The origins and evolution of freeze-etch electron microscopy

Abstract

The introduction of the Balzers freeze-fracture machine by Moor in 1961 had a much greater impact on the advancement of electron microscopy than he could have imagined. Devised originally to circumvent the dangers of classical thin-section techniques, as well as to provide unique en face views of cell membranes, freeze-fracturing proved to be crucial for developing modern concepts of how biological membranes are organized and proved that membranes are bilayers of lipids within which proteins float and self-assemble. Later, when freeze-fracturing was combined with methods for freezing cells that avoided the fixation and cryoprotection steps that Moor still had to use to prepare the samples for his original invention, it became a means for capturing membrane dynamics on the millisecond time-scale, thus allowing a deeper understanding of the functions of biological membranes in living cells as well as their static ultrastructure. Finally, the realization that unfixed, non-cryoprotected samples could be deeply vacuum-etched or even freeze-dried after freeze-fracturing opened up a whole new way to image all the other molecular components of cells besides their membranes and also provided a powerful means to image the interactions of all the cytoplasmic components with the various membranes of the cell. The purpose of this review is to outline the history of these technical developments, to describe how they are being used in electron microscopy today and to suggest how they can be improved in order to further their utility for biological electron microscopy in the future.

Fig. 1.

Classical freeze-fracturing of the frog neuromuscular junction, circa 1974 [54], showing the elongated presynaptic nerve terminal (pale green) punctuated every micron or so by transverse bands (orange) which represent the presynaptic ‘active zones’ where synaptic vesicles undergo exocytosis (captured as a great abundance of ‘pores’ adjacent to IMPs that likely represent voltage-activated calcium channels). The nerve terminal is surrounded by Schwann cells (purple) and is separated by the synaptic cleft (blue) from the surface of the muscle fiber (dark green). This sample was chemically fixed during nerve stimulation, impregnated with antifreeze (25% glycerol), and slowly frozen by ‘quenching’ in Freon 22 before freeze-fracturing and unidirectional ‘shadow casting’ with platinum. This was the standard approach used before the ‘quick-freeze’ technique was invented. A few years after having learning how to generate such images, we developed a liquid helium-cooled ‘Cryopress’ slam-freezer [3], which made it possible to obtain comparable freeze-fracture images of nerves that received only one single stimulus and were then ‘quick-frozen’ immediately thereafter. This permitted us to capture individual synaptic vesicle openings at the presynaptic “active zones” more definitively, and then to correlate their exact abundance with the number of transmitter quanta that had been discharged, thus proving Katz's vesicle hypothesis [52].

Fig. 2.

Quick-freeze, deep-etch, rotary replication (QF-DE-RR) view of the inside surface of a cell infected with the SSPE version of measles virus, circa 1984 (unpublished), illustrating the helical viral nucleocapsid (yellow) strongly and specifically attached to an orthogonal scaffold of M-protein (or matrix-protein), which this virus assembles onto the plasmalemma in order to bud new viruses. Because this was a budding mutant of the virus, derived from an SSPE patient (subacute sclerosing panencephalitis), the bud site is usually large and flat and is less closely associated with surrounding subcortical actin filaments (green) than usual. Such images originally explained the mechanism of viral latency in SSPE [163,164]. A ∼0.1 μm diameter clathrin-coated pit in the lower left of the field (purple) is shown for a sense of scale.

Fig. 3.

View of the inside surface of the plasmalemma of a HELA cell, prepared as in Fig. 2 by ‘unroofing’ a monolayer culture grown on a glass coverslip before quick-freezing and freeze-drying it according to our standard procedures [165]. Here, we focus on the variety of clathrin lattices that are seen on all cells and illustrate the stages in their evolution from totally flat lattices to fully curved ones, ready to pinch off from the cell surface during endocytosis. Such images were the first to illustrate that actin filaments (purple) become involved in the later stages of the budding process, a fact now well established [166,167], and remain behind as circular ‘scars’ after the coated vesicles have left the surface (arrow). The opportunity to view such expanses of the plasma membrane, at such high resolution, was a lucky outcome of being able to freeze at speeds high enough to avoid ice-crystal formation and then rotary replicate with platinum for TEM without melting these quick-frozen samples.

Fig. 4.

First-time published image of a brand new variation of the QF-DE-RR procedure for imaging the interiors of unroofed cells, which involves doing everything on a pre-formed carbon substrate rather than on glass, so that the platinum replica does not ever need to be separated from its substrate before TEM viewing. This eliminates the huge problem of replica breakage that has always plagued the field – the breakage occurring when the replica is separated from the underlying cells and substrate, in order to be viewed in the TEM. Also, because freeze-fracture replicas were always so fragile, they formerly had to be supported by carbon films deposited on top of them, which tended to confuse or obscure the final images. Now, with the carbon substrate underneath the cells, this ‘backing’ or support film can be eliminated, and vastly larger and more stable replicas can be obtained, even so. This permits approaches possible never before, as in this experiment, where HeLa cells that were in suspension culture, and thus were highly ‘blebby’, were exposed to glow-discharged carbon and allowed to attach to it for only 1 min before they were ‘unroofed’, thereby yielding images of the inside surface of such highly dynamic cell processes as blebs (the obvious, micron-sized circular domains in the cell) where actin polymerization is known to be going on actively [168,169] but could never before be visualized at this resolution. Close inspection of the circular domains illustrates that actin polymerization in situ is via the formation of y-shaped branches from the existing actin filaments, exactly as was predicted from the ‘dendritic branching’ model we derived earlier from our in vitro imaging of actin and Arp2/3 [170].

Fig. 5.

Illustration of the three basic approaches to in vitro imaging of macromolecules permitted by QF-DE-RR, using molluscan hemocyanin as an example. This barrel-shaped didecamer (molecular weight ∼4 mDa) appears in replicas as having a diameter of ∼28 nm and normally a length of ∼30 nm (except in certain snails, as in the central panel, where lengths are multiples of the 30 nm didecamer). Classical ‘shadow casting’ of platinum from a fixed angle (11° above the horizontal in the upper panels) yields dramatic shadows whose lengths are direct measures of the hemocyanin's elevation above the mica. The newer approach of rotary replication with platinum (center and lower panels) yields better overall delineation of surface architecture, in this case showing the basic helical construction of hemocyanin particularly clearly in molecules loosely adsorbed to mica flakes (center panels, derived from our original 1983 introduction of this procedure) [171]. More recently, we have learned that the carbon substrate approach described in Fig. 4 also works extremely well for molecules as well as for cells and cell derivatives, allowing controlled adsorption of nearly anything, almost as on a Biacore SPR chip. Shown on the lower panel is how uniformly such spreads of molecules can be made on carbon, which makes them particularly suitable for contemporary single-particle analyses.

Fig. 6.

Freeze substitution and plastic thin-sectioning samples frozen with our ‘Cryopress’ illustrates how good freezing can be (but only in the uppermost layer of cells that come into direct contact with the liquid helium-cooled copper block). The upper two fields in the figure show sections, about a half-micron apart, of a cultured cell infected with vaccinia virus that was serially sectioned, and the left section is enlarged below to illustrate the characteristic poxvirus factories of cytoplasmic DNA (highlighted in yellow) surrounded by dark spherical virions in various stages of maturation and compaction. Barely visible at this low magnification are numerous granular deposits of proteins and lipids that are depots for supporting the growing, initially crescent-shaped membranous envelopes of the viruses (highlighted in green). Without nearly perfect freezing, none of these cytoplasmic inclusions and differentiations could be properly distinguished nor could viral morphogenesis be properly determined.

Fig. 7.

Platinum replica of freeze-fractures through cells infected with vaccinia virus and then quick-frozen, exactly as was done in Fig. 6, permit imaging of individual viruses in their cytoplasmic context, at much higher magnification. This illustrates that the developing poxvirus crescents, as well as the complete spheres, have a distinctive geodetic [172] or ‘honeycomb’ scaffold on their external surface, not unlike the lattices on clathrin-coated vesicles (Fig. 3) but more than twice as fine (only 7 mm vertex to vertex). This lattice is highlighted in yellow to illustrate how freeze-fracturing breaks through it and travels along the bumpy membrane of the developing virion (the dome-shaped examples) or ‘cross-fractures’ the virions to expose their inner core of protein and DNA (the flatter examples). One of the granular ‘supply depots’ of membrane and protein for making the envelopes of these virions, which are barely visible as the green-highlighted entities in Fig. 6, is seen distinctly in the lower panel (it is highlighted in green). It is composed of a scrambled collection of 7 mm granules that we have shown in earlier work [147,148] to represent trimers of a 63 kDa self-assembling coat protein (made by the ORF ‘D-13 L’ in vaccinia nomenclature). Comparison of the granular texture of this viral inclusion with the rest of the surrounding cytoplasm illustrates why QF-DE-RR is a particularly useful technique for discerning and mapping all kinds of cytoplasmic inclusions in health and disease: (i) the imaging is unperturbed by any chemical cross-linking or dehydration; (ii) it does not require any differential staining of the substances in the inclusion and (iii) it relies entirely on the natural surface texture of the inclusion, which is readily apparent in such 3D views.

Fig. 8.

Thawed Tokuyasu cryosections of vaccinia-infected cells decorated with anti-D-13, the honeycomb scaffold protein [148], and imaged using the standard method (upper two panels) or by re-freezing the thawed and decorated cryosection and then freeze-drying and platinum replicating it (lower three panels). In keeping with the fact that nearly anything can be put on a substrate, and then quick-frozen and freeze-dried for platinum replication, a particularly useful application of this is to image Tokuyasu-type immunogold-decorated cryosections. The Tokuyasu technique is widely used and highly successful [–146], but has always suffered from the exceedingly low e-contrast of the sections, making it particularly difficult to discern cellular membranes in any detail. This can be overcome by making ‘deep-etch’ replicas of these sections after they have been antibody- and gold-decorated. This provides remarkably clear 3D information about the location of gold relative to cellular structures. (To make the comparison more obvious, the upper left panel is contrast-reversed in the upper right panel, to make the gold dots look white, as they do in our platinum replicas). (Actually, all the white dots have been artificially highlighted yellow here, for ease of interpretation.) Readily apparent in the 3D views of the platinum replicas, below is that the ‘honeycomb’ lattice labels only on its uppermost cut edges, where apparently the D13 epitope must be exposed. This immediately explains why earlier Tokuyasu cryosection EM with anti-D-13 antibodies incorrectly assigned this protein to an internal location in the virion [173]. (Note how the left-most virion in the upper panel would seem to suggest this.) Namely, all the virions that are just barely cut open (e.g. are ‘scalped’), or are otherwise not cut exactly through their equators during the cryosectioning, stain only on their uppermost cut edges, which are invariably inside the widest circumference of the virion. The lack of this 3D information, plus the lack of clear-cut imaging of the surfaces of membranes in Tokuyasu cryosections, can lead to such misinterpretations.