pp60v-src association with the cytoskeleton induces actin reorganization without affecting polymerization status

The mechanism by which Rous sarcoma virus (RSV) induces a reorganization of actin and its associated proteins and a reduction in microfilament bundles is at present poorly understood. To examine the relationship between the organization of the microfilament system and the polymerization state of actin after transformation, we have investigated these changes in a Rat-1 cell line transformed by LA29, a temperature-sensitive (ts) mutant of RSV. Parallel immunofluorescence and biochemical analysis demonstrated that LA29 pp60v-src was ts for tyrosine kinase activity and cytoskeletal association. Changes in the distribution and organization of actin, alpha-actinin and vinculin were dependent on the association of a kinase-active pp60v-src molecule with the detergent-insoluble cytoskeleton. Whilst there was a transformation-dependent loss of microfilament bundles, biochemical quantitation demonstrated that the polymerization state of the actin in both detergent-soluble and insoluble fractions of these cells grown at temperatures either permissive or restrictive for transformation was quantitatively unchanged. These results indicate that the loss of microfilament bundles after transformation is not due to a net depolymerization of filamentous actin but rather to a reorganization of polymeric actin from microfilament bundles and stress fibers to other polymeric forms within the cell. The polymeric nature of the actin in these cells was confirmed by electron microscopy of cytoskeletons and substrate-adherent membranes.     

Organization of stress fibers in cultured fibroblasts after extraction of actin with bovine brain gelsolin-like protein

Mouse and quail embryo fibroblasts were extracted with Triton X-100 and the resulting cytoskeletons were treated with gelsolin-like actin-capping protein (the 90-kDa protein-actin complex isolated from bovine brain). Staining of cells with rhodamine-conjugated phalloin or an antibody to actin did not reveal any actin-containing structures after treatment with the 90-kDa protein-actin complex. Extraction of actin was confirmed by SDS-gel electrophoresis. Immunofluorescence microscopy showed that vinculin and α-actinin were released from the cytoskeletons together with actin. However, myosin remained associated with the cytoskeleton after treatment with the 90-kDa protein-actin complex. The distribution of myosin in treated cells showed no significant difference from that in control cells: in both cases myosin was localized mainly in the stress fibers. Double-fluorescence staining showed the absence of actin in myosin-containing stress fibers of treated cells. The ultrastructural organization of actin-depleted stress fibers was studied by transmission electron microscopy of platinum replicas. On electron micrographs these fibers appeared as bundles of filaments containing clusters of globular material. It is concluded that myosin localization in stress fibers does not depend on actin.     

Direct visualization of bipolar myosin filaments in stress fibers of cultured fibroblasts

The authors examined the molecular organization of myosin in stress fibers (microfilament bundles) of cultured mouse embryo fibroblasts. To visualize the organization of myosin filaments in these cells, fibroblast cytoskeletons were treated with gelsolin-like protein from bovine brain (hereafter called brain gelsolin), which selectively disrupts actin filaments. As shown earlier [Verkhovsky et al., 1987], this treatment did not remove myosin from the stress fibers. The actin-free cytoskeletons then were lightly sonicated to loosen the packing of the remaining stress fiber components and fixed with glutaraldehyde. Electron microscopy of platinum replicas of these preparations revealed dumbbell-shaped structures of approximately 0.28 micron in length, which were identified as bipolar myosin filaments by using antibodies to fragments of myosin molecule (subfragment 1 and light meromyosin) and colloidal gold label. Bipolar filaments of myosin in actin-free cytoskeletons were often organized in chains and lattices formed by end-to-end contacts of individual filaments at their head-containing regions. Therefore, after extraction of actin, it was possible for the first time to display bipolar myosin filaments in the stress fibers of cultured cells.     

Visualization of the same PtK2 cytoskeletons by both immunofluorescence and low power electron microscopy.

A procedure is described which allows the examination of the cytoskeleton of a single PtK2 cell first by immunofluorescence and then by electron microscopy after staining with uranyl acetate. The immunofluorescent patterns of these detergent resistant cytoskeletons elicited with various monospecific antibodies closely resemble the patterns found in whole cells. Comparison of the immunofluorescence and electron micrographs directly supports the previous assignments of actin, myosin, filamin, α-actinin and tropomyosin as proteins associated with microfilament bundles in non-muscle cells. Actin is also found associated with a fine lattice-like structure present both in the ruffles and lying above the microfilament bundles in the cell body. The tonofilament bundles present in PtK2 cytoskeletons are not decorated by antibodies directed against the proteins associated with microfilament bundles. Antibodies directed against tonofilaments decorate specifically this system and not the microfilament bundles.     

Differences in the organization of actin in the growth cones compared with the neurites of cultured neurons from chick embryos

Sensory neurons from chick embryos were cultured on substrata that support neurite growth, and were fixed and prepared for both cytochemical localization of actin and electron microscopic observation of actin filaments in whole-mounted specimens. Samples of cells were treated with the detergent Triton X-100 before, during, or after fixation with glutaraldehyde to determine the organization of actin in simpler preparations of extracted cytoskeletons. Antibodies to actin and a fluorescent derivative of phallacidin bound strongly to the leading margins of growth cones, but in neurites the binding of these markers for actin was very weak. This was true in all cases of Triton X- 100 treatment, even when cells were extracted for 4 min before fixation. In whole-mounted cytoskeletons there were bundles and networks of 6-7-nm filaments in leading edges of growth cones but very few 6-7-n filaments were present among the microtubules and neurofilaments in the cytoskeletons of neurites. These filaments, which are prominent in growth cones, were identified as actin because they were stabilized against detergent extraction by the presence of phallacidin or the heavy meromyosin and S1 fragments of myosin. In addition, heavy meromyosin and S1 decorated these filaments as expected for binding to F-actin. Microtubules extended into growth cone margins and terminated within the network of actin filaments and bundles. Interactions between microtubule ends and these actin filaments may account for the frequently observed alignment of microtubules with filopodia at the growth cone margins.     

Actin bundles in human hair follicles as revealed by confocal laser microscopy

A confocal laser microscope was used to examine the distribution pattern of actin bundles in whole-mounts of human hair follicles stained with fluorescently labeled phalloidin. Actin bundles were found exclusively in the epithelial outer root sheath of the lower and middle portions of the follicle. In the growth stage, the lower follicle was characterized by well-developed actin bundles arranged circumferentially in the innermost and outermost cell layers of the outer root sheath. Actin bundles in the innermost cells were aligned end-to-end so that they formed complete circular bands surrounding the inner root sheath. In the outermost cells, actin bundles ran underneath the basal plasma membrane to which they attached at both ends. In contrast, in the quiescent stage, actin bundles in the lower follicle were disposed radially toward the follicle surface where they terminated perpendicular to the basal plasma membrane. In the middle follicle, circumferential actin bundles were found only in the intermediate layer of the outer root sheath throughout the hair cycle. Immunofluorescent anti-myosin and anti-α-actinin staining showed a striated pattern along actin bundles. Vinculin was localized at both ends of actin bundles, corresponding to the cell-to-cell or cell-to-substrate adherens junctions. Glycerinated follicles changed in shape on the addition of MgATP, suggesting a contraction of actin bundles. From these observations, we conclude that actin bundles in the hair follicle are comparable to stress fibers and that they serve as a tensile scaffold for the growth and integrity of the follicle. Received: 6 May 1995 / Accepted: 25 October 1995     

Immunocytochemical localization of fodrin and ankyrin in bovine chromaffin cells in vitro.

We raised antibodies to brain fodrin and erythrocyte ankyrin and examined the distribution of the antigens in cultured bovine chromaffin cells by immunocytochemical techniques. Immunofluorescence microscopy of whole cells showed intense labeling for both proteins, but fine localization could not be determined. In contrast, in cell specimens mechanically unroofed before fixation, the distribution of the two proteins revealed an apparent difference in the ventral plasma membrane: immunofluorescence for fodrin was dense and mostly even, whereas that for ankyrin appeared as scattered dots. Immunogold electron microscopy of the unroofed cells showed that labeling for fodrin was localized in a network of thin filaments, the diameter of which was 2-3 nm at the thinnest portion. Ankyrin labeling was mostly associated with filaments 5-10 nm in diameter. Notably, labeling for both fodrin and ankyrin was found over the coated membrane. The present results indicate that fodrin and ankyrin in the chromaffin cell do not constitute a submembranous network as spectrin and ankyrin do in the erythrocyte; whereas fodrin is closely associated with the plasma membrane, ankyrin is mostly linked to the cytoskeleton. The existence of both proteins in the coated region implies that they are functionally related to exocytosis and/or to ensuing membrane retrieval in the chromaffin cell.     

Mechanisms of actin rearrangements mediating platelet activation.

The detergent-insoluble cytoskeleton of the resting human blood platelet contains approximately 2,000 actin filaments approximately 1 micron in length crosslinked at high angles by actin-binding protein and which bind to a spectrin-rich submembrane lamina (Fox, J., J. Boyles, M. Berndt, P. Steffen, and L. Anderson. 1988. J. Cell Biol. 106:1525-1538; Hartwig, J., and M. DeSisto. 1991. J. Cell Biol. 112:407-425). Activation of the platelets by contact with glass results within 30 s in a doubling of the polymerized actin content of the cytoskeleton and the appearance of two distinct new actin structures: bundles of long filaments within filopodia that end at the filopodial tips (filopodial bundles) and a circumferential zone of orthogonally arrayed short filaments within lamellipodia (lamellipodial network). Neither of these structures appears in cells exposed to glass with cytochalasin B present; instead the cytoskeletons have numerous 0.1-0.3-microns-long actin filament fragments attached to the membrane lamina. With the same time course as the glass-induced morphological changes, cytochalasin-sensitive actin nucleating activity, initially low in cytoskeletons of resting platelets, increases 10-fold in cytoskeletons of thrombin-activated platelets. This activity decays with a time course consistent with depolymerization of 0.1-0.3-microns-long actin filaments, and phalloidin inhibits this decay. Cytochalasin-insensitive and calcium-dependent nucleation activity also increases markedly in platelet extracts after thrombin activation of the cells. Prevention of the rise in cytosolic Ca2+ normally associated with platelet activation with the permeant Ca2+ chelator, Quin-2, inhibits formation of lamellipodial networks but not filopodial bundles after glass contact and reduces the cytochalasin B-sensitive nucleation activity by 60% after thrombin treatment. The filopodial bundles, however, are abnormal in that they do not end at the filopodial tips but form loops and return to the cell body. Addition of calcium to chelated cells restores lamellipodial networks, and calcium plus A23187 results in cytoskeletons with highly fragmented actin filaments within seconds. Immunogold labeling with antibodies against gelsolin reveals gelsolin molecules at the ends of filaments attached to the submembrane lamina of resting cytoskeletons and at the ends of some filaments in the lamellipodial networks and filopodial bundles of activated cytoskeletons. Addition of monomeric actin to myosin subfragment 1-labeled activated cytoskeletons leads to new (undecorated) filament growth off the ends of filaments in the filopodial bundles and the lamellipodial network. The simplest explanation for these findings is that gelsolin caps the barbed ends of the filaments in the resting platelet. Uncapping some of these filaments after activation leads to filopodial bundles.(ABSTRACT TRUNCATED AT 400 WORDS)     

Fodrin: axonally transported polypeptides associated with the internal periphery of many cells

Fodrin (formerly designated 26 and 27) comprises two polypeptides (250,000 and 240,000 mol wt) that are axonally transported at a maximum time-averaged velocity of 40 mm/d--slower than the most rapidly moving axonally transported proteins, but faster than at least three additional groups of proteins. In this communication, we report the intracellular distribution of fodrin. Fodrin was purified from guinea pig brain, and a specific antifodrin antibody was produced in rabbit and used to localize fodrin in tissue sections and cultured cells by means of indirect immunofluorescence. Fodrin antigens were highly concentrated in the cortical cytoplasm of neurons and also nonneuronal tissues (e.g., skeletal muscle, uterus, intestinal epithelium). Their disposition resembles a lining of the cell: hence, the designation fodrin (from Greek fodros, lining). In cultured fibroblasts, immunofluorescently labeled fodrin antigens were arranged in parallel arrays of bands in the plane of the plasma membrane, possibly reflecting an exclusion of labeled fodrin from some areas occupied by stress fibers. The distribution of fodrin antigens in mouse 3T3 cells transformed with simian virus 40 was more diffuse, indicating that the disposition of fodrin is responsive to altered physiological states of the cell. When mixtures of fodrin and F-actin were centrifuged, fodrin cosedimented with the actin, indicating that these proteins interact in vitro. We conclude that fodrin is a specific component of the cortical cytoplasm of many cells and consider the possibilities: (a) that fodrin may be indirectly attached to the plasma membrane via cortical actin filaments; (b) that fodrin may be mobile within the cortical cytoplasm and that, in axons, a cortical lining may be in constant motion relative to the internal cytoplasm; and (c) that fodrin could serve to link other proteins and organelles to a submembrane force-generating system.     

The submembranous location of p11 and its interaction with the p36 substrate of pp60 src kinase in situ

p36, a major cytoplasmic substrate of pp60 src kinase, is present beneath the plasma membrane. It can be isolated either as a monomer or as a heterotetramer (protein I) containing two copies each of p36 and a unique p11 polypeptide. To compare the expression rules of p36 and p11 as well as their cellular distributions, monoclonal antibodies to the two porcine proteins were isolated. In tissue culture cells p11-specific antibodies decorated the same submembranous compartment previously seen with antibodies to p36 and fodrin or spectrin and followed the p36 images under all fixation/extraction conditions tested. Immunofluorescence microscopy on tissue sections showed coincident expression patterns of both proteins confirming and extending previous results with p36 antibodies. Antibodies with limited cross-species reaction have been used to trace the fate of porcine p11 and p36 injected into cultured cells. Both proteins are incorporated in the submembranous compartment, where they remain in Triton cytoskeletons prepared in the presence but not in the absence of Ca2+. The incorporation of p36 in vivo conforms with its Ca2+-dependent binding to actin, fodrin, and certain phospholipids in vitro. In contrast, the incorporation of p11 seems to depend on an in situ interaction with p36 or an exchange with endogenous p11 present on p36. The combined results indicate a strong coupling of p11 and p36 in cellular compartmentalization and tissue differentiation.     

α-Actinin: Immunofluorescent localization of a muscle structural protein in nonmuscle cells

Antibodies specific for the skeletal muscle structural protein α-actinin are used to localize this protein by indirect immunofluorescence in nonmuscle cells. In cultured nonmuscle cells, α-actinin is localized along or between actin filament bundles producing an almost regular periodicity. The protein is also detected in the form of fluorescent plaques at some ends of actin filament bundles, as well as in a filamentous form in some overlap areas of cells. In spreading rat embryo cells, α-actinin assumes a focal distribution which corresponds to the vertices of a highly regular actin filament network. The results suggest that α-actinin may be involved in the organization of actin filament bundles, in the attachment of actin filaments to the plasma membrane, and in the assembly of actin filaments in areas of cell to cell contact.     

Influence of calmodulin on the human red cell membrane skeleton

The calcium receptor calmodulin interacts with components of the human red cell membrane skeleton as well as with the membrane. Under physiological salt conditions, calmodulin has a calcium-dependent affinity for spectrin, one of the major components of the membrane skeleton. It is apparent from our results that calmodulin inhibits the ability of erythrocyte spectrin (when preincubated with filamentous actin) to create nucleation centers and thereby to seed actin polymerization. The gelation of filamentous actin induced by spectrin tetramers is also inhibited by calmodulin. The inhibition is calcium dependent and decreases with increasing pH, similar to the binding of calmodulin to spectrin. Direct binding studies using aqueous two-phase partition indicate that calmodulin interferes with the binding of actin to spectrin. Even in the presence of protein 4.1, which is believed to stabilize the ternary complex, calmodulin has an inhibitory effect. Since calmodulin also inhibits the corresponding activities of brain spectrin (fodrin), it appears likely that calmodulin may modulate the organization of cytoskeletons containing actin and spectrin or spectrin analogues.     

Further characterization of a brain high molecular weight actin-binding protein (BABP): interaction with brain actin and ultrastructural studies

Crude actomyosin fraction from porcine brain contained a large amount of high molecular weight actin-binding protein (BABP). The molar ratio of BABP to actin (BABP/actin) in the fraction was estimated to be 0.22. From this fraction, BABP and actin were solubilized with a molar ratio of 0.25, suggesting the existence of an interaction between BABP and brain actin. BABP was finally purified to 90% purity. The purified BABP was negatively stained and observed by electron microscopy; it appeared to be a slender, flexible, two-stranded molecule whose contour length was about 200 nm. The structure was very similar to those of fodrin and other high molecular weight actin-binding proteins such as filamin, spectrin, and ABP. Lattice cage-like structures composed of BABP molecules were occasionally observed at high BABP concentrations. The addition of BABP to actin filaments resulted in the appearance of many branching, filamentous bundles. The electron microscopic observations suggested that a single BABP molecule could crosslink actin filaments, that is, one BABP molecule has two actin binding sites.     

Some of eukaryotic elongation factor 2 is colocalized with actin microfilament bundles in mouse embryo fibroblasts

Indirect immunofluorescent microscopy was used to study the distribution of eukaryotic elongation factor 2 (EF-2) in cultured mouse embryo fibroblasts. The perinuclear area (endoplasm) of all the cells and many straight cables running along the whole cytoplasm were stained with monospecific goat or rabbit antibodies to rat liver EF-2. Double staining of the cells with antibodies to EF-2 and rhodaminyl-phalloidin (used for actin microfilament detection) showed that EF-2 containing cables coincided with bundles of actin microfilaments. Not all actin microfilament bundles contained EF-2: sometimes EF-2 was not observed in bundles running along the cell edges or in actin microfilament junctions. Triton X-100 extracted most of EF-2 from the cells and no actin microfilament bundles were stained with the EF-2 antibodies in the Triton-extracted cells. Thus, in mouse embryo fibroblasts EF-2 can be found along actin microfilament bundles, but it is unlikely to be their integral protein.     

Organization of actin in the leading edge of cultured cells: influence of osmium tetroxide and dehydration on the ultrastructure of actin meshworks

The ordered structure of the leading edge (lamellipodium) of cultured fibroblasts is readily revealed in cells extracted briefly in Triton X- 100-glutaraldehyde mixtures, fixed further in glutaraldehyde, and then negatively stained for electron microscopy. By this procedure, the leading edge regions show a highly organised, three-dimensional network of actin filaments together with variable numbers of radiating actin filament bundles or microspikes. The use of Phalloidin after glutaraldehyde fixation resulted in a marginal improvement in filament order. Processing of the cytoskeletons though the additional steps generally employed for conventional electron microscopy resulted in a marked deterioration or complete disruption of the order of the actin filament networks. In contrast, the actin filaments of the stress fiber bundles were essentially unaffected. Thus, postfixation in osmium tetroxide (1% for 7 min at room temperature) transformed the networks to a reticulum of kinked fibers, resembling those produced by the exposure of muscle F-actin to OsO4 in vitro (P. Maupin-Szamier and T. D. Pollard. 1978. J. Cell Biol. 77:837--852). While limited exposure to OsO4 (0.2+ for 20 min at 0 degrees C) obviated this destruction, dehydration in acetone or ethanol, with or without post-osmication, caused a further and unavoidable disordering and aggregation of the meshwork filaments. The meshwork regions of the leading edge then showed a striking resemblance to the networks hitherto described in critical point-dried preparations of cultured cells. I conclude that much of the "microtrabecular lattice" described by Wolosewick and Porter (1979. J. Cell Biol. 82:114--139) in the latter preparations constitutes actin meshworks and actin filament arrays, with their associated components, that have been distorted and aggregated by the preparative procedures employed.     

Role of actin filaments in shape formation of mesenteric mesothelial cells of the bullfrog

The role of actin filaments in the development of cellular shape in the mesenteric mesothelium of the bullfrog was studied by using a simple, new technique for making en face preparations of mesothelial sheets. By using these mesothelial cell preparations, the distribution of actin was determined by means of fluorescence microscopy with 7-nitrobenz-2-oxa-1,3-diazole (NBD)-phallacidin and that of myosin by means of immunofluorescence microscopy. Although fluorescence produced by both NBD-phallacidin and antimyosin staining was found exclusively along the margins of the cells, its intensity was altered in correspondence with changes in cell shape. For instance, tadpole-type mesothelial cells with either an irregular or very slender cell shape showed very weak fluorescence. On the other hand, frog-type mesothelial cells with a polygonal shape showed intense fluorescence at their margins and had circumferential bundles of actin filaments at their apices. Furthermore, intercellular junctions between the mesothelial cells developed as the cell shape became polygonal during metamorphosis. The present study showed that development of circumferential bundles of actin filaments and intercellular junctions may serve to establish and maintain the definitive polygonal cellular pattern in the mesenteric mesothelium of the bullfrog.     

Organization of stress fibrils in cultured fibroblasts studied by using an actin filament-fragmenting protein

Earlier we isolated a 1:1 complex of 90 kD-protein and actin from bovine brain. This complex was able to fragment actin filaments. Effects of this complex on the cytoskeleton of mouse and quail embryo fibroblasts are described. The cells were extracted with Triton X-100, and the resulting cytoskeletons were treated with the complex. Tetramethylrhodaminylphalloin and actin antibody staining failed to detect actin in preparations treated with the 90 kD protein-actin complex. Electrophoretic data confirmed actin solubilization upon this treatment. Immunofluorescent microscopy showed that actin solubilization was accompanied by extraction of vinculin and alpha-actinin from focal contacts and stress-fibers. In contrast, myosin distribution in treated cytoskeletons did not differ significantly from that in control extracted cells: in both the cases myosin was found mainly in the stress-fibers. Thus, myosin localization in stress-fibers does not depend on actin and is probably controlled by some other cytoskeletal component(s).     

The epidermal growth factor receptor is associated with actin filaments.

In this paper we describe our investigations on the association of receptors for the epidermal growth factor (EGF) with the cytoskeleton of A431 cells. In order to determine which filamentous system the EGF receptors are associated to, the cytoskeletal fraction to which these receptors bind was isolated. Second, the possible colocalization of EGF receptors with different cytoskeletal elements was examined in A431 cells. By selective extractions of the A431 cytoskeletons, it is shown that more than 90% of the cytoskeleton-associated EGF receptors are removed from the cytoskeletons together with the actin filamentous system. During several cycles of poly- and depolymerization of actin isolated from A431 cells, the EGF receptor precipitates together with the actin containing filaments, indicating that EGF receptors are able to bind in vitro to actin filaments. With immunofluorescence studies we show that EGF receptors especially colocalize with actin filaments. These results demonstrate that the EGF receptor is associated specifically with actin filaments in A431 cells.     

Changes in the levels of human tropomyosins IEF 52,55, and 56 do not correlate with the loss of actin cables observed in SV 40 transformed MRC-5 fibroblasts

Summary A mouse monoclonal antibody (mAb 1D122G9) raised against human tropomyosin IEF 52 (HeLa protein catalogue number, Mr=35 kd) has been characterized both in terms of specificity and patterns of immunofluorescence staining in Triton extracted cultured cells. As determined by two dimensional gel immunoblotting of HeLa cell proteins the antibody recognized IEF 52 and two other acidic proteins (IEF 55, Mr=31.8 kd; IEF 56, Mr=31 kd) previously identified as putative tropomyosin-like proteins. Immunofluorescence staining of Triton extracted cultured cells revealed the striated or interrupted pattern on the actin cables characteristic of tropomyosin staining. Quantitation of the three tropomyosins in Triton cytoskeletons from normal and SV 40 transformed human MRC-5 fibroblasts showed that the latter contained significantly less of tropomyosin IEF's 52 (52%) and 56 (72%) as compared to their normal counterparts. The ratios of these two tropomyosins to actin however was very similar for both types of cytoskeletons. This was not the case for tropomyosin IEF 55, which was present in nearly twice the amount in the cytoskeletons from the SV 40 transformed cells. The ratio of actin to total tropomyosin for whole cells was found to be unchanged on transformation. This ratio however was 31% lower in the cytoskeletons from the transformed cells. These and other results presented here suggest that changes in the levels of these three tropomyosins are not enough to account for the magnitude of the loss of actin cables observed in the transformed cells.Abbreviations IEF isoelectric focusing - mAb monoclonal antibody - NEPHGE non equilibrium pH gradient electrophoresis     

Identification of a membrane skeleton in platelets

Platelets have previously been shown to contain actin filaments that are linked, through actin-binding protein, to the glycoprotein (GP) Ib-IX complex, GP Ia, GP IIa, and an unidentified GP of Mr 250,000 on the plasma membrane. The objective of the present study was to use a morphological approach to examine the distribution of these membrane-bound filaments within platelets. Preliminary experiments showed that the Triton X-100 lysis buffers used previously to solubilize platelets completely disrupt the three-dimensional organization of the cytoskeletons. Conditions were established that minimized these postlysis changes. The cytoskeletons remained as platelet-shaped structures. These structures consisted of a network of long actin filaments and a more amorphous layer that outlined the periphery. When Ca2+ was present, the long actin filaments were lost but the amorphous layer at the periphery remained; conditions were established in which this amorphous layer retained the outline of the platelet from which it originated. Immunocytochemical experiments showed that the GP Ib-IX complex and actin-binding protein were associated with the amorphous layer. Analysis of the amorphous material on SDS-polyacrylamide gels showed that it contained actin, actin-binding protein, and all actin-bound GP Ib-IX. Although actin filaments could not be visualized in thin section, the actin presumably was in a filamentous form because it was solubilized by DNase I and bound phalloidin. These studies show that platelets contain a membrane skeleton and suggest that it is distinct from the network of cytoplasmic actin filaments. This membrane skeleton exists as a submembranous lining that, by analogy to the erythrocyte membrane skeleton, may stabilize the plasma membrane and contribute to determining its shape.