Non-redundant Roles of Phosphoinositide 3-Kinase Isoforms α and β in Glycoprotein VI-induced Platelet Signaling and Thrombus Formation*

Platelets are activated by adhesion to vascular collagen via the immunoglobulin receptor, glycoprotein VI (GPVI). This causes potent signaling toward activation of phospholipase Cγ2, which bears similarity to the signaling pathway evoked by T- and B-cell receptors. Phosphoinositide 3-kinase (PI3K) plays an important role in collagen-induced platelet activation, because this activity modulates the autocrine effects of secreted ADP. Here, we identified the PI3K isoforms directly downstream of GPVI in human and mouse platelets and determined their role in GPVI-dependent thrombus formation. The targeting of platelet PI3Kα or -β strongly and selectively suppressed GPVI-induced Ca2+ mobilization and inositol 1,4,5-triphosphate production, thus demonstrating enhancement of phospholipase Cγ2 by PI3Kα/β. That PI3Kα and -β have a non-redundant function in GPVI-induced platelet activation and thrombus formation was concluded from measurements of: (i) serine phosphorylation of Akt, (ii) dense granule secretion, (iii) intracellular Ca2+ increases and surface expression of phosphatidylserine under flow, and (iv) thrombus formation, under conditions where PI3Kα/β was blocked or p85α was deficient. In contrast, GPVI-induced platelet activation was insensitive to inhibition or deficiency of PI3Kδ or -γ. Furthermore, PI3Kα/β, but not PI3Kγ, contributed to GPVI-induced Rap1b activation and, surprisingly, also to Rap1b-independent platelet activation via GPVI. Together, these findings demonstrate that both PI3Kα and -β isoforms are required for full GPVI-dependent platelet Ca2+ signaling and thrombus formation, partly independently of Rap1b. This provides a new mechanistic explanation for the anti-thrombotic effect of PI3K inhibition and makes PI3Kα an interesting new target for anti-platelet therapy.

, whereas p110␦ has a more important role in adaptive immunity, e.g. in T and B cells (14). Human and mouse platelets contain four different PI3K isoforms, among which are the class IA catalytic subunits, p110␣, -␤, and -␦ (PI3K␣, -␤, and -␦), and the class IB catalytic subunit, p110␥ (PI3K␥) (15)(16)(17). For class IA, the corresponding regulatory subunits are p85␣/␤, p55␣/␥, and p50␣, whereas for class IB the regulatory subunit is p101␥. Structural studies in other cells have indicated that the regulatory class IA subunits, particularly p85␣, can interact with tyrosine kinase-linked receptors via the SH2 domains (18). In contrast, class IB isoforms may rather interact with G-proteincoupled receptors (16). This concept was recently challenged by the observation that, in platelets, both PI3K␤ and -␥ are activated via the P2Y 12 receptor for ADP, which is coupled to G i , and that both isoforms contribute to integrin ␣ IIb ␤ 3 activation and platelet aggregation (17, 19 -21). Hence, it is clear that PI3K isoforms can be activated by other platelet receptors than only GPVI.
To date, it is debated which of the PI3K isoforms become directly activated by GPVI signaling, and which are activated indirectly, e.g. following ADP receptor stimulation. Also unclear is which are the downstream events mediated by the various isoforms. Reasons for this lack of clarity include: (i) the large contribution to GPVI-induced responses of secondary, autocrine stimulators, particularly ADP and thromboxane (1,17,22); (ii) the proposed stimulation by PI3K and its product PI(3,4,5)P 3 to Ca 2ϩ entry rather than to PLC activity (23,24); (iii) the limited knowledge on the effector targets of PI3K and PI(3,4,5)P 3 in platelets, of which only phosphoinositidedependent kinase and Akt are well studied (21,25); (iv) the observation that p110␦ is not a major isoform implicated in GPVI-induced platelet activation (26); and (v) the limited availability of mice deficient in PI3K subunits.
One possible, poorly explored target of PI3K in platelets is the small GTPase Rap1b, which is highly expressed in these cells and is considered to play a key role in the activation process particularly toward ␣ IIb ␤ 3 activation. Platelet agonists such as ADP and thrombin produce the active GTP-bound form of Rap1b, partly in a PI3K-dependent manner (27,28). Recent findings suggest that PI3K␤ is the main isoform responsible for the ADP-induced activation of Rap1b and Akt, whereas PI3K␥ contributes to ␣ IIb ␤ 3 activation by a separate pathway (21). However, whether and how Rap1b is activated by GPVI is unresolved, as is the role of different PI3K isoforms herein.
In the present report, we studied the identity of the PI3K isoforms downstream of GPVI and determined their role in platelet activation and thrombus formation. To discriminate between direct and indirect GPVI-induced effects, the cells were stimulated in the presence of autocrine stimulation inhibitors, blocking the signaling contributions of both ADP and thromboxane. For the studies, we combined a pharmacological approach using a panel of isoform-specific PI3K inhibitors with experiments using mice deficient in distinct PI3K catalytic or regulatory subunits. We found that PI3K␣ and -␤ are key mediators of GPVI-induced thrombus formation and show that these isoforms affect the activity of both PLC␥2 and Rap1b.

EXPERIMENTAL PROCEDURES
Mouse Strains-Mice deficient in the p85␣ regulatory PI3K subunit had a C57BL/6 genetic background (11,29). Mice deficient in p110␥ or p110␦ (17) or deficient in Rap1b (30) were from sources previously described. All knock-out mice had normal platelet counts. Wild-type mice were used of the same background and same breeding program. Animal experiments were approved by the local animal experimental committees.
Blood Collection and Platelet Isolation-For platelet isolation, blood was collected into one-sixth volume of acid-citratedextrose anticoagulant (85 mM sodium citrate, 78 mM citric acid, and 11 mM D-glucose). Donors gave full informed consent according to the Helsinki declaration and had not taken medications for 2 weeks. Platelet-rich plasma (PRP) and washed platelets were obtained by centrifugation, as described for human (7) and mouse (32,33) platelets. Washed human platelets were resuspended in Hepes buffer, pH 7.45 (137 mM NaCl, 10 mM Hepes, 2.7 mM KCl, 2 mM MgSO 4 , 0.42 mM D-glucose, 0.2 unit/ml apyrase, and 0.1% bovine serum albumin, pH 7.45). Mouse platelets were resuspended in a modified Hepes buffer (33). Platelets were counted with a Coulter counter. For whole blood flow experiments, human blood was collected into 40 M PPACK (34), whereas mouse blood was collected into 40 M PPACK plus 5 units/ml heparin (32).
Intracellular Ca 2ϩ Measurement in Platelet Suspensions-Human PRP (2 ϫ 10 8 platelets/ml) was loaded with 2.5 M Fura-2 acetoxymethyl ester in the presence of aspirin (100 M) and apyrase (0.2 unit of ADPase/ml) at 37°C for 45 min (35). The loaded platelets were resuspended in Hepes buffer, pH 7.45 (1 ϫ 10 8 /ml), and were used within 90 min. Before addition of agonist, the cells were preincubated with apyrase (0.1 unit/ml) and ADP receptor blockers (40 M MRS-2179 for P2Y 1 and 10 M ARC-69931MX for P2Y 12 ) to inhibit autocrine stimulation. In some experiments, apyrase was added at a high concentration (0.6 unit/ml). Platelet preincubation with indicated PI3K inhibitors or Me 2 SO vehicle was for 10 min (37°C). Calcium responses were recorded under stirring with an SLM-Aminco or a Cairn Research spectrofluorometer, at alternate excitation wavelengths of 340 and 380 nm (37°C). The 340/380 nm ratio values were converted into nanomolar concentrations of [Ca 2ϩ ] i , as described (35). Separate calibrations were done for incubations containing colored PI3K inhibitors. Rises in [Ca 2ϩ ] i were expressed as 5-min time integrals, to quantify prolonged Ca 2ϩ -signaling effects (36).
Platelet Aggregation and Flow Cytometry-Human PRP was pretreated with aspirin, and platelets were washed. The washed platelets were preincubated for 10 min with autocrine stimulation inhibitors (see above) and PI3K blockers or Me 2 SO vehicle. After 10 min of activation, samples were analyzed by flow cytometry for P-selectin expression using fluorescein isothiocyanate-anti-CD62 monoclonal antibody (1:100) or for ␣ IIb ␤ 3 activation using fluorescein isothiocyanate-labeled PAC1 monoclonal antibody (37).
InsP Measurement-Accumulation of InsP 1 due to InsP 3 production was measured with an IP-One ELISA kit from Cis-Bio, determining InsP 1 levels. Aspirin-treated platelets (1 ϫ 10 8 /ml) were preincubated with ADP receptor blockers, LiCl (1 mM), and Me 2 SO vehicle or PI3K blocker. After stimulation for 5-10 min with convulxin or thrombin, the platelets were lysed. Cell lysates were incubated with InsP 1 -horseradish peroxidase conjugate and anti-InsP 1 monoclonal antibody, according to the manufacturer's instructions.
Activation of Akt and Rap1 by Western Blotting-Akt activation was measured by Western blot analysis of platelet lysates. A polyclonal anti-phosphoserine-473 Akt antibody was used to detect active Akt and a polyclonal anti-Akt antibody to determine total Akt (19). Activation of Rap1 was measured in platelet lysates by selective precipitation of GTP-bound Rap1 using glutathione S-transferase-RalGDS bound to glutathione-Sepharose (28). Western blotting was performed with anti-Rap1 antibody (250 ng/ml) and a sheep anti-mouse horseradish peroxidase-coupled secondary antibody (1:5000).
Thrombus Formation and Procoagulant Activity under Flow-Perfusion of PPACK-anti-coagulated human blood (34) and of PPACK/heparin-anti-coagulated mouse blood (32) was done in the absence of coagulation as described. Blood samples were preincubated with PI3K inhibitor or Me 2 SO vehicle for 10 min. Blood was flowed through a transparent, parallel-plate flow chamber, containing a collagen-coated coverslip, at a defined shear rate for 4 min. The thrombi in flow chambers were stained by rinse with Hepes buffer, pH 7.45, supplemented with 1 unit/ml heparin, 2 mM CaCl 2 , and OG488-annexin A5 (1 g/ml). Microscopic images were recorded in real-time using a Visitech imaging system, equipped with two intensified chargecoupled device cameras (34). Bright-field contrast and fluorescence digital images were taken from Ͼ10 randomly chosen fields. ImagePro software (Media Cybernetics) was used to quantify platelet deposition (38).
Intracellular Ca 2ϩ Measurement under Flow-Human PRP (2 ϫ 10 8 /ml) was incubated with 7 M Fluo-3 acetoxymethyl ester for 45 min at 20°C under gentle rotation. Washed mouse platelets (2 ϫ 10 8 /ml) were incubated with 5 M Fluo-3 acetoxymethyl ester plus 0.2 mg/ml Pluronic F-127. The dyeloaded cells were added to autologous blood (10% labeled platelets) and used for flow experiments within 1 h. Fluorescence images were recorded during high speed (5 Hz) perfusion of blood over a collagen surface (39). Off-line, regions of interest representing one adhered cell were analyzed for fluorescence changes. Raw fluorescence data were converted into F/F o values by pseudo-ratio analysis, and then into nanomolar concentrations of [Ca 2ϩ ] i (40). For quantification, traces from individual cells were superimposed, so that frame numbers of initial [Ca 2ϩ ] i increases coincided.
Statistics-Differences between groups were tested with a Mann-Whitney U test or analysis of variance. Effects of inhibitors were tested with a Student's paired t test. The statistical package for social sciences was used (SPSS 15.0).

Prominent Roles of PI3K␣ and -␤ Isoforms in GPVI-induced Ca 2ϩ
Responses-Although PI3K activity has been implicated in GPVI-mediated platelet activation, it is unclear whether its role is direct or indirect, e.g. via autocrine ADP release and P2Y 12 signaling. Another proposal is that PI3K may specifically stimulate GPVI-induced Ca 2ϩ entry rather than mobilization of Ca 2ϩ from internal stores (23). To investigate this further, human Fura-2-loaded platelets were activated with GPVI agonist under conditions where indirect effects of autocrine stimulators (thromboxane A 2 and ADP) were prevented with aspi- rin and a high concentration of ADP-degrading apyrase (0.6 unit/ml). When platelets were stimulated with CRP in the presence of extracellular EGTA (where Ca 2ϩ is only mobilized from internal stores), the general PI3K inhibitor wortmannin (0.1 M) caused ϳ60% suppression of the Ca 2ϩ signal (Fig. 1A). In the presence of extracellular CaCl 2 (where Ca 2ϩ entry takes place in addition), wortmannin had a similar inhibitory effect on the Ca 2ϩ signal. The treatment of platelets with another general PI3K inhibitor, LY-294002 (25 M), was as effective as wortmannin. Furthermore, similar effects were obtained with both inhibitors, when platelets were stimulated with the GPVI ligand, convulxin (Fig. 1B). To directly determine the inhibitor effect on Ca 2ϩ entry, platelets were first stimulated with GPVI ligand causing Ca 2ϩ mobilization, after which CaCl 2 was added and Ca 2ϩ entry was allowed. Strikingly, wortmannin suppressed both parts of the Ca 2ϩ signal at similar extent (Fig. 1C). A control experiment indicated that blocking anti-GPIb antibodies did not affect the convulxin-induced Ca 2ϩ response (not shown), thus excluding involvement of GPIb-induced Ca 2ϩ signal generation with this agonist.
Interestingly, treatment of the platelets with wortmannin or LY-294002 caused only small inhibition of the thrombinevoked Ca 2ϩ responses in both EGTA and CaCl 2 medium (Fig.   1B). Together, these data suggest that PI3K activity primarily enhances PLC␥2-evoked Ca 2ϩ mobilization from stores, whereas it only secondarily affects Ca 2ϩ entry.
Various new isoform-specific catalytic site inhibitors of class I PI3Ks have been developed and used for functional studies in other cell types (31,41). The selectivity of these inhibitors for PI3K␣ through ␦ isoforms is mostly determined in cell-free in vitro studies using isolated enzymes (IC 50 values are given in supplemental Table I). The compound PIK-75 has a far higher affinity for PI3K␣ (IC 50 8 nM) than for other isoforms. At nanomolar concentrations, the compound PI-103 inhibits both PI3K␣ and -␤ (PIK-112 is an inactive analog). YM-024 has a relatively high affinity for PI3K␣ and -␦, whereas TGX-211 is most active against PI3K␤ (IC 50 7 nM). The compounds AS-252424 and IC-87114 show highest activity toward PI3K␥ and PI3K␦ isoforms, respectively. It should be remarked that effective doses of these compounds for complete inhibition of the kinase activity in adipocytes, hepatoma cell lines, and platelets are reported to be in the low micromolar range (17,21,41). Hence, for adequate inhibition of PI3K activities in intact platelets, higher doses are needed than apparent from the IC 50 values. Reasons for this are the high platelet counts

PI3K Isoforms in Glycoprotein VI-induced Platelet Activation
in experiments and the high ATP concentration in cells (at least 10-fold higher than of in vitro kinase assays).
For the above-mentioned inhibitors, dose-effect relations were determined for convulxin-induced Ca 2ϩ responses in CaCl 2 medium. To prevent autocrine stimulatory effects of thromboxane and ADP, all experiments were carried out with aspirin-treated platelets, and blockers of both ADP receptors were added. Time integrals of increases in [Ca 2ϩ ] i served as a read-out of the extent of GPVI-induced PLC activation (36). Strikingly, those inhibitors with a high affinity toward PI3K␣ (PIK-75, PI-103, and YM-024, 0.5-1 M) reduced the Ca 2ϩ signal with ϳ50%, whereas the control substance PIK-112 was completely inactive (Fig. 2, A and B). In addition, the PI3K␤ selective inhibitor TGX-221 (0.1-2.5 M) caused a similar reduction in Ca 2ϩ signal. Combined application of PIK-75/ TGX-221 or PI-103/TGX-221 did not further suppress the response. In general, the isoform-specific PI3K␣ or -␤ inhibitors were similarly effective as the general PI3K inhibitors, wortmannin and LY-294002. On the other hand, AS-252424 (0.1-1 M), inhibiting PI3K␥, was without any effect, whereas the PI3K␦ inhibitor IC-87114 only slightly reduced the Ca 2ϩ response.
For comparison, the same panel of inhibitors was tested in measurements of thrombin-induced Ca 2ϩ responses. Here, only compounds with a high affinity toward PI3K␤ (i.e. PI-103 and TGX-221) and the general PI3K inhibitors (wortmannin and LY-294002) reduced the Ca 2ϩ signal by 15-20% (supplemental Fig. 1). In this case, the PI3K␣ inhibitor PIK-75 was without effect. Together, these results point to high and non-redundant roles of the PI3K␣ and -␤ isoforms in Ca 2ϩ signaling directly downstream of GPVI, but not of thrombin receptors.
Control experiments were performed to exclude that the high affinity PI3K␣ inhibitor, PIK-75, acted by residual inhibition of PI3K␤. We examined the effect of PIK-75 (0.5-1 M) on P2Y 12 -induced platelet responses and integrin ␣ IIb ␤ 3 activation, because these are exclusively mediated by the PI3K␤ and -␥ isoforms (20,21). As shown in supplemental Fig.  2, PIK-75 was completely inactive in both platelet aggregation and integrin activation in response to ADP, in marked contrast to the established PI3K␤ inhibitor, TGX-221. On the other hand, TGX-221 did not affect Akt phosphorylation at Ser 473 evoked by insulin-like growth factor-1, which is known to be mediated by only the PI3K␣ isoform (S. Kim, data not shown, but see Ref. 42).
Contribution of Both PI3K␣ and -␤ Isoforms to GPVI-induced Signaling-In platelets, the GPVI receptor activates the PHdomain containing PLC␥2, in contrast to thrombin receptors, which activate PLC␤ isoforms. For other cells, it was hypothesized that an increase in PI(3,4,5)P 3 level caused by PI3K pulls PLC␥2 to the membrane and increases its action (18). To verify this for GPVI-stimulated platelets (again with blocked autocrine responses), we determined overall PLC activity by measuring the accumulation of InsP 1 due to InsP 3 production. Platelet stimulation with convulxin led to a production of InsP 3 for up to 10 min (Fig. 3A). Wortmannin treatment reduced the InsP 3 formation and moreover confined it to 5 min. Such a shortening of InsP 3 production was not seen in thrombin-stimulated platelets. The isoform-specific inhibitors, PIK-75 or TGX-221 (targeting at PI3K␣ and -␤, respectively) suppressed the InsP 3 production to a similar extent as wortmannin, whereas the combination of the two did not have an additional effect (Fig. 3B). In contrast, inhibition of PI3K␥ (AS-252424) or PI3K␦ (IC-87114) was  DECEMBER 4, 2009 • VOLUME 284 • NUMBER 49 not effective on InsP 3 amounts. Hence, these results point to a GPVI-induced prolongation of PLC activity by PI3K␣/␤.
A key downstream effect of [Ca 2ϩ ] i elevation in platelets is the secretion of ␣-granules, which is measured as expression of granular P-selectin at the platelet surface. Platelet treatment with PIK-75 or TGX-221, alone or in combination, suppressed convulxin-or CRP-induced P-selectin expression with 40 -50%, i.e. to the same extent as did wortmannin or LY-294002 (Fig. 4). This points to a non-additive effect of both isoform-selective compounds on GPVI-induced secretion.

Roles of PI3K␣ and -␤ Isoforms in GPVI-induced Platelet Activation and Thrombus Formation on Collagen under Flow-
The pharmacological evidence so far points to a common role of the PI3K␣/␤ isoforms in GPVI-induced Ca 2ϩ signaling and downstream platelet activation events. We then performed studies to determine the importance of these isoforms in flowdependent thrombus formation at physiological shear rates. Earlier work has shown that, during platelet interaction with collagen under flow, GPVI-induced increases in [Ca 2ϩ ] i are important for granule secretion, activation of integrins, and exposure of PS (22,34,43,44). In whole blood, supplemented with Fluo-3-loaded platelets, [Ca 2ϩ ] i increases were measured of individual platelets upon adhesion to collagen under flow. Interestingly, pretreatment with either PIK-75, TGX-221, or wortmannin led to a marked suppression in single cell Ca 2ϩ responses, so that the adhered platelets showed low amplitude spiking changes in [Ca 2ϩ ] i (Fig. 5). Control experiments indicated that the Ca 2ϩ responses under flow were not influenced by aspirin and blockage of ADP receptors (not shown), thus confirming that autocrine ADP release did not contribute to this initial platelet response. Other controls indicated that these [Ca 2ϩ ] i increases were completely suppressed by blocking GPVI (22).
The effects of inhibitors on thrombus formation were assessed from measurements of the deposition of platelets on the collagen surface (Fig. 6A). In addition, the exposure of PS of collagen-adhered platelets (detected with OG-annexin A5) was measured, the response for which is shown by a subpopulation of the platelets in direct contact with collagen due to GPVI activation (38). Both parameters were substantially diminished with wortmannin, with the PI3K␣ inhibitors PIK-75 and YM-024, or with the PI3K␤ inhibitor TGX-221. In the presence of all inhibitors, platelet deposition on collagen was substantially reduced in comparison to the vehicle control (Fig. 6B). In addition, the number of PS-exposing platelets was reduced to ϳ50% by PIK-75, YM-024, or TGX-221. On the other hand, inhibition of PI3K␥ (AS-252424) or -␦ (IC-87114) was without effect. Together, these results suggest that inhibition of only PI3K␣ or -␤ isoforms suppresses flow-dependent thrombus formation on collagen, by reducing GPVI-dependent Ca 2ϩ responses and downstream activation processes leading to platelet aggregation and PS exposure.

Roles of Murine PI3K Isoforms in GPVI-induced Platelet Activation and Thrombus
Formation-Platelets from mice lacking the p85␣ regulatory PI3K subunit are impaired in collageninduced aggregation, which points to diminished GPVI signaling (11). In the p85␣ Ϫ/Ϫ platelets, expression of the p110␣ subunit (PI3K␣) is almost undetectable, while both p85␤ and p110␤ are still present at low levels (11,29,45). We used blood from p85␣ Ϫ/Ϫ mice to measure collagen-dependent [Ca 2ϩ ] i increases in adhered platelets, thrombus formation, and PS exposure under flow. In contrast to the potent and prolonged Ca 2ϩ responses of p85␣ ϩ/ϩ platelets, p85␣-deficient platelets showed a markedly lower Ca 2ϩ signal, which in 36% of the cells showed spiking Ca 2ϩ responses (Fig. 7A). Treatment of p85␣ ϩ/ϩ blood with TGX-221 reduced the average Ca 2ϩ signal to the level seen in p85␣ Ϫ/Ϫ platelets. Treatment of wild-type blood with YM-024 had a similar suppressive effect on the Ca 2ϩ response (p ϭ 0.012). As a comparison, blood was used from p110␥ Ϫ/Ϫ mice, in which case Ca 2ϩ responses under flow were not different from those of corresponding wild-type platelets. 4 Measurements of flow-dependent thrombus formation on collagen showed a reduced deposition of p85␣ Ϫ/Ϫ platelets, as well as a reduced PS exposure of the cells in comparison to wild-type blood (Fig.  7B). Similarly, addition of TGX-221 to p85␣ ϩ/ϩ blood, but not to p85␣ Ϫ/Ϫ blood, significantly decreased thrombus formation and number of PS-exposing platelets. A role for PI3K␣ and -␤ in murine thrombus formation was further supported by inhibitor studies using blood from wild-type mice. Treatment with PIK-75 and/or TGX-221 led to a similar reduction in platelet deposition and PS exposure as treatment with the general PI3K inhibitor, wortmannin (Fig. 7C). Together, these results demonstrated that, in mouse, the regulatory p85␣ subunit and the catalytic PI3K␣ and -␤ subunits are essential in flow-dependent Ca 2ϩ signaling, PS exposure, and thrombus formation on collagen.
Experiments carried out with washed mouse platelets, in the presence of autocrine stimulation inhibitors, confirmed that GPVI-induced activation relied on PI3K␣/␤ activity, because LY-294002, PIK-75, and TGX-221 all inhibited CRP-induced aggregation (not shown). In addition, platelets from mice lacking the catalytic p110␥ or p110␦ subunits were essentially unaltered in aggregation, when compared with wild-type platelets. 5 Together, this points to major functions of murine p110␣ and p110␤ PI3K isoforms and to minor roles for p110␥ and p110␦ in the aggregation response.
Roles of PI3K Isoforms in GPVI-induced Rap1b Activation in Human and Mouse Platelets-In platelets, the low molecular weight GTPase, Rap1b, is a well characterized effector downstream of PI3K (27). However, there is no conclusive evidence how Rap1b functions directly downstream of GPVI. Platelets from mice deficient in Rap1b were used to investigate this. In the presence of autocrine stimulation inhibitors, Rap1b Ϫ/Ϫ platelets showed a 36% reduction in CRP-induced aggregation in comparison to Rap1b ϩ/ϩ wild types (Fig. 8, A and B). This suggested a direct role for this GTPase in GPVI-induced platelet aggregation. Inhibition of PI3K with LY-294002 markedly 4 I. Munnix, unpublished data. 5 P. Mangin, unpublished data.

PI3K Isoforms in Glycoprotein VI-induced Platelet Activation
reduced the aggregation of Rap1b ϩ/ϩ platelets and, surprisingly, of Rap1b Ϫ/Ϫ platelets, with reductions of 89 and 86%, respectively, in comparison to vehicle-treated controls. The observation that LY-294002 further reduced the aggregation of Rap1b Ϫ/Ϫ platelets indicated that, even though Rap1b is involved in GPVI-mediated activation, it is not the only effector of PI3K in this response.
To elucidate the PI3K isoforms implicated in Rap1b activation directly downstream of GPVI, the effect of CRP was evaluated on Rap1b-GTP levels in platelets treated with autocrine stimulation inhibitors. An inhibitor of ␣ IIb ␤ 3 of human (40 g/ml Reopro) or mouse (100 M GPI-562) platelets served to suppress platelet aggregation, which process interferes with the assay. In both human and mouse platelets, CRP induced a potent activation of Rap1b (Fig. 9, A and B). Treatment with LY-294002 strongly suppressed Rap1b activation with Rap1-GTP levels being reduced in human and mouse platelets by 72 and 69%, respectively. Markedly, treatment with TGX-221 or PIK-75 resulted in a similar degree of reduction in Rap1b activation, both in mouse or human platelets. In contrast, effects of AS-252424 treatment were negligible. These findings demonstrate a major role for PI3K␣ and -␤ but not -␥ in Rap1b activation downstream of GPVI.

DISCUSSION
The present report provides the first data showing the contribution of multiple PI3K isoforms in GPVI-induced activation of (human) platelets and the role of these isoforms in GPVI-dependent thrombus formation. Our findings significantly extend the earlier evidence that GPVI agonists increase the formation of PI(3,4,5)P 3 and its derivative phosphoinositide 3,4-bisphosphate via PI3K activation (9,10,46). A marked finding is that both PI3K␣ and -␤ isoforms are needed for full InsP 3 formation and Ca 2ϩ signal generation, thus suggesting that both activities are necessary for optimal activation of the key effector enzyme of GPVI, PLC␥2. Detailed Ca 2ϩ measurements in the presence of extracellular EGTA and CaCl 2 pointed to a marked contribution of PI3K to Ca 2ϩ mobilization from intracellular stores, which is a direct read-out of PLC, and not to a (store-independent) effect on Ca 2ϩ entry. Hence, the earlier suggested regulation of Ca 2ϩ entry by PI3K or its product PI(3,4,5)P 3 (23,24) seems to be secondary to the effect of PI3K regulation of Ca 2ϩ store depletion.
Several observations point to a non-redundant contribution of PI3K␣ and -␤ isoforms in GPVI-induced platelet activation. These include the inhibitory effects on human platelets of a panel of compounds with high affinity to PI3K␣ (PIK-75, PI-103, and YM-024) or to PI3K␤ (TGX-221 and also PI-103). Other support comes from studies with mouse platelets, which responded similarly to the isoform-specific inhibitors as did human platelets. Furthermore, we found that mouse platelets lacking p85␣ were strikingly impaired in collageninduced Ca 2ϩ signaling and thrombus formation. These p85␣ Ϫ/Ϫ platelets are fully depleted in p110␣, but still express residual p85␤ and p110␤ (11). However, treatment of these platelets with TGX-221 did not further suppress Ca 2ϩ signaling and downstream responses such as PS exposure, although this compound did insignificantly reduce p85␣ Ϫ/Ϫ platelet deposition under flow. This suggested a function of the residual p110␤ in ADP-dependent platelet aggregate formation.
In human platelets, inhibition of PI3K␥ (AS-252424) failed to affect the GPVI-induced Ca 2ϩ response, whereas inhibition of PI3K␦ (IC-87114) caused a minor reduction, in contrast to the marked effect of PI3K␣/␤ inhibition. Similarly, mouse platelets, which were deficient in p110␥ or p110␦, showed undiminished activation, when stimulated with GPVI agonist under conditions where autocrine stimulators were blocked. These results are supported by data from others, showing that deficiency in murine p110␥ (19) or p110␦ (26) has no more than a modest role in GPVI-induced platelet activation.  The role of PI3K␣ and -␤ isoforms was particularly clear in measurements of GPVI-dependent thrombus formation, where platelets were flowed over collagen at a defined physiological shear rate. Blocking or deficiency in PI3K␣/␤ resulted in suppression of Ca 2ϩ increases and PS exposure of platelets adhered to collagen, as well as in reduced formation of aggregate formation. For both human and mouse blood, it is known that these responses fully rely on GPVI signaling via Src family kinases to LAT and PLC␥2 (34,47). Hence, the present findings support a concept that full PLC␥2-evoked activity is required for optimal platelet activation and thrombus formation under flow. Earlier, it was established that at the same flow conditions different populations of platelets assemble into aggregates or expose PS (38). Although this heterogeneity in platelet responses was maintained with PI3K␣/␤ inhibition or absence, the present results also suggest that the reduced PS exposure in inhibited platelets is a direct consequence of the reduced Ca 2ϩ signal of those platelets that are adhered to collagen.
Earlier, we and others have shown that two PI3K isoforms, namely PI3K␤ and -␥, are implicated in ADP/P2Y 12 -induced integrin ␣ IIb ␤ 3 activation and subsequent thrombus stabilization (17,19,20). Recently, these observations were extended to in vivo thrombosis models, where wortmannin treatment of p110␥-deficient mice was found to cause a dramatic defect in initial thrombus growth following vascular damage (21). The present data shed a new light on these in vivo observations, because it now appears that both collagen-and ADPdependent platelet activation processes are affected by PI3K inhibition with a likely key role of the PI3K␣/␤ isoforms.
There is a growing body of evidence that, in many cell systems individual PI3K isoforms, although involved in specific cellular functions, show functional redundancy, e.g. as noted for the additive contribution of PI3K␤ and -␥ in complement 5a-stimulated macrophages (48). So far, non-redundancy in PI3K function has been described only in particular cases. For example, in neutrophils producing inflammatory reactive oxygen species, both PI3K␥ and -␦ isoforms contribute to PI(3,4,5)P 3 formation, but in this case the role of PI3K␥ may be prior to that of PI3K␦ (31,49). In comparison to platelets, this is reminiscent of the early contribution of PI3K␣/␤ in GPVI stimulation, and the later involvement of PI3K␤/␥ in responses to autocrine produced ADP.
In the classic scheme of immunoglobulin receptor-linked PLC␥ activity, as proposed for lymphocytes, accumulation of PI(3,4,5)P 3 in the plasma membrane may function as an anchoring site for the N-terminal pleckstrin homology domain of PLC␥ isoforms (18,50). This scheme can be applicable to GPVI-stimulated platelets as well. Given the high turnover of phosphoinositides in the platelet plasma membrane, we hypothesize that the stimulation and anchoring of PLC␥2 requires a threshold elevation of PI(3,4,5)P 3 that can only be achieved by simultaneous activity of PI3K␣ and -␤. In other words, activation of either PI3K␣ or -␤ alone may be insufficient for PLC␥2 targeting to membrane via its PH domain. It is conceivable that this is a consequence of the need of PLC␥2 to compete with other PH-domain containing signaling proteins such as Akt isoforms. Such a mechanism would provide an attractive, but not the only, explanation for the identified non-redundant contribution to GPVI-induced signaling of both PI3K isoforms. One possible scenario is that one isoform may produce an initial small amount of PI(3,4,5)P 3 , which in turn promotes activation of the other isoform through PI(3,4,5)P 3 regulation of p85 subunits. Future studies examining temporal signaling by individual PI3K isoforms, as previously performed in neutrophils (31), will be needed to address this issue. The enhancing effect of PI3K activity on PLC␥2 is in contrast to the essential contribution of Src family kinases to GPVI-induced activation, because this is apparent from the complete abrogation of GPVI-induced Ca 2ϩ increases by Src family kinase inhibition (39).
The literature contains solid evidence that platelet Rap1b is activated in a PI3K-dependent manner, in particular after occupation of G i -coupled receptors such as the P2Y 12 receptor for ADP (27,28). By using autocrine stimulation inhibitors, where the P2Y 12 -dependent component was blocked, we could identify, for both human and mouse platelets, the existence of a direct GPVI-stimulated and PI3K-dependent component of Rap1b activation. Use of the isoform-specific inhibitors revealed involvement of PI3K␣/␤ in this activation of Rap1b. However, the prominent suppressive effect of PI3K inhibitors on ␣ IIb ␤ 3 activation and aggregation seen in Rap1b Ϫ/Ϫ platelets indicates that the PI3K isoforms contribute to GPVI-induced integrin activation also through a pathway that is independent of Rap1b. This pathway may involve PI3K␣/␤-dependent stimulation of PLC␥2, which can contribute to the Ca 2ϩ -and protein kinase C-mediated activation of ␣ IIb ␤ 3 .
PI3K␣ is the most frequently mutated kinase in human cancer. Hence, this isoform has raised great interest as a target for novel anti-tumor drugs, some of which are currently in early stage clinical trials (51). Given the likely important function of GPVI in (experimental) thrombosis, the present data suggest another important application for PI3K␣ and PI3K␤ inhibitors, namely in anti-platelet therapy and arterial thrombosis. In this context, it is relevant to note that arterial thrombosis is a known companion of some cancers. The availability of selective pharmacological agents against specific isoforms may thus offer new potential approaches for prevention of these diseases.