MICROSCOPY RESEARCH AND TECHNIQUE 42:4358 (1998)

Pre-Embedding Immunolabeling for Electron Microscopy: An Evaluation of Permeabilization Methods and Markers
BRUNO M. HUMBEL,* MARGO D.M. DE JONG, WALLY H. MULLER, AND ARIE J. VERKLEIJ
Department of Molecular Cell Biology, Institute for Biomembranes, Utrecht University, Utrecht, The Netherlands

KEY WORDS

Brij; colloidal gold; diaminobenzidine; uorochrome; peroxidase; Saponin; streptolysin; Triton; tyramide; ultrasmalls gold

ABSTRACT For scarce antigens or antigens which are embedded in a dense macromolecular structure, on-section labeling, the rst method of choice, is not always successful. Often, the antigen can be localized by immunouorescence microscopy, usually by a pre-embedding labeling method. Most of these methods lead to loss of ultrastructural details and, hence, labeling at electron microscope resolution does not add essential information. The scope of this paper is to compare ve permeabilization methods for pre-embedding labelling for electron microscopy. We aim for a method that is easy to use and suitable for routine investigations. For our ongoing work, special attention is given to labeling of the cell nucleus. Accessibility of cytoplasmic and nuclear antigens is monitored with a set of different marker antibodies. From this investigation, we suggest that prexation with formaldehyde/glutaraldehyde is necessary to stabilize the ultrastructure before using a detergent (Triton X-100 or Brij 58) to permeabilize or remove the membranes. The experimental conditions for labeling should be checked rst with uorescence or uorescence-gold markers by uorescence microscopy. Then either ultrasmall gold particles (with or without uorochrome) with silver enhancement or, if the ultrasmall gold particles are obstructed, peroxidase markers are advised. The most promising technique to localize scarce antigens with good contrast is the combination of a pre-embedding peroxidase/tyramide-FITC or -biotin labeling followed by an on-section colloidal gold detection. 1998 Wiley-Liss, Inc. Microsc. Res. Tech. 42:4358, 1998. INTRODUCTION In cell biology and biomedical sciences, immunolabeling is an important tool to study the relationship of structure and function. For light microscopy, often a pre-embedding immunolabeling method is followed either on permeabilized cells, thawed cryostat sections, rehydrated paraffin sections, or deplasticized methacrylate sections. This is true even for samples to be analyzed by confocal (scanning laser) light microscopy. The location of the protein in question is visualized either by enzyme reaction products, e.g., peroxidase/ diaminobenzidine, silver-enhanced gold particles, or uorescent labels. In general, the uorescence label gives a more brilliant picture; often it also seems to be more sensitive and is the method of choice for confocal light microscopy. The other markers have the advantage that the biological material can be counterstained, hence the label and the cellular structure can be seen simultaneously. Further, these preparations can be stored without fading of the label. Usually, the results of such pre-embedding techniques serve as the basis for subsequent electron microscope investigations. In electron microscopy, post-embedding immunolabeling methods, i.e., on-section labeling, are preferred. For on-section labeling, the material can be prepared with optimum preservation of the cellular ultrastructure if, e.g., preparation techniques based on cryoxation are used (Hohenberg et al., 1994; Humbel et al., 1983; Humbel and Muller, 1986; Menco, 1986; Muller and Moor, 1984; Robards and Sleytr, 1985; Steinbrecht and
1998 WILEY-LISS, INC.

Zierold, 1987; Studer et al., 1995). Gold markers are now preferred for electron microscope localization studies (Hayat, 19891991; Roth, 1989; Verkleij and Leunissen, 1989) because they are clearly distinguishable from biological structures. Electron microscopic results, however, often do not reect the label intensity expected from light micrographs. In pre-embedding methods, the structures are accessible in three dimensions, whereas on sections only the section surface is exposed to the antibodies (Stierhof et al., 1986). Further, uorescent antibodies seem to be more efficient in labeling than gold-tagged antibodies, even if the identical on-section preparation is used, e.g., cryosections (according to Tokuyasu, 1973, 1986) either mounted on coverslips and labeled with uorescent antibodies or mounted on grids and labeled with colloidal gold particles (Humbel, unpublished results). Figure 1 shows an example of ultrathin resin sections of Leishmania mexicana (Stierhof et al., 1991b) labeled for tubulin and either detected with a uorescent secondary antibody or the identical uorescent antibody additionally tagged with ultrasmall gold particles. The gold-tagged antibody exhibits a less intense and a more patchy staining (Fig. 1: Humbel and Stierhof, unpublished). The size of the gold particles

*Correspondence to: Bruno M. Humbel, Department of Molecular Cell Biology, Utrecht University, Padualaan 8, NL-3584 CH Utrecht, The Netherlands. E-mail: bruno@accu.uu.nl Received 18 March 1998; Accepted 1 April 1998

44

B.M. HUMBEL ET AL.

Fig. 1. Ultrathin sections of Leishmania mexicana xed with 2% (w/v) formaldehyde and 0.05% (v/v) glutaraldehyde and embedded in Lowicryl HM20 (Stierhof et al., 1991b; gift of Dr. Y.-D. Stierhof) either mounted on glass coverslips or grids, labeled for tubulin. The primary antibody was detected with a goat anti-mouse FITC antibody (A) or with the same antibody to which ultrasmall gold particles were bound (B,C). The labeling properties were imaged by uorescence microscopy (A,B) or after silver enhancement by electron microscopy (C). Note that the uorescence signal of the Au-FITC antibody (B) is less intense than of the identical FITC-antibody without gold tag (A). Furthermore, the labeling pattern is more patchy, which correlates well with the electron micrograph (C). These are unpublished results of Humbel and Stierhof (1997). A, B: Bar 10 m; C: bar 1 m.

also seems to be important for labeling (Humbel and Biegelmann, 1992). The higher label density of ultrasmall gold compared to larger gold particles suggests that the gold particles repulse each other. The larger the particles, the stronger the repulsion force. In pre-embedding methods, the cells have to be permeabilized either mechanically by sectioning or freezing/thawing, or chemically by pore-forming proteins, detergents, or organic solvents to give the mark-

ers access to cytoplasmic or nuclear antigens. During permeabilization processes the cellular membranes are perforated or completely removed and an uncontrolled number of so-called soluble proteins may be redistributed within the cell (Melan and Sluder, 1992) or lost. The consequence is a false labeling pattern or the protein of interest escapes detection. On the other hand, the loss of a part of the soluble proteins may uncover an epitope of a structural protein resulting in a positive label. As an example, a lamin A/C antibody only labeled the nuclear lamina of Triton X-100 permeabilized cells (De Graaf et al., 1991) and not on Tokuyasu cryosections (Tokuyasu, 1973, 1986) (Humbel, unpublished observations). For in situ hybridization studies, three-dimensional access to nucleic acids is very important. High copy nucleic acids like 28S rRNA can clearly be labeled on sections (Sibon et al., 1995), whereas others are more difficult or even impossible to detect, as, e.g., the EGF-receptor transcript (Sibon et al., 1994). Hence, although on-section labeling approaches are preferred due to the superiority in preservation of the ultrastructure, pre-embedding techniques are needed to locate scarce or masked antigens. In addition, pre-embedding labeling is the method of choice to study cellular interactions in three dimensions by electron tomography (Koster et al., 1997; Mehlin et al., 1992). Methods for pre-embedding immunolabeling are widely used for light and electron microscopic investigations, especially for biomedical studies. An example of mechanical permeabilization is cryostat sectioning (Matsuno et al., 1994; Yazama et al., 1997). During freezing of organs or pieces of tissue, ice crystals are formed, which disrupt the cell membranes. In light microscopy, Triton X-100 or Nonidet P40 are frequently used (De Graaf et al., 1992; Langanger et al., 1984; Martone et al., 1996; Satijn et al., 1997; Sibon et al., 1994). Both detergents disrupt the bilayer structure by dissolving the lipids and membrane proteins (Helenius and Simons, 1975). For electron microscopy, less stringent detergents such as Saponin are preferred (Andersson et al., 1990; Brown and Farquhar, 1989; Heuser, 1981; Louvard et al., 1982; Macville et al., 1995; Tanner et al., 1996). Saponin forms complexes with cholesterol that leads to reversible pores in membranes (Brown and Farquhar, 1989). To avoid lipid removal by detergents, cells can be permeabilized by treatment with sodium borohydride (Van Lookeren Campagne et al., 1992; Young and Furness, 1995). A method to label living cells would be of special interest. Streptolysin, a cytolytic toxin of Streptococcus pyogenes, integrates into the cellular membranes and forms pores in vivo (Alouf, 1980). On ice, streptolysin-O will bind to plasma membrane-associated cholesterol. Then excessive toxin is removed. At higher temperature the toxin-cholesterol complexes associate via lateral diffusion to form oligomers. These oligomers form pores of about 35 nm in diameter (Leno et al., 1992). The pores are large enough to allow free passage of cytoplasmic proteins (Krijnse-Locker et al., 1994). Dundr and Raska (1993) applied the streptolysin method to label active transcription sites within the nucleolus. They permeabilized the plasma membrane of HeLa cells and added bromouridine, which the RNA polymerases incorporate into nascent RNA (Wansink et al., 1993).

PRE-EMBEDDING IMMUNOLABELING FOR EM

45

We also considered the method published by Schliwa et al. (1981) for labeling living cells. They have shown that, at least temporarily, the morphological integrity of cells can be maintained (Schliwa et al., 1981). After a short Brij 58 treatment of living cells, only a few so-called soluble proteins are lost (Schliwa et al., 1987) and even the loss of potassium ions is retarded in Brij 58-opened cells (Kellermayer et al., 1986). The scope of this article is to compare ve permeabilization methods for pre-embedding labeling on the electron microscope level based on: 1) pre-xation/ Triton X-100 (De Graaf et al., 1992); 2) Saponin (Macville et al., 1995); 3) sodium borohydride (Van Lookeren Campagne et al., 1992); 4) streptolysin-O (Dundr and Raska, 1993); and 5) Brij-opening (Schliwa et al., 1987). We assessed the accessibility of cytoplasmic and nuclear proteins (De Graaf, et al., 1991; Fu and Maniatis, 1990; Satijn, et al., 1997) for a set of different marker antibodies. The results should provide a basis for choosing the most suitable method. MATERIALS AND METHODS Chemicals and Antibodies Brij 58 (16004) was purchased from Fluka (Fluka Chemicals, Buchs, Switzerland), Saponin (S-1252) and Triton X-100 (X-100) from Sigma (Sigma Chemical Co., St. Louis, MO, USA) and streptolysin-O from Wellcome (Wellcome Laboratories, Beckenham, UK). Ethylene glycol-bis( -aminoethyl ether)N,N,N ,N ,-tetraacetic acid (EGTA; E-4378) was from Sigma, 2-amino-2(hydroxymethyl)-1,3-propandiol (Tris; 708976) from Boehringer (Boehringer Mannheim GmbH, Mannheim, Germany), piperazine-1,4-bis(2-ethanesulfonic acid) (PIPES; 10220) and 2-[4-(2-hydroxyethyl)-1-piperazinyl](2-ethanesulfonic acid) (HEPES; 1.10110.) both from Merck (Merck, Darmstadt, Germany). Cell culture media and plastic ware were purchased from Gibco BRL (Life Technologies, Breda, The Netherlands). All other chemicals used were pro analysis grade from either Merck or Fluka. The monoclonal mouse anti- -tubulin antibodies were from Sigma (T-5168), supernatant of monoclonal mouse anti-SC-35 antibody from the group of Dr. T. Maniatis (Fu and Maniatis, 1990), the monoclonal mouse antiAM88 was a gift of Dr. R van Driel (De Graaf et al., 1991), and the rabbit anti-polycomb antibodies (bmi/ ring) were a gift from Drs. D. Satijn and A. Otte (Satijn et al., 1997). The rabbit anti-biotin antibodies were from Enzo (#43861; Enzo Diagnostics, Farmingdale, NY). The following secondary antibodies were from Jackson (Jackson Immunoresearch Laboratories, West Grove, PA): goat anti-mouse-DTAF (dichlorotiazinyl amino uorescein; 11015044), goat anti-rabbit-FITC (Coons, 1956) (uorescein isothiocyanate; 111095 003), rabbit anti-mouse-peroxidase (Nakane and Pierce, 1967) (315035044) and goat anti-rabbit-peroxidase (111035045). The ultrasmall gold particles, goat antimouse (880305) and goat anti-rabbit (880304) were from an experimental batch (Janssen Life Sciences Products, Olen, Belgium, gift of Dr. J. Leunissen) and goat anti-mouse Nanogold (#2001; Nanoprobe, Stony Brook, NY). The FITC-ultrasmall gold conjugated antibodies (Au-FITC) were from Nanoprobe (Powell et al., 1997) (goat anti-rabbit, #7004) and a custom-tailored

goat anti-mouse was a gift from Dr. J. Leunissen (Aurion, Wageningen, The Netherlands). All the other gold-tagged antibodies, goat anti-mouse 6 nm (GAM EM-6 nm), goat anti-rabbit 6 nm (GAR EM-6 nm), goat anti-mouse 10 nm (GAM EM-10 nm), and goat antirabbit 10 nm (GAR EM-10 nm), were purchased from Aurion. The FITC-tyramide and the biotin-tyramide were a gift of R. van Gijlswijk and Dr. A. Raap. Cell Culture Osteosarcoma cells (U2OS) were cultured in 75 cm2 cell culture asks in Dulbeccos modied Eagles medium (DMEM) supplemented with 10% fetal bovine serum in a 7% humidied CO2 atmosphere. For the labeling experiments, the cells were seeded on 12-mm glass coverslips in 12-well culture dishes and grown overnight to 5070% conuency. The cells were further treated as described below. If not stated otherwise, all incubations and washing steps were done in the 12-well culture dishes at room temperature.

Permeabilization Methods
Prexation and Triton X-100 Permeabilization. The cells were rinsed twice with PBS (137 mM NaCl, 2.7 mM KCl, 8.1 mM Na2HPO4, 1.5 mM KH2PO4, pH 7.4) and xed with 2% (w/v) formaldehyde in PBS (FA prexed) for 15 minutes or with 2% (w/v) formaldehyde and 0.02% (v/v) glutaraldehyde in 200 mM HEPES (pH 7.2; FA-GA prexed) (Griffiths, 1993) for 30 minutes. After permeabilization with 0.5% (w/v) Triton X-100 in PBS for 5 minutes the cells were washed twice with PBS. In one experiment, Triton X-100 was replaced by 0.2% (w/v) Brij 58. Saponin Permeabilization (Macville et al., 1995). The coverslips were rinsed twice with PBS and xed with 1% (w/v) formaldehyde and 0.02% (v/v) glutaraldehyde for 30 minutes. Permeabilization was done with 0.1% (w/v) Saponin in PBS for 30 minutes. Sodium Borohydride Permeabilization (Van Lookeren Campagne et al., 1992). The cells were rinsed twice with PBS, then xed with 4% (w/v) formaldehyde in 0.1 M sodium cacodylate buffer, pH 7.6, for 30 minutes at room temperature and permeabilized twice with 0.1% (w/v) sodium borohydride in PBS for 5 minutes at room temperature. In the experiment to check which of the components was responsible for perforating the plasma membrane, incubations were done with either of the reactants sodium cacodylate buffer, formaldehyde in sodium cacodylate buffer, or sodium borohydride in PBS. Membrane leakage was monitored with propidium iodide, a stain for nucleic acids (Tas and Westerneng, 1981). Streptolysin-O Permeabilization (Dundr and Raska, 1993; Krijnse-Locker et al., 1994). Cells grown on coverslips were rinsed once with PBS and once with SLO-buffer (25 mM HEPES, pH 7.0, 50 mM KCl, 5 mM MgCl2, 2 mM EGTA). Then they were incubated with 1 U/ml streptolysin-O in SLO-buffer supplemented with a cocktail of protease inhibitors (Complete, Boehringer) and 7 mM mercaptoethanol for 30 minutes at 37C. Thereafter, the cells were xed for 30 minutes in 4% (w/v) formaldehyde at room temperature.

46

B.M. HUMBEL ET AL.

Brij 58-Opening. Cells grown on coverslips were washed twice with PBS and incubated for 5 minutes at room temperature with 0.2% (w/v) Brij 58 in PHEM buffer (60 mM PIPES, pH 6.9, 25 mM HEPES, 10 mM EGTA, 2 mM MgCl2) supplemented with the cocktail of protease inhibitors and 7 mM mercaptoethanol. The cells were xed with 4% (w/v) formaldehyde in PHEMbuffer for 30 minutes at room temperature (Schliwa et al., 1981). In all experiments formaldehyde was freshly prepared by de-polymerizing para-formaldehyde. Immunolabeling The permeabilized and xed cells were washed twice with PBS and incubated twice with PBG (PBS, 0.5% (w/v) bovine serum albumin, 0.045% (w/v) cold-water sh gelatin (Birrell et al., 1987)) for 5 minutes to block protein binding sites. The coverslips were removed from the 12-well dishes and placed on a piece of paralm in a moist chamber. The diluted primary antibodies mouse anti-tubulin (1:1,000), mouse antiAM88 (1:1,000), mouse anti-SC35 (1:5), rabbit anti-bmi/ ring (1:500), or PBG (as a control) were pipetted onto the coverslips and incubated for 1 hour at room temperature. The Saponin preparations were incubated for 2 hours at 37C. Then the coverslips were put back on the 12-well plates and washed six times for at least 5 minutes. The plates were rocked on a rotary shaker at 150 rpm. The coverslips were put back on the paralm and incubated with the secondary antibodies overnight in the moist chamber. Up to this point, 0.1% Saponin was added to all the solutions of Saponin-permeabilized cells. The coverslips were put back in the 12-well plates and washed four times with PBG and four times with PBS for at least 15 minutes each step on a rotary shaker. Visualization for Light Microscopy A coverslip (24x50 mm) with on-grown living cells was inverted on a glass slide and immediately photographed. The FITC and Au-FITC labeled cells were directly mounted on glass slides with moviol (Rodriguez and Deinhardt, 1960) containing para-phenylene diamine (Johnson and de C. Nogueira Araujo, 1981) and photographed with a Leitz Orthoplan (Ernst Leitz Wetzlar GmbH, Wetzlar, Germany) microscope using a PLANAPO 63/1.4 OEL PHACO 4 objective and the I3 lter set (BP450490, RKP510; LP515). The gold-labeled cells and Au-FITC-labeled cells (after being examined with uorescence microscopy) were xed with 1% (v/v) glutaraldehyde in PBS for 30 minutes at room temperature. Silver enhancement was done according to the method of Danscher (1981, 1984; Stierhof et al., 1991a, 1995) for 40 minutes. The peroxidase-labeled cells were rinsed in 0.1 M Tris/HCl, pH 8.8, and 10 mM imidazole and incubated in the same buffer supplemented with 0.001% (v/v) H2O2 and 0.5 mg/ml 3,3 diaminobenzidine (DAB) for 20 minutes in the dark (Macville et al., 1995). DAB can be stored in the freezer as a stock solution of 100 mg/ml in dimethyl formamid. Alternatively, peroxidase can also be developed with FITC-tyramide or biotin-tyramide (Van Gijlswijk et al., 1996, 1997) under the same conditions as DAB. Rinsing once with the buffer and once with PBS stopped the reaction. The coverslips with the gold and

peroxidase/DAB labeled cells were mounted on glass slides with moviol and photographed with a Leitz Orthoplan microscope in the bright eld and phase contrast mode. The FITC-tyramide labeled probes were photographed in the uorescence mode and, thereafter, together with the tyramide-biotin probes embedded in Lowicryl HM20 (see below). Electron Microscopy After light microscopical investigations, the coverslips were soaked in PBS for 28 hours and removed from the glass slides. The peroxidase/DAB-labeled samples were postxed with 1% (v/v) glutaraldehyde in PBS and 1% (w/v) osmium tetroxide in double-distilled water. After dehydration in a graded series of ethanol the cells were embedded in Epon. For at embedding, the bottom of BEEM capsules 3 (G362; PLANO, W. Plannet GmbH, Marburg, Germany) was cut off and the capsules were put on the cell side of the coverslips. Overnight a thin layer of Epon was polymerized to seal the capsules to the coverslips. After introducing a sample label the capsules were lled with Epon and polymerized for another 2 days. The dish containing the coverslips was immersed in liquid nitrogen until cracking, then the capsules were broken off. Remaining glass was removed with 40% (v/v) hydrouoric acid. For Lowicryl embedding (Carlemalm et al., 1982; Simon et al., 1987), after dehydration the cells were inltrated with Lowicryl HM20 (80% (w/w) monomer, 20% (w/w) cross-linker and 0.5% (w/w) initiator). The coverslips were placed on a glass slide wrapped in aluminum foil (Fig. 2). A BEEM capsule 3 was lled with fresh Lowicryl and frozen in liquid nitrogen and put upside down onto the coverslip (Fig. 2). The glass slide was placed in a CS Auto freeze-substitution apparatus (Reichert-Jung, Vienna, Austria) cooled to -40C. Two hours later, after adaptation to the temperature, the resin was polymerized under UV illumination for 1 day at -40C and 1 day at room temperature. Thereafter, the capsules were removed from the glass slide by immersion in liquid nitrogen. Under these conditions, the glass cracks and the Lowicryl blocks can be lifted off. The biotin-tyramide polymer was detected with a rabbit anti-biotin antibody and an ultrasmall gold or 10 nm colloidal gold goat anti-rabbit antibody on ultrathin Lowicryl sections. Sections of 90 nm or 300 nm were cut parallel to the cells. Except for the peroxidase/DAB-labeled cells, sections were stained with 2% (w/v) aqueous uranyl acetate and 0.4% (w/v) lead citrate (Venable and Coggeshall, 1965) 10 minutes each. The sections were examined with an EM 420 electron microscope (Philips Electron Optics, Eindhoven, The Netherlands) at 60 kV acceleration voltage using a 20 m aperture. RESULTS Structural Preservation Evaluated by Light Microscopy The structural appearance of the cells was evaluated by phase contrast light microscopy and by the appearance of the labeled microtubules by uorescence microscopy. The morphology of cells prexed with formaldehyde or formaldehyde/glutaraldehyde and permeabilized with Triton X-100 (Fig. 4A,B) or after Saponin permeabiliza-

PRE-EMBEDDING IMMUNOLABELING FOR EM

47

Fig. 3. Living cells photographed in phase contrast. Two micrographs showing the same group of cells, focused on the nuclei (A) and on the surface of the coverslip (B). Different features, such as ruffling cell membrane, vesicles, the nuclear envelope, nucleoli and large nuclear bodies, are clearly seen. Bar 20 m.

Fig. 2. (A) BEEM capsules were lled with fresh resin and frozen in liquid nitrogen so that the capsules could be safely turned upside down and placed with the open side on the coverslips. (B) For at embedding into Lowicryl, the coverslips with resin-inltrated cells were placed on a glass slide, which is wrapped in a trough of aluminium foil. (C) A bottom view of the polymerized sample on the coverslip.

tion (Fig. 4C) did not change signicantly compared to living cells (Fig. 3). The organelles were smaller and the cytoskeleton was more distinct. The nucleoplasm of the FA prexed/Triton X-100 (Fig. 4A) and the Saponin (Fig. 4C) permeabilized cells was slightly granular. The very sensitive microtubular network in the three prepa-

rations show a nice, smooth curvature, as expected (Fig. 5A,B,C). To establish which component of the sodium borohydride method (Van Lookeren Campagne et al., 1992) causes the permeabilization, cells grown on coverslips were exposed to the individual components. Permeabilization was assessed by penetration of propidium iodide. Buffer-treated cells were impermeable (not shown). After formaldehyde xation, some propidium iodide could pass the membrane barrier, indicated by a light labeling of the nucleic acids (Fig. 6A). Only if followed by sodium borohydride treatment were the cells completely permeable for the dye (Fig. 6B). Treatment with sodium borohydride without prexation was so rough that all the cells were removed from the coverslips (not shown). Hence, only the combination of xation followed by sodium borohydride treatment led to permeabilized cells. These results indicate that borohydride treatment opens the plasma membrane, but not formaldehyde. Sodium borohydride permeabilized cells still have the contours of living cells but fewer organelles are visible and the cytoplasm has become granular around the nucleus (Fig. 4D). It looks as if the cytoplas-

48

B.M. HUMBEL ET AL.

Fig. 4. Cells prepared using the different permeabilization methods were investigated by phase contrast microscopy. (A) Formaldehyde and (B) formaldehyde/glutaraldehyde prexation followed by Triton X-100, (C) Saponin, (D) sodium borohydride, (E) streptolysin-O, and (F) Brij 58-opened. Bar 20 m.

mic structures have been dug up. The tubulin labeling shown in Figure 5D and labeling of the nuclear protein SC-35 (not shown) demonstrate that the holes are large enough to let antibodies pass, but permeabilization is not equally efficient for different cells. Even in one individual cell, only a part of the tubulin network is labeled (Fig. 5D). Most of the permeabilized cells at room temperature with streptolysin-O were collapsed (Fig. 4E). Only the subpopulation of large at cells seemed to have retained some nuclear structures. The microtubuli network appeared crumpled (Fig. 5E). In some preparations of Brij 58-opened cells (Fig. 4F), the structural appearance is close to that of prexed and Triton X-100 or Saponin-treated cells. In many cells, a halo around the nucleus could be observed,

suggesting that the cytoskeleton has come off the cell nucleus (Fig. 4F). This impression was conrmed by electron microscopy (data not shown). The microtubular network shows little sign of collapse (Fig. 5F): it still extends through the whole cell, but the individual microtubules are wavy, rather than having a taut curvature. In addition, the preparation is not easily reproducible. From these results, we conclude that prexation is favorable, as it leads to more reproducible results. Accessibility of Cytoplasmic and Nuclear Proteins In Table 1, the labeling results of tubulin and the nuclear antigens are summarized. The cytoplasm and the nucleus of cells xed with formaldehyde and perme-

PRE-EMBEDDING IMMUNOLABELING FOR EM

49

Fig. 5. Cells prepared using the different permeabilization methods were labeled for tubulin and visualized with an FITC-tagged secondary antibody. (A) Formaldehyde and (B) formaldehyde/ glutaraldehyde prexation followed by Triton X-100, (C) Saponin, (D) sodium borohydride, (E) streptolysin-O, and (F) Brij 58-opened. Bar 20 m.

abilized with Triton X-100 were easily accessible for the primary antibodies and the marker antibodies. The 10 nm gold probe showed a reduced labeling intensity. The same is true for cells xed with formaldehyde and glutaraldehyde (Fig. 7). Only the larger gold probes, 6 nm and 10 nm, were no longer able to access the nucleus (Fig. 7E,F). The microtubular network was weakly labeled and, in the case of 10 nm colloidal gold particles, the access of the network was clearly hampered. Saponin-permeabilized cells gave essentially the same results in terms of tubulin labeling as described for Triton X-100. The access of the nucleus was clearly reduced for FITC-tagged antibodies. Only in one case did the peroxidase-tagged antibodies label a nuclear

antigen. All the gold-tagged antibodies were excluded from the nucleus. From the labeling pattern obtained in sodium borohydride-permeabilized cells, we conclude that the cell membrane of only about half of the cells is perforated enough for antibody penetration. The labeling of the microtubular network is more or less the same with all the markers. Hence, if permeabilization is sufficient for antibodies to pass, it is also sufficient for the marker. Only FITC-tagged antibodies can penetrate the nucleus and localize nuclear proteins. Even then, only a few cells showed a good labeling pattern; most, however, did not label at all, supporting the notion that only about half of the cells are permeable enough for antibody penetration.

50

B.M. HUMBEL ET AL.

TABLE 1. Accessibility of cytoplasmic and nucleus proteins Marker FITC PO Au-FITC US 6 nm 10 nm

Triton X-100 permeabilization (formaldehyde pre-xed) cytoplasm tubulin nucleus AM88 bmi/ring Triton X-100 permeabilization (formaldehyde/glutaraldehyde prexed) cytoplasm tubulin / nucleus SC-35 bmi/ring / / Saponin permeabilization cytoplasm tubulin / nucleus SC-35 bmi/ring / Sodium borohydride permeabilization cytoplasm tubulin / / / / / / nucleus SC-35 / / bmi/ring / Brij 58-opened cytoplasm tubulin nucleus SC-35 bmi/ring
Cells were prepared according to the different permeabilization methods described in Materials and Methods. To evaluate the accessibility within the cells, one cytoplasmic antigen: tubulin; and three nuclear antigens: a nuclear matrix protein AM88, the essential splicing factor SC-35, and a member of the polycomb group bmi/ring; were labeled. The label efficiency of the different markers was scored by light microscopy. high label intensity, good label efficiency, sufficient label efficiency, / debatable labeling, no label.

Fig. 6. Cells were xed with formaldehyde in cacodylate buffer (A) and treated with sodium borohydride (B). Permeabilization was monitored with propidium iodide. Bar 20 m.

In our hands, streptolysin-O treatment at 37C had a deleterious effect on the cellular structure and, therefore, the labeling was not further evaluated. In Brij 58-opened cells all the markers label the microtubular network. However, only FITC- and peroxidase-tagged antibodies can access nuclear proteins; but, all the gold probes were excluded. From Table 1 some conclusions on the labeling properties of the secondary antibodies can be drawn. FITCtagged antibodies are obviously so small that they can access all the primary antibodies. Peroxidase-tagged antibodies, increasing in size by about 40 kD (Griffiths, 1993), have a somewhat reduced penetration capacity compared to FITC-tagged antibodies. Still, they are more efficient than the ultrasmall gold probes. Thus, in critical cases peroxidase labels are preferred to ultrasmall gold labels. The real cut-off, however, is at the level of the ultrasmall gold particles. The detection efficiency of larger colloidal gold particles is clearly inferior; therefore, they are not suited for pre-embedding labeling of nuclear proteins. Ultrastructural Preservation and Labeling Evaluated by Electron Microscopy In our hands, the most successful and reproducible method is based on prexation with formaldehyde/

glutaraldehyde (FA-GA prexed) and permeabilization with Triton X-100 (Fig. 8AC) or Brij 58 (Fig. 8DF). The cellular organelles can still be recognized; hence, a label relates to ultrastructural details. The cytoplasm appears to be riddled with holes after detergent treatment (Fig. 8B,E). The holes seem to be somewhat larger in regions with few organelles and very ne in the area of mitochondria (Fig. 8A,B). In prexed and Brij 58treated cells, the holes also seem somewhat larger (Fig. 8E, note also the inverted contrast of the mitochondria and that the membranes of the cristae are still visible). The state of preservation varies from cell to cell; thus, extraction is not even for all the cells. The nucleoplasm of prexed and Triton X-100-permeabilized cells becomes grainy and the area of heterochromatin can no longer be seen. Other features, such as nucleoli (Nu) with the brillar centers and nuclear bodies (Nb), are still recognizable. The nucleoplasm of prexed and Brij 58-permeabilized cells is denser and seems better preserved than that of Triton X-100-treated cells (Fig. 8C,F). Again, the contrast is inverted, which renders the structural details less recognizable (Fig. 8F), e.g., for heterochromatin. Histochemical staining techniques for DNA (Testillano et al., 1991) and RNA (Bernhard, 1969; Monneron and Bernhard, 1969) would be needed. Structures such as nucleolus (Nu) with brillar centers and clusters of interchromatin granules (Ig) can easily be recognized (Fig 8C,F).

PRE-EMBEDDING IMMUNOLABELING FOR EM

51

Fig. 7. Cells were prexed with formaldehyde and glutaraldehyde, then permeabilized with Triton X-100. A nuclear protein, the essential splicing factor SC-35, was identically labeled with a monoclonal primary antibody. The labeling was visualized with FITC (A), peroxi-

dase (B), Au-FITC (C), ultrasmall gold (D), 6 nm (E) and 10 nm gold (F) tagged secondary antibodies. The gold particles were enlarged by silver enhancement. A and C are uorescence and B, DF bright-eld images. Bar 20 m.

A further proof for the denser nucleoplasm is shown in Figure 9. Cells prexed with formaldehyde/glutaraldehyde were permeabilized with Brij 58 and labeled for the splicing factor SC-35. The label pattern could be established with FITC- (Fig. 9A) and peroxidase-tagged secondary antibodies (Fig. 9B); however, not with the ultrasmall gold particles (Fig. 9C). Again, the gold particles prove to be bulkier than peroxidase. Pre-embedding is the method of choice for the study of cellular interactions in three dimensions by electron tomography. Here, we labeled the essential splicing factor SC-35 and visualized the primary antibodies either with peroxidase/DAB (Fig. 10A) or silverenhanced ultrasmall gold particles (Fig. 10B). The most obvious difference is the labeling density. The clusters

of interchromatin granules are completely covered with DAB, while the individual gold particles do not obscure underlying structures. For the peroxidase/DAB method, the sections cannot be stained, as stain obscures the label, and the ultrastructural details are less clearly visible. The distribution pattern of gold particles in three dimensions (Fig. 10B) demonstrates why a much lower label density for on-section labeling methods has to be expected. In one plane of the example shown, only a few particles are present. Three-dimensional viewing also reveals much more clearly the diffusion gradient of the DAB polymerization product (Fig. 10A). Both of the micrographs show that the label can be related to nuclear domains, e.g., nucleolus and nuclear bodies.

Fig. 8. The two most successful methods were based on formaldehyde/glutaraldehyde prexation followed by permeabilization with either Triton X-100 (AC) or Brij 58 (DF). In both the preparations the cytoplasm and the nucleus are rather dense and mitochondria (m) and vesicles (v) are visible. In the nucleus the nucleoli (Nu) with the

brillar center, clusters of interchromatin granules (Ig) and nuclear bodies (Nb) are observed. In Brij 58-treated cells the contrast in the nucleus and the mitochondria is inverted. (A,D) Whole cell at low magnication, bar 5 m. (B,E) A portion of the cytoplasm; (C,F) a portion of the nucleus at higher magnication; bar 1 m.

PRE-EMBEDDING IMMUNOLABELING FOR EM

53

Fig. 9. Cells prexed with formaldehyde/glutaraldehyde and permeabilized with Brij 58 were labeled for SC-35. The primary antibody was detected with (A) FITC, (B) peroxidase DAB, and (C) ultrasmall gold particles. Bar 20 m.

The intense label after pre-embedding labeling with peroxidase suggests combining pre-embedding labeling with post-embedding detection. In a mixed technique, one would prot from the higher labeling efficiency without the disadvantage of the low contrast of the DAB reaction product. DAB is replaced by the recently established tyramide probes (Van Gijlswijk et al., 1996, 1997). In whole mount preparations the label pattern and efficiency is checked by light microscopy using FITC-tyramide (Fig. 11A). For electron microscopy, biotin-tyramide was used. The labeling procedure is controlled stepwise. First, the distribution of the biotin is shown on ultrathin Lowicryl sections by uorescence microscopy (Fig. 11B); thereafter, subsequent sections were labeled with an ultrasmall (Fig 11C) or a 10 nm gold-tagged antibody (Fig. 11D) in two steps. After gold labeling, the sections can be stained with uranyl acetate and lead citrate without fear of covering the weak contrast of the label product (Fig. 11C,D). DISCUSSION In this article, we discuss different permeabilization methods in respect to their ultrastructural quality and the labeling efficiency. The most appealing methods are those that open living cells and make them permeable to drugs and, hopefully, for antibodies and markers. In our investigation, we concluded that both methods investigated, streptolysin-O pore formation and Brij 58-opening (without prexation), are not suited for routine use. In fact, one has to expect that if the cells are as well preserved as described by Schliwa et al. (1981, 1987), movement of the antibodies in the cell is greatly obstructed. Indeed, the nucleus was less accessible after protease inhibitors were added to the extraction buffer. Dundr and Raska (1993) obtained excellent results using the streptolysin-O technique. They labeled nascent transcript with bromouridine. Nuclear activity, e.g., RNA synthesis, could continue, although

the plasma membrane was porous. The nucleoplasm was well preserved and ne localization of rRNA synthesis in the nucleoli could be demonstrated. As reported by Leno et al. (1992) the nuclear membranes were still intact and non-nuclear proteins excluded. The streptolysin-O method undoubtedly has its merits for cytoplasmic localization studies (Krijnse-Locker et al., 1994). For immunolabeling of nuclear proteins, the streptolysin method should be used at higher temperatures to access nuclear membranes as well. In addition, the microtubules disassemble at 0C. The cellular integrity is destroyed if the streptolysin method is applied at higher temperature. From our results, we conclude that both methods to permeabilize or open living cells are not suited for immunolabeling of nuclear antigens for each cell type the method has to be optimized and under optimum conditions the cell nucleus is not accessible for large molecules, e.g., antibodies. For routine applications, xation prior to permeabilization is favorable, as the structures can be xed to a degree dened by the xation process. Very strong xatives, e.g., high concentrations of glutaraldehyde, likely also prevent free movement of antibodies within the cells. Formaldehyde xation is widely used for (confocal) light microscope investigations. Under optimum conditions, i.e., high concentrations, formaldehyde can preserve ultrastructural details and is used for cryosection labeling (Griffiths, 1993). In combination with Triton X-100 as a detergent, cytoplasmic proteins are lost or can be relocated (Melan and Sluder, 1992). In addition, the nucleoplasm loses its ne structure and changes into a brillar network (data not shown). We suspect that not just soluble proteins are lost to reveal the nucleic acid-protein network, but also that the stretches of DNA and RNA are dislocated. One should bear in mind that these displacements might be signicant enough to inuence the labeling pattern even at the level of uorescence microscopy.

54

B.M. HUMBEL ET AL.

Fig. 10. Stereo micrographs of cells prexed with formaldehyde/ glutaraldehyde and permeabilized with Triton X-100 and labeled for SC-35. The primary antibody was detected either with peroxidase/ DAB (A) or silver-enhanced ultrasmall gold markers (B). In the peroxidase/DAB preparation (A), the whole volume of the clusters of interchromatin granules is intensely labeled. Note the diffusion halo

between two clusters of interchromatin granules and the blurring of the ultrastructure. In the colloidal gold-labeled sections (B) ultrastructural details can be seen clearly without obscuring the distinct silver-enhanced gold marker. Careful inspection of the stereo image shows that only a few gold particles lay within one horizontal plane. Bar 1 m.

PRE-EMBEDDING IMMUNOLABELING FOR EM

55

Fig. 11. Cells prexed with formaldehyde/glutaraldehyde and permeabilized with Triton X-100 were labeled for SC-35. The primary antibody was revealed by peroxidase/FITC-tyramide (A) or peroxidase/ biotin-tyramide (BD) reaction. Ultrathin sections of biotin-tyramide

developed cells were mounted on glass coverslips (B) or grids and visualized by FITC (B), ultrasmall gold (C), or 10 nm colloidal gold particles (D). Note the dense labeling on the interchromatin granules. Bars 20 m and 1 m, respectively.

56

B.M. HUMBEL ET AL.

In general, mild xation with formaldehyde is preferred because it interferes least with antibody binding. Therefore, we investigated the combination of formaldehyde xation with borohydride as an agent to pierce the membranes. First, it had to be established which of the components of the method described by van Lookeren Campagne et al. (1992) caused leakage. It was suspected that formaldehyde in cacodylate buffer might be sufficient to render plasma membranes permeable. As indicated by propidium iodide labeling of nucleic acids (Fig. 6A), the membranes indeed become slightly permeable, but only borohydride treatment makes holes large enough to give efficient access to the dye. Antibodies can also reach cytoplasmic structures, which seem to lay on top of the cells under the surface of the membranes (Fig. 5D). Deeper layers or those close to the coverslip are not accessible. Close inspection of the cells in phase contrast microscopy suggest that borohydride digs up the cytoplasm from the apical side. The borohydride method may well be suited for the long and thin dendrites or nerve growth cones (Van Lookeren Campagne et al., 1992) but not for bulkier specimens, except perhaps in combination with an ethanol wash (Young and Furness, 1995). Saponin is suggested as the method of choice for electron microscope pre-embedding labeling and many excellent studies have been published (Louvard et al., 1982; Macville et al., 1995; Griffiths, 1993, and references therein). As demonstrated (Figs. 4, 5C), the cellular structure is well preserved and cytoplasmic antigens are easily accessed. The ultrastructural morphology was considerably improved compared to the formaldehyde prexed and Triton X-100-extracted cells. The nuclear antigens, however, were not accessible (Table 1) for most of the markers. The reason for this is not yet clear, considering the excellent detection of a viral transcript by Macville et al. (1995). There are two explanations to consider: either the in situ hybridization process opens the nuclear structures more or the proteins labeled in this report are tightly bound to other macromolecules, which rst have to be removed before they can be accessed by an antibody. In our hands, the best and easiest to use method is prexation with formaldehyde and glutaraldehyde. Permeabilization is either done with Triton X-100 or Brij 58. In both cases, preservation of the ultrastructure is good enough to correlate a label with cellular components (Fig. 8). Most of the features of the nucleus, such as nucleolus with its substructures, clusters of interchromatin granules, and nuclear bodies, can be identied. Still, the nuclear meshwork is loose enough to give access to the antibodies. After Triton X-100 permeabilization, all small markers (FITC, peroxidase, and ultrasmall colloidal gold particles) can enter the nucleus (Figs. 7, 10); only the 6 and 10 nm particles are excluded, conrming previously published results (De Graaf et al., 1991). As suggested by the electron micrographs (Fig. 8F), the nucleoplasm of Brij 58-permeabilized cells is denser. The splicing factor SC-35 could only be detected with FITC- or peroxidase-tagged antibodies and not with ultrasmall gold particles (Fig. 9). From these results, we conclude that even the smallest gold markers cannot move within a cell as easily as a uorescence or peroxidase marker. Therefore, in criti-

cal situations the use of peroxidase/DAB might reveal a labeling which otherwise would escape our attention. Both of the markers, peroxidase/DAB and ultrasmall gold particles, may be used for three-dimensional localization studies. The stereo micrographs (Fig. 10) clearly show the advantages and disadvantages of the two markers. Peroxidase/DAB seems more efficient. The entire volume of the clusters of interchromatin granules is intensely labeled. On the other hand, there is diffusion of the polymer (note the less dark area between two clusters of interchromatin granules). Viewed in stereo, the diffusion halo is more visible. The nucleoplasm is weakly stained. In the stereo view, the structures have a slightly blurred appearance. Peroxidase/ DAB preparation should not be stained with uranium and lead ions in order not to obscure the label. In contrast, the colloidal gold-labeled sections may be contrasted. The ultrastructural details can clearly be seen without obscuring the distinct silver-enhanced gold marker. Viewed in stereo, the structural details are clearly visible without blurring. On the other hand, one has to accept reduced label density. The location of the clusters of interchromatin granules is unambiguously revealed. Careful inspection of the stereo image shows that only a few gold particles lay within one horizontal plane. This observation coincides with and may explain the low label efficiency observed by onsection labeling experiments. The described results suggest combining high label efficiency and better penetration of enzyme markers with the clear distinction of colloidal gold markers on well-contrasted sections. In addition, the enzyme label can be used as an enhancement step. From one bound peroxidase molecule, a controllable amount of secondary antigen can be produced. The recently introduced tyramide technique (Van Gijlswijk et al., 1996) can serve as the link between the two labeling methods. In a pre-embedding approach, the antigen is detected by peroxidase antibodies and a polymer of tyramide molecules, to which either biotin or FITC is bound. FITC has the advantage that the label pattern, the amount of tyramide deposition, and the degree of diffusion can be directly monitored with the uorescence microscope. The reaction time determines the label intensity and the viscosity of the development solution controls the diffusion radius (Van Gijlswijk et al., 1997). Thereafter, the biotin or FITC molecules instead of the antigen can be visualized by on-section labeling with a colloidal gold marker. Thus, high label efficiency even of scarce antigens and good visualization can be combined. Of course, labeling with enzyme products will compromise the spatial resolution. However, we should not forget that pre-embedding labeling as such is already a compromisethe cellular ultrastructure is inuenced in that it has to give way to antibodies and markers. Proteins may be relocated or lost. In terms of spatial resolution, the pre-embedding labeling technique is somewhere between light and electron microscopy because the ultrastructure can not be preserved to high resolution. Often, however, pre-embedding labeling is the only method of choice to locate scarce or masked antigens and it opens the possibility of studying cellular interactions in three dimensions. In addition, it can link light and electron microscopic investigations.

PRE-EMBEDDING IMMUNOLABELING FOR EM

57

Based on our results, the routine pre-embedding labeling method of choice is prexation with formaldehyde/glutaraldehyde and permeabilization with Triton X-100 or Brij 58. For detection, either ultrasmall gold particles (with or without uorochrome) with silver enhancement or if the ultrasmall gold particles are obstructed, peroxidase markers are advised. The most promising technique to localize scarce antigens with good contrast is the combination of a pre-embedding peroxidase/tyramide-FITC or -biotin labeling followed by an on-section colloidal gold detection. ACKNOWLEDGMENTS We thank Drs. R. van Driel, D. Satijn and A. Otte and Dr. J. Leunissen for their generous gift of antibodies and Dr. T. Raap and R. van Gijlswijk for the tyramide conjugates. We thank Ronald Leito for Figure 2 and Frits Kindt and his colleagues for fast and professional photographic work. We thank Dr. H. Schwarz and Dr. Y.-D. Stierhof Schwarz for their continuous interest and discussions and for careful reading of the manuscript. Dr. Stierhof also supplied the Lowicryl-embedded Leishmania mexicana cells. REFERENCES
Alouf, J.E. (1980) Streptococcal toxins (streptolysin O, streptolysin S, erythrogenic toxin). Pharmacol. Ther., 11:661717. Andersson, G.L., Ericson, L.E., and Jennische, E. (1990) Ultrastructural localization of IGF-I in the rat kidney; an immunocytochemical study. Histochemistry, 94:263267. Bernhard, W. (1969) A new staining procedure for electron microscopical cytology. J. Ultrastruct. Res., 27:250265. Birrell, G.B., Hedberg, K.K., and Griffith, O.H. (1987) Pitfalls of immunogold labeling: Analysis by light microscopy, transmission electron microscopy and photoelectron microscopy. J. Histochem. Cytochem., 35:843853. Brown, W.J., and Farquhar, M.G. (1989) Immunoperoxidase methods for the localization of antigens in cultured cells and tissue sections by electron microscopy. In: Methods in Cell Biology. A. Tartakoff, ed. Academic Press, New York, pp. 553569. Carlemalm, E., Garavito, R.M., and Villiger, W. (1982) Resin development for electron microscopy and an analysis of embedding at low temperature. J. Microsc., 126:123143. Coons, A.H. (1956) Histochemistry with labeled antibody. Int. Rev. Cytol., 5:123. Danscher, G. (1981) Histochemical demonstration of heavy metals. A revised version of the silver method suitable for both light and electron microscopy. Histochemistry, 71:116. Danscher, G. (1984) Autometallography. A new technique for light and electron microscopic visualization of metals in biological tissues (gold, silver, metal sulphides and metal selenides). Histochemistry, 81:331335. De Graaf, A., van Bergen en Henegouwen, P.M.P., Meijne, A.M.L., van Driel, R., and Verkleij, A.J. (1991) Ultrastructural localization of nuclear matrix proteins in HeLa cells using silver-enhanced ultrasmall gold probes. J. Histochem. Cytochem., 39:10351045. De Graaf, A., Humbel, B.M., Stuurman, N., van Bergen en Henegouwen, P.M.P., and Verkleij, A.J. (1992) Three-dimensional immunogold labeling of nuclear matrix proteins in permeabilized cells. Cell Biol. Int. Rep., 16:827836. Dundr, M., and Raska, I. (1993) Nonisotopic ultrastructural mapping of transcription sites within the nucleolus. Exp. Cell Res., 208:275 281. Fu, X.D., and Maniatis, T. (1990) Factor required for mammalian spliceosome assembly is localized to discrete regions in the nucleus. Nature, 343:437441. Griffiths, G. (1993) Fine Structure Immunocytochemistry. SpringerVerlag, Berlin/Heidelberg. Hayat, M.A. (ed.) (19891991) Colloidal Gold: Principles, Methods, and Applications. Academic Press, San Diego. Helenius, A., and Simons, K. (1975) Solubilization of membranes by detergents. Biochim. Biophys. Acta, 415:2979. Heuser, J.E. (1981) Comparing several detergents commenly used to visualize the cytoskeleton. J. Cell Biol., 91:302a.

Hohenberg, H., Mannweiler, K., and Muller, M. (1994) High-pressure freezing of cell suspensions in cellulose capillary tubes. J. Microsc., 175:3443. Humbel, B.M., and Biegelmann, E. (1992) A preparation protocol for postembedding immunoelectron microscopy of Dictyostelium discoideum cells with monoclonal antibodies. Scanning Microsc., 6:817 825. Humbel, B., and Muller, M. (1986) Freeze substitution and low temperature embedding. In: The Science of Biological Specimen Preparation 1985. M. Muller, R.P. Becker, A. Boyde, and J.J. Wolosewick, eds. SEM, AMF OHare, pp. 175183. Humbel, B., Marti, T., and Muller, M. (1983) Improved structural preservation by combining freeze substitution and low temperature embedding. In: Beitrage zur elektronenmikroskopischen Direktab bildung von Oberachen. Vol. 16. G. Pfefferkorn, ed. Antwerpen, pp. 585594. Johnson, G.D., and de C. Nogueira Araujo, G.M. (1981) A simple method of reducing the fading of immunouorescence during microscopy. J. Immunol. Meth., 43:349 350. Kellermayer, M., Ludany, A., Jobst, K., Szucs, G., Trombitas, K., and Hazlewood, C.F. (1986) Cocompartmentation of proteins and K within the living cell. Proc. Natl. Acad. Sci. USA, 83:10111015. Koster, A.J., Grimm, R., Typke, D., Hegerl, R., Stoschek, A., Walz, J., and Baumeister, W. (1997) Perspectives of molecular and cellular electron tomography. J. Struct. Biol., 120:276308. Krijnse-Locker, J., Ericsson, M., Rottier, P.J.M., and Griffiths, G. (1994) Characterization of the budding compartment of mouse hepatitis virus: Evidence that transport from RER to the Golgi complex requires only one vesicular transport step. J. Cell Biol., 124:55 70. Langanger, G., De Mey, J., Moeremans, M., Daneels, G., De Brabander, M., and Small, J.V. (1984) Ultrastuctural localization of -actinin and lamin in cultured cells with the immunogold staining (IGS) method. J. Cell Biol., 99:13241334. Leno, G.H., Downes, C.S., and Laskey, R.A. (1992) The nuclear membrane prevents replication of human G2 nuclei but not G1 nuclei in Xenopus egg extract. Cell, 69:151158. Louvard, D., Reggio, H., and Warren, G. (1982) Antibodies to the Golgi complex and the rough endoplasmic reticulum. J. Cell Biol., 92:92 107. Macville, M.V.E., Wiesmeijer, K.C., Dirks, R.W., Fransen, J.A.M., and Raap, A.K. (1995) Saponin pre-treatment in pre-embedding electron microscopic in situ hybridization for the detection of specic RNA sequences in cultured cells: A methodological study. J. Histochem. Cytochem., 43:10051018. Martone, M.E., Pollock, J.A., Jones, Y.Z., and Ellisman, M.H. (1996) Ultrastructural localization of dendritic messenger RNA in adult rat hippocampus. J. Neurosci., 16:74377446. Matsuno, A., Ohsugi, Y., Utsunomiya, H., Takekoshi, S., Osamura, R.Y., Watanabe, K., Teramoto, A., and Kirino, T. (1994) Ultrastructural distribution of growth hormone and prolactin mRNAs in normal rat pituitary cells: A comparison between pre-embedding and postembedding methods. Histochemistry, 102:265270. Mehlin, H., Daneholt, B., and Skoglund, U. (1992) Translocation of a specic premessenger ribonucleoprotein particle through the nuclear pore studied with electron microscope tomography. Cell, 69:605 613. Melan, M.A., and Sluder, G. (1992) Redistribution and differential extraction of soluble proteins in permeabilized cultured cells. Implications for immunouorescence microscopy. J. Cell Sci., 101:731 743. Menco, B.P.M. (1986) A survey of ultra-rapid cryoxation methods with particular emphasis on applications to freeze-fracturing, freezeetching, and freeze-substitution. J. Electron Microsc. Tech., 4:177 240. Monneron, A., and Bernhard, W. (1969) Fine structural organization of the interphase nucleus in some mammalian cells. J. Ultrastruct. Res., 27:266288. Muller, M., and Moor, H. (1984) Cryoxation of thick specimens by high pressure freezing. In: Science of Biological Specimen Preparation 1983. J.P. Revel, T. Barnard, and G.H. Haggis, eds. SEM, AMF OHare, pp. 131138. Nakane, P.K., and Pierce, G.B. (1967) Enzyme-labelled antibodies: Preparation and application for the localization of antigens. J. Histochem. Cytochem., 14:929931. Powell, R.D., Halsey, C.M.R., Spector, D.L., Kaurin, S.L., McCann, J., and Hainfeld, J.F. (1997) A covalent uorescent-gold immunoprobe: Simultaneous detection of a pre-mRNA splicing factor by light and electron microscopy. J. Histochem. Cytochem., 45:947956. Robards, A.W., and Sleytr, U.B. (eds.) (1985) Low Temperature Meth-

58

B.M. HUMBEL ET AL.

particles. In: Immunogold-Silver Staining: Principles, Methods, and Applications. M.A. Hayat, ed. CRC Press, Boca Raton, pp. 97118. Studer, D., Michel, M., Wohlwend, M., Hunziker, E.B., and Buschmann, M.D. (1995) Vitrication of articular cartilage by highpressure freezing. J. Microsc., 179:321332. Tanner, V.A., Ploug, T., and Tao Cheng, J.H. (1996) Subcellular localization of SV2 and other secretory vesicle components in PC12 cells by an efficient method of pre-embedding EM immunocytochemistry for cell cultures. J. Histochem. Cytochem., 44:14811488. Tas, J., and Westerneng, G. (1981) Fundamental aspects of the interaction of propidium diiodide with nucleic acids studied in a model system of polyacrylamide lms. J. Histochem. Cytochem., 29:929936. Testillano, P.S., Sanchez-Pina, M.A., Olmedilla, A., Ollacarizqueta, M.A., Tandler, C.J., and Risueno, M.C. (1991) A specic ultrastruc tural method to reveal DNA: The NAMAUr. J. Histochem. Cytochem., 39:14271438. Tokuyasu, K.T. (1973) A technique for ultracyotomy of cell suspensions and tissues. J. Cell Biol., 57:551565. Tokuyasu, K.T. (1986) Application of cryoultramicrotomy to immunocytochemistry. J. Microsc., 143:139149. Van Gijlswijk, R.P.M., Wiegant, J., Raap, A.K., and Tanke, H.J. (1996) Improved localization of uorescent tyramides for uorescence in situ hybridization using Dextran sulfate and polyvinyl alcohol. J. Histochem. Cytochem., 44:389392. Van Gijlswijk, R.P.M., Zijlmans, H.J.M.A.A., Wiegant, J., Bobrow, M.N., Erickson, T.J., Adler, K.E., Tanke, H.J., and Raap, A.K. (1997) Fluorochrome-labeled tyramides: Use in immunocytochemistry and uorescence in situ hybridization. J. Histochem. Cytochem., 45:375 382. Van Lookeren Campagne, M., Dotti, C.G., Jap Tjoen San, E.R.A., Verkleij, A.J., Gispen, W.H., and Oestreicher, A.B. (1992) B-50/ GAP43 localization in polarized hippocampal neurons in vitro: An ultrastructural quantitative study. Neuroscience, 50:3552. Venable, J.H., and Coggeshall, R. (1965) A simplied lead citrate stain for use in electron microscopy. J. Cell Biol., 25:407408. Verkleij, A.J., and Leunissen, J.L.M. (eds.) (1989) Immuno-Gold Labeling in Cell Biology. CRC Press, Boca Raton, FL. Wansink, D.G., Schul, W., van der Kraan, I., van Steensel, B., van Driel, R., and de Jong, L. (1993) Fluorescent labeling of nascent RNA reveals transcription by RNA polymerase-II in domains throughout the nucleus. J. Cell Biol., 122:283293. Yazama, F., Esaki, M., and Sawada, H. (1997) Immunocytochemistry of extracellular matrix components in the rat seminiferous tubule: Electron microscopic localization with improved methodology. Anat. Rec., 248:5162. Young, H.M., and Furness, J.B. (1995) Ultrastructural examination of the targets of serotonin-immunoreactive descending interneurons in the guinea pig small intestine. J. Comp. Neurol., 356:101114.

ods in Biological Electron Microscopy. Elsevier, Amsterdam/New York/Oxford. Rodriguez, J., and Deinhardt, F. (1960) Preparation of a semipermanent mounting medium for uorescent antibody studies. Virology, 12:316317. Roth, J. (1989) Postembedding labeling on Lowicryl K4M tissue sections: Detection and modication of cellular components. In: Methods in Cell Biology. Vol. 31. A.M. Tartakoff, ed. Academic Press, New York, pp. 513551. Satijn, D.P.E., M.J., G., van der Vlag, J., Hamer, K.M., Schul, W., Alkema, M.J., Saurin, A.J., Freemont, P.S., van Driel, R., and Otte, A.P. (1997) RING1 is associated with the polycomb group protein complex and acts as a transcriptional repressor. Mol. Cell. Biol., 17:41054113. Schliwa, M., van Blerkom, J., and Porter, K.R. (1981) Stabilization of the cytoplasmic ground substance in detergent-opened cells and a structural and biochemical analysis of its composition. Proc. Natl. Acad. Sci. USA, 78:43294333. Schliwa, M., Euteneuer, U., and Porter, K.R. (1987) Release of enzymes of intermediary metabolism from permeabilized cells: Further evidence in support of a structural organization of the cytoplasmic matrix. Eur. J. Cell Biol., 44:214218. Sibon, O.C.M., Humbel, B.M., de Graaf, A., Verkleij, A.J., and Cremers, F.F.M. (1994) Ultrastructural localization of epidermal growth factor (EGF)-receptor transcripts in the cell nucleus using preembedding in situ hybridization in combination with ultra-small gold probes and silver enhancement. Histochemistry, 101:223232. Sibon, O.C.M., Cremers, F.F.M., Humbel, B.M., Boonstra, J., and Verkleij, A.J. (1995) Localization of nuclear RNA by pre- and post-embedding in situ hybridization using different gold probes. Histochem. J., 27:3545. Simon, G.T., Thomas, J.A., Chorneyko, K.A., and Carlemalm, E. (1987) Rapid embedding in Lowicryl K4M for immunoelectron microscopic studies. J. Electron Microsc. Tech., 6:317324. Steinbrecht, R.A., and Zierold, K. (eds.) (1987) Cryotechniques in Biological Electron Microscopy. Springer-Verlag, Berlin/Heidelberg. Stierhof, Y.-D., Schwarz, H., and Frank, H. (1986) Transverse sectioning of plastic-embedded immunolabeled cryosections: Morphology and permeability to protein A-colloidal gold complexes. J. Ultrastruct. Mol. Struct. Res., 97:187196. Stierhof, Y.-D., Humbel, B.M., and Schwarz, H. (1991a) Suitability of different silver enhancement methods applied to 1 nm colloidal gold particles: An immunoelectron microscopy study. J. Electron Microsc. Tech., 17:336343. Stierhof, Y.-D., Schwarz, H., Menz, B., Russell, D.G., Quinten, M., and Overath, P. (1991b) Monoclonal antibodies to Leishmania mexicana promastigote antigens. II. Cellular localization of antigens in promastigotes and infected macrophages. J. Cell Sci., 99:181186. Stierhof, Y.-D., Hermann, R., Humbel, B.M., and Schwarz, H. (1995) Use of TEM, SEM, and STEM in imaging 1-nm colloidal gold