Reversible translocation of phosphoinositide 3-kinase to the cytoskeleton of ADP-aggregated human platelets occurs independently of Rho A and without synthesis of phosphatidylinositol (3,4)-bisphosphate.

The aim of our study was to evaluate the effect of ADP and the role of cytoskeleton reorganization during reversible and irreversible platelet aggregation induced by ADP and thrombin, respectively, on the heterodimeric (p85α-p110) phosphoinositide 3-kinase translocation to the cytoskeleton and its activation. Reversible ADP-induced aggregation was accompanied by a reversible reorganization of the cytoskeleton and an increase in levels of the regulatory subunit p85α in this cytoskeleton similar to the increase observed in thrombin-activated platelets. This translocation followed a course parallel to the amplitude of aggregation. No increase in levels of both phosphatidylinositol (3,4)-bisphosphate (PtdIns(3,4)P2) and phosphatidylinositol-(3,4,5)P3 could, however, be detected even at the maximum aggregation and PI 3-kinase α translocation. Moreover, in contrast to the situation for thrombin stimulation, the GTP-binding protein RhoA was hardly translocated to the cytoskeleton when platelets were stimulated with ADP, whereas translocation of pp60c-src and focal adhesion kinase did occur. These results suggest (i) translocation of signaling enzymes does not necessarily imply their activation, (ii) the reversibility of ADP-induced platelet aggregation may be the cause or the result of a lack of PI 3-kinase activation and hence of PtdIns(3,4)P2 production, and (iii) RhoA does not seem to be involved in the ADP activation pathway of platelets. Whether PtdIns(3,4)P2 or RhoA may contribute to the stabilization of platelet aggregates remains to be established.

Phosphoinositide 3-kinase (PI 3-kinases) 1 (1) are enzymes involved in growth factor signal transduction through association with receptor and nonreceptor tyrosine kinases and with G-protein-coupled receptors such as the fMet-Leu-Phe receptor in neutrophils or the thrombin receptor in platelets (1)(2)(3). Phos-phoinositide kinases and their products have been implicated in the reorganization of the cytoskeleton, and PI 3-kinase is known to be directly involved in platelet-derived growth factor, insulin-like growth factor-1, and insulin-induced membrane ruffling (4,5). Blood platelets also undergo morphological changes in response to stimulation, in particular shape change, extension of pseudopods, secretion of granule contents, aggregation, and contraction, all of which are linked to cytoskeletal modifications. Thrombin activation of human platelets leads to cytoskeletal translocation of the heterodimeric (p85␣-p110) PI 3-kinase (PI 3-kinase ␣) and accumulation of PtdIns(3,4)P 2 in an integrin ␣ IIb ␤ 3 -dependent manner (6,7).
Integrins are transmembrane heterodimers mediating cellmatrix and cell-cell interactions. The platelet ␣ IIb ␤ 3 integrin serves as an activation-dependent receptor for the adhesive proteins fibrinogen, fibronectin, and von Willebrand factor. In patients with Glanzmann's thrombasthenia (8), an inherited hemorrhagic disorder where the ␣ IIb ␤ 3 integrin is absent, reduced, or abnormal, platelets are unable to bind fibrinogen upon activation and consequently do not aggregate. In platelets, as in other cells, integrin ligation triggers the assembly of specific cytoskeletal proteins and enzymes into structures termed focal adhesion sites (9,10). These focal adhesion structures comprise proteins such as vinculin, talin, and the integrin ␣ IIb ␤ 3 itself, enzymes such as PI 3-kinase ␣, phospholipase C, protein kinase C, the tyrosine kinases pp60 c-src , pp72 SYK , and focal adhesion kinase (FAK) or the small GTP-binding proteins, RhoA and Cdc42Hs (11)(12)(13)(14)(15)(16). Rho, Rac, and Cdc42Hs are all members of the Ras superfamily of small GTP-binding proteins. These molecules are important regulators of the cytoskeleton, and evidence is now accumulating that Rho promotes the formation of focal adhesions and their anchoring to stress fibers (17)(18)(19).
Depending on the cell type studied, the heterodimeric PI 3-kinase ␣ has been found to be activated by several pathways including interactions of the p85␣ regulatory subunit with phosphorylated receptor tyrosine kinases, tyrosine kinases of the src family (20), p21 ras (21,22), RhoA (23,24), Cdc42Hs (25), or FAK (16). In platelets stimulated by thrombin, both RhoA and FAK could participate to the activation of PI 3-kinase ␣, the former presumably by an indirect mechanism (24) and the latter by a direct interaction with the SH 3 domain of the p85␣ subunit (16). In addition, heterotrimeric G-protein ␤␥ complexes may be involved in the stimulation of a second isoform of PI 3-kinase present in platelets. This form is immunologically related to a recently cloned monomeric PI 3-kinase, which was designated PI 3-kinase ␥ and found to be activated in vitro by * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  ␤␥ subunits (24,26,27).
A close relationship between integrin ␣ IIb ␤ 3 -dependent p85␣ translocation to the cytoskeleton and PtdIns(3,4)P 2 accumulation has been demonstrated in thrombin-stimulated platelets (16), suggestive of a role of the heterodimeric PI 3-kinase at this stage of 3-phosphoinositide synthesis. However, true assessment of the direct action of thrombin on platelets is difficult due to the presence of several mediators, in particular ADP and serotonin, released from the platelet-dense granules and of mediators such as thromboxane A 2 resulting from activation of the arachidonic pathway, all of which interact with their own specific receptors on the platelet membrane.
Platelet aggregation by ADP plays a key role in the development and extension of arterial thrombosis (28). Specific inhibitors of the ADP activation pathway such as the anti-aggregatory thienopyridine compounds ticlopidine and clopidogrel (29) markedly prolong the bleeding time and are used clinically as antithrombotic drugs. Furthermore, a rare congenital bleeding disorder with impairment of ADP-induced platelet aggregation (30,31) strikingly resembles the acquired thrombopathy resulting from ticlopidine or clopidogrel intake (32). Contained at very high concentrations in the platelet-dense granules, ADP is released when platelets are stimulated by other aggregating agents such as thrombin or collagen and thus contributes to and reinforces platelet aggregation. Low concentrations of ADP also potentiate or amplify the effects of all other agents, even weak agonists such as epinephrine or serotonin. Addition of ADP to washed human platelet suspensions results in shape change, exposure of the fibrinogen binding site on the ␣ IIb ␤ 3 integrin, and in contrast to other agonists such as thrombin, reversible aggregation in the presence of fibrinogen and physiological concentrations of Ca 2ϩ . At the intracellular level, platelet activation following ADP binding to its receptor leads to a transient rise in free cytoplasmic Ca 2ϩ , resulting from both Ca 2ϩ influx and mobilization of internal stores, without apparent activation of phospholipase C or D-myo-inositol 1,4,5trisphosphate (33,34). ADP also inhibits stimulated adenylyl cyclase (35). On the basis of its agonist selectivity and signaling properties, the platelet receptor for ADP has been classified as a P 2T receptor of the P 2 purinoceptor family (36). Although the biochemical structure of this receptor remains unknown, it may belong to the seven transmembrane domain G-proteincoupled receptor family since ADP has been found to activate the G i2 subtype of the heterotrimeric G-protein family (37,38).
The aim of the present study was to evaluate the direct effect of ADP on PI 3-kinase activation and to compare PtdIns(3,4)P 2 accumulation during reversible and irreversible aggregation. ADP was found to induce a reversible modification of the cytoskeleton which paralleled aggregation. The regulatory subunit p85␣ of PI 3-kinase ␣ and FAK reversibly translocated to the cytoskeleton, and this effect was dependent on the presence of ␣ IIb ␤ 3 and on the binding of fibrinogen to its receptor. However, significant accumulation of PtdIns(3,4)P 2 did not occur, indicating that although translocation of the heterodimeric PI 3-kinase ␣ occurred, activation did not. Moreover, in contrast to thrombin stimulation, the small GTP-binding protein RhoA was not significantly translocated to the cytoskeleton when platelets were stimulated with ADP, adding further support to a functional relationship between RhoA and PI 3-kinase.

EXPERIMENTAL PROCEDURES
Materials-The rabbit anti-p85␣ antibody was from Upstate Biotechnology Inc. (Lake Placid, NY), and rabbit anti-FAK and anti-RhoA antibodies were from Tebu (Santa Cruz Biotechnology Inc., Santa Cruz, CA), and an affinity purified sheep polyclonal antibody against pp60 c-src was from Cambridge Research Biochemicals Inc. (Cambridge, UK).
Preparation of Washed Human Platelets-Human blood was collected from a forearm vein (6 blood volumes into 1 volume of acid/ citrate/dextrose anticoagulant), and twice-washed platelet suspensions were prepared as described previously (39). In some experiments, platelets were labeled with sodium [ 32 P]orthophosphate (200 Ci/ml) for 1 h at 37°C during a first washing step in Tyrode's buffer containing no phosphate. The final resuspending medium, pH 7.35, was Tyrode's buffer containing 2 mM Ca 2ϩ , 1 mM Mg 2ϩ , 0.35% human serum albumin (Etablissement de Transfusion Sanguine, Strasbourg, France), and apyrase (2 g/ml, a concentration which converted 0.25 M ATP to AMP within 2 min at 37°C). Platelets were stored at 37°C throughout experiments, and cell count was adjusted in the final suspension to 7.5 ϫ 10 5 /l using a Sysmex 100 particle counter (Merck Clevenot, Nogent-sur-Marne, France).
Platelet Aggregation Studies-Aggregation was measured at 37°C by a turbidimetric method in a dual-channel Payton aggregometer (Payton Associates, Scarborough, Ontario, Canada). A 1.45-ml aliquot of nonlabeled or 32 P-labeled platelet suspension was stirred at 1,100 rpm and activated by addition of ADP in the absence or presence of human fibrinogen (0.8 mg/ml), or of thrombin in the absence of fibrinogen. The extent of aggregation was estimated quantitatively by measuring the maximum curve height above base-line level. At predetermined times, the reaction was stopped by addition of 1 ml of chloroform/methanol (v/v) for lipid extraction and analysis or by addition of an equal volume of CSK buffer (50 mM Tris-HCl, pH 7.4, 10 mM EGTA, 1 mM Na 3 VO 4 , 4 g/ml aprotinin, 4 g/ml leupeptin, 100 g/ml phenylmethylsulfonyl fluoride, 2% (v/v) Triton X-100).
Lipid Extraction and Analysis-Lipids were extracted and analyzed as described previously (2).
Cytoskeleton Extraction-Unlabeled platelets, activated or nonactivated, were mixed with 1 volume of CSK buffer and incubated successively for 5 min at room temperature and for 10 min at 4°C under shaking. Cytoskeletal material was collected by centrifugation (12,000 ϫ g, 10 min, 4°C), washed once with 2 volumes of CSK buffer containing 1% (v/v) Triton X-100, and then washed five times with 2 volumes of CSK buffer containing no Triton.

RESULTS
Reversible Modification of the Cytoskeleton-Washed human platelets were stimulated with 10 M ADP in the presence of fibrinogen or with 1 unit/ml thrombin in the absence of added fibrinogen. Typical aggregation curves were obtained (Fig. 1A), ADP inducing reversible and thrombin irreversible aggregation. In some experiments, the reaction was stopped at predetermined time points, the cytoskeleton was extracted and cytoskeletal proteins were solubilized, separated by SDS-PAGE, and stained with Coomassie Blue. Small amounts of actin binding protein (250 kDa), ␣-actinin (100 kDa), and F-actin (45 kDa) were found in the cytoskeleton of resting platelets. After stimulation with ADP, actin binding protein, myosin (200 kDa), and actin were reversibly translocated to the cytoskeleton, maximum incorporation corresponding to the maximum amplitude of platelet aggregation (Fig. 1B, left panel). Myosin was only weakly incorporated into the cytoskeleton of ADP-stimulated platelets. When platelets were stimulated with thrombin, translocation of actin binding protein, myosin, and actin to the cytoskeleton was not reversible (Fig. 1B, right panel). Actin polymerization induced by ADP was reversible and followed a course parallel to the amplitude of aggregation (Fig. 1C).
Reversible Translocation of the Regulatory p85␣ Subunit of PI 3-Kinase ␣ without [ 32 P]PtdIns(3,4)P 2 Accumulation during Platelet Aggregation-Using a polyclonal antibody, we found an increase in amounts of the regulatory subunit p85␣ in the cytoskeleton during both ADP-and thrombin-induced platelet aggregation ( Fig. 2A). In the case of ADP-stimulated platelets, translocation was reversible and followed the amplitude of aggregation. However, 20 and 40 s after stimulation, levels of p85␣ in the cytoskeleton were identical using either ADP or thrombin, and PI 3-kinase activity measured in the cytoskele-ton followed the same time course (not shown). In order to measure [ 32 P]PtdIns(3,4)P 2 accumulation during platelet aggregation, 32 P-labeled washed human platelets were stimulated with 10 M ADP in the absence or presence of fibrinogen or with 1 unit/ml thrombin in the absence of added fibrinogen. Thrombin stimulation gave rise to the expected time-dependent accumulation of [ 32 P]PtdIns(3,4)P 2 (Fig. 2B, right panel), whereas ADP did not induce any significant synthesis of [ 32 P]PtdIns(3,4)P 2 (Fig. 2B, left panel). In 5 of 7 experiments, where 32 P incorporation into lipids was especially high, we could detect only transient trace amounts of radioactivity in both PtdIns(3,4)P 2 and PtdIns(3,4,5)P 3 at short times (20 s) following ADP stimulation. Addition of fibrinogen did not increase this labeling. Similar results were obtained using 100 M ADP (data not shown).
p85␣ Translocation Requires Integrin ␣ IIb ␤ 3 and Fibrinogen Binding-ADP-induced translocation of PI 3-kinase ␣ to the cytoskeleton was dependent on the presence of the integrin ␣ IIb ␤ 3 , since this effect was not detectable using platelets from a type I Glanzmann's thrombasthenia patient (40) (Fig. 3, right  panels). Translocation was also dependent on the binding of fibrinogen to its receptor and was clearly reduced when fibrinogen was omitted (Fig. 3, left panels). The residual translocation observed is probably due to secreted fibrinogen.
Translocation of pp60 c-src , FAK, and RhoA to the Cytoskeleton-Washed human platelets were stimulated with 10 M ADP in the presence of fibrinogen or with 1 unit/ml thrombin in the absence of added fibrinogen; the cytoskeleton was extracted at indicated time points and analyzed by Western blotting. The tyrosine kinases pp60 c-src and FAK were translocated to the cytoskeleton in a similar manner to the PI 3-kinase regulatory subunit p85␣ (Fig. 4). In contrast, although thrombin induced clearly detectable translocation of RhoA to the cytoskeleton (Fig. 4, right panel), ADP did not (Fig. 4, left panel), and we could distinguish only a faint band at 20 s. DISCUSSION As is now well established (7,16), thrombin activation of washed human platelets resulted in translocation of the heterodimeric PI 3-kinase ␣ to the actin-rich cytoskeleton, together with production and accumulation of PtdIns(3,4)P 2 (Fig.  2B, right panel). These effects of thrombin are dependent on the presence of functional ␣ IIb ␤ 3 integrin (6,41). In general terms, coordinated signaling through agonist receptors and integrins results in reorganization of the cytoskeleton and formation of focal adhesion structures with translocation and activation of signaling proteins and enzymes (9). The aim of our study was to assess the specific role of ADP in these events and to investigate PI 3-kinase ␣ activation during reversible aggregation. So far, the molecular mechanisms leading to platelet aggregation by ADP and its typical feature of reversibility are not well understood. Our results showed an increase in levels of the regulatory subunit p85␣ in the cytoskeleton of ADP-stimulated platelets equivalent to the increase observed in thrombin-stimulated platelets ( Fig. 2A). This translocation was reversible and followed a course parallel to the amplitude of aggregation. Furthermore, this effect of ADP was dependent on the presence of the integrin ␣ IIb ␤ 3 and on the binding of fibrin-FIG. 1. Differential aggregation and cytoskeletal reorganization induced by ADP and thrombin. A, platelet aggregation was induced by ADP (10 M) in the presence of fibrinogen or by thrombin (1 unit/ml) in the absence of added fibrinogen and followed as described under "Experimental Procedures." Curves are representative of five independent experiments giving very similar results. B, in parallel, cytoskeletons were isolated from ADP-(10 M ϩ fibrinogen) or thrombin (1 unit/ml) -stimulated platelets at the indicated times. Cytoskeletal proteins (corresponding to 7.5 ϫ 10 7 platelets) were separated by SDS-PAGE (7.5%) and revealed by Coomassie Blue staining, actin, and the major actin-binding proteins being identified on the right side of the gel. Data are representative of two independent experiments giving very similar results. C, the F-actin content of ADP-or thrombin-stimulated platelets was quantified by densitometric analysis (ScanMaker IIHR, Microtek, Germany) of the Coomassie Blue-stained gel.

FIG. 2. Reversible translocation of p85␣ to the cytoskeleton without PtdIns(3,4)P 2 accumulation in ADP-stimulated platelets as compared with time-dependent PtdIns(3,4)P 2 accumulation after thrombin stimulation.
A, cytoskeletons were extracted from ADP-(10 M ϩ fibrinogen) or thrombin (1 unit/ml) -stimulated platelets at the indicated times. Proteins of the cytoskeleton (corresponding to 7.5 ϫ 10 7 platelets) were separated by SDS-PAGE (7.5%), blotted onto nitrocellulose, and tested for reactivity with an anti-p85␣ antibody using enhanced chemiluminescence. B, the time course of PtdIns(3,4)P 2 accumulation in washed platelets stimulated by ADP (10 M ϩ fibrinogen) or thrombin (1 unit/ml) was followed as indicated under "Experimental Procedures." ogen to its receptor (Fig. 3), clearly indicating an "outside-in" signaling event involving the "ADP-activated" integrin. Reversibility of the association of transduction proteins with the cytoskeleton of platelets activated by a thrombin receptor agonist peptide (TRAP) has already been reported to occur 15 min after stimulation in a Ca 2ϩ -dependent but aggregation-independent manner (42). Our results demonstrate the rapidly reversible association of these proteins in a manner differing according to the agonist used and the aggregation response.
Surprisingly, PtdIns(3,4)P 2 did not accumulate even at the maximum amplitude of aggregation, although PI 3-kinase ␣ was translocated to the cytoskeleton in amounts comparable with those found in thrombin-aggregated platelets. This result demonstrates that translocation of the enzyme is an aggregationdependent event but is not sufficient for activation of PI 3-kinase. At least three pathways of activation of PI 3-kinases have been reported in platelets, involving the small GTP-binding protein RhoA (23,24) and the tyrosine kinase FAK (16) for the (p85␣-p110) enzyme, or the ␤␥ subunit complex of heterotrimeric G-proteins for the p110 PI 3-kinase ␥ (24,26,27). In the case of ADP-induced platelet aggregation, we found pp60 c-src and FAK to be reversibly translocated to the cytoskeleton in a manner similar to p85␣, whereas RhoA was not. These observations deserve several comments. (i) G i2 proteins involved in the ADP signaling pathway do not seem to provide PI 3-kinase activating ␤␥ subunits which would otherwise have activated such an enzyme. (ii) The clear-cut difference between ADP and thrombin in inducing translocation of RhoA suggests that PI 3-kinase ␣ could be regulated by this small G-protein as previously reported (23,24). Interestingly, using lysophosphatidic acid as an agonist of a G-protein-coupled receptor, evidence has been provided that Rho-dependent assembly of an actin-based signaling complex linked to integrins was stimulated downstream of G q (43,44). In contrast to thrombin, ADP activates only G i2 with no effect on G q (38). Our data would fit with this scheme, underlying a possible role of RhoA in regulating PI 3-kinase activity and stabilization of the cytoskeleton. (iii) One cannot exclude a direct role of FAK, since its translocation to the cytoskeleton does not necessarily imply activation. This point could be clarified by looking at tyrosine phosphorylation of FAK during ADP-induced platelet aggregation. Nevertheless, the ADP scavenger apyrase has been shown to prevent tyrosine phosphorylation of FAK and the spreading of platelets on immobilized fibrinogen, which suggests that ADP in fact induces this phosphorylation (45).
Inhibition of PI 3-kinase by wortmannin or LY294002 has been reported to reverse the platelet aggregation induced by agonists such as TRAP (46). These authors suggested that PI 3-kinase activation may be necessary for prolonged ␣ IIb ␤ 3 activation and irreversible platelet aggregation. However, using wortmannin up to 100 nM, we were unable to reverse the irreversible aggregation induced by thrombin, even though the aggregates were smaller (data not shown). Aggregation was also found to be necessary for the late accumulation of PtdIns(3,4)P 2 measured 3-5 min after thrombin addition (16). Furthermore, previous studies have clearly established that thrombin-induced translocation to the cytoskeleton of several signaling proteins including ␣ IIb ␤ 3 , pp60 c-src , PI 3-kinase ␣ (12,16), as well as FAK tyrosine phosphorylation (47) require platelet to platelet contacts. Hence, the current data suggest that irreversible aggregation is necessary to initiate a "mechanical" transduction pathway leading to PI 3-kinase ␣ activation, whereas reversible aggregation would appear to be insufficient. Thrombospondin, a large trimeric adhesive molecule released from the ␣ granules of thrombin-but not ADP-stimulated platelets, could contribute to such a late signaling event by binding to the plasma membrane through its receptor CD 36, which has been shown to be linked to tyrosine kinases of the src family (48). On the other hand, when platelets were stimulated with lysophosphatidic acid (49) or concanavalin A (50), aggregation did not appear to be necessary for the synthesis of PtdIns(3,4)P 2 . Identification of the PI 3-kinase isoforms involved in these processes would contribute to a better understanding of the different pathways and stages of 3-phosphoinositide synthesis in platelets. Nevertheless, the physiological significance of the synthesis of PtdIns(3,4)P 2 depending on irreversible platelet aggregation remains to be established. The microvesiculation and clot retraction occurring at this stage of platelet activation are controlled by mechanisms involving reorganization of the membrane and cytoskeleton, which could be regulated by 3-phosphoinositide synthesis.
Alternatively, the results presented here are also consistent with the hypothesis that the typical reversibility of ADP-induced platelet aggregation finds its origin in the lack of PI FIG. 3. Integrin ␣ IIb ␤ 3 and fibrinogen binding dependence of p85␣ translocation. Control and type I Glanzmann's thrombasthenia (GT) platelets were stimulated with ADP (10 M) in the presence or absence of added fibrinogen. At the indicated time points, the reaction was stopped and translocation of p85␣ was evaluated as in Fig. 2.

FIG. 4. Reversible translocation of pp60 c-src and FAK but not
RhoA to the cytoskeleton of ADP (10 M ؉ fibrinogen) -stimulated platelets, as compared with irreversible translocation of all three proteins after thrombin stimulation. Platelets were stimulated with ADP (10 M ϩ fibrinogen) or thrombin (1 unit/ml) for increasing periods as indicated on the figure. Reactions were then stopped; the cytoskeletons were immediately extracted, and cytoskeletal proteins (corresponding to 7.5 ϫ 10 7 platelets) were separated by SDS-PAGE (7.5% for p125 FAK and pp60 c-src or 12% for RhoA), blotted onto nitrocellulose, and examined for reactivity with the indicated antibodies. Alkaline phosphatase detection was used for pp60 c-src analysis and the enhanced chemiluminescence system for p125 FAK and RhoA.
3-kinase activation and hence of PtdIns(3,4)P 2 formation. Recent data on the effects of wortmannin on platelet aggregation induced by TRAP (46) may indicate that PI 3-kinase products are necessary to stabilize platelet aggregates. Moreover, it has been demonstrated that PtdIns(3,4,5)P 3 is able to bind to the SH2 domains of several proteins including p85␣ (51), thereby blocking the binding of PI 3-kinase to tyrosine phosphorylated proteins, which would suggest direct involvement of these D3phosphorylated phosphoinositides. Whether PtdIns(3,4)P 2 , among other components of assembled signaling complexes, contributes directly to the stability of platelet aggregates is not known. The differential effects of ADP and thrombin on platelet activation and aggregation may thus provide a physiological model leading to the improvement of our understanding of PI 3-kinase pathways, at least in platelets.