Harnessing the Platelet Signaling Network to Produce an Optimal Hemostatic Response

INTRODUCTION: THE CONTRIBUTION OF PLATELETS TO THE HEMOSTATIC RESPONSE

Platelets have evolved to be the primary cellular component of the hemostatic response to injury, replacing the nucleated thrombocytes found in most nonmammalian species. After maturing from megakaryocyte-derived proplatelets, platelets circulate in an essentially quiescent state for 10 days in humans, and about half of that in mice. During that time, the quiescent state is maintained by extrinsic regulators such as nitrogen monoxide (NO) and prostacyclin (PGI₂) by the presence of healthy endothelium, which separates platelets from activators within the vessel wall; and, perhaps most critically, by the absence of physiologic platelet activators such as collagen, thrombin, and adenosine diphosphate (ADP). Hemostatic thrombi form when the vessel wall has been breached. Pathologic thrombi form in response to vessel wall disease in the arterial system, stasis and inflammation in the venous system, and the inappropriate presence of platelet activators anywhere in the vascular system. Even in a healthy vascular system, differences in local conditions within arteries and veins affect platelet activation. Because pressures and flow rates are higher, hemostasis in the arterial system requires activated platelets to form the initial obstacle to blood loss, accelerate the production of thrombin, and provide a stable base and protected environment in which fibrin can accumulate. However, platelet activation is not limited to the arterial system; it occurs in the venous system and in capillaries as well and contributes to the hemostatic response in those locations. Patients with marked thrombocytopenia develop petechiae because of small leaks in the microcirculation that are otherwise prevented by platelets.

The primary goal of the hemostatic response is to limit blood loss when penetrating injuries occur. Many of the attributes of platelets that allow them to sense and respond rapidly to vascular injury are the same attributes that contribute to heart attacks and strokes when misdirected. The hemostatic response is sometimes modeled as occurring in 3 overlapping phases: initiation, extension, and stabilization. Initiation occurs when moving platelets become tethered to von Willebrand factor (VWF)/collagen complexes within the injured vessel wall and remain in place long enough to become activated by collagen. Extension occurs when additional platelets adhere to the initial monolayer, extending it in the lateral and luminal directions. Locally generated thrombin plus ADP and thromboxane A₂ (TXA₂) released by platelets play an important role in this step, activating platelets via G protein–coupled receptors (GPCRs). Subsequent intracellular signaling activates the integrin α_IIbβ₃ on the platelet surface. Activated platelets stick to each other via bridges formed by the binding of fibrinogen, fibrin or VWF to activated α_IIbβ₃. Stabilization refers to the events that help to consolidate the platelet plug and prevent premature disaggregation, in part by amplifying signaling within the platelet. The net result of these 3 phases is a hemostatic plug comprising activated platelets and cross-linked fibrin, a structure stable enough to withstand the forces generated by flowing blood in the arterial circulation.

This model for thinking about the hemostatic response can be helpful, because it emphasizes the initial role of collagen and VWF in anchoring platelets to a site of injury and the subsequent roles of thrombin, ADP, and TxA₂. Platelets require some form of initial tethering to remain in place long enough to become fully activated. Moving platelets are pushed to the periphery of the blood stream by the flow properties of blood and the presence of red cells at the center of the blood stream. In the first seconds after injury, plasma VWF joins VWF already present in the vascular wall in becoming anchored to collagen. Flow forces working on the anchored VWF expose cryptic binding sites for platelet glycoprotein (GP) Ibα, allowing platelets to become tethered without requiring prior activation. In the absence of functional VWF, platelets can still be activated by collagen, thrombin, ADP, and TxA₂, but in vivo, they are unable to remain at the site of injury long enough for the initial layers of the hemostatic thrombus to form. Hence, VWF deficiency or dysfunction results in a bleeding disorder, as do disorders affecting secretion. As platelet activation progresses, additional VWF is released locally from platelet storage granules.¹

However, although useful in some respects, a limitation of thinking about platelet activation as successive waves of initiation, extension, and stabilization is that it does not completely describe what has been observed at sites of injury. Recent studies of platelet activation after penetrating injuries in the microcirculation show that platelet activation is not uniform throughout the hemostatic mass.^2–8 Instead of a homogeneous mass, there is a core of fully activated, stably adherent platelets overlaid by a shell of less-activated, mostly unstable platelets (Fig. 1). As the response to injury continues, the shell persists even as the stable core expands, suggesting that not all of the platelets associated with the hemostatic response become fully activated. Rather than a homogeneous mass in which all platelets are activated to the same extent and fibrin is dispersed throughout the mass, the hemostatic thrombus retains regional heterogeneity and a hierarchical structure. Fibrin is found primarily in the parts of the thrombus core that are closest to the site of injury. At present, observations such as these are still largely limited to flow chambers and the cremasteric and mesenteric microcirculations in rodent models, but they raise several questions. Do the same events occur when penetrating injuries occur in larger vessels? What normally limits the growth of the hemostatic response and prevents inappropriate vascular occlusion? What goes wrong in pathologic settings such as dyslipidemia or plaque rupture? How is the hemostatic response regulated to produce an optimal outcome?

Fig. 1.

Structural heterogeneity in the hemostatic plug. Recent observations of the platelet response to acute vascular injury in vivo show that there are distinct regions within the hemostatic thrombus where platelets are either more or less activated and where fibrin tends to accumulate. A stable core of closely packed, fully activated platelets is overlaid by an outer shell of less-stable, less-activated platelets, with a boundary zone in between. Fibrin is found within the thrombus core and in a fibrin network formed from fibrinogen that escapes before hemostasis can be achieved.

THE PLATELET SIGNALING NETWORK: THE FOREST AND THE TREES

The molecular mechanisms that underlie the hemostatic response have been the subject of intense scrutiny for many years, reflecting the relevance of this process to clinical medicine and the lives of humans. To the extent that space allows, the remainder of this article summarizes what has been learned. However, before diving into the trees, it is worth considering the forest. Largely for experimental reasons, it has generally been more practical to tease apart platelet activation 1 molecule at a time. Studies with genetically modified mice have helped this process considerably, even although mouse platelets turn out to be different from human platelets in several critical ways. The results make it possible to draw models such as the one in Fig. 2, which summarizes key steps in platelet activation. Knowledge of platelet signaling pathways has aided the development of several antiplatelet agents that have entered clinical practice or are in clinical trials.

Fig. 2.

An overview of some of the major signaling pathways of platelet activation. Targets for antiplatelet agents that are in clinical use or in clinical trials are indicated in blue. Inhibitors are in red. AA, arachidonic acid; IP₃, inositol-1,4,5-trisphosphate; PAR, protease-activated receptor; PGI₂, prostaglandin I₂; PKA, protein kinase A; PKC, protein kinase C; PLC, phospholipase C; RGS, regulator of G protein signaling; TXA₂, thromboxane A₂.

However, Fig. 2 has important limitations. One limitation is that it has little to say about the events that lead to granule secretion in activated platelets. Resting platelets store and activated platelets secrete numerous molecules that support platelet activation and the hemostatic response. Some of these molecules also help to drive events such as wound healing and angiogenesis. Patients lacking storage granules have a clinical bleeding disorder. The presence of both proangiogenic and antiangiogenic factors within platelet storage granules has fueled recent debates about differential selectivity in either the storage or the release of these molecules in vivo.^9–12

A second limitation of models such as the one shown in Fig. 2 is that it falls short in answering questions that are critical for understanding platelet activation. Why does platelet activation require such complexity? Why are there so many platelet agonists and so many signaling pathways? If they are merely redundant, why does it seem that genetic or pharmacologic ablation of so many individual molecules has a meaningful effect on platelet function? An answer to some of these questions may come from thinking about signaling networks rather than signaling pathways.

One way of illustrating such networks is shown in Fig. 3. In a network-centric view of platelet activation, instead of responding to a single agonist via a single signaling pathway, platelets accumulating at a site of injury respond to the mixture of inputs (agonists) to which they are exposed. This mixture varies over both time and space, creating agonist concentration gradients. Positive and negative feedback in the form of soluble molecules (some released by platelets, others by damaged cells) plus contact-dependent interactions between adjacent platelets play an important role in modulating platelet reactivity. Information flow within the platelet sums the effects of these inputs and nodal points in the signaling network provide links between them. This view of platelet activation can be tied to the structure shown in Fig. 1 by considering, first, how declining gradients of different platelet agonists originating near the point of injury differentially affect the platelet activation state as the distance from the point of injury grows and, second, how the closer proximity of platelets in the thrombus core facilitates the binding of cell surface ligands to receptors on the surface of adjacent platelets.

Fig. 3.

A network-centric view of platelet activation. Signaling in platelets is represented as a balanced network, in which the effects of multiple inputs (mainly agonists) are modulated by positive and negative feedback. Signaling pathways communicate with each other and share nodal points in the network. The extent of activation of any given platelet varies according to the combination and concentration of the agonists to which the platelet is exposed and the sum of the feedback to which it is subjected. Gradients of agonist concentration are expected to be highest at the vessel wall near the site of injury (bottom) and lowest in the thrombus shell (top). Referring to the model in Fig. 1, platelets in the core of the thrombus are most subject to thrombin, collagen, and contact-dependent feedback, whereas those in the shell are affected primarily by soluble mediators such as ADP and TxA₂.

NORMAL PLATELET ACTIVATORS IN VIVO

Once vascular injury has occurred, platelets are principally activated by locally exposed collagen, locally generated thrombin, platelet-derived TXA₂, and ADP that has either been released from damaged cells or secreted from platelet dense granules. The flow pattern of circulating red cells facilitates platelet adhesion to collagen by pushing platelets closer to the vessel wall, allowing GPIbα on the platelet surface to be snared by the A1 domain of VWF bound to collagen. Human platelets express 15,000 to 25,000 copies of GPIbα in a multiprotein complex with GPIbβ, GPIX, and GPV. Mutations in GPIbα that prevent its expression on the platelet surface or impair its receptor function produce an inherited recessive bleeding disorder (Bernard-Soulier syndrome) because platelet adhesion to the vessel wall is impaired. Macrothrombocytopenia arises in this disorder in part from a failure to form the normal linkages between GPIbα complex and filamin in the platelet membrane cytoskeleton.¹³

Once platelets are captured at the damaged vessel wall, the primary drivers for platelet activation are collagen, thrombin, ADP, TXA₂, and, to a more limited extent, epinephrine (see Fig. 2). With the exception of collagen, each of these agonists signals through 1 or more members of the GPCR superfamily. Properties that are common to this class of receptors make them well suited for their tasks in platelets. Most bind their ligands with high affinity and, because they act as exchange factors, each occupied receptor can theoretically activate multiple G proteins, amplifying the initial signal. Binding studies show that agonist GPCRs are expressed on the platelet surface in low numbers, ranging from a few hundred (ADP and epinephrine) to a few thousand (thrombin) per cell.^14–20 The duration of GPCR signaling is subject to receptor internalization, receptor desensitization, and the accelerated inactivation of G proteins by members of the RGS (regulator of G protein signaling) family. This multiplicity of mechanisms allows for tight regulation of platelet activation.

Human platelets express at least 10 heterotrimeric G proteins, including at least 1 G_s, 4 G_i (G_i1, G_i2, G_i3 and G_z), 3 G_q family members (G_q, G₁₁ and G₁₆), and 2 G₁₂ family members (G₁₂ and G₁₃). Much has been learned about the critical role of these G proteins by knocking out the genes encoding their α subunits in mice.²¹ Less is known about the G_βγ isoforms that are expressed in platelets and the selectivity (if any) of their individual contributions to platelet activation.

Most of the time agonist-initiated platelet activation begins with the activation of a phospholipase C (PLC) isoform, which by hydrolyzing membrane phosphatidylinositol-4,5-bisphosphate (PIP₂) produces the second messenger (inositol-1,4,5-trisphosphate [IP₃]) needed to increase the cytosolic Ca²⁺ concentration. This situation leads to integrin activation via an intracellular pathway that includes a Ca²⁺-dependent exchange factor (CalDAG-GEF), a switch (predominantly Rap1b), an adaptor (RIAM), and proteins that interact directly with the integrin cytosolic domains (kindlin and talin).²² Which isoform of PLC is activated depends on the agonist. Collagen activates PLCγ₂ using a mechanism that depends on scaffold molecules and protein tyrosine kinases. Thrombin, ADP, and TxA₂ activate PLCβ using G_q as an intermediary. The a subunit of G_q binds directly to the phospholipase, activating it.

It is the binding of bivalent fibrinogen to α_IIbβ₃ that enables platelets to stick to one another, forming an aggregate. Other proteins that can substitute for fibrinogen include fibrin, VWF, and fibronectin. Average expression levels of α_IIbβ₃ range from approximately 50,000 per cell on resting platelets to 80,000 on activated platelets. Mutations in α_IIbβ₃ that prevent its expression or suppress its function produce a bleeding disorder (Glanzmann thrombasthenia) because platelets are unable to form stable aggregates. Antiplatelet agents such as eptifibatide (Integrilin), tirofiban (Aggrastat), and abciximab (ReoPro) take advantage of this situation by blocking α_IIbβ₃.

Platelet Activation by Collagen

Under static conditions, collagen can activate platelets without the assistance of co-factors, but under arterial flow conditions, VWF plays an essential role. Collagen polymers are better platelet agonists than collagen monomers.^23,24 Platelets can adhere to monomeric collagen but require the more complex structure found in fibrillar collagen for optimal activation.²⁵ Four collagen receptors have been identified on human and mouse platelets. Two bind directly to collagen (integrin α₂β₁ and GPVI); the other 2 (α_IIbβ₃ and GPIbα) bind to collagen via VWF (Fig. 4). Of these receptors, GPVI is the most potent signaling receptor.²⁶ The structure of the GPVI extracellular domain places it in the immunoglobulin superfamily. Its ability to generate signals rests on its constitutive association with the immunoreceptor tyrosine-based activation domain (ITAM)-containing Fc receptor γ-chain (FcRγ). Platelets from mice that lack either GPVI or FcRγ have impaired responses to collagen, as do platelets in which GPVI has been depleted or blocked.^27–30 Loss of FcRγ impairs collagen responses in part because of loss of a necessary signaling element and in part because of its role in helping GPVI reach the platelet surface. α₂β₁ supports adhesion to collagen and acting as a source of further signaling after engagement.^31,32 Human platelets with reduced expression of α₂β₁ have impaired collagen responses,^33,34 as do mouse platelets that lack β₁ integrins when the ability of these platelets to bind to collagen is tested at high shear.^30,35

Fig. 4.

Platelet activation by collagen. Platelets use several different molecular complexes to support platelet activation by collagen. These complexes include (1) VWF-mediated binding of collagen to the GPIb-IX-V complex and integrin α_IIbβ₃ and (2) a direct interaction between collagen and both the integrin α₂β₁ and the GPVI/FcRγ-chain complex. Clustering of GPVI results in the phosphorylation of tyrosine residues in the FcRγ cytoplasmic domain, followed by the activation of the tyrosine kinase, Syk. One consequence of Syk activation is the activation of PLCγ₂, leading to phosphoinositide hydrolysis, secretion of ADP, and the production of TXA₂. COX-1, cyclooxygenase 1; DAG, diacylglycerol; PG, prostaglandin; PI3K, phosphatidylinositol 3-kinase; PKC, protein kinase C; PLA₂, phospholipase A₂.

Signaling through GPVI can be studied in isolation with the snake venom protein, convulxin, or with synthetic collagen-related peptides, both of which bind to GPVI, but not to other collagen receptors. Each of these entities causes clustering of GPVI, leading to the phosphorylation of FcRγ by Src family tyrosine kinases constitutively associated with a proline-rich domain in GPVI.³⁶ Phosphorylation creates an ITAM motif recognized by the tandem SH2 domains of the tyrosine kinase Syk. Association of Syk with the GPVI/FcRγ-chain complex activates Syk, produces signaling complexes based on the scaffold proteins LAT and SLP-76, and leads to the activation of PLCγ₂. Loss of Syk impairs collagen responses.²⁹ PLCγ₂ hydrolyzes PIP₂ to form IP₃ and diacylglycerol (DAG). IP₃ opens Ca²⁺ channels in the platelet dense tubular system, increasing the cytosolic Ca²⁺ concentration through passive efflux. Depletion of the Ca²⁺ stores within the dense tubular system triggers Ca²⁺ influx across the platelet plasma membrane. The changes in the cytosolic Ca²⁺ concentration that occur when platelets adhere to collagen under flow can be visualized in real time.^37,38 Diacylglycerol activates the more common protein kinase C (PKC) isoforms that are expressed in platelets, allowing the serine/threonine phosphorylation events that are needed for platelet activation.

Collectively, collagen receptors support the capture of fast-moving platelets at sites of injury, cause activation of the captured platelets, and stimulate the cytoskeletal reorganization that allows the previously discoid platelets to flatten out and adhere more closely to the exposed vessel wall. As noted earlier, VWF supports this process. It increases the density of potential binding sites for collagen per platelet, because the number of copies of GPIbα and α_IIbβ3 greatly exceeds the number of copies of GPVI and α₂β₁. Because VWF is highly multimeric, it also increases the number of binding sites per collagen molecule. This situation increases the likelihood that platelets encounter an available binding site and, once bound, are able to increase the number of contacts with collagen by bringing additional receptors into play. GPVI and GPIbα are able to bind collagen and VWF without prior platelet activation, but once activation begins, α₂β₁ and α_IIbβ₃ are able to bind their respective ligands as well. Some of the integrin-activating signaling occurs downstream of GPVI, but there is evidence that the GPIb-IX-V complex can signal as well, as can α₂β₁ and α_IIbβ₃ once they are engaged.

Platelet Activation by Thrombin

Thrombin is able to activate platelets at concentrations as low as 0.1 nM. Although other platelet agonists can also cause phosphoinositide hydrolysis, none seems to be as efficiently coupled to PLC as thrombin. Within seconds of the addition of thrombin, the cytosolic Ca²⁺ concentration increases 10-fold, triggering downstream Ca²⁺-dependent events, including the activation of phospholipase A₂ and integrin activation via the Rap1b-mediated pathway. All of these responses, but not shape change, are abolished in platelets from mice lacking G_qα.³⁹ Thrombin also activates Rho in platelets, leading to rearrangement of the actin cytoskeleton and shape change, responses that are greatly reduced or absent in mouse platelets that lack G_13α.⁴⁰ Thrombin is able to inhibit adenylyl cyclase activity in human platelets, either directly (via a G_i family member) or indirectly (via released ADP).^41,42

Platelet responses to thrombin are mediated by members of the protease-activated receptor (PAR) family of GPCRs. There are 4 members of this family, 3 of which (PAR1, PAR3, and PAR4) can be activated by thrombin. PAR1 and PAR4 are expressed on human platelets; mouse platelets express PAR3 and PAR4. Receptor activation occurs when thrombin cleaves the extended N terminus of each of these receptors, exposing a new N terminus that serves as a tethered ligand.⁴³ Synthetic peptides based on the sequence of the tethered ligand domain of PAR1 and PAR4 are able to activate the receptors, mimicking at least some of the actions of thrombin. PAR3 was originally identified after gene ablation studies showed that platelets from mice lacking PAR1 were still fully responsive to thrombin.⁴⁴ Approximately half of PAR1^−/− mice die in utero, but this seems to be caused by loss of receptor expression in the vasculature, rather than in platelets.⁴⁵ Whereas human PAR3 has been shown to signal in response to thrombin, PAR3 on mouse platelets primarily serves to facilitate PAR4 cleavage.⁴⁶ Activation of PAR4 requires higher concentrations of thrombin than are required for activation of PAR1, apparently because it lacks the hirudinlike sequences that can interact with the anion-binding exosite of thrombin and facilitate receptor cleavage.^46–49 Kinetic studies in human platelets suggest that thrombin signals first through PAR1 and subsequently through PAR4.^50,51

Considerable evidence shows that PAR family members are sufficient to activate platelets. Peptide agonists for either PAR1 or PAR4 cause platelet aggregation and secretion.^43,48 Conversely, simultaneous inhibition of human PAR1 and PAR4 abolishes responses to thrombin,⁵² as does deletion of the gene encoding PAR4 in mice.⁵³ However, there are differences between mice and humans: optimal thrombin responses in human platelets require both PAR1 and PAR4, whereas those in mouse platelets are mediated by PAR3-facilitated cleavage of PAR4. PAR1 and PAR4 are coupled to at least G_q and G₁₃. There is conflicting evidence about whether G_i-dependent signaling in thrombin-activated platelets is entirely mediated by secreted ADP or occurs in part by a direct interaction between PARs and G_i2.^54–56 However, given the ability of PAR1 to directly couple to G_i2 in cells other than platelets, it is likely that both occur. This hypothesis is consistent with a recent observation that expression of an RGS-resistant variant of G_i2α enhances thrombin responses in mouse platelets, even when the contribution of secreted ADP is blocked.⁵⁷ On the other hand, a requirement for PAR family members does not preclude the involvement of other participants in platelet responses to thrombin. GPIbα has a high affinity thrombin binding site within residues 268 to 287.^58,59 Deletion or blockade of this site reduces platelet responses to thrombin, particularly at low thrombin concentrations,^60–64 and impairs PAR1 cleavage on human platelets.⁶³ PAR1 antagonists have been developed and are undergoing clinical trials.⁶⁵ The early results show that blocking PAR1 may have some clinical usefulness, but can also increase bleeding risk in patients receiving other antiplatelet agents.^66–68

Platelet Activation by ADP

ADP is stored in platelet dense granules and released on platelet activation. It is also released from damaged cells at sites of vascular injury, serving as a stimulus for activating tethered platelets and stabilizing the hemostatic plug. Aggregation studies show that all of the other platelet agonists are dependent to some extent on released ADP to elicit maximal platelet aggregation, although this dependence varies with the agonist and is dose related. Drugs that block platelet ADP P2Y₁₂ receptors have proved to be effective antiplatelet agents despite the fact that ADP, when added alone, is a less potent platelet agonist than thrombin.

When added to platelets in vitro, ADP causes an increase in cytosolic Ca²⁺, TXA₂ formation, protein phosphorylation, shape change, aggregation, and secretion. It also inhibits cyclic adenosine monophosphate (cAMP) formation. These responses are half-maximal at approximately 1 μM ADP. However, even at high concentrations, ADP is a comparatively weak activator of PLC. Instead, its usefulness as a platelet agonist rests more on its ability to activate other pathways.^69,70 Human and mouse platelets express 2 distinct receptors for ADP, denoted P2Y₁ and P2Y₁₂. Both receptors are members of the purinergic class of GPCRs.^{17–19,71,72} P2Y₁ receptors couple to G_q. P2Y₁₂ receptors couple to G_i family members. Optimal activation of platelets by ADP alone requires activation of both receptors, but the potentiation by ADP of platelet responses to other agonists seems to be mainly caused by P2Y₁₂. Knockouts of P2Y₁ and P2Y₁₂ in mice produce effects consistent with those predicted by pharmacologic studies on human platelets.^17,54 A third purinergic receptor on platelets, P2X₁, is an adenosine trisphosphate (ATP)-gated Ca²⁺ channel.^73–76 Platelet dense granules contain ATP as well as ADP, and studies suggest that there are conditions in which P2X₁ activity is essential for platelet activation.^77–79 P2X₁ is also functional on megakaryocytes.⁸⁰

When P2Y₁ is blocked or deleted, ADP is still able to inhibit cAMP formation, but its ability to cause an increase in cytosolic Ca²⁺, shape change, and aggregation is greatly impaired, as it is in platelets from mice that lack G_qα.³⁹ P2Y₁^−/− mice have a minimal increase in bleeding time and show some resistance to thromboembolic mortality after injection of ADP, but no predisposition to spontaneous hemorrhage. Primary responses to platelet agonists other than ADP are unaffected and when combined with serotonin, which is a weak stimulus for PLC in platelets, ADP can still cause aggregation of P2Y₁^−/− platelets. Taken together, these results show that platelet P2Y₁ receptors are coupled to G_qα and responsible for activation of PLC. P2Y₁ receptors can also activate Rac and the Rac effector, p21-activated kinase (PAK),⁸¹ but do not seem to be coupled to G_i family members.

P2Y₁₂ was independently identified by 2 groups.^19,72 As had been predicted by inhibitor studies and by the phenotype of a patient lacking functional P2Y₁₂,⁸² platelets from P2Y₁₂^−/− mice do not aggregate normally in response to ADP.⁸³ P2Y₁₂^−/− platelets retain P2Y₁-associated responses, including shape change and PLC activation, but lack the ability to inhibit cAMP formation in response to ADP. The G_i family member associated with P2Y₁₂ seems to be primarily G_i2, because platelets from G_i2α^−/− mice have an impaired response to ADP,^84,85 whereas those lacking G_i3α or G_zα do not.^85,86 Conversely, expression of a G_i2α variant that is resistant to the inhibitory effects of RGS proteins produces a gain of function in mouse platelets stimulated with ADP.⁵⁷ Absence of P2Y₁₂ produces a hemorrhagic phenotype in humans, albeit a mild one.^19,82,87 Deletion of either P2Y₁ or P2Y₁₂ in mice prolongs the bleeding time and impairs platelet responses not only to ADP but also to thrombin and TXA₂, particularly at low concentrations.^83,88,89 Because the receptors for thrombin and TXA₂ can cause robust activation of PLC, the contribution of ADP when thrombin or TXA₂ are present seems to be largely because of its ability to activate G_i2.⁹⁰ Downstream effectors for G_i2α include Src family members⁹¹ as well as PI3 kinase and Rap1b.

Platelet Activation by TXA₂

TXA₂ is produced from arachidonate in platelets by the aspirin-sensitive cyclooxygenase 1 (COX-1) pathway. When added to platelets in vitro, stable TXA₂ analogues such as U46619 cause shape change, aggregation, secretion, phosphoinositide hydrolysis, protein phosphorylation, and an increase in cytosolic Ca²⁺ and have little, if any, direct effect on cAMP formation. Similar responses are seen when platelets are incubated with exogenous arachidonate.⁹² TXA₂ can diffuse across the plasma membrane and activate nearby platelets.⁹³ Like secreted ADP, release of TXA₂ amplifies the initial stimulus for platelet activation and helps to recruit additional platelets.⁹⁴ This process is effective locally, but is limited by the brief (~30 second) half-life of TXA₂ in solution, helping to confine the spread of platelet activation to the original area of injury.

Only 1 gene encodes TXA₂ receptors, but 2 splice variants are produced that differ in their cytoplasmic tails. Human platelets express both.⁹⁵ Loss of G_qα abolishes U46619-induced IP₃ formation, but does not prevent shape change,³⁹ which can be blocked by knocking out G_13α.⁴⁰ Although in cells other than platelets, TXA₂ receptors have been shown to couple to G_i family members,⁹⁶ in platelets the inhibitory effects of U46619 on cAMP formation seem to be largely mediated by secreted ADP. These observations have previously been interpreted to mean that platelet TXA₂ receptors are coupled to G_q and G_12/13, but not to G_i family members. However, the enhanced response observed in mouse platelets carrying an RGS protein-resistant G_i2α variant suggests that this is still an open issue.⁵⁷

The biological relevance of TXA₂ to normal platelet function is supported by genetic and pharmacologic evidence. TP^−/− mice have a prolonged bleeding time. Their platelets are unable to aggregate in response to TXA₂ agonists and show delayed aggregation with collagen, presumably reflecting the role of TXA₂ in platelet responses to collagen.⁹⁷ A group of Japanese patients with impaired platelet responses to TXA₂ analogues have proved to be either homozygous or heterozygous for an R60L mutation in the first cytoplasmic loop of TP.⁹⁸ However, the most compelling case for the contribution of TXA₂ signaling in human platelets comes from the successful use of aspirin as an antiplatelet agent. When added to platelets in vitro, aspirin abolishes TXA₂ generation. Aspirin also blocks platelet activation by arachidonate and impairs responses to thrombin and ADP. The defect in thrombin responses appears as a shift in the dose/response curve, indicating that TXA₂ generation is supportive of platelet activation by thrombin, but not essential.

Platelet Activation by Epinephrine

Compared with thrombin, epinephrine is a weak activator of human platelets when added on its own. Nonetheless, there are reports of human families in which a mild bleeding disorder is associated with impaired epinephrine-induced aggregation and reduced numbers of catecholamine receptors.^99,100 Epinephrine responses in platelets are mediated by α_2A-adrenergic receptors.^20,101,102 In both mice and humans, epinephrine is able to potentiate the effects of other agonists so that the combination is a stronger stimulus for platelet aggregation than either agonist alone. Potentiation is often attributed to the ability of epinephrine to inhibit cAMP formation, but as is discussed later, there are clearly other effects as well. In contrast to other platelet agonists, epinephrine has no detectable direct effect on PLC and does not cause shape change, although it can trigger phosphoinositide hydrolysis indirectly by stimulating TXA₂ formation.¹⁰³ Taken together, these results suggest that platelet α_2A-adrenergic receptors are coupled to G_i, but not G_q or G₁₂ family members. Human and mouse platelets express 4 members of the G_i family: G_i1, G_i2, G_i3, and G_z.¹⁰⁴ Of these members, G_zα has the slowest rate of intrinsic guanosine triphosphate (GTP) hydrolysis and is the only one that is not a substrate for pertussis toxin. Knockout studies show that epinephrine responses in platelets are abolished when G_zα is deleted. Loss of G_i2α or G_i3α has no effect.^85,86 G_z also seems to be responsible for the ability of epinephrine to activate Rap1b.^85,105 Therefore, it seems that in mouse platelets, α_2A-adrenergic receptors couple to G_z, but not G_i2 or G_i3.

CRITICAL EVENTS

Thus far, this review has focused on platelet activation mechanisms from the perspective of individual agonists, many of which trigger a common set of pathways. In this section, some of those pathways are considered.

Phosphoinositide Hydrolysis

With the exception of epinephrine, platelet activation begins with the activation of PLC which, by hydrolyzing membrane phosphatidylinositol-4,5-bisphosphate (PIP₂), produces IP₃ and (DAG), the second messengers needed to increase the cytosolic Ca²⁺ concentration and activate some of the PKC isoforms found in platelets. As already noted, how PLC is activated varies with the agonist. Collagen activates PLCγ₂ using adaptor molecules and tyrosine kinases. Thrombin, ADP, and TXA₂ activate PLCβ isoforms using G_qα and (less efficiently if at all) G_βγ derived from G_i. Regardless of which PLC isoform is activated, the subsequent increase in the Ca²⁺ concentration triggers downstream events, including integrin activation and TXA₂ formation. Thrombin provides a robust stimulus for phosphoinositide hydrolysis and causes the largest and fastest increase in cytosolic Ca²⁺. Collagen and ADP are more dependent on the synthesis and release of TXA₂ to achieve a maximal response.

Cytosolic Ca²⁺ and Integrin Activation

Resting platelets maintain their cytosolic free Ca²⁺ concentration at approximately 0.1 μM by (1) limiting Ca²⁺ influx and (2) pumping Ca²⁺ out of the cytosol across the plasma membrane or into the dense tubular system (smooth endoplasmic reticulum). This action consumes ATP. In activated platelets, [Ca²⁺]_i can spike 10-fold to greater than 1 μM with potent agonists like thrombin. Measurements made of large populations of platelets suggest that this increase occurs rapidly and uniformly in response to potent agonists such as thrombin and less potently with agonists such as ADP and collagen. However, observations made with single platelets suggest that the Ca²⁺ response is heterogeneous, occurring to a different extent in different platelets, with some showing spiking behavior rather than a sustained increase and some not responding at all.^106–108 The molecular basis of this heterogeneity is not fully understood, although simulation studies suggest that platelet size is a factor.¹⁰⁸

The increase in cytosolic Ca²⁺ that occurs during platelet activation derives from 2 sources. The first part is caused by the IP₃-mediated release of Ca²⁺ from the platelet dense tubular system. The second part occurs when depletion of the dense tubular system Ca²⁺ pool triggers store operated Ca²⁺ entry through a conformational change in STIM1, a protein in the dense tubular system membrane, binding of STIM1 to Orai1 in the plasma membrane, allowing an influx of plasma Ca²⁺.¹⁰⁹ These events can be separated by chelating extracellular Ca²⁺, thereby precluding Ca²⁺ influx. The increase in [Ca²⁺]_i activates the GTP-binding protein, Rap1b, via the Ca²⁺-dependent guanine nucleotide exchange factor (GEF), CalDAG-GEF.¹¹⁰ Activated Rap1b has been shown to bind to RIAM, bringing it to the plasma membrane, where it can bind talin. This situation allows talin to bind to the cytoplasmic domain of α_IIbβ₃, triggering integrin activation and exposing a binding site for fibrinogen.²²

TXA₂ Production

The initial events of platelet activation lead to the activation of phospholipase A₂, which cleaves phosphatidylcholine and other membrane phospholipids, liberating arachidonate from the C2 position of the glycerol backbone. Arachidonate can be transformed into many bioactive compounds, but in platelets, TXA₂ is the key product. TXA₂ synthesis begins with COX-1, which forms prostaglandin (PG) G₂ and PGH₂ from arachidonate in a 2-step process (see Fig. 4). The PGH₂ is then metabolized to TXA₂ by thromboxane synthetase.¹¹¹ Evidence suggests that phospholipase A₂ activation in platelets can occur in more than 1 way. It can clearly happen in response to an increase in cytosolic Ca²⁺, because the addition of a Ca²⁺ ionophore is sufficient to cause phospholipase A₂ activation.¹¹² There is also evidence that mitogen-activated protein kinase pathway signaling activates phospholipase A₂, although not necessarily by direct phosphorylation of phospholipase A₂.^113,114 TXA₂ formation is promoted by platelet aggregation and may, therefore, also occur as a consequence of integrin signaling, although this has not been tested directly.¹¹⁵ TXA₂ can diffuse out of the platelet, activating nearby receptors in an autocrine or paracrine fashion before it is hydrolyzed to inactive TXB₂. Even although TXA₂ is a potent platelet activator, its half-life in aqueous solution is short (about 30 seconds), which has implications for both its duration of action and its impact downstream from a growing thrombus. This situation may be especially relevant for the model of the hemostatic response shown in Fig. 1, which postulates that declining gradients of soluble platelet agonists originating in the thrombus core contribute to the heterogeneity in platelet activation that we and others have observed within the hemostatic mass.

Shape Change and Rearrangement of the Actin Cytoskeleton

Activated platelets lose their characteristic discoid shape, either assuming a globular shape with filopodial extensions if they are activated in suspension or flattening out and developing lamellopodia if they are activated on a suitable surface. Shape change reflects loss of the circumferential microtubule ring and rearrangement of the actin cytoskeleton of the platelet, a process mediated by monomeric G proteins in the Rho and Rac families. When platelets in suspension are activated by soluble agonists, shape change precedes platelet aggregation. The soluble agonists (thrombin, ADP, and TXA₂) that trigger shape change typically act through receptors that are coupled to members of the G_q and G₁₂ families. Epinephrine is unable to cause shape change, because its receptors are coupled solely to G_z.

At least 2 pathways are involved in the reorganization of the actin cytoskeleton: Ca²⁺-dependent activation of myosin light-chain kinase downstream of G_q family members and activation of Rho family members downstream of G₁₃.^21,116 Several proteins having both G_α-interacting domains and GEF domains can link G₁₂ family members to Rho family members, including p115RhoGEF.¹¹⁷ With the exception of ADP, shape change persists in platelets from mice that lack G_qα³⁹ but is lost when G_13α expression is suppressed, alone or in combination with G_12α.⁴⁰ A combination of inhibitor and genetic approaches suggests that G₁₃-dependent Rho activation leads to shape change via pathways that include the Rho-activated kinase (p160ROCK) and LIM kinase.^116,118,119 Activation of these kinases results in phosphorylation of myosin light-chain kinase and cofilin, helping to regulate both actin filament formation and myosin. ADP, on the other hand, depends more heavily on G_q-dependent activation of PLC to produce shape change and is able to activate G₁₃ only as a consequence of TXA₂ generation; hence, the loss of ADP-induced shape change when G_q signaling is suppressed.

PI3 Kinase Activation

An additional category of critical events needed for robust platelet activation involves activation of the PI3-kinase (PI3K) isoforms expressed in platelets, either by G_i family members (PI3Kγ) or by phosphotyrosine-dependent signaling pathways downstream of collagen receptors (PI3Kαβδ). PI3-kinases phosphorylate PI-4-P and PI-4,5-P₂ to produce PI-3,4-P₂ and PI-3,4,5-P₃, respectively. Among the best-described consequence of PI3K activation in platelets is the activation of the protein kinase, Akt. Knockout and inhibitor studies show that all 3 Akt isoforms are necessary for normal platelet activation.^120–122 Knockout and inhibitor studies also show a G_i/PI3K-dependent mechanism for activating Rap1b,^85,105 which, given the clinical usefulness of P2Y₁₂ antagonists as antiplatelet agents, seems to be a mechanism that is important for stabilizing platelet aggregates. However, aside from Rap1b and GSK3β, a kinase the activity of which is negatively regulated by Akt, little is known about Akt-dependent signaling pathways in platelets. To our knowledge, the Akt-dependent GEF for Rap1b has yet to be identified. PI3Kβ has been proposed as a target for antiplatelet agents.¹²³

CONTACT-DEPENDENT AND CONTACT-FACILITATED SIGNALING

As platelet activation proceeds, platelets that were previously mobile come into increasingly stable contact with each other, eventually with sufficient proximity that molecules on the surface of 1 platelet can interact directly with molecules on the surface of adjacent platelets.¹²⁴ Although in theory this situation could occur anywhere within the growing hemostatic thrombus, it is likely to occur most readily in the thrombus core, where platelets are most densely packed (see Fig. 1).⁸ Stable cohesive contacts between platelets require engagement of α_IIbβ₃ with 1 of its ligands, after which inward-directed (ie, outside-in) signaling occurs through the integrin. Inward-directed signaling also occurs through other surface molecules, which can engage their counterparts on the surface of adjoining platelets. Some of these molecules have been shown to affect platelet activation and thrombus stability, either positively or negatively. Others serve primarily to help form contacts between platelets and create a protected space in which soluble molecules, including agonists, can accumulate. In the next section, several examples are considered in which contact-dependent signaling has been studied in platelets and shown to have a meaningful impact.

Outside-In Signaling by Integrins

Activated α_IIbβ₃ bound to fibrinogen, fibrin, or VWF provides the dominant cohesive force that holds platelet aggregates together. It also contributes a further impetus for sustained platelet activation by serving as a scaffold for the assembly of signaling molecules.^125–127 The term outside-in signaling refers to the effects of these molecules.¹²⁸ Integrin signaling depends in large part on the formation of protein complexes that link to the integrin cytoplasmic domain. Some of the protein-protein interactions that involve the cytoplasmic domains of α_IIbβ₃ help regulate integrin activation; others participate in outside-in signaling and clot retraction. Proteins that are reportedly capable of binding directly to the cytoplasmic domains of α_IIbβ₃ include β₃-endonexin, CIB1, talin, kindlin, myosin, Shc, and the tyrosine kinases, Src, Fyn, and Syk. As noted earlier, the binding of talin and kindlin is believed to be one of the final events in the allosteric regulation of integrin activation.^129–132 Some other interactions require the phosphorylation of tyrosine residues Y773 and Y785 (Y747 and Y759 in mice) in the β₃ cytoplasmic domain by Src family members. Substitution of Y747 and Y759 with phenylalanine produces mice with platelets that tend to disaggregate and that show impaired clot retraction and a tendency to rebleed from tail bleeding time sites.¹³³ Fibrinogen binding to the extracellular domain of activated α_IIbβ₃ stimulates an increase in the activity of Src family kinases and Syk.^133,134 Studies of platelets from mice lacking these kinases suggest that these events are required for the initiation of outside-in signaling and for full platelet spreading, irreversible aggregation, and clot retraction.^135–138 There is evidence that the ITAM-containing receptor, FcRγIIA, is the link between Src family kinase and Syk activation in human platelets activated by α_IIbβ₃, although if so, a different receptor must perform this role in mouse platelets, which do not express FcRγIIA.¹³⁹

Contact-Dependent Ligand-Receptor Interactions

The proximity of 1 platelet to another can also permit the direct binding of cell surface ligands to cell surface receptors on adjacent platelets. Two examples that we have studied are the interactions of ephrinB1 and sema4D with their cognate receptors. Ephrins are cell surface molecules attached by either a GPI anchor (ephrin [Eph] A family members) or a single transmembrane domain (ephrin B family members) that serve as ligands for cell surface receptor tyrosine kinases in the Eph A and B families. Eph kinases, like Eph B family members, have a single transmembrane domain that includes the catalytic domain and protein/protein interaction motifs. Human platelets express at least 2 Eph kinases, EphA4 and EphB1, and their ligand, ephrinB1.¹⁴⁰ Forced clustering of either EphA4 or ephrinB1 results in cytoskeletal changes leading to platelet spreading, as well as increased adhesion to fibrinogen, Rap1b activation, and granule secretion.^140,141 EphA4 associates with α_IIbβ₃ and Eph/ephrin interactions promote phosphorylation of the β₃ cytoplasmic domain. Conversely, blockade of Eph/ephrin interactions impairs clot retraction and causes platelet disaggregation at low agonist concentrations, resulting in impaired thrombus growth.^140,142 It also inhibits platelet accumulation on collagen under flow.^140,142 These observations suggest that close contacts between platelets during the early stages of thrombus formation allow the binding of ephinB1 to EphA4 and EphB1 in trans to provide signaling for sustained integrin activation.

Semaphorin 4D (Sema4D, CD100) provides another example of a ligand that is involved in contact-dependent signaling in platelets. Like the ephrins, semaphorins are best known for their role in the developing nervous system,¹⁴³ but they have also been implicated in organogenesis, vascularization, and immune cell regulation.¹⁴⁴ The 25 members of the semaphorin family are defined by a large sema domain¹⁴⁵ and divided into 8 classes differing in part by how (and whether) they are anchored to the surface of the cell that expresses them.¹⁴⁶ Sema4D, which has a transmembrane domain, is expressed on the surface of both mouse and human platelets. Platelet activation causes a rapid increase in surface expression, after which there is a gradual decline below the original expression level as the sema4D exodomain is cleaved and shed by the metalloprotease, ADAM17.¹⁴⁷ Inhibiting or deleting ADAM17 blocks this process. Sema4D^−/− mouse platelets have a defect in their responses to collagen and convulxin in vitro, and a reduced response to vascular injury in vivo.¹⁴⁷ Responses to thrombin, ADP, and TXA₂ mimetics are normal.^147,148 The collagen defect has been mapped to a failure to maximally activate Syk downstream of the collagen receptor, GPVI. Events in the pathway upstream of Syk occur normally in sema4D^−/− platelets.^147,148 These defects are observed only when platelets come in contact with each other and can be reversed by adding soluble recombinant sema4D. Thus, the evidence suggests that sema4D provides a contact-dependent boost in collagen signaling. The platelets that are most likely to be activated via GPVI are those in contact with collagen. If so, then the sema4D signaling is likely most relevant for the platelets closest to the vessel wall at the site of injury. Human platelets express at least 2 receptors for sema4D: CD72 and a member of the plexin B family.¹⁴⁷ Mouse platelets express the plexin, but not CD72. However, although the presence of these Sema4D receptors has been established, the extent to which they mediate Sema4D signaling in platelets remains to be determined.

OBTAINING AN OPTIMAL RESPONSE TO INJURY

The molecular mechanisms that drive platelet activation reflect an evolutionary compromise. This compromise can be thought of as establishing a threshold for platelet activation. If the threshold is too high, then platelets become useless for hemostasis. If too low, then opportunities for unwarranted platelet activation increase (Fig. 5). An optimal platelet response to injury can be defined as one in which blood loss is restrained and hemostasis is achieved without the penalty of further tissue damage caused by unwarranted vascular occlusion. The set point is established by balancing the intracellular signaling mechanisms that drive platelet activation in response to injury with regulatory mechanisms that either dampen those responses or prevent their initiation in the first place. The normal environment of the platelet includes the other blood cells, the soluble molecules found in plasma and, most critically, the vascular wall and the endothelial cells that line it, all of which are subject to change in the face of injury, disease, circadian rhythms, and aging. Many of these entities serve as extrinsic regulators of platelet activation. A healthy endothelium provides a physical barrier that limits platelet activation. It also produces inhibitors of platelet activation, including NO,¹⁴⁹ PGI₂,^150–152 and the ecto-ADPase, CD39, which hydrolyzes plasma ADP that would otherwise sensitize platelets to activation by other agonists.¹⁵³ In addition to these extrinsic regulators, several intrinsic regulators of platelet activation have been identified. A few of them are considered here.

Fig. 5.

Formation of an optimal platelet plug. Vascular injury produces a hemostatic response that can be too aggressive (leading to occlusion and ischemia), inadequate (leading to further bleeding), or optimal. This model suggests that an optimal response varies in detail (hence, a range of normal), but is best viewed as a response that results in hemostasis with a minimum of blood loss and an avoidance of unwarranted vascular occlusion. In the setting of a vascular wall disease such as atherosclerosis, the rapid accumulation of platelets on top of a ruptured plaque represents an escape from normal restraints.

Regulation of G Protein-Dependent Signaling

As noted earlier, most of the platelet agonists act via GPCRs. Because mechanisms exist that can limit the activation of GPCRs, platelet activation can be tightly regulated. The role of RGS proteins in platelets is just beginning to be explored. RGS proteins limit signaling intensity and duration by accelerating the intrinsic hydrolysis of GTP by activated G protein α subunits.¹⁵⁴ As many as 10 RGS proteins have been identified in platelets at the RNA level, but to our knowledge, only RGS10 and RGS18 have been confirmed at the protein level.^155–160 The evidence that RGS proteins are biologically relevant in platelets comes from studies on mice in which glycine 184 in the α subunit of G_i2 has been replaced with serine, rendering it unable to interact with RGS proteins without impairing the ability of G_i2 to interact with either receptors or downstream effectors. This substitution produces the predicted gain of platelet function in vitro and in vivo, even in the heterozygous state.⁵⁷ At the molecular level, the G_i2α(G184S) allele causes an attenuated increase in cAMP levels in response to PGI₂ and a substantial increase in basal Akt activation, 2 events that occur downstream of G_i2 in platelets. In contrast, agonist-stimulated increases in [Ca²⁺]_i and Rap1 activation are unaffected, indicating no crossover into G_qα-dependent signaling pathways.

These results show that removing the restraining hand of RGS proteins on G_i2 in platelets is sufficient to produce a pronounced gain of function, arguing that the normal role of the RGS proteins is to limit platelet activation. If so, then the timing of the interaction of RGS proteins with activated G proteins becomes critical as well. If signaling is constrained too early, then platelet activation might never occur, whereas if signaling is shut down too late, more platelets might accumulate than are needed. Two recent studies show that RGS proteins interact with other molecules the role of which may be to regulate RGS protein availability. The first of these proteins is the scaffold protein, spinophilin, which binds RGS10 and RGS18 in resting platelets, and then releases them on platelet activation. Knocking out spinophilin in mice produces a loss of function in platelets, as would be expected if RGS10 and RGS18 interact prematurely with activated G proteins when spinophilin is not present.¹⁶¹ A second group of RGS protein-interacting proteins in platelets includes members of the 14-3-3 scaffold protein family. 14-3-3 family members can bind to phosphorylated serine residues in RGS18, an interaction that is displaced when cAMP-dependent protein kinase A phosphorylates RGS18.¹⁶² Taken together, these recent observations suggest that spinophilin serves as a critical node within the platelet signaling network.

cAMP and Protein Kinase A

Rising cAMP levels turn off signaling in platelets. Regulatory molecules released from endothelial cells cause G_sα-mediated increases in adenylyl cyclase activity (PGI₂) and inhibit the hydrolysis of cAMP by phosphodiesterases (NO).¹⁶³ When added to platelets in vitro, PGI₂ can cause a greater than 10-fold increase in the platelet cAMP concentration, but even small increases in cAMP levels (2-fold or less) impair platelet activation.¹⁶⁴ Deletion of either G_i2α or G_zα causes an increase in the basal cAMP concentration in mouse platelets.⁸⁵ cAMP phosphodiesterase inhibitors such as dipyridamole act as antiplatelet agents by raising cAMP levels. Conversely, loss of PGI₂ receptor (IP) expression causes a decrease in basal cAMP levels, an enhanced response to agonists, and a predisposition to thrombosis in murine arterial injury models.^85,165 Despite ample evidence that cAMP inhibits platelet activation, the mechanism by which it does this is not fully defined. cAMP-dependent protein kinase A is believed to be involved, but other mechanisms may be as well. Substrates for the kinase in platelets include serine and threonine residues in some GPCRs, IP₃ receptors, GPIbβ, vasodilator-stimulated phosphoprotein, RGS18, and Rap1b, but it remains unclear whether a single substrate or small group of substrates accounts for most of the effect or whether it is the accumulated effect of phosphorylation of many substrates.^{162,166–174}

Adhesion/Junction Receptors Contribute to Contact-Dependent Signaling

In addition to amplifying platelet activation, contact-dependent signaling can also help to limit thrombus growth and stability. Examples of this phenomenon include platelet/endothelial cell adhesion molecule-1 (PECAM-1), carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1) and at least 2 members of the CTX family of adhesion molecules, ESAM and JAM-A. PECAM-1 is a type 1 transmembrane protein with 6 extracellular domains, the most distal of which can form homophilic interactions in trans.¹⁷⁵ The cytoplasmic domain contains 2 immunoreceptor tyrosine inhibitory motifs (ITIMs) that can bind the tyrosine phosphatase, SHP-2.¹⁷⁶ PECAM-1-deficient platelets show enhanced responses to collagen in vitro and in vivo.^177–179 CEACAM1 is a second ITIM family member expressed on the platelet surface that can form homophilic and heterophilic interactions with other CEACAM superfamily members. In contrast to PECAM-1, the CEACAM1 ITIMs prefer SHP-1 over SHP-2, although either can become bound.^180,181 The CEACAM1 knockout, like the PECAM-1 knockout, produces an enhanced platelet response, showing increased platelet activation in vitro in response to collagen and increased thrombus formation in a FeCl₃ injury model.¹⁸² ESAM and JAM-A have 2 extracellular immunoglobulin domains, a single transmembrane region and a cytoplasmic tail of varying length that terminates in a binding site for PDZ domain-containing proteins. ESAM is associated with a-granules in resting platelets, but localizes to junctions during platelet activation.¹⁸³ ESAM^−/− platelets show increased aggregation in response to low doses of GPCR agonists and are more resistant to disaggregation compared with wild-type platelets. In vivo, ESAM^−/− mice form larger and more stable thrombi compared with their wild-type counterparts.¹⁸³ Collectively, these studies suggest that ESAM negatively regulates platelet function through contact-dependent homophilic interactions, although the mechanistic basis for these observations remains to be identified. JAM-A^−/− mice also show a gain of function phenotype.¹⁸⁴

SUMMARY

Platelets are part of the initial response to vascular injury intended to resist further blood loss by producing a hemostatic plug or thrombus. Conventional models portray such thrombi as relatively uniform masses of fully activated platelets intermixed with fibrin. Recent observational studies suggest that reality is more complex in at least the microcirculation, with fibrin deposition concentrated close to the vessel wall at the site of injury and the extent of platelet activation varying inversely with distance from the injury such that a hierarchical structure is formed. Underlying platelet activation is flexible signaling network within the platelets that sums the response to multiple platelet agonists, including collagen, thrombin, ADP, TxA₂, and epinephrine. Critical events in this network include the activation of PLC, an increase in cytosolic Ca²⁺, reorganization of the platelet cytoskeleton, granule secretion, and the steps that link Ca²⁺ to activation of the platelet fibrinogen receptor, α_IIbβ₃, so that platelet aggregates can form. Extrinsic regulators provided by a healthy endothelium dampen platelet responsiveness and resist unwarranted platelet activation, whereas intrinsic regulators such as RGS proteins limit signaling downstream of platelet agonists once it has been initiated. As activated platelets come into increasingly stable contact with each other, molecules on their surface are able to interact directly, generating a wave of contact-dependent signaling events. Such events can amplify platelet activation, but can also help to mold the growing thrombus by limiting the extent of platelet activation.

KEY POINTS.

• Circulating platelets possess the ability to undergo rapid activation at sites of vascular injury and resist unwarranted activation, which can lead to heart attacks and strokes.

• The signaling mechanisms underlying the regulation of platelet activation have been approached as a collection of individual pathways unique to agonist.

• This review takes a different approach, casting platelet activation as the product of a signaling network, in which activating and restraining mechanisms interact in a flexible network that regulates platelet adhesiveness, cohesion between platelets, granule secretion, and the formation of a stable hemostatic thrombus.