Tyrosine Phosphorylation of the β3 Cytoplasmic Domain Mediates Integrin-Cytoskeletal Interactions*

Tyrosine phosphorylation of the β3 subunit of the major platelet integrin αIIbβ3 has been shown to occur during thrombin-induced platelet aggregation (1). We now show that a wide variety of platelet stimuli induced β3 tyrosine phosphorylation, but that this phosphorylation occurred only following platelet aggregation. Several lines of evidence suggest that the β3 cytoplasmic domain tyrosine residues and/or their phosphorylation function to mediate interactions between β3 integrins and cytoskeletal proteins. First, phospho-β3 was retained preferentially in a Triton X-100 insoluble cytoskeletal fraction of thrombin-aggregated platelets. Second, in vitro experiments show that the cytoskeletal protein, myosin, associated in a phosphotyrosine-dependent manner with a diphosphorylated peptide corresponding to residues 740–762 of β3. Third, mutation of both tyrosines in the β3 cytoplasmic domain to phenylalanines markedly reduced β3-dependent fibrin clot retraction. Thus, our data indicate that platelet aggregation is both necessary and sufficient for β3 tyrosine phosphorylation, and this phosphorylation results in the physical linkage of αIIbβ3 to the cytoskeleton. We hypothesize that this linkage may involve direct binding of the phosphorylated integrin to the contractile protein myosin in order to mediate transmission of force to the fibrin clot during the process of clot retraction.

Integrins are a family of heterodimeric transmembrane proteins that link the extracellular matrix to the cytoskeletal/ contractile apparatus within a cell (2). These cytoskeletal linkages are characteristically induced by integrin clustering that can occur by the binding of multivalent or immobilized extracellular ligands, often resulting in the assembly of "focal contacts" in cultured cells. Several cytoskeletal proteins (e.g. talin, actin binding protein, ␣-actinin) directly bind to integrin cytoplasmic domains (3)(4)(5)(6), indicating that integrins may interact by multiple mechanisms for focal contact assembly. Focal contact assembly is often followed by signal transduction events such as induction of gene transcription (7) and prevention of apoptosis (8,9), regulating a diversity of cellular functions from embryonic development to hemostasis (10). Although cytoskel-etal and signaling proteins have been identified which bind integrin cytoplasmic domains, major unsolved questions persist. For example, what is the identity of the proteins responsible for the initial interactions of integrins with the cytoskeletal structures? How are these interactions regulated? In this regard, it is relevant to the present study that focal contacts are major sites for protein tyrosine phosphorylation, one of the earliest signaling events observed upon integrin ligation (11).
On platelets, the interaction of the integrin ␣ IIb ␤ 3 with its adhesive ligands, fibrinogen, or von Willebrand factor leads to platelet aggregation and association of ␣ IIb ␤ 3 with the cytoskeleton (12,13). Under normal conditions, platelet aggregation is the desired response to external trauma, allowing for hemostasis. However, inappropriate platelet aggregation does occur, as in ruptured artherosclerotic plaques, resulting in the formation of occlusive thrombi leading to myocardial infarction or thrombolytic stroke (14). The importance of ␣ IIb ␤ 3 -mediated events in both hemostasis and thrombosis is underscored in two ways. First, patients who lack, or have mutated, ␣ IIb ␤ 3 , a condition known as Glanzmann's thrombasthenia, have a bleeding disorder that arises from the failure of the platelets to aggregate (15). Second, clinical trials have shown that antagonists for ␣ IIb ␤ 3 ligand binding are effective antithrombotics (16).
␣ IIb ␤ 3 is involved in both "inside-out" and "outside-in" signaling pathways during platelet aggregation (12). In order to bind soluble forms of its adhesive protein ligands, ␣ IIb ␤ 3 on resting platelets has to undergo a conformational change. This process, the consequence of "inside-out" ␣ IIb ␤ 3 signaling, occurs when agonists such as ADP or thrombin activate platelets. Binding of fibrinogen and von Willebrand factor to ␣ IIb ␤ 3 induces platelet aggregation and ␣ IIb ␤ 3 clustering: the signals transduced by this process are referred to as "outside-in" signaling events. The cytoplasmic domains of the integrin are thought to play a critical role in these signaling events (17)(18)(19)(20). In addition, platelet aggregation induces the direct interaction of ␣ IIb ␤ 3 with the cytoskeleton (21,22). The cytoskeletal proteins talin and ␣-actinin have been found to act directly with ␣ IIb ␤ 3 (4,6). Along with ␣ IIb ␤ 3 , many other intracellular proteins, including Src and FAK, redistribute to the cytoskeleton of aggregated platelets (22,23). In these ways, the integrin may play a direct role not only in organizing the cytoskeleton but also in transducing signals to elicit cellular responses. Although it is clear that the cytoplasmic domains of ␣ IIb ␤ 3 are involved in signal transduction and cytoskeletal reorganization events, the precise mechanisms regulating these processes remain to be discovered.
Previously, we showed that tyrosine phosporylation of ␤ 3 occurs upon thrombin-induced platelet aggregation, indicating a potential role for integrin cytoplasmic tyrosine residues in outside-in ␣ IIb ␤ 3 signaling (1). In support of this hypothesis, we observed that the signaling adaptor proteins SHC and Grb2 interacted with peptides corresponding to the tyrosine phosphorylated cytoplasmic domain of ␤ 3 (1). The present study shows that tyrosine phosphorylation of ␤ 3 is a unifying event of platelet aggregation and provides in vitro evidence that tyrosine phosphorylation of this integrin subunit may direct its binding to myosin, a specific element contained within the platelet cytoskeleton.
Platelet Lysate Preparation-For two-dimensional gel analysis of detergent-soluble and cytoskeletal fractions, platelets were lysed immediately after aggregation by the addition of an equal volume of ice-cold 2ϫ Triton X-100 lysis buffer (1% (v/v) Triton X-100, 100 mM NaCl, 20 mM Tris, pH 7.0, 2 mM EDTA, 2 mM ethyleneglycol-bis(␤-aminoethyl ether)-N,N,NЈ,NЈ-tetraacetic acid, 20 g/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride, 200 M leupeptin, 4 mM sodium orthovanadate, 2 mM benzamidine, 0.27 mM diisopropyl fluorophosphate, 5 mM sodium pyrophosphate (final concentrations)). The lysate was then centrifuged for 6 min at 15,000 ϫ g to remove the Triton X-100 insoluble cytoskeletons formed during aggregation (22). The supernatant was reserved and 100 l of 2ϫ RIPA buffer (see below) was added to the pellet and sonicated for 20 min at room temperature in a Branson 5120 Sonicator to resolubilize the pellet. Nonreducing sample buffer (as described above) was added to each of the samples (supernatant and resolubilized pellet) and boiled for 5 min.
Determination of ␤ 3 Phosphorylation Level-Nonreduced-reduced two-dimensional gel electrophoresis was performed to visualize the characteristic migration of ␤ 3 and assess its phosphorylation state as described previously (1,29). The two-dimensional gels were transferred to nitrocellulose, and blots were probed with anti-phosphotyrosine antibodies PY-20 and 4G10. The blots were washed and incubated with horseradish peroxidase-conjugated sheep anti-mouse Ig and developed using the Enhanced Chemiluminescent (ECL) System. The level of phosphorylation was determined by densitometry using Imagequant software on a Molecular Dynamics densitometer. In each case, three to five different ECL exposures were subjected to densitometry analysis to reduce the risk of erroneous results from nonlinear signals. The blots were then stripped (according to ECL protocol, Amersham Pharmacia Biotech) and reprobed with the ␤ 3 antibody C3a.19.5 to determine ␤ 3 protein content for each sample. Phosphorylation results were normalized for the total amount of ␤ 3 protein present.
Preparation of Proteins-Myosin was purified from human platelets as described, and purification yielded myosin heavy chain and light chains (30). Controlled proteolytic digests of platelet myosin with papain or chymotrypsin were performed as described (31) except that myosin was not phosphorylated prior to chymotryptic digestion. Papain was activated according to the instructions of the manufacturer. Digests were run on 4 -20% SDS-PAGE and subjected to Coomassie Blue staining or transferred to nitrocellulose for ligand blotting.
Ligand Blot Analysis-Platelet lysates or purified myosin were reduced, separated by SDS-PAGE, and transferred to nitrocellulose. The blots were wet briefly in HEPES blot buffer (HBB) (25 mM HEPES, 25 mM NaCl, 5 mM MgCl 2 , 1 mM dithiothreitol) at 4°C. The transferred proteins were denatured by 6 M guanidine HCl in HBB for 10 min at 4°C and renatured by 2-fold dilution of guanidine HCl (10-min incubations each with 3 M, 1.5 M, 0.75 M, 0.38 M, 0.19 m, and 0 M guanidine HCl in HBB). The blot was blocked in HBB containing 4% bovine serum albumin overnight at 4°C and probed with 1 M biotinylated peptide in HBB containing 0.5% bovine serum albumin for 3 h at room temperature. After washing in Tris-buffered saline (20 mM Tris, 150 mM NaCl, pH 7.4)/ 0.01% Nonidet P-40 three times at 4°C, peptide-reactive bands were visualized by incubating the blots in horseradish peroxidaseconjugated streptavidin and employing ECL detection.
CHO Cell Generation and Flow Cytometry-␤ 3 (Y747F, Y759F) was generated and stably transfected into CHO cells as described (32). For flow cytometric analysis, cells were detached with trypsin, washed in Dulbecco's modified Eagle's medium ϩ 25 mM HEPES once and resuspended at 3 ϫ 10 6 cells/ml in FACS buffer (Hanks buffered saline solution with 3% heat-inactivated fetal bovine serum, 1% bovine serum albumin, 1% normal goat serum, 0.1% Gamimmune N, 0.03% sodium azide). The cells were then seeded at 200 l/well, pelleted, and incubated with 5 g/ml primary antibodies LM609 or control mouse IgG for 1 h at 4°C. After two washes, the cells were incubated with 1:200 goat anti-mouse fluorescein isothiocyanate-conjugated F(abЈ) 2 for 30 min at 4°C. The cells were washed and resuspended in FACS buffer, and the samples were analyzed by flow cytometry on a FACSort (Becton Dickinson).
Clot Retraction Assays-Clot retraction experiments were performed as described with minor modifications (33). In brief, cells were trypsinized, washed twice, and resuspended in Dulbecco's modified Eagle's medium ϩ 25 mM HEPES. 0.5 ml of cell suspension containing 5 ϫ 10 6 cells was mixed with 0.1 ml of fibronectin-depleted plasma in a 12 ϫ 70-mm glass tube treated with Sigmacote. Fibrin clots were formed by adding 1 unit/ml thrombin and allowed to retract at 37°C over a 2-3-h period. The extent of clot retraction was measured by removing and weighing the clot. 1 The abbreviations used are: PAGE, polyacrylamide gel electrophoresis; HBB, HEPES blot buffer; CHO, Chinese hamster ovary; FACS, fluorescence-activated cell sorter; ITAM, immune receptor tyrosine-based activation motif.

␤ 3 Tyrosine Phosphorylation Is a General Consequence of
Platelet Aggregation-We previously reported that aggregation of platelets by thrombin in a stirred suspension induced a marked increase in the tyrosine phosphorylation of the ␤ 3 subunit of ␣ IIb ␤ 3 (1). To determine whether this effect was thrombin-specific, we examined ␤ 3 tyrosine phosphorylation in response to various agonists. Adding ADP or ADP ϩ epinephrine to a stirred suspension of platelets in the presence of added fibrinogen induced platelet aggregation and an increase in ␤ 3 phosphorylation, similar to that seen in thrombin-aggregated platelets (Table I). In contrast, when ADP was added in the absence of added fibrinogen or was added without stirring, no platelet aggregation occurred and no increase in ␤ 3 tyrosine phosphorylation was seen. ADP added in this manner did, however, induce platelet stimulation since other substrates were tyrosine phosphorylated and fibrinogen binding on unstirred preparations occurred (data not shown). An illustration of the increase in ␤ 3 tyrosine phosphorylation upon thrombinor ADP-induced platelet aggregation is shown in Fig. 1A. Thus, ADP and ADP ϩ epinephrine induced tyrosine phosphorylation of ␤ 3 in an aggregation-dependent manner.
Threshold concentrations of ADP can result in reversible platelet aggregation (34,35). We next determined the effect of reversal of ADP-induced aggregation on ␤ 3 tyrosine phosphorylation. Washed platelets, resuspended in a buffer containing fibrinogen and Ca 2ϩ , were stirred with ADP. Platelet aggregation occurred but, as illustrated in Fig. 1C, reversed with time.
The reactions were terminated by the addition of SDS-sample buffer to the aggregometer tubes either at maximal ADP-induced aggregation or after aggregation was fully reversed (Fig.  1C). As described above, ADP-induced aggregation led to an increase in ␤ 3 phosphorylation, as well as the phosphorylation of a number of other substrates, similar to the response induced by thrombin (Fig. 1A). When samples were obtained in which ADP-induced platelet aggregation had reversed, no tyrosine phosphorylation of ␤ 3 was observed (Fig. 1A). In contrast, when thrombin-induced platelet aggregates were maintained in suspension for up to 10 min prior to addition of SDS-sample buffer, no reversal of ␤ 3 tyrosine phosphorylation was observed (data not shown).
Platelets aggregated in response to stimulation through the collagen or thromboxane A 2 receptors also showed a marked increase in ␤ 3 tyrosine phosphorylation (Table I). However, if platelet aggregation was prevented by cyclic RGD peptide (a competitive inhibitor of ␣ IIb ␤ 3 ), ␤ 3 phosphorylation was not observed (Table I). Together, these data indicate that many platelet agonists can induce aggregation-dependent tyrosine phosphorylation of ␤ 3 .
In the previous experiments, platelet agonists were used to induce fibrinogen binding and subsequent aggregation. To determine whether ␤ 3 tyrosine phosphorylation can occur in the absence of an agonist, platelet aggregation was induced by anti-LIBS6, a ␤ 3 -specific antibody that, upon binding to ␤ 3 , activates the receptor such that it becomes competent to bind fibrinogen. This activation occurs even in the absence of detectable platelet stimulation (27). The LIBS6 antibody induced smaller platelet aggregates, and aggregation took slightly longer than that induced by classical platelet agonists. Nevertheless, an average 4-fold increase in the level of ␤ 3 phosphorylation was still observed (Table I). Again, the addition of cyclic RGD peptide inhibited both platelet aggregation and ␤ 3 phosphorylation. Thus, ␣ IIb ␤ 3 -dependent platelet aggregation is both necessary and sufficient for tyrosine phosphorylation of ␤ 3 .
Tyrosine Phosphorylated ␤ 3 Preferentially Redistributes to the Cytoskeleton in Aggregated Platelets-To gain insight into the functional significance of tyrosine phosphorylation of ␤ 3 , we assessed its effects on the partitioning of the receptor with the cytoskeleton, a process known to occur upon platelet aggregation (21,22). Platelets were aggregated by the addition of thrombin and lysed with Triton X-100 lysis buffer. Under the solubilization conditions described under "Experimental Procedures," approximately 34% (p ϭ 0.002) of the total ␤ 3 protein associated with the cytoskeletal fraction upon aggregation, in agreement with earlier studies (21,22). Notably, previous work has established that only about 5% of the ␣ IIb ␤ 3 is found in the Triton X-100-insoluble residue of unstimulated platelets (21). Densitometry of anti-phosphotyrosine immunoblots indicated that approximately 72% of the tyrosine-phosphorylated ␤ 3 redistributes to the cytoskeletal fraction (Fig. 2). Thus, tyrosinephosphorylated ␤ 3 is more than twice as likely to become associated with the cytoskeleton (p ϭ 0.018), which may indicate a pivotal role for this ␤ 3 modification in linking a ligand-occupied receptor on the surface of aggregated platelets to the cytoskeletal/contractile apparatus within.
Doubly Phosphorylated Cytoplasmic Domain ␤ 3 Peptide Binds a 200-kDa Protein Identified as Platelet Myosin-The above observation, that tyrosine-phosphorylated ␤ 3 preferentially partitions with the cytoskeleton, highlighted the possibility that phosphorylation of ␤ 3 could be involved in mediating interactions between ␤ 3 and elements within the cytoskeleton. We have found, as have others (3,4,6), that the poor detergent solubility of many of the components of the cytoskeleton makes

tyrosine phosphorylation is induced by platelet aggregation
Washed platelets were prepared as described under "Experimental Procedures," except that calcium was omitted from the final buffer in the collagen sample and magnesium was omitted and 100 M aspirin and 0.33 units/ml apyrase were added to the final buffer in the LIBS6 samples. Platelets (approximately 5 ϫ 10 8 /ml) were stimulated at 37°C and allowed to stir in the aggregometer for 0.5-8 min with the following agonist concentrations: 0.1 units/ml thrombin, 1 M TRAP, 4 -16 M ADP, 2 M epinephrine, 0.2 mg/ml collagen, 5 l LIBS6 ascites, 1 M U44619. Fibrinogen was added to the indicated samples at a final concentration of 0.25-0.5 mg/ml. In the indicated samples, the cyclic RGD peptide (1 M) was added to inhibit aggregation. Nonreducing sample buffer containing vanadate was added immediately after aggregation to terminate the reaction. The level of ␤ 3 phosphorylation was assessed by densitometry after transferring nonreduced-reduced twodimensional gels to nitrocellulose and probing with anti-phosphotyrosine antibodies. The blots were stripped and reprobed with the anti-␤ 3 monoclonal antibody C3a.19.5 to determine protein content. The key to the -fold increase of ␤ 3 phosphorylation is as follows: 0, no increase in phosphorylation; ϩ, 2-6-fold increase; ϩϩ, 7-14-fold increase in phosphorylation as compared with control unstimulated platelets. Densitometry was performed on multiple exposures of the same blot, and similar results were obtained in at least two separate experiments. it technically difficult to observe interactions between cytoskeletal and other proteins in vivo. Thus, to address this issue, we employed an in vitro ligand blotting approach in an attempt to identify candidate proteins that, by binding to tyrosine-phosphorylated cytoplasmic domain of ␤ 3, could direct association of ␣ IIb ␤ 3 to the cytoskeleton. Proteins from platelet lysates were separated by SDS-PAGE, transferred to nitrocellulose and renatured on the blot. Synthetic peptides corresponding to the cytoplasmic domain of ␤ 3 , which contained biotin at their amino termini, were used to probe the nitrocellulose blot (Fig.  3A). The direct binding of peptide to renatured proteins was visualized by the addition of streptavidin-horseradish peroxidase and detected by chemiluminescence. As illustrated in

FIG. 2. ␤ 3 preferentially redistributes to the cytoskeletal fraction following thrombin-induced platelet aggregation.
Lysates were prepared from platelets stimulated with 0.1 units/ml thrombin that were allowed to aggregate for 8 min at 37°C, and Triton X-100insoluble platelet cytoskeletons were isolated as described under "Experimental Procedures." The supernatant fraction or the insoluble pellet (which was aggressively solubilized) from thrombin-aggregated platelets were subjected to two-dimensional gel analysis as described. This graph depicts the average of the ratio of ␤ 3 phosphorylation (with basal phosphorylation subtracted)/total amount of ␤ 3 present in samples as determined by densitometry of five separate experiments. A Student's t test (1-tailed) yielded a p value of 0.018. Triton X-100 insoluble cytoskeletal fraction of thrombin-aggregated platelets (data not shown) and appeared, therefore, to be an integral component of the platelet cytoskeleton. Together, its cytoskeletal distribution and apparent molecular mass suggested that the 200-kDa protein was the heavy chain of platelet myosin. To test this hypothesis, myosin was purified from platelets and examined in the ligand blot assay. Myosin heavy chain displayed the same peptide binding specificity as the 200-kDa protein in platelet lysates in that it bound the diphosphorylated ␤ 3 peptide but not the nonphosphorylated ␤ 3 peptide, the diphosphorylated S752P peptide or the diphosphorylated ITAM peptide ( Fig. 4B and data not shown). A similar peptide-binding specificity was also observed in solid phase assays, in which platelet myosin was coated on plates and binding of the various biotinylated peptides were detected enzymatically with horseradish peroxidase-conjugated streptavidin (data not shown). Phenylphosphate, a compound known to compete for phosphotyrosine binding sites (38), was used to further establish that the ␤ 3 phosphotyrosines were required for the interaction of ␤ 3 with myosin. 10 mM phenylphosphate inhibited completely the binding of the diphosphorylated ␤ 3 peptide to purified myosin in renatured blots (Fig. 4C), indicating that the ␤ 3 -myosin interaction was phosphotyrosine-dependent under these binding conditions. To further demonstrate the specificity of binding and to rule out nonspecific charge effects, a ␤ 3 peptide with tyrosines 747 and 759 replaced by glutamates also failed to bind purified myosin (data not shown). Taken together, these observations establish the identity of the 200-kDa protein as myosin heavy chain and demonstrate that its direct interaction with the ␤ 3 cytoplasmic domain peptide is phosphotyrosine-dependent.
Further ligand blotting experiments were performed using fragments of myosin generated by controlled proteolytic cleavage (31). Cleavage of myosin with papain yields single-headed soluble subfragment-1 (S-1) and an insoluble coiled-coil rod fragment, whereas cleavage with chymotrypsin results in double-headed heavy meromyosin and coiled-coil light meromyosin (31). We found that the doubly phosphorylated ␤ 3 peptide bound to the rod portion of the papain digest and the light meromyosin fragment in chymotrypsin-digested myosin (not shown). Neither the unphosphorylated ␤ 3 peptide nor the diphosphorylated S752P ␤ 3 peptide bound to any of these myosin fragments. Since the chymotryptic light meromyosin fragment and the rod portion generated by papain cleavage both contain overlapping sequences within the tail region of myosin (31), the data indicate that this region is responsible for binding the diphosphorylated ␤ 3 cytoplasmic domain.
The Tyrosine Residues within the ␤ 3 Cytoplasmic Domain Are Important for ␤ 3 -dependent Fibrin Clot Retraction-The results reported above indicate that the phosphorylation of the ␤ 3 cytoplasmic tyrosine residues might influence integrin-cytoskeletal interactions. This finding predicts that mutating the tyrosine residues of ␤ 3 should affect cellular functions dependent upon integrin-cytoskeletal interactions. One such function is the ␤ 3 -dependent retraction of fibrin clots, where the integrin is believed to function as a transmembrane linkage between extracellular adhesion proteins and the cytoskeleton. Due to the difficulty of genetically manipulating platelets, we used a CHO cell system that has proven useful in the study of integrin function (19,39) to directly address this issue. It has previously been shown that CHO cells transfected with wild-type ␤ 3 will express the ␤ 3 on the cell surface in conjunction with endogenous ␣ v chains (19). In contrast to nontransfected CHO cells, the ␤ 3 -transfected CHO cells gain the ability to retract fibrin clots (Ref. 19, and data not shown). We generated stable CHO cell lines expressing either wild-type ␤ 3 or ␤ 3 bearing the con- servative Y747F and Y759F mutations. As illustrated in Fig.  5A, FACS analysis with the ␣ v ␤ 3 -specific antibody LM609, confirmed that these transfectants expressed similar levels of ␣ v ␤ 3 or ␣ v ␤ 3 (Y747F, Y759F) at the cell surface. When the two CHO cell lines were used in the fibrin clot retraction assay, it was found that clot retraction was reduced markedly in the Y747F, Y759F transfectants, as demonstrated by a 50 Ϯ 11.9% increase in clot weight compared with the clot weights obtained with the wild-type ␤ 3-expressing CHO cells (Fig. 5B).
To assess whether the presence of ␣ IIb would alter the effect of the double tyrosine to phenylalanine mutations on ␤ 3 -mediated fibrin clot retraction, similar experiments were performed with CHO cells that had been co-transfected with ␣ IIb in addition to the mutant or wild-type ␤ 3 cDNA. These cells expressed both ␣ v ␤ 3 and ␣ IIb ␤ 3 on their surface, and essentially the same results were obtained: the cells bearing Y747F and Y759F ␤ 3 showed about a 50% decrease in their ability to retract fibrin clots when compared with those cells expressing wild-type ␤ 3 (data not shown). Thus, the tyrosine residues within ␤ 3 do indeed play a critical role in ␤ 3 integrin-mediated fibrin clot retraction. DISCUSSION A well established function of integrin cytoplasmic domains is as a bridge between extracellular matrix proteins and the cytoskeletal/contractile machinery within a cell. The data presented in this study indicate that the two tyrosine residues within the ␤ 3 cytoplasmic domain may be important in mediating some of these interactions and suggest a way in which a modification of these residues, namely by phosphorylation, may also be involved in these bridging processes. We have found that the tyrosine phosphorylation of the ␤ 3 subunit of ␣ IIb ␤ 3 occurs as a general consequence of platelet aggregation. That this phosphorylation may in turn affect the interaction of the ␤ 3 integrin with the platelet cytoskeleton is indicated by the following data. First, phosphorylated ␤ 3 is located preferentially within the detergent-insoluble cytoskeletal fraction of aggregated platelets; and second, the contractile protein myosin can bind directly, in a phosphotyrosine-dependent manner, to peptide corresponding to the cytoplasmic domain of ␤ 3 . Furthermore, functional data obtained by analysis of CHO cells bearing a ␤ 3 in which both cytoplasmic tyrosine residues were mutated to phenylalanines also indicates the importance of these ␤ 3 tyrosines in the ␤ 3 -dependent retraction of fibrin clots. We have previously demonstrated the interaction of tyrosinephosphorylated ␤ 3 with the signaling proteins SHC and Grb2 (1) and hypothesized that tyrosine phosphorylation of ␤ 3 allowed for the recruitment of signaling complexes to the membrane. Our new studies indicate that in addition to the role of ␤ 3 tyrosine phosphorylation in binding signaling proteins, the phosphorylation may also influence the interaction of ␤ 3 with the myosin-based contractile apparatus and in doing so play an important role in integrin-dependent functions involving cytoskeletal reorganization.
The present data demonstrate that platelet aggregation is both necessary and sufficient to induce tyrosine phosphorylation of ␤ 3 . First, conditions that only induced the active conformation of ␣ IIb ␤ 3 and did not allow for aggregation to occur, such as ADP stimulation in the absence of fibrinogen or in the presence of an inhibitory RGD peptide, did not induce ␤ 3 tyrosine phosphorylation. Second, platelet aggregation induced by LIBS6 independent of platelet stimulation also resulted in ␤ 3 tyrosine phosphorylation. Third, ␤ 3 tyrosine phosphorylation did not occur under conditions that induced ligand occupancy of ␣ IIb ␤ 3 but not platelet aggregation, such as the addition of ADP and fibrinogen in the absence of stirring. Last, a reversal of aggregation-induced ␤ 3 tyrosine phosphorylation was observed upon reversal of platelet aggregation. In all instances, platelet aggregation, with the subsequent platelet-platelet interactions, was absolutely required for ␤ 3 tyrosine phosphorylation.
Early studies established that a significant portion of ␣ IIb and ␤ 3 could be isolated with cytoskeletal structures in thrombin-aggregated, but not activated, platelets (21), suggesting that platelet aggregation induces the association of ␣ IIb ␤ 3 with the platelet cytoskeleton. It was proposed that this integrin became Triton X-100 detergent-insoluble because of the macromolecular associations between the platelet membrane surfaces and actin filaments. Morphological studies have also demonstrated that fibrinogen binding to ␣ IIb ␤ 3 induced its interaction with the cytoskeleton since the membrane-bound integrin appeared to be co-aligned with cytoskeletal structures of the platelet (40). Further, Fox and co-workers demonstrated an aggregation-dependent redistribution of ␣ IIb ␤ 3 from the membrane skeleton to the Triton X-100-insoluble fraction of platelets (22). Several tyrosine kinases and other tyrosine-phosphorylated proteins also redistributed to the cytoskeleton upon platelet aggregation (22). The present data points to a possible mechanism for the redistribution of ␣ IIb ␤ 3 to the cytoskeleton in aggregated platelets. Examination of the phospho-␤ 3 distribution between Triton X-100 soluble and insoluble fractions of aggregated platelets demonstrated that phosphorylated ␤ 3 preferentially redistributes to the cytoskeletal fraction. The observations that tyrosine phosphorylation of the ␤ 3 cytoplasmic domain is a common consequence of aggregation by a wide range of platelet agonists and is a potential player in driving the redistribution of ␤ 3 to the cytoskeleton prompted us to examine biochemically whether ␤ 3 phosphorylation plays a part in mediating novel interactions of the receptor with the platelet cytoskeleton.
The binding of integrin cytoplasmic domains to cytoskeletal proteins is not unprecedented, and although little in vivo data exist, due to the technical difficulties of working with poorly soluble cytoskeletal proteins, a variety of in vitro strategies have been used to discover and characterize such interactions. Interactions between talin and ␤ 1 integrin cytoplasmic domain were first studied using equilibrium gel filtration of purified proteins (3). In solid phase binding assays, talin was found to bind directly with ␣ IIb ␤ 3 integrin cytoplasmic tail sequences and to purified ␣ IIb ␤ 3 (4). ␣-Actinin has been shown to bind directly to the cytoplasmic domain of ␤ 1 as well as to purified ␣ IIb ␤ 3 (6). In another study, the cytoskeletal protein skelemin interacted with the ␤ 3 cytoplasmic domain in a yeast twohybrid screen and with peptides corresponding to the membrane proximal regions of ␤ 1 and ␤ 3 (41). Actin binding protein has also been demonstrated to bind directly to the cytoplasmic domain of ␤ 2 integrin using peptide affinity chromatography (42) and to the dimerized ␤ 1 cytoplasmic domains using a novel experimental strategy (5). However, the mechanisms that regulate these integrin-cytoskeletal interactions are unknown.
Given that tyrosine phosphorylation of ␤ 3 is a general consequence of platelet aggregation and appears to direct the redistribution of ␣ IIb ␤ 3 to the cytoskeleton, we postulate that ␤ 3 tyrosine phosphorylation could be a general mechanism for regulating integrin-cytoskeletal interactions. Fittingly, members of the Src family of tyrosine kinases are known to selectively redistribute with a subpopulation of ␣ IIb ␤ 3 to the actin cytoskeleton in aggregated platelets (22,23). This redistribution is reduced by treatment of platelets with tyrosine kinase inhibitors, suggesting that tyrosine kinases, either directly or through the phosphorylation of other proteins, may regulate the cytoskeletal attachment of ␣ IIb ␤ 3 (43). Further, ␣ IIb ␤ 3 -mediated clot retraction is inhibited by tyrosine kinase inhibitors (43). Although there is circumstantial evidence that certain integrin-cytoskeletal interactions are phosphotyrosine-dependent, experimental data directly supporting this hypothesis are lacking.
In the present work, we used in vitro ligand binding methodology to observe a novel interaction between myosin and a ␤ 3 integrin cytoplasmic domain peptide that was regulated by tyrosine phosphorylation. Phosphorylated and nonphosphorylated integrin cytoplasmic domain peptides were synthesized and used to identify a direct and tyrosine phosphorylation-dependent interaction between the ␤ 3 cytoplasmic domain peptide and platelet myosin heavy chain. Since the doubly phosphorylated ␤ 3 peptide bound specifically to myosin, this contractile protein may possess tandem phosphotyrosine binding regions analagous to the tandem SH2 domains of the tyrosine kinases ZAP-70 or Syk, which bind ITAM domains in immune receptor complexes (44,45). However, to our knowledge, classic phosphotyrosine binding motifs in platelet myosin heavy chain have not yet been identified. We further observed that the tail domain of myosin was responsible for its interaction with ␤ 3 . Interestingly, the tail region of myosin serves as an anchor so that it can translocate actin and has been hypothesized to bind certain myosin isoforms to cell or organelle membranes (46). Together these data suggest that the tyrosine phosphorylated ␤ 3 binding domain of myosin exists on the tail region of myosin heavy chain and that this domain contains previously unrecognized phosphotyrosine binding motifs. These binding motifs may allow for the interaction of phosphorylated ␤ 3 with the cytoskeletons of aggregated platelets, providing alignment for certain postaggregation contractile events, such as clot retraction, to occur.
Although our in vitro data strongly suggest that tyrosine phosphorylation of the ␤ 3 cytoplasmic domain can allow for ␣ IIb ␤ 3 interaction with myosin, this conclusion was not possible to confirm in vivo because of the aforementioned problems associated with working with many cytoskeletal proteins. Indeed, myosin is insoluble at physiological salt concentrations; only highly stringent co-immunoprecipitation conditions could be employed using detergent lysates that are well known to disrupt protein-protein interactions. In this case, the problem is compounded by the fact that the major portion of tyrosine phosphorylated ␣ IIb ␤ 3 does itself translocate to the insoluble cytoskeleton in aggregated platelets. Also, robust tyrosine dephosphorylation mechanisms are present in platelets (47), which make it difficult to preserve tyrosine phosphorylation of ␤ 3 except under denaturing conditions (e.g. by the addition of SDS-containing sample buffer) or in rapid postlytic fractionations (e.g. cytoskeleton isolation) hampering immunoprecipitation experiments. Therefore, our data do not preclude other, possibly phosphotyrosine-independent, interactions between the cytoplasmic domains of ␣ IIb ␤ 3 and myosin. If other such interactions do indeed exist, it is attractive to hypothesize that tyrosine phosphorylation of ␤ 3 cytoplasmic domain, possibly at only one of the tyrosine residues, could induce a more stable and avid interaction between previously-associated ␣ IIb ␤ 3 and myosin.
Thus, our data suggest that tyrosine phosphorylation of the ␤ 3 cytoplasmic tail may regulate a direct association with myosin, providing anchorage of surface ␤ 3 integrins to the contractile apparatus. A possible consequence of this interaction is to allow for ␣ IIb ␤ 3 -mediated clot retraction in platelets. We addressed the role of the ␤ 3 cytoplasmic tyrosines in clot retraction using CHO cells transfected with ␤ 3 . Previous studies using such a CHO cell expression system have proven useful for analyzing the role of both ␣ IIb and ␤ 3 integrin cytoplasmic domains in ␣ IIb ␤ 3 signaling and adhesive functions (17)(18)(19)36). In particular, CHO cells transfected with ␣ IIb ␤ 3 gain the ability to contract fibrin clots, whereas both untransfected CHO cells and cells expressing the S752P Glanzmann's mutant ␣ IIb ␤ 3 fail to do so (36). Another expression system, in which CS-1 melanoma cells are transfected with a cDNA encoding the integrin ␤ 3 subunit, has been used to characterize the roles of the ␣ v ␤ 3 integrin cytoplasmic domains in adhesion, spreading, and migration on vitronectin (20). Although clot retraction was not addressed in this study, mutating either tyrosine 747 or 759 on ␤ 3 to phenylalanine had little or no effect on other ␣ v ␤ 3 adhesive events (20). In the present work, CHO cells bearing the Y747F and Y759F ␤ 3 cDNA displayed a pronounced defect in fibrin clot retraction: the first demonstration of an effect of ␤ 3 tyrosine to phenylalanine mutations on a biologically relevant event. Since integrins are believed to support clot retraction by providing the transmembrane linkage between extracellular adhesive proteins and the contractile cytoskeleton (19), it is interesting to hypothesize that, by mutating the tyrosine residues within the ␤ 3 cytoplasmic domain, we have disrupted the phosphotyrosine-dependent integrin-myosin interaction and that this could account for the defective clot retraction observed in the mutant CHO cell transfectants.
Our working hypothesis of the role of ␤ 3 cytoplasmic domain tyrosine phosphorylation in platelet function can be summarized as follows: the receptor is phosphorylated as a common consequence of aggregation in response to a number of platelet agonists. Although direct associations of known tyrosine kinases with ␣ IIb ␤ 3 have not yet been detected, members of the Src family of tyrosine kinases are capable of phosphorylating the receptor in vitro (1) and Src and Lyn can be cross-linked to ␤ 3 in intact platelets treated with chemical cross-linking agents (48). Once phosphorylated, the ␤ 3 integrin tails are capable of associating with signaling proteins SHC and Grb2 to potentially initiate outside-in signaling cascades (1). In addition to providing a scaffold for the recruitment of signaling complexes to the membrane, the doubly phosphorylated cytoplasmic domain of ␤ 3 can also bind to cytoskeletal proteins. In particular, the present work demonstrates direct binding of a doubly tyrosine-phosphorylated ␤ 3 integrin cytoplasmic domain peptide to myosin and further reveals that replacement of these tyrosine residues with phenylalanines in ␣ IIb ␤ 3 -transfected CHO cells results in defective ␤ 3 integrin-mediated retraction of fibrin clots. In light of these data, we postulate that phosphorylation of ␤ 3 integrin cytoplasmic domain may be an important mechanism for regulating a direct myosin-integrin interaction. Inhibition of this interaction may interfere with the transmission of the mechanical forces that regulate processes such as clot retraction and cell motility. Proving or disproving such hypotheses will be the focus of future work.