By continuing to browse this site you agree to us using cookies as described in About Cookies
Wiley Online Library is migrating to a new platform powered by Atypon, the leading provider of scholarly publishing platforms. The new Wiley Online Library will be migrated over the weekend of March 17 & 18. You should not experience any issues or loss of access during this time. For more information, please visit our migration page: http://www.wileyactual.com/WOLMigration/
Three cell-permeant compounds, cytochalasin D, latrunculin A and jasplakinolide, which perturb intracellular actin dynamics by distinct mechanisms, were used to probe the role of filamentous actin and actin assembly in clathrin-mediated endocytosis in mammalian cells. These compounds had variable effects on receptor-mediated endocytosis of transferrin that depended on both the cell line and the experimental protocol employed. Endocytosis in A431 cells assayed in suspension was inhibited by latrunculin A and jasplakinolide, but resistant to cytochalasin D, whereas neither compound inhibited endocytosis in adherent A431 cells. In contrast, endocytosis in adherent CHO cells was more sensitive to disruption of the actin cytoskeleton than endocytosis in CHO cells grown or assayed in suspension. Endocytosis in other cell types, including nonadherent K562 human erythroleukemic cells or adherent Cos-7 cells was unaffected by disruption of the actin cytoskeleton. While it remains possible that actin filaments can play an accessory role in receptor-mediated endocytosis, these discordant results indicate that actin assembly does not play an obligatory role in endocytic coated vesicle formation in cultured mammalian cells.
Receptor-mediated endocytosis in mammalian cells requires clathrin, adaptors and dynamin[1–3]. In contrast, endocytic vesicle formation in the yeast, Saccharomyces cerevisae, is only partially inhibited by mutations in clathrin[4,5] and occurs independently of the yeast adaptor- or dynamin-related proteins. Instead, both receptor-mediated and fluid-phase endocytosis in yeast appear to be functionally coupled to organization and assembly of the actin cytoskeleton[8–10]. Mutations in yeast actin and in several actin binding proteins inhibit endocytosis[11–13]. Correspondingly, many of the so-called ‘end’ mutations or ‘dim’ mutations[15,16], isolated based on their effects on endocytosis, also disrupt the organization of the yeast cortical actin cytoskeleton. Thus, it remains possible that distinct mechanisms mediate endocytosis in yeast as compared to mammalian cells.
Nonetheless, a functional link between clathrin-mediated endocytosis and the actin cytoskeleton in mammalian cells is being forged by recent findings that several components of the mammalian endocytic machinery interact, either directly or indirectly, with the actin cytoskeleton. Overexpression of a dominant-negative mutant of dynamin-1 in HeLa cells leads to redistribution of actin stress fibers to the cell cortex. These effects may be mediated by dynamin binding partners, because the neuronal isoform of dynamin interacts both in vivo and in vitro with profilin and with syndapin, an SH3 domain containing protein that binds to the actin regulatory protein, N-WASp (neuronal Wiskott-Aldrich syndrome protein). Two other proteins involved in clathrin-mediated endocytosis in mammalian cells, eps15 and amphiphysin, have been reported to also interact with the actin cytoskeleton[21,22]. Endocytosis is inhibited by constitutively active mutants of Rho and Rac, although these effects appeared to be independent of actin assembly.
Whether either filamentous actin or actin assembly play a direct role in clathrin-mediated endocytosis in mammalian cells has yet to be established. In contrast to the well-defined and rapid onset temperature-sensitive genetic lesions used to study actin and endocytosis in yeast, studies in mammalian cells have relied on three cell-permeant toxins which perturb actin assembly and disassembly by distinct mechanisms, namely: 1) cytochalasin D (cytoD) which caps actin filaments preventing assembly and, owing to the dynamic nature of actin filaments, ultimately leads to actin filament disassembly; 2) latrunculin A (latA) which also causes actin filament disassembly, in this case by sequestering actin monomers; and 3) jasplakinolide (jas) which, like phalloidin, binds to and stabilizes actin filaments promoting their assembly. These toxins appear to have variable effects on clathrin-mediated endocytosis in mammalian cells. Several studies have reported that early endocytic events are at least partially inhibited by cytochalasins[27–29]; however, others have reported no effect on receptor-mediated endocytosis[23,30–32]. CytoD treatment selectively inhibited fluid-phase uptake at the apical surface of polarized MDCK and Caco-2 cells without affecting endocytosis at the basolateral surface. In contrast, treatment with jas stimulated the uptake and accumulation of fluid-phase endocytic tracers at the basolateral surface of MDCK cells without affecting endocytosis at the apical surface. Receptor-mediated endocytosis of transferrin (Tfn) was also unaffected. Finally, although both cytoD and latA cause actin filament disassembly, only latA inhibited endocytosis in human adenocarcinoma, A431 cells.
A role for actin in endocytosis was also demonstrated in vitro using a perforated cell assay. Several proteins that sequester actin monomers, including DNAse I, gelsolin fragments and, in particular, β-thymosins, were shown to inhibit endocytic vesicle formation as measured by the receptor-mediated endocytosis of Tfn. In contrast, neither cytoD nor phalloidin (which stabilizes actin filaments) had any effect. While these studies suggested that active actin assembly might be required for endocytosis in mammalian cells, the possibility remained that actin plays a more passive role, for example, in simply maintaining the structural integrity of the plasma membrane.
Given these discrepancies, we have reinvestigated the role of actin in endocytic coated vesicle formation in mammalian cells using several methods to perturb actin dynamics. The results reported here suggest that actin assembly is not directly required for endocytic coated vesicle formation. Instead, they suggest that actin assembly and the organization of a cortical actin network can play, at best, a modulatory role that is more pronounced under certain conditions of growth and membrane differentiation.
We used a well-established perforated cell assay for receptor-mediated endocytosis of Tfn[2,36] to test the hypothesis that de novo assembly of actin filaments might play an active role in endocytic vesicle formation. To this end, we assessed the ability of actin-depleted cytosol to support endocytic coated vesicle formation. As assessed by western blotting,> 95% of actin was depleted from bovine brain cytosol by passage over DNase I-Sepharose. Unexpectedly, the actin-depleted cytosol supported the sequestration of biotinylated-Tfn from exogenously added avidin indistinguishably from untreated cytosol (data not shown). This was true even when assays were performed in the presence of 1 μM phalloidin to stabilize any actin filaments associated with the perforated cell preparations, as these might have provided an alternate source of monomeric actin.
This inability to detect a direct requirement for actin assembly in endocytic coated vesicle formation in perforated cells, coupled to inconsistencies in the reported effects of actin disrupting drugs on endocytosis in intact cells led us to re-examine the requirements for actin filament assembly in endocytosis in vivo using cytochalasin D and latrunculin A. The effects of preincubating A431 cells for 30 min at 37°C with either buffer alone (dashed lines), 5 μM cytoD (open circles) or 10 μM latA (open squares) on internalization of Tfn are shown inFig. 1. Preincubation at these concentrations of cytoD and latA resulted in complete disruption of the actin cytoskeleton, as assessed by Alexa488-phalloidin staining or indirect immunofluorescence using anti-actin antibodies (Fig. 2). Endocytosis assays were performed in three ways: 1) cells were gently removed from tissue culture dishes with PBS/EDTA and both preincubation and internalization assays were performed in suspension (panel a); 2) adherent cells were preincubated with toxins and then released by PBS/EDTA treatment on ice for internalization assays in suspension in the continuous presence of the drugs (panel b); and 3) both preincubation and internalization assays were performed on adherent cells grown on 35 mm tissue culture dishes (panel c). Importantly, the rates (∼20%/min) and extents (∼150% at steady-state) of Tfn internalization in untreated cells were indistinguishable, regardless of the protocol used (compare dashed lines in panels a,b and c). (The fact that the extent of internalization of Tfn exceeded the amount bound on the surface at 4°C reflected the recycling and reinternalization of the substantial intracellular pool of Tfn receptors.) In contrast, the effects of actin filament disruption on endocytosis varied depending on the experimental conditions. Consistent with our previous findings obtained with A431 cells pretreated and assayed in suspension, latA inhibited the rate of endocytosis by ∼40%, whereas cytoD had little or no effect (Fig. 1a). The opposite result was obtained when A431 cells were preincubated on plates and then assayed in suspension (Fig. 1b). Under these conditions cytoD inhibited endocytosis by 60–70%, while latA had little effect. Finally, neither actin filament disrupting drug inhibited endocytosis when A431 cells were pretreated and assayed while remaining adherent (Fig. 1c), even though actin filaments were dramatically disrupted by both toxins under these conditions (Fig. 2, panels b and c).
Similar inconsistencies in the effects of actin disrupting drugs on endocytosis were observed in TRVb-1 CHO cells which express the human transferrin receptor (Fig. 3), although with these cells cytoD and latA had similar effects. When CHO cells were preincubated either in suspension or on plates with cytoD and latA and then assayed in suspension (panels a and b), both the rate and extent of Tfn endocytosis were partially inhibited. In contrast to results obtained with A431 cells, adherent CHO cells treated with latA or cytoD and assayed on plates showed significant (>70%) inhibition in both the rate and extent of Tfn endocytosis. The structural changes in the actin cytoskeleton induced in CHO cells treated with these toxins (Fig. 6, panels b and c) were indistinguishable from those seen in similarly treated A431 cells (Fig. 2b,c). As for A431 cells, the efficiency of endocytosis in untreated CHO cells was not significantly different when assayed in suspension or on plates (compare dashed lines inFig. 3b,c), although a reduction in endocytosis rates was observed during more prolonged incubations in suspension (compareFig. 3a with b and c, see alsoFig. 4a). For both A431 cells and CHO cells, decreases in the extent of Tfn accumulation at steady-state reflect alterations in the relative cell surface versus intracellular pool sizes of TfnR after treatment with the toxins. Inhibition of endocytosis invariably led to increases in surface bound TfnR (not shown), presumably at the expense of intracellular receptors. A decrease in the extent of Tfn accumulation at steady-state might also reflect effects of the actin cytoskeleton on later stages of trafficking through endosomal compartments, as previously reported.
These data suggest that the sensitivity of endocytosis to actin disruption is affected by the adherence properties of each cell type. Because CHO cells can be cultured both on plates and in suspension, we could examine the effects of actin disruption on cells cultured in suspension as compared to those grown in adherent cultures. Although TRVb-1 CHO cells cultured in suspension internalized Tfn less efficiently than when grown on plates, receptor-mediated endocytosis in suspension-grown CHO cells was completely unaffected by either latA or cytoD (Fig. 4a). Similarly, the efficient endocytosis of Tfn in human K562 erythroleukemic cells, which grow only in suspension, was unaffected by preincubation with either latA or cytoD (Fig. 4b).
Jasplakinolide (jas) is a recently discovered membrane permeant compound that binds to and stabilizes actin filaments much like phalloidin[26,38]. Interestingly, cytoD and jas have opposing effects on fluid-phase endocytosis from distinct plasma membrane domains in polarized MDCK cells: cytoD inhibits endocytosis selectively at the apical surface, whereas jas enhances endocytosis selectively at the basolateral surface. We therefore tested the effects of jas on endocytosis, assayed both in suspension and on plates, in several cell lines. Incubation of cells with 1 μM jas for 45 min at 37°C resulted in dramatic changes in the morphology of actin filaments and thick, short actin bundles were detected by indirect immunofluorescence using anti-actin antibodies (Fig. 2, panel d, not shown for CHO cells). No detectable labeling was observed using Alexa-phalloidin, indicating that the phalloidin binding sites were now occupied with jas (Fig. 6d, not shown for A431 cells).Fig. 5 shows that as for latA and cytoD, treatment with jas gave variable effects on endocytosis depending on the cell line and assay conditions. Endocytosis in A431 cells was sensitive to jas only when assayed in suspension (compareFig. 5 a with b). A complementary pattern of inhibition and resistance was seen in 3T3-L1 fibroblasts (Fig. 5c,d). Consistent with results obtained using actin disrupting drugs (Fig. 3) endocytosis in CHO cells cultured on plates was sensitive to jas under either assay condition (Fig. 5e,f). In contrast, endocytosis in the Cos-7 fibroblast cell lines was not significantly affected by jas under either condition (Fig. 5g,h), even when treated with 4 μM jas (not shown). Endocytosis in K562 and CHO cells grown in suspension was also unaffected by 1 μM jas (Fig. 4, triangles).
Together these data suggest that neither filamentous actin nor actin assembly are directly required for clathrin-mediated endocytosis in mammalian cells. In yeast, cortical actin patches were found associated with invaginations of the plasma membrane and it has been suggested that these might correspond to sites of endocytosis. In mammalian cells, disruption of the actin cytoskeleton by latA leads to partial dispersal of highly localized endocytic ‘hot-spots’ . We therefore examined the effect of actin disruption on the distribution of the plasma membrane-specific adaptor protein, AP2, in adherent TRVb-1 CHO cells in which receptor-mediated endocytosis was most strongly affected (Fig. 6). In control cells, AP2 is distributed in small puncta at the plasma membrane and a diffuse cytosolic pool is also detectable (Fig. 6a). Cells treated with either cytoD (Fig. 6b), latA (Fig. 6c) or jas (Fig. 6d) appear to retract towards their nucleii leaving thin membrane sheets adherent to the glass surface. The AP2-staining puncta on these flattened membranes appeared larger than in control cells. Whether these slight changes in AP2 staining patterns relate to the decreased rates of endocytosis is unknown.
Using quick freeze deep-etch methodology, we examined the ultra-structural relationship between cortical actin filaments and clathrin-coated pits. We focused our attention on adherent A431 cells (Fig. 7), Cos cells and CHO cells (Fig. 8a and b, respectively) as receptor-mediated endocytosis in these cell types was differentially sensitive to disruption of the actin cytoskeleton. Endocytosis in adherent CHO cells was more sensitive to disruption of actin cytoskeleton than that in A431 cells or Cos cells (compareFig. 1c withFig. 3c orFig. 5 panels a, f and h). The images inFigs 7 and 8 are displayed as three dimensional anaglyphs that should be viewed with red/green or red/blue glasses. Several coated pits at various stages of invagination can be seen in A431 cells (Fig. 7). While actin filaments are sometimes seen around the base of coated pits (top panel), several others appeared to bud from areas devoid of actin filaments (bottom panel). In contrast, caveolae, which appear as smaller diameter vesicles covered by a ‘whorling’ coat structure and whose uptake has been shown to be actin dependent[41,42], appear to line up along actin filaments. Cos-7 cells have an elaborate cortical actin cytoskeleton (Fig. 8, top) and coated pits can be seen to bud from areas devoid of actin. At higher magnification in CHO cells (Fig. 8, bottom), actin filaments can be seen to run adjacent to the coated pit on the left and above the coated pit on the right.
Much remains to be understood regarding the mechanisms of clathrin-mediated endocytosis. While the roles played by adaptors, clathrin and dynamin are being elucidated, the function of other proteins implicated in endocytosis, including amphiphysin, eps15, endophilin, etc. remain largely obscure. Several workers have shown that clathrin-mediated endocytosis is inhibited upon disruption of the actin cytoskeleton[27,28,31,33,43], suggesting a role for actin in endocytosis. However, given previous conflicting results[23,29,30,33,34], we sought to determine whether filamentous actin or actin assembly play direct roles in endocytic vesicle formation. Assuming that the fundamental machinery driving coat assembly, receptor recruitment and vesicle detachment at the plasma membrane is the same in all mammalian cells, then our results suggest that neither filamentous actin nor actin assembly/disassembly are obligatorily required for clathrin coated vesicle formation. Consistent with this, electron microscopic analysis failed to detect a specific structural relationship between actin filaments and endocytic coated pits at any stage of vesicle formation. Instead, these data would suggest that actin filaments have, at most, an accessory function in endocytic vesicle formation required only under select conditions.
A more direct functional link between endocytosis and actin assembly and organization has been suggested to occur in S. cerevisiae. However, even in yeast, the correlation between actin disruption and inhibition of endocytosis is not absolute. For example, there exist mutant alleles of myo5 (affecting a type I unconventional myosin)[44,45], pfy1 (affecting profilin) and tpm1 (affecting tropomyosin I) that disrupt actin organization in yeast but do not inhibit endocytosis (AL Munn, E Kubler and H Reizman, unpublished results, but see[8,14]). Recent genetic analysis of the end5-1 allele has yielded similar results (SN Naqvi and AL Munn, personal communication). The end5-1 mutation in yeast disrupts both actin organization and endocytosis. Cloning and further characterization revealed end5p to be a proline-rich protein weakly homologous to WIP, a human WASp-interacting protein and two hybrid analysis confirmed interactions between end5p and las17p, the yeast homologue of WASp, which is also necessary for endocytosis. Importantly, some mutations that suppress the temperature-sensitive growth defect associated with end5-1 have been identified which selectively suppress either the actin organization defect or the endocytic defect of end5-1 (SN Naqvi and AL Munn, personal communication). These results provide further evidence that actin assembly and cortical actin organization might not be obligatorily coupled to all forms of endocytosis, even in yeast.
If not directly required for endocytic clathrin coated vesicle formation, what accessory role(s) might actin be playing? Actin filaments might be required when endocytosis must be highly localized at specific sites; for example at the base of microvilli in polarized epithelial cells[33,49,50], at endocytic hot-spots adjacent of active sites in the synapse[51,52], or after patching and capping of receptors on lymphocytes. Alternatively, actin might be differentially required when endocytosis occurs from the adherent versus nonadherent surfaces of cells, depending on the complexity of the cortical actin cytoskeleton underlying the plasma membrane. Actin might have a more dominant role in endocytosis from membranes that are made more rigid by contacts with the extracellular matrix or from membranes that have more elaborate cortical networks than those found in the cultured cells studied here. Actin and perhaps actin based motors might also be required to facilitate movement of detached vesicles through the cell cortex, especially where there is a well-developed structure such as at the terminal web or within microvilli in polarized epithelial cells. Indeed, actin tails were seen to assemble onto newly formed pinocytic vesicles in cultured mast cells. This role might be analogous to the facilatory role of microtubules and microtubule-based motors, whose requirements for vesicular trafficking are most apparent over longer distances. Importantly, our biochemical measurements assess the formation of sealed endocytic vesicles: subsequent events such as the movement of newly formed endocytic vesicles through the cell cortex or trafficking through the endosomal compartment are not assessed. It is likely, based on results of others, that actin plays a role in these later events associated with endocytosis and endosomal trafficking[28,55].
Our results provide insight into the discrepancies in the published literature regarding the effects of actin disruption on clathrin-mediated endocytosis in that endocytosis in different cell types or the same cell type assayed under different conditions, was found to be differentially sensitive to actin disrupting toxins. Moreover, they suggest that perturbations of the actin cytoskeleton can impact the rates and extents of clathrin-mediated endocytosis under conditions related to the adherent properties of a given cell line. Based on these variable effects of actin disruption on endocytosis, we conclude that neither actin assembly nor actin filament organization play an obligatory role in endocytic coated vesicle formation in mammalian cells.
Materials and Methods
Cells and reagents
Human adenocarcinoma A431 cells and human erythroleukemic K562 cells were cultured as previously described. Cos-7 cells were maintained in DMEM containing penicillin, streptomycin and 10% fetal calf serum (Hyclone). TTA-HeLa cells were maintained as previously described. Chinese hamster ovary cells expressing human Tfn receptors, designated TRVb-1, were obtained from Tim McGraw (Columbia University, NY) and were maintained as previously described. Mouse 3T3-L1 preadipocytes were obtained from Arie Verkleij (Utrecht University, The Netherlands) and cultured as described.
Cytochalasin D (Calbiochem, CA), Latrunculin A (Molecular Probes, OR) and Jasplakinolide (Molecular Probes, OR) were stored at −20°C as 1 mM stock solutions in DMSO. Control experiments confirmed that DMSO alone had no effect on cells at the concentrations used. Biotinylated Tfn, B-XX-Tfn was prepared as previously described[31,57]. All other chemicals were reagent grade.
Actin depletion and endocytosis in perforated cells
Bovine brain cytosol was prepared as previously described and incubated with DNase-Sepharose for 30 min at 4°C with end-over-end mixing. Mock treated cytosol was incubated with unconjugated Sepharose and served as a control. The extent of actin depletion was assessed by western blotting and determined to be> 95%. Endocytosis in perforated A431 cells was performed exactly as described.
Endocytosis assays and toxin treatment
Condition 1: Adherent cells, plated the previous day on 15 cm dishes, were brought into suspension by incubation for 5 min at room temperature in PBS containing 5 mM EDTA, collected by centrifugation for 3–5 min at 1 500 rpm and resuspended in ∼300 μl of ice cold SFM (DMEM containing 0.2% BSA, 20 mM Hepes pH 7.4) to a density of ∼2×107 cells/ml. Cells cultured in suspension were washed with SFM, collected by centrifugation and resuspended to 2×107 cells/ml in SFM. For each assay, 50 μl of suspension cells were added to 450 μl SFM and the appropriate concentration of toxin in 13×100 mm borosilicate tubes and the samples preincubated at 37°C with or without toxin (30 min for cytoD and latA, 45 min with jas). Tubes were gently shaken every 3–5 min to maintain even suspension of cells. Assay tubes were briefly returned to ice, 10 μl of 100 μg/ml B-XX-Tfn were added. Two 50 μl aliquots were removed to determine total bound and 4°C incubation backgrounds. The remaining samples were incubated at 37°C and 50 μl aliquots were removed to ice at each time point after gentle mixing to ensure uniform cell suspension. Internalization of receptor-bound B-XX-Tfn was determined by its inaccessibility to exogenously added avidin using an ELISA-based assay exactly as previously described[31,57].
Condition 2: Cells were plated the previous day on 10-cm dishes. Plates were washed with SFM on ice and then incubated for 30–45 min with or without the indicated concentration of toxin at 37°C. Cells were then brought into suspension, harvested and assayed for Tfn uptake as above, except in that toxin levels were kept constant in all buffers.
Condition 3: Cells were plated the previous day on 35 mm dishes, one plate was used for each time point and for determining total bound at 4°C controls. Adherent cells were washed with PBS and then preincubated with 500 μl SFM with or without toxin as indicated for 30–35 min at 37°C. Plates were placed on ice cold metal trays over ice and media was replaced with SFM containing 2 μg/ml B-XX-Tfn. Plates were ‘floated’ in a 37°C water bath and incubated for the indicated times in the continuous presence of the toxin. After incubation, cells were removed by incubation with PBS/EDTA for 15–30 min on ice, collected by centrifugation in a microfuge and processed to determine avidin inaccessible B-XX-Tfn as previously described[31,57].
Curves shown in all the figures are averages of 3–4 independent experiments which varied by <10%. The standard deviation for these experimental variations are indicated by the error bars in controls which are representative of the slight variations obtained with treated samples.
Cells, plated on coverslips the previous day, were incubated with toxins at indicated concentrations for 30 (cytoD and latA) or 45 min (jas) and then fixed for 20 min at room temperature with 4% PFA in PBS++ (containing 1 mM MgCl2 and 1 mM CaCl2). Cells were permeabilized and blocked with PBS++ containing 4% normal goat serum and 0.1% TX-100. Labeling was performed with 1:40 dilution of Alexa488-phalloidin (Molecular Probes, OR) according to the manufacturers instructions. AP.6 monoclonal antibody against α-adaptin was used at 40 μg/ml and detected with Texas Red conjugated goat anti-mouse (Molecular Probes, OR). Cells were viewed and photographed through a 63X epifluorescence objective lens and each panel is shown at identical magnification.
For electron microscopy, HeLa cells were grown as above, but on 3×3 mm pieces of glass generated by scoring and breaking standard #1 glass coverslips. (The tiny size of these coverslips is to facilitate their later quick-freezing.) The cells are ‘unroofed’ by exposure to a 1 s ultrasonic burst from a probe-type sonicator held ∼3 mm above the coverslip. This is performed in a ‘stabilization buffer’ consisting of 70 mM KCl, 30 mM HEPES buffer brought to pH 7.4 with KOH, 5 mM MgCl2, 3 mM EGTA, 1 mM DTT and 0.1 mM AEBSF as a protease inhibitor. To facilitate ‘unroofing’ and to decrease the proportion of cells that are completely torn off the glass during the moment of sonication, cells are prewashed for 15–30 min in warm PBS containing 2 mM CaCl2 and 1 mM MgCl2 to remove all tissue culture medium and then pretreated immediately before sonication with a 10 s exposure to 0.4 mg/ml of polylysine (Sigma, MO; molecular weight ∼40–70kDa) followed by three 5 s rinses in the above ‘stabilization buffer’ diluted 1:3 with distilled water. This ‘tacks down’ the edges of the cells and swells them gently, thereby greatly increasing the yield of properly unroofed cells. Immediately after ‘unroofing’, the coverslips are transferred to 2% glutaraldehyde (EM Grade, from EMS) dissolved in the same ‘stabilization buffer’ as above. After 2–4 h of fixation, the coverslips are then picked out of the fixative with fine forceps and washed by brief immersion in several different dishes of distilled water, using extreme care to insure that no oil films are generated on the dishes of water and transferred onto the coverslips. (The reason for this 30–90 s water-wash, the most problematic step in the whole procedure, is that any residual salts or organic molecules left on the coverslips at the time of freezing will appear on their surfaces as an unattractive ‘scum’ after freeze drying.) Next, without allowing any time for air-drying, the water-washed coverslips are mounted on 3×3 mm slabs of aldehyde-fixed rabbit lung (0.8 mm thick) as a cushion and quick-frozen by abrupt contact with an ultrapure copper block cooled to 4° above absolute zero by a spray of liquid helium. Thereafter, the coverslips are stored in liquid nitrogen until mounting in a Balzers’ Model 301 vacuum-evaporator. In this device, they are freeze-dried by warming to −80°C for 15 min and then rotary-replicated with a thin (∼2 nm) film of platinum evaporated over 5–10 s from an electron beam gun mounted 15–20° above the horizontal, all while the coverslip is rotated at 5 Hz. The replica is then immediately supported or ‘backed’ by evaporating ∼10 nm of pure carbon onto it, using a standard carbon-arc supply mounted approximately 10° off the vertical, while the sample continues to rotate to insure the generation of a uniform, strong film of carbon. Coverslips are then removed from the Balzers, allowed to thaw and the replica is floated off by immersion at an ∼45° angle into full strength (47%) hydrofluoric acid (HF). Immediately thereafter, the replica is picked up on the surface of the HF with a glass rod and transferred via the rod through several washes of distilled water, then a brief wash with standard household bleach and more distilled water washes, before being picked up on a 75 mesh formvar-coated EM grid. For electron microscopy, the grid is mounted in a eucentric side-entry goniometer stage of a JEOL 200CX electron microscope, imaged at 30–70 K magnification and photographed in stereo at ±10° of tilt off the vertical axis.
For the production of final ‘anaglyph’ stereo images, the two micrographs representing each field are placed in proper register on a Bessler copy stand and photographed at an additional 3–6×magnification with a Kodak 520 digital camera, producing a 2 000×1 200 pixel, ∼1.5 MB B & W file for each view. The files are next sorted into left and right views by direct inspection on the computer screen and then using Adobe Photoshop®, the left view is converted to a pure red-channel RGB image and the right view to a blue+green-channel RGB image. Next, either one of these colored images is copied directly onto the other and the two are imaged simultaneously by selecting the ‘screen’ command in the ‘Layers’ menu of Photoshop. This creates an anaglyph stereo image of the original field, in a roughly 3–4 MB file. The anaglyph is finally brought into perfect alignment by using the ‘free translate’ command in Photoshop, on one of the two layers. (This even allows for correction of slight mismatches in magnification between the two original electron micrographs.) The final digital anaglyph stereo image is then printed on a standard dye-sublimation printer for publication and viewed with anaglyph glasses wearing the red lens on the left and the green or blue lens on the right.
We thank Dr Alan Munn for helpful discussion and for communicating his unpublished results. This work was supported by USPHS grants CA69099 to SLS and GM29647 to JEH.