Developed over a period of years, the standard procedure is to chemically fix small tissue pieces with aldehydes, glycerinate for cryoprotection, freeze in a metal specimen carrier, and mount on a brass table that is inserted into the freeze fracture unit. An excellent overview of standard tissue preparation techniques for freeze fracture is available [22]. Because freezing of tissue leads to growth of ice crystals that distort cellular structure at the electron microscopic level, tissues must be cryoprotected by infiltration with high concentrations of glycerol (25–35 % v/v), with the result that many more ice crystals are nucleated but their growth is slow thereby resulting in an extremely fine granularity that does not disrupt structure. It was quickly learned that glycerination without prior chemical fixation of the tissue results in artifactual membrane blebbing and IMP redistribution [23]. Therefore, tissue is typically fixed with glutaraldehyde before glycerination and freezing. Use of osmium tetroxide as a fixative, while common in material embedded in epoxy resin and sectioned, is not compatible with the freeze fracture process because it prevents membrane splitting and formation of panoramic membrane faces. Small tissue cubes or pellets of isolated cells, fixed and glycerinated, are placed on gold–nickel alloy carriers shaped like top hats, and then frozen in a liquid cryogen. Liquid nitrogen is unsuitable as a cryogen because its ability to absorb heat is small and its boiling point is too low. As a result, it boils and forms an insulating layer of gas around the specimen resulting in very slow freezing and growth of relatively large ice crystals even in glycerinated specimens. For many years dichlorodifluoro-methane (Freon 12) was the cryogen of choice until its discovery as an ozone layer depleting agent and removal from the market. Although, Freon replacements such as R-22 and R134a are in use, R-22 is also ozone depleting and due to be phased out by 2020 and R134a is thought to contribute to greenhouse warming; both have gone largely untested in their possible use as a cryogen for freeze fracture work. Thus, the two most common cryogens currently in use are small hydrocarbons such as liquid propane or a nitrogen slush. Traditional freezing of specimens by plunging into liquid nitrogen-cooled cryogen is done by handheld tweezers. Alternatively, machines are now available such as the Leica EM GP, the Leica EM CPC, the EMS-002 or the Gatan Cryoplunge 3 that provide for automated plunging of grids and freeze fracture specimens into liquid propane. Alternatively, nitrogen slush can be generated by repeated evacuation of liquid nitrogen with a either a common mechanical pump or a dedicated instrument such as the Vit-Master in which boiling of the liquid extracts heat to produce a semisolid slush.

Fracturing of the frozen specimen is typically carried out in a vacuum of 2 × 10⁻⁶ mbar with the specimen temperature being held at −130 °C. Immediate shadowing of the freshly cleaved specimen from an angle of 45° with platinum (“unidirectional shadowing”) followed by carbon to strengthen the replica, results in clean membrane faces framed by smooth ice surfaces that have not been etched as seen in Fig. 4a. In this figure, note that some membrane surfaces can be difficult to distinguish from the surrounding cytoplasmic ice. In contrast, if the specimen is held at −110 °C and allowed to “etch” for 15–30 s between fracturing and shadowing, a modest amount of ice is sublimated leaving a rough patina on the non-membrane surfaces. This allows one to easily distinguish smooth membrane surfaces from roughened ice surfaces. As shown in Fig. 4b, in which a sea urchin egg cortical granule (cg) is undergoing fusion with the plasma membrane (pm), this rough patina identifies the aqueous pores formed during exocytosis. In Fig. 4c, these roughened aqueous pores can be distinguished from pores covered by a membrane diaphragm (arrows) which represent an interesting intermediate structure in the pore forming process [24]. The usefulness of etching standard specimens is limited, however, because deep etching of specimens prepared by the standard protocol is not possible due to the low vapor pressure of the glycerol cryoprotectant.

In order to avoid ice crystal damage of electron microscopic specimens that have not been cryoprotected, freezing rates of at least −50,000 °C per second must be used [25]. These rates can be achieved by several different means. Very thin specimens such as cells cultured on dialysis membrane can be frozen by manually plunging into liquid propane (see Fig. 5a) or by using an automated machine as described in the previous section. Alternatively, cell suspensions or tissues no greater than 50 μm thick can be housed between copper planchets that are sprayed with liquid propane in a propane jet freezer such as the RMC MF-7200 (Fig. 5b) designed by Staehelin and Gilkey [26] or the Balzers JFD30. In contrast, relatively thick specimens can be frozen by being pressed against a liquid-helium or liquid nitrogen-cooled copper block, a process known as “slam freezing” or “cold metal block freezing.” This process can be carried out by any one of several machines built for this purpose: the “CryoPress” designed by John Heuser [27], the Life-Cell metal block freezer (now out of production) and the Leica EM MM80 E that is still produced (see Fig. 5c). Finally, one can utilize high pressure freezing in which specimens up to 1 mm³ are enclosed in a planchet that is subjected to 2,100 bar pressure while being frozen with liquid nitrogen [28, 29]. This process can be carried out by a number of commercial units such as the Balzers/Baltec HPM 010, now the Leica EM HPM100 as described in the companion chapter on high pressure freezing.

Specimens rapidly frozen without chemical fixation or glycerination, when fractured and replicated unidirectionally with platinum, exhibit smooth membranes that are free of membrane blebs and wrinkles that are generally seen in glutaraldehyde-fixed specimens and considered artifacts [30, 31]. In addition, since specimens are immobilized within milliseconds, ultrarapidly frozen specimens are capable of capturing extremely fast biological processes such as synaptic vesicle exocytosis and membrane fusion [24, 32–34].

Since rapidly frozen specimens are not glycerinated, extensive sublimation of ice can be achieved thereby revealing micron-deep layers of biological structure that are particularly useful in visualizing macromolecular assemblies in situ—that is, within the context of normal tissue structure. Although applicable to a wide variety of structural features, the organization of the cytoskeleton and the extracellular matrix are particularly striking examples of how this technique has contributed to our understanding of functional macromolecular assemblies [35–39]. The depth of etching in a specimen is related to the temperature and duration of etching. Quick-frozen specimens fractured at −110 °C and etched briefly will exhibit a roughened, somewhat granular ice surface in the cytoplasm such as seen in the cortex of a mast cell between the plasma membrane and the underlying secretory granule (see Fig. 6a). In contrast, etching for 5 min at −90 °C may reveal an extensive layer of intertwined microfilaments whose relative three-dimensional architecture demonstrates functional relationships (see Fig. 6b). Additional etching (up to 10 min at −85 °C), can remove up to 1 μm of ice from all aqueous surfaces, revealing embedded biological structures. In Fig. 6c, for example, deep etching has revealed the intricate network of extracellular filaments that make up the vitelline layer of the sea urchin egg, a layer that is instrumental in binding sperm just before they enter and fertilize the egg. If etching had not been carried out, all the structures above the fracture line (arrow) would have been hidden by extracellular ice. Deep etching is equally effective at revealing cytoskeletal networks in the cytoplasm. Figure 6d illustrates the apical brush border of the intestinal epithelium in which myosin-head decorated microfilaments forming the core of microvilli are anchored into a web of intermediate filaments below. Deep etching is usually combined with rotary shadowing in which the specimen is rotated by an electric motor at about 3 Hz while platinum is being applied at a relatively low angle, about 20°–30°, so as to get platinum into every nook and cranny of its complex topology.

Cell fragments adhering to a glass or mica surface can be freeze dried by extensive sublimation of ice carried out for long periods of time (10 min), at relatively high temperatures (−85 to −90 °C) in a freeze fracture unit at high vacuum. Cell fragments are typically produced by cell adhesion to a surface then the cell sheared open by a stream of buffer under pressure. Cell adhesion may be natural as in the case of epithelial or migrating cells, or may be induced by plating cells on a polylysine coated glass, plastic or mica surface [40, 41]. This approach is particularly useful in visualizing events at the cell cortex such as exocytosis, endocytosis and cytoskeletal reorganization [42–44]. A typical preparation—the true inner surface of a neutrophil plasma membrane—is shown in Fig. 7a. Seen attached to the plasma membrane are a number of microfilament clusters as well as clathrin baskets marking endocytic vesicles in various stages of budding.

Macromolecules, generally greater than 50 kDa, can be purified and adhered to the surface of mica flakes and then quick frozen and either deeply etched or freeze dried to reveal their architecture. Fixation or cryoprotection is not required providing the proteins adsorbed to the mica flakes are rapidly frozen usually by a cold metal block procedure [45, 46]. Optimal images are obtained by rotary shadowing with platinum at 25° although lower angles are sometimes used. This procedure also works well for macromolecular assemblies such as vesicle coat proteins and viruses [47, 48]. Classic examples of quick frozen and freeze dried proteins and molecular assemblies are shown in Fig. 7a. Low magnification images reveal the mica flake substrate (see Fig. 7b) with adherent macromolecules—in this case polymerized Rec A, a DNA binding protein isolated from Escherichia coli. At high magnifications, individual Rec A polymers can be seen to be a linear chain of monomers (see Fig. 7c). Using horseshoe crab hemocyanin as a test specimen, Heuser demonstrated this protein to be a hollow cylinder when one compares proteins adsorbed lengthwise to the mica (see Fig. 7d) with proteins that stand “stump-like” on the mica surface after being cross-fractured at a variety of levels (see Fig. 7e). This method was developed and perfected by Heuser whose review articles [45, 46] discuss the details of mica preparation for optimal protein absorption, a step critical to its success.

Traditional cytochemistry involves the localization of enzymatic activities of tissues in situ by supplying a substrate that can be turned into a product capable of being imaged in the electron microscope. Classic examples include the localization of acid phosphatases to lysosomes by their production of phosphate which is then precipitated with lead or cerium and converted to a sulfide which is imaged as an electron dense deposit. Similar cytochemical procedures have been used successfully in tissue that is subsequently processed by freeze fracture and replicated [49]. Of more widespread use are reagents that complex with the tissue constituent of choice to produce large particles or topological features in the replica itself. A common example of this is the use of filipin or tomatine, sterols that bind to cholesterol in membranes to form large complexes easily seen in a freeze fracture replica (see Fig. 8a). This technique, pioneered by Friend and colleagues [50, 51], has been used extensively to demonstrate cholesterol-containing domains in membranes [52].

Of greater impact has been the ability to use immunocytochemistry in conjunction with freeze fracture, first achieved by Pedro Pinto da Silva in a process called “fracture-label” [53, 54]. In this method, tissues were fixed, rapidly frozen and fractured and only then thawed in aldehyde fixative, washed and subjected to colloidal gold immunolabeling. Finally the labeled fracture surfaces could be visualized either by embedding in epoxy resin and sectioning or by critical point drying followed by platinum–carbon shadowing either in a freeze fracture unit or in a standard vacuum evaporator [53, 54]. As shown in Fig. 8b, the colloidal gold markers are easily seen on fracture surfaces in sectioned material while in platinum replicas, the markers are visible but the preservation of the fracture faces is compromised by both aldehyde fixation and dehydration during critical point drying (see Fig. 8c). Subsequently, Pinto da Silva modified this procedure to one of “label-fracture” in which the tissue, usually a cell suspension, was carried through immunocytochemical labeling first and then frozen, fractured and shadowed with platinum-carbon in a freeze fracture unit [55, 56]. Next, the specimen was washed, mounted on a grid and observed by TEM directly without having cleaned away the tissue. Although views of many areas of the replicated surface were obliterated by the presence of tissue, regions containing E faces with attached label were visible since the biological material present was limited to that of a single, transparent monolayer. For example, in studies of antigen-induced capping of lymphocytes, aggregates of colloidal gold-labeled HLA I proteins were easily seen in sections (see Fig. 8d). Specimens processed in parallel using label-fracture methods revealed remarkable replicas of “E” fracture faces with colloidal gold labeling that defined the extent of the cap in two dimensions (see Fig. 8e).

The label-fracture approach led naturally to the “fracture-flip” method also developed by Pinto da Silva [57, 58] and the “SDS freeze-fracture replica immuno-labeling” (FRIL) method developed by Fujimoto [59, 60]. The FRIL method, in common use today, is an extension of the label-fracture method. The only additional step is that the labeled, fractured and replicated tissue is cleaned with sodium dodecyl sulfate and water before viewing, thereby solubilizing most of the tissue and allowing clear visualization of both E and P fracture faces. As shown in Fig. 8f, the colloidal gold markers identify connexins 26 and 32 within the gap junctions of mouse liver cells [60]. The combination of well preserved structure and clean fracture faces in these replicas make this the current method of choice for freeze fracture immunocytochemistry.

Fracture faces of tissues are three dimensional with topology in the optical axis that is easily perceived by shadowing of this surface at an angle. The perception of depth can be direct by using stereology in which two images of the replica taken at different angles are used to form a stereo pair, each image projected to a different eye by use of appropriate optics. This requires a goniometer stage to tilt the specimen typically at ±6° to 12° from the optical axis as images are recorded. Detailed methods for doing this have been described by Heuser in several articles [61, 62]. As an example, the stereo pair of images in Fig. 9 reveals the topology of the vitelline layer on the sea urchin egg surface in a specimen that was quick-frozen, deep-etched and rotary-shadowed [63]. The depth afforded in this stereo pair reveals numerous fibrous polymers connecting the tips of microvilli as well as a sheet of networked fibers draped over the egg surface and connected by short anchoring posts to the plasma membrane. Below the fracture line (arrow), the etched “P” face of the plasma membrane reveals the stumps of several microvilli that have been cleaved.

The depth inherent in a replica can be quantitatively examined using tomography methods in which a film-supported replica is rotated through a large angular sweep with images being acquired at 1° intervals. As described by Morone et al. [64, 65], one can use appropriate software and image information to reconstruct a three-dimensional model of the replica with calibrated dimensions. Although most studies require only the qualitative depth perception that stereology provides, any quantitative studies of depth might best use the tomographic approach.