Identification of a Unique Co-operative Phosphoinositide 3-Kinase Signaling Mechanism Regulating Integrin αIIbβ3 Adhesive Function in Platelets*

Phosphoinositide (PI) 3-kinases play an important role in regulating the adhesive function of a variety of cell types through affinity modulation of integrins. Two type I PI 3-kinase isoforms (p110β and p110γ) have been implicated in Gi-dependent integrin αIIbβ3 regulation in platelets, however, the mechanisms by which they coordinate their signaling function remains unknown. By employing isoform-selective PI 3-kinase inhibitors and knock-out mouse models we have identified a unique mechanism of PI 3-kinase signaling co-operativity in platelets. We demonstrate that p110β is primarily responsible for Gi-dependent phosphatidylinositol 3,4-bisphosphate (PI(3,4)P2) production in ADP-stimulated platelets and is linked to the activation of Rap1b and AKT. In contrast, defective integrin αIIbβ3 activation in p110γ-/- platelets was not associated with alterations in the levels of PI(3,4)P2 or active Rap1b/AKT. Analysis of the effects of active site pharmacological inhibitors confirmed that p110γ principally regulated integrin αIIbβ3 activation through a non-catalytic signaling mechanism. Inhibition of the kinase function of PI 3-kinases, combined with deletion of p110γ, led to a major reduction in integrin αIIbβ3 activation, resulting in a profound defect in platelet aggregation, hemostatic plug formation, and arterial thrombosis. These studies demonstrate a kinase-independent signaling function for p110γ in platelets. Moreover, they demonstrate that the combined catalytic and non-catalytic signaling function of p110β and p110γ is critical for P2Y12/Gi-dependent integrinαIIbβ3 regulation. These findings have potentially important implications for the rationale design of novel antiplatelet therapies targeting PI 3-kinase signaling pathways.

The phosphoinositide (PI) 2 3-kinases are a well defined family of lipid kinases that participate in a broad range of signaling processes downstream of growth factor, antigen, hormone, and adhesion receptors (1,2). They are classified into several distinct groups (types I-III), based on their primary structure, mode of regulation, and substrate specificity (3,4). The most intensely studied members of the PI 3-kinase superfamily are the type I PI 3-kinases, due to their involvement in the regulation of fundamental cell processes, including proliferation, glucose metabolism, survival, and migration (1). PI 3-kinases principally transduce signals through the catalytic generation of phosphatidylinositol 3,4,5-trisphosphate (PI(3,4,5)P 3 ) and phosphatidylinositol 3,4-bisphosphate (PI(3,4)P 2 ); second messengers that facilitate the recruitment of pleckstrin homology domain-containing signaling proteins to the plasma membrane (5). The type I enzymes are divided into two subtypes, Ia and Ib; type Ia isoforms include p110␣, -␤, and -␦ and type Ib includes a single isoform p110␥. p110␣, -␤, and -␦ share common regulatory subunits (p85␣, p85␤, p55␣, p55␥, p50␣) and are classically regulated by tyrosine kinase-linked receptors, although G protein-coupled receptor-mediated activation of p110␤ has also been demonstrated (6,7). PI 3-kinase p110␥ contains structurally distinct p84 and p101 subunits that are typically regulated by G protein-coupled receptors, particularly G␤␥ i subunits (4).
There is a growing body of evidence demonstrating important functional specialization of individual type I PI 3-kinase isoforms. For example, p110␣ plays a key role in promoting cell growth and survival in response to growth factor and oncogenic stimulation (8,9). Moreover, gain-of-function mutations in p110␣ have oncogenic potential (10,11), and isoform-selective p110␣ inhibitors are effective at reducing tumor formation in vivo (12). The type Ib isoform, p110␥, plays an important role in innate immunity and in various inflammatory reactions linked to arthritis, pancreatitis, pulmonary inflammation, glomerulonephritis, and systemic lupus erythematosus (13). As a result, isoform-selective p110␥ inhibitors are currently being evaluated in a broad range of inflammatory disorders (13). p110␦ appears to have an important role in adaptive immunity, particularly in T-and B-cell, and mast cell signaling (14) with evidence that inhibition of p110␦ reduces allergic responses (15).
The role of p110␤ has been less clearly defined although based on studies using isoform-selective inhibitors p110␤ appears to play a major role in regulating integrin adhesive function in platelets, necessary for arterial thrombus formation (16). PI 3-kinase regulation of integrins, including ␣ IIb ␤ 3 (17,18), (19,20), ␣ 4 ␤ l (VLA-4) (21)(22)(23)(24)(25)(26), and ␤ 2 integrins (27) has been described in platelets, leukocytes, smooth muscle (ASMC) and melanoma (A2058, C32) cell lines. Multiple PI 3-kinase isoforms (p110␥, p110␤, and p110␦) and their downstream signaling molecules (Rap1b (28,29), AKT (30), and MAP kinases) have been implicated in integrin regulation. In platelets, both p110␥ and p110␤ signal downstream of the purinergic P2Y 12 / G i -coupled receptor and presumably regulate integrin ␣ IIb ␤ 3 adhesive function in a co-operative manner, although the basis for this has not been defined. In fact, one of the least clearly defined aspects of PI 3-kinase signaling are the mechanisms underlying signaling co-operativity between individual isoforms. In neutrophils, p110␥, p110␤, and p110␦ contribute to fMLP-induced superoxide generation (31), whereas in hepatoma cells and macrophages efficient insulin-stimulated glucose transport requires the co-operative input from multiple type Ia PI 3-kinases (p110␣, -␤, and -␦) (32). Similarly, cell migratory (33) and proliferative responses (9,34) appear to be regulated by multiple PI 3-kinases, with the contribution of individual isoforms depending on the cell type and context of stimulation. The most straightforward explanation for co-operative PI 3-kinase isoform signaling is a simple amplifier effect, i.e. enhanced second messenger formation (3-phosphoinositides) through the cumulative impact of multiple enzymes. However, it is also possible that individual PI 3-kinases generate spatially and temporally distinct pools of 3-phosphoinositides that act to assemble discrete signaling complexes with complementary signaling functions. In addition, individual family members may also amplify the signals other PI 3-kinases, i.e. the initial production of a small amount of 3-phosphoinositides may enhance subsequent activation of other family members. This may involve PI(3,4,5)P 3 binding to p85 (35) or through recruitment of pleckstrin homology domain containing proteins that modulate PI 3-kinase activity. These mechanisms are not mutually exclusive and it is possible that multiple co-operative signaling processes operate in the cell.
In the current study we have investigated the mechanisms underlying signaling co-operativity between p110␥ and p110␤ in platelets. Our studies indicate that p110␤ predominately regulates G i -dependent integrin ␣ IIb ␤ 3 activation through a classical lipid kinase-dependent mechanism, involving Rap1 and AKT, whereas p110␥ appears to regulate integrin ␣ IIb ␤ 3 principally through a non-catalytic signaling mechanism. Moreover, inhibiting the kinase function of PI 3-kinases, in combination with p110␥ deletion, led to a much greater defect in platelet adhesive function than inhibition of the catalytic function of PI 3-kinases alone. These studies demonstrate the existence of a co-operative PI 3-kinase signaling mechanism, involving the non-catalytic and catalytic function of p110␥ and p110␤, respectively. Furthermore, they demonstrate that this signaling mechanism is the predominant pathway utilized by the G i -coupled P2Y 12 receptor to regulate integrin ␣ IIb ␤ 3 adhesive function. These findings have potentially important impli-cations for the antithrombotic potential of isoform-selective PI 3-kinase inhibitors.
Preparation of Washed Platelets-Preparation of washed murine platelets was performed as described previously (39).
Measurement of Integrin ␣ IIb ␤ 3 Activation-Integrin ␣ IIb ␤ 3 activation was assessed by measurement of Oregon Green-labeled fibrinogen binding, as described previously (16). Background (nonspecific) Oregon Green-labeled fibrinogen binding was determined on ADP-stimulated platelets pre-treated with GPI562 (100 M). Specific Oregon Green-labeled fibrinogen binding was determined by subtracting nonspecific fluorescence readings from total fluorescence obtained from ADPstimulated platelets.
HPLC-based Phospholipid Analysis-Washed mouse platelets were labeled with 0.5 mCi/ml inorganic [ 32 P]H 3 PO 4 , as described previously (39) and stimulated with ADP (10 M) for the indicated time periods. Lipids were extracted and separated by HPLC according to a modified method of Stephens et al. (41). PI peaks co-eluting with commercially available PI(3,4)P 2 and PI(3,4,5)P 3 standards were integrated and normalized to the total lipid applied.
Measurement of Rap1 Activation-The level of GTP-bound Rap1 in mouse platelet lysates was measured as described previously (28,29).
Measurement of AKT Activation-AKT activation was measured by Western blot analysis, using a phosphoserine-473 AKT polyclonal antibody to detect active AKT, and an AKT polyclonal antibody (Cell Signaling Technology, MA) to determine total AKT, as described previously (42).
PI 3-Kinase Lipid Kinase Assays-Lipid kinase assays were performed to determine the level of PI 3-kinase inhibition achieved in platelets following preincubation of mouse whole blood with vehicle/wortmannin, or following administration (intravenous bolus) of vehicle/wortmannin in mice. Platelets were isolated from whole blood as described above, lysed, and PI 3-kinase immunoprecipitated from platelet lysates using an anti-p85 polyclonal antibody (Upstate Biotechnology). Kinase assays were performed on immunoprecipitated PI 3-kinase according to a previously described method (43). Immunoprecipitations were also were performed in the absence of the anti-p85 polyclonal antibody, and the resulting PI 3-kinase activity subtracted as background.
Platelet Aggregation Studies-For aggregation studies, washed mouse platelets (2.0 ϫ 10 8 /ml) were preincubated with vehicle alone, LY294002 (25 M), TGX221 (0.5 M), D-010 (5 M), or AS252424 (1-10 M) prior to stimulation with the indicated concentrations of ADP in the presence of calcium (1 mM) and fibrinogen (500 g/ml). All aggregations were initiated by stirring the suspensions at 600 rpm for 10 min at 37°C in a four-channel automated platelet analyzer (AggRAM, Haem). The extent of platelet aggregation was defined as the percentage change in optical density as measured by the automated platelet analyzer.
In Vitro Thrombus Formation under Flow-Flow-based thrombus formation assays on a bovine fibrillar type I collagen matrix were performed as described previously (44). Anticoagulated whole blood (100 g/ml lepirudin) collected from p110␥ ϩ/ϩ or p110␥ Ϫ/Ϫ mice was labeled with DiOC 6 (1 M) and preincubated with vehicle (Me 2 SO) or wortmannin (400 nM) (10 min, 37°C) prior to perfusion through fibrillar type I collagen-coated microcapillary tubes (2.5 mg/ml) at 1800 s Ϫ1 for up to 2.5 min. Thrombus formation was observed using an inverted Leica DMIRB microscope (Leica Microsystems, Wetzlar, Germany) and a ϫ63, 1.2 numeric aperture water objective. Images were acquired using a Dage-MTI charge-coupled device camera 300 ETRCX (Dage-MTI, Michigan City, IN). Analysis of thrombi volume and height was performed after 2.5 min (where thrombi reached a maximal volume in vehicle-treated p110␥ ϩ/ϩ platelets), as described previously. Total thrombus volume and maximum height of thrombi in a given field (covering 25,058.89 m 2 ) were calculated using Metamorph 6.
Electrolytic Model of Occlusive Thrombus Formation-An in vivo electrolytic model of occlusive thrombus formation was performed as described previously (16,44). p110␥ ϩ/ϩ and p110␥ Ϫ/Ϫ mice were administered (intravenous bolus) vehicle (10% Me 2 SO) or the indicated concentrations of wortmannin, prior to induction of electrolytic injury. Blood flow was monitored for 60 min, and changes in blood flow recorded offline for analysis. The total amount of blood flowing through the injured artery following vascular injury was determined by calculating the area under the blood flow curve, corrected for body weight.
Tail Bleeding Studies in Mice-Tail bleeding time was measured in anesthetized (pentobarbitone 60 mg/kg, intraperitoneal), ventilated p110␥ ϩ/ϩ and p110␥ Ϫ/Ϫ mice, with body temperature maintained at 37°C throughout the experiment (16,44). Tail bleeding was measured before and 10 min after administration of vehicle or wortmannin (0.1 mg/kg). Incisions 5-mm long and 1-mm deep were made, 10 mm from the tip of the tail, and bleeding was monitored by blotting the edge of the incision with a tissue every 30 s until it had ceased (ϭtail bleeding time).
Histology-Carotid arteries were paraffin-embedded, processed, and stained using the Carstair stain, as described previously (44). Using this staining technique, fibrin appears red, platelets appear blue/purple, red blood cells appear yellow, and collagen staining in the vessel wall appears blue.
Statistical Analysis-Data are presented as mean Ϯ S.E. Average S.E. for carotid blood flow over time was calculated from repeated measures of analysis variance. When comparing matched values between two treatment groups, an unpaired Student's t test with two-way analysis of variance was used (Prism software, GraphPAD Software for Science, San Diego, CA) (not significant; p Ͼ 0.05; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001).

Relative Roles of Type I PI 3-Kinase Isoforms in G i -dependent
Integrin ␣ IIb ␤ 3 Activation-ADP is an important platelet agonist stimulating integrin ␣ IIb ␤ 3 activation through two purinergic receptors, P2Y 1 and P2Y 12 (45). P2Y 1 is a G q -coupled receptor that induces activation of phospholipase C␤ and transient calcium mobilization, whereas P2Y 12 is a G i -coupled receptor linked to the inhibition of adenylyl cyclase and activation of PI 3-kinase (45). ADP binding to P2Y 1 stimulates transient integrin ␣ IIb ␤ 3 activation, whereas P2Y 12 amplifies and sustains integrin ␣ IIb ␤ 3 activation, necessary for stable platelet aggregation. To examine the contribution of PI 3-kinases to integrin ␣ IIb ␤ 3 activation over a broad range of ADP concentrations, fibrinogen binding studies were performed on murine platelets in the presence of the pan-PI 3-kinase inhibitors, LY294002 or wortmannin. As demonstrated in Fig. 1, stimulation of platelets with increasing concentrations of ADP led to a dose-dependent increase in fibrinogen binding (Fig. 1). This increase was specific to integrin ␣ IIb ␤ 3 as it was completely eliminated by an anti-integrin ␣ IIb ␤ 3 receptor antagonist (data not shown). Pretreating platelets with LY294002 decreased fibrinogen binding by 75-85% over a broad range of ADP concentrations (Fig. 1). Similar findings were observed with wortmannin (data not shown), confirming a major role for PI 3-kinases in this process.
To investigate the relative contribution of type I PI 3-kinase isoforms to ADP-dependent integrin ␣ IIb ␤ 3 regulation, fibrinogen binding studies were performed on p110␥ Ϫ/Ϫ and p110␦ Ϫ/Ϫ platelets and on wild type platelets treated with the isoform-selective pharmacological inhibitors against p110␤ (TGX221) and p110␣ (PIK-75). Deficiency of p110␦ (p110␦ Ϫ/Ϫ ) or inhibition of p110␣ had no significant inhibitory effect on ADP-induced fibrinogen binding to platelets or on platelet aggregation (supplementary Fig. S1, A and B, and data not shown), excluding a significant contribution of these isoforms to P2Y 12 /G i -dependent ␣ IIb ␤ 3 regulation. In contrast, inhibiting p110␤ produced a major defect in fibrinogen binding over a wide range of ADP concentrations ( Fig. 1) that accounted for up to 90% of the inhibitory effects observed with LY294002. In control studies, we confirmed that combining LY294002 with TGX221 had no further inhibitory effect above that observed with LY294002 alone (data not shown), confirming that the inhibitory effects of TGX221 were primarily linked to PI 3-kinase signaling processes. Furthermore, in P2Y 1 -deficient murine platelets, TGX221 inhibited the P2Y 12 /PI 3-kinase-dependent platelet aggregation response induced by high concentrations of ADP (supplementary Fig. S2) (46). The effects of TGX221 were dose-dependent (supplementary Fig. S2) with an IC 50 (50 nM) that correlated well with the inhibition of p110␤ at physiological ATP concentrations.
The ability of TGX221 to inhibit ADP-induced fibrinogen binding with only slightly less potency than LY294002 was unexpected given previous findings for an important role for p110␥ in ADP-dependent integrin ␣ IIb ␤ 3 activation (42). As demonstrated in Fig. 1, we confirmed an approximate 35% reduction in fibrinogen binding to ADP-stimulated p110␥ Ϫ/Ϫ platelets. This effect was observed over a broad range of ADP concentrations (Fig. 1). In control studies we confirmed that inhibition of p110␤ or deficiency of p110␥ had no inhibitory effect on P2Y 1 -dependent calcium flux or platelet shape change (data not shown), confirming that these isoforms were principally signaling downstream of P2Y 12 . Interestingly, pretreating p110␥ Ϫ/Ϫ platelets with LY294002 produced a greater reduction in fibrinogen binding relative to p110␥ ϩ/ϩ platelets treated with LY294002 ( Fig. 1), an unexpected finding if p110␥ regulated integrin ␣ IIb ␤ 3 principally through a kinase-dependent mechanism. This differential effect of LY294002 on p110␥ ϩ/ϩ and p110␥ Ϫ/Ϫ platelets was also observed with wortmannin (data not shown) and furthermore, was specific to p110␥ Ϫ/Ϫ platelets as it was not apparent with p110␦ Ϫ/Ϫ platelets (supplementary Fig. S1B). These initial findings raised the possibility that p110␥ Ϫ/Ϫ may contribute to integrin ␣ IIb ␤ 3 activation through a kinase-independent signaling mechanism.
p110␤ Is Primarily Responsible for Promoting G i -dependent Rap1b Activation, AKT Phosphorylation, and PI(3,4)P 2 Production in Platelets-The possibility that p110␥ may signal downstream of P2Y 12 , at least partially through a non-catalytic signaling mechanism, has not previously been investigated. One of the downstream signaling proteins activated by PI 3-kinases and linked to the regulation of integrin ␣ IIb ␤ 3 is the small GTPase Rap1b (47,48). Both p110␥ (48) and p110␤ (16) have been implicated in Rap1b activation, however, their relative contribution to this process remains controversial (28,29). To examine changes in Rap1 activation in platelets we utilized the GST-RalGDS fusion protein to selectively precipitate active Rap1 from platelet lysates. Consistent with previous findings, Rap1 activation occurred rapidly and transiently in ADP-stimulated mouse platelets, peaking within 30 min of stimulation and declining by 60 min (Fig. 2A, (i). Under these experimental conditions, Rap1 activation occurred independent of ligand binding to integrin ␣ IIb ␤ 3 (data not shown). Pretreating platelets with TGX221 resulted in a marked reduction in the activa-tion of Rap1b ( Fig. 2A (ii)), and time course studies revealed a strong correlation between the reduction in active Rap1 and the subsequent disaggregation of platelets (data not shown). Similar findings were apparent with LY294002 (data not shown). Notably, Rap1 activation in p110␥ Ϫ/Ϫ platelets was normal at all time points examined ( Fig. 2A, (i) and (ii)). Relative roles of p110␤ and p110␥ in ADP-induced Rap1 activation, AKT activation, and PI(3,4)P 2 production. Washed platelets (5.0 ϫ 10 8 /ml) were isolated from p110␥ ϩ/ϩ and 110␥ Ϫ/Ϫ mice. In some experiments, platelets were preincubated with vehicle (DMSO), TGX221 (221, 0.5 M), or AR-C69931MX (10 M) for 10 min at 37°C. A and B, platelets were stimulated with ADP in the presence of stirring. A, following stimulation, platelets were lysed and active Rap1b precipitated as described under "Experimental Procedures." Results depict p110␥ ϩ/ϩ and p110␥ Ϫ/Ϫ platelets stimulated with 2 M ADP for (i) the indicated times, or (ii) 30 s. B, following stimulation with 10 M ADP for 10 min, platelets were lysed, and AKT phosphorylation detected as described under "Experimental Procedures." C, platelets were loaded with 32 P, preincubated with the indicated inhibitors, and stimulated with ADP (10 M) for the indicated time periods. Following this, platelets were lysed, and 32 P-labeled phospholipids (i, PI(3,4)P 2 ; ii, PI(4,5)P 2 ) were extracted and analyzed by SAX-HPLC, as described under "Experimental Procedures." A-C, results represent the mean Ϯ S.E. from three independent experiments; immunoblots are taken from one representative of three experiments. DMSO, dimethyl sulfoxide.
Our demonstration of minimal changes in Rap1 activation and AKT phosphorylation in p110␥ Ϫ/Ϫ platelets raised the provocative possibility that p110␥ may not play a major role in generating PI(3,4)P 2 and PI(3,4,5)P 3 in ADP-stimulated platelets. Definitive conclusions on the activation state of PI 3-kinases requires direct measurement of their phospholipid products in vivo, however, to date there have been no studies quantifying PI(3,4)P 2 and PI(3,4,5)P 3 levels in p110␥ Ϫ/Ϫ platelets. As demonstrated in Fig. 2C, ADP stimulated a time-dependent increase in the cellular levels of PI(3,4)P 2 in p110␥ ϩ/ϩ platelets (Fig. 2C, i). This increase was P2Y 12 -dependent as it was completely inhibited by the P2Y 12 receptor antagonist AR-C69931MX (Fig. 2C, i). A transient increase in PI(3,4,5)P 3 was also noted (data not shown), however, its levels were very low relative to PI(3,4)P 2 and therefore was not a reliable marker for assessing the relative contribution of individual PI 3-kinase isoforms to the lipid response. Analysis of PI(3,4)P 2 levels in ADP-stimulated p110␥ Ϫ/Ϫ platelets demonstrated no significant difference from p110␥ ϩ/ϩ controls (Fig. 2C, i), suggesting very little role for p110␥ in this response. In contrast, pretreating p110␥ ϩ/ϩ murine platelets with TGX221 inhibited PI(3,4)P 2 generation by (Ͼ90%) at all time points examined ( Fig. 2C (i)). The effects of TGX221 were selective, as it had no significant inhibitory effect on non-PI 3-kinase lipid products such as PI(4,5)P 2 (Fig. 2C (ii)), PI(3)P, or PI(4)P (data not shown). These studies support our findings from the AKT/ RapIb assays that the catalytic function of p110␤ is primarily responsible for G i /P2Y 12 -dependent PI(3,4)P 2 production.
Effect of an Isoform-selective p110␥ Kinase Inhibitor on ADPstimulated Integrin ␣ IIb ␤ 3 Activation-To investigate further the relationship between the kinase function of p110␥ and integrin ␣ IIb ␤ 3 regulation in ADP-stimulated platelets we examined the effect of an active site pharmacological inhibitor against p110␥ (AS252424). This inhibitor has 100-fold selectivity for p110␥ over p110␤ (IC 50 , 12 versus 1,220 nM in in vitro kinase assays, respectively) (31) and inhibits the production of PI 3-kinase lipid products at low micromolar concentrations in neutrophils (31,36,50). We confirmed previous findings (50) that AS252424 at concentrations as low as 1 M partially inhibits platelet thrombus formation on a collagen substrate under flow (data not shown), confirming its effectiveness in platelets. At these concentrations, AS252524 has minimal off-target effects on other PI 3-kinase isoforms (see supplementary Fig.  S3). As demonstrated in Fig. 3, pretreating mouse platelets with AS252424 resulted in a non-significant reduction (p Ͼ 0.1) in ADP-induced integrin ␣ IIb ␤ 3 activation (Fig. 3A) and platelet aggregation ( Fig. 3B (i) and (ii)). Even with high concentrations of AS252424 (up to 10 M), no significant reduction in integrin ␣ IIb ␤ 3 activation and platelet aggregation was observed (data not shown). These findings provide further evidence suggesting that loss of the kinase function of p110␥ is unlikely to fully explain the defects in integrin ␣ IIb ␤ 3 regulation in p110␥-deficient platelets.

The Combined Catalytic and Non-catalytic Signaling Function of PI 3-Kinases Is Critical for Platelet Aggregation and Thrombus
Growth-To gain insight into the potential functional significance of co-operative (catalytic and non-catalytic) PI 3-kinase signaling for integrin ␣ IIb ␤ 3 -dependent functional responses, we initially examined the aggregation response of p110␥ ϩ/ϩ and p110␥ Ϫ/Ϫ platelets treated with LY294002 or wortmannin. As demonstrated in Fig. 4, deficiency of p110␥ alone resulted in a significant defect in ADPinduced platelet aggregation, characterized by a 20 -30% decrease in both the extent and duration of platelet aggregation (Fig. 4, i and ii). Pretreating p110␥ Ϫ/Ϫ platelets with either  Fig. 4A (i) and (ii)) or wortmannin (data not shown) resulted in a significantly greater defect in platelet aggregation relative to p110␥ ϩ/ϩ controls. Similarly, blocking the kinase function of p110␤ with TGX221 produced a more severe defect in the aggregation response of p110␥ Ϫ/Ϫ platelets relative to p110␥ ϩ/ϩ platelets ( Fig. 4 (ii)). In fact, inhibiting p110␤ signaling in p110␥ Ϫ/Ϫ platelets reduced integrin ␣ IIb ␤ 3 activation by up to 90% (Fig. 4B), leading to a major defect in sustained platelet aggregation (Fig. 4A (ii)). These findings suggest that the combined signaling function of p110␤ and p110␥ is the principal mechanism by which the P2Y 12 receptor regulates integrin ␣ IIb ␤ 3 activation in platelets.

LY294002 (
It is well established that the platelet P2Y 12 receptor plays a major role in regulating platelet thrombus formation in vitro and in vivo (45,51). To investigate the significance of co-operative PI 3-kinase signaling for platelet thrombus development, we performed platelet perfusion experiments using p110␥ ϩ/ϩ and p110␥ Ϫ/Ϫ mouse blood treated with wortmannin (400 nM, see Fig. 5A). Real-time analysis of platelet thrombi formed from p110␥ ϩ/ϩ anticoagulated whole blood on a collagen substrate (1,800 s Ϫ1 ) revealed the rapid formation of large, stable thrombi within the first 2.5 min of flow (Fig. 5, B and C, and supplementary video). Pretreating p110␥ ϩ/ϩ whole blood with wortmannin lead to the development of unstable thrombi with constant embolization from the thrombus surface (see supplementary video). Volumetric analysis of formed thrombi by confocal microscopy confirmed that wortmannin reduced thrombus volume and height, by approximate 45 and 30%, respectively (Fig. 5, B and C). Consistent with previous reports, platelets from p110␥ Ϫ/Ϫ mice also demonstrated a defect in thrombus formation (50) (Fig. 5, B and C), with reduced recruitment of platelets onto the surface of developing thrombi (data not shown) leading to a ϳ10% reduction in thrombus height and an ϳ30% decrease in thrombus volume (Fig. 5B). Significantly, deficiency of p110␥ in combination with wortmannin resulted in a profound defect in thrombus growth (Fig. 5B). Thrombi formed under these conditions were very fragile with continuous detachment of platelet aggregates from the thrombus surface (see supplementary video), resulting in a ϳ90% reduction in thrombus volume and a 70% decrease in thrombus height. Moreover, the defect in thrombus development in these mice was strikingly different from that observed with wortmannin-treated p110␥ ϩ/ϩ platelets, in that the former had a profound defect in initial thrombus growth, whereas the latter had a primary defect in thrombus stability. In control studies we confirmed that the concentrations of wortmannin used in whole blood studies (400 nM) effectively inhibited the catalytic function of type I PI 3-kinases (Fig. 5A) and even at concentrations as high as 2 M, wortmannin was unable to produce a thrombosis defect similar to that observed in wortmannintreated p110␥ Ϫ/Ϫ platelets (data not shown). Overall, these studies demonstrate that inhibition of the catalytic function of PI 3-kinases, in combination with p110␥ deficiency, leads to a profound defect in platelet aggregation and thrombus development.

Role of Combined Catalytic and Non-catalytic PI 3-Kinase Signaling in Regulating Platelet Adhesive Function in Vivo-To
investigate the significance of the combined catalytic and noncatalytic signaling function of PI 3-kinases for platelet function in vivo, we assessed arterial thrombus formation in p110␥ ϩ/ϩ and p110␥ Ϫ/Ϫ mice, following treatment with wortmannin (for full details on the antithrombotic effects of p110␥ deficiency and PI 3-kinase inhibitors, see Fig. 6 and additional supplementary information in supplementary Figs. S4 and S5). We utilized a carotid artery thrombosis model that employs electrolytic injury to induce the formation of fibrin-and platelet-rich occlusive thrombi (Fig. 6). This model was chosen as it is able to discern the relative potency of different antithrombotic approaches. For example, occlusive thrombus formation is effectively inhibited by potent anti-platelet agents, including integrin ␣ IIb ␤ 3 , P2Y 12 receptor antagonists (44), and high dose thrombin inhibitors (data not shown), but not by weaker antithrombotic approaches, such as aspirin (16) or GPVI deficiency (44). To investigate the effect of inhibiting the catalytic function of PI 3-kinases on thrombus formation in the electrolytic model we administered wortmannin to p110␥ ϩ/ϩ mice (0.1 mg/kg (intravenously); a dose that completely inhibited PI 3-kinase activity in platelets, see supplementary Fig. S5A). Wortmannin decreased platelet accumulation at areas of vascular injury and prevented arterial thrombotic occlusion, however, the residual thrombi formed under these conditions were fibrin-rich and as a consequence complete restoration of blood flow to pre-injury levels was not observed (see Fig. 6B, and supplementary Fig. S5B, i and ii). Under experimental conditions in which thrombin is partially inhibited with enoxaparine, wortmannin markedly inhibited arterial thrombus formation and increased blood flow from 17.32 Ϯ 4.6 ml/min/100 g for 60 min to 48.18 Ϯ 5.9 ml/min/100 g for 60 min (Fig. 6, A (i) and (ii),  and B). Significantly, inhibiting the catalytic function of PI 3-kinases in p110␥ Ϫ/Ϫ mice eliminated arterial thrombotic occlusion completely in both untreated (Fig. 6B) and enoxaparine (Fig.  6B)-treated animals. Blood flow was maintained at preinjury levels throughout the entire 60 min postinjury period (Fig. 6A; supplementary Fig. S5B). Histological analysis revealed a complete inhibition of thrombus formation, regardless of whether thrombin was partially inhibited or not (Fig. 6B). Similar to our in vitro findings, exposing platelets in vivo to progressively higher wortmannin concentrations (up to 1 mg/kg in p110␥ ϩ/ϩ mice) did not reproduce the potent antithrombotic effects observed in wortmannin (0.1 mg/kg)-treated p110␥ Ϫ/Ϫ mice (data not shown).
Additional evidence that loss of p110␥ further undermined platelet adhesive function in wortmannintreated mice was derived from tail bleeding time studies. Using a modified template bleeding time assay we observed no significant difference in the tail bleeding time of p110␥ ϩ/ϩ or p110␥ Ϫ/Ϫ mice (data not shown). In contrast, infusing wortmannin at concentrations that effectively inhibited the catalytic function of type I PI 3-kinases in p110␥ ϩ/ϩ mice (supplementary Fig.  S4) increased the bleeding time ϳ1.8-fold (n ϭ 7, p Ͻ 0.05), whereas inhibition of the catalytic function of PI 3-kinases, combined with p110␥ deficiency, lead to a further 65% increase in bleeding time (ϳ2.8-fold greater than p110␥ null mice treated with vehicle alone) (n ϭ 8, p Ͻ 0.001). Collectively, these studies suggest an important co-operative catalytic and non-catalytic signaling role for PI 3-ki-FIGURE 5. The non-catalytic signaling function of p110␥ contributes to regulation of thrombus formation. A-C, anticoagulated whole blood collected from p110␥ ϩ/ϩ or p110␥ Ϫ/Ϫ mice was preincubated with the membrane lipid dye DiOC 6 and the indicated concentrations of vehicle control or wortmannin (WN) (10 min, 37°C). A, platelets were isolated from whole blood, lysed, and immunoprecipitated PI 3-kinase assayed for lipid kinase activity as described under "Experimental Procedures." Results depict the mean Ϯ S.E. (n ϭ 3) of PI 3-kinase activity (corrected for background count) expressed as a percentage of vehicle-treated whole blood (100%). Inset, thin layer chromatography analysis of one such PI 3-kinase activity assay, representative of three independent experiments. B and C, whole blood pretreated with vehicle or wortmannin (WN, 400 nM) was perfused through microcapillary tubes (coated with 2.5 mg/ml type I fibrillar collagen) at 1800 s Ϫ1 for 2.5 min. Thrombi were then fixed, and imaged by confocal microscopy (Leica Microsystems, Wetzlar, Germany), using a ϫ63, 1.2 numeric aperture water objective, as described under "Experimental Procedures." B, thrombus volume and maximum height of thrombi in a given field (covering 25,058.89 m 2 ) were calculated using Metamorph (Molecular Signaling). Histograms represent the mean Ϯ S.E. of three independent experiments. C, thrombi were reconstructed in Voxblast (Vaytek Inc., Fairfield, IA). Images are taken from one representative of three independent experiments. DMSO, dimethyl sulfoxide. nases in regulating both the hemostatic and prothrombotic properties of platelets in vivo.

DISCUSSION
The studies reported here demonstrate a unique co-operative signaling mechanism operating between type Ia and Ib PI 3-kinases that plays a major role in regulating the adhesive function of platelets. We have demonstrated that p110␤ is the dominant PI 3-kinase isoform responsible for G idependent integrin ␣ IIb ␤ 3 activation, regulating platelet aggregation through the catalytic modulation of Rap1b and/or AKT. In contrast, p110␥ appears to regulate integrin ␣ IIb ␤ 3 adhesive function downstream of P2Y 12 through a non-catalytic signaling mechanism. We demonstrate that this co-operative signaling function of PI 3-kinase isoforms plays a major role in regulating the formation of stable platelet-platelet adhesive interactions under flow, and as a conse-quence, is essential for platelet thrombus growth. Thus, the combined signaling function of p110␤ and p110␥ appears to be the principal mechanism by which P2Y 12 regulates the adhesive function of integrin ␣ IIb ␤ 3 . Moreover, based on our in vivo studies, this co-operative PI 3-kinase signaling mechanism appears indispensable for the normal hemostatic function of platelets.
A major finding in the current study was the demonstration that p110␥ can regulate integrin ␣ IIb ␤ 3 adhesion function through a noncatalytic signaling mechanism. Numerous independent lines of experimental evidence suggest a minor kinase-dependent signaling function for p110␥ downstream of P2Y 12 in platelets. First, we have demonstrated minimal reduction in ADP-mediated P2Y 12 -dependent PI(3,4)P 2 production in p110␥-deficient mouse platelets. Second, P2Y 12 -dependent Rap1b and AKT activation, which are principally regulated by PI 3-kinase lipid products, are minimally affected by p110␥ deficiency. Third, pretreating platelets with the isoform-selective p110␥ inhibitor, AS252424, had a minimal effect on P2Y 12 -dependent integrin ␣ IIb ␤ 3 activation and platelet aggregation. Fourth, inhibiting the kinase function of all type I PI 3-kinases lead to a much greater defect in platelet aggregation and thrombus formation in p110␥ Ϫ/Ϫ platelets relative to p110␥ ϩ/ϩ platelets. Finally and perhaps most unexpectantly, the nature of the thrombosis defect observed in wortmannin-treated p110␥ Ϫ/Ϫ mice was distinct from that observed in wortmannin-treated p110␥ ϩ/ϩ mice, in that the former had a profound defect in initial thrombus growth, whereas the latter had a primary defect in thrombus stability.
One of the most significant observations in the current study was the extent of the thrombosis defect observed in p110␥ Ϫ/Ϫ mice treated with PI 3-kinase inhibitors, relative to PI 3-kinase catalytic inhibition alone, suggesting a fundamental role for the combined catalytic and non-catalytic signaling function of PI 3-kinases in integrin ␣ IIb ␤ 3 regulation. In fact, this is the first report demonstrating that inhibition of PI 3-kinase signaling processes can cause a significant hemostatic defect in vivo. To date, the only signaling pathway that has been demonstrated to play such a fundamental role in integrin ␣ IIb ␤ 3 regulation is the phosphoinositide-specific phospholipase C signaling pathway, whereby defects in G q (52,53), phospholipase C␤ 2/3 (54), Rap1b (47), and CalDAG-GEF1 (55) lead to a major reduction in platelet adhesive function and thrombus formation. These latter signaling molecules are critical for integrin ␣ IIb ␤ 3 activation in response to all major soluble agonists, however, in isolation, are ineffective at sustaining integrin ␣ IIb ␤ 3 activation and stable platelet aggregation. Our studies suggest that the combined signaling function of PI 3-kinases is likely to be the predominant pathway maintaining integrin ␣ IIb ␤ 3 activation in vivo, necessary for both hemostatic plug formation and thrombus growth.
Although our studies demonstrate a non-catalytic signaling function for p110␥ in platelets, we wish to emphasize that our results do not exclude an important kinase-dependent signaling function for this isoform as well. For example, it has previously been demonstrated that p110␥ kinase inhibitors can reduce thrombus formation on a collagen substrate under flow (50), a finding confirmed in the current study. In addition, a previous report has demonstrated a role for p110␥ kinase function in promoting P2Y 12 -dependent AKT phosphorylation (42). Other studies have suggested that p110␥ kinase function can partly contribute to G i -dependent Rap1 activation (28), although this remains controversial (29). The reason for this discrepancy is not immediately apparent but presumably reflects methodological differences. For example, it remains possible that p110␥ kinase function may be functionally relevant at certain agonist concentrations or time points of platelet stimulation. Thus, it is reasonable to assume that p110␥ signals through both kinase-dependent and -independent mechanisms in platelets, with the studies presented here demonstrating that the latter plays an important co-operative role with p110␤ to regulate integrin ␣ IIb ␤ 3 adhesive function.
One of the major findings in the current study is the demonstration that the combined (catalytic and non-catalytic) signaling function of p110␤ and p110␥ is the major mechanism by which G i /P2Y 12 regulates integrin ␣ IIb ␤ 3 adhesive function. The importance of the P2Y 12 receptor in supporting the hemostatic function of platelets and in promoting arterial thrombosis is well defined, and as a consequence, this receptor is a major target for antithrombotic therapies (45). This receptor can regulate multiple G i -coupled signaling pathways (reviewed in Ref. 56), including adenylyl cyclase, Src kinases, PI 3-kinases, MAP kinases, and potentially G-protein inward rectifying potassium channels (57), although which of these represents the dominant pathway regulating integrin ␣ IIb ␤ 3 adhesive function has remained elusive. It is likely that the major effector pathway utilized by p110␤ to regulate integrin ␣ IIb ␤ 3 activation is via activation of Rap1b, although it is possible that AKT may also play a role. The mechanism by which p110␥ regulates integrin ␣ IIb ␤ 3 remains unclear. The only non-catalytic signaling mechanism identified to date for p110␥ is through the regulation of cAMP in cardiac cells, potentially through modulation of cAMP phosphodiesterase (PDE3) (58). Although cAMP has a well defined role in negatively regulating integrin ␣ IIb ␤ 3 adhesive function in platelets, partially through regulation of VASP (59), we and others (42) have found no evidence of abnormal cAMP levels, nor did we detect any differences in VASP phos-phorylation or PKA activation in p110␥ null platelets. 3 There is also evidence that p110␥ regulates both calcium mobilization and influx in multiple cell types including cardiac myocytes (60,61), skeletal myotubes (62), pancreatic acinar cells (63), and mast cells (64). Several groups have also implicated PI 3-kinase in the regulation of calcium signaling in platelets (65,66), although the involvement of p110␥ in this response remains controversial (42,54), with our own preliminary studies revealing no significant difference in the calcium response in ADPstimulated p110␥ Ϫ/Ϫ versus p110␥ ϩ/ϩ platelets. 3 Future studies will be required to define the mechanism by which p110␥ regulates integrin ␣ IIb ␤ 3 adhesive function.
The demonstration for a major role for p110␤ in G i signaling in platelets is consistent with findings in other cell types, including NIH3T3 cells (67), C6 glioma cells (68), and COS-7 cells (69). In fact, given the limited tissue distribution of p110␥ it has been proposed that p110␤ is likely to be the major PI 3-kinase isoform transducing G i -dependent signals in most cell types (67). However, the situation in hemopoietic cells might be different as these cells contain abundant levels of both p110␤ and p110␥ and based on studies in fMLP-stimulated p110␥ Ϫ/Ϫ neutrophils (31), p110␥ appears to play the major role in G i -dependent superoxide formation. The G i -dependent signaling function of p110␥ in neutrophils appears to be primarily dependent on its catalytic function as fMLP-stimulated PI(3,4,5)P 3 production is completely eliminated in p110␥ Ϫ/Ϫ neutrophils and following treatment of human and mouse neutrophils with AS252424 (31). In neutrophils, inhibiting the catalytic function of p110␤ has only a minor effect on fMLPstimulated PI(3,4,5)P 3 production and superoxide generation suggesting that p110␥ rather than p110␤ is playing the major role in G i signaling (31). The situation in platelets appears to be different, in that p110␤ appears to be the dominant PI 3-kinase isoform responsible for G i -dependent PI(3,4)P 2 accumulation. Our finding that p110␥ contributes minimally to PI(3,4)P 2 production was unexpected, particularly given the well defined ability of G␤␥ i dimers to promote p110␥ catalytic function (6). These dimers can also stimulate p110␤ catalytic function through a non-tyrosine kinase-dependent manner (6). It is noteworthy that the level of PI 3-kinase lipid products induced by G i -coupled agonists in platelets is much lower than in neutrophils, perhaps reflecting a reduced level of G i stimulation in the former cell types. It is possible that the number of free G␤␥ i dimers is the critical determinant controlling the catalytic function of p110␥ and p110␤, however, we cannot rule out the possibility that other cell type-specific adaptor molecules and/or spatial signaling mechanisms may influence the activation state of individual PI 3-kinase isoforms.
The findings presented here demonstrating that p110␥ primarily regulates integrin function through non-catalytic signaling mechanisms has potentially important therapeutic implications. It has previously been proposed (42) that targeting the catalytic function of p110␥ may represent a potentially useful method of reducing integrin ␣ IIb ␤ 3 adhesive function and thrombus formation in vivo. This is based on the observations that p110␥ Ϫ/Ϫ mice have less fatal thromboembolic complications from intravenous infusion of ADP and collagen (42) and form less occlusive thrombi following chemical injury of mouse carotid arteries (54). Whereas it remains to be established whether the defective thrombotic response observed in p110␥ Ϫ/Ϫ mice can be principally attributed to defects in G i /P2Y 12 -dependent integrin ␣ IIb ␤ 3 regulation, our studies indicate that p110␥ kinase inhibitors may have less effect on platelet adhesive function than previously anticipated. Our studies also suggest that the thrombosis defect in p110␥ Ϫ/Ϫ mice may be context dependent, i.e. in the setting of severe vascular damage (such as electrolytic injury) defects in the thrombotic response were only observed following concurrent administration of thrombin inhibitors (data in supplementary Fig. S4). These findings raise the interesting possibility that there may be a hierarchy of antithrombotic effects when targeting components of the PI 3-kinase signaling processes in platelets. For example, targeting individual signaling processes linked to either p110␥ or p110␤ may produce a weaker, albeit perhaps better tolerated antithrombotic method, whereas strategies targeting both type Ia and Ib PI 3-kinase signaling components may represent a new approach to develop potent antithrombotic agents.