Platelet-collagen interaction: is GPVI the central receptor?

Vessel wall injury triggers sudden platelet activation and platelet plug formation, followed by coagulant activity and the formation of fibrin-containing thrombi that occlude the site of injury. These events are crucial to limit blood loss at sites of tissue trauma but may also block diseased vessels, leading to ischemia and infarction of vital organs. One of the major clinical problems in the developed world is arterial thrombosis caused by rupture or erosion of an atherosclerotic plaque, leading to platelet adhesion and subsequent thrombus formation in coronary and cerebral arteries causing myocardial infarction and stroke, respectively. Therefore, a detailed understanding of the mechanisms underlying the formation of the atherosclerotic plaque as well as (arterial) thrombosis is required in order to control ischemic cardiovascular diseases while retaining hemostasis.

The first step in the hemostatic cascade is platelet interaction with the exposed extracellular matrix (ECM) at sites of injury. Among the macromolecular constituents of the ECM, collagen is considered to play a major role in this process, as in vitro it not only supports platelet adhesion through direct and indirect pathways but it also directly activates the cells initiating aggregation and coagulant activity.¹  Platelet adhesion and aggregation on collagen is an integrated process that involves several platelet agonists that act through a variety of surface receptors, including integrins, immunoglobulin (Ig)–like receptors, and G-protein–coupled receptors. Over the last 20 years, immense effort has been spent on the identification of these receptors and their individual contribution to the complex processes of platelet tethering, adhesion, secretion, aggregation, and coagulant activity on collagen. However, the multiplicity of candidate collagen receptors and lack of detailed knowledge of the molecular events that underlie these responses have severely hampered developments in this field.

Platelet-collagen interactions are believed to have the greatest significance at the medium and high shear rates found in arteries and diseased vessels. At the very high shear rates found in small arteries and arterioles, the rapid onset of interaction between glycoprotein Ib-V-IX (GPIb-V-IX) and von Willebrand factor (VWF) immobilized on collagen is crucial for the initial tethering (or capture) of flowing platelets.^2,3  The interaction between VWF and GPIb-IX-V, however, is rapidly reversible and insufficient for stable adhesion. This can be illustrated by the rolling of GPIb-IX–expressing cells or platelets on a VWF monolayer at high shear. Rapid conversion to stable adhesion requires additional contacts between the platelet and the ECM. Integrins are recognized as the major class of surface receptor mediating stable adhesion at high shear in hematopoietic cells. Integrins are heterodimeric proteins consisting of α and β subunits. They are present on the surface of most cells in a resting conformation, which has low affinity for their natural ligand, but they can be converted to a state of high affinity through “inside-out” signals generated by other surface receptors. Collagen binds directly or indirectly to 2 platelet integrins namely α₂β₁ and α_IIbβ₃ (via VWF), respectively. Either of these interactions is sufficient to convert rolling of platelets on collagen/VWF to stable adhesion. However, for this to take place, at least one of these integrins must undergo conversion to the high-affinity conformation in response to inside-out signals. Although there are many stimuli that have the potential to mediate this activation, it is noteworthy that collagen is the most reactive component of the ECM inducing integrin activation through the Ig receptor, GPVI. Thus, it is now recognized that firm adhesion on collagen under high shear requires intracellular signals from GPVI, and that this is reinforced by release of soluble mediators, the most important of which are adenosine diphosphate (ADP) and thromboxane A₂ (TxA₂), and by the generation of thrombin.

This new appreciation of the role of GPVI in mediating integrin activation has led to revision of the so-called “2-site, 2-step” model, which now places a critical role for GPVI in the initial interaction with collagen and upstream of adhesion. This review will discuss this new role of GPVI in platelet activation and define outstanding questions and future directions in this rapidly progressing field.

A thorough understanding of the molecular events that underlie platelet activation by collagen can be achieved only by consideration of the available experimental tools. This includes many different types, preparations, and species forms of collagen, receptor-specific synthetic collagens, snake toxins, and a rapidly increasing number of antibodies with the ability to inhibit or mimic collagen-receptor interactions. Each of these tools has important advantages and disadvantages that must be appreciated in using them to evaluate the role of collagen receptors in platelet-collagen interaction.

There are genes for more than 20 forms of collagen in the human genome of which 9 have been identified to be expressed in the vessel wall, namely types I, III, IV, V, VI, VIII, XII, XIII, and XIV. Fibrillar types I and III are the major constituents of the ECM of blood vessels and have been the focus of most attention. In addition, the network-forming type IV collagen is the major form in the subendothelial basement membrane. Collagens consist of repeat GXY motifs where G is glycine and X and Y are frequently proline (amino acid code, P) and hydroxyproline (amino acid code, O). The sequence GPO makes up approximately 10% of collagens I and III. The GXY repeat sequence forms a single left-handed helix that associates with 2 other chains to form a right-handed super-helix. In collagens I and III, the chains have approximately 1000 amino acids flanked by short nonhelical N- and C-terminal telopeptide extensions. The cross-linking of these monomeric collagen structures forms fibrillar collagen, the predominant structure that platelets come into contact with in the ECM. Fibrillar collagen usually consists of more than one collagen type along with other matrix components. The utility of preparations of fibrillar collagen to simulate the behavior of subendothelial matrix collagen is limited because of this imprecise composition. Collagen fibers can be broken down to their monomeric forms by peptidases such as pepsin that target the nonhelical sequences in collagens, whereas specific collagenases are required to solubilize the parent molecule. The pepsin-generated monomers can be cross-linked to form collagens that lack the nonhelical regions, although it is important to recognize that the degree of cross-linking can have a marked influence on the biologic activity. Collagens are isolated from tissues and therefore represent a mixture of various types, although they are usually enriched in 1 or 2 particular forms as determined by the tissue source. Collagens are also prepared from several species, usually bovine, equine, and rat (but seldom human), and the same type may therefore have important differences in sequence that could influence activity. The most commonly used preparation of collagen for platelet studies, “Horm” collagen, is a suspension of fibrils made up of equine collagen type I and a small amount of equine type III, along with low levels of other ECM proteins.

The differences between the various preparations of collagens can have important experimental implications. In addition, the mode of presentation of collagen can also influence the biologic activity. For example, monomeric collagen interacts selectively with the integrin α₂β₁ in suspension and requires cross-linking or immobilization to a surface at a sufficient density to stimulate platelet activation via GPVI. This is probably due to the low affinity of the GPVI-binding motif, GPO (see the next paragraph), for the glycoprotein receptor and the relatively low frequency of this motif in monomeric collagen. When presented as a monolayer, platelets may be able to bind to the GPO motif in more than one molecule of the monomeric peptide, bringing about a net increase in the interaction with collagen. These considerations illustrate the need for the use of appropriate physiologic systems and reagents to establish the molecular basis and significance of platelet-collagen interactions.

There is considerable interest in the development of synthetic collagens of defined composition and in the synthesis of receptor-selective peptides. This is not a trivial matter in that it is generally recognized that the peptides need to be present in helical form to maintain activity at collagen receptors. A breakthrough in this area came from the observation that GPP or GPO repeats of 5 or more spontaneously form the helical structures that are present in collagen chains. The synthesis of peptides based on these sequences by Morton et al in Cambridge led to the unexpected observation that a peptide with a repeat GPO motif, cross-linked by N- and C-terminal cysteine or lysine residues, was a powerful platelet agonist, whereas cross-linked GPP was inactive.⁴  The GPO-containing peptide was termed collagen-related peptide (CRP) with CRP-XL being recommended for the cross-linked form. Significantly, CRP-XL is unable to support adhesion under the same stringent conditions used to show adhesion to collagen, but can induce platelet activation in the presence of α₂β₁-blocking antibodies.⁴  CRP-XL was therefore the first selective agonist to be identified for the major collagen signaling receptor in the platelet, which was later recognized to be GPVI.^5-7  Confirmation that the activity of CRP-XL was due to the GPO repeat motif and independent of the N- and C-terminal cross-linking groups was subsequently shown by the synthesis of a GPO repeat peptide that lacked these regions.⁸ 

Knight et al later synthesized an α₂β₁-specific peptide by introducing the α₂β₁-reactive sequence, GFOGER, into a backbone of GPP to confer a helical sequence. GPP itself is inactive at α₂β₁ and GPVI.⁹  The activity at α₂β₁ is critically dependent on the presence of the GER group, which can also be preceded by other sequences within collagen, although these combinations have a lower affinity for the integrin.¹⁰  Importantly, the GFOGER helical peptide supports α₂β₁-mediated adhesion.¹¹  Commercial preparations of the receptor-specific peptides are not available, but several groups have had the collagen peptides made and performed the cross-linking themselves.

A number of snake venom peptides that mediate their actions through GPVI have been identified in recent years and have proved to be powerful tools in the study of the Ig receptor. The snake C-type lectin convulxin was the first identified member of this group and was instrumental in the cloning of the protein.^12,13  A number of additional C-type lectins from snake venoms have since been shown to activate GPVI, along with the metalloproteinase alborhagin, which binds to a distinct site on the glycoprotein.¹⁴  The snake venom toxins are multimeric and induce platelet activation by clustering the receptor. Several of these snake toxins bind to a second surface glycoprotein in addition to GPVI. For example, alboaggregin A binds to GPVI and GPIb, inducing powerful activation.^15,16  The interaction with GPIb, however, is not essential for activation, as alboaggregin A activates Bernard-Soulier platelets and cell lines transfected with GPVI.¹⁶  Nevertheless, it may be the norm that snake venom toxins mediate their effects through binding to more than one platelet surface glycoprotein. The reader is referred to a recent review for further information on the action of snake venom toxins on GPVI and other platelet glycoproteins (Andrews et al¹⁷ ).

A number of antibodies to GPVI have been raised within the last few years. GPVI-specific antibodies are powerful platelet agonists when cross-linked by secondary antibodies, but on their own several of them can serve as receptor antagonists. The antibodies are powerful tools used to identify the sequences within the extracellular domain of GPVI that confer binding to collagen.

GPVI was first identified as a 60- to 65-kDa platelet glycoprotein by 2-D gel electrophoresis more than 20 years ago.¹⁸  The first indication that GPVI may be an important platelet receptor for collagen, however, came from studies on a patient who presented to the clinic with an autoimmune thrombocytopenia caused by autoantibodies to a 65-kDa protein that was present in healthy individuals but absent in the patient.¹⁹  Gel electrophoresis (2-D) was used to demonstrate that the antiserum recognized GPVI. Platelets from this patient were unresponsive to collagen, whereas activation by other stimuli was normal. A F(ab)₂ preparation of the IgG fraction from the patient was found to strongly activate platelets from healthy individuals, whereas monovalent F(ab) fragments inhibited collagen-induced activation. A small number of additional patients with low levels of GPVI have been described.^6,20,21  In most cases, the patients display a mild bleeding phenotype and their platelets exhibit defective aggregation to collagen. The early studies on the GPVI-deficient patients provided compelling evidence for a key role of the glycoprotein in platelet activation by collagen, but this was not initially recognized because of the multiplicity of other candidates for this role, most notably α₂β₁.

A new era on collagen receptors followed the discovery of the GPO-based CRP-XL and the demonstration that the peptide and also collagen induce platelet activation through a tyrosine kinase–based signaling pathway that involves the kinase Syk and phospholipase Cγ2 (PLCγ2).^22-27  This work led to the discovery that collagen and the CRPs stimulate tyrosine phosphorylation of the Fc receptor (FcR) γ-chain that contains an immunoreceptor tyrosine-based activation motif (ITAM)²⁸  and that this is present as a complex with GPVI.^7,29  The importance of the FcR γ-chain and Syk in platelet activation by collagen and CRP-XL was shown by the loss of activation of platelets by collagen and CRP-XL in “knock-out” murine platelets that lacked either of the 2 signaling proteins.³⁰  Together, this work identified collagen as a unique platelet agonist in that it is the only stimulus that activates platelets at sites of vascular injury through an ITAM-regulated pathway. The snake venom toxin convulxin was subsequently recognized to activate platelets through GPVI, supported by the observation that it stimulates tyrosine phosphorylation of the FcR γ-chain and Syk.^12,13  Convulxin was first shown to be a powerful platelet agonist almost 20 years before,³¹  but its mode of action had remained unclear despite the observation that its action resembled that of collagen.

The first antimouse GPVI monoclonal antibody (mAb), JAQ1, was reported in 2000.³²  JAQ1 was detected during the screening of a panel of rat mAbs that had been generated against murine platelet membranes. The antibodies were screened against FcR γ-chain–deficient platelets in anticipation that the glycoprotein receptor would be absent from the surface of the platelets, as shown for other receptors that couple to the ITAM-containing protein.^33,34  In confirmation, the rat mAb JAQ1 recognizes a single band of 65 kDa by Western blot that comigrates with GPVI and is absent in FcR γ-chain–deficient murine platelets.³²  JAQ1 inhibits aggregation of normal mouse platelets induced by collagen and CRP-XL but rapidly induces aggregation when cross-linked by a secondary antibody.³²  Together, these findings established GPVI as the major activating collagen receptor on mouse platelets.

Several groups have now confirmed the inactivity of Horm collagen on FcR γ-chain–deficient platelets.^30,35,36  More recently, Jarvis et al have shown that platelets deficient in FcR γ-chain are refractory to stimulation by collagen types I to V, strongly suggesting that the GPVI/FcR γ-chain complex is a key receptor for all types of collagen.³⁷  However, the defects in these platelets cannot be directly related to the absence of GPVI, as the function of other receptors may also be affected by the absence of the FcR γ-chain, most notably GPIb.^36,38  More direct evidence for an essential role of GPVI in platelet activation by collagen comes from studies with mice in which GPVI was depleted by injection of JAQ1 in vivo.³⁹  This treatment leads to a transient GPVI deficiency while not affecting other receptors including α₂β₁, α_IIbβ₃, GPIb, and GPV. Initial studies indicated that this GPVI loss occurs through internalization and proteolytic degradation of the receptor,³⁹  but further studies will be required to confirm this hypothesis. Like FcR γ-chain–deficient platelets, such GPVI-depleted platelets are refractory to activation by CRP-XL and collagen.

The phosphorylation events evoked by CRPs, snake toxins, and cross-linked GPVI antibodies are qualitatively similar to those induced by collagen, although the intensity of response is considerably greater. The GPVI-specific stimuli are all multimeric in nature and have the capacity to induce the formation of clusters of GPVI on the platelet surface, thereby generating a powerful intracellular signal. On the other hand, the GPVI-specific motif, GPO, is present at a level of approximately 10% in collagens and may not be appropriately spaced to induce an equivalent degree of clustering. In this context, it is noteworthy that F(ab′)₂ fragments of JAQ1, which can induce only receptor dimerization, stimulate a weak intracellular signal that is sufficient to synergize with G_i-based signaling pathways but alone is unable to induce shape change, aggregation, or release.⁴⁰  These findings question the extent to which these GPVI-specific stimuli can be used to mimic signaling by collagen, an issue that is discussed later in the review.

In October 1999, Clemetson et al reported the cloning of human GPVI, revealing it to be a member of the Ig receptor superfamily.⁴¹  The sequence of GPVI is most closely related to human FcαR and natural killer (NK) cell receptors as well as to polymorphic mouse receptors known as paired Ig-like receptors (PIRs).⁴¹  Human GPVI is composed of 339 amino acids but displays an apparent molecular weight of 62 kDa in sodium dodecyl sulfate–polyacrylamide gel electrophoresis due to glycosylation. Human GPVI contains 2 Ig-C2–like extracellular domains formed by disulfide bonds, a mucinlike stalk, a transmembrane region, and a short 51–amino acid cytoplasmic tail (Figure 1). In a single publication, a splice variant, GPVI-II, has been described in human erythroleukemic megakaryocyte-like cells (HEL cells) that lack 18 amino acids in the mucin stalk.⁴²  A further splice variant, GPVI-III, that lacks the transmembrane domain was also reported in this study. A number of polymorphic variations at the human GPVI locus have been identified (Figure 1),⁴³  but it is unclear whether these variations significantly influence structure and function of the receptor, although there is evidence that at least one polymorphic form influences susceptibility to myocardial infarction as discussed in “The clinical significance of GPVI and α₂β₁ in hemostasis and thrombosis.” Mouse GPVI has 319 amino acids and shares 64.4% and 67.3% identity with human GPVI at the protein and nucleotide levels, respectively.⁴⁴  Human GPVI has an intracellular tail of 51 amino acids. Murine GPVI has an intracellular tail of 27 amino acids that lacks the 24 amino acids that lie C-terminal to the proline-rich region in human GPVI (Figure 1).⁴⁴ 

The transmembrane and cytoplasmic tail of human GPVI has been shown to contain distinct regions that mediate association with other proteins. GPVI has a positively charged arginine in its transmembrane region that is essential for association with the FcR γ-chain.^45,46  A transmembrane arginine is also found in other receptors that associate with the FcR γ-chain including FcγRI, FcγRIII, FcϵRI, FcαR, PIR-A, and NKR-P1 (CD161).^33,47,48  In addition, there is evidence that the first 6 juxtamembrane amino acids are essential for the interaction with the FcR γ-chain⁴⁹  (Figure 1). The GPVI cytosolic tail contains a proline-rich motif that binds selectively to the Src homology 3 (SH3) domain of the Src family tyrosine kinases, Fyn and Lyn.⁵⁰  This domain has recently been shown to be necessary for intracellular signaling through GPVI in Jurkat and COS-7 cells.⁵⁰  It is proposed that cross-linking of GPVI brings SH3-associated Fyn or Lyn to the FcR γ-chain, enabling both phosphorylation of the ITAM to take place and initiation of the signaling cascade. The constitutive association of Fyn and Lyn with GPVI may place the receptor in a “ready-to-go” state, enabling rapid activation on exposure to collagen as a similar mechanism for activation does not seem to be present in most other ITAM-coupled receptors. The cytoplasmic part of GPVI also contains a calmodulin-binding domain.⁵¹  Calmodulin is constitutively associated with GPVI in platelets and undergoes delayed dissociation upon activation, although the functional significance of this is not known.

Signaling through GPVI/FcR γ-chain occurs via the same pathway as that used by immunoreceptors, with the FcR γ-chain serving as the signal transducing part of the receptor complex. Crosslinking of GPVI leads to tyrosine phosphorylation of the FcR γ-chain on its ITAM by the Src kinases Fyn and Lyn (Figure 2).^52-54  This leads to binding and subsequent activation of the tandem SH2 domain–containing tyrosine kinase, Syk, which initiates a downstream signaling cascade that culminates in activation of a number of effector enzymes including PLCγ2, small G-proteins, and phosphoinositide-3 kinase. The adapters LAT and SLP-76 play critical roles in this signaling cascade. Several groups have shown that GPVI signaling cascade takes place in cholesterol-rich membrane domains known as Gems or rafts.^55-57 

Direct confirmation of GPVI as a collagen receptor has come from studies on transfected cell lines. Clemetson et al first showed that DAMI cells, a megakaryocytic cell line with a low level of GPVI expression, become responsive to collagen when transfected with GPVI and show enhanced responses to the more powerful convulxin.⁴¹  Interestingly, however, several other groups were unable to show activation of GPVI by collagen in megakaryocytic cells and in other transfected lines despite obtaining robust responses to convulxin.^45,46,58  This was partially due to a low level of expression of GPVI, as RBL-2H3 cells, which express a comparable level of GPVI to that found in platelets, were subsequently shown to be responsive to collagen.⁵⁹  The ability to obtain activation by convulxin in cells that express a low level of the receptor can be explained by its much greater affinity for GPVI and its multimeric nature. It is important to stress that RBL-2H3 cells do not express α₂β₁ as well as a number of other putative collagen receptors and that this is also likely to have influenced the need for such a high level of expression, bearing in mind that murine platelets with 20% of endogenous level of GPVI can be fully activated by collagen.⁶⁰ 

There is evidence that collagen interacts with GPVI at 2 distinct sites.⁶¹  JAQ1 completely blocks activation of platelets by the GPO-rich peptide CRP, suggesting that the antibody occupies the “CRP binding site,” which is very likely identical to the major collagen binding site on the receptor. However, the blocking action of JAQ1 can be overcome at high collagen concentrations indicating a second site of interaction. When one considers that FcRγ-deficient platelets do not respond to collagen, this finding suggests that there is a second FcRγ-coupled receptor on the platelet surface or that collagen interacts with a second site on GPVI. Of these, the latter appears to be the more likely, as platelets in which GPVI has been depleted by JAQ1 in vivo are unresponsive to collagen.³⁹  An alternative explanation for these results that is consistent with a single binding site for collagen on GPVI is that the inhibitory effect of JAQ1 is mediated through an allosteric effect that reduces the net affinity of GPVI for collagen and CRP. This would be equivalent to lowering the number of GPVI receptors on the platelet surface. This has been shown to have a much greater effect on responses to CRP-XL than to collagen.⁶⁰ 

Convulxin, CRP-XL, and cross-linked antibodies such as JAQ1 elicit rapid, powerful aggregation of platelets and are able to induce full aggregation in the presence of inhibitors of cyclo-oxygenase and antagonists of the 2 major ADP receptors, P2Y₁ and P2Y₁₂.^40,62,63  In contrast, collagen induces aggregation after a delay of at least 20 to 30 seconds through a pathway that is primarily mediated by release of ADP and TxA₂.^40,63  This is illustrated by the inhibitory effects of cyclo-oxygenase inhibitors such as aspirin⁶⁴  and the ADP receptor antagonists clopidogrel⁶⁵  and the AR-C series of compounds.⁶⁶  Furthermore, a major characteristic of patients with storage pool disease, in which ADP is lacking in the dense granules, is an impaired aggregation to collagen.⁶⁷  Similarly, Pearl mice, which are deficient in dense granules, also exhibit a marked inhibition of aggregation to collagen.⁶³ 

The receptors for ADP and TxA₂ couple to several heterotrimeric G-proteins, namely G_i,G_q,G₁₂, and G₁₃.⁶⁸  The importance of G_q-coupled receptors for collagen-induced platelet aggregation is highlighted by the finding that platelets from Gα_q-deficient mice display a severely impaired aggregation response when stimulated with low or intermediate concentrations of collagen.⁶⁹  However, aggregation induced by GPVI-specific agonists is only marginally inhibited by the absence of Gα_q.⁴⁰  Interestingly, Gα_q-deficient platelets are able to aggregate in response to very high collagen concentrations, but this process is also strictly dependent on the release of ADP, acting via P2Y₁₂,⁷⁰  and TxA₂, acting via G₁₂/G₁₃.⁷¹  Thus, the robust aggregation induced by collagen is blocked in the presence of inhibitors of secondary mediators and is replaced by a slow increase in light transmission, which most likely represents a combination of α₂β₁-mediated adhesion and α_IIbβ₃-dependent aggregation.⁷² 

It is important to consider why collagen but not GPVI-specific agonists are dependent on the release of these mediators to induce aggregation. There are likely to be several factors that contribute to this: (1) Collagen is unable to induce the same strength of intracellular signal as that induced by powerful GPVI-specific agonists such as CRP-XL, convulxin, and cross-linked antibodies,^5,61,73  most likely because of the infrequent spacing of GPO throughout its sequence. On the other hand, it stimulates a similar level of tyrosine phosphorylation to that induced by concentrations of the GPVI-selective agonists that are sufficient to induce aggregation, suggesting that this alone cannot account for the differential reliance on secondary mediators. (2) Fibrillar collagen is a relatively bulky material that is presented to platelets in vivo as a monolayer. When presented in suspension, platelets do not gain uniform access to the fibrillar collagen in contrast to soluble released mediators or the GPVI-specific agonists. This can be revealed by flow cytometric analysis of collagen-stimulated platelets, which allows examination of the activation state of single cells in diluted suspensions while excluding the accumulation of released mediators. Under these conditions, only 10% to 15% of normal platelets are directly activated (as shown by α_IIbβ₃ activation and P-selectin expression) upon incubation with concentrations of fibrillar collagen (50 μg/mL) that are 50-fold in excess of those required to induce aggregation (1 μg/mL).^35,71  A similar result is obtained with Gα_q-deficient platelets, demonstrating that G_q-mediated signaling is not required for activation of platelets on collagen.⁷¹  Thus, even at very high concentrations, collagen activates only a fraction of platelets directly and is reliant on release of secondary mediators to activate further cells in the suspension. (3) A major difference compared with the GPVI-specific agonists is the presence of additional receptors for collagen on the cell surface, most notably the integrin α₂β₁ and also the integrin α_IIbβ₃, which serves as an indirect receptor via VWF. It is significant that the affinity of α_IIbβ₃ and α₂β₁ for their ligands is increased by inside-out signals from a variety of agonists including ADP, TxA₂, and collagen (via GPVI).^35,74-76  Activation of α₂β₁ and α_IIbβ₃ in this way promotes firm platelet-collagen interactions and therefore stabilization of GPVI-collagen interactions leading to enhanced signaling.^77,78  The ability of secondary mediators and the integrin α₂β₁ to potentiate signals from GPVI in this way is illustrated by measurement of protein tyrosine phosphorylation in washed platelets. The stimulation of increases in tyrosine phosphorylation by collagen is inhibited at early times by the presence of inhibitors of thromboxane formation, receptor antagonists at the P2Y₁ and P2Y₁₂ receptors, and by α₂β₁-blockade, with recovery seen at later times. The significance of this observation should be seen in the context that none of these receptors stimulates a comparable increase in tyrosine phosphorylation on its own, suggesting that the modulation of the response to collagen is indirect, most likely through activation of α₂β₁. Thus, secondary mediators increase the rate at which GPVI binds to collagen. This is discussed further in the context of the modified 2-site, 2-step model in “Conclusions.”

Together, these observations suggest that the dependency of aggregation by collagen on secondary mediators reflects 2 independent processes. First, both ADP and TxA₂ activate platelets that do not come into direct contact with the collagen in suspension. This process mimics the situation of in vivo thrombus formation, in which only the first layer of platelets is activated by collagen, whereas thrombus growth is predominantly mediated by soluble agonists in combination with VWF and fibrinogen. Second, the mediators may potentiate the activation of integrins α₂β₁ and α_IIbβ₃ on collagen-adherent platelets, thereby leading to robust association of collagen with GPVI and enhanced signaling. The G_i-coupled P2Y₁₂ receptor plays a particularly important role in this process, as its signaling pathway synergizes with signals from GPVI to promote integrin activation.⁴⁰ 

It has been accepted for many years that GPVI is essential for platelet activation by collagen but that α₂β₁ is required for adhesion. In support of this, platelets adhere extremely weakly to CRP-XL under flow but adhere strongly to collagen through an integrin-dependent pathway.⁷⁹  It was therefore an unexpected finding that FcR γ-chain–deficient and GPVI-depleted mouse platelets show virtually no adhesion to collagen under static and flow conditions even though they express normal levels of the major adhesion receptors α₂β₁, GPIb-V-IX, and α_IIbβ₃.³⁵  Defective adhesion of GPVI/FcRγ-deficient platelets to collagen under static conditions through α₂β₁ and α_IIbβ₃/VWF was restored in the presence of Mn²⁺, which is known to directly activate integrins,³⁵  and by agonists that activate integrins, such as ADP, via inside-out signaling (B.N. et al, unpublished observations, May 2002; and Inoue et al⁸⁰ ). This demonstrates that the role of GPVI is to generate intracellular signals that promote integrin activation rather than to serve as an adhesion receptor. This is supported by studies in mice in which adhesion to collagen under static conditions is maintained in platelets that express 50% or 20% of the control level of GPVI.⁶⁰  Low levels of adhesion are also observed in platelets in which GPVI has been substantially (> 98%), but not completely, depleted from the surface by in vivo administration of JAQ1.³⁹  These platelets retain only trace amounts of GPVI on their surface, and yet this is sufficient to mediate adhesion under static conditions, demonstrating that even a very low copy number of the glycoprotein is sufficient to induce integrin activation, most likely through release of ADP and thromboxanes. Importantly, this residual adhesion is blocked in the presence of JAQ1.³⁵  Under moderate to high shear flow conditions, however, GPVI-depleted platelets are unable to adhere to collagen, demonstrating that strong integrin activation is essential for shear-resistant adhesion.³⁵ 

Initial studies with human platelets did not identify a critical role for GPVI in platelet adhesion to collagen, reporting only moderate adhesion defects of GPVI-deficient platelets.^20,81  Goto et al, however, have recently reported complete abolition of adhesion of platelets from 2 GPVI-deficient patients to collagen at high shear.⁸²  A possible explanation for these discrepancies is that the GPVI-deficient platelets used in the former studies expressed sufficient levels of the receptor to support adhesion. In support of this, we see substantial adhesion of murine platelets that express 20% of the normal level of GPVI at an intermediate rate of shear (800 s^–1) (D. Best and S. P. W. unpublished data, June 2002).

Together, these observations provide compelling evidence that cellular activation is a prerequisite for platelet adhesion to collagen and that GPVI plays the central role in this process, but that GPVI-mediated activation is reinforced by ADP and thromboxanes in vivo.

Integrin α₂β₁ (also known as platelet GPIa/IIa or lymphocyte VLA-2) was the first collagen receptor to be identified on platelets^83,84  and is known to mediate adhesion in a Mg²⁺-dependent manner.⁸⁵  For a long time, α₂β₁ was considered to be the major receptor for collagen on the platelet surface supporting adhesion and activation, and considered to play a key role in hemostasis. This was largely the result of studies on 2 patients with reduced levels of α₂β₁ who suffered from posttraumatic bleeding and excessive menorrhagia and whose platelets failed to respond to collagen.^83,86  However, aggregation to collagen in one patient could be restored by the addition of thrombospondin-1,⁸⁶  and the second was later shown to have defective adhesion to a number of adhesion molecules, demonstrating that the defect was not specific to collagen.⁸⁷  Although it is unclear why addition of thrombospondin-1 was able to restore collagen responses, this observation demonstrated that α2 deficiency was not responsible for the loss of collagen responses.

Studies in which α₂β₁ has been inhibited by antibodies or proteolytic snake toxins have produced conflicting results on the role of the integrin. While some authors reported that inhibition of α₂β₁ markedly reduced or abolished adhesion and aggregate formation in stasis and flow,^88-90  others found only minor effects of such treatment on adhesion to collagen^91-93  and collagen-induced aggregation.⁹⁴  Several factors are likely to have contributed to these differences, including the nature of the response under investigation; the experimental conditions, notably the presence or absence of plasma; and the preparation of collagen. These are each discussed in further detail in this section. Additionally, the mode of α₂β₁ inhibition may have influenced the results. For example, a number of anti-α₂β₁ agents have also been reported to weakly inhibit platelet activation by GPVI-specific agonists such as convulxin and CRP,^13,89,94  suggesting either an inhibitory effect of the integrin on GPVI-dependent signals or a nonspecific action. Perhaps the clearest indication of a nonspecific inhibitory effect of anti-α₂β₁ treatment came from studies with the snake venom toxin rhodocytin, a powerful platelet agonist. Several investigators demonstrated that an α₂β₁-blocking antibody inhibits rhodocytininduced platelet activation and aggregation, suggesting that α₂β₁ is crucial in this activation process.^95,96  However, it was later shown that α₂β₁-deficient platelets respond normally with unaltered concentration-response relationships to rhodocytin, thereby excluding a role for the integrin.⁹⁷ 

It is now recognized that the type of collagen largely determines the requirement for α₂β₁. Whereas α₂β₁ is essential for platelet adhesion and activation on monomeric type I collagen in stasis and flow, it is dispensable for these processes on native fibrillar collagen.^3,93,98  This is elegantly exemplified by Savage et al who found no effect of α₂β₁ inhibition on adhesion of human platelets under low and high shear flow conditions to fibrillar type I collagen, whereas an absolute dependence of α₂β₁ function was found for adhesion and thrombus growth on pepsin-solubilized type I collagen under the same conditions.⁹²  Similar observations have been reported by Nieswandt et al using β₁-deficient murine platelets.³⁵ 

Many studies on platelet-collagen interactions have been performed with washed platelets in order to control experimental conditions and to avoid problems encountered with plasma proteins such as nonspecific protein binding of pharmacologic agents and difficulties in biochemical extractions. However, the absence of plasma limits the role of other adhesion receptors for collagen, namely GPIb and α_IIbβ₃, which bind indirectly to collagen via VWF.³  VWF is present at a high level in plasma and is also released from α granules during platelet activation, and both of these sources play a significant role in tethering and firm adhesion during thrombus formation.^35,99  These differences therefore explain why α₂β₁ is of greater importance for adhesion of washed platelets to collagen under flow, whereas the integrin is less important in whole blood in the presence of circulating VWF.^3,100 

A different picture is found under static conditions in that washed platelets can firmly adhere to fibrillar collagen by α₂β₁-independent mechanisms, although the integrin also contributes to this process depending on the extent of washing and density of collagen.^{35,93,98,101,102}  α₂β₁-Independent adhesion can also be supported by the interaction of α_IIbβ₃ with VWF released from α granules, but this requires prior generation of strong intracellular signals mediated by other receptors, notably GPVI. This requirement may contribute to the absolute dependency on α₂β₁ for adhesion to monomeric versus fibrillar collagen, in that monomeric collagen is a weak agonist at GPVI.

A key development in the understanding of the role of α₂β₁ in platelet activation came through the demonstration by Jung and Moroi^74,75  that the affinity of the integrin for collagen is regulated by intracellular signals. Using a novel collagen-binding assay, they showed that several platelet agonists, including ADP, thromboxanes, and GPVI-specific stimuli, increase the affinity of α₂β₁ for monomeric or soluble collagen from a low to an intermediate or high-affinity state.¹⁰³  This observation was extended by the demonstration that the defective adhesion of GPVI-deficient platelets to collagen can be restored by direct activation of α₂β₁ by Mn²⁺,³⁵  or by exogenously added agonists (B.N. et al, unpublished data May 2002; and Inoue et al⁸⁰ ). In retrospect, the finding that α₂β₁ exists in more than one affinity state was not surprising, as stimulus-dependent affinity modulation is a general property of many integrins, although in many cases this is assumed rather than supported by experimental evidence. Importantly, this work led to revision of the original 2-site, 2-step model and the proposal that the initial interaction of collagen through GPVI leads to activation of integrins α₂β₁ and α_IIbβ₃, and that this in turn mediates stable adhesion to collagen and thereby reinforces the signaling through GPVI.^35,58 

The generation of mice deficient in the integrin β₁ subunit (and therefore α₂β₁, α₅β₁, and α₆β₁)orthe α₂ subunit has shed new light on this field.^35,100,102  While these studies confirmed previous reports of a delay in platelet aggregation to collagen in plasma and a more pronounced phenotype under washed conditions, these mouse strains have normal tail-bleeding times, indicating that α₂β₁ may not be not essential for hemostasis. However, the utility of tail-bleeding assays in predicting the events that underlie normal hemostasis is not established. There is also no clear correlation between bleeding time and bleeding risk.¹⁰⁴ 

In vitro whole blood perfusion studies showed that α₂β₁-deficient platelets adhere and aggregate on fibrillar (Horm) collagen at low and high rates of shear to a similar extent as wild-type controls, presumably through α_IIbβ₃/VWF interactions.^3,35  In contrast, adhesion of α₂β₁-deficient platelets to monomeric collagen, which binds weakly to GPVI, is abolished under the same experimental conditions, confirming that the nature of the collagen preparation determines the significance of the integrin.⁹²  In the absence of plasma (and therefore extracellular VWF), the adhesion of α₂β₁-deficient platelets to collagen under flow is largely inhibited, confirming that the integrin serves as an adhesion collagen receptor.¹⁰⁰  In static conditions, however, washed α₂β₁-deficient platelets can adhere to fibrillar but not monomeric collagen,^35,102  reflecting the Mg²⁺-independent adhesion previously reported by others.^93,98,101  Together, these findings demonstrate that α₂β₁ plays a crucial role for platelet adhesion to collagen in washed platelets under flow conditions, whereas it may not be strictly required for these processes under physiologic conditions (ie, in plasma) because of the presence of extracellular VWF.

It is well established that integrins generate intracellular signals in a wide variety of cells, including platelets, as exemplified by α_IIbβ₃.¹⁰⁵  It is therefore important to consider whether the role of α₂β₁ is limited solely to adhesion or whether it also generates intracellular signals that regulate activation. This is difficult to study using collagen because of the powerful stimulatory role of GPVI and the need for prior activation of the integrin to promote stable binding to collagen. Studies using α₂β₁-blocking antibodies or platelets deficient in either subunit of the integrin are also difficult to interpret because of the effect on adhesion.

It is possible to overcome these issues to a large degree using α₂β₁-specific ligands. Knight et al¹¹  used the α₂β₁-specific GFOGER peptide described in “Collagens, synthetic collagens, and GPVI-specific ligands” to probe the role of the integrin in platelets. Surprisingly, they reported that the peptide has no effect on functional responses and tyrosine phosphorylation in platelet suspensions despite being able to bind to the integrin under these conditions.¹⁰⁶  Hers et al⁹⁴  made a similar conclusion through examination of the actions of a number of α₂β₁-specific antibodies in platelet suspensions, many of which have been shown to activate the integrin in other systems.

In contrast to these studies, Suzuki-Inoue et al described the spreading of platelets on F(ab′)₂ fragments of the activating antibody TS2/16.¹⁰⁷  Adhesion to TS2/16 led to formation of filopodia and lamellapodia through a pathway that involves activation of Rac and possibly CDC42.¹⁰⁷  More recently, platelets were shown to generate filopodia and lamellapodia on a monolayer of GFOGER peptide in the presence of inhibitors of secondary mediators, even though the peptide had no effect on platelets in suspension.⁸⁰  A similar set of observations was also made following adhesion of GPVI-FcR γ-chain–deficient platelets to collagen following exposure of the platelets to ADP or the phorbol ester phorbol myristate to activate α2β1 and thereby promote binding to the ECM protein. The α2β1-blocking antibody 6F1 was used to confirm that adhesion to collagen under these conditions was mediated through the integrin. The GFOGER peptide induces spreading through the same signaling pathway as that used by the integrin α_IIbβ₃ in platelets,^57,108  namely through activation of Src and Syk family kinases leading to tyrosine phosphorylation of PLCγ2.⁸⁰  This pathway has many similarities with the signaling pathway used by GPVI in platelets including a pivotal role for PLCγ2.

We speculate that these α₂β₁-mediated signals may be important in stabilizing the growing thrombus on collagen. In support of this, Kuijpers et al⁷⁷  found that although β₁-deficient platelets can adhere to collagen at high shear, the newly generated aggregates display a looser structure and embolize more frequently compared with controls. Furthermore, both collagen-induced calcium mobilization and the development of procoagulant activity are signifi-cantly reduced in β₁-deficient platelets. These results strongly suggest that the role of α₂β₁ is not limited solely to adhesive interactions but that it also generates intracellular signals that help to stabilize the thrombus.

Platelet adhesion to collagen under high shear (> 600 s^–1) requires VWF immobilized on collagen via its A3 domain. This interaction is mediated through the N-terminal 45-kDa domain of GPIbα, which carries the binding site for VWF as shown by monitoring adhesion of cells and coated beads to a VWF-coated surface.^109,110  This interaction is essential for the initial capture or tethering of platelets by VWF and is critically dependent on the fast on-rate of association between VWF and GPIbα.³  This interaction, however, also has a fast off-rate of association that leads to rolling of platelets on a VWF surface for several minutes until α_IIbβ₃-mediated stable adhesion (via VWF) is seen.¹¹¹  In contrast, stable adhesion occurs rapidly on a collagen-coated surface through integrins α₂β₁ and α_IIbβ₃.^3,35 

These observations raise the question as to the receptors and their pathways that mediate activation of α_IIbβ₃ and α₂β₁ by the VWF and collagen/VWF surfaces. GPIb has been reported to induce activation of α_IIbβ₃ in platelets and in transfected CHO cells, although it is a very weak agonist in platelets compared with other stimuli.^112,113  Importantly, GPVI/FcR γ-chain–deficient mouse platelets, which express a normal level of GPIb-IX-V, fail to adhere on immobilized collagen at low and high shear rates because of the absence of integrin activation.³⁵  Consistent with this, Goto et al reported a complete lack of adhesion of GPVI-deficient human platelets to collagen under the same conditions.⁸²  Thus, it appears that signaling events mediated by GPIb-IX-V are insufficient to promote rapid stable adhesion on a collagen surface or that they are lost in the absence of the GPVI/FcR γ-chain complex. In this context, it is noteworthy that GPIb-IX-V has been reported to induce tyrosine phosphorylation of the FcR γ-chain in platelets along with a number of proteins in the GPVI-signaling cascade including Syk and PLCγ2, although the significance of this remains to be clarified.^36,38,114 

The significance of these observations has been addressed in vivo. Surprisingly, platelet-tethering/slow-surface translocation at sites of arterial injury, as well as firm adhesion, is virtually abolished in the absence of functional GPVI in mice.¹¹⁵  This observation provides strong evidence that GPIb and GPVI must act in concert to recruit platelets to the exposed subendothelial matrix under high shear. One explanation for the involvement of GPVI in the process of tethering in vivo could be that the VWF layer is inhomogeneous and interrupted such that the collagen is not covered with VWF. In this model, direct platelet interaction with collagen (involving GPVI) would be required for efficient GPIb/VWF interactions. Alternatively, the concerted action of GPIb-IX-V and GPVI/FcRγ complexes may be operative independently of collagen, as suggested by the study of Goto et al who reported reduced adhesion of GPVI-deficient human platelets to immobilized VWF under high shear.⁸²  The authors speculated that the GPVI/FcR γ-chain complex may be involved in the process of platelet activation, leading to firm adhesion to VWF through activated α_IIbβ₃. Potentially, this could be mediated through the direct activation of GPVI by VWF or by passive clustering on the assumption that GPIb-IX-V and GPVI/FcR γ-chain are in some way associated in the membrane. It is of interest that Shrimpton et al have recently presented evidence that a small proportion of the GPIb-IX-V complex is found in lipid rafts and that this may contribute to shear-dependent platelet adhesion and tyrosine phosphorylation in response to VWF.¹¹⁶  Since the GPVI/FcR γ-chain complex also signals through lipid rafts,^55-57  it may be that GPIb-IX-V and GPVI/FcR γ-chain associate in these domains and that clustering of GPIbα leads to ligand-independent signaling through GPVI/FcR γ-chain. Others, however, have been unable to detect the presence of GPIb-IX-V complex in membrane rafts and have reported that GPIb-IX-V induces tyrosine phosphorylation of Syk and PLCγ2 but not the raft adapter protein LAT,^36,114  which is the most prominently tyrosine phosphorylated protein in GPVI-activated platelets and is critical for activation.¹¹⁷ 

Besides its crucial role for platelet tethering, VWF also plays an important role as a substrate for firm platelet adhesion to collagen through α_IIbβ₃.^3,35  Therefore, in plasma or whole blood, α_IIbβ₃ and α₂β₁ have largely redundant roles with respect to mediating shear-resistant adhesion. Importantly, as in the case of α₂β₁ and collagen, the interaction of α_IIbβ₃ with VWF requires prior activation of the integrin.¹⁰⁵  On collagen-coated surfaces in vitro, this activation is critically dependent on functional GPVI/FcR γ-chain as shown by the absence of α_IIbβ₃ activation and adhesion of GPVI/FcR γ-chain–deficient platelets under both static and flow conditions.³⁵  Together, these observations show an important role for the interaction of VWF with GPIb and α_IIbβ₃ in platelet adhesion to collagen that is largely dependent on functional GPVI.

Over the years, a significant number of other proteins have been proposed as receptors for collagen in platelets. A number of these have been described through studies on patients with unexplained bleeding defects or through the differential pharmacology of collagens and collagen-based peptides. The absence of knowledge of the genetic defect in the patient studies combined with the low number of cases, however, makes it difficult to assess this work. On the other hand, it seems likely the GPVI/FcR γ-chain receptor complex plays a critical role in platelet activation by most if not all collagens as is illustrated by the abolition of aggregation to collagen types I to V in GPVI/FcR γ-chain–deficient platelets.³⁷  Small differences between collagen types in terms of their dose-response relationships and sensitivity to α₂β₁ blockade may be due to variations in cross-linking rather than receptor differences, as similar differences can be seen for individual preparations of the same collagen type as discussed in the section “Collagens, synthetic collagens, and GPVI-specific ligands.”³⁷  The 2 reports that collagen is able to stimulate tyrosine phosphorylation and weak aggregation in platelets from GPVI-deficient patients should be treated with caution.^118,119  The low level of expression of the glycoprotein in the platelets from these patients may have been sufficient to mediate the increase in phosphorylation and weak activation that was described. Significantly, Snell et al have shown that collagen, but not CRP-XL, stimulates activation of murine platelets that express 20% of the endogenous level of GPVI/FcR γ-chain.⁶⁰ 

These factors have important implications for the existence of other collagen receptors in platelets. Chiang et al have cloned proteins of 65 and 47 kDa and have shown that they are expressed in platelets and bind collagen types I¹²⁰  and III,¹²¹  respectively, while Monnet and Fauvel-Lafeve¹²²  and Monnet et al¹²³  have described a novel p68 type III collagen-binding protein in platelets. At best, therefore, it would seem that the role of these receptors would be to modulate (either positively or negatively) signals from GPVI/FcR γ-chain rather than to induce activation in their own right. These studies require, however, independent confirmation as well as additional experimentation (which in the case of p68 includes receptor cloning) to establish the role of these receptors in platelet activation by collagen. CD36 (or GPIV) has also been proposed as a collagen receptor in the late 1980s^124-127  and more recently as a specific receptor for collagen type V.¹²⁷  CD36 is no longer considered to be a functional receptor for collagen, however, as human and mouse platelets that lack the glycoprotein respond normally to collagens types I to V.^37,128-130 

One of the strongest candidates as an additional collagen receptor is GPV. GPV is noncovalently linked to the GPIb-IX complex in the platelet membrane¹³¹  and its function is poorly defined. The generation of GPV knock-out mice has provided a novel way to examine its function.^132,133  GPV^–/– mice display no Bernard-Soulier phenotype or any other obvious defects, but their platelets show a moderate decrease in collagen-dependent adhesion and aggregation.¹³⁴  Further, additional studies demonstrated that GPV specifically binds to collagen and that soluble GPV blocks collagen-induced platelet aggregation, confirming that it can function as a direct collagen receptor.¹³⁴  Importantly, GPV^–/– platelets are more sensitive to inhibition of collagen-induced aggregation by the anti-GPVI mAb JAQ1, suggesting that GPV facilitates the collagen-GPVI interaction in a manner that may be analogous to that of α₂β₁.¹³⁴  Determination of the functional significance of GPV in vivo, however, is hampered by its role in platelet activation by thrombin. GPV^–/– mice exhibit an increased sensitivity to thrombin, and thus these mice have altered responses to the 2 strongest known platelet agonists. Initial in vivo studies in models of arterial thrombosis have yielded conflicting results on the significance of GPV deficiency,^134,135  and this may be explained by different experimental conditions favoring either collagen- or thrombin-dependent thrombus formation.

It therefore remains an open question as to whether there is a third, functional collagen receptor, although a number of lines of indirect evidence suggest that this is the case. This includes the observation that collagen stimulates a weak increase in protein tyrosine phosphorylation in GPVI/FcR γ-chain–deficient platelets.³⁰  In addition, we have shown that collagen, but not CRP-XL, induces activation of murine platelets that express 20% of the normal levels of the GPVI/FcR γ-chain in the presence of blocking antibodies to α₂β₁ and GPV.⁶⁰  This suggests either that collagen retains a greater affinity for the GPVI/FcR γ-chain complex than CRP or that its binding to the glycoprotein complex is supported by an unidentified third receptor.

It is assumed that, despite the complexity of the ECM, platelet-collagen interactions play a pivotal role in the initiation of hemostasis and thrombosis in vivo. The small number of patients with defects in GPVI and α₂β₁, and the presence of additional complications in these patients, however, has prevented a direct assessment of the role of the 2 glycoproteins in platelet-collagen interactions in vivo. The generation of mice with targeted disruption of the 2 integrin subunits and in FcR γ-chain, and the availability of antibodies to block or deplete GPVI in vivo, has now provided a means to directly assess their clinical significance.

It is noteworthy that ablation of GPVI function in mice leads only to moderately increased tail-bleeding times and that this response is unaltered in α₂β₁-deficient mice. Furthermore, the loss of function of either of these collagen receptors does not lead to an increase in spontaneous bleeding tendency in the mouse.^39,100,102  While it will be important to study hemostasis in mice lacking both GPVI and α₂β₁, the lack of bleeding in collagen receptor–deficient individuals suggests that other agonists such as thrombin, ADP, or thromboxanes can substitute for GPVI in mediating platelet activation and that other adhesive receptors, such as integrins α_IIbβ₃, α₅β₁, and α₆β₁, can mediate firm adhesion and spreading in the absence of the 2 collagen receptors. Nevertheless, GPVI, together with GPIb, has been shown to play a crucial role in the very first step of platelet attachment to the injured arterial wall in vivo,¹¹⁵  suggesting that GPVI deficiency should have an effect similar to that of defective GPIb/VWF interactions. However, while impaired GPIb/VWF interactions lead to a pronounced bleeding phenotype, only a moderate bleeding tendency is seen in GPVI/FcRγ-deficient individuals. One possible explanation for this discrepancy may be that GPVI is involved only in the attachment of the first platelet layer to the ECM, whereas GPIb/VWF interactions are also involved in recruitment of additional platelets into the growing thrombus.⁹⁹  On the other hand, these observations demonstrate that GPVI-independent mechanisms can mediate platelet attachment and activation at sites of vascular injury during normal hemostasis. Further studies will be required to identify these mechanisms.

Although individual collagen receptors may not be essential to arrest bleeding, there is strong evidence that they contribute to pathologic thrombosis, notably following rupture of the atherosclerotic plaque in a stenosed artery. In the process of atherogenesis, enhanced collagen synthesis by intimal smooth muscle cells and fibroblasts has been shown to significantly contribute to luminal narrowing.^136,137  Thus, the matrix exposed on plaque rupture is enriched in collagens and is therefore likely to be a major player in arterial thrombus formation. A possible role of α₂β₁ in cardiovascular disease has long been proposed and has been intensively studied through analysis of genetic polymorphisms. The expression level of α₂β₁ on platelets varies by up to a factor of 5, and this variation is strongly linked to a polymorphism in the noncoding region of the α₂ gene.¹³⁸  A large number of clinical studies have examined a possible association of these polymorphisms with cardiovascular disease but has failed to reach an overall consensus, with approximately half of the studies reporting a significant but minor role for the integrin.^139-143  Several factors are likely to have contributed to this ambiguity, including the complex nature of vascular disease and the relatively small variation in the level of the integrin. The analysis of in vivo studies in integrin α₂–deficient mice in models of cardiovascular disease, such as arterial thrombosis or atherogenesis, is warranted to provide a more detailed understanding of the role of this receptor in pathologic conditions.

GPVI has recently been shown to be highly polymorphic, although the functional significance of this is yet to be clarified. In a first report, Croft et al described a link between the GPVI 13254CC genotype and an increased risk of myocardial infarction in association with age, sex, and fibrinogen genotype,⁴³  although on its own it is not an independent risk factor. This polymorphism replaces a serine with a proline in the mucin-rich domain in GPVI and therefore might change the orientation of the 2 Ig domains in the outer leaflet. This mutation causes a small (∼ 20%) reduction in the expression of GPVI but has little effect on platelet activation by collagen using a shear-based model of platelet activation (D. Best and S.P.W., unpublished data, June 2002).

Furihata et al reported a 5-fold range of expression levels of GPVI using a semiquantitative ligand-blotting method in a panel of 23 healthy individuals that directly correlated with GPVI-specific prothrombinase activity.¹⁴⁴  In addition, the level of GPVI was found to correlate with expression of α₂β₁ in a group of 15 donors. However, the level of expression of GPVI was found to vary by less than 2-fold in 102 patients when measured by flow cytometry using 2 distinct antibodies to the glycoprotein, and there was also only a very weak correlation with the level of expression of α₂β₁ (D. Best and S.P.W., unpublished data, June 2002). Minor variations in GPVI expression levels on human platelets were also reported by Chen et al⁵⁹  using a smaller study group. It will be essential to extend the initial clinical studies to larger groups of patients and to the other polymorphisms to confirm the link between GPVI polymorphisms and the risk of cardiovascular disease, although the low frequency of the 13254CC genotype (∼ 1%) will hamper further investigation.

There is compelling evidence for a crucial role of GPVI in arterial thrombosis from studies in mice. Massberg et al demonstrated that thrombus formation in the injured carotid artery in mice is virtually abolished in the absence of functional GPVI.¹¹⁵  This agrees with a recent study by Konishi et al¹⁴⁵  who found markedly reduced platelet attachment and subsequent neointimal hyperplasia at sites of vascular injury in FcR γ-chain–deficient mice that lack GPVI.³²  These developments implicate GPVI as a potential pharmacologic target for the treatment of ischemic cardiovascular disease. Such a strategy might have a number of advantages. First, GPVI is exclusively expressed on platelets and megakaryocytes,⁴⁴  which prevents the risk of side effects of anti-GPVI agents on other cell types. Second, GPVI deficiency is not associated with a major bleeding risk in humans and mice, suggesting that anti-GPVI therapy might be well tolerated. Thus, there is considerable interest in developing novel tools, such as receptor antagonists, or depleting antibodies to evaluate GPVI function in humans. The second of these procedures is a unique way to abolish collagen responses for the lifetime of the platelet. This has been demonstrated in mice in which the injection of the anti-GPVI antibody JAQ1 leads to internalization and degradation of GPVI in circulating platelets resulting in a prolonged “GPVI knock-out”–like phenotype.³⁹  A similar mechanism of GPVI down-regulation appears to exist in humans, as one GPVI-deficient patient had developed antibodies against the apparently absent receptor, suggesting that she suffered from an acquired GPVI deficiency based on antibody-induced clearing of the receptor from her platelets.¹⁹  Recent studies in mice demonstrate that the in vivo depletion of GPVI can be induced through more than one site and does not require prior receptor activation.¹⁴⁶  This strongly suggests that anti-GPVI agents that do not directly activate the platelet may generally induce down-regulation of the receptor in vivo, resulting in long-term, powerful antithrombotic protection.

The last few years have seen major advances in understanding how platelets interact with collagen and the events that initiate primary hemostasis and arterial thrombosis. The cloning of GPVI, the generation of mice deficient in collagen receptors and their associated signaling molecules, and the availability of new antibodies and collagen-derived peptides allowed detailed in vitro and in vivo studies on the role of the individual candidate receptors in the complex process of platelet tethering, activation, adhesion, aggregation, and procoagulant activity. These studies have changed the long-standing concept of platelet-collagen interactions, the socalled 2-site, 2-step model, which proposed α₂β₁ as the major collagen receptor in hemostasis and thrombosis. This hypothesis had been based on the assumption that the integrin is constitutively in a high-affinity conformation and is essential for the initial firm arrest of platelets on collagen.¹⁴⁷  However, it is now recognized that α₂β₁, like the other integrins, is in a low-affinity state on resting platelets and requires inside-out signals to efficiently bind to collagen. It is now established that the initial platelet contact with collagen and the subsequent initiation of integrin activation (ie, adhesion and thrombus growth) are strictly dependent on functional GPVI. These developments identify a new sequence of events in the initial phase of hemostasis and thrombosis and place GPVI in a central position in this complex process (Figure 3). It is now proposed that under high shear flow conditions, GPIbα and GPVI act in concert to tether platelets to the ECM through their respective ligands, VWF and collagen. The fast off-rate of these interactions prevents the rapid onset of stable adhesion. The generation of intracellular signals from GPVI, and possibly GPIb, converts β₁ and β₃ integrins to a high-affinity state and induces the release of soluble agonists, most importantly ADP and TxA₂ (which also induce integrin activation). Activated α₂β₁ and α_IIbβ₃ integrins now initiate firm adhesion by binding to collagen and VWF, respectively, and this process is reinforced by the autocrine action of the released mediators. In turn, integrin-mediated adhesion strengthens GPVI-collagen interactions, leading to enhanced signaling and further up-regulation of integrin activity, enhanced release, and the development of procoagulant activity. Integrinmediated signaling events are also likely to contribute to these processes. Finally, the accumulation of released ADP and TxA₂ results in the activation of further platelets (ie, thrombus growth).

Despite the major steps that have been made in understanding platelet-collagen interactions, there are still many fundamental questions that need to be addressed both with respect to platelet-collagen interactions and to the events that give rise to hemostasis in vivo. These questions include not only a precise description of the roles of GPIb, GPVI, and α₂β₁, but also a more complete assessment of the role of other components of the ECM and additional platelet integrins. Finally, and most importantly, we need to understand the value of these events in the context of development of novel antithrombotics. It is our belief that, in this context, GPVI is a major target for development of novel pharmaceuticals that are likely to be highly effective and well tolerated.

We are pleased to acknowledge the present and past members of our laboratories for their work on platelet-collagen interactions and for many of the ideas and concepts that have been discussed in this review.