Journal of Structural Biology 175 (2011) 62–72

Contents lists available at ScienceDirect

Journal of Structural Biology

journal homepage: www.elsevier .com/ locate/y jsbi

A new approach to improve the quality of ultrathin cryo-sections; its usefor immunogold EM and correlative electron cryo-tomography

Erik Bos a,b,d, Celso Sant́Anna b, Helmut Gnaegi c, Roberta F. Pinto b, Raimond B.G. Ravelli d,Abraham J. Koster d, Wanderley de Souza b, Peter J. Peters a,e,⇑a Netherlands Cancer Institute, Antoni van Leeuwenhoek Hospital, Amsterdam, The Netherlandsb Inmetro, Av. Nossa Senhore das Graças 50, CEP 25250-020, Duque de Caxias, Rio de Janeiro, Brazilc Diatome Ltd, Helmstrasse 1, CH-2560 Nidau, Switzerlandd LUMC, Postal zone S1-P, P.O. Box 9600, 2300 RC Leiden, The Netherlandse Kavli Institute of Nanoscience, Delft University of Technology, Delft, The Netherlands

a r t i c l e i n f o a b s t r a c t

Article history:Received 25 January 2011Received in revised form 30 March 2011Accepted 31 March 2011Available online 5 April 2011

Keywords:Cryo-sectionsCryo-ultramicrotomyImmunogold labelingVitrificationCorrelative electron cryo-tomography

1047-8477/$ - see front matter � 2011 Elsevier Inc. Adoi:10.1016/j.jsb.2011.03.022

Abbreviations: EM, electron microscopy.⇑ Corresponding author at: Division of Cell Biolo

Institute, Antoni van Leeuwenhoek, Hospital, Plesmadam, The Netherlands.

E-mail address: p.peters@nki.nl (P.J. Peters).URL: http://www.nki.nl/research/peters (P.J. Peter

Cryo-ultramicrotomy can be used to obtain ultrathin cryo-sections from cryo-fixed or aldehyde-fixedcryo-protected vitreous biologic samples. For immuno-gold EM, cryo-sections are retrieved from thecryo-chamber on a droplet of a pick-up solution (paste-like and almost frozen) to which the sectionsattach. The sections are then placed on an EM specimen grid at room temperature. This procedure com-promises the ultrastructure, resulting in folds, holes, and loss of the original material. In this paper weshow the critical influence of humidity, stretching, and relief of compression during thawing of the sec-tions. We show a new lift-up hinge device for semi-automated retrieval of cryo-sections that results insignificantly improved section quality. This approach was also applied successfully to vitreous sectionsfrom high pressure frozen samples. An important advance is that these vitreous cryo-sections can nowsuccessfully be post-fixed and immunolabelled after thawing; this allows cryo-EM comparison with adja-cent ribbons of sections still in the frozen hydrated state. These findings call for technical innovationsaiming at automated cryo-ultramicrotomy in a fully controlled environment for improved localizationof proteins within their ‘close to native’ cellular context and correlative electron cryo-tomography of con-secutive ribbons of sections of one frozen hydrated sample.

� 2011 Elsevier Inc. All rights reserved.

1. Introduction

The cellular nanocosm is made up of many kinds of macromo-lecular complexes or biologic nanomachines. Information on thestructure of these nanomachines has largely been obtained byanalyzing isolated structures, using mass spectrometry, X-raycrystallography, NMR, or single particle electron microscopy. Togain a better understanding of their functions, biologic complexesshould be imaged in a native state and within a cellular environ-ment. However, for the gene products that make up nanoma-chines, their proper localization within subcellular organellesneeds to be determined first by markers. These antigens can belocalized in their cellular context by several different immunoflu-

ll rights reserved.

gy, The Netherlands Cancernlaan 121, 1066 CX Amster-

s).

orescence techniques (Johnson and Nogueira Araujo, 1981), orusing constructs of the green fluorescent protein (GFP) family asmarkers for gene expression (Tsien, 1998, Watanabe et al.,2011). However, due to the limited resolution, many ultrastruc-tural details within the fluorescently-labeled nanomachines can-not be observed. Identification and observation of thesestructures in cells at higher resolution requires labeling experi-ments with electron dense nano particles and transmission elec-tron microscopy and even more advanced tomographicrecording with cryo electron microscopy. Thin samples are re-quired for clear visualization of cellular structures, and this canbe achieved by ultrathin sectioning in an ultramicrotome. Section-ing also gives antibodies access to the intracellular antigens. Thisavoids the use of detergents for membrane permeabilization,which affect ultrastructural integrity. Biologic samples are too softfor sectioning and need to be hardened. This can be done either byinfiltrating samples with resins, or by freezing. The demand forimmunolabeling limits the use of resin infiltration because thisfrequently hampers the accessibility of antibodies to the antigens.The efficiency of immunolabeling, the possibility of quantitation

E. Bos et al. / Journal of Structural Biology 175 (2011) 62–72 63

and of performing correlative electron cryo-tomography is mostprobably better on cryo-sections for the vast majority of antigens(Slot et al., 1989; Zeuschner et al., 2006; Slot and Geuze, 2007).

A major difficulty associated with freezing is that water in thesample crystallizes and can cause freeze damage and difficultieswith sectioning. It is possible, however, to freeze biologic samplesin a ‘close to native’ state (Vanhecke et al., 2008) with such a highcooling rate that the water solidifies without crystallization (vitri-fication) using high pressure freezing (Studer et al., 2001). The vit-rification depth in these samples is about 200 lm. Ultrastructuraldetails in sections from these cryo-fixed samples can be visualizedin a cryo EM (Dubochet et al., 1988), but cannot be immunolabeled,since the integrity of the sections is not maintained after thawing.

In order to produce cryo-sections suitable for immunolabelingthe samples are either aldehyde fixed, before cooling them down,or after high pressure freezing under native conditions (Van Dons-elaar et al., 2007). Chemical fixation also renders the sample suit-able for infiltration with a cryo-protectant such as sucrose, whichreduces the water content in the cell. As a result larger sample vol-umes can be vitrified using lower cooling rates. Typically, sucrosecryo-protected samples of 1–2mm3 can be vitrified by plungingin liquid nitrogen and successfully sectioned at low temperature(160 K). The integrity of subcelluar structures in cryo-sections fromthese samples is relatively well preserved upon warming-up due tothe chemical fixation. The thawed sections can be placed on an EMspecimen grid and immunolabeled at room temperature. By bind-ing an electron-dense gold nano particle to the antibody (Slot andGeuze, 1985) bound antigens in cryo-sections can be localized inthe EM. These sections need heavy metal staining to delineateultrastructural details. This technique, pioneered by Tokuyasu(1973) and refined by Slot and Geuze (2007) and their lab mem-bers, is often referred to as the ‘‘Tokuyasu technique’’. Based onmany years of experience it is by us and others considered a verysensitive method for detection of intracellular antigens by EM (Pe-ters and Pierson, 2008).

Successful immunogold labeling depends on the preservationand accessibility of antigens, but also on the preservation of cellu-lar ultra-structure, since well defined morphological details are re-quired to identify the location of immunogold labeled antigens.The quality of the immunogold-labeled section depends on severalfactors such as sample preparation, vitrification, ultrathin section-ing, and thawing of the sections. Vitreous sections of superior qual-ity can be obtained, but when thawing is carried out improperly,the end result shows poor ultrastructural detail. The original wayto thaw cryo-sections is to attach them to a sucrose pick-up solu-tion in a metal loop and subsequently withdraw them from thecryo-chamber (Tokuyasu, 1973). Methylcellulose was introducedto the sucrose solution to prevent overstretching (Liou et al.,1996). Exposed to the air the sections will thaw while floating onthe melting solution. At an early stage of this process the about-to-thaw sections will experience condensation of atmosphericwater that will freeze and thaw on their surface. Together withstretching due to the surface tension of the pick-up solution thisleads to damage of the sections. Although excellent results canbe obtained with this approach, sections typically show stretching,folds and ultrastructural damage. This results in difficulties inobserving and identifying structures associated with immunogoldlabel at high resolution. In an attempt to reduce ultrastructuraldamage we subjected thawing of sections to different approachesand present here a solution that may work in many cases. In addi-tion, we present the same approach for successful chemical fixa-tion of frozen hydrated sections from high pressure frozensamples, so that in future studies two consecutive ribbons of cryo-sections can be used for correlative immunogold-labeling and elec-tron cryo-tomography of vitreous sections, as recently furtherdeveloped by Pierson et al. (2010, 2011a,b).

2. Methods

2.1. Cells

The unicellular alga Ankistrodesmus sp. was acquired from a cul-ture collection of the Federal University of Rio de Janeiro (UFRJ).The original sample was obtained in Funil Reservoir, RJ, Brazil.The cells were cultured in ASM-1 medium (Gorham et al., 1964),pH 8.0, at 24 �C, under a photoperiod of 12 h at artificial illumina-tion intensity of 150 lmol photons m2 s�1. Epithelial cells (LLC-MK2) were cultured in DMEM. Other cells were cultured as de-scribed elsewhere: human umbilical vein endothelial cells (Valent-ijn et al., 2008), purified schizont stages of Plasmodium berghei(Janse et al., 2006), differentiated 3T3-L1 cells (adipocytes) (Xuand Kandror, 2002), 9L3.9 (Högemann-Savellano et al., 2003), Hu-man dendritic cells (van der Wel et al., 2007), brain (Mironov et al.,2003) and intestinal (Sato et al., 2009) tissue, Jurkat cells (Mercantiet al., 2010).

2.2. Vitrification of samples

Fixed cells and tissue were frozen in liquid nitrogen as de-scribed elsewhere (Peters et al., 2006). Briefly, the samples werefixed in PHEM buffer (Schliwa and van Blerkon, 1981) containingeither freshly prepared 2% formaldehyde (EM grade, EMS) or 2%formaldehyde and 0.2% glutaraldehyde (EM grade, EMS) for 18 or2 h, respectively. Brain tissue came from mice that were perfusedwith fixative. The fixed samples were embedded in 12% gelatin(type A, bloom 300, Sigma) and cut with a razor blade into cubes,0.5 mm3. The sample blocks were infiltrated in phosphate buffercontaining 2.3 M sucrose for 3 h. The sucrose-infiltrated sampleblocks were mounted on aluminum pins and plunged in liquidnitrogen. The frozen samples were stored under liquid nitrogen.High pressure freezing of unfixed material was performed as de-scribed elsewhere (Pierson et al., 2010). Briefly, the cells grownin standard growth medium with 20% dextran in suspension werespun down at 380 g and resuspended in growth medium contain-ing 20% dextran. This cell suspension was frozen in a copper tubeof 300 lm inner diameter in a high pressure freezer (EM pact II,Leica). The copper tubes containing frozen cells were stored underliquid nitrogen.

2.3. Cryo-ultramicrotomy

2.3.1. Conventional methodUltrathin sections were cut as described elsewhere (Peters et al.,

2006). Briefly, the frozen sample was mounted in a cryo-ultrami-crotome (Leica) at 158 K. The sample was trimmed to yield asquared block with a front face of about 200 � 200 lm. Using adiamond knife (Diatome) and antistatic devise (Leica) a ribbon of60 nm thick sections was produced that was retrieved from thecryo-chamber with a droplet of 1.15 M sucrose containing 1%methylcellulose. In some cases 2.3 M sucrose was used for sectionretrieval. After thawing the sections were transferred to a speci-men grid previously coated with formvar/carbon. In another seriesof experiments the frozen sections were attached to the specimengrid by charging (Pierson et al., 2010) and stored under liquidnitrogen.

2.3.2. Lift-up hinge methodThe handle for the metal wire loop and the lifting tool were

made by the instrumentation group of the Netherlands CancerInstitute. The lifting tool (Supplement, Movie 2; technical drawingavailable on request) was mounted on the cryo-chamber just be-fore sectioning. After sectioning, the ribbon of sections was

64 E. Bos et al. / Journal of Structural Biology 175 (2011) 62–72

attached to a nearly frozen droplet of 2.3 M sucrose in phosphatebuffer in a wire loop connected to a handle that contains a hinge.Bending the hinge the frozen droplet with the attached sectionswas flipped, such that the sections were facing upwards, and thehandle was fixed to the lifting tool. A formvar/carbon coated spec-imen grid was placed on top of the sections with the support filmdownwards. This system was lifted up by changing the position ofthe lifting tool, moving the sections and grid to a level just abovethe cryo-chamber (Fig. 5). This resulted in thawing of the sectionsand subsequent attachment to the carbon-coated formvar film onthe specimen grid. The grid with the sections was removed fromthe wire loop with a pair of tweezers. With this procedure a smallamount of sucrose remained attached to the grid and covered thesections, thus preventing drying out.

In some experiments a ribbon of vitreous sections from cellswas picked up together with a ribbon of sections from a vitrifiedsolution of 12% gelatin in phosphate buffer, such that one ribbonoverlapped the other. In this way part of the gelatin sections be-came ‘‘sandwiched’’ between the carbon-coated formvar film andcell sections.

Cryo-sectioning of frozen hydrated cells at 123 K and attach-ment of the sections to the specimen grid by charging was per-formed as described elsewhere (Pierson et al., 2010). Briefly, thecopper tube containing the cells was mounted in a cryo-ultrami-crotome that was covered by an anti-contamination glove box. Asquared block was trimmed to give a block face of 100 � 100 lmusing a diamond trimming tool (Diatome). Using a diamond knife(Diatome) and an anti static device (Leica) a ribbon of 60 nm sec-tions was produced. This ribbon of sections was then electro-stat-ically attached to a carbon-coated specimen grid (C-flat, EMS).Using the handle with hinge a droplet of 1% glutaraldehyde in2.3 M sucrose in PHEM buffer was frozen in a metal loop by press-ing the solution to a metal surface inside the cryo-chamber. Thiscreated a flat surface on one side of the frozen droplet, on top ofwhich the grid with the attached sections was positioned. Usingthe lift-up hinge, the sections were allowed to thaw and storedas described above.

2.4. Preparation for TEM observation

The sections attached to the grid and covered with pick-up solu-tion were floated on 2% gelatin in phosphate buffer for 30 min at37 �C. The sections were then rinsed on droplets of PBS at roomtemperature. In some cases incubation on 2% gelatin was skippedand the sections were placed directly on droplets of PBS at 4 �Cin order to preserve the 12% gelatin in the sections. Rinsing onPBS droplets was also skipped in the case of sections from frozenhydrated samples. Following rinsing on distilled water, grids withsections were placed on droplets of 2% methylcellulose in distilledwater containing 0.4% uranyl acetate at 4 �C for 5 min. The pH ofthe methylcellulose/uranyl acetate solution was previously ad-justed to between 4.5 and 5 to increase contrast. Each grid withsections was picked up from the methylcellulose/uranyl acetatedroplet with a metal wire loop. Excess of methylcellulose wasdrained off onto filter paper, after which the grid was allowed toair-dry. Grids with embedded sections were removed from thewire loops and stored in a grid box.

2.5. Transmission electron microscopy

Sections were examined in a Tecnai 12 transmission electronmicroscope (FEI) at 80 kV acceleration voltage. In some experi-ments, frozen sections from sucrose-protected samples were ob-served in a cryo-holder (Gatan) with low electron dose(20 e� �2 s�1) at 120 kV. Images were made with a FEI Eagle 4 Kdigital camera or with a Megaview 3 (Olympus SIS).

2.6. Measurements

Measurements were made on low magnification images of sec-tions from monocytes. The cells were assumed to be spherical. Foreach group, 20 cell profiles were measured on 4 different images.Stretching was estimated by measuring the diameter of the cellprofiles and calculating the average. For each cell, measurementswere made in a direction perpendicular to the cutting direction,which is not affected by compression, as well as parallel to the cut-ting direction, where compression is maximal. Compression wasestimated using the following equation: C = 100% � (L1�L2)/L1,where C = compression, L1 = cell profile length perpendicular tothe cutting direction, L2 = cell profile length parallel to the cuttingdirection. C was calculated for each cell profile and the average wasdetermined.

3. Results

3.1. Influence of humidity on section integrity

Thawed cryo-sections often show folds (Fig. 1a), which impairthe observation of ultrastructural details. The extent of foldingseems to depend on the type of sample. We carried out experi-ments where we picked up on one droplet two ribbons of sectionsfrom different samples and found that only one showed consis-tently severe folding (Supplement, Fig. 1). The extent of folding in-creased when the pick-up solution with the attached sections wasleft in the cryo-chamber for longer time. These findings suggestthat folding was not just related to inadequate picking up, but alsoto some additional factors. Evidence for such a factor came fromobserving under the stereo microscope that frozen sections be-came covered by frost followed by melted water when retrievedfrom the cryo-chamber (Supplement, Movie 1). To reduce deposi-tion of water vapor from saturated air during warming up, sectionswere allowed to thaw inside a small cup either held by hand ormounted on the cryo-chamber (Fig. 1a0). This approach eliminatedmost of the folds either with (Fig. 1b) or without (not shown) addi-tion of methylcellulose to the sucrose solution. Using the recentlydeveloped glove box that is filled with nitrogen gas to create a dryenvironment around the microtome (9) we easily obtained sec-tions with a minimum amount of folds similar to the results shownin Fig. 1. From these experiments we conclude that condensing ofatmospheric water on a frozen retrieved droplet resulted in ultra-structural damage.

3.2. Influence of stretching on ultrastructural integrity

Ultrathin sections that float on a pick-up solution stretch due tosurface tension. This stretching was shown for sections of mousebrain, where the width of the sections expanded to a size almosttwice the size of the trimmed sample block (Fig. 2a). Stretching af-fects the ultrastructural integrity (Fig. 3a and b) as shown by Liouet al. (1996). The effects of stretching were more severe on sectionsof mildly fixed samples such as adipocytes (Fig. 4a) and algae(Fig. 4c), which were fixed in formaldehyde without addition ofglutaraldehyde, resulting in mild fixation to preserve antigenicityof certain antigens. Observation under the stereo microscopeshowed that stretching during thawing took place in our setupwithin a few seconds (Supplement, Movie 1). This suggests thatthe section quality would benefit from a rapid transfer to the spec-imen grid. Therefore, the specimen grid was brought into closecontact with the sections while they were still frozen, such thatthe sections would attach to the grid immediately upon thawing.For this purpose a tool was devised that allowed the frozen dropletwith the attached sections to be inverted such that the grid could

Fig.1. Protection from moisture in the air during thawing reduces folding. Frozen sections of epithelial cells were picked up with sucrose/methylcellulose and allowed tothaw exposed to the air. Section profiles show many folds (a, arrows). To prevent precipitation of moisture from the air frozen sections from the same vitreous sample wereallowed to thaw on a droplet of sucrose/methylcellulose inside a small cup that was mounted on the cryo-chamber (a0). Thawing of the sections inside the cup resulted in flatcell profiles (b). Asterisks in b indicate cell nuclei. Bar is 5 lm.

E. Bos et al. / Journal of Structural Biology 175 (2011) 62–72 65

be placed on top of the frozen sections inside the cryo-chamber(Supplement, Movie 2). This tool consists of two components, ahandle for the attachment of the metal loop, and a lifting tool(Fig. 5). Using this tool, the grid and vitreous sections were liftedup to a level just above the cryo-chamber to initiate thawing. Thisapproach, referred to as the lift-up hinge method, considerably re-duced the stretching of sections from mildly fixed tissue (Fig. 2b).The lift-up hinge method also showed improved structural integ-rity for the afore-mentioned adipocytes and algae (Fig. 4b and d).In general, the cytosol matrix appeared denser with the lift-uphinge method (Figs. 3c and 4b and d), indicating that the cytosolexperienced less stretching. A quantitative relationship betweenreduction of stretching and improvement of ultrastructural integ-rity was found by measuring the size of cell profiles on sectionsover a length perpendicular to the cutting direction (Fig. 2c andd). With the lift-up hinge method, stretching was reduced by 21%compared to conventional picking up. From these experimentswe conclude that minimizing the stretching time reduces ultra-structural damage.

3.3. Relief of compression

Sections attached to the grid with the lift-up hinge method fre-quently showed electron-dense areas (Fig. 6a). The center of theseareas showed less contrast (Fig. 6a0) indicating that these areas didnot represent thicker section areas but wrinkles. These wrinkleswere different from the folds seen on sections thawed in a humidenvironment, e.g. wrinkles were always perpendicular to the cut-ting direction. These anisotropic wrinkles appeared after thawing,since they were not seen on vitreous sections in the cryo EM. Fur-thermore, when frozen sections were attached to the support filmby electrostatic adhesion (Pierson et al., 2010) and subsequentlythawed in a dry environment they did not show wrinkles. How-ever, wrinkles did appear when these thawed sections were ex-posed to a sucrose solution (Supplement, Fig. 2). Apparently, theanisotropic wrinkles represented an increase of section surfacearea resulting from relief of compression. This explanation wassupported by measuring the ratio of cell profile length (parallelto the cutting direction) and width (perpendicular to the cutting

direction), showing that with the lift-up hinge method sectionswere more compressed as compared to conventional picking uptechniques (Fig. 2c and d). A series of experiments showed thatthe extent of wrinkling depends on the quality of the surface ofthe support film. As shown here (Fig. 6b) we found that coatingof the formvar film with gelatin favored wrinkling. But we alsofound that wrinkling increased when carbon-coating of the form-var support film was omitted. The quality of the carbon layer aswell showed an effect, such that wrinkling was favored by glow-discharging or aging of the carbon layer. Moreover, we saw varia-tions between different samples. For instance, wrinkling was oftenmore pronounced on sections from tissues compared to sectionsfrom single cells (data not shown). These findings indicate that re-lief of compression was a result of poor attachment of the sectionsto the support film. We also found that wrinkling was favored bysectioning with a diamond knife having inferior gliding properties.In this respect, it is interesting that the periodicity of the wrinklescorresponded to the frequency of chatter (cyclic variations in sec-tion thickness) seen on frozen sections in the cryo-electron micro-scope (Fig. 6c and d). This frequency was 5000–7000 Hz. Thefrequency decreased when the section thickness increased (Sup-plement, Fig. 3) but was not affected by sectioning speed, samplesize, clearance angle and knife angle. From these experiments weconclude that anisotropic wrinkles, which frequently appear withthe lift-up hinge method, represented an increase of section sur-face area resulting from relief of compression caused by thesectioning.

3.4. Influence of methylcellulose on section integrity

Using the lift-up hinge method there was more space betweencell profiles and features like cell adhesion sites on the petri dishand the trimmed block edge could be recognized (Fig. 7b). Thistherefore truly represents a replica of the block-face of the sample.These features could not be observed with conventional picking upusing methylcellulose (Fig. 7a). To find an explanation for this phe-nomenon, sections of cells that were embedded in gelatin beforevitrification, were picked up either conventionally with sucrose/methylcellulose or with the lift-up hinge method using sucrose

Fig.3. Comparison of different picking up approaches. Sections from one vitreous sample of 9L3.9 cells were picked up conventionally (a and b) using sucrose (a) or sucrose/methylcellulose (b), or picked up using the lift-up hinge method (c). Sections picked up conventionally with sucrose show severe ultrastructural damage (a, arrows).Ultrastructural preservation is improved when methylcellulose is added to the sucrose, but there are still signs of damage due to overstretching (b, arrows). With the lift-uphinge method the cell content is more dense, membranes become better outlined, and there is less ultrastructural damage (c). N = nucleus, G = Golgi system, andM = mitochondrion. Bar is 500 nm.

Fig.2. Stretching of sections during thawing. Sections from mouse brain tissue were picked up conventionally (a) or with the lift-up hinge method (b). The width of thesample block was 150 lm. Unlike the lift-up hinge method, conventional picking up resulted in sections that stretched to a width of 300 lm. Arrows in a and b indicate thebeginning and ending of one section. In another experiment sections from dendritic cells were picked up conventionally (c), or with the lift-up hinge method (d). Bymeasuring the cell profile width (l1) it was found that the lift-up hinge method reduced stretching by 21% as compared to conventional picking up. This was quantified bycalculating the ratio of the cell profile length (l2) and width (l1). After section retrieval with the sucrose/lift-up hinge method, compression of 28% was still present.Compression was only 13% using conventional section retrieval. Bar is 100 lm (a and b) or 20 lm (c and d).

66 E. Bos et al. / Journal of Structural Biology 175 (2011) 62–72

without methylcellulose and processed for microscopy at 4 �C tokeep the gelatin in the sections. Unlike with the lift-up hinge meth-

od, sections picked-up conventionally with methylcelluloseshowed areas of heavily crumpled gelatin between the cell profiles

Fig.4. Improved ultrastructural integrity with the lift-up hinge method. Adipocytes (a and b) unicellular alga (c and d) were picked up conventionally (a and c) or with the lift-up hinge method (b and d). Conventional picking up resulted in poor ultrastructural preservation. Adipocytes show empty fat vacuoles (a, arrows). The alga shows emptystarch granules (arrows in c) and the cytosol and cell wall are disrupted (c, arrowheads). With the sucrose/lift-up hinge method the content of fat vacuoles (b, arrows) andstarch granules (d, arrows) is better preserved and the cells show overall improved ultrastructural integrity. Bar is 1 lm.

Fig.5. Demonstration of the lift-up hinge method. The device consists of a handle with a hinge to which the metal loop for picking up is connected (a), and a lifting tool thatholds the handle (a, arrow without text). The lifting tool is mounted on the cryo-chamber with a screw (b, arrow). Sections are attached to a frozen droplet of sucrose with thehinged loop down (c) after which the loop is turned 180� so that the sections face upwards. Then the hinge is bent 90� such that the sections are horizontal when the handle ismounted on the lifting tool (d). A formvar-coated grid is placed on top of the frozen sections with the aid of a pair of tweezers (e). The frozen sections with the grid on top arethen lifted up to initiate thawing (f). Bar is 3 cm (a, b, and f) or 3 mm (c, d, and e).

E. Bos et al. / Journal of Structural Biology 175 (2011) 62–72 67

(Fig. 7a0). Furthermore, methylcellulose affected the geometry ofloose tissues that were embedded in gelatin, making it difficultto recognize some morphological features (Fig. 7c). From theseexperiments we conclude that methylcellulose induced crumplingof gelatin, which resulted in packing and distortion of cell profiles.

3.5. Frozen hydrated sections

The principle of the lift-up hinge method was also tested on‘close to native state’ frozen hydrated sections of cells that werevitrified in copper tubes by high-pressure freezing. These cryo-sec-

Fig.6. Relief of compression associated with wrinkle formation with the lift-up hinge method. In some experiments 9L3.9 cells showed electron-dense areas (a, arrows). Lowcontrast in the middle (a0 , arrows) indicates that these areas are wrinkles. When half of the formvar support film was covered with gelatin, the periodicity and size of thewrinkles changed at the interface of formvar and gelatin (b, white line). The frequency of the wrinkles on thawed sections (d) was compared with the frequency of chatter onfrozen sections from the same sample seen in the cryo TEM (d). This showed similar periodicities (c and d, vertical bars, indicating the length of two periods). Sectioning speedwas 2.5 mm/s. Bar is 1 lm.

Fig.7. Demonstration of the crumpling effects of methylcellulose. Jurkat cells (a and b) and intestine (c and d) were embedded in gelatin before sucrose infiltration andsections were picked up in a conventional way with sucrose/methylcellulose (a and c) or with the lift-up hinge method (b and d). Conventional picking up with sucrose/methylcellulose resulted in packing of cell profiles (a) as compared to the lift-up hinge method (b). Sections picked up conventionally showed heavily crumpled gelatinbetween the cell profiles when they were processed at 4 �C for EM observation (a0 , arrows). With the lift-up hinge method the trimmed edge on cell profiles (b, arrowheads),as well as adhesion sites where the cells were attached to the culture plate (b, arrows) can be distinguished. Also sections from mouse intestine tissue crumpled when pickedup with sucrose/methylcellulose (c). An open duct clearly visible with the lift-up hinge method (arrows in d) appears as a blocked duct when using conventional picking up (c,arrows). Bar is 10 lm.

68 E. Bos et al. / Journal of Structural Biology 175 (2011) 62–72

tions were first electro-statically attached to the EM specimen gridfor electron cryo-tomography (Pierson et al., 2010). In this study aglutaraldehyde/sucrose solution was pre-frozen in the metal loopand the grid with the sections was put on top of the frozen droplet

such that the sections were facing the frozen glutaraldehyde/su-crose solution. Using the lifting tool the sections were raised to alevel just above the cryo-chamber to initiate thawing. This resultedin chemical fixation of the sections on the EM grid (Fig. 8). Cell

Fig.8. Chemical fixation of frozen hydrated sections. Frozen hydrated sections of rat erythrocytes infected with P. berghei (a) and endothelial cells (b) were fixed on a frozendroplet of sucrose solution containing 1% glutaraldehyde using the handle with hinge and lifting tool. After thawing the sections were embedded in methylcellulose anduranyl acetate. P = parasite, H = hemozoin crystals, G = Golgi system, N = nucleus, M = mitochondrion, and W = Weibel-Palade body. Bar is 500 nm.

E. Bos et al. / Journal of Structural Biology 175 (2011) 62–72 69

profiles showed ultra-structural details of similar quality com-pared to sections from chemically fixed cells. From these experi-ments we conclude that retrieval with the semi-automated lift-up hinge method results in better stabilization of frozen hydratedsections on a pick-up droplet.

4. Discussion

Ultrathin sectioning of sucrose cryo-protected vitrified samplesand subsequent thawing are critical steps in cryo-immunogold EMlabeling. Whereas other steps in this process, i.e. vitrification of thesamples and staining of ultrathin sections on a specimen grid, aremore straightforward, sectioning and thawing require a high levelof experience. Successful ultrathin sectioning requires a well per-forming cryo-ultramicrotome, a perfect diamond knife, and, usu-ally, an antistatic device. With these tools and proper handling,ultrathin sections of excellent quality can be obtained from chem-ically fixed vitrified samples. Nevertheless, experienced cryo-ultra-microtomists find themselves sometimes surprised by the poorquality of thawed sections. Perfectly flat sections seen in the micro-tome appear terribly affected by folds and ultrastructural damagein the electron microscope. Several factors seem to play a role inthe prevalence of thawing artifacts seen with conventional pickingup. First, the temperature of the pick-up droplet at the moment itcontacts the frozen sections seems to affect the extent of folds onthe thawed sections. The right timing is difficult to indicate andhas to be found by trial and error since section pick-up is not auto-mated yet. Second, the composition of the pick-up solution affectsultrastructural integrity. A 2.3 M sucrose solution is often used forsection retrieval since it becomes vitreous upon freezing in thecryo-chamber. This approach originates from the pioneering workof Tokuyasu (1973). Addition of methylcellulose results in betterultrastructural preservation (Liou et al., 1996), but the pick-up solu-tion crystallizes quickly when cooled down and is therefore moredifficult to handle. Third, successful thawing depends on the typeof sample. When different cell types are subjected to identical fixa-tion and vitrification protocols, they may show variations in the ex-tent and frequency of thawing artifacts.

A persistent artifact seen on thawed cryo-sections with con-ventional picking up is folding. Folds can cause problems byoverlapping structures of interest and accumulating immunogoldlabel. Moreover, the increased electron density of folds results in

dark areas. The finding that humidity has an influence on sectionintegrity may explain why folds appeared in an unpredictablemanner. Humidity in the lab changes from day to day and isnever as low as desired for cryo-sectioning. In addition, theamount of moisture that precipitates on the sections dependson the temperature of the frozen pick-up solution. Thus, the ex-tent of folds increases when the pick-up droplet with the at-tached sections is left in the cryo-chamber for extended periodsof time. An almost frozen droplet when picking up the sectionsand quick withdrawal from the cryo-chamber as soon as the sec-tions are attached to the pick-up solution will help to reducefolds. But the best way to reduce folds is to protect them fromprecipitation of moisture, as shown in this study. We initiallymounted a small cup of about 2 ml on the cryo-chamber. We ob-served that sections picked up conventionally with sucrose be-came attached to the surface of the frozen droplet and could bedetached easily with an eyelash (Supplement, Movie 3), whereaswith sucrose/methylcellulose, one side of the sections becamefirmly embedded in the frozen droplet (Supplement, Movie 4).In both cases we see that there are gaps between non-flat sectionareas and the surface of the frozen solution. To explain the effectof moisture on folding we assume that moisture from the airaccumulates over these gaps and cause the non-flat areas to col-lapse, creating folds. The formation of gaps may depend upon theflatness of the sections as well as the timing of picking up. Thaw-ing in a dry environment reduces folding and eliminates the fac-tor of timing for successful picking up. Another argument forthawing in a dry environment is that the deposits of minute icecrystals (microscopic frozen rain) that form when atmosphericwater vapor precipitates results in a melted layer of water ontop of the sections. This water is absorbed by the sections dueto the high sucrose concentration in both the sections and thepick-up solution. The resulting water transport through the sec-tion may facilitate extraction of weakly fixed cellular componentsfrom the section into the pick-up solution.

A better way to eliminate folds is represented by the lift-uphinge method, where the grid with the support film covers thesections during thawing. With this approach we see a minimalamount of folding. This approach also results in less ultrastruc-tural damage. Ultrastructural damage seen with conventionalpicking up results from overstretching and impedes the observa-tion of structures associated with immunogold label. Overstret-ching is caused by the surface tension of the pick-up solution

70 E. Bos et al. / Journal of Structural Biology 175 (2011) 62–72

and leads to a loss of coherence in membranes and other cellstructures. The longer sections are exposed to surface tensionthe more they stretch and suffer damage of fine details. With con-ventional picking up it is therefore important to attach the sec-tions to the support film as fast as possible upon thawing. Agood moment for transfer to the grid is when the surface of themelting solution becomes shiny. However, even then the timeinterval where surface tension may act on the sections is stilllarge as judged by the ultrastructural damage. Therefore, we usedanother approach, where the sections were allowed to thaw withthe grid on top of them. In this arrangement the sections attachto the support film immediately upon thawing. However, the gridon top of the frozen droplet is an unstable situation and the grideasily falls off when lifted up. To face this problem we devised ahandle with a hinge and a lifting tool that guides the withdrawalof the frozen droplet and grid from the cryo-chamber in a smoothmanner. This semi-automated approach shows remarkable re-sults. The cytosol appears denser and fine membrane structuresare better outlined. Sections from samples that were thought tobe of poor fixation quality now show beautiful sections. Theimprovements with the lift-up hinge method are apparently theresult of a reduction of the stretching time. The larger surfacearea of cell profiles seen with conventional picking up is due toincreased spacing between cell components. The looser cytosolmatrix facilitates loss of cellular material such as is often seenat the sites of endosomes, multi-vesicular bodies, the Golgi com-plex and TGN. Overall we see better definition of negativelystained membranes with the lift-up hinge method., This is prob-ably due to a higher density of proteins and consequently a den-ser uranyl acetate staining. Good definition of membranes is alsoaffected by section thickness. When the stretching time is re-duced with the lift-up hinge method thinner sections (silverinterference color, instead of light yellow) can be cut. Thinnersections show less overlap of structures and therefore revealmore fine details.

Although reduction of the stretching time results in improvedultrastructural integrity, we still see signs of ultrastructural dam-age. It is possible that stretching is not the only mechanical forcethat acts on the sections. Perhaps crystallization of the vitreouswater present in the sections also plays a role. We found that adroplet of sucrose solution vitrified by plunging in liquid nitro-gen, transforms into crystalline ice before melting. Also sucrose-protected samples vitrified by plunging in liquid nitrogen andsubsequently immersed in warm sucrose re-crystallize beforethawing. These observations indicate that the sucrose-infiltratedvitreous sample contains microscopic ice crystals. Perhapsultrathin sections experience ice crystal growth during warmingup, which may result in mechanical damage to some cellularstructures.

Reduction of the stretching time with the lift-up hinge methodimproves preservation of cellular ultrastructure, but also intro-duces a new artifact, namely anisotropic wrinkles. The periodicityof the wrinkles corresponds to high frequency chatter sometimesseen on frozen sections in the cryo electron microscope, whichindicates a causal relationship. The frequency of 5000–7000 Hz isin the range of a fricative sound (Ladefoged, 2005) and is muchhigher than is observed occasionally when vibrations are gener-ated by the instrument itself. This suggests that it arises from fric-tion between the sample and the knife surface. We presume thatduring sectioning the nascent section initially sticks to the knife-edge. When the knife intrudes further into the sample, the sectionmaterial will become compressed while stuck to the knife. Thiscontinues until the build up pressure forces the section to slip, thuscompleting one period in the frequency pattern. This stick and slippattern on a section may affect the surface of the section in its abil-ity to attach firmly to the support film. Since sections become com-

pressed during sectioning (Studer and Gnaegi, 2000), we presumethat the observed wrinkles represent section areas that are notwell attached to the support film and that at these sites relief ofcompression occurs. Relief of compression restores the plasticdeformation that occurs when sections are compressed and there-fore would result in stretching. From this point of view perfect-gliding properties of the knife are mandatory for firm attachmentof the sections to the support film. Perhaps an oscillating knife(Studer and Gnaegi, 2000) that works at cryogenic conditionsmay help to study improved gliding and reduce section compres-sion. Firm attachment necessary to hold compression is also influ-enced by the properties of the support film. Relief of compressionalone would not result in negative effects on the ultrastructure if itwere assumed that compression is a plastic deformation. But reliefof compression seen with the lift-up hinge method is not homoge-neous and therefore undesired. The resulting wrinkles may impedeobservation of affected details on the section. Wrinkles due to re-lief of compression should not be confused with folds that resultfrom thawing in a humid environment. Although both representnon-flat section areas, the findings described here indicate thatthey are invoked by different mechanisms. Anisotropic wrinklesrepresent relief of compression induced by a repeating pattern ofthick and thin section areas (chatter). Folds represent crumplingof sections resulting from non-flat section areas induced by mois-ture from the air.

By measuring the ratio of cell profile length and width we foundthat with conventional picking up cell profiles are less compressedas compared to the lift-up hinge method. This indicates that sec-tions experience both relief of compression and stretching whenfloating on a melting solution. These processes will occur simulta-neously and their dissociation is usually not observed. With the re-sults obtained here dissociation is needed to describe thecoincidence of wrinkling and improved ultrastructural preservationas seen with the lift-up hinge method (Fig. 9). Both relief of com-pression and overstretching contribute to an increase of the sectionsurface area, but ultrastructural damage should be attributedmainly to overstretching. In the present work we show that overst-retching can be minimized with the semi-automated lift-up hingemethod. With this approach it is important that sections attachfirmly to the support film in order to maintain compression, other-wise contact with the melting solution will induce wrinkles. Thetype of sample, the gliding properties of the knife and the surfaceproperties of the support film affect firm attachment. We see fewerwrinkles when carbon coating of the support film is done just be-fore use and without glow-discharging.

Addition of methylcellulose to the sucrose solution for conven-tional picking up results in better ultrastructural preservation, butalso introduces a persistent artifact, which is folding. When foldsappear on a section they are usually numerous and affect most ofthe cell profiles. This crumpling of sections is probably the resultof heterogeneous consistencies in the sample and of the way sec-tions become embedded in the pick-up solution as describedabove. As shown here, addition of methylcellulose also results incrumpling of the gelatin present in between the fixed cells. The un-fixed gelatin is used to give support to pellets of single cells in or-der to facilitate cutting of the pellet into small blocks using a razorblade (Peters et al., 1991). Gelatin is also used to fill up clefts andcavities in blocks of tissue to improve sectioning. The gelatin dis-solves at room temperature when the sections are processed forimmunogold labeling before microscopical observation. As shownhere, severe crumpling of gelatin seen with conventional pickingup is responsible for distortion of cell profiles and affects their spa-tial distribution in the section.

The mechanism responsible for reduction of stretching whenmethylcellulose is used is not well understood. A possible explana-tion is that the forces responsible for gelatin crumpling also act on

Fig.9. Schematic presentation of the dissociation of relief of compression andoverstretching. The dark gray area (I) represents a section cut from a sphericalobject. When the section is thawed on a pick-up solution the section will stretch inthe direction parallel to the cutting direction (solid arrows) due to relief ofcompression, and the surface area of the section will become the sum of the darkgray area (I) and the white area (II). The surface area of the section will also expandin all directions (dashed arrows) as a result of overstretching represented by thelight gray area (III). With the lift-up hinge method relief of compression andoverstretching are reduced. When attachment to the support film is not optimal thesection surface area represented by the white area (II) will be seen as anisotropicwrinkles.

E. Bos et al. / Journal of Structural Biology 175 (2011) 62–72 71

the more rigid fixed cellular material and that it counteracts thestretching induced by surface tension, thus stabilizing cell struc-tures. This crumpling became very noticeable when a sample ofpure gelatin without cells was cut and observed (not shown). Theadvantage of methylcellulose is no longer present with the lift-uphinge method. In fact we found that, picking up with sucrose usingthe lift-up hinge method results in better preservation of ultra-structural details as compared to conventional picking up usingmethylcellulose/sucrose mixture. Apparently, the reduced timeinterval between thawing and attachment to the grid leaves littlespace for surface tension to act on the sections, and the stabilizingeffect of methylcellulose becomes redundant. Omission of methyl-cellulose gives the additional advantages that the solution freezesslower, thus facilitating picking up, and that it results in less frostprecipitation on the knife.

5. Conclusions

We studied several factors that have a negative effect on thepreservation of the ultrastructure of biologic cryo-sections uponthawing and came to the following conclusions:

(1) Thawing of cryo-sections on a droplet in a dry environmentwith the lift-up hinge method improves the ultrastructuralintegrity of sections.

(2) The time interval between thawing and subsequent attach-ment of the sections to the support film should be mini-mized to reduce ultrastructural damage.

(3) The gliding properties of the knife and the adhesion proper-ties of the support film should be optimized to reduce wrin-kles resulting from relief of compression.

(4) Sections from gelatin-embedded samples or tissues thathave a lot of collagen should be retrieved without methylcel-lulose in the pick-up solution if the original spatial distribu-tion of the cells (stereology and quantitive evaluation) has tobe preserved.

6. Future outlook: Frozen hydrated sections for immunogoldlabeling

The improved quality of the sections with the lift-up hingemethod opened the door to chemical fixation of frozen hydratedsections from high pressure frozen samples. Earlier attempts madeshowed unsatisfactory results when frozen hydrated sections werepicked up with about-to-freeze fixation solutions (Möbius et al.,2002; van Donselaar et al., 2007). We found that reduction of thestretching time results in retrieval of well preserved ultrastructuraldetails of samples that exhibited bad ultrastructure with conven-tional picking up due to weak fixation. Good quality of ultrastruc-ture shows that chemical fixation of the vulnerable frozenhydrated sections may benefit from a reduction of the stretchingtime using the lift-up hinge method, in a way that this helps tomaintain coherency of thawed cellular structures while they arein the process of cross linking through contact with the melting fix-ative solution. In some experiments ultrastructural preservationwas of less quality indicating that the protocol for chemical fixa-tion of frozen hydrated sections needs fine-tuning, for instancewith the type and concentration of fixatives. With this work it be-comes evident that automated ultrathin cryo-sectioning in anatmosphere-controlled cryo-chamber with robotic devices hookedup to software such as from the computer game entertainmentindustry are essential to get the highest preservation. This wouldgive us the opportunity to correlate two consecutive ribbons ofcryo-sections, one for immunogold label followed by room temper-ature EM or vitrification of the immunolabeled cryo-section fol-lowed by cryo light microscopy and electron cryo-tomographyand the other for direct electron cryo-tomography of the same cell.Options for such correlative approaches are currently being inves-tigated in our laboratories.

Acknowledgments

The authors thank Sannie Kraan (NKI-AVL, Amsterdam) for herpractical ideas for the lifting tool, Jason Pierson for help with cryoelectron microscopy, Isabel Porto-Carreira (UFRJ, Rio de Janeiro) forsharing her tools, Chris Janse and Blandine Franke-Fayard (LUMC,Leiden) for kindly providing the infected erythrocytes, ElizabethRibeiro da Silva (Centro de Microscopia da UFMG, Belo Horizonte)for kindly providing the LLC-MK2 cells, Sandra Azevedo (UFRJ, Riode Janeiro) for kindly providing the unicellular alga cultures, Vâniada Silva Vieira and Paulo Freitas-Junior (Inmetro, Rio de Janeiro) fortheir support in the lab, Luis Sérgio Cordeiro (Inmetro, Rio de Ja-neiro) for help with the microscope, Sue Godsave (NKI-AVL), KentMcDonald (UC Berkely) and Carmen Lopez-Iglesias (Scientific Parkof Barcelona) for critical reading of the manuscript, and KiyoteruTokuyasu (UC San Diego) and all members of the Peters lab forhelpful discussions. We also appreciate the feedback from theDutch cryo EM team organized in www.necen.nl for their feedback.This work was supported by FAPERJ/Inmetro.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.jsb.2011.03.022.

References

Dubochet, J., Adrian, M., Chang, J.J., Homo, J.C., Lepault, J., et al., 1988. Cryo-electronmicroscopy of vitrified specimens. Q. Rev. Biophys. 21 (2), 129–228.

Gorham, P.R., McLachlan, J., Hammer, U.T., Kim, W.K., 1964. Isolation and culture oftoxic strains of Anabaena flosaquae (Lyngb.). Breb. Verh. Int. Ver. Limnol. 15,796–804.

72 E. Bos et al. / Journal of Structural Biology 175 (2011) 62–72

Högemann-Savellano, D., Bos, E., Blondet, C., Sato, F., Abe, T., et al., 2003. Thetransferrin receptor: a potential molecular imaging marker for human cancer.Neoplasia 5 (6), 495–506.

Janse, C.J., Ramesar, J., Waters, A.P., 2006. High-efficiency transfection and drugselection of genetically transformed blood stages of the rodent malaria parasitePlasmodium berghei. Nat. Protoc. 1 (1), 346–356.

Johnson, G.D., Nogueira Araujo, G.M., 1981. A simple method of reducing the fadingof immunofluorescence during microscopy. J. Immunol. Methods 43 (3), 349–350.

Ladefoged, P., 2005. Vowels and Consonants: An Introduction to the Sounds ofLanguages, vol. 1. Wiley-Blackwell, Oxford.

Liou, W., Geuze, H.J., Slot, J.W., 1996. Improving structural integrity of cryosectionsfor immunogold labeling. Histochem. Cell Biol. 106 (1), 41–58.

Mercanti, V., Marchetti, A., Lelong, E., Perez, F., Orci, L., et al., 2010. Transmembranedomains control exclusion of membrane proteins from clathrin-coated pits. J.Cell Sci. 123, 3329–3335.

Mironov Jr., A., Latawiec, D., Wille, H., Bouzamondo-Bernstein, E., Legname, G., et al.,2003. Cytosolic prion protein in neurons. J. Neurosci. 23 (18), 7183–7193.

Möbius, W., Ohno-Iwashita, Y., van Donselaar, E.G., Oorschot, V.M., Shimada, Y.,et al., 2002. Immunoelectron microscopic localization of cholesterol usingbiotinylated and non-cytolytic perfringolysin O. J. Histochem. Cytochem. 50 (1),43–55.

Peters, P.J., Borst, J., Oorschot, V., Fukuda, M., Krähenbühl, O., Tschopp, J., et al., 1991.Cytotoxic T lymphocyte granules are secretory lysosomes, containing bothperforin and granzymes. J. Exp. Med. 173 (5), 1099–1109.

Peters, P.J., Bos, E., Griekspoor, A., 2006. Cryoimmunogold electron microscopy.Curr. Protoc. Cell Biol. (Chapter 4:Unit 4.7).

Peters, P.J., Pierson, J., 2008. Immunogold labeling of thawed cryosections. MethodsCell Biol. 88, 131–149.

Pierson, J., Fernández, J.J., Bos, E., Amini, S., Gnaegi, H., et al., 2010. Improving thetechnique of vitreous cryosectioning for cryoelectron tomography: electrostaticcharging for section attachment and implementation of an anti-contaminationglove box. J. Struct. Biol. 169 (2), 219–225.

Pierson, J., Ziese, U., Sani, M., Peters, P.J., 2011a. Exploring vitreous cryo-section-induced compression at the macromolecular level using electron cryo-tomography; 80S yeast ribosomes appear unaffected. J. Struct. Biol.doi:10.1016/j.jsb.2010.09.017.

Pierson, J., Vos, M., McIntosh, J.R., Peters, P.J., 2011b. Perspectives on electron cryotomography of vitreus cryo sections. J. Elect. Microsc.

Sato, T., Vries, R.G., Snippert, H.J., van de Wetering, M., Barker, N., et al., 2009. SingleLgr5 stem cells build crypt-villus structures in vitro without a mesemchymalniche. Nature 459, 262–266.

Schliwa, M., van Blerkon, J., 1981. Structural interaction of cytoskeletal components.J. Cell Biol. 90, 222–235.

Slot, J.W., Geuze, H.J., 1985. A new method of preparing gold probes for multiple-labeling cytochemistry. Eur. J. Cell Biol 38 (1), 87–93.

Slot, J.W., Geuze, H.J., 2007. Cryosectioning and immunolabelling. Nat. Protoc. 2(10), 2489–2491.

Slot, J.W., Posthuma, G., Chang, L.Y., Crapo, J.D., Geuze, H.J., 1989. Quantitativeaspects of immunogold labeling in embedded and in nonembedded sections.Am. J. Anat. 185 (2–3), 271–281.

Studer, D., Gnaegi, H., 2000. Minimal compression of ultrathin sections with use ofan oscillating diamond knife. J. Microsc. 197 (Pt 1), 94–100.

Studer, D., Graber, W., Al-Amoudi, A., Eggli, P., 2001. A new approach forcryofixation by high-pressure freezing. J. Microsc. 203 (Pt 3), 285–294.

Tokuyasu, K., 1973. A Technique for ultracryotomy of cell suspensions and tissue. J.Cell Biol. 57 (2), 551–565.

Tsien, R.Y., 1998. The green fluorescent protein. Annu. Rev. Biochem. 67, 509–544.Valentijn, K.M., Valentijn, J.A., Jansen, K.A., Koster, A.J., 2008. A new look at Weibel-

Palade body structure in endothelial cells using electron tomography. J. Struct.Biol. 161 (3), 447–458.

Van der Wel, N., Hava, D., Houben, D., Fluitsma, D., van Zon, M., et al., 2007. M.tuberculosis and M leprae translocate from the phagolysosome to the cytosol inMyeloid cells. Cell 129, 1287–1298.

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.

Vanhecke, D., Graber, W., Studer, D., 2008. Close-to-native ultrastructuralpreservation by high pressure freezing. Methods Cell Biol. 88, 151–164.

Watanabe, S., Punge, A., Hollopeter, G., Willig, K.I., Hobson, R.J., et al., 2011. Proteinlocalization in electron micrographs using fluorescence nanoscopy. Naturemethods 8, 80–84.

Xu, Z., Kandror, K.V., 2002. Translocation of small preformed vesicles is responsiblefor the insulin activation of glucose transport in adipose cells. Evidence from thein vitro reconstitution assay. J. Biol. Chem. 277, 47972–47975.

Zeuschner, D., Geerts, W.J., van Donselaar, E., Humbel, B.M., Slot, J.W., et al., 2006.Immuno-electron tomography of ER exit sites reveals the existence of freeCOPII-coated transport carriers. Nat. Cell Biol. 8 (4), 377–383.