The platelet cytoskeleton stabilizes the interaction between alphaIIbbeta3 and its ligand and induces selective movements of ligand-occupied integrin.

Previously, we showed that a subpopulation of the major platelet integrin, alphaIIbbeta3, co-sediments from detergent lysates with talin and other membrane skeleton proteins. Once alphaIIbbeta3 has bound adhesive ligand in a platelet aggregate, the detergent-insoluble alphaIIbbeta3 redistributes (along with the detergent-insoluble membrane skeleton proteins and a variety of signaling molecules) to a fraction that contains cytoplasmic actin filaments. Concomitantly, certain signaling molecules are activated. The present study shows that, in intact platelets, alphaIIbbeta3 forms clusters when occupied by ligand and is selectively moved into the open canalicular system; alphaIIbbeta3 that has not bound ligand remains diffusely distributed at the periphery of the cell. When cytoplasmic actin filaments are depolymerized by cytochalasins, the ability of alphaIIbbeta3 to bind ligand is decreased, and the movement of ligand-occupied alphaIIbbeta3 is prevented. Together with the previous findings, these results suggest that (i) membrane skeleton-associated alphaIIbbeta3 is selectively induced to bind ligand in activated platelets, (ii) ligand-induced transmembrane signaling causes an altered association of membrane skeleton-associated alphaIIbbeta3 with the cytoplasmic component of the cytoskeleton, (iii) ligand-induced cytoskeletal reorganizations stabilize the interaction between ligand and integrin, and (iv) ligand-occupancy triggers cytoskeletal reorganizations that result in selective movements of occupied ligand.

Integrins are a family of transmembrane glycoproteins that bind adhesive molecules and play important roles in mediating cell-cell and cell-matrix interactions (1). Several integrins are unable to bind their ligand unless the cells are activated (2)(3)(4)(5)(6). One such integrin is the glycoprotein IIb-IIIa complex (␣ IIb ␤ 3 ) on platelets (7). When platelets are activated, unidentified intracellular events act on ␣ IIb ␤ 3 to induce binding of fibrinogen to the extracellular domain of the receptor; by cross-linking ␣ IIb ␤ 3 molecules on adjacent platelets, fibrinogen is thought to mediate the formation of platelet aggregates. Although binding of the adhesive ligand is initially reversible, it becomes irreversible after several minutes (8,9). The binding of adhesive ligand to ␣ IIb ␤ 3 in a platelet aggregate induces a number of intracellular events, including phosphorylation of specific proteins on tyrosine residues (10 -12), activation of calpain (13), the calpain-induced hydrolysis of cytoskeletal proteins (13)(14)(15)(16), and activation of Na ϩ /H ϩ exchange (17). How ␣ IIb ␤ 3 is induced to bind adhesive ligand, how binding is rendered irreversible, or how binding of adhesive ligand to the extracellular domain of ␣ IIb ␤ 3 activates intracellular signaling molecules are all unanswered questions.
The cytoplasmic face of the plasma membrane of platelets is coated by a membrane skeleton that associates with the cytoplasmic domains of transmembrane glycoproteins and with underlying cytoplasmic actin filaments (18 -20). When platelets are lysed with Triton X-100, the cytoplasmic actin filaments can be sedimented by low speed centrifugation. The membrane skeleton separates from the cytoplasmic filaments; the fragments of membrane skeleton and associated membrane glycoproteins require higher g forces to be sedimented (19). Recently, we showed that some ␣ IIb ␤ 3 sedimented with fragments of the membrane skeleton from lysates of unstimulated platelets (21). Tyrosine kinases (pp60 c-src and pp62 c-yes ) were also recovered in this detergent-insoluble fraction (21). Once ␣ IIb ␤ 3 had bound to fibrinogen in a platelet aggregate, ␣ IIb ␤ 3 along with membrane skeleton proteins (21), tyrosine kinases (21), tyrosine-phosphorylated proteins (21), and a variety of additional signaling molecules (e.g. phosphoinositide 3-kinase and protein kinase C) (22) were recovered in the detergentinsoluble fraction that contained the cytoplasmic component of the cytoskeleton. We suggested that the membrane skeleton may play a role in regulating the activation of ␣ IIb ␤ 3 , that binding of ligand to membrane skeleton-associated ␣ IIb ␤ 3 causes the integrin-skeleton complexes to undergo an altered association with cytoplasmic actin forming "focal contact-like" structures, and that these focal contact-like structures may in turn play an important role in positioning signaling molecules and in regulating the ability of such molecules to mediate integrin-induced signaling events (21,23).
Our previous study (21) was based on the co-sedimentation of the integrin with detergent-insoluble fractions. In the present study, we provide evidence that membrane skeleton-associated ␣ IIb ␤ 3 binds ligand and is incorporated into complexes with cytoplasmic actin filaments in intact cells, that the formation of these integrin-cytoskeletal complexes results in stabilization of integrin-ligand interactions and a selective movement of ligand-occupied integrin into the surface-connected open canalicular system.

MATERIALS AND METHODS
Antibody Production and Characterization-PAC-1 and A5G8 are both monoclonal antibodies of the immunoglobulin (Ig) 1 M class that recognize epitopes on ␣ IIb ␤ 3 . PAC-1 is similar to fibrinogen in that it binds only to activated platelets (4). A5G8 binds to both unstimulated and activated platelets and does not inhibit fibrinogen binding. 2 Polyclonal and monoclonal antibodies against ␣ IIb ␤ 3 were provided by Dr. David Phillips of COR Therapeutics (South San Francisco, CA), polyclonal antibodies against fibrinogen were from Calbiochem (San Diego, CA) and those against glycoprotein Ib ␣ were raised as described previously (18).
Isolation and Analysis of Platelet Suspensions-The venous blood of healthy adult donors was collected into acid citrate/dextrose solution, and platelet-rich plasma was prepared (19). Platelets were isolated by gel filtration (24) or by centrifugation (19) and resuspended at a concentration of 1 ϫ 10 8 to 1 ϫ 10 9 platelets/ml in Tyrode's buffer (19). Platelets were activated with thrombin (a gift of Dr. John W. Fenton II of the New York Department of Health, Albany, NY) or ADP (Sigma). Unless otherwise stated, activation was induced without stirring. In some experiments, platelets were preincubated with cytochalasins (Sigma), which were added in a final concentration of 0.1% dimethyl sulfoxide (Me 2 SO). Control incubations also contained Me 2 SO. Platelet aggregation and secretion of ATP were assessed in a lumiaggregometer (Chrono-Log Corporation, Havertown, PA) (25). Western blotting was performed by the method of Towbin et al. (26). Antigen-antibody complexes were detected with 125 I-labeled anti-IgG (Dupont NEN).
Binding of PAC-1 to Platelets-Monoclonal antibody PAC-1 was labeled with fluorescein isothiocyanate (FITC) (24) and incubated with platelets in the presence of an agonist. At intervals, platelets were diluted into suspension buffer, and the amount of bound PAC-1 was determined by flow cytometry (24) (Becton Dickinson FACS 440 cytometer, San Jose, CA). For analysis of the reversible and irreversible components of PAC-1 binding, 45-l aliquots were added to 50 l of buffer or to the same buffer containing 100 mM EDTA; after 10 min, samples were diluted 3-fold, and the amount of bound PAC-1 was determined.
Localization of Proteins by Immunofluorescence-Platelets were activated in the presence of monoclonal antibody PAC-1 or A5G8. Incubations were terminated by addition of 9 volumes of a solution containing 4% paraformaldehyde in 150 mM sodium chloride, 10 mM Tris-HCl, pH 7.4, and platelets were allowed to settle onto poly-L-lysine-coated glass slides (19). Fixed platelets were permeabilized by addition of a buffer containing 0.5% Triton X-100, 0.1% Carnation milk, 150 mM ammonium acetate, 150 mM sodium chloride, 10 mM Tris-HCl, pH 7.4, then washed three times in a buffer that contained a 1:50 dilution of sheep serum (Sigma). Specimens were then incubated with biotinylated sheep anti-mouse IgM (Amersham Corp.) for 3 h, washed five times, incubated with Texas Red-labeled streptavidin (Amersham) for 30 -60 min, washed another five times, and mounted.
In dual-label experiments, platelets that had been activated in the presence of PAC-1, A5G8, or fibrinogen (a kind gift of Dr. Leslie Parise of the University of North Carolina, Chapel Hill, NC) were fixed and permeabilized with Triton X-100. Platelets that had been activated in the presence of PAC-1 or A5G8 were incubated for approximately 16 h with rabbit polyclonal antibodies against talin, or ␣ IIb ␤ 3 ; those that had been incubated with fibrinogen were incubated with polyclonal antibodies against fibrinogen and a monoclonal antibody against ␣ IIb ␤ 3 (that binds to a site other than that occupied by fibrinogen). Specimens were washed five times, incubated with biotinylated sheep anti-mouse IgM to label the monoclonal antibody, washed five times, and incubated with FITC-labeled streptavidin to detect monoclonal antibody and with Texas Red-labeled anti-rabbit IgG to detect polyclonal antibodies. Specimens were examined with a Zeiss Universal microscope or with a Bio-Rad MRC-600 confocal microscope (both equipped with dual fluorescence dichroic filters) and photographed.
Electron Microscopy of Activated Platelets-Platelets were activated with thrombin in the presence of PAC-1. After 15 min, suspensions were fixed with 4% paraformaldehyde in 0.1 M sodium cacodylate buffer, pH 7.4, and centrifuged. The pellets were dehydrated in a graded ethanol series, embedded in LR White resin, and 60 -90-nm thick sections collected on carbon-Formvar-coated nickel grids. Sections were incubated for 4 h with goat anti-mouse IgG and IgM conjugated to 5-nm diameter gold colloid (Janssen Life Sciences Products, Beerse, Belgium). The colloidal gold was silver-enhanced (27). The sections were stained with uranyl acetate and lead citrate and viewed in a Philips 301 transmission electron microscope.
Electron Microscopy of Platelet Cytoskeletons-Platelets were lysed by the addition of a Triton X-100 containing buffer and anti-glycoprotein Ib ␣ (13 g/ml) and Protein A coupled to 15-nm diameter gold colloid (200 l) (Janssen) were added (19). Lysates were incubated at 37°C for 60 min and then fixed in glutaraldehyde. Specimens were processed for electron microscopy and photographed in a JEM 100 CX II microscope (JEOL, Peabody, MA) (19).

RESULTS
Ligand-induced Association of ␣ IIb ␤ 3 with Cytoplasmic Actin in the Intact Platelet-Based on studies in detergent-lysed platelets, we have previously suggested that ␣ IIb ␤ 3 can associate with a submembranous component of the cytoskeleton and that binding of adhesive ligand to ␣ IIb ␤ 3 leads to altered association of the integrin and membrane skeleton proteins with cytoplasmic actin filaments. If this model is correct, then ␣ IIb ␤ 3 that had bound its adhesive ligand would be expected to show a different distribution within intact platelets from that of ␣ IIb ␤ 3 that had not bound ligand. Further, any ligand-induced change in distribution that occurred as a consequence of association of the integrin with cytoplasmic actin would be prevented by disrupting the network of cytoplasmic actin filaments.
To determine whether ligand-occupied ␣ IIb ␤ 3 had a different distribution from unoccupied integrin, platelets were incubated with the fibrinogen-mimetic monoclonal antibody, PAC-1. At intervals following thrombin addition, platelets were fixed, permeabilized with Triton X-100, and incubated with a polyclonal ␣ IIb ␤ 3 antibody that bound to both ligand-occupied and -unoccupied ␣ IIb ␤ 3 . The distribution of PAC-1 -occupied and total ␣ IIb ␤ 3 was visualized by dual-label confocal microscopy. As reported previously (4), virtually no PAC-1 bound to unstimulated, discoid platelets ( Fig. 1A) even though ␣ IIb ␤ 3 was over all of the surface of the platelets (Fig. 1B). As the platelets were activated with thrombin, PAC-1 binding occurred. At early times after platelet activation (e.g. 60 s), PAC-1-occupied ␣ IIb ␤ 3 was clustered in a few discrete areas (Fig. 1C). However, ␣ IIb ␤ 3 antibodies revealed that the rest of the ␣ IIb ␤ 3 was still present over all of the surface of the activated platelet ( Fig. 1D). At later times after platelet activation, clusters of PAC-1-occupied ␣ IIb ␤ 3 became concentrated toward the center of the platelet (Fig. 1E). Polyclonal antibodies revealed that ␣ IIb ␤ 3 that had not bound PAC-1 was still present in a relatively uniform distribution at the periphery of these platelets (Fig. 1F). The diameters of two representative platelet profiles (1 and 2) in Fig. 1, E and F, have been marked by pairs of arrows. When comparing, profiles 1 ϩ 2 (PAC-1 distribution) in Fig. 1E with profiles 1 ϩ 2 (␣ IIb ␤ 3 distribution) in Fig. 1F, it is obvious that the PAC-1 labeling pattern has a much smaller diameter than the ␣ IIb ␤ 3 labeling pattern in each case, small enough, in fact, to fit within the peripheral ␣ IIb ␤ 3 labeling pattern.
The platelets shown in Fig. 1 had been activated with thrombin at a platelet concentration of 1 ϫ 10 9 platelets/ml. A similar distribution of PAC-1-occupied ␣ IIb ␤ 3 was detected on platelets that were activated at a concentration of 1 ϫ 10 8 platelets/ml (data not shown). Another monoclonal antibody of the IgM class, A5G8, binds to a site on ␣ IIb ␤ 3 other than that to which adhesive ligand binds; it recognized ␣ IIb ␤ 3 on both unstimulated and activated platelets and showed a distribution similar to that detected with the polyclonal ␣ IIb ␤ 3 antibody (data not shown). Thus, the clustering and centralization of PAC-1-occupied ␣ IIb ␤ 3 presumably results from occupancy of the ligand binding site rather than from the binding of an IgM. Further evidence for this came from experiments in which platelets were incubated with the natural ligand for ␣ IIb ␤ 3 , fibrinogen. As with PAC-1, fibrinogen had a patchy distribution in activated platelets ( Fig. 2A). The simultaneous use of an ␣ IIb ␤ 3 antibody to reveal the outline of the platelet in a dual label experiment (Fig. 2B) showed that, like PAC-1, fibrinogen (shown in Fig. 2A) was concentrated toward the center of activated platelets.
To determine whether the clustering and centralization of ␣ IIb ␤ 3 was induced by cytoplasmic actin filaments, platelets were preincubated with cytochalasins. Cytochalasins inhibit the burst of actin polymerization that occurs when platelets are activated (28,29). Thin section electron microscopy ( Fig. 3) revealed that concentrations of cytochalasins higher than those needed to inhibit actin polymerization also disrupted the preexisting network of cytoplasmic actin filaments. Fig. 3A shows an untreated platelet with intact cytoplasmic actin filaments. Fig. 3B shows that the membrane skeleton remained intact after cytochalasin treatment but that the cytoplasmic actin network was disrupted. In addition, biochemical experiments revealed that, while the amount of actin sedimenting from detergent lysates at low g forces (i.e. networks of cytoplasmic actin) was markedly reduced in cytochalasin-treated cells, the amount of membrane skeleton proteins and ␣ IIb ␤ 3 that sedi-FIG. 1. Immunofluorescence confocal images showing the distribution of ligand-occupied ␣ IIb ␤ 3 (left column) and total ␣ IIb ␤ 3 (right column) in thrombin-stimulated platelets. Platelet suspensions were incubated with 40 g/ml PAC-1 for 2 min and then with 1.0 unit/ml thrombin for the indicated times. Incubations were terminated by the addition of paraformaldehyde. Platelets were permeabilized and incubated with polyclonal anti-␣ IIb ␤ 3 antibodies. The distribution of PAC-1 and polyclonal antibodies were detected by confocal microscopy as described under "Materials and Methods." The images in the righthand panels show the same platelets as are shown in the corresponding left-hand panels. In Panel F the outline of the platelets is revealed by labeling with ␣ IIb ␤ 3 antibodies; in Panel E, the location of PAC-1 in the same platelets is shown; comparison of the platelets indicated with arrows in the two panels shows that PAC-1 is concentrated toward the center of the platelets. mented at high g forces (i.e. in the membrane skeleton fraction) was not detectably altered (data not shown). To determine the effect of disruption of the network of cytoplasmic actin filaments on the ligand-induced redistribution of ␣ IIb ␤ 3 , cytochalasin-treated platelets were incubated with thrombin in the presence of PAC-1, and the distribution of PAC-1 was determined. Ligand-occupied ␣ IIb ␤ 3 still appeared to be present in clusters in the cytochalasin-treated cells (Fig. 4A). However, in contrast to the non-cytochalasin-treated cells (as seen in Figs.  1 and 2), the clusters of ligand-occupied integrin did not move inward but remained at the periphery of the cell (Fig. 4A). Fig.  4B shows the same field as seen in Fig. 4A, but now the platelet periphery is delineated by immunolabeled total ␣ IIb ␤ 3 . These results suggest that in non-cytochalasin-treated platelets, ligand-occupied ␣ IIb ␤ 3 associates with cytoplasmic actin and that this association leads to the inwards movement of the occupied integrin.
Localization of Ligand-occupied ␣ IIb ␤ 3 -The images shown in Figs. 1 and 2 were obtained when platelets were permeabilized with Triton X-100 prior to the addition of the secondary antibodies. Although the intensity of the fluorescence was often weaker, the same clustering and centralization of ligand-occupied ␣ IIb ␤ 3 was detected when the permeabilization step was omitted (Fig. 5) (control experiments using antibodies to actinbinding protein, a protein known to be present only on the cytoplasmic side of the membrane, confirmed that the platelets were not permeabilized inadvertently in these experiments) (data not shown). These results suggested that the membranes of the surface-connected open canalicular system were the most likely location of the centralized receptor since the immunolabeling antibody could have gained access via the open channels.
To determine if this was the case, thin-section electron microscopy of platelets that had been activated in the presence of PAC-1 and subsequently fixed, sectioned, and incubated with anti-mouse antibodies coupled to 5 nm of colloidal gold was performed. Because the PAC-1 was added prior to sectioning the amount of label in any one section was small. However, some PAC-1 was detected on the surface of the cell (where it would initially bind); any nonperipheral PAC-1 was present in the open canalicular system (Fig. 6, Panels A and B, arrows). When 125 gold particles were counted (representing staining of 127 platelets), 21 were present on the surface of the platelets, 93 in the open canalicular system, and 11 in intracellular areas. In platelets that had not been activated, a total of 13 gold particles were counted in 130 platelets; 3 particles on the surface, 3 in the open canalicular system, and 7 in the cytoplasm (data not shown). These data suggest that the cytoplasmic labeling represents background levels. The finding that ligand-occupied ␣ IIb ␤ 3 is present primarily in the open canalicular system, is consistent with the observation by others that fibrinogen is localized in the open canalicular system of activated platelets (30).

Regulation of Fibrinogen Binding by the Cytoskeleton-To determine whether the cytoplasmic actin filaments play a role FIG. 4. Immunofluorescence confocal images showing the distribution of ligand-occupied ␣ IIb ␤ 3 and total ␣ IIb ␤ 3 in cytochalasin-treated platelets. A platelet suspension was incubated with 10 Ϫ5
M cytochalasin E for 10 min. PAC-1 was then added to a concentration of 40 g/ml. After 2 min, 1.0 unit/ml thrombin was added, and incubation was continued for another 5 min. The platelets were then fixed by the addition of paraformaldehyde, permeabilized, and incubated with polyclonal ␣ IIb ␤ 3 antibodies. The distribution of PAC-1 and the polyclonal ␣ IIb ␤ 3 antibodies were detected by confocal microscopy as described under "Materials and Methods." Panel A shows the distribution of PAC-1, and Panel B shows polyclonal antibody-labeled ␣ IIb ␤ 3 in the same field of platelets as is shown in Panel A.

FIG. 5. Immunofluorescence images
showing the distribution of ligandoccupied ␣ IIb ␤ 3 in thrombin-treated platelets. Platelet suspensions were incubated with 20 g/ml PAC-1 for 2 min and then incubated alone or with 0.05 unit/ml thrombin for the indicated times. Incubations were terminated by the addition of paraformaldehyde and the distribution of PAC-1 detected with fluorescent secondary antibodies as described under "Materials and Methods." Panels A-C, platelets were exposed to Triton X-100 prior to the addition of the fluorescent secondary antibodies; Panels D-F, the platelets were not exposed to detergent. in regulating the binding of ligand to ␣ IIb ␤ 3 , platelets were incubated with cytochalasin under conditions in which the network of cytoplasmic actin filaments was disrupted and the subsequent ADP-induced binding of PAC-1 to platelets visualized by immunofluorescence (Fig. 7). The images in Panels A, B, and C were generated using similar exposures so that the relative amounts of PAC-1 binding to the platelets in cytochalasin-treated (Panel C) and -untreated cells (Panel B) could be visualized. Comparison of these images revealed that cytochalasin decreased the binding of PAC-1 to ADP-activated platelets. Similar results were obtained when platelets were activated with thrombin (data not shown).
The binding of PAC-1 was quantitated by flow cytometry. As shown in Fig. 8, cytochalasin inhibited the binding of PAC-1 to thrombin-activated platelets in a dose-dependent manner. Cytochalasin also inhibited binding of the natural ligand, fibrinogen, as shown by an inhibition of the thrombin-induced aggregation of a stirred platelet suspension (Fig. 9, top panel). This was a specific inhibitory effect, as shown by the lack of an inhibitory effect on the secretion of ATP (Fig. 9, bottom panel). The concentrations of cytochalasins required to inhibit aggregation ( Fig. 9) were comparable to those required to inhibit PAC-1 binding (Fig. 8).
The binding of ligand to ␣ IIb ␤ 3 initially occurs in a reversible manner (it can be reversed by addition of EDTA), but with time it becomes irreversible (8,9). To determine whether the cytoplasmic actin filaments are involved in regulating the reversible or irreversible component of binding, platelets were preincubated in the presence or absence of cytochalasins and activated with thrombin in the presence of PAC-1; after either 5 or 45 min, one aliquot was diluted into buffer while another was diluted into buffer containing EDTA. The irreversible component of binding was defined as the PAC-1 that remained bound to the platelets following dilution in EDTA, while the reversible binding was the difference between that which remained bound in EDTA and that which remained bound in buffer alone. As shown in Fig. 10, cytochalasin E inhibited the reversible binding of PAC-1 by approximately 50%. However, it almost completely inhibited the irreversible binding of PAC-1 to activated platelets. DISCUSSION The transmembrane signaling that occurs across integrins is of critical importance in a variety of events such as inflammation, embryonic development, arterial thrombosis, and hemo-stasis. The mechanisms involved in the two-way signaling across integrins are not well understood. In a previous study (21), we provided evidence that the integrin ␣ IIb ␤ 3 cosedimented with membrane skeleton proteins (e.g. spectrin, talin, and vinculin) from unstimulated platelets, that signaling molecules were associated with the membrane skeleton, and that binding of ligand to ␣ IIb ␤ 3 caused the membrane skeleton together with associated integrin and signaling molecules to redistribute to the low-speed detergent insoluble fraction. We suggested that ␣ IIb ␤ 3 was associated with the membrane skeleton, that binding of ligand to ␣ IIb ␤ 3 caused the membrane skeleton and associated integrin to become incorporated into complexes containing cytoplasmic actin filaments, and that the cytoskeleton might play a role in regulating the two-way signaling across ␣ IIb ␤ 3 . The evidence for these suggestions came entirely from co-sedimentation of proteins in detergent lysates.
The goal of the present study was to determine whether ␣ IIb ␤ 3 and the membrane skeleton become associated with cytoplasmic actin filaments in intact platelets and to determine the consequence of integrin-induced cytoskeletal reorganizations in intact cells. The studies suggest that membrane-skeleton associated ␣ IIb ␤ 3 binds ligand, clusters, and associates with cytoplasmic actin filaments. The resulting integrin-cytoskeletal complexes allow ␣ IIb ␤ 3 that has bound ligand to be selectively moved inward into the open canalicular system. Further, the integrin-cytoskeletal complexes play a role in sta- bilizing the integrin-ligand interaction. Taken together with our previous results in detergent lysates, these studies indicate that the cytoskeleton plays an important role in regulating the two-way signaling across ␣ IIb ␤ 3 in platelets.
Association of ␣ IIb ␤ 3 with the Cytoskeleton in Intact Cells-It has been known for many years that a subpopulation of ␣ IIb ␤ 3 cosediments with cytoplasmic actin filaments from detergent lysates of platelets that have aggregated (34,35). It has been assumed that this sedimentation results from an association of ␣ IIb ␤ 3 with cytoplasmic actin filaments. The present study shows that this is the case. Thus, a subpopulation of ␣ IIb ␤ 3 clustered and moved into the depths of the open canalicular system in activated platelets; this movement was induced by the cytoplasmic actin filaments because if these filaments were disrupted with cytochalasins, the movement was prevented.
In our previous study, we found that ␣ IIb ␤ 3 only redistributed to the low speed detergent-insoluble fraction in platelets in which ligand had bound to the integrin; moreover, ligand binding alone was not sufficient and the platelets needed to aggregate, presumably because ligand-induced cross-linking of the integrin was needed (36). In the present study, we used the fibrinogen mimetic monoclonal antibody PAC-1 (which is a pentameric molecule and therefore induces signaling comparable to that induced by fibrinogen in a platelet aggregate rather than that induced by fibrinogen in a nonaggregating suspension) (37). Dual-labeled immunofluorescence allowed us to show that only ␣ IIb ␤ 3 that had bound ligand moved inward. The rest of the ␣ IIb ␤ 3 remained at the periphery of the cell. We conclude, therefore, that the association of an ␣ IIb ␤ 3 molecule with cytoplasmic actin in aggregating platelets is a direct consequence of ligand binding to that molecule of integrin. One can envisage a physiological mechanism in which the selective movement of ligand-occupied integrin inward into the open canalicular system allows the externally bound fibrin clot to be pulled inward.
Others have used fibrinogen-coated gold beads or soluble fibringen to study movements of ␣ IIb ␤ 3 on adherent platelets (38 -49). The distribution of gold beads on platelets activated in suspension has also been studied (40, 41, 50 -52). These studies have revealed that ␣ IIb ␤ 3 that had bound gold beads moves inward. Although the physiological relevance of the movement of gold beads on the platelet surface has been questioned (53) and the functional state of integrins or signaling molecules on surface-activated platelets as compared to platelets activated with a physiological agonist is completely unknown, it is of interest that in all cases only the occupied receptor moves inward. The finding that receptor that has bound gold beads moves into the open canalicular system if it exists (40 -43, 45), that it co-localizes with cytoskeletal proteins (38), and that cytochalasins can inhibit some of the movements (42,44), suggests that, as with ␣ IIb ␤ 3 that has bound soluble ligand in platelet suspensions, the consequence of receptor occupancy is that the occupied receptor is selectively pulled by the cytoskeleton toward the most central regions of the platelet. Selective Association of Membrane Skeleton-associated ␣ IIb ␤ 3 with Cytoplasmic Actin-In our previous biochemical experiments (21), we noticed that in unstimulated platelets a subpopulation of ␣ IIb ␤ 3 co-sedimented with membrane skeleton proteins at high g forces (about 30% of the total platelet ␣ IIb ␤ 3 ). As platelets bound ligand, it was this population that initially redistributed to the low speed detergent-insoluble fraction. The present study supports the idea that only a subpopulation of ␣ IIb ␤ 3 is induced to bind ligand and to associate with cytoplasmic actin filaments in the intact cell. Taken together with the previous biochemical findings, this suggests that the ␣ IIb ␤ 3 that sediments with membrane skeleton proteins from unstimulated platelets represents a pool of integrin that can be selectively induced to bind ligand. The finding that the ␣ IIb ␤ 3 that bound ligand became clustered even when the networks of cytoplasmic filaments were depolymerized with cytochalasins (which did not disrupt the membrane skeleton) is also consistent with the possibility that it is the ␣ IIb ␤ 3 that is associated with membrane skeleton proteins that is selectively induced to bind ligand. Because the membrane skeleton is in close contact with the plasma membrane and associates with both signaling molecules (21) and ␣ IIb ␤ 3 (21), it appears possible that this structure serves to localize signaling molecules that are involved in the activation of the integrin. Similarly, the association of the integrin with this structure may be important in allowing the selective association of ligand-occupied integrin with underlying cytoplasmic actin filaments. Future work will be needed to find out how the integrin associates with submembranous skeletal proteins and to directly test the hypothesis that this association allows the activation and ligand-induced association of the integrin with cytoplasmic actin.
Immunofluorescence experiments have revealed that in cultured cells, binding of an integrin to its ligand in the extracellular matrix causes it to cluster and become incorporated into complexes of cytoskeletal proteins and signaling molecules known as focal contacts (54). The selective clustering of ligandoccupied ␣ IIb ␤ 3 and its association with cytoplasmic actin filaments in platelets is reminiscent of the formation of focal contacts. Several of the proteins that co-sediment with ␣ IIb ␤ 3 from detergent lysates of unstimulated platelets and redistribute with ␣ IIb ␤ 3 into the low-speed detergent-insoluble fraction from activated platelets are proteins that have been found to co-localize with integrins in focal contacts (e.g. vinculin, talin, and pp60 c-src ) (54). Additional components of focal contacts (e.g. protein kinase C) incorporate into the low speed pellet in aggregating platelets (22). Thus, the cytoskeletal reorganizations that are induced as a consequence of ligand binding to ␣ IIb ␤ 3 in platelets may be similar to those that form as a consequence of integrin-ligand interactions in cultured cells, and similar signaling mechanisms may be involved.
Regulation of Transmembrane Signaling-␣ IIb ␤ 3 can exist in a number of different affinity states, and several steps may be involved in the activation of the integrin and subsequent stabilization of ligand binding (55-57). As discussed above, the present immunofluorescence findings combined with the previous findings in detergent lysates (21) indicate that the membrane-skeleton associated ␣ IIb ␤ 3 is selectively induced to bind ligand. However, the finding that concentrations of cytochalasins that disrupted the cytoplasmic actin filaments inhibited ligand binding suggests that the cytoplasmic actin filaments also play an important role in regulating ligand binding to ␣ IIb ␤ 3 . Previously, variable effects of cytochalasins on fibrinogen binding (58,59), aggregation (58), and fibrinogen-gold distribution (40,41,44) have been reported. These variations may have resulted from the use of cytochalasins at concentrations that had variable effects on the cytoplasmic actin filaments. In the present study, we determined concentrations of cytochalasins that were needed not only to inhibit the increased polymerization of actin that occurs when platelets were activated (28,29) but also to disrupt the preexisting networks of cytoplasmic actin (60). These concentrations of cytochalasins almost totally prevented the conversion of the binding of fibrinogen from reversible to irreversible. In addition, they partially inhibited the reversible component of the binding. In cultured cells, the formation of focal contacts plays a role in stabilizing integrinligand interactions (54). Thus, in platelets, the formation of the "focal contact-like" integrin-cytoskeletal complexes may be important in rendering the binding of ligand irreversible. Associations between the cytoplasmic actin filaments and components of the membrane skeleton probably exist even in the unstimulated platelets (19,20). Thus, one possible mechanism by which the cytoplasmic actin filaments could exert their regulatory influence on the earlier, reversible stages of ligand binding might be their association with membrane skeleton FIG. 10. Effect of cytochalasin E on the reversible (A) and irreversible (B) binding of PAC-1 to thrombin-stimulated platelets. Suspensions of gel-filtered platelets were preincubated with 0.2% Me 2 SO or 10 Ϫ5 M cytochalasin E for 10 min. FITC-labeled PAC-1 was then added to a final concentration of 40 g/ml and thrombin to 0.1 unit/ml. After 5 or 45 min, 45-l aliquots of each incubation were added to 5 l of either buffer or 100 mM EDTA. The amount of FITC-labeled PAC-1 bound to platelets was analyzed by flow cytometry 10 min later. The amount of PAC-1 that was reversibly bound to platelets was determined as the amount that was displaced by EDTA (i.e. the difference between plus and minus EDTA samples). The amount of PAC-1 that was irreversibly bound was defined as that which remained bound to the platelets following the 10-min incubation in EDTA. The values shown represent the mean Ϯ S.D. obtained from three separate incubations.
proteins which in turn might associate with ␣ IIb ␤ 3 and signaling molecules.
Summary-Taken together with the results of our previous biochemical studies, the present studies indicate that the subpopulation of ␣ IIb ␤ 3 that is associated with the membrane skeleton is preferentially activated, that it subsequently becomes incorporated into complexes with cytoplasmic actin filaments, and that the formation of these integrin-rich cytoskeletal complexes plays a role in stabilizing the ligand-integrin interactions, inducing a selective redistribution of occupied integrin, and inducing activation of signaling molecules. Future studies will be needed to elucidate the molecular nature of the interactions between ␣ IIb ␤ 3 and the membrane skeleton and to identify the mechanisms by which integrin-cytoskeletal complexes regulate the post-occupancy events in activated platelets.