Ultrastructural immunocytochemistry requires that minimal damage to antigens is imposed by the processing methods. Immersion fixation in cross-linking fixatives with their potential to damage antigens is not an ideal approach and rapid freezing as an alternative sample-stabilization step has a number of advantages. Rapid freezing at ambient pressure restricts the thickness of well-frozen material obtainable to ≈ 15 μm or less. In contrast, high-pressure freezing has been demonstrated to provide ice-crystal-artefact-free freezing of samples up to 200 μm in thickness. There have been few reports of high-pressure freezing for immunocytochemical studies and there is no consensus on the choice of post-freezing sample preparation. A range of freeze-substitution time and temperature protocols were compared with improved tissue architecture as the primary goal, but also to compare ease of resin-embedding, polymerization and immunocytochemical labelling. Freeze-substitution in acetone containing 2% osmium tetroxide followed by epoxy-resin embedding at room temperature gave optimum morphology. Freeze-substitution in methanol was completed within 18 h and in tetrahydrofuran within 48 h but the cellular morphology of the Lowicryl-embedded samples was not as good as when samples were substituted in pure acetone. Acetone freeze-substitution was slow, taking at least 6 days to complete, and gave blocks which were difficult to embed in Lowicryl HM20. Careful handling of frozen samples avoiding rapid temperature changes reduced apparent ice-crystal damage in sections of embedded material. Thus a slow warm-up to freeze-substitution temperature and a long substitution time in acetone gave the best results in terms of freezing quality and cellular morphology. No clear differences emerged between the different freeze-substitution media from immunocytochemical labelling experiments.
Protein detection by immunocytochemistry at the ultrastructural level requires that the processing method does not alter an antigen beyond recognition by its specific antibody. For the majority of antigens, routine fixation and processing protocols for electron microscopy will not allow subsequent immunocytochemistry. A number of methods has been devised to try to reduce the damage to antigens in the tissue while still retaining, as near as possible, the familiar appearance of aldehyde/osmium-stabilized cells. The simplest approach is to reduce exposure to cross-linking fixatives to a minimum and it is possible to detect many antigens following mild fixation (e.g. low concentrations of paraformaldehyde with or without very low concentrations of glutaraldehyde) for relatively short times (30–60 min). Both low-temperature embedding in Lowicryl resin by the progressive lowering of temperature (PLT) technique ( Carlemalm et al., 1982 ; Robertson et al., 1992 ) and thawed cryosections ( Tokuyasu, 1986; reviewed by Griffiths, 1993) rely on such tissue stabilization. There remain two potential problems with immersion fixation in aldehydes. The first is that with some antigen/antibody combinations the antibody binding is reduced below the level of detection, and even if an antigen is detected following fixation there remains uncertainty over how much antigen has been lost. The second problem is that immersion fixation is slow, giving the possibility of redistribution of cellular constituents during the fixation process.
Rapid freezing followed by freeze-substitution overcomes these two problems and offers an effective alternative to immersion fixation techniques. Rapid-freezing methods, which are suitable for immunocytochemical studies (reviewed by Echlin, 1992), include impact freezing, jet freezing and plunge freezing. However, all are restricted in the depth of satisfactory freezing obtainable. A maximum of 10–15 μm of well-frozen sample in contact with the cryogen is the most that can be achieved. In contrast, high-pressure freezing has the potential to freeze up to 200 μm of biological samples without apparent ice-crystal artefacts ( Studer et al., 1989 ; Sartori et al., 1993 ) and offers the morphologist the ability to retain excellent cellular ultrastructure ( McDonald & Morphew, 1993; Benchimol, 1994) and the immunocytochemist a way to avoid aldehyde fixatives ( Young et al., 1995 ). However, surprisingly few reports show immunocytochemical labelling of high-pressure frozen material. One of the main uncertainties in this area is not so much the freezing process, where several approaches have been devised to improve tissue handling prior to freezing ( Hohenberg et al., 1994 , 1996), but the definition of protocols for resin embedding which are optimal for immunolabelling.
In an ideally frozen sample, the water in the specimen will be in the form of vitreous or amorphous ice. Ice artefact-free stabilization of a wide range of samples has been demonstrated in high-pressure frozen samples ( Studer et al., 1989 , 1992; Zhang et al., 1993 ; Szczesny et al., 1996 ;) and the presence of vitreous ice has been confirmed by examination of the electron diffraction pattern of the ice form in cryosections ( Michel et al., 1991 ). In practice, samples will contain a combination of vitreous and crystalline ice ( Richter, 1994) but as long as ice crystals are smaller than a few nanometres, the ultrastructure of the cells following freeze-substitution and embedding appears undamaged. Following freezing, the frozen water must be removed if the sample is to be resin embedded and this must be done without permitting the regrowth of ice crystals. This can be achieved by freeze-substitution or freeze drying and for immunolabelling studies, by low temperature embedding. The concept of combining freeze-substitution and low-temperature embedding was introduced by Humbel & Müller (1983, 1986), and at around the same time freeze drying was given some impetus with the development of methods described by Linner et al. (1986 ), but the majority of publications have relied upon freeze-substitution.
During freeze-substitution, a technique which was originally developed for morphological studies ( Van Harreveld & Crowell, 1964), the ice in the frozen sample is removed by a solvent at low temperature. For these morphological studies, the substitution solvent will contain fixatives, usually osmium tetroxide, and after removal of ice is complete, the sample is warmed to room temperature and embedded by routine methods. The most commonly used substitution protocol uses acetone containing 2% osmium tetroxide. In general, although the effects of fixatives at substitution temperatures are uncertain, for immunocytochemical studies fixatives are best avoided. This can be achieved by freeze-substitution in solvent without fixatives ( Monaghan & Robertson, 1990) followed by embedding in a low-temperature resin which is polymerised at low temperature by UV light. The Lowicryl resins are ideal in this context.
In our initial high-pressure freezing experiments, the yield of well-frozen samples was variable and the freezing quality varied within samples. While it is clear that the quality of freezing is influenced by the nature of the sample being frozen ( Studer et al., 1995 ; Erk et al., 1996 ), a further potential problem is the post-freezing handling of the frozen (hopefully vitrified) sample when it is brought to substitution temperature where there is the risk of ice recrystallization. We have investigated the role of sample handling during freeze-substitution and also a number of different freeze-substitution protocols on the yield of well-frozen and embedded samples. The aim was to define a protocol which gave the best retention of cellular structures and antigenicity and to identify procedures which might influence the proportion of well-frozen samples obtained.
Materials and methods
Samples of solid tissues were obtained from human and mouse biopsy. Lymphoma cell lines and human breast tumour cells were obtained from tissue culture. Samples of normal human breast tissue were obtained from cosmetic reduction mammoplasty. Mice were killed by cervical dislocation and samples of heart, kidney, skeletal muscle and lactating mammary gland were rapidly removed, and to prevent dehydration were transferred to either an air-buffered tissue culture medium (L15, Gibco, BRL) or 1-heaxdecene (Sigma, Poole, U.K.), ( Studer et al., 1989 ). They were sliced into ≈ 200 μm slices, discs of tissue 3 mm in diameter were cut with the end of a sharpened stainless-steel tube and these were then placed into the 200 μm depression of a high-pressure freezing planchette (Bal-Tec, Liechtenstein). The sample was enclosed by the flat back of a 300 μm planchette, giving a sample chamber depth of 200 μm. The two planchettes were then placed in the sample holder of a Bal-Tec HPM010 high-pressure freezer and frozen.
The human lymphoma cell lines K562 and W1 L2 and the human breast tumour cell line PMC42 ( Whitehead et al., 1983 ) were maintained in tissue culture and cells were introduced into 200 μm capillary tubing as described by Hohenberg et al. (1994 ). Retention of the cells within the tubes through substitution and embedding steps was not always complete, and in order to overcome this W1 L2 cells were grown inside the tubes. The tubes were sterilized in alcohol and, under sterile conditions, the cells were introduced as described previously and the tubes replaced in tissue culture medium for 24–48 h. The cells appeared by phase-contrast microscopy to continue to divide and filled the tubes. These were then frozen as before, surrounded by 1-hexadecene. Cell loss during processing was much reduced by this approach. Following freezing, the planchettes were split apart under liquid nitrogen and the sample planchette containing the tissue was transferred to a Leica (Milton Keynes, U.K.) AFS freeze-substitution system. Transfer was either directly to the substitution temperature (– 90 °C) or to − 160 °C (the lowest setting for the AFS) with a subsequent warming at 5 °C h−1 to − 90 °C. Frozen samples were either left in the sample planchettes or carefully separated from the planchettes. Freeze-substitution protocols were as follows. For morphological studies, samples were substituted in acetone containing 2% osmium tetroxide for 8 h at − 90 °C, 8 h at − 60 °C and 8 h at − 30 °C ( Müller et al., 1980 ). Following a change to pure acetone the samples were warmed to room temperature and embedded in epoxy resin as described previously ( Monaghan et al., 1985 ). For immunocytochemical studies, all substitution was in solvent without fixative. Methanol substitution was for 36, 24, 18 or 8 h at − 90 °C followed by warm-up at 10 °C h−1 to − 50 °C. Tetrahydrofuran substitution was as described by Palsgard et al. (1994 ) and was for 48 h at − 80 °C. Acetone substitution protocols were 6 days at − 90 °C, followed by warm-up at 10 °C h−1 to − 50 °C ( Edelmann, 1991), or 24 h at − 90 °C, warmed at 4 °C h−1 to − 80 °C for 5 days, followed by 4 °C h−1 warm-up to − 60 °C for 24 h ( Kaneko & Walther, 1995). Acetone was dried over an activated molecular sieve which had been dried in an oven prior to use, and the substitution containers had a small quantity of molecular sieve at the base, with the samples supported on a layer of filter paper.
At the end of each substitution protocol, the samples were warmed to − 50 °C for embedding in Lowicryl HM20. Lowicryl HM20 was chosen for this work as it has proven successful in previous low-temperature embedding studies ( Kellenberger et al., 1987 ; Monaghan & Robertson, 1990; Robertson et al., 1992 ). Following acetone substitution, samples were either infiltrated directly into HM20, which resulted in variable embedding quality, or transferred to ethanol at − 50 °C for 60 min prior to infiltration. This latter step was introduced as ethanol appears to be more compatible with Lowicryl HM20 when compared with acetone. Substituted samples were infiltrated in HM20 broadly as described in the manufacturer's information. Samples were infiltrated in 3:1 ethanol:HM20, 1:1 ethanol:HM20, 1:3 ethanol:HM20, all for 60 min each, followed by HM20 for 2 × 60 min. This was followed by 18 h in HM20 and one further resin change before embedding. Polymerization was for 48 h at − 50 °C and a further 48 h at room temperature after a 10 °C h−1 warming.
Resin blocks were sectioned on diamond knives and the sections collected on formvar-coated grids and contrasted with uranyl acetate and lead citrate in a Leica Ultrostainer or EM Stain. They were examined in a Philips CM100 equipped with a Biotwin high-contrast lens at 80 kV.
For immunocytochemistry, sections were collected on formvar-coated gold grids. They were immunolabelled as described previously ( Monaghan & Robertson, 1990). Briefly, the grids were floated on drops of a blocking buffer solution consisting of phosphate-buffered saline (PBS), 0.8% bovine serum albumin (BSA), 5% fetal calf serum (FCS) and 0.1% fish gelatin for 30 min. They were transferred to primary antibody and incubated at room temperature overnight. A range of antigens has been successfully demonstrated on samples substituted in methanol, acetone and tetrahydrofuran. The primary antibodies diluted in blocking buffer as indicated include anti-vimentin (1:10, Dako, High Wycombe, U.K.), anti-keratin 14 (1:400, Sigma), anti-collagen IV (1:10 Dako), anti-protein disulfide isomerase (PDI 1:5000, Bioquote, York, U.K.), anti-α-tubulin (1:250 Sigma), anti-β-tubulin (1:1000, Sigma), anti-connexin26 (1:100 Monaghan et al., 1994 ), anti-connexin43 (1:100, Cambridge Bioscience, Cambridge, U.K.), anti-skeletal muscle actin (1:50). After incubation, the grids were washed on three separate droplets of washing buffer (blocking buffer without FCS) for 5 min each and transferred to droplets of antirabbit or antimouse IgG conjugated to 5 nm colloidal gold (Amersham International, Little Chalfont, U.K.) diluted 1:40 in blocking buffer for 60 min. Following three more 5-min-washing-buffer washes the grids were washed on droplets of pure water (e.g. double glass distilled or reverse osmosis 18 mΩ), again for 3 × 5 min. The colloidal gold was silver enhanced for 6 or 12 min using IntenSe M (Amersham International) and the grids were washed in pure water and air dried. The dried grids were contrasted as described above.
Freezing quality was assessed at the ultrastructural level as almost all blocks appeared well frozen at the light microscope level. Cell nuclei were often the first cellular component to show ice-crystal damage, and red blood cells were also places to look for ice-crystal growth. The overall yield of well-frozen samples was increased to more than 80% when the samples were warmed slowly from liquid nitrogen temperature to substitution temperature. When the samples were substituted without removal from the planchette, they often showed ice-crystal artefacts on one side whereas the other side showed significantly fewer ice-crystal artefacts. This effect was reduced when the samples were removed from the planchettes prior to substitution. This is, however, a difficult procedure and it is easy to damage the sample during removal from the planchette. When comparing the results obtained with 1-hexadecene and L15 tissue culture medium as the fluid surrounding the sample in the planchette, less ice crystal damage was obtained with 1-hexadecene, and that was used for the majority of the experiments.
No difficulties were encountered in infiltration and embedding in HM20 following methanol- and tetrahydofuran substitution. Blocks were generally well infiltrated and the resin fully polymerized. Following acetone substitution, satisfactory infiltration and embedding in Lowicryl HM20 proved difficult. Nuclei in particular were not well infiltrated and the resin was not uniformly polymerized. Ultimately, the solution adopted was to introduce a 60 min change in ethanol at − 50 °C immediately prior to resin infiltration. All illustrations are from samples processed in this way.
Substitution in methanol was rapid, and resulted in well-infiltrated and polymerized blocks. Following initial substitution for 36 h, the substitution times were progressively reduced from 36 h to determine minimum substitution times. Less than 18 h substitution gave variable results and this was chosen as a standard methanol-substitution protocol. Samples freeze-substituted for 18 h or longer in methanol embedded well in Lowicryl HM20. The contrast of the samples was acceptable, but the cells had an extracted appearance with a lack of definition of cytoplasmic membranes and organelles ( Fig. 1). Overall there was considerable variation in the appearance of different samples; there was often a lack of electron density in the tissue and the cytoplasm appeared to lack the full complement of organelles. A common but not universal observation was an empty halo around mitochondria in methanol freeze-substituted samples ( Fig. 1). Immunolabelling was successfully undertaken on a number of samples and an example of labelling for the intermediate filament protein vimentin in a human breast tumour cell line (PMC42) is illustrated in Fig. 2.
Only one freeze-substitution protocol was used, and this resulted in well-infiltrated and embedded blocks. The overall appearance of samples substituted in tetrahydrofuran was similar to that seen in methanol-substituted blocks. The samples appeared extracted and generally gave images of very low contrast, although this appearance was variable. With the exception of mitochondria, rough endoplasmic reticulum and cytoskeletal elements, the cytoplasmic organelles were difficult to recognize ( Fig. 3). The section has been labelled with antibodies to α-tubulin, and the colloidal gold is present over the cytoplasm. In tetrahydrofuran-substituted samples the mitochondria often showed increased electron density compared with the rest of the cell organelles.
When samples were freeze-substituted in acetone in the presence of 2% osmium tetroxide, followed by epoxy resin embedding, the contrast and cytoplasmic detail were similar to that seen in material that has been fixed in aldehyde and osmium tetroxide and epoxy resin embedded. Representative samples are illustrated in Fig. 4 (skeletal muscle) and Fig. 6 (lactating mouse mammary gland) for comparison with the ‘acetone only’ freeze-substituted samples (Figs. 5 and 7). In samples substituted in acetone without fixative (Figs. 5, 7–9), using either of the substitution protocols, the morphology was significantly better than that seen after methanol or tetrahydrofuran freeze-substitution. The cell cytoplasm did not appear extracted and cellular organelles were readily observed although there were subtle differences between the acetone/osmium- and pure-acetone-substituted samples. Much of this was due to the variable appearance of membranes, which tended to appear in negative contrast. Examples are illustrated in Fig. 5 (skeletal muscle), Fig. 7 (lactating mouse mammary gland), Fig. 8 (mouse kidney) and Fig. 9 (tissue culture cell line W1 L2). The ultrastructural improvement over other substitution media was seen in all samples, but tissue culture cells showed less intrinsic contrast than solid tissues. Overall, the results were marginally better with the freeze-substitution protocol of Kaneko & Walther (1995) and all figures are from samples processed by this method.
W1 L2 cells introduced under sterile conditions into capillary tubes appeared to continue to divide as normal and using phase contrast microscopy were seen to fill the tubes after 48 h in culture. Following processing, the cells showed no difference in morphology when compared with cells frozen directly from tissue culture, and there were no areas in the tubes showing increased cell death. The ultrastructure of these cells substituted in acetone followed by embedding in Lowicryl HM20 is seen in Fig. 9. In these cells, the mitochondria show comparatively low contrast compared with the rest of the cell cytoplasm.
A number of antigens have been successfully detected on sections of methanol ( Fig. 2) tetrahydrofuran ( Fig. 3) and acetone (Figs. 5, 7 and 8) freeze-substituted specimens. The sections from acetone freeze-substituted samples showed a good combination of immunolabelling and section contrast, allowing unequivocal identification of cytoplasmic organelles. Skeletal muscle myosin filaments are labelled with an antimyosin antibody ( Fig. 5 inset), connexin26 (a gap junction protein) has been localized to a gap junction plaque ( Fig. 7 inset), and protein disulfide isomerase (PDI) is identified in the rough endoplasmic reticulum of mouse kidney tubule ( Fig. 8). Background labelling was low, irrespective of the freeze-substitution protocol used, and the best combination of sensitivity and visibility was obtained by utilizing a 5 nm colloidal gold conjugate with short silver enhancement. Control sections, labelled in the absence of primary antibody, or with an irrelevant antibody against a protein not present in the sample were not labelled. Not all antibodies used gave positive labelling.
Rapid freezing followed by freeze-substitution offers the advantage for ultrastructural immunocytochemistry that cross-linking fixatives can be completely avoided without compromising morphology. High-pressure freezing extends that advantage in that significantly larger samples can be adequately frozen without ice-crystal artefacts. We have addressed the processes which follow high-pressure freezing from an immunocytochemist's perspective in order to determine protocols which will give maximum ice-crystal-artefact-free samples with minimum loss of morphology. The freeze-substitution, embedding and polymerization protocols are lengthy, making the development process highly protracted. In addition, almost every high-pressure frozen block looks well-frozen in 1 μm toluidine blue sections, so that all quality control must be undertaken on thin sections. Despite these difficulties, a number of factors have been identified which appear to influence the yield of well-frozen and embedded samples and the quality of cellular preservation achieved.
One of the unexpected results of this work was the influence of the rate of change of warming from liquid nitrogen temperature to freeze-substitution temperature. The slow warm-up for the frozen samples to freeze-substitution temperature was introduced for two reasons: first, to determine whether this could be a factor in overall sample freezing appearance by influencing ice recrystallization, and secondly to attempt to reduce the cracks that are seen in samples transferred directly from liquid nitrogen to freeze-substitution temperature. Comparisons were made of samples from the same freezing run where half were transferred directly to freeze-substitution temperature and half were slowly warmed followed by identical freeze-substitution and embedding. Analysis of the embedded blocks from these experiments made it clear that ice-crystal damage (due presumably to ice recrystallization) was less prominent in the slowly warmed samples and the cracks in the samples were less common.
The morphology of samples freeze-substituted in acetone containing osmium tetroxide was excellent as has been demonstrated in a number of previous reports ( Studer et al., 1989 ; Eggli & Graber, 1994) and initial experiments used pure acetone freeze-substitution for immunocytochemical studies. Considerable difficulties were encountered, particularly with the embedding process, and these led to the investigation of methanol and tetrahydrofuran as alternatives. Initial results with the longer freeze-substitution times in methanol gave poor retention of cytoplasmic detail and the freeze-substitution times were reduced to determine if shorter freeze-substitution times would improve the cellular appearance. Even the shortest freeze-substitution times did not give improved morphology. Tetrahydrofuran was reported to be a superior freeze-substitution medium for subsequent X-ray microanalysis ( Palsgard et al., 1994 ) and it was tested to determine whether it would prove suitable for immunocytochemical studies. Overall, the results with tetrahydrofuran freeze-substitution gave satisfactory Lowicryl HM20 embedding, but morphology similar to that obtained with methanol, which was of poorer quality than required.
It is not certain at present whether the choice of freeze-substitution medium has any effect upon the level of immunocytochemical labelling on the section, but from a morphological point of view, although methanol freeze-substitution is quick, reproducible and leads to satisfactory embedding, the structural detail retained in the samples is poor. Nonetheless, immunocytochemical localization was satisfactory on sections of methanol-substituted samples. Similar results were seen in tetrahydrofuran-substituted samples.
The morphology of samples substituted with pure acetone was the nearest to that seen in acetone/osmium freeze-substituted samples. This would be predicted from the observations of Weibull & Christiansson (1986), where extraction of cellular lipids was greater with methanol than with acetone. Membrane contrast was clearly inferior when compared with acetone/osmium freeze-substituted samples, being seen sometimes in reverse contrast, but when compared with methanol or tetrahydrofuran freeze-substitution, the samples had a far less apparent extraction of cytoplasmic contents. However, freeze-substitution with acetone requires care in that it must be dried over a molecular sieve or similar dehydration agent to achieve adequate freeze-substitution. Acetone containing more than 1% water will not freeze-substitute samples fully ( Humbel et al., 1983 ). The difficulties in obtaining satisfactory Lowicryl embedding following acetone freeze-substitution were resolved by the short change into ethanol. Clearly this is not an ideal solution and further work is needed to circumvent this problem. Nonetheless, a slow warm-up to substitution temperature followed by a long freeze-substitution in acetone must remain the method of choice at present.
Many questions remain to be resolved. Perhaps the outstanding one is the question of freeze-substitution protocol. There is no doubt that freeze-substitution in acetone/osmium at − 90 °C, − 60 °C and − 30 °C in a relatively short protocol ( Müller et al., 1980 ) yields excellent results. The same protocol with pure acetone was not successful, giving blocks that were very badly infiltrated. Increased freeze-substitution times in pure acetone have been recommended for impact-frozen samples ( Edelmann, 1991) and these increased freeze-substitution times were successful for high-pressure frozen samples. The use of freeze-substitution media without additional fixatives was aimed at the establishment of a protocol which would work in as wide a range of situations as possible. Nonetheless, the activity of fixatives in organic solvents at low temperatures is almost certainly reduced ( Steinbrecht & Müller, 1987), but by no means fully understood. Indeed, recent publications indicate that some antigens are detectable in high-pressure frozen samples substituted in acetone containing osmium tetroxide ( Favre et al., 1995 ) or uranyl acetate ( Audit et al., 1996 ). For any one antigen/antibody combination, investigating the effects of freeze-substitution media containing low concentrations of fixatives may prove beneficial ( Eppenberger-Eberhardt et al., 1997 ) but at present, this approach must be seen as trial and error. With the choice of antigens detected in this study, no discernible difference was detected in the level of immunocytochemical labelling between samples substituted in acetone, methanol or tetrahydrofuran. Similarly several antibodies failed to react with sections of material freeze-substituted in all three solvents. Whether freeze drying followed by low-temperature embedding will prove advantageous for some antigens remains to be determined.
This work was supported by the Cancer Research Campaign.