundlagen zur Zell-Zell Interaktion bei adulten Herzmus-
kelzellen in Kultur. Dissertation, Eidgenössische Technische Hochschule ETH Zürich. DISS. ETH
No. 9195.
Schwarz, H., & Humbel, B. M. (1989). Influence of fixatives and embedding media on immunolabelling of
freeze-substituted cells. Scanning of Microscopy (Suppl. 3), 57–63. discussion 63–54.
Schwarz, H., Hohenberg, H., & Humbel, B. M. (1993). Freeze-substitution in virus research: a preview.
Immuno-Gold Electron Microscopy in Virus Diagnosis and Research, 349–376.
Studer, D., Michel, M., & Muller, M. (1989). High-pressure freezing comes of age. Scanning Microscopy
(S3), 253–269.
Studer, D., Hennecke, H., & Muller, M. (1992). High-pressure freezing of soybean nodules leads to an
improved preservation of ultrastructure. Planta, 188, 155–163.
Studer, D., Michel, M., Wohlwend, M., Hunziker, E. B., & Buschmann, M. D. (1995). Vitrification of articu-
lar cartilage by high-pressure freezing. Journal of Microscopy, 179, 321–332.
Szczesny, P. J., Walther, P., & Muller, M. (1996). Light damage in rod outer segments: the effects of fixation
on ultrastructural alterations. Current Eye Research, 15, 807–814.
Tiedemann, J., Hohenberg, H., & Kollmann, R. (1998). High-pressure freezing of plant cells cultured in
cellulose microcapillaries. Journal of Microscopy, 189, 163–171.
Trump, B. F., Smuckler, E. A., & Benditt, E. P. (1961). A method for staining epoxy sections for light micros-
copy. Journal of Ultrastructure Research, 5, 343–348.
Verkleij, A. J., Humbel, B., Studer, D., & Muller, M. (1985). Lipidic particle-systems as visualized by thin-
section electron-microscopy. Biochimica et Biophysica Acta, 812, 591–594.
Vielhaber, G., Pfeiffer, S., Brade, L., Lindner, B., Goldmann, T., Vollmer, E., et al. (2001). Localization of
ceramide and glucosylceramide in human epidermis by immunogold electron microscopy. Journal of
Investigative Dermatology, 117, 1126–1136.
Vielhaber, G., Brade, L., Lindner, B., Pfeiffer, S., Wepf, R., Hintze, U., et al. (2001). Mouse anti-ceramide
antiserum: a specific tool for the detection of endogenous ceramide. Glycobiology, 11, 451–457.
Walter, T., Shattuck, D. W., Baldock, R., Bastin, M. E., Carpenter, A. E., Duce, S., et al. (2010). Visualization
of image data from cells to organisms. Nature of Methods, 7, S26–S41.
Walther, P., & Ziegler, A. (2002). Freeze substitution of high-pressure frozen samples: the visibility of bio-
logical membranes is improved when the substitution medium contains water. Journal of Microscopy,
208, 3–10.
De Winter, D. A., Schneijdenberg, C. T., Lebbink, M. N., Lich, B., Verkleij, A. J., Drury, M. R., et al. (2009).
Tomography of insulating biological and geological materials using focused ion beam (FIB) sectioning
and low-kV BSE imaging. Journal of Microscopy, 233, 372–383.
Young, R. J., Dingle, T., Robinson, K., & Pugh, P. J. A. (1993). An application of scanned focused ion-beam
milling to studies on the internal morphology of small arthropods. Journal Microscopy-Oxford, 172,
81–88.
CHAPTER 18
Correlative Light and Volume Electron

Microscopy: Using Focused Ion Beam
Scanning Electron Microscopy to Image
Transient Events in Model Organisms
Andrew J. Bushby*, Giovanni Mariggi†, Hannah E. J. Armer‡
and Lucy M. Collinson§
* The
Nanovision Centre, School of Engineering and Materials Science, Queen Mary University of London,
London, UK
† Vascular Biology Laboratory, London Research Institute, Cancer Research UK, London, UK
‡ Imaging Suite, Institute of Ophthalmology, University College London, London, UK
§ Electron Microscopy Unit, London Research Institute, Cancer Research UK, London, UK
Abstract
I. Introduction
II. Rationale
III. Materials
A. Equipment
B. Materials
IV. Methods
A. Video Microscopy
B. Sample Preparation for FIB/SEM
C. FIB/SEM Data Collection
D. Image Analysis
V. Discussion
VI. Summary
Acknowledgments
References

358 Andrew J. Bushby et al.
Abstract
The study of a biological event within a live model organism has become rou-
tine through the use of fluorescent labeling of specific proteins in conjunction with
laser confocal imaging. These methods allow 3D visualization of temporal events that
can elucidate biological function but cannot resolve the tissue organization, extracel-
lular and subcellular details of the tissues. Here, we present a method for correlat-
ing electron microscopy image data with the light microscopy data from the same
sample volume to reveal the 3D structural information: “correlative light and volume
electron microscopy.” The methods for live video confocal microscopy, fixation and
embedding of the tissue for electron microscopy, the focused ion beam scanning elec-
tron microscopy method for sequentially slicing and imaging the volume of interest,
and the treatment of the resulting 3D dataset are presented. The method is illustrated
with data collected during the angiogenesis of blood vessels in a transgenic zebrafish
embryo.
I. Introduction
Complex networks are found in all organisms, across many different scales. From
the endoplasmic reticulum and the Golgi apparatus that are responsible for manufactur-
ing the molecules of life in individual cells, to the vascular systems that supply blood
to tissues in vertebrates. From subcellular factories where viruses hide from the host
immune system and replicate, to the neural networks that control movement, pain,
and consciousness. Until now, the study of these networks has been hampered by our
inability to image across scales, from individual subcellular processes through to the
whole organisms in which they occur.
Recent innovations in electron microscopy (EM) have led to a paradigm shift in
high-resolution imaging of large volumes of biological samples. These techniques
exploit different sectioning methods to cut and image slices of material automatically,
thus minimizing the artifacts associated with manual serial sectioning, speeding up the
process, and freeing the operator to perform other work. Serial block face/scanning
electron microscopy (SBF/SEM) (Armer et al., 2009; Briggman & Denk, 2006; Denk
& Horstmann, 2004; Mun, Jeong, Kim, Han, & Kim, 2011; O’Connell et al., 2008)
and focused ion beam/scanning electron microscopy (FIB/SEM) (Armer et al., 2009;
Bushby et al., 2011; De Winter et al., 2009; Heymann et al., 2009; Knott, Marchman,
Wall, & Lich, 2008; Schmidt et al., 2011; Schneider, Meier, Wepf, & Muller, 2011)
allow the automated removal of a section of the sample in situ in the SEM chamber, fol-
lowed by high-resolution electron imaging of the resulting block surface. Cutting and
imaging are repeated sequentially to automatically collect a stack of images through
the sample. In the SBF/SEM, a slice of material is removed using a diamond knife held
in a miniaturized ultramicrotome attached to the chamber door. In the FIB/SEM, cut-
ting is performed by a gallium ion beam, which is inherently destructive and can mill
away thin slices of material from the block face. The revealed surface is imaged using
18. Correlative light and FIB/SEM 359
a backscattered electron (BSE) detector. Collectively, we call these in situ SEM serial
imaging techniques “volume electron microscopy” (volume EM) (Armer et al., 2009).
We have proven the principle of correlative light and volume EM (CLVEM) in the
studies of the highly complex 3D network that forms the developing circulatory system
in zebrafish (Armer et al., 2009). Here, traditional light and electron microscopy cannot
resolve the conflict between volume imaging of a large branching blood vessel system
in vivo and high resolution ultrastructural imaging of vessel formation. The formation
of new blood vessels, angiogenesis, is crucial in the patterning of the vascular sys-
tem during vertebrate embryonic development in normal physiology and in pathologi-
cal settings such as chronic inflammation, tumor progression, and metastasis. During
angiogenesis, endothelial cells (ECs) extend long filopodia to sense their environment
and display a highly migratory behavior toward morphogen gradients (Risau, 1997).
Filopodia from two ECs must make contact and connect, an event known as anasto-
mosis, giving rise to lumenized vessels able to carry blood flow. Anastomosis is a key
step that has received little attention but is critical for the formation of new functional
vessels and could yield novel specific targets for the inhibition of angiogenesis during
tumor formation.
Live confocal microscopy can be used to image anastomosis in the Tg( f li1:EGFP)y1
transgenic zebrafish embryo, where ECs express green fluorescent protein (GFP)
(Isogai et al., 2003; Lawson & Weinstein, 2002). We correlated this with FIB/SEM to
locate and image a transient blood vessel fusion event of only ∼3 µm3 in a total tissue
volume of ∼12,600,000 µm3 (Armer et al., 2009). This CLVEM technique could be
applied to any model organism, tissue, or cell, where small areas of interest need to be
located within a large integrated network.
II. Rationale
Fluorescence microscopy is a powerful tool for localizing proteins within cells and
tissues, giving information as to their function in biological processes. Although tradi-
tional brightfield light microscopes are limited in resolution by the wavelength of light,
super-resolution techniques have recently broken this limit while imaging proteins
tagged with fluorescent markers in thin samples (PALM, STORM, OMX) (Schermel-
leh, Heintzmann, & Leonhardt, 2010). However, as with any fluorescence microscopy
technique, information is limited to the distribution of the tagged protein and tells us
little about the ultrastructure of the surrounding cells and tissues, which may be inti-
mately involved in the biological process under study. Larger specimens can be imaged
using newly developed optical 3D volume techniques based on light-sheet microscopy
(Keller & Stelzer, 2008, 2010), but again with limited resolution.
Due to the properties of the electron beam, transmission electron microscopy (TEM)
of resin-embedded biological samples offers improved lateral resolution in the order of
1–5 nm. However, the axial resolution is limited by the poor material penetration of the
electron beam to around 80 nm (at an acceleration voltage of 120 kV). Until recently,
the only way to image through larger volumes at nanometer resolution has been to
examine the tissue slice-by-slice using a technique called serial-section TEM (ssTEM)
(White, Southgate, Thomson, Brenner, 1986), where each section is produced, col-
lected, and imaged in order. ssTEM is prone to errors and artifacts including missing
sections, tears or folds in sections, stain precipitation, section rotation, grid rotation,
and section shrinkage or expansion. Realignment of images postacquisition is chal-
lenging and leads to a reduction in the final field of view (FOV). In addition, the larger
the section, the harder it is to form a stable ribbon, and the fewer sections can be fitted
onto each TEM grid leading to a greater chance of artifacts. It is, therefore, preferable
to limit the block surface area, and thus the FOV, to less than about 500 × 300 µm2.
Considering that even a single cell of 10 µm depth would require ∼140 sections to be
cut, imaged, and digitally reconstructed, an entire tissue or model organism would
require thousands of serial images, taking months or years of effort (White et al., 1986).
Correlative light and electron microscopy (CLEM) combines the benefits of both
techniques. The fluorescent tag is localized, imaged, and marked in some manner, and
tracked through sample preparation for TEM. The surface area of the block can thus
be minimized by trimming closely around the volume of interest (VOI), but serial
sectioning is still required as location of the VOI within the depth of the sample is
less certain due to the limited axial resolution of the light microscope. Correlative
light and FIB/SEM extend the benefits of CLEM to whole cells, tissues, and model
organisms by automating the serial imaging process. The technique results in large,
registered 3D cubes of data targeted to specific VOIs that can be virtually sliced in
any orientation, revealing the overall architecture of cells and tissues as well as sub-
cellular detail at nanometer resolution. As the entire cutting and imaging process is
performed within the SEM chamber and on the block face, many of the artifacts
associated with serial sectioning are avoided. However, it is worth noting that vol-
ume EM techniques are destructive, and that the sections (SBF/SEM) or sputtered
material (FIB/SEM) are lost and cannot be reimaged at a later date. In addition, each
technique has associated artifacts, which need to be properly characterized and taken
into account while processing and interpreting the data (Bassim et al., 2011; Bushby
et al., 2011; Drobne, Milani, Leser, & Tatti, 2007; Heymann et al., 2006; Munroe,
2009; Prenitzer et al., 2003).
III. Materials
A. Equipment
• Glass knifemaker EM KMR3 (Leica Microsystems, Vienna, Austria)
• Leica ultramicrotome EM UC7 (Leica Microsystems, Vienna, Austria)
• Sputter-carbon coating unit Q150 EM (Quorum Technologies, West Sussex, UK)
• Zeiss Axiovert 200 M fitted with a LSM 5 Pascal system and 488 nm laser emission
supplied by an Argon laser (Carl Zeiss MicroImaging GmbH, Jena, Germany)
• Tecnai Spirit Biotwin 120 keV TEM (FEI Company, Eindhoven, the Netherlands)
• Orius CCD and Digital Micrograph software (GATAN Inc., California, USA)
• Focused ion beam/scanning electron microscope (FIB/SEM) with a field emission

electron source, secondary electron imaging (SEI, Everhart–Thornley Detector),
a BSE detector, a platinum (Pt) gas injection system, and a liquid gallium metal
source
• Workstation for 3D reconstruction with 96GB RAM, NVidia Quadro 6000 graphics
card and 2TB data storage
• Wacom Cintiq 21UX interactive pen display (Wacom Technology Corp.,
Vancouver, USA)
B. Materials
• Tg( f li1:eGFP)y1 zebrafish embryos (Lawson & Weinstein, 2002)
• Formaldehyde EM (TAAB Laboratories Equipment Ltd., Berkshire, UK)
• Silver paint (Agar Scientific)
• Low melting point agarose (Sigma Aldrich)
• Tricaine (Sigma Aldrich)
• Glass coverslip-bottom culture dishes (MatTek Corp., Ashland, USA)
• Glass for glass knives (Agar Scientific, Stansted, UK)
• Cryotrim 90° diamond knife (Diatome, Biel, Switzerland)
• Ultra 45° room temperature diamond knife (Diatome, Biel, Switzerland)
• Razor blades, single edge heavy duty carbon steel (Agar Scientific, Stansted, UK)
• FIB/SEM embedding molds (TAAB Laboratories Equipment Ltd., Berkshire, UK)
• SEM specimen stubs appropriate to FIB/SEM (Agar Scientific, Stansted, UK)
• Leit-C conducting carbon cement (Agar Scientific, Stansted, UK)
• Glutaraldehyde (TAAB Laboratories Equipment Ltd., Berkshire, UK)
• Sodium cacodylate (TAAB Laboratories Equipment Ltd., Berkshire, UK)
• Osmium tetroxide (TAAB Laboratories Equipment Ltd., Berkshire, UK)
• Potassium ferricyanide (TAAB Laboratories Equipment Ltd., Berkshire, UK)
• Ethanol series (Sigma Aldrich)
• Acetone (Sigma Aldrich)
• Durcupan ACM Epoxy Kit (TAAB Laboratories Equipment Ltd., Berkshire, UK)
• Uranyl acetate (Agar Scientific, Stansted, UK)
IV. Methods
A. Video Microscopy
1. Principle
The zebrafish has become part of the basic toolbox for the study of vascular biology
and angiogenesis, thanks to its optical clarity and cell type–specific fluorescent protein
expression, allowing for live cell imaging. Most of the work has concentrated on study-
ing the sprouting of intersegmental vessels (ISVs) that grow in the trunk of the embryo,
starting from 24 h postfertilization (hpf). One part of the angiogenic process is anasto-
mosis, the fusion of two sprouts to form a functional vessel loop, taking about 8 h from
initial sprouting for fusion to occur. The sites of anastomosis can be found near the top
of the neural tube at the level of the middle of the somites. The exact location cannot
be predicted nor can the time when anastomosis occurs, even though there is a rostral
to caudal direction to the process. Live confocal microscopy can be used to image this
process in Tg(f li1:eGFP)y1 embryos (Lawson & Weinstein, 2002) with reasonable reso-
lution, but not giving either subcellular ultrastructure or detail of the surrounding tissue.
Therefore, measurements are taken to record the initial site of contact between the ECs,
formed by thin filopodial structures, for subsequent imaging using volume EM.
2. Protocol
1. Maintain zebrafish in standard conditions (28°C, 14/10 h light/dark cycle). Fish
are fed daily, twice prior to mating. Embryos at 28 hpf are anesthetized using 1×
tricaine and dechorionated. Immobilize the embryos in a coverslip dish (MatTek
Corp.) using 0.2% low melting point agarose at 37°C, taking care to place the
embryo on its side to allow visualization of the ISVs. At this developmental stage,
ISVs lie within a tissue volume of ∼100 µm × 100 µm × 40 µm.
2. Live confocal microscopy is used to image blood vessel growth and development in
the Tg( f li1: EGFP)y1 transgenic zebrafish embryo (Lawson & Weinstein, 2002) in
which ECs express GFP (Fig. 1(A)). During angiogenesis, ECs from adjacent ISVs
extend long filopodia, which eventually make contact and fuse in an event known
as anastomosis (Fig. 1(B–F)). Throughout the imaging process, embryos are main-
tained at a constant 28.0°C in an environmental chamber. Time-lapse microscopy is
carried out on an inverted Zeiss Axiovert 200 M fitted with an LSM 5 Pascal system.
Images are captured using a 488 nm laser and 510 nm BP filter, until the filopodia of
adjacent ISVs connect (Fig. 1, arrows), at which point the embryos are imaged with
the 10× and 40× objectives. Embryos are then removed from the agarose and fixed
for 1 h in 4% formaldehyde (FA) in 0.1 M phosphate buffer (pH 7.4). Glutaralde-
hyde is not used in this step as it can quench the GFP signal.
3. A confocal stack of images is then collected through the depth of the VOI (Fig.
2(A)). A 2D projection of the fluorescence data is overlaid onto a brightfield image
of the zebrafish (Fig. 2(B)). Distance and angular measurements are taken to aid
correlation between light and electron microscopy in the VOI containing the
anastomosing blood vessels with the Zeiss LSM image viewer software. Gross
morphological markers including the yolk sac (YS), yolk tube, and undulations
along the back of the zebrafish embryo are used as landmarks (Fig. 2(B)). These
landmarks must be visible once the zebrafish has been osmicated and embedded,
at which point the internal structure will be masked.
B. Sample Preparation for FIB/SEM

1. Principle
In general, sample preparation for volume EM is similar to that for traditional TEM.
In FIB/SEM, the image contrast is formed by the backscattered electrons (BSEs),
which correlate with atomic number within the sample. Advances in sample preparation
Fig. 1 Live confocal imaging of angiogenesis in zebrafish embryos. (A) A Tg( f li1: EGFP) y1 transgenic zebrafish embryo (Lawson
& Weinstein, 2002) expressing green fluorescent protein in the endothelial cells of the developing blood vessels. Intersegmental vessels
(ISV) arise alternately from the dorsal aorta (DA) and the posterior cardinal vein (PCV). The tip cells of the intersegmental vessels have
long thin filopodia which project and retract (B, arrows) until they encounter an adjacent tip cell filopodium (C), at which point they
make contact and fuse in a process called anastomosis (D), to form the dorsal longitudinal anastomotic vessel (DLAV; E, F).
Fig. 2 Recording the position of the VOI in the light microscope. (A) After fixation, a confocal z-stack is acquired in the spinning
disc microscope. (B) A projection of the image stack is overlaid onto a brightfield image of the zebrafish embryo. Distance and angular
measurements are taken from the VOI to gross morphological markers including the YS and the yolk tube. ISV – intersegmental vessels,
DA – dorsal aorta, PCV – posterior cardinal vein.
techniques have concentrated on introducing as much heavy metal into the biological
sample as possible, which has the additional advantage of reducing charging at the block
face during electron imaging. The protocol we describe here uses two osmium steps as
well as an en bloc uranyl acetate stain (Knott, Rosset, & Cantoni, 2011). Initial tests on
resin compatibility for ion beam milling led us to use Durcupan, a hard epoxy resin, for
FIB/SEM. Softer, more hydrophilic resins such as Lowicryls performed less well.
2. Protocol
Fixation and Embedding of the Specimen
1. The zebrafish embryos are fixed again in 2% FA/1.5% glutaraldehyde in 0.1 M
phosphate buffer (pH 7.4) for 1 h at room temperature, followed by secondary
fixation in 1.5% (w/v) potassium ferricyanide/1% (w/v) osmium tetroxide in
0.1 M sodium cacodylate (pH7.4) for 1 h. Samples are further stained with 1%
(w/v) osmium tetroxide in 0.1 M sodium cacodylate for 30 min, followed by an
en bloc treatment with 1% (w/v) uranyl acetate in double-distilled water for 30 min.
Samples are then rinsed in water, dehydrated through a graded ethanol series
(2 × 70%, 2 × 90%, 2 × 100% for 10 min each change), and transferred to acetone
(2 × 10 min). Samples are then passed through a series of Durcupan:acetone (1:2
for 1 h, 1:1 for 1 h, 2:1 for 1 h), followed by 100% Durcupan for 2 h. Samples are
placed into specially designed FIB/SEM molds filled with fresh 100% Durcupan
as described in detail below, and polymerized overnight at 60°C.
2. Preparation of good blocks is critical to the success of CLVEM. We designed

special molds specifically for this purpose (FIB/SEM molds, TAAB, Fig. 3(A)).
The zebrafish is placed in the bottom corner of the mold so that, once embedded,
it lies in an overhang of resin, which allows sputtered material to collect under the
sample during milling (rather than on the imaging face) (Fig. 3(C)). The sample
is placed in the preferred orientation, in this case with the tail parallel to the front
face of the block (Fig. 3(D and E)). This is determined during preliminary runs to
maximize data collection in the region of interest, while minimizing milling. All
faces of the resulting block are perpendicular or parallel (Fig. 3(B and C; arrows)),
which speeds up the trimming process and allows the block to be mounted onto
the stub in the correct orientation.
Preparation of the block face and mounting the block for FIB/SEM
3. Final preparation of the resin block is important for successful 3D image data collec-
tion. The block is placed into an ultramicrotome and the front face is initially trimmed
with a razor blade around the VOI, in this case the tail of the zebrafish embryo (Fig.
3(D)). The front face of the block is then polished at 1.0 mm/s cutting speed and
200 nm advance using a glass knife, as close to the sample as possible without cutting
into the tissue. At this point, the glass knife is exchanged for a diamond knife and the
front face is polished at 0.8 mm/s cutting speed and 100 nm advance until the first hint
of the tissue is seen in the sections. Trimming the block close to the VOI is important
in saving FIB/SEM time while locating the region of interest within the resin. If the
VOI is not immediately below the surface of the tissue then 70 nm sections can be
collected for examination in the TEM, which can aid in relocalization of the VOI.
However, great care must be taken not to cut too close to the VOI, as some further
polishing will take place in the FIB/SEM prior to data collection.
4. An eyepiece graticule is used to assess the front face of the block, cross-refer-
encing with the original light microscope measurements, and the block is further
trimmed as small as possible around the VOI so that all edges are parallel or
perpendicular using a 90° diamond trimming knife (Diatome). All edges must be
sharp and all faces must be smooth (Fig. 3(E)).
5. Transfer the block to a vice and using a fine hacksaw remove excess resin from
the top and back surfaces to reduce the volume of nonconductive material and
hence minimize charging in the SEM. Do not trim the base of the block as this is
already flat and parallel to the resin overhang. Take care to avoid contamination of
the block. The nature of the polished block surface and the edges of the imaging
face has an important influence on subsequent Pt coating, ion beam milling, and,
in turn, the image quality. Therefore, these need to be kept defect- and debris-free
prior to collection of the image series (Fig. 4).
6. Glue the flat base of the block to the surface of an aluminum SEM specimen stub
(appropriate to the FIB/SEM being used) using Leit-C carbon cement such that
the polished face is perpendicular to the stub surface. The rigidity of the speci-
men mounting is important, as any mechanical drift of the specimen will create
artifacts or jumps that degrade the dataset. Silver paste may also be used and sets
Fig. 3 Embedding and trimming of the sample for FIB/SEM. (A) Specially designed molds are used to embed the sample for FIB/
SEM. The zebrafish embryo is placed in the bottom corner of the mold so that the VOI, in this case the tail, lies against the front face of
the block. This is shown from underneath (B) and from the side (C). Arrows indicate parallel faces. The resin overhang is designed so
that sputtered material collects away from the front face during milling. (D) The front face is trimmed close to the sample using a razor
blade and polished using a glass knife in an ultramicrotome. Using the measurements from the light microscope, the front face is further
trimmed and polished as close as possible to the VOI using a 90° diamond trimming knife (E). yolk sac (YS).
Fig. 4 Secondary electron image (SEI) of the block trimmed close to the VOI.
more rapidly but carbon cement has been found to give a more rigid bond and so
is the material of choice.
7. Sculpt the carbon cement high up the sides of the block to construct conductive
paths from the top surface of the block to the aluminum stub. Allow the cement
to dry thoroughly, preferably overnight. The conductive paths should be created
as uniformly as possible by sculpting carbon cement up the block to just below
its top edges so that charge dissipation is uniform. However, care must be taken
that the cement does not contaminate the top or front face of the block. The stub-
mounted block can be stored in a stub box over desiccant until coating.
8. Carbon coat the block in a standard laboratory vacuum evaporation unit, such as a
Balzers CED030 or Leica EMSCD050, or any other carbon evaporator designed
for SEM specimen preparation.
C. FIB/SEM Data Collection

1. Principle
The aim is now to find the same features identified in the light microscope and
image the VOI at EM resolution. This is not as easy as perhaps it sounds, since the
scale and resolution of the features are tiny compared to the size of the block of tissue,
the imaging contrast is now quite different, and the FIB/SEM process is, by its nature,
destructive so that care must be taken not to inadvertently destroy the features without
recording them. It can be appreciated that this is rather like looking for the proverbial
needle in a haystack, and so some prior knowledge of the anatomy of the tissue is desir-
able before commencing the experiment, either from a preexisting atlas of anatomy or
from a previous attempt at the same experiment.
There are several FIB/SEM instruments commercially available, any of which
would be suitable for this process. A field emission electron source is highly desirable
to provide an intense electron beam but must be capable of delivering high current.
Most modern instruments use a Schottky field emitter [hot field emission gun (FEG)]
that is more than adequate. A BSE detector is required. This is typically a four-segment
annular solid-state detector mounted below the pole-piece of the electron column,
inside the sample chamber. Alternatively, some instruments have an in-lens detector
system that allows for shorter working distances to be used.
The FIB column typically uses a liquid gallium metal source, although other sources
are now becoming available, and operates at 30 kV accelerating voltage. The ion beam
current that is available is an important factor. High ion currents (64 nA or greater) pro-
vide much faster rough milling to rapidly remove material. This can be useful in cutting
away excess resin to approach the VOI. The position of the ion beam column relative
to the electron beam column is determined by the design of the instrument. The angle
between the two beams is typically around 50° but some instruments are now being
designed with the beams at 90° to obviate the need for tilting the sample. The conver-
gence point of the two beams dictates the working distance. A short working distance is
desirable for improved resolution but can impose restriction of sample size when high
tilt angles are required (Fig. 5(A)).
Fig. 5 The focused ion beam scanning electron microscope. (A) Typical FIB/SEM instrument showing the
positioning of the major components. (B) Geometrical arrangement of the sample block, BSE detector, and ion
and electron beams.
The key stages in the process are preparing and mounting the tissue block, setting
the block at the eucentric position in the microscope, identifying the approximate posi-
tion of the VOI, rough cutting the block until some recognizable anatomical features
are revealed, and then setting up the image collection run. The image collection relies
on BSE imaging to reveal the subcellular and extracellular features, which in turn is
dependent on successful tissue staining during the preparation described above.
2. Protocol
Preparation of the block face in the FIB/SEM
The sample is usually tilted so that the ion beam is perpendicular to the top surface
of the block and hence the image face is viewed by the electron beam at an apparent
tilt angle of ∼40° (Fig. 5(B)). It is essential therefore that the front face of the block is
positioned with the convergence of the two beams at the eucentric position of the stage
so that the milling and imaging positions are the same when the specimen is tilted.
The eucentric position is dictated by the design of the microscope but is typically at
a working distance of 10 mm. The position is found by the following procedures:
1. Focus the electron beam in secondary electron imaging (SEI) mode and identify
a feature on the surface at about 5,000×. Set the cross hairs on the feature.
2. Tilt to 5°. If the feature moves, return it to the original position by raising the
stage (Z-height) using the manual stage controls (on the microscope column, see
Fig. 5(A)).
3. Return to zero tilt and check the position of the feature.
4. Tilt to 15°. Adjust the Z-height to return the feature to the original position.
5. Tilt to 52° (or similar so that the ion beam is perpendicular to the top face of the
block depending on the angle between the electron beam and the ion beam on the
instrument in question). Adjust the Z-height to return the feature to the original
position. At high tilt angles, it is possible that the feature may appear differently
in the image.
6. Check the eucentricity by returning to zero tilt and ensuring the feature is still at
the centre of the cross-wires. If necessary, repeat steps 2–6 again.
7. Tilt to 52°. If a new region of interest is selected or the sample is removed and
replaced in the microscope chamber, then the procedure (1–7) will need to be
repeated.
8. Imaging the tilted block in BSE mode, identify any features of anatomy visible
on the block surface (Fig. 6(A and B)).
9. Identify the approximate position of the VOI within the block by taking mea-
surements from the edges of the block or from any features visible on the sur-
face.
10. Having identified the probable position of the VOI, return the stage to the 0° tilt
position. The top surface of the block can now be coated with a dense layer of
Pt. This protects the milled edge of the block face and helps to reduce curtain-
ing artifacts on the image face. If an in-chamber BSE detector is being used, it
should be removed from the chamber during Pt deposition, replacing it after
rough milling of the block face and repeating steps 1–7 above.
Fig. 6 Relocating the VOI in the FIB/SEM. (A) The trimmed and polished block is mounted on a stub and viewed using the BSE
detector. The head and YS of the zebrafish can be seen due to the heavy metal staining of the tissue. The front face has been trimmed so
that the scales on the tail are just visible. (B) When the side of the resin overhang is imaged, a lateral view of the YS and trunk is exposed,
giving further landmarks for relocation of the VOI. (C) A 1 µm thick layer of platinum is deposited above the tissue to protect the top
edge during milling, and 150 nm thick slices are milled into the tissue to expose the muscle blocks. (D) Light microscopy measurements
are compared with the block and exposed tissue to determine the location of the VOI. (E) Trenches are placed on either side of this
region to aid in removal of sputtered material from the front face of the block. (F) The image is focused and positioned for imaging, and
the top face measured to determine the possible run length.
11. Warm up the Pt deposition gas injector and insert into the column. Select an
area on the top face of the block that covers all of the probable area above the
VOI and deposit 1 µm of dense Pt from the gas injection system (Fig. 6(C)).
The metal-organic precursor vapor is injected into the microscope chamber
close to the sample surface and the ion beam used to break down the vapor
and deposit Pt on the surface in a defined area scanned by the ion beam. The
ion current required depends on the area over which Pt is to be deposited. As
a rule-of-thumb, 5–10 pA/µm2 is required. The Pt will also tend to deposit
down the front (imaging) face of the block but this is easily removed during
ion milling.
12. The front face of the block should now be milled away and polished to reveal
the underlying tissue. If the VOI is close to the front face of the block, great care
should be taken in this initial milling step and low ion beam currents should be
used. Otherwise, if there is a large amount of resin or tissue to remove, the high-
est ion beam current available should be used to remove material quickly. At
the highest ion current, there is no aperture in the ion column and this saves the
aperture strip from erosion. The milled area should be considerably wider than
the expected VOI, in this case ∼400 µm in width (Fig. 6(C)).
13. With the tissue now exposed, the light microscopy measurements should be
overlaid onto images of the front and top faces of the block, taking into account
any resin already trimmed away manually in the ultramicrotome. A decision
must be made at this point on the size of the front face to be milled, and a balance
achieved between surface area to be milled and time taken to mill each slice,
particularly with respect to the overall run time. In this case, a large FOV of
∼300 µm × 200 µm was chosen, revealing at least four somites (muscle blocks).
Trenches need to be cut down each side of the VOI to prevent sputtered material
from redepositing onto the front face of the block (Fig. 6(D and E)).
14. Polish the front face of the block at successively lower ion beam currents (e.g.,
30 nA, 15 nA, and 5 nA) to produce a smooth final block face in the region where
the VOI is expected.
15. Image the block face in BSE mode and focus the features at the front face of the
block. In addition, check how many slices can be milled and imaged before the
front face moves out of the FOV by measuring the top face of the block (Fig.
6(F)). Before starting the collection of the full dataset, it may be necessary to
begin collecting data with relatively thick slices between image frames to help
recognize further features within the tissue block. This can be done by setting up
the instruments’ automated tomography software (e.g., “Slice & View” on FEI
instruments, “Nanotomography” on Carl Zeiss instruments, etc.).
16. Start the automated tomography software on the instrument and select an area
that is considerably wider on the block face than the intended imaging area and
select the number of slices to give a slice thickness of about 150 nm (Fig. 7(C)).
Start the run in front of the current image face so that the new surface will be
well polished and there is no chance of an under-cut (milling away behind the
current image face).
Fig. 7 Schematic view of top face of the block: (A) resin block, (B) Pt deposited area, (C) initial investiga-
tive slices, (D) trenches, (E) serial imaging slices, and (F) VOI.
17. Set the ion beam current to 15 nA at maximum accelerating voltage (usually
30 kV). The highest accelerating voltage is suitable because the ion beam is at its
finest spot size and most destructive, which are both highly desirable features for
FIB/SEM tomography applications. Focus the beam on the Pt deposit or an area
close to but separate from the image face. The beam must be well-focused and
free from astigmatism at the beam current that will be used for slicing. However,
imaging at this beam current will be destructive. The Pt deposit will erode more
slowly than the resin block.
18. The electron imaging conditions can now be optimized. Sufficient electron beam
current is required to give a detectable intensity of BSEs. This typically requires
an electron beam current greater than 1 nA and sometimes as high as 4 nA. The
image resolution is determined by the accelerating voltage of the electron beam,
lower accelerating voltages giving increased resolution. Most solid-state detec-
tors do not detect BSEs at accelerating voltages less than 5 kV. Specialist low kV
detectors and in-lens systems are now capable of detecting BSEs at accelerating
voltages as low as 1 kV. These detectors will provide a significant improvement in
image resolution, since at such low voltages only the heavy metal stain elements
in the sample cause significant backscattering of electrons so that the image con-
trast represents the distribution of stain in the sample. In-lens detectors usually
allow some energy filtering of the electrons by rejecting low energy secondary
electrons. A similar effect can be achieved for in-chamber detectors by simultane-
ously operating the Everhart–Thornley detector to attract the secondary electrons.
19. As soon as it is clear where the imaged surface lies in relation to the VOI, it is
possible to begin a full data collection run.
4. Image Data Collection: Imaging Conditions and Setup for Serial Imaging
The image data collection process can now begin in earnest by setting up the tomog-
raphy software and imaging conditions that will collect data throughout the VOI. This
requires consideration of image collection strategy and microscopic conditions. The
process will generate many hundreds of images, possibly more than a thousand, so it
is important that there is enough disc space available to store all of the images. Also, a
decision must be made on the FOV and image resolution. The larger the FOV, the more
likely that the VOI will be present within the data, but the lower the pixel resolution of
the images.
20. Focus the ion beam at the milling current, typically 15 nA. In the tomography
software, select an area on the top surface of the block that you expect to include
all of the VOI, starting just in front of the current imaging face. The milling area
should in any case be made somewhat wider than the image area so that any
redeposition of sputtered materials or electron shadowing from the sides of
the trench does not affect the image area (Fig. 7(E)). The position of the VOI
and its orientation in the tissue block may not be known to better than ±20 µm
(Fig. 7(F)).
21. In choosing imaging conditions, the field width (in pixels), dwell time per
pixel, and slice thickness are the important parameters. The pixel size does
not need to be smaller than the resolution and so determines the pixel width
of the image field. The longer the dwell time at each pixel, the greater the
signal-to-noise ratio and improved image contrast. However, longer dwell
time also increases the image frame collection time and the possibility of
charging artifacts. There is some advantage in keeping the slice thickness
the same as the pixel size so that the voxels are cubic. This can aid in feature
recognition and accurate partitioning of the 3D image dataset, since it is
possible to interrogate the data in any direction without distortion (Bushby
et al., 2011).
22. The imaging conditions can now be optimized. Fine focus the electron beam
in SEI mode and manually adjust the stage Z-height so that the FOV fills the
electron image and the Pt-coated surface is toward the top of the electron image.
Viewing at a tilt angle, the image will appear to move up as the “slice and view”
process progresses. The VOI should be centered on the screen, then lock all
electronic stage controls.
23. Set the contrast and brightness so that as much as possible of the gray scale range
is used; white for heavily stained tissue material and black for resin containing
no tissue. However, the white levels should not oversaturate causing the white to
bleed into neighboring pixels.
24. Start the automated tomography process. Begin a few slices in front of the
block face and observe the process until the ion beam begins to cut the
block face. Monitor the first few slices to assess if the position of the slice is
correct. The contrast and brightness of the image should also be checked and
may even be fine-tuned without stopping the auto “slice and view” run. It is
also possible to refocus and realign the electron beam, but if the ion beam
setting is wrong or if it is milling the wrong position, then it is better to abort
the run, fix the problem, and restart the run. A typical sequence of images is
shown in Fig. 8.
Fig. 8 Sequential images from a FIB/SEM serial imaging run with a slice thickness of 51nm. (A–D) The endothelial cells of the
intersegmental vessel (ISV) can be seen arising from the dorsal aorta (DA). Resolution is sufficient to see subcellular detail. M––muscle
block, PCV––posterior cardinal vein. Scale bar: 50 µm.
D. Image Analysis
1. Principle
One of the major advantages of automatic serial imaging using FIB/SEM is that
the raw images are nearly aligned. Small shifts of a few pixels between images may
be present, and occasionally larger shifts, often due to the environmental conditions
including thermal drift and acoustic or mechanical vibration. It is therefore advisable
to perform a fine alignment of the images after data collection. Following alignment,
the volume can be viewed in a variety of ways using free and/or commercial 3D visu-
alization packages.
It is then necessary to locate and extract the VOI from the rest of the data. The
volume of data is effectively opaque, that is, the slice being viewed forms a surface,
which masks the data underneath. In order to view a 3D structure within the data, it
must be highlighted in a process called segmentation. In light microscopy and medical
imaging, images tend to contain either dark pixels (no tissue) or bright pixels (tissue),
and so computer algorithms have been developed to recognize and segment the images
automatically using thresholding. “Tissue” pixels are kept whilst “no tissue” pixels are
made transparent so that the 3D tissue structure is revealed.
Due to the range of gray levels within an EM image, it is extremely difficult to
automatically segment specific structures using thresholding. Exceptions to this occur
when a structure is electron-dense, either due to preferential heavy metal staining dur-
ing sample preparation (e.g., myelin sheaths surrounding axons, which are highly
osmiophilic) or where a specific stain has been introduced to highlight a structure [e.g.,
biocytin staining of dendritic trees (Kubota et al., 2011) or photoconversion of the
genetically encoded tag miniSOG (Shu et al., 2011)]. Recently, new algorithms have
been developed for semiautomatic or automatic segmentation of the tissue, again con-
centrated in the field of neurobiology. These have used Kalman Snakes (Jurrus et al.,
2009) or contour propagation (Macke et al., 2008) to segment axons and automatic
segmentation of synapses using machine learning (Kreshuk et al., 2011).
However, other non-neuronal tissues have different characteristics and neuronal seg-
mentation algorithms may not be applicable. The ultrastructure of non-neuronal VOIs,
their location with the cell or tissue in relation to other structures, and their appearance in
2D slices compared with their entire 3D structure is often unknown. For this reason, and in
most cases, the data must be manually segmented. The process of alignment, manual seg-
mentation, 3D visualization, and analysis of the structure of interest are discussed further.
2. Software
There are many software packages available for image alignment, segmentation, 3D
rendering, visualization and analysis. Freeware packages include the following:
• Fiji (Fiji Is Just ImageJ)––a distribution of ImageJ optimized for life sciences
(http://fiji.sc/wiki/index.php/Fiji)
• IMOD––a set of image processing, modeling, and display programs used for tomo-
graphic reconstruction and for 3D reconstruction of EM serial sections (http://bio3
d.colorado.edu/imod/)
• Drishti––a volume exploration and presentation tool (http://anusf.anu.edu.au/Vizl
ab/drishti/)
• BioVis3D––software for 3D reconstruction from histological serial sections (http:
//www.biovis3d.com/)
• Ilastik––a tool for image classification and segmentation (http://www.ilastik.org/)
• ESPINA––a tool for the automated segmentation and counting of synapses in large
stacks of electron microscopy images (http://cajalbbp.cesvima.upm.es/espina)
(Morales et al., 2011)
• NeuroStruct––a built-in toolbox for the tracing and analysis of neuroanatomy from
nanoscale imaging (http://www.neurostruct.org/) (Lang, Drouvelis, Tafaj, Bastian,
& Sakmann,2011)
Commercial packages include the following:

• Amira––a software platform for visualizing, manipulating, and understanding life
science and bio-medical data (Visage Imaging Inc., http://www.amira.com/)
• Imaris––for 3D and 4D real-time interactive data visualization (Bitplane Scientific
Software, http://www.bitplane.com/go/products/imaris)
We use the TrakEM2 plug-in in Fiji for montaging images, the image alignment
and manual segmentation tools in Amira, and the moviemaker in Amira or Imaris with
QuickTime 7 Pro for movie editing, and STOMP (Shinywhitebox Ltd.) for high-quality
movie compression. For the purposes of this protocol, we will describe our work using
Amira, but many of the other softwares have similar capabilities.
3. Hardware
It is not unusual to collect a thousand images during a volume EM run. Assum-
ing an image size of 4096 × 4096 pixels, datasets can easily exceed 10 GB, and have
reached 100 GB while montaging images. Due to the requirement for a large FOV dur-
ing CLVEM data collection, the structures of interest are at or near the resolution limit
of the images, and so datasets must be loaded into the software in their entirety (i.e.,
unbinned). Gaming technology has led to massive advances in image handling and
GPUs. At present, we use 96 GB RAM and a 6 GB graphics card, although no doubt
this will soon appear outdated. In addition, an interactive pen display is highly advis-
able for accurate segmentation whilst minimizing the risk of repetitive strain injury—
manual segmentation can take weeks or months depending on the data.
4. Image Alignment
In addition to removing minor pixel shifts, image movement due to the tilt angle of
the sample must be corrected.
1. Using Amira, load the entire dataset into memory.
2. Attach a “Bounding Box” module and an “OrthoSlice” (scroll through to view
individual slices).
3. Attach an “AlignSlices” module. Edit slices manually to minimize shifts or use
the automatic alignment tools, with or without landmarks. Check the alignment—
be aware that the samples that have an inherent direction of movement or rotation
to their structure may cause aberrant shifts in automatic mode. “Resample” the
data, attach a “Bounding Box” module and add three “OrthoSlice” modules (xy,
yz, xz) to visualize the aligned dataset (Fig. 9(A)).
5. Manual Segmentation
4. Attach a “Labelfield” module to the aligned dataset. Open the segmentation editor
(Fig. 9(B)). In the left panel, new “Materials” can be created and named. As fea-
tures are segmented, they are assigned to a material, i.e. a particular type of tissue
or structure. Pixels may be selected using various tools including a brush, a lasso,
a magic wand, and a blow tool. It is often easier to segment in a plane perpendicu-
lar to the plane of data collection, or at least to cross-check the segmentation in the
other orthoslices. In this case, the zebrafish cells and tissues have been segmented
Fig. 9 The VOI can only be viewed using image analysis techniques. (A) Aligned and cropped dataset of
280 slices (of 780 original slices) at 72 nm3 voxel resolution containing the VOI. Scale bar:10 µm. (B) Segmen
tation editor in Amira software (Visage Imaging Inc.) used to assign structures to different materials. (C) 3D
rendering of segmented materials. Neural tube––dark blue, notocord––cyan, cells surrounding the anastomotic
event––purple, lilac, green, endothelial tip cells––yellow, orange. (D, E) The 3D model can be overlaid onto
the aligned data to show the anastomotic point (arrow) with surrounding tissue ultrastructure. (F) The model
shows the proximity of a granular cell to the fusion point (arrow), detail that cannot be observed in the original
fluorescence microscope but which may have significance for the biological process. (G) The 3D model of the
endothelial cells demonstrates why CLVEM is necessary––the cells are only 200 nm thick at some points and
the fusion point itself is approximately 3 µm3. (See color plate.)
into the following materials: neural tube (dark blue), notocord (cyan), cells in the
VOI (purple, lilac), granular cell in the VOI (green), and endothelial tip cells (yel-
low and orange). Crucially, it was not possible to identify the anastomotic ECs
until the manual segmentation had been completed and the structure of each cell
in the VOI had been compared with the original light microscopy data. Therefore,
the manual segmentation was done “blind.”
5. Following segmentation, a 3D representation of the structures can be observed in
the segmentation editor. A corresponding polygonal surface model may then be
created from the 2D segmentation slices using a “SurfaceGen” module. It is likely
that the data will have to be downsampled at this point as the surface generation is
extremely computationally expensive. To do this, attach a “Resample” module to
the Labelfield and downsample by at least 2,2,1 (x, y, z).
6. Visualization of the complete VOI in the original data

6. The 3D model can be displayed at the same time as the aligned orthoslices so as to
view the ultrastructure of the VOI (Fig. 9(D–F)). However, due to the 3D nature
of the VOI, it is impossible to view the entire structure in any one orthoslice.
7. If the voxels are isotropic, the data can be viewed from any direction without
distortion using the “ObliqueSlice” module. Movies can be made where the
“ObliqueSlice” is moved within the VOI to visualize the entire structure within
the aligned data (the same can be done with a “Clipping Plane” in Imaris).
8. While working with tissues, the VOI is highly unlikely to be flat, and therefore
it cannot be viewed in its entirety within any one plane, whether orthogonal or
oblique. In order to view a structure that winds through the tissue, in this case
the endothelial tip cell, a “CurvedSlice” must be created (Fig. 10(A–C). First,
an “ObliqueSlice” is placed in the VOI. This is moved, using the 3D model as a
reference, and each time the “ObliqueSlice” touches the VOI a marker is placed in
the data. These markers are then used to create a curving plane, which appears as
a 2D projection of the structure from the original data.
9. Display and presentation of CLVEM data is challenging. The raw datasets are usually
in excess of 10 GB, whereas publications and presentations usually require a maxi-
mum file size of around 10 MB. Snapshots of the orthoslices and 3D reconstructions
are useful, but must be displayed as 2D projections on paper and so lose informa-
tion and interactivity. To show the aligned data, we first make a movie which scrolls
through the orthoslices in each plane (xy, yz, zx), turning the data between each direc-
tion, using the Imaris software (Bitplane Scientific Software). We then use Quick-
Time 7 Pro to edit the movie, and the H264 compression in STOMP (Shinywhitebox
Ltd.) to reduce the file size whilst maintaining detail in the images. Movies are also
made from the 3D rendered surfaces, and edited and compressed in the same way.
V. Discussion
There is currently no single method for visualizing the complete internal ultrastructure
of large biological samples at nanometer resolution. In general, the larger the sample,
Fig. 10 Viewing the VOI in the aligned dataset. The 3D model shows the overall structure of the endothelial
cells and surrounding structures, but not the ultrastructure. The cells wind through the space between the neural
tube and the muscle blocks, and cannot be viewed in any single orthogonal or oblique plane. In order to view
the cells in Amira, an oblique plane is selected perpendicular to the required final view (A). The oblique slice
is moved through the data and each time the endothelial cells are seen, a marker is placed (B). The markers are
then used to create a CurvedSlice (A, C) in which the VOI can be followed and ultrastructure analyzed.
the lower the imaging resolution. This holds true for microscopes that use X-rays, light,
or electrons as the imaging medium. Correlating light and volume EM overcomes this
problem to reveal the ultrastructure of transient events occurring in situ in whole tissues
and model organisms.
The CLVEM technique described here uses FIB/SEM as the volume EM method of
choice. Serial block face SEM could also be used to produce a stack of serial images
through the sample. However, although FIB/SEM and SBF/SEM give similar results,
each technique has distinct advantages and disadvantages depending on the application.
Cutting with the diamond knife in SBF/SEM is fast and can be applied to large areas
(>0.25 mm2) although slice thickness, and thus axial resolution, is currently limited to
around 25 nm. This makes SBF/SEM highly suitable for creating initial microanatomi-
cal maps of samples (Armer et al., 2009). In many cases, biological samples have not
been viewed in three dimensions at nanometer resolution over hundreds of microns3,
and the resulting datasets are often informative and revealing in themselves.
Milling in the FIB/SEM is more time consuming than cutting with a diamond knife,
and is much more limited in the surface area that can be cut (<0.25 mm2). However, the
ion beam can be targeted to the VOI whilst leaving other areas on the sample surface
uncut. If the VOI is missed during CLVEM, it may be still be possible to investigate an
adjacent region. FIB/SEM also has a higher axial resolution, in the order of 10 nm or
less, although the imaging voltage must be carefully set to ensure that the BSE signal
is only collected from the slice depth.
Advances in sample preparation will improve the morphology of the sample and aid
correlation between imaging modalities. High-pressure freezing and freeze substitu-
tion would give faster fixation, improved preservation, and reduce shrinkage during
dehydration (Muller-Reichert, Mancuso, Lich, & McDonald, 2010). We are currently
using X-ray microscopy to create virtual 3D models of intact samples prior to resin-
embedding to create more accurate 3D maps of the VOI, allowing the FOV to be
reduced and the pixel resolution to be increased during data collection.
In addition, CLVEM will reach its full potential when fluorescent markers for pro-
teins of interest can be detected directly in the SEM chamber, generating registered
3D fluorescence and electron datasets and removing ambiguity from the correlative
process. Advances have already been made in integrating fluorescence detection with
TEM, in a technique called integrated laser electron microscopy or ILEM (Karreman
et al., 2009).
As well as providing structural information on the VOI, datasets generated using
CLVEM also contain a wealth of information regarding other processes or functions
of the organism. We have demonstrated that we can identify, follow, and reconstruct
neurons in a dataset originally collected to analyze blood vessel formation (Armer
et al., 2009). The creation of an open access repository to make high-resolution volume
image data available to the wider research community would maximize the potential of
these techniques across multiple disciplines.
VI. Summary
Currently, correlation of light microscopy and volume EM is the only method by

which we can trace and visualize the 3D ultrastructure of complex networks at nano-
meter resolution within whole tissues and organisms, forming a basis for understanding
genetic and proteomic information in vivo.
Acknowledgments
LC, GM, and HA were supported by Cancer Research UK. Access to the FEI Quanta
3D microscope at Queen Mary University of London was funded for this project by an
EPSRC Access to Equipment Scheme grant [EP/F019882/1].
References
Armer, H. E., Mariggi, G., Png, K. M., Genoud, C., Monteith, A. G., Bushby, A. J., et al. (2009). Imaging
transient blood vessel fusion events in zebrafish by correlative volume electron microscopy. PLoS One,
4, e7716.
Bassim, N. D., De Gregorio, B. T., Kilcoyne, A. L., Scott, K., Chou, T., Wirick, S., et al. (2011). Minimizing
damage during FIB sample preparation of soft materials. Journal of Microscopy, 245, 288–301.
Briggman, K. L., & Denk, W. (2006). Towards neural circuit reconstruction with volume electron micros-
copy techniques. Current Opinion in Neurobiology, 16, 562–570.
Bushby, A. J., P’Ng, K. M., Young, R. D., Pinali, C., Knupp, C., & Quantock, A. J. (2011). Imaging three-
dimensional tissue architectures by focused ion beam scanning electron microscopy. Nature Protocols,
6, 845–858.
De Winter, D. A., Schneijdenberg, C. T., Lebbink, M. N., Lich, B., Verkleij, A. J., Drury, M. R., et al. (2009).
Tomography of insulating biological and geological materials using focused ion beam (FIB) sectioning
and low-kV BSE imaging. Journal of Microscopy, 233, 372–383.
dimensional tissue nanostructure. PLoS Biology, 2, e329.
Drobne, D., Milani, M., Leser, V., & Tatti, F. (2007). Surface damage induced by FIB milling and imaging of
biological samples is controllable. Microscopy Research and Technique, 70, 895–903.
Heymann, J. A., Hayles, M., Gestmann, I., Giannuzzi, L. A., Lich, B., & Subramaniam, S. (2006). Site-
155, 63–73.
Heymann, J. A., Shi, D., Kim, S., Bliss, D., Milne, J. L., & Subramaniam, S. (2009). 3D imaging of mam-
malian cells with ion-abrasion scanning electron microscopy. Journal of Structural Biology, 166, 1–7.
Isogai, S., Lawson, N. D., Torrealday, S., Horiguchi, M., & Weinstein, B. M. (2003). Angiogenic network
formation in the developing vertebrate trunk. Development, 130, 5281–5290.
Jurrus, E., Hardy, M., Tasdizen, T., Fletcher, P. T., Koshevoy, P., Chien, C. B., et al. (2009). Axon tracking in
serial block-face scanning electron microscopy. Medical Image Analysis, 13, 180–188.
Karreman, M. A., Agronskaia, A. V., Verkleij, A. J., Cremers, F. F., Gerritsen, H. C., & Humbel, B. M. (2009).
Discovery of a new RNA-containing nuclear structure in UVC-induced apoptotic cells by integrated laser
electron microscopy. Biology of the Cell, 101, 287–299.
Keller, P. J., & Stelzer, E. H. (2008). Quantitative in vivo imaging of entire embryos with digital scanned
laser light sheet fluorescence microscopy. Current Opinion in Neurobiology, 18, 624–632.
Keller, P. J., & Stelzer, E. H. (2010). Digital scanned laser light sheet fluorescence microscopy. Cold Spring
Harbor Protocols 2010, pdb top78.
Knott, G., Rosset, S., & Cantoni, M. (2011). Focussed ion beam milling and scanning electron microscopy
of brain tissue. Journal of Visualized Experiments, 53, e2588. doi:10.3791/2588.
Kreshuk, A., Straehle, C. N., Sommer, C., Koethe, U., Cantoni, M., Knott, G., et al. (2011). Automated
detection and segmentation of synaptic contacts in nearly isotropic serial electron microscopy images.
PLoS One, 6, e24899.
Kubota, Y., Karube, F., Nomura, M., Gulledge, A. T., Mochizuki, A., Schertel, A., et al. (2011). Conserved
properties of dendritic trees in four cortical interneuron subtypes. Scientific Reports, 1.
Lang, S., Drouvelis, P., Tafaj, E., Bastian, P., & Sakmann, B. (2011). Fast extraction of neuron morphologies
from large-scale SBFSEM image stacks. Journal of Computational Neuroscience, 31, 533–545.
Lawson, N. D., & Weinstein, B. M. (2002). In vivo imaging of embryonic vascular development using trans-
genic zebrafish. Developmental Biology, 248, 307–318.
Macke, J. H., Maack, N., Gupta, R., Denk, W., Scholkopf, B., & Borst, A. (2008). Contour-propagation
algorithms for semi-automated reconstruction of neural processes. Journal of Neuroscience Methods,
167, 349–357.
Morales, J., Alonso-Nanclares, L., Rodriguez, J. R., Defelipe, J., Rodriguez, A., & Merchan-Perez, A.
(2011). Espina: a tool for the automated segmentation and counting of synapses in large stacks of electron
microscopy images. Frontiers in Neuroanatomy, 5, 18.
Mun, J. Y., Jeong, S. Y., Kim, J. H., Han, S. S., & Kim, I. H. (2011). A low fluence Q-switched Nd:YAG
laser modifies the 3D structure of melanocyte and ultrastructure of melanosome by subcellular-selective
photothermolysis. Journal of Electron Microscopy (Tokyo), 60, 11–18.
Munroe, P. R. (2009). The application of focused ion beam microscopy in the material sciences. Materials
Characterization, 60, 2–13.
O’Connell, M. K., Murthy, S., Phan, S., Xu, C., Buchanan, J., Spilker, R., et al. (2008). The three-
dimensional micro- and nanostructure of the aortic medial lamellar unit measured using 3D confocal
and electron microscopy imaging. Matrix Biology, 27, 171–181.
Prenitzer, B. I., Urbanik-Shannon, C. A., Giannuzzi, L. A., Brown, S. R., Irwin, R. B., Shofner, T. L., et al.
(2003). The correlation between ion beam/material interactions and practical FIB specimen preparation.
Risau, W. (1997). Mechanisms of angiogenesis. Nature, 386, 671–674.
Schmidt, F., Kuhbacher, M., Gross, U., Kyriakopoulos, A., Schubert, H., & Zehbe, R. (2011). From 2D slices
to 3D volumes: image based reconstruction and morphological characterization of hippocampal cells
on charged and uncharged surfaces using FIB/SEM serial sectioning. Ultramicroscopy, 111, 259–266.
Schneider, P., Meier, M., Wepf, R., & Muller, R. (2011). Serial FIB/SEM imaging for quantitative 3D assess-
ment of the osteocyte lacuno-canalicular network. Bone, 49, 304–311.
Biology, 9, e1001041.
White, J., Southgate, E., Thomson, J., & Brenner, S. (1986). The structure of the nervous system of the
nematode C. elegans. Philosophical Transactions of the Royal Society of London. Series B, Biological
Sciences, 314 (1165), 1–340.
INDEX
A Bradyrhizobium japonicum, 340–342

Aequorea victoria, 122 British anti-Lewisite (BAL), 146
O6-Alkylguanine-DNA-alkyltransferase (hAGT), BSA-Au labeling, of endocytic pathway, 64–65, 64f
148
American Type Culture Collection (ATCC), 7 C
Amira software, 31, 340, 346, 376
Antibody particle labeling versus antibody fluores- Caenorhabditis elegans (C. elegans), 76, 204–206,
cent photo-oxidation labeling, 142–143, 143f 212–213, 215–216, 219–221, 288–289, 331
Arabidopsis thaliana, 149 early embryo, meiotic spindle assembly interme-
ATCC. See American Type Culture Collection diate stages in, 223–234
(ATCC) electron microscopy, 229, 230f, 231
ATLAS system, 343 freeze-substitution, 228–229
Atmospheric scanning electron microscope high-pressure freezing, 228, 230
(ASEM), 307–324 isolated meiotic embryos, staging of, 226–227,
application studies, 315–320 227f, 230
cell surface receptors, 316–317, 316f, 318f prescreening of samples, 229, 231
cells in culture or tissue sections, 315–316 serial sectioning, 229, 231
lipid rafts, 316–317 thin-layer embedding, 229, 231
mycoplasma in solution, examination of, female meiosis in, 225–226, 225f
317–320, 319f male meiosis in, 225–226, 225f
phosphor particles, cathodoluminescence of, Cells
320, 321f crowded, 140–141
using heated or electrochemical ClairScope culture, 7, 8f, 13–14, 315–316
dishes, 315 in ClairScope dishes, 311b
cell stabilization, 311 of green fluorescent proteins, 126, 128
discussion, 321–323 surface receptors, 316–317, 316f, 318f
dish preparation, 311 in tissue sections, 315–316
heavy metal staining, 312–315, 314f Cellular compartment analysis, 95–115
immunogold labeling, 312 hybrid morphometry, 107–110
JEOL JASM-6200, designing of, 309–311 surface-area estimation (EM image analysis),
sample preparation, 311–315 110
staining of samples, 312 surface-area estimation (FLM image analysis),
109
systematic error calculation, 110
B
materials, 98–100
BAL. See British anti-Lewisite (BAL) antibodies, 99
Biomolecules, immunolocalization of, 348 reagents, 98–99
BioVis3D software, 375 sectioning and imaging equipments, 99
Bleaching, 118, 119f software, 100
of green fluorescent proteins, 126–127, 129. reagents preparation, 100–110
See also Photobleaching 3D HDO-CLEM, 97, 103–107
Block-face imaging, 336 CLEM image processing, 107
BODIPY. See Boron dipyrromethene difluoride double labeling, 102–103, 103f
(BODIPY) electron tomography, 107
Boron dipyrromethene difluoride (BODIPY), EM imaging, 107
119–121 FLM imaging, 104, 105f
383
384 Index
grids preparation, for FLM observations, 103, FIB-SEM, 3D correlation of, 349–350
112 freeze-substitution, 349
grids retrieval and EM preparation, 103 high-pressure freezing, 348–349
immunofluorescence, 102–103, 112 serial-section SEM, 350
immunogold labeling, 102, 112 methods, 331–348
registering serial ROIs, 104–105, 106f correlative array tomography, 345f, 346
3D surface rendering, 106–107 embedding, 332
ultrathin cryosection preparation for, 101–102 en-bloc CLSM, 3D correlation of, 334–342
Cellular cryoelectron tomography, 259–281, 261f FIB-SEM, 3D correlation of, 334–342
discussion and outlook, 277–278 freeze-substitution, 332
instrumentation and materials, 275–277 high-pressure freezing, 331–332
cells, preparation and vitrification of, 275–276 preembedding labeling, fluorescent dyes for,
cryoelectron microscopy, 277 332–334, 333t
cryoelectron tomography, 277 serial-section SEM, 342–346, 344f
cryofluorescence correlative microscopy, 276 rationale, 330–331
cryofocused ion beam milling, 277 sample preparation, 329–330
methods, 263–275 Confocal acquisition, 207, 218
cryocorrelative microscopy, 263–264 Confocal laser optical imaging, 4f
cryoelectron microscopy, 273–274 4-D, 17
cryoelectron tomography, 273–274, 274f for imaging fluorescently labeled drug complexes,
cryofluorescence correlative microscopy, 7–10, 11f, 14
264–266, 265f Confocal laser scanning microscopy (CLSM), 76
cryofocused ion beam instrumentation, 268–270, Correlative array tomography, 48–49, 330
269f of complex biological structures, 345f, 346, 352
cryofocused ion beam milling, 266–268, 267f materials and instrumentation, 350
quantitative data analysis, from cellular Correlative light and electron microscopy (CLEM),
tomograms, 274–275 37–57, 157–173, 4f
thin cellular specimen preparation, experimental of colloidal gold, 41
workflow for, 270–273, 270f–272f combined with fluorescence light microscopy,
rationale, 262. See also Electron tomography 159–161
Chemical fixation, 177–180. See also Fixation combinatorial probes, 160–161
CheY protein, 151 high resolution LM markers, 160
Cisplatin–FITC, 9, 10f–11f of sparse events and dynamic processes, 160
CLEM. See Correlative light and electron microscopy conventional use with transmission light
(CLEM) microscopy, 159
CLIP-tag system, 148 correlation using block face LM image, 164
CLSM. See Confocal laser scanning microscopy of 3,3′-diaminobenzidine tetrahydrochloride, 39–40
(CLSM) of FluoroNanogold, 41–42
CLVEM. See Correlative light and volume electron future perspective of, 170
microscopy (CLVEM) genetics labels, for protein identification, 139–155
Colloidal gold of green fluorescent proteins, 117–138
correlative light-electron microscopy of, 41, 45 for imaging fluorescently labeled drug complexes
markers, 142–144 cell culture for, 7, 8f, 13–14
Combinatorial probes, 160–163 drug effect on cells, determination of, 5
Complex biological structures, 3D correlative light relocation studies, sample preparation for,
and electron microscopy of, 325–356, 327f 10–12, 14
biomolecules, immunolocalization of, 348 integrated fluorescence and electron microscopy,
fluorescence staining, during freeze-substitution, 42–43
350–352 integrated microscopes, 164–165
materials and instrumentation, 348–350 intracellular organelles with preembedding, 21–35
correlative array tomography, 350 labeling of ultrathin resin sections for, 75–93
en-bloc CLSM, 3D correlation of, 349–350 large-scale, 161
Index 385
live-cell imaging, 163–164 Cryofixing CLEM protocol, 23, 27–28

of mammalian culture cells, 262f Cryofluorescence correlative microscopy, 264–266,
marks, 162 265f
mosaic-based back track, flowchart for, 168f instrumentation and materials, 276
need for, 3 Cryofocused ion beam instrumentation, 268–270,
in parasite research, 59–73 269f
photoconversion, 162 Cryofocused ion beam milling, 266–268, 267f
principle of, 179f instrumentation and materials, 277
of quantum dots, 40–41 Cryosectioning, 23. See also Sectioning
of reporter systems, 39–43
of small model organisms, 203–222 D
super resolution microscopy, 164
3,3′-Diaminobenzidine tetrahydrochloride (DAB),
stage holders for, 164–165
286
3D HDO-CLEM. See 3D HDO-CLEM
correlative light-electron microscopy of, 39–40
of tissues, 44–50
fluorescence photo-oxidation of, 143–144
array tomography, 48–49
imaging, preembedding, 286
serial block face scanning electron microscopy,
photoconversion reaction of, 44
50
photo-oxidation of, 118, 118f, 122, 123f
ultrathin cryosections, 44–48, 46f–47f, 49f
DIC. See Differential interference contrast (DIC)
utilization in optical microscopes, limitations of,
imaging
38–39
Differential interference contrast (DIC) imaging, 9
virtual reality overlay during preparation, 165–169
DiI. See 1,1′-Dioctadecyl-3,3,3′,3′-
aim, 165
tetramethylindocarbocyanine perchlorate (DiI)
principle, 166
Dimercaprol, 146
requirements, 165
2,3-Dimercaptopropanol, 146
workflow, 166–169
1,1′-Dioctadecyl-3,3,3′,3′-
Correlative light and volume electron microscopy
tetramethylindocarbocyanine perchlorate (DiI),
(CLVEM), 357–382
119–121
discussion, 378–380
Direct stochastic optical reconstruction microscopy
equipment, 360–361
(dSTORM), 285–286
materials, 361
Drishti software, 375
methods
Drug effect on cells, determination of, 5
FIB/SEM, sample preparation for, 362–367,
dSTORM. See Direct stochastic optical reconstruction
366f
microscopy (dSTORM)
FIB/SEM data collection, 367–373, 368f, 370f,
372f, 374f
E
image analysis, 374–378, 377f, 379f
video microscopy, 361–362, 363f–364f ECFP. See Enhanced cyan fluorescent protein (ECFP)
rationale, 359–360 EDT. See 1,2-Ethanedithiol (EDT)
Correlative multidimensional light microscopy, 4f EGFP. See Enhanced green fluorescent protein
need for, 2–3 (EGFP)
Covalent labeling, 148 Electron microscopy (EM)
Crowded cell, 140–141 capturing endocytic segregation events with
Cryocorrelative microscopy, 263–264 data correlation, 195, 196f
Cryoelectron microscopy, 273–274 sample processing for, 193–194, 194f
instrumentation and materials, 277 cellular compartment analysis using, 96–97, 107,
Cryoelectron tomography, 273–274, 274f 108f
instrumentation and materials, 277 surface-area estimation, 110
Cryo-FIB milling and vitreous cryosectioning, of early C. elegans embryo, 229, 230f, 231
comparison between, 272f, 273–274 fixation for, 177–180
Cryo-FIB thinning technique, 267 fluorescence, 286–287
Cryofixation, 180. See also Fixation nano-FEM, 286, 288f
386 Index
postembedding, 286 Eosin-based fluorescence photo-oxidation, 119–121,

preembedding, 286 142–144
of fluorescent proteins, 302–303 Epidermal growth factor receptor (EGFr), 177
genetics labels, for protein identification, 141–142 Escherichia coli (E. coli), 288–289
of green fluorescent proteins, 127–128 FtnA protein, 151
photo-oxidation, evaluation of, 126–127, 129 ESPINA software, 375
grids, 250 ET. See Electron tomography (ET)
immunoelectron microscopy, 51f 1,2-Ethanedithiol (EDT), 146
ultrathin tissue cryosections, 45 eTomo, 13, 229, 233
integration with fluorescence microscopy, 42–43
Lowicryl resin sections
F
application to thin section 2D, 248, 249f
ROI localization, 215–216 FACS cell sorting, 71–72
serial section, 152 fEM. See Fluorescence electron microscopy (fEM)
structure of interest, 198f Ferritin, 45, 151
Electron tomography (ET) FIB-SEM. See Focused ion beam-scanning electron
capturing endocytic segregation events with, microscopy (FIB-SEM)
197–198 Fiji software, 340, 375
cellular compartment analysis using, 96–97, 107 Fixation, 78
genetically appended peroxidase-based labeling, 152 chemical fixation, 177–180
of Golgi stack, 132f cryofixation, 180
of green fluorescent proteins, 128–129 for electron microscopy, 177–180
Lowicryl resin sections, 244 of fluorescent proteins, 301
instrumentation and materials, 250. See also of green fluorescent proteins, 126–127, 129
Cellular cryoelectron tomography of membrane tubules, 180–181, 182f
EM. See Electron microscopy (EM) FlAsH ligand, 146, 181–183
Embedding, 29–30, 78 Flavin mononucleotide (FMN), 149
complex biological structures, 332 FLM. See Fluorescence light microscopy (FLM)
epoxy resin versus HM20, 209t Flow cytometry, infected cell sorted by, 66–67
flat, 209–210, 210f, 218–219 Fluorescence confocal microscopy
with MatTek Petri dishes, 63–64, 63f L. donovani experiments, 65
LMP agarose, 206–207, 206f, 216–218 Fluorescence electron microscopy (fEM), 286–287
Lowicryl resin sections, 210, 210f, 239–241 nano-FEM, 286, 288f
instrumentation and materials, 249 postembedding, 286
plastic, 291–292, 300, 302 preembedding, 286. See also Electron microscopy
progressive lowering of temperature, 77–88, (EM)
326–328, 328f Fluorescence image shifts, compensating, 247–248
thin-layer, 229, 231 Fluorescence light microscopy (FLM), 76
EMPACT2, 184–185, 207 cellular compartment analysis using, 96–97,
rapid transfer system of, 176, 181, 189, 192f, 104–107, 105f–106f
226–228, 231, 239 surface-area estimation, 109
En-bloc CLEM combinatorial probes, 162–163
challenges to, 329–330 combined with CLEM, 159–161
3D correlation of, 334–342 combinatorial probes, 160–161
materials and instrumentation, 349–350 high resolution LM markers, 160
Endocytic pathway, BSA-Au labeling of, 64–65, 64f of sparse events and dynamic processes, 160
Endocytic segregation, live cell imaging of, 192f immunilabeling, 162–163
Enhanced cyan fluorescent protein (ECFP), 120f– widefield, 76
121f, 123f–124f, 125, 129, 133–134 Fluorescence microscopy
Enhanced green fluorescent protein (EGFP), 123f, integration with electron microscopy, 42–43
125–127, 129, 133–134, 135f Lowicryl resin sections, 241–244, 242f–243f,
photobleaching of, 253–254, 254f. See also Green 245f
fluorescent proteins (GFPs) instrumentation and materials, 250
Index 387
Fluorescence photoactivation localization microscopy fPALM. See Fluorescence photoactivation localiza-

(fPALM), 285–286 tion microscopy (fPALM)
Fluorescent fiducial-based correlation procedure, Freeze substitution, 208–209, 218, 228–229
245f, 246–247 complex biological structures, 332
instrumentation and materials, 250 materials and instrumentation, 349
Fluorescently labeled drug complexes, imaging of, fluorescence staining during, 350–352
1–20 fluorescent proteins, 289–290, 289f, 290t, 291f
confocal laser optical imaging for, 7–10, 11f, 14 instrumentation and materials, 299
correlative light and electron microscopy for, 5 Lowicryl resin sections, 239–241, 249
cell culture for, 7, 8f, 13–14
relocation studies, sample preparation for,
10–12, 14 G
correlative multidimensional light microscopy, Genetically appended peroxidase-based labels,
need for, 2–3 151–152
data processing and correlative analysis, 13, 15 Genetically appended protein tags, 143–144, 145t
discussion, 15–18 Genetics labels, for protein identification, 139–155
4-dimensional future for, 17 challenges to, 152–153
live-cell microscopy for, 7–10, 10f, 14 covalent labeling, 148
rationale, 5–7 crowded cell, 140–141
2-D/3-D transmission electron microscopy for, electron microscopy, 141–142
13–14 future directions of, 152–153
Fluorochrome-conjugated antibodies genetically appended peroxidase-based labels,
CLEM of ultrathin resin sections using 151–152
GFP-labeled rhodopsin, in transgenic Rho-GFP genetically appended protein tags, 143–144, 145t
mouse retina, 85–86 green fluorescent protein, 144–146
α-tubulin, in mouse spermatids microtubular immunomarkers, 142–143
manchette, 87 metal ligand-based labels, 150–151
FluoroNanogold (FNG), 48, 142–143, 151, 160–161 miniSOG, 148–149, 150f
CLEM of ultrathin resin sections using, 85f ReAsH ligand, 146–148, 147f
α-tubulin, in mouse spermatids microtubular spatiotemporal proteomics, 140–141
manchette, 87–88 tetracysteine domain, 146–148, 147f
correlative light-electron microscopy of, 41–42 GFP-GGA1 transport carriers ultrastructure, CLEM
FluoSpheres, 250 analysis of, 25f
FM1-43, 119–121 GFP recognition after bleaching (GRAB), 23–24
FMN. See Flavin mononucleotide (FMN) GFPs. See Green fluorescent proteins (GFPs)
FNG. See FluoroNanogold (FNG) Glutaraldehyde, 142
Focused ion beam (FIB) instrumentation, 261 Glycol methacrylate (GMA), 291
Focused ion beam-scanning electron microscopy autofluorescence in, 294–295
(FIB-SEM), 50 sectioning of, 292–294, 293f
applications of, 340–342, 341f Gold-enhancement reaction, 29
challenges to, 329–330 Golgi apparatus, correlation light and electron
of complex biological structures, 336–337 microscopy of, 120f–121f
data collection, 367–373 GRAB. See GFP recognition after bleaching (GRAB)
principle, 367–369 Green fluorescent proteins (GFPs), 117–138, 140,
protocol, 368f, 369–373, 370f, 372f, 374f 144–146, 158–159, 204–205, 264, 285
membrane staining, optimization of, 337, 338f -based video microscopy, 22
sample preparation for, 362–367 bleaching, 23–24
principle, 362–364 3,3′-Diaminobenzidine photo-oxidation through,
protocol, 364–367, 366f 118, 118f, 122, 123f
signal-to-ratio in, 330 discussion, 129–135
3D correlation of, 334–342 enhanced, 123f, 125–127, 129, 133–134, 135f
materials and instrumentation, 349–350. See also photobleaching of, 253–254, 254f
Scanning electron microscopy (SEM) -labeled rhodopsin
388 Index
in resin-embedded mouse retina, 80f membrane-targeted, 152

in transgenic Rho-GFP mouse retina, 85–86 recombinant, 152
materials, 125–128 single sequence, 151–152
bleaching, 129 HPF. See High-pressure freezing (HPF)
cell culture, 128 HPF-CLEM, capturing endocytic segregation events
electron microscopy, 129 with, 175–201
electron tomography, 129 electron microscopy
EM Preparation, 129 data correlation, 195, 196f
fixation, 129 sample processing for, 193–194, 194f
methods, 125–128 electron tomography, 197–198
bleaching, 126–127 image acquisition, improvement of, 189, 190f
cell culture, 126 intracellular trafficking, 181
electron microscopy, 127–128 live cell carriers, preparation of, 184–185,
electron tomography, 128 186f–187f
EM Preparation, 126–127 live cell imaging, 191–193, 192f
fixation, 126–127 markers, 181–183, 183f
3D reconstruction, 128 modification to rapid loader, 189
peroxisomal-targeted GFP, photo-oxidation of, 122 modified carrier
photoconversion reaction of, 44 development of, 185–187, 186f–187f
rationale, 124–125 preparation of, 186f–187f, 187–188
signal-to-noise ratio, quantitative analysis of, 122, specimen carriers, 183–184
124f stage inserts, 190f
Groundstate depletion with single-molecule return HRP. See Horseradish peroxidase (HRP)
microscopy (GSDIM), 285–286
fluorescent proteins, 287, 295, 297–298, 299f, 302 I
IEM. See Immunoelectron microscopy (IEM)
H
Ilastik software, 375
hAGT. See O6-Alkylguanine-DNA-alkyltransferase ILEM. See Integrated laser and electron microscopy
(hAGT) (ILEM)
HALOtag™, 148 Image analysis, in CLVEM, 374–378, 377f, 379f
Heavy metal staining, 312–315, 314f. See also complete VOI visualization, in original data, 378
Staining hardware, 376
HepG2 infection, 66 image alignment, 376
Hexamethyldisilazane (HMDS), 292–294 manual segmentation, 376–378
High-pressure freezing (HPF), 23, 78, 163–164, principle, 374–375
207–208, 260, 326–329, 328f software, 375–376
carrier, specimen removal from, 209–210 ImageJ, 97–98, 100, 107, 109–110, 340
complex biological structures, 330–332 Imaris software, 31, 376
materials and instrumentation, 348–349 Immunoelectron microscopy (IEM), 51f
of early C. elegans embryo, 228, 230 ultrathin tissue cryosections, 45. See also Electron
flat embedding after, 209–210, 210f microscopy (EM)
of fluorescent proteins, 288–289 Immunoenzymatic methods, 142
instrumentation and materials, 298–299 Immunoglobulin G (IgG) gold
instrumentation and materials, 218 CLEM of ultrathin resin sections using, 79–82,
of Lowicryl resin sections, 239 84f
of mammalian cells, 248 α-tubulin, in mouse spermatids microtubular
of yeast cells, 248. See also HPF-CLEM. See also manchette, 87
capturing endocytic segregation events with Immunogold labeling, 28, 312
High-throughput biomolecular screening assays, 5 immunofluorescence with, 204–205
HMDS. See Hexamethyldisilazane (HMDS) on ultrathin cryosections, 102, 112
Horseradish peroxidase (HRP) double labeling, 102–103, 103f
-mediated DAB participation, 131–132 Immunolabeling, 28–29, 162–163
Index 389
complex biological structures, 342–343, 347f specimen preparation for, 206–207

fluorescent protein with 20 nm gold particles, 313b confocal acquisition, 207
Immunomarkers, 142–143 LMP agarose embedding, 206–207, 206f
IMOD software, 100, 107, 111, 128, 229, 244, 375 Lipid rafts, 316–317
Integrated laser and electron microscopy (ILEM), 380 Live cell carriers, preparation of, 184–185,
Intracellular organelles with preembedding CLEM, 186f–187f
21–35, 26f Live-cell imaging, 2–3, 163–164
discussion, 32–33 of endocytic segregation, 192f
embedding, 29–30 of fluorescently labeled drug complexes, 7–10,
immuno-labeling, 28–29 10f, 14
materials, 24–27 HPF-CLEM, 191–193, 192f
methods, 27–32 LM. See Light microscopy (LM)
cell transfection, 27–28 Lowicryl embedding, 210, 210f
fixation, 27–28 Lowicryl resin sections, using fluorescent fiducial
observation of structures, 27–28 markers, 235–257, 240f
rationale, 22–24 applications of, 241, 254–255
sectioning, 30–31 correlation accuracy, estimation of, 253
serial-section analysis, 31–32 EGFP, photobleaching of, 253–254, 254f
3D reconstruction, 31–32 EM grids, choice of, 251–252
Intracellular trafficking, 177, 178f FluoSpheres, choice of, 252
HPF-CLEM for, 181 instrumentation and materials, 248–250
iTEM platform, 100, 107 electron tomography, 250
embedding, 249
EM grids, 250
J
fluorescence microscopy, 250
JEOL 1400 TEM, 13. See also Transmission electron fluorescent fiducial-based correlation, 250
microscopy (TEM) FluoSpheres, 250
JEOL 2100 TEM, 13. See also Transmission electron freeze-substitution, 249
microscopy (TEM) mammalian cells, high-pressure freezing of,
JEOL JASM-6200, 308, 309f 248
designing of, 309–311, 310f ultramicrotomy, 250
dishes, culturing cells in, 311b yeast cells, high-pressure freezing of, 248
methods, 239–248
K application to thin section 2D electron micros-
copy, 248, 249f
Kaede, photoconversion of, 121–122
electron tomography, 244
Kodak KAI-2020 sensor, 310–311
embedding, 239–241
fluorescence image shifts, compensating,
L 247–248
LAMP-1, 23, 66, 68–69, 70f fluorescence microscopy, 241–244, 242f–243f,
Laser carving, 219 245f
Laser etching, ROI marking by, 210–211 fluorescent fiducial-based correlation, 245f,
Leishmania donovani (L. donovani) 246–247
experiments, 65–66 freeze-substitution, 239–241
fluorescence confocal microscopy, 65 high-pressure freezing, 239
macrophages infection assay, 65 sectioning, 241
mice and macrophages, 65 mCherry, photobleaching of, 253–254, 254f
TIRF microscopy, 66 rationale, 239
transmission electron microscopy, 66 sample preparation protocol, flexibility in, 251
promastigote–host cell interaction, 60–61, 68 Low melting point (LMP) agarose embedding,
Light microscopy (LM) 206–207, 206f, 216–218
isolated meiotic embryos, staging of, 226–227, LysoTracker labeling, 62, 62f, 67, 67f
227f LysoTracker Red DND-99, 9
390 Index
M flat embedding procedure with MatTek Petri

Macrophages infection assay, 65 dish, 63–64, 63f
Mammalian cells Plasmodium berghei experiments, 66–67
correlative light and electron microscopy of, 262f HepG2 infection, 66
high-pressure freezing of, 248 infected cell sorted by flow cytometry, 66–67
MATLAB, 246–247 parasites and mosquitoes, 66
MatTek Petri dishes, 27–28, 30–31, 63–64, 63f, 67, 125 transmission electron microscopy, 67
mCherry, photobleaching of, 253–254, 254f rationale, 61–62
Membrane tubules, fixation of, 180–181, 182f results and discussion, 67–70, 67f–70f
Mercaptopropionic acid-functionalized quantum dots, Patterning, 271
41 Peroxisomal-targeted GFP, photo-oxidation of,
Metal ligand-based labels, 150–151 122
Metallothionein (MTH), 151 Phase contrast microscopy, 60–61
MicroScoBioJ, 97–98, 100, 109–110 Phosphor particles, cathodoluminescence of, 320,
Minimal Edgar’s growth medium (MEGM), 226, 228 321f
Mini singlet-oxygen generator (miniSOG), 23–24, Photoactivatable localization microscopy (PALM),
32–33, 40, 148–149, 160–161, 181–183 76, 160, 164, 263, 278, 285–286
fusion with SynCAM, 149 fluorescence, 285–286
labeled α-actinin, fluorescence photo-oxidation of, of fluorescence proteins, 287, 292–295, 297–298,
150f 298f, 300, 302–303
MosaicJ, 100, 104, 105f, 111 Photoactivation localization microscopy, 38–39
MPO. See Myeloperoxidase (MPO) Photobleaching
Multiple image alignment (MIA) software, 9 of EGFP, 253–254, 254f
Mycoplasma in solution, examination of, 317–320, 319f of mCherry, 253–254, 254f. See also Bleaching
Mycoplasma pneumoniae, 317 Photoconversion, 121–122, 162
Myeloperoxidase (MPO) of Kaede, 121–122
localization in human neutrophils, 42, 43f Photo-oxidation, 119–121
of 3,3′-Diaminobenzidine, 118, 118f
of FlAsH, 119–121
N
of ReAsH, 119–121
Nano-FEM, 286, 288f, 298f–299f Plasmodium berghei experiments, 66–67
NeuroStruct software, 375 HepG2 infection, 66
infected cell sorted by flow cytometry, 66–67
O parasites and mosquitoes, 66
Olympus CellR live-cell microscope, 9. See also transmission electron microscopy, 67
Live-cell microscopy Plastic embedding, 291–292, 293f, 302
Open Tomo program, 128 instrumentation and materials, 300. See also
Embedding
Polymerization, 78
P
Preembedding labeling, fluorescent dyes for,
PALM. See Photoactivatable localization microscopy 332–334, 333t
(PALM) Progressive lowering of temperature (PLT)-
Paraformaldehyde, 142 embedding, 77–88, 326–328, 328f. See also
Parasite research, CLEM in, 59–73 Embedding
L. donovani experiments, 65–66 Protein A-gold (PAG), 97–98
fluorescence confocal microscopy, 65 CLEM of ultrathin resin sections using, 79–82,
macrophages infection assay, 65 82f–83f
mice and macrophages, 65 GFP-labeled rhodopsin, in transgenic Rho-GFP
TIRF microscopy, 66 mouse retina, 85–86
transmission electron microscopy, 66 α-tubulin, in mouse spermatids microtubular
methods, 63–65 manchette, 87
endocytic pathway, BSA-Au labeling of, 64–65, Proteins visualization, in electron micrographs at
64f nanometer resolution, 283–306
Index 391
alignment, 303 focused ion beam. See Focused ion beam-scanning

electron microscopy imaging, 302–303 electron microscopy
fixation, 301 serial block face, 50, 152, 267–268
instrumentations and materials, 298–301 of tissues, 38
coverslip cleaning, 300 Sectioning, 30–31, 78, 99, 213–215
freeze-substitution, 299 cellular compartment analysis, 99
high-pressure freezing, 298–299 in coverglass, mounting, 294, 294f
image alignment, 301 cryosectioning, 23
PALM imaging, 300 of early C. elegans embryo
plastic embedding, 300 serial, 229, 231
sectioning, 300 glycol methacrylate, 292–294, 293f
SEM imaging, 301 of GMA plastic, 292–294, 293f
staining, 300 instrumentation and materials, 300
methods, 287–298 intracellular organelles, 30–31
fluorescence imaging, 294–296 Lowicryl resin sections, 241
freeze-substitution, 289–290, 289f, 290t, 291f pickup, support film on metallic slides and, 215,
high-pressure freezing, 288–289 215f
image alignment, 297–298 serial
plastic embedding, 291–292, 293f in early C. elegans embryo, 229, 231
sectioning, 292–294, 293f–294f in small model organisms, 213–214, 219
SEM imaging, 296 ultrathin resin sections, 78
plastic embedding, 302 SEM. See Scanning electron microscopy (SEM)
quantification, 303 Serial block face scanning electron microscopy
rationale, 286–287 (SBFSEM), 50, 267–268
super-resolution fluorescence imaging, 302 genetically appended peroxidase-based labeling,
152. See also Scanning electron microscopy
Q (SEM)
SerialEM software, 13, 128
Quantum dots, 40–41, 142–144, 160–161 Serial-section analysis of intracellular organelles,
QuickTime 7 software, 376 31–32
SNAP-tag system, 148
R Spatial cell biology, 141
Rapid loader, modification to, 189 Spatiotemporal proteomics, 140–141
Rapid transfer system (RTS) Specimen carriers, 183–184
of EMPACT2, 176, 228, 239 Staining, 312
ReAsH ligand, 146, 147f, 148, 181–183 biomaterials with platinum blue, 313b
Reduced osmium method, 122 fluorescence, during freeze substitution, 350–352
Region of interest (ROI), 161–165, 329–330, 334 heavy metal, 312–315
localization, 215–216 membrane staining, optimization of, 337
marking by laser etching, 210–211 Stimulated emission depletion (STED) microscopy,
registering serial, 104–105, 106f 38–39, 76, 160, 164, 285, 287
RTS. See Rapid transfer system (RTS) Stochastic optical reconstruction microscopy
Russell bodies, 98, 112 (STORM), 38–39, 263, 278, 285–286
rough, 98 direct, 285–286
smooth, 98 STOMP software, 376
Structured illumination microscopy, 38–39
Styryl dyes, 119–121
S Sucrose-embedded tissues, immunoelectron
SBFSEM. See Serial block face scanning electron microscopy of, 45
microscopy (SBFSEM) Super-resolution fluorescence microscopy, 38–39, 76
Scanning electron microscopy (SEM), 15 of fluorescent proteins, 302
of fluorescent proteins, 296, 297f Super-resolution optical microscopy, 2–3
instrumentation and materials, 301 SynCAM–miniSOG fusion, 149
392 Index
T immunofluorescence, 102–103, 112

Targeted ultramicrotomy, of small model organisms, immunogold labeling, 102, 112
203–222, 210–213 registering serial ROIs, 104–105, 106f
discussion, 219–221 3D surface rendering, 106–107
instrumentation and materials, 216–219 ultrathin cryosections preparation for, 101–102. See
confocal imaging, 218 also Correlative light and electron microscopy
flat embedding, 218–219 (CLEM)
freeze substitution, 218 3D reconstruction
high-pressure freezing, 218 of green fluorescent proteins, 128
laser carving, 219 of intracellular organelles, 31–32
LMP agarose embedding, 216–218 3D surface rendering, 106–107
serial sectioning, 219 TIRF. See Total internal reflection fluorescence
TEM analysis, 219 (TIRF) microscopy
methods, 206–216 Tissues, correlative light-electron microscopy of,
flat embedding, 209–210, 210f 44–50
freeze substitution, 208–209 ultrathin cryosections, 44–48
high-pressure freezing, 207–208 Total internal reflection fluorescence (TIRF)
internal landmark determination, 212–213 microscopy, 60–61, 68
light microscopy, specimen preparation for, L. donovani experiments, 66
206–207 TrakEM2 software, 346, 376
measurements and positioning, 211, 211f Transferrin receptor (Tfr), 177
ROI localization, 215–216 Trans-Golgi network (TGN), 177
ROI marking by laser etching, 210–211 Transmission electron microscopy (TEM), 4f
section pickup, support film on metallic slides of fluorescent proteins, 302–303
and, 215, 215f 4-D, 17
serial sectioning, 213–214 for imaging fluorescently labeled drug complexes
specimen removal from HPF carrier, 209–210 2-D, 13–14
TEM analysis, 216 3-D, 13–14
trimming precision, 212–213 JEOL 1400, 13
rationale, 205–206 JEOL 2100, 13
TEM. See Transmission electron microscopy (TEM) L. donovani experiments, 66
TEP. See Triethylphosphine (TEP) need for, 3
Tetracysteine domain, 146–148, 147f Plasmodium berghei experiments, 67
Thin cellular specimen preparation, experimental serial-section, 359–360
workflow for, 270–273, 270f–272f small model organisms, analysis of, 216
3D correlative light and electron microscopy, 329 instrumentation and materials, 219
applications of, 340–342 of tissues, 38
block-face imaging by, 336 Transmission light microscopy
for cell culture applications, 339f conventional use with CLEM, 159
of complex biological structures, 325–356, 327f, Triethylphosphine (TEP), 148
334–342, 352 α-Tubulin, in mouse spermatids microtubular
materials and instrumentation, 348–349 manchette, 87–88
membrane staining, optimization of, 337 on-section labeling
3D HDO-CLEM, cellular compartment analysis with FluoNanogold, 87–88
using, 97, 103–107 with fluorochrome-conjugated antibodies, 87
CLEM image processing, 107 with IgG gold, 87
double labeling, 102–103, 103f with Prot A gold, 87
electron tomography, 107 TurboReg, 100, 104–105, 106f, 107, 110
EM imaging, 107
FLM imaging, 104, 105f U
grids preparation, for FLM observations, 103, 112 Ultramicrotomy, of Lowicryl resin sections, 250
grids retrieval and EM preparation, 103 Ultrathin resin sections, labeling of, 75–93, 79f
Index 393
GFP-labeled rhodopsin, in transgenic Rho-GFP Ultrathin tissue cryosections, for correlative light-
mouse retina, 85–86 electron microscopy, 44–48, 46f–47f, 49f
on-section labeling with fluorochrome- UPLAN Super Apochromat NA 0.3, 9
conjugated antibodies, 85–86 UPLAN Super Apochromat NA 0.9, 9
on-section labeling with Prot A gold, 85–86
materials, 88–89 V
PLT-embedding, 77–88
Video microscopy, 361–362, 363f–364f
dehydration, 78
principle, 361–362
embedding, 78
protocol, 361–362
fixation, 78
Vigna radiata, 331
infiltration, 78
Vitreous cryosectioning and cryo-FIB milling,
polymerization, 78
comparison between, 272f, 273–274
sectioning, 78
Volocity software, 31
protocol, 79–84
rationale, 77
summary and outlook, 89–90 Y
α-tubulin, in mouse spermatids microtubular Yeast cells, high-pressure freezing of, 248
manchette, 87–88 Yellow fluorescent protein, 133–134
on-section labeling with FluoNanogold, 87–88
on-section labeling with fluorochrome- Z
conjugated antibodies, 87
on-section labeling with IgG gold, 87 ZapA protein, 151
on-section labeling with Prot A gold, 87 Zebrafish embryos, angiogenesis in, 363f
using immunoglobulin G gold, 79–82, 84f
using protein A (Prot A) gold, 79–82, 82f–83f
using FluoNanogold, 85f
VOLUMES IN SERIES
Founding Series Editor

DAVID M. PRESCOTT
Volume 1 (1964)
Methods in Cell Physiology
Edited by David M. Prescott
Volume 2 (1966)
Volume 3 (1968)
Volume 4 (1970)
Volume 5 (1972)
Volume 6 (1973)
Volume 7 (1973)
Volume 8 (1974)
Volume 9 (1975)
Volume 10 (1975)
395
396 Volumes in Series
Volume 11 (1975)
Yeast Cells
Volume 12 (1975)
Yeast Cells
Volume 13 (1976)
Volume 14 (1976)
Volume 15 (1977)
Volume 16 (1977)
Chromatin and Chromosomal Protein Research I
Edited by Gary Stein, Janet Stein, and Lewis J. Kleinsmith
Volume 17 (1978)
Chromatin and Chromosomal Protein Research II
Volume 18 (1978)
Chromatin and Chromosomal Protein Research III
Volume 19 (1978)
Chromatin and Chromosomal Protein Research IV
Volume 20 (1978)
Advisory Board Chairman

KEITH R. PORTER
Volume 21A (1980)
Normal Human Tissue and Cell Culture, Part A: Respiratory, Cardiovascular,
and Integumentary Systems
Edited by Curtis C. Harris, Benjamin F.Trump, and Gary D. Stoner
Volumes in Series 397
Volume 21B (1980)

Normal Human Tissue and Cell Culture, Part B: Endocrine, Urogenital, and
Gastrointestinal Systems
Edited by Curtis C. Harris, Benjamin F.Trump, and Gray D. Stoner
Volume 22 (1981)
Three-Dimensional Ultrastructure in Biology
Edited by James N.Turner
Volume 23 (1981)
Basic Mechanisms of Cellular Secretion
Edited by Arthur R. Hand and Constance Oliver
Volume 24 (1982)
The Cytoskeleton, Part A: Cytoskeletal Proteins, Isolation and
Characterization
Edited by Leslie Wilson
Volume 25 (1982)
The Cytoskeleton, Part B: Biological Systems and In Vitro Models
Edited by Leslie Wilson
Volume 26 (1982)
Prenatal Diagnosis: Cell Biological Approaches
Edited by Samuel A. Latt and Gretchen J. Darlington
Series Editor
LESLIE WILSON
Volume 27 (1986)
Echinoderm Gametes and Embryos
Edited by Thomas E. Schroeder
Volume 28 (1987)
Dictyostelium discoideum: Molecular Approaches to Cell Biology
Edited by James A. Spudich
Volume 29 (1989)
Fluorescence Microscopy of Living Cells in Culture, Part A: Fluorescent
Analogs, Labeling Cells, and Basic Microscopy
Edited by Yu-Li Wang and D. Lansing Taylor
Volume 30 (1989)
Fluorescence Microscopy of Living Cells in Culture, Part B: Quantitative
Fluorescence Microscopy—Imaging and Spectroscopy
Edited by D. Lansing Taylor and Yu-Li Wang
Volume 31 (1989)
Vesicular Transport, Part A
Edited by Alan M.Tartakoff
Volume 32 (1989)
Vesicular Transport, Part B
Volume 33 (1990)
Flow Cytometry
Edited by Zbigniew Darzynkiewicz and Harry A. Crissman
Volume 34 (1991)
Vectorial Transport of Proteins into and across Membranes
Selected from Volumes 31, 32, and 34 (1991)
Laboratory Methods for Vesicular and Vectorial Transport
Volume 35 (1991)
Functional Organization of the Nucleus: A Laboratory Guide
Edited by Barbara A. Hamkalo and Sarah C. R. Elgin
Volume 36 (1991)
Xenopus laevis: Practical Uses in Cell and Molecular Biology
Edited by Brian K. Kay and H. Benjamin Peng
Series Editors
LESLIE WILSON AND PAUL MATSUDAIRA
Volume 37 (1993)
Antibodies in Cell Biology
Edited by David J. Asai
Volume 38 (1993)
Cell Biological Applications of Confocal Microscopy
Edited by Brian Matsumoto
Volume 39 (1993)
Motility Assays for Motor Proteins
Edited by Jonathan M. Scholey
Volume 40 (1994)
A Practical Guide to the Study of Calcium in Living Cells
Edited by Richard Nuccitelli
Volume 41 (1994)
Flow Cytometry, Second Edition, Part A
Edited by Zbigniew Darzynkiewicz, J. Paul Robinson, and Harry A. Crissman
Volume 42 (1994)
Flow Cytometry, Second Edition, Part B
Edited by Zbigniew Darzynkiewicz, J. Paul Robinson, and Harry A. Crissman
Volume 43 (1994)
Protein Expression in Animal Cells
Edited by Michael G. Roth
Volume 44 (1994)
Drosophila melanogaster: Practical Uses in Cell and Molecular Biology
Edited by Lawrence S. B. Goldstein and Eric A. Fyrberg
Volume 45 (1994)
Microbes as Tools for Cell Biology
Edited by David G. Russell
Volume 46 (1995)
Cell Death
Edited by Lawrence M. Schwartz and Barbara A. Osborne
Volume 47 (1995)
Cilia and Flagella
Edited by William Dentler and George Witman
Volume 48 (1995)
Caenorhabditis elegans: Modern Biological Analysis of an Organism
Edited by Henry F. Epstein and Diane C. Shakes
Volume 49 (1995)
Methods in Plant Cell Biology, Part A
Edited by David W. Galbraith, Hans J. Bohnert, and Don P. Bourque
Volume 50 (1995)
Methods in Plant Cell Biology, Part B
Edited by David W. Galbraith, Don P. Bourque, and Hans J. Bohnert
Volume 51 (1996)
Methods in Avian Embryology
Edited by Marianne Bronner-Fraser
Volume 52 (1997)
Methods in Muscle Biology
Edited by Charles P. Emerson, Jr. and H. Lee Sweeney
Volume 53 (1997)
Nuclear Structure and Function
Edited by Miguel Berrios
Volume 54 (1997)
Cumulative Index
Volume 55 (1997)
Laser Tweezers in Cell Biology
Edited by Michael P. Sheetz
Volume 56 (1998)
Video Microscopy
Edited by Greenfield Sluder and David E.Wolf
Volume 57 (1998)
Animal Cell Culture Methods
Edited by Jennie P. Mather and David Barnes
Volume 58 (1998)
Green Fluorescent Protein
Edited by Kevin F. Sullivan and Steve A. Kay
Volume 59 (1998)
The Zebrafish: Biology
Edited by H.William Detrich III, Monte Westerfield, and Leonard I. Zon
Volume 60 (1998)
The Zebrafish: Genetics and Genomics
Volume 61 (1998)
Mitosis and Meiosis
Edited by Conly L. Rieder
Volume 62 (1999)
Tetrahymena thermophila
Edited by David J. Asai and James D. Forney
Volume 63 (2000)
Cytometry, Third Edition, Part A
Edited by Zbigniew Darzynkiewicz, J. Paul Robinson, and Harry Crissman
Volume 64 (2000)
Cytometry, Third Edition, Part B
Edited by Zbigniew Darzynkiewicz, J. Paul Robinson, and Harry Crissman
Volume 65 (2001)
Mitochondria
Edited by Liza A. Pon and Eric A. Schon
Volume 66 (2001)
Apoptosis
Edited by Lawrence M. Schwartz and Jonathan D. Ashwell
Volume 67 (2001)
Centrosomes and Spindle Pole Bodies
Edited by Robert E. Palazzo and Trisha N. Davis
Volume 68 (2002)
Atomic Force Microscopy in Cell Biology
Edited by Bhanu P. Jena and J. K. Heinrich Hörber
Volume 69 (2002)
Methods in Cell–Matrix Adhesion
Edited by Josephine C. Adams
Volume 70 (2002)
Cell Biological Applications of Confocal Microscopy
Edited by Brian Matsumoto
Volume 71 (2003)
Neurons: Methods and Applications for Cell Biologist
Edited by Peter J. Hollenbeck and James R. Bamburg
Volume 72 (2003)
Digital Microscopy: A Second Edition of Video Microscopy
Volume 73 (2003)
Cumulative Index
Volume 74 (2004)
Development of Sea Urchins, Ascidians, and Other Invertebrate
Deuterostomes: Experimental Approaches
Edited by Charles A. Ettensohn, Gary M.Wessel, and Gregory A.Wray
Volume 75 (2004)
Cytometry, 4th Edition: New Developments
Edited by Zbigniew Darzynkiewicz, Mario Roederer, and Hans Tanke
Volume 76 (2004)
The Zebrafish: Cellular and Developmental Biology
Edited by H.William Detrich, III, Monte Westerfield, and Leonard I. Zon
Volume 77 (2004)
The Zebrafish: Genetics, Genomics, and Informatics
Edited by William H. Detrich, III, Monte Westerfield, and Leonard I. Zon
Volume 78 (2004)
Intermediate Filament Cytoskeleton
Edited by M. Bishr Omary and Pierre A. Coulombe
Volume 79 (2007)
Cellular Electron Microscopy
Edited by J. Richard McIntosh
Volume 80 (2007)
Mitochondria, 2nd Edition
Edited by Liza A. Pon and Eric A. Schon
Volume 81 (2007)
Digital Microscopy, 3rd Edition
Volume 82 (2007)
Laser Manipulation of Cells and Tissues
Edited by Michael W. Berns and Karl Otto Greulich
Volume 83 (2007)
Cell Mechanics
Edited by Yu-Li Wang and Dennis E. Discher
Volume 84 (2007)
Biophysical Tools for Biologists, Volume One: In Vitro Techniques
Edited by John J. Correia and H.William Detrich, III
Volume 85 (2008)
Fluorescent Proteins
Edited by Kevin F. Sullivan
Volume 86 (2008)
Stem Cell Culture
Edited by Dr. Jennie P. Mather
Volume 87 (2008)
Avian Embryology, 2nd Edition
Edited by Dr. Marianne Bronner-Fraser
Volume 88 (2008)
Introduction to Electron Microscopy for Biologists
Edited by Prof.Terence D. Allen
Volume 89 (2008)
Biophysical Tools for Biologists, Volume Two: In Vivo Techniques
Edited by Dr. John J. Correia and Dr. H.William Detrich, III
Volume 90 (2008)
Methods in Nano Cell Biology
Edited by Bhanu P. Jena
Volume 91 (2009)
Cilia: Structure and Motility
Edited by Stephen M. King and Gregory J. Pazour
Volume 92 (2009)
Cilia: Motors and Regulation
Volume 93 (2009)
Cilia: Model Organisms and Intraflagellar Transport
Volume 94 (2009)
Primary Cilia
Edited by Roger D. Sloboda
Volume 95 (2010)
Microtubules, in vitro
Edited by Leslie Wilson and John J. Correia
Volume 96 (2010)
Electron Microscopy of Model Systems
Edited by Thomas Müeller-Reichert
Volume 97 (2010)
Microtubules: In Vivo
Edited by Lynne Cassimeris and Phong Tran
Volume 98 (2010)
Nuclear Mechanics & Genome Regulation
Edited by G.V. Shivashankar
Volume 99 (2010)
Calcium in Living Cells
Edited by Michael Whitaker
Volume 100 (2010)
The Zebrafish: Cellular and Developmental Biology, Part A
Edited by: H.William Detrich III, Monte Westerfield and Leonard I. Zon
Volume 101 (2011)

The Zebrafish: Cellular and Developmental Biology, Part B
Edited by: H.William Detrich III, Monte Westerfield and Leonard I. Zon
Volume 102 (2011)
Recent Advances in Cytometry, Part A: Instrumentation, Methods
Edited by Zbigniew Darzynkiewicz, Elena Holden, Alberto Orfao,William Telford and
Donald Wlodkowic
Volume 103 (2011)
Recent Advances in Cytometry, Part B: Advances in Applications
Edited by Zbigniew Darzynkiewicz, Elena Holden, Alberto Orfao, Alberto Orfao and
Donald Wlodkowic
Volume 104 (2011)
The Zebrafish: Genetics, Genomics and Informatics 3rd Edition
Volume 105 (2011)
The Zebrafish: Disease Models and Chemical Screens 3rd Edition
Volume 106 (2011)
Caenorhabditis elegans: Molecular Genetics and Development 2nd Edition
Edited by Joel H. Rothman and Andrew Singson
Volume 107 (2011)
Caenorhabditis elegans: Cell Biology and Physiology 2nd Edition
Edited by Joel H. Rothman and Andrew Singson
Volume 108 (2012)
Lipids
Edited by Gilbert Di Paolo and Markus R Wenk
Volume 109 (2012)
Tetrahymena thermophila
Edited by Kathleen Collins
Volume 110 (2012)
Edited by Anand R. Asthagiri and Adam P. Arkin
Plate 1 (Fig. 3 on page 27 of this volume).




EE
ation
tur
ma TEN
RE
maturation
TGN
LE/MVB
G
ER
L
ERGIC
TfR
EGFR
Coat



A B C

input for fiducial-based correlation procedure
fluorescence microscopy electron microsocopy
1. assigning fiducial markers in FM and EM

A fiducial fluorescence B C
2. assigning FP spot of interest

calculate optimal
D RFP fluorescence E centroid of FP spot transformation between
of interest fiducial and EM image
based on fiducial positions
apply transformation to
fluorescent spot coordinates
G
3. shift between fluorescent images
F identify fiducial fluorescent
signal bleedthrough in RFP image
calculate shift
and apply
to fluorescent
spot coordinates

coordinates: x=1106 y=1210
A C

TOM-20-tdEos TOM-20-tdEos
A No fixatives B With fixatives
50 µm
C Quantification of fluorescence intensity from

animals treated with or without fixatives
Fluorescence intensity
200
(arbitary unit)
100
0
Control (no fixatives, Fixed (0.1% KMnO4 + 0.001% OsO4,
embedded in GMA) en bloc UA staining, embedded in GMA)
A Picking up sections using a loop B Picking up sections directly onto a coverslip



Correlative Light and Electron MIcroscopy (Thomas Müller-Reichert and Paul Verkade

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Correlative Light and Electron MIcroscopy (Thomas Müller-Reichert and Paul Verkade

Uploaded by

Copyright:

Available Formats

AMSTERDAM • BOSTON • HEIDELBERG • LONDON

The Boulevard, Langford Lane, Kidlington, Oxford, OX51GB, UK

First edition 2012

Copyright © 2012 Elsevier Inc. All rights reserved

For information on all Academic Press publications

Printed and bound in USA

Numbers in parentheses indicate the pages on which the author’s contributions

Marius Ader (75), Center for Regenerative Therapies, TU Dresden, Fetscherstraße

Clare L. Dennison (307), Technology Facility, Biology Department, University of

Elena V. Polishchuk (21), Institute of Protein Biochemistry, Naples, Italy

Paul Verkade (175), Department of Biochemistry, School of Medical Sciences,

Introduction to Correlative Light and Electron Microscopy

Correlative microscopy, as an approach to combine methods of different modalities,

In parallel to this volume of Methods in Cell Biology, we have designed an annual

Thomas Müller-Reichert, Paul Verkade

March 31st, 2012

Imaging Fluorescently Labeled Complexes

METHODS IN CELL BIOLOGY, VOL 111 0091-679X

A. Need for Correlative Microscopy

B. Determining the Effects of Drugs on Cells: an Application for CLEM

cross-correlate structure–function information. The first route is based on 3-D live-cell

A. Cell Culture for Correlative Light and Electron Microscopy Imaging

B. Live-Cell Microscopy and Confocal Laser Imaging

C. Sample Preparation for Relocation Studies by Electron Microscopy

D. 2-D and 3-D Transmission Electron Microscopy

E. Data Processing and Correlative Analysis

IV. Instrumentation and Materials

A. Cell Culture for Correlative Light and Electron Microscopy Imaging

B. Live-Cell Microscopy and Confocal Laser Imaging

C. Sample Preparation for Relocation Studies by Electron Microscopy

D. 2-D and 3-D Transmission Electron Microscopy

E. Data Processing and Correlative Analysis

A. Making the Most of This Approach

apply high-quality EM sample preparation techniques to the sample. This knowledge

B. A 4-Dimensional Future

preparation protocols for electron microscopy. This approach allows researchers to

Visualizing Live Dynamics

One of the very effective methods to perform correlative light-electron microscopy

pattern. The process resembles taking a high-resolution snapshot of an interesting

I. Introduction and Rationale

Correlative microscopy emerged decades ago when morphologists started to com-

these transport organelles appeared as significantly larger structures, with a complex

(1) Thirty-five-millimeter, glass-grid-bottomed dishes (MatTek Corporation, P35G-

A. Cell Transfection, Observation, and Fixation

E. Serial-Section Analysis and 3D Reconstruction

more accurate reconstruction of 3D organization of organelles (with 3–4 nm resolution)

CLEM remains a unique technology to visualize structures of interest or specific/

Correlative Fluorescence and Transmission

d­ ramatic increases in the resolving power in fluorescence microscopy. Several differ-

II. Correlative Microscopy: Reporter Systems

Light from fluorescence sources, observable in the fluorescence microscope, is

E. Integrated Fluorescence and Electron Microscopy

III. Correlative Microscopy of Tissues

spaced membranes were often difficult to distinguish in confocal optical sections,

for 3-dimensional reconstruction is more readily achieved with resin-embedded tissue

C. Serial Block Face Scanning Electron Microscopy

A number of investigators have developed procedures to combine imaging systems,

Fluorescence-based imaging methods are widely used in cell biological research.

Correlative Light and Electron Microscopy

B. BSA-Au Labeling of the Endocytic Pathway

IV. Experiments and Materials

A. L. donovani Experiments

3. Macrophages Infection Assay

4. Fluorescence Confocal Microscopy

A. Need for Correlative Microscopy

B. Determining the Effects of Drugs on Cells: an Application for CLEM

A. Cell Culture for Correlative Light and Electron Microscopy Imaging

B. Live-Cell Microscopy and Confocal Laser Imaging

C. Sample Preparation for Relocation Studies by Electron Microscopy

D. 2-D and 3-D Transmission Electron Microscopy

E. Data Processing and Correlative Analysis

IV. Instrumentation and Materials

A. Cell Culture for Correlative Light and Electron Microscopy Imaging

B. Live-Cell Microscopy and Confocal Laser Imaging

C. Sample Preparation for Relocation Studies by Electron Microscopy

D. 2-D and 3-D Transmission Electron Microscopy

E. Data Processing and Correlative Analysis

A. Making the Most of This Approach

B. A 4-Dimensional Future

I. Introduction and Rationale

(1) Thirty-five-millimeter, glass-grid-bottomed dishes (MatTek Corporation, P35G-

A. Cell Transfection, Observation, and Fixation

E. Serial-Section Analysis and 3D Reconstruction

d ramatic increases in the resolving power in fluorescence microscopy. Several differ-

II. Correlative Microscopy: Reporter Systems

E. Integrated Fluorescence and Electron Microscopy

III. Correlative Microscopy of Tissues

C. Serial Block Face Scanning Electron Microscopy

B. BSA-Au Labeling of the Endocytic Pathway

IV. Experiments and Materials

A. L. donovani Experiments

3. Macrophages Infection Assay

4. Fluorescence Confocal Microscopy

6. Transmission Electron Microscopy

B. Plasmodium Berghei Experiments

3. Infected Cell Sorted by Flow Cytometry

4. Transmission Electron Microscopy

V. Results and Discussion

A. Fixation and PLT-Embedding of Samples for On-Section CLEM

B. General Labeling Protocol

8. 4 × 2 min PBS

1 9. Optional: after silver enhancement: 5 × 3 min water

C. CLEM of GFP-Labeled Rhodopsin in Retinas of Mice Using Prot A Gold

D. CLEM of α-Tubulin in the Microtubular Manchette of Mouse Spermatids

1. On-Section Labeling with Prot A Gold and Fluorochrome-Conjugated Antibodies

2. On-Section Labeling with IgG Gold and Fluorochrome-Conjugated Antibodies

3. On-section Labeling with FluoroNanogold

• Fixatives: For immunocytochemistry: 4% PFA in 0.1 M PB pH 7.4 (PB; 0.2 M

V. Summary and Outlook

1 2. 0.3M oxalic acid

C. Sectioning and Imaging Equipments

B. Preparation of Ultrathin Cryosections for HDO-CLEM

C. Immunofluorescence and Immunogold Labeling for HDO-CLEM (see Note 14)

1. Immunolabeling (see Note 15)

2. Double Labeling (see Fig. 1)

6. Incubate with PAG-15 nm for 20 min.

3. Preparing Grids for FLM Observations

4. Retrieval of Grids and EM Preparation

2. FLM Image Processing: Registering Serial ROIs

3. FLM Image Processing: 3D Surface Rendering

5. EM image Processing: Electron Tomography

6. CLEM image Processing

E. Hybrid Morphometry: A Systematic Error-Based Approach

1. FLM Image Analysis: Surface-Area Estimation

2. EM Image Analysis: Surface-Area Estimation

3. Calculating Systematic Error

1. Each objective lens is designed and assembled to achieve optimum performance

5. MosaicJ is a plugin designed to work in the ImageJ software package. The

A. Day 1––Cell Culture

B. Day 2––Fixation, Bleaching, and EM Preparation

C. Day 3––Electron Microscopy

D. Electron Tomography, 3D Reconstruction, and Modeling

B. Fixation, Bleaching, and EM Preparation

C. Electron Microscopy and Electron Tomography

II. The Crowded Cell and Spatiotemporal Proteomics

III. What EM Has to Offer

V. Genetically Appended or Inserted Protein Tags