Study of vitrified, unstained frozen tissue sections by cryoimmunoelectron microscopy

The ultrastructural mapping of biological macromolecules in cells and tissues is primarily based on immunoelectron microscopy (Griffiths, 1991). In preparing specimens for such analysis, three distinct and often conflicting demands need to be met: the preservation of cellular ultrastructure, the retention of antigenicity of the molecules of interest and their accessibility to the immunochemical reagents. The stabilization of cellular substructure usually requires extensive fixation, which may adversely affect the exposure of antigenic moieties, and may often modify them (Kyte, 1976; Kraehenbuhl et al. 1977; Geiger et al. 1981; Griffiths, 1991). Thus, for high-resolution immuno-EM labeling an optimal compromise between these different requirements has to be established.

Immunolabeling of thawed ultrathin frozen sections (Tokuyasu, 1973; Tokuyasu and Singer, 1976; Singer et al. 1982) is to date the method of choice for locating biological macromolecules within a tissue, despite the fact that some structural details are lost, distorted or remain invisible (Griffiths, 1991). Improved preservation of gross cellular ultrastructure is usually attainable by the embedding of extensively fixed tissues in polymeric resins, yet such specimens are of limited use for immunological localization. In essentially all immuno-EM techniques involving the study of tissues the samples are inevitably dehydrated either before embedding or following sectioning and immunolabeling. It is conceivable that much of the distortion of cellular substructure is introduced during this step. A way of avoiding dehydration in the study of biological specimens is to embed the tissue, the section or both in a vitreous layer. This involves cooling of the sample so rapidly that the water molecules do not crystallize, but remain in an amorphous or vitrified state (Adrian et al. 1984). The sample is then examined in the TEM at temperatures around −170 °C under conditions where the vapor pressure of water is negligible, and thus they remain fully hydrated. To date, this approach has been confined to samples that can be dispersed into a fine suspension.

We present here a novel methodology, namely, the use of vitrified frozen sections (VFS), that enables direct examination of unstained, hydrated tissue sections in conjunction with immunogold labeling. The method involves preparation of ultrathin frozen sections of fixed tissues, recovery and immunolabeling of these sections on EM grids, embedding of the labeled but unstained sections in vitreous (amorphous) ice, and examination of the fully hydrated sections at near liquid nitrogen temperatures by cryo-EM. The study of smooth muscle, heart and kidney tissue using this technique showed excellent preservation of fine structural details, observed directly without any additional staining.

We present here the major features of this methodology and discuss its potential advantages for the ultrastructural analysis and immunocytochemical molecular mapping of cells and tissues.

Muscular tissues were freshly dissected from 3-to 4-week-old chickens and incubated for 3 mm in relaxation buffer (Fawcett and McNutt, 1969). Tissue blocks were then transferred and further dissected in the fixative solution. Unless stated otherwise, we have fixed for 24 h with Karnovsky’s fixative (3% formaldehyde, 2% glutaraldehyde, 5mM CaCl₂ in 100mM cacodylate buffer, pH 7.4) at room temperature and 10 h at 4°C. Lighter fixation was carried out with a mixture of 3% formaldehyde and 0.1% glutaraldehyde in the same buffer or with 3% formaldehyde alone. Tissues were either processed for Epon embedding (Polybed 812, PolySciences, USA) or impregnated in 100 mM cacodylate buffer, with 2.3 M sucrose, pH 7.4. The latter samples were quick-frozen in liquid nitrogen (close to its freezing temperature) and ultrathin frozen sections (500–1000 Å; 1Å=0.1nm) were cut at −115°C using a Reichert-Jung FS-4D ultracryo-microtome. The sections were recovered from the knife in a 2.3 M sucrose droplet according to the method of Tokuyasu and Singer (1976), and transferred to either uncoated copper grids (600–1000 mesh) or Formvar-coated 300 mesh grids. The grids were floated on water in a Petri dish at 4°C. Duration of the water rinse was between 1 h and 16 h, without any appreciable difference in the resulting ultrastructure. The sections were then either processed for methyl-cellulose mounting according to a modification of the “Tokuyasu method’ (Griffiths et al. 1984) or prepared for vitrification (see below). Some of the sections were immunolabelled using a-actinin antibodies followed by gold-conjugated goat anti-rabbit IgG (60R G-10, Janssen Life Sci., Belgium) (Slot and Geuze, 1981).

Grids were removed from the water with fine tweezers and most of the water was blotted off both sides with filter paper. The grids were rapidly plunged into liquid ethane, pre-cooled with liquid nitrogen close to its freezing temperature (Adrian et al. 1984). A Gatan cryo-transfer chamber was used to transfer the samples to the Gatan cold stage, and the specimens were examined in a Philips EM400T electron microscope, operated at 100 kV. Grids were maintained at −170 °C throughout their examination. Micrographs were taken using standard low-dose precautions and Kodak SO-163 plates at a nominal magnification of × 17 000 or ×22 000. Plates were developed for 12min in full-strength D-19 developer (Kodak, USA).

Chicken gizzard smooth muscle was used as our primary model system for the development of the new VFS technique, as it has previously been examined using a large variety of preparation techniques including ‘conventional’ Epon sections (Bagby, 1990; Somlyo et al. 1977), conventional frozen sections (Geiger et al. 1980, 1981; Volberg et al. 1986), freeze substitution (Tsukita et al. 1982), quick freezing/deep etching (Somlyo and Franzini-Armstrong, 1985) and polyvinyl alcohol embedding (Small et al. 1986).

During the development of the VFS procedure the following technical details are worth noting: (1) the frozen sections should be preferably between 500 Å and 700 Å. (2) Sections may be collected on either membrane-coated (200–300 mesh) or uncoated (600–1000 mesh) grids. The micrographs of the latter had somewhat better resolution and lower beam damage. (3) Blotting of most of the water from the grid prior to vitrification should not be too extensive. A thick layer of amorphous ice (about 1000–1500 Å above the section) appeared to enhance the stability of the sections and minimize radiation damage.

Examination of unstained VFS revealed an elaborate network of cytoplasmic filaments, which was visualized both in longitudinal (Fig. 1A) and in cross-sections (Figs IB and 2C). Most abundant were thin filaments, with an average diameter of 70–90 Å (tf in Fig. 2C), usually arranged in groups or bundles. These filaments, clearly related to F-actin visualized by other techniques (see, for comparison, Fig. 6 of Tsukita et al. (1982) and Fig. 4b of Small et al. (1986)), were periodically associated with elongated electron-dense bodies (Fig. 1A and B; Fig. 2C matched arrows in the stereo pairs, Fig. 5B and B ‘). Cross-sections of these bodies showed that filaments of the same diameter continue through their cores (Fig. 2C). In addition to the actin filaments, many denser fibers (approximate dimensions in cross-sections: 200ÅX600Å) were detected along the thin filament arrays, apparently corresponding to the-thick myosin filaments of smooth muscle (Fig. 2C, m). A third class of sarcoplasmic filaments, best visualized in cross-section, corresponded to intermediate-sized filaments with a diameter of 120–170 Å (mt in Fig. 2C). In addition, we occasionally detected cross-sectioned microtubules (270 Å) (mt in Fig. 2E). Both intermediate filaments and microtubules were commonly surrounded by an electron-lucid region of 100–150 Å. At the cell periphery, membrane-bound dense plaques were observed with an approximate thickness of 400 Å (Fig. IA, Fig. 2E). These structures were associated with the cytoplasmic filaments, as well as with a fine layer of membrane-bound electron-dense extracellular material (double arrowhead in Fig. 2E). Numerous sub-sarcolemmal vesicles (çaveolae) were found between the dense plaques (matched small arrowheads in Fig. 5B/B’). Membrane-bound structures were well preserved in the VFS, including the nucleus (Fig. 1A), the plasmalemma (Fig. 2B) and mitochondria (Fig. 1A, Fig. 7). In the extracellular space, connective tissue elements (most conspicuously, collagen) were detected, often running into the dense plaque region along the plasma membrane (Fig. 2A,C in association site, marked with an arrow). The collagen periodicity was clearly apparent in these specimens (Fig. 2D), measuring approximately 680 Å, and was distinctly different with respect to both the periodic pattern and the repeating distance from that of dehydrated, uranyl acetate-stained, methyl-cellulose-embedded samples (Fig. 2D’), which is around 640 Å. Note that X-ray diffraction of hydrated rat-tail tendon collagen fibrils yields a repeating distance of 670 Å, whereas dehydrated samples measured by ‘conventional’ TEM have values in the range of 550–650 (Chapman et al. 1990).

Epon sections, shown here for comparison, revealed most of the structures found in VFS, although the exact dimensions were often somewhat different (Fig. 3A and B). It is noteworthy that visualization of cytoplasmic structures in Epon sections requires staining with uranyl acetate and/or lead salts, while the VFS were unstained. Moreover, examination of sections prepared in a manner similar to that for VFS, but embedded in methyl-cellulose, stained and dried (the classical “Tokuyasu’ technique), showed that many of the fine cytoplasmic details, especially fine filaments, were not comparably visible in the latter specimens (Fig. 3C).

We also examined the extent of fixation needed to preserve adequately the cellular sub-structure of VFS. In a preliminary set of experiments it was established that some fixation was, indeed, essential for the preservation of cytoplasmic material, and was particularly important for the stabilization of sections during the rinsing steps on the grid. Fig. 4 shows typical examples of the ultrastructure of gizzard smooth muscle subjected to different fixation procedures: (a) moderate fixation with 0.1% glutaraldehyde and 3% formaldehyde (Fig. 4A); and (b) light formaldehyde (3%) fixation-(Fig. 4B). Both specimens yielded large, rather uniformly preserved sections similar to the well-fixed samples. Nevertheless, significant differences were noted in the fine preservation of cytoplasmic structures, especially between the glutaraldehyde-fixed specimens (Figs 1, 2 and 4B) and the formaldehyde-(1ightly) fixed specimen (Fig. 4B). This was manifested by a partial loss of cytoplasmic material and apparent distortion of cytoplasmic filaments in the latter specimen.

Since the VFS remain fully hydrated throughout their processing on the grid, including their embedding in vitrified ice (see Discussion), it was of interest to determine whether the 3-dimensional features of the cells were retained. For this purpose stereo-pairs of photographs (taken at ±6°) were obtained (Fig. 5). They show that the 3-dimensional structure and filament disposition are well preserved throughout the thickness of the section. This is apparent from both low-power and higher-power photomicrographs (Fig. 5A/A’ and B/B’, respectively). Examination of the stereo-pairs also improves the apparent resolution of fine details such as the sub-structure of filaments, the shape of caveolae, organization of dense bodies and plaques, etc. The stereoscopic examination also reveals that the distal surfaces of the sections were covered with a rather thick layer of amorphous ice. The absence of crystalline ice was confirmed by electron diffraction analysis (not shown).

One of the major objectives in developing the VFS method is to use these specimens for high-resolution immunolocalization of intra- and extracellular molecules. To determine whether or not this is possible, we chose to localize the protein a-actinin in gizzard sections using indirect immunogold labeling. The immunolabeling step was performed following the rinsing of the section and prior to its embedding in ice. Fig. 6A shows specific gold labeling along the cytoplasmic dense bodies similar to that observed in previous studies (Schollmeyer and Furcht, 1976; Geiger et al. 1981; Volberg et al. 1986). The extent of labeling is comparable to that obtained with methylcellulose-embedded sections (Fig. 6B).

To evaluate the potential of this new method for more general applications, we examined two additional tissues: namely, chicken heart muscle (Fig. 7A) and rat kidney (Fig. 7B-E). Both tissues have distinct structures and their consistencies are quite different from that of smooth muscle. The former tissue is dominated by well-organized sarcomeric structures, displaying thin and thick filaments, Z-disks etc. The kidney tissue consists mainly of polar epithelial structures, including podocyte processes (shown in stereo in Fig. 7B/B’), microvilli (Fig. 7C) and intercellular junctional complexes (Fig. 7E). VFS retained cellular ultrastructure and often revealed details that could barely be detected using methyl-cellulose-embedded, or even Epon sections, following heavy-metal staining.

Electron microscopy combined with immunolabeling has provided invaluable information on the molecular topology of subcellular and extracellular structures. This information has some significant intrinsic limitations: (a) the tissues are almost inevitably dehydrated either before or after sectioning. This undoubtedly results in significant structural alteration, particularly at the molecular level, (b) Visualization of structures is primarily based on staining or shadowing with heavy metals, rather than on direct examination of the structure itself. This can result in significant biases depending upon the affinity of the various sub-cellular structures for the metal, (c) Optimal conditions for immunolabeling are often in conflict with the most suitable procedures for structural preservation. Efficient labeling depends upon the use of relatively light fixation, allowing for both improved antibody and gold complex penetration, as well as preservation of antigenicity. Such fixation conditions, however, may result in poor structural integrity (see, for discussion of this point: Griffiths, 1991).

The development of the new approach reported here using vitrified frozen sections (VFS), with or without immunolabeling, was attempted in order to overcome at least some of these problems. Perhaps the most important potential advantage of VFS is that the specimens remain hydrated during both sample preparation and examination. This was verified by the controlled removal of liquid from the grid prior to vitrification and from 3-D examination of the specimen. Furthermore, excessive drying of sections invariably resulted in their deterioration, which could easily be recognized. Moreover, it was found that the thickness of vitrified ice layer did not change in the microscope column, even following prolonged examination. To improve structural preservation, it was important to prevent ice crystals from damaging the tissue during the initial freezing step by introducing sucrose as a cryoprotectant, and in the second freezing step by cooling the tissue section so rapidly that the water molecules cannot crystallize, but remain in an amorphous state.

A similar approach to this problem has previously been reported (Chang et al. 1981; McDowall et al. 1983, 1989), in which unfixed tissue sections were prepared from the vitrified surfaces of frozen tissues. The sections were examined under frozen conditions without allowing them to thaw, disclosing such structural details as the Golgi complex, nuclear pores, endosomes, cytoskeletal filaments etc. This approach, when compared with our VFS method, has the intrinsic advantage of not using a chemical fixative, yet it is noteworthy that this approach is incompatible with post-sectioning immunolabeling of the tissues; a major consideration in the present study. Fixation of the tissue is unavoidable for the preparation of VFS according to the procedure described here. Unfixed frozen sections readily deteriorate during sample preparation, whereas sections from fixed tissues, especially glutaraldehyde-fixed tissues, are relatively stable. Even very lightly fixed tissues (0.1% glutaraldehyde or 3% formaldehyde) show a well-preserved overall structure, with only limited distortion of cytoplasmic filaments.

VFS are compatible with immunolabeling, as shown here by the localization of α-actinin in the cytoplasmic dense bodies, in accord with previous studies (Schollmeyer and Furcht, 1976; Geiger et al. 1980, 1981). Immunolabeled VFS also show extensive labeling comparable to that obtained with the Tokuyasu method, disclosing only low background staining. One important problem associated with the VFS method as well as essentially all postsectioning labeling procedures, is that the immuno-gold particles are largely confined to the tissue surface. This is true for both VFS and methyl-cellulose-embedded sections, as verified by stereoscopic examination (not shown). We are currently examining the possibility of enhancing the electron density of the specific antibodies or their Fab fragments to allow for their direct visualization, and in so doing significantly increase the resolution of the technique.

The contrast observed in the VFS is obtained by photographing the image under slightly underfocused conditions. Such contrast is not obtained in unstained sections embedded in polymeric materials such as Epon or methyl-cellulose. Interestingly, the observed contrast in VFS is reminiscent of positively stained specimens. For example, cytoplasmic fibers, membrane profiles and chromatin appear as electron-dense structures. There are, however, some distinct differences. Basement membranes, which, in stained sections, appear as a dense layer, were not readily apparent in VFS (Fig. 7B). In contrast, the nucleus (Figs 1A and 7D) clearly revealed a dense layer corresponding to the fibrous lamina; a structure not easily visualized in stained sections (Fawcett, 1966; Franke, 1974). The physical basis for the generation of contrast in VFS is not entirely clear, although it seems to be, at least in part, proportional to the actual density of mass and possibly to the differential distribution of bound ions. A good example of that is the revelation of the banding structure of collagen without stain. Obviously the fact that soluble unfixed macromolecules are partially extracted from the sections during the rinsing steps could contribute to the relative apparent density and contrast of the insoluble structure. This cannot, however, be a major factor as both lightly and heavily fixed tissues show similar degrees of contrast.

In this study we examined only three tissues: namely, smooth muscle, heart and kidney. They have very different cytoplasmic structures and consistencies, yet all show a high degree of structural preservation in VFS. This suggests that the method may be readily applicable to a broad spectrum of tissues. In view of the novelty of the method, it may be important to review some of the difficulties we encountered during the preparation and examination of VFS. (a) It is essential to prepare ultrathin frozen sections with approximate thicknesses of 700 Å. Thicker sections tended to display poor morphology, possibly due to non-uniform freezing, as well as higher susceptibility to radiation damage, (b) Blotting of the excess water on the grid prior to plunging is difficult to control and reproduce. Experience and sometimes patience help. Achieving an optimally thick vitreous ice film on grids containing tissue sections was much easier than for samples containing particles in suspension, probably because of the flat surface of the sections themselves. A relatively thick vitreous ice layer above the section (1000–1500 Å) is important for reducing radiation damage. This observation is consistent with previous studies discussed by Talmon (1987). (c) We have examined tissue sections placed directly on high-mesh uncoated grids, as well as on grids containing carbon films. The former were found to be superior, (d) Radiation damage is a major problem and low-dose working conditions should be used. To date, this has restricted our ability to collect multiple photographs of the same area, and in some cases, even for obtaining two undamaged photographs for stereoscopic analysis at low magnifications (1ess than ×20 000). Obtaining photographs at higher magnifications may well require the extensive use of highresolution image intensifiers and computerized imageprocessing systems.

The use of vitrified frozen sections for structural and immunolabeling studies of tissues using the electron microscope potentially reduces the extent to which preparation artifacts can alter the in vivo structure. This consideration should more than offset the difficulties involved in preparing and examining the specimens. Furthermore, there are a variety of ways of extending the potential of the technique, such as the development of small immunoprobes making possible 3-dimensional labelling of the tissue, preparation of 3-dimensional images from multiple photographs and the improvement of the resolution of the method by better sample treatment and electronic image processing.

We thunk Professor Y. Talmon, Technion, Haifa, for introducing us to cryoelectron microscopy and for illuminating discussions. This study was supported in part by the Helen and Milton Kimmelman Center for Biomolecular Structure and Assembly, by the Revson Foundation, administered by the Israel Academy of Sciences and Humanities and by a US-PSH grant no. DE09654 to S.W. B. Geiger is the E. Neter Professor for Cell and Tumor Biology.