Over the years many studies have identified potential platelet receptors for collagen. This is a difficult interaction to dissect because: of the 18 described collagen types at least 7 are present in the vessel wall [ 19]; collagen is an insoluble macromolecular protein at physiologic pH making analysis of cell–protein interactions difficult [ 20]; and many activation end-points, including membrane GP trafficking, granule secretion and aggregation, progress simultaneously when platelets bind collagen [ 12]. In spite of these obstacles, several proteins on the platelet surface, both integrin and nonintegrin, have been shown to possess characteristics consistent with being specific collagen receptors.
Integrin α2β1 (glycoprotein Ia-IIa)
Integrin α2β1, which is identical with the platelet membrane GP Ia-IIa, the very late activation antigen-2 on activated T cells [ 21] and the class II extracellular matrix receptor on fibroblasts [ 22], has recently been reviewed in depth [ 23]. The α2 (GPIa) and β1 (GPIIa) subunits have apparent molecular weights of 165 kDa (reduced) and 130 kDa, respectively. The aminoacid (aa) sequences were deduced from their corresponding cDNA [ 24, 25]. Both subunits possess a short C-terminal cytoplasmic tail, a single hydrophobic membrane spanning region and a large extracellular domain. The α2 chain, like most other integrin α subunits, does not undergo post-translational cleavage. Its extracellular domain contains a 7-fold repeat segment which includes a EF hand motif with three cation binding sites thought to be involved in ligand binding [ 26]. Just distal to these, there is a segment of 191 aa called the I (inserted)-domain, which is homologous to the collagen-binding A domain of vWF [ 24]. In fact, the α2 I-domain has been shown to bind collagen [ 27, 28], and a recombinant α2 I-domain fusion protein inhibits collagen-induced platelet adhesion [ 29]. While the α2 I-domain is sufficient and essential for platelet-collagen binding, other structures, such as the adjacent EF hand motif can optimize it [ 26]. In addition, the extracellular domain of the β1 subunit, which is composed of a proximal portion of four internally folded cysteine-rich repeat units and a distal portion of highly conserved sequence shared with other β integrins, is thought to regulate the binding affinity of the α2 I-domain [ 30, 31].
The first indication that integrin α2β1 might be a physiologically relevant collagen receptor came from the observation of a female patient with excessive post-traumatic bleeding and menorrhagia [ 32]. This patient's platelets, which selectively failed to aggregate or undergo shape change in response to collagen, were found to contain only 15–20% of the normal amount of GPIa [ 32]. Furthermore, under flow conditions these platelets exhibited a markedly decreased adhesion to and failure to spread on a subendothelial surface [ 33]. A second patient with GPIa deficiency has been reported, whose haemorrhagic symptoms and biochemical defect surprisingly disappeared when she entered menopause [ 34]. In addition, a patient with an acquired bleeding disorder associated with an autoantibody against GPIa [ 35] and another with a complete deficiency of GPIa in the context of a myeloproliferative disorder [ 36] have been described.
The interaction of integrin α2β1 with collagen requires Mg2+ (which can be substituted by Mn2+, Co2+, Cu2+, Fe2+ and Zn2+) and is inhibited by Ca2+ [ 37, 38], thus explaining the observation that platelet adhesion to collagen is markedly increased in the presence of Mg2+ [ 39]. Saelman et al. demonstrated that collagens I through VIII are variously able to support platelet adhesion and that this can be completely inhibited, under conditions of both stasis and flow, by a monoclonal antibody directed against GPIa [ 40]. Noteworthy, collagen type V does not support adhesion under flow, and at shear rates of 1600/s adhesion to collagen type III is only inhibited by 85% [ 40]. Interestingly, at high shear rates inhibition of platelet adhesion to the most reactive collagens, types I through IV, requires three times more antibody than needed at low shear [ 40], suggesting that the role of integrin α2β1 is more critical at high shear. Consistent with this are studies showing that the binding avidity of integrin α2β1 can be modulated with antibodies against the β1 subunit [ 30, 31], and enhanced following activation of protein kinase C [ 41]. Studies with soluble collagen indicate that integrin α2β1 becomes activated following platelet interaction with thrombin, ADP and a GPVI specific agonist [ 42]. Finally, CD47/IAP (integrin-associated protein), a receptor for the cell binding domain of thrombospondin-1, has recently been shown to coimmunoprecipitate with integrin α2β1 and to augment its function [ 43]. These data suggest the interesting hypothesis that variations in integrin α2β1 binding avidity may have physiological significance.
The number of integrin α2β1 molecules on the platelet surface significantly varies between normal individuals [ 44], ranging between 800 and 1800 [ 21, 44, 45]. While platelet activation does not increase the number of surface molecules by more than 5% [ 44], the interindividual variation of integrin α2β1 surface levels by itself correlates with platelets' ability to adhere to collagen types I and III under static conditions [ 44] and to collagen type I under flow [ 46]. This heterogeneity is associated with three α2 gene alleles, defined by eight nucleotide polymorphisms [ 46, 47]. Initially two silent, linked polymorphisms located at nucleotides 807 (TTT/TTC at codon Phe224) and 873 (ACA/ACG at codon Thr246) were described [ 47]. Although the aa sequence of the α2 subunit is not affected by the polymorphisms, the 807T/873A pair is associated with higher surface levels of integrin α2β1 than the 807C/873G pair [ 47]. Subsequently, a similarly silent but much rarer polymorphism located at nucleotide 837 (C or T) and linked to the Br polymorphism [ 48] was also identified [ 46]. Allele 1 (807T/837T/873A/Brb) has a frequency of ≈ 39% and is associated with increased levels of integrin α2β1, while allele 2 (807C/837T/873G/Brb) with a frequency of ≈ 53% and allele 3 (807C/837C/873G/Bra) with a frequency of ≈ 8% are both associated with lower levels of integrin α2β1 [ 46]. In a case control study, a significantly higher prevalence of individuals homozygous for 807T/873A were found among patients with myocardial infarction [ 49], and multivariate analysis confirmed this genotype as an independent risk factor for myocardial infarction [ 49]. Recently, the 807T allele was found to strongly correlate with the development of nonfatal myocardial infarction [ 50] and stroke [ 51] in younger patients. These three studies indicate that the inherited variation of platelet surface integrin α2β1 is a significant risk factor for acute thrombotic events. Interestingly, the frequency of the 807C allele, which is associated with a lower integrin α2β1 density and a decreased haemostatic function, has been shown to be significantly higher among type 1 von Willebrand disease patients than among normal individuals [ 52]. This observation confirms that integrin α2β1 is a significant contributor to the haemostatic process, and suggests that its surface variability among patients with similar vWF levels may account for the very different bleeding phenotypes.
In summary, integrin α2β1 is considered to be the major receptor mediating direct and permanent platelet adhesion to collagen [ 11], thus perfecting the initial vWF-dependent tethering at high shear rates [ 6] and facilitating engagement of lower affinity receptors, such as GPVI. The extent of integrin α2β1 contribution to subsequent platelet activation is less clear (see below). The clinical relevance of integrin α2β1 polymorphisms with its variable expression on the platelet surface has been addressed in a recent review [ 53].
Glycoprotein VI (p62)
GPVI, a not yet cloned 62 kDa (reduced) platelet membrane protein, was first described 20 years ago [ 54], and its involvement in platelet–collagen interactions was postulated one decade later based on the following observations. An antibody that recognized a 62/57 kDa platelet membrane protein was identified in the serum of a patient with autoimmune thrombocytopenia whose platelets were selectively defective in collagen-induced aggregation [ 55]. This antibody could recognize a 62 kDa protein and induce aggregation of normal platelets [ 55], but did not react with platelets from a GPVI-deficient patient [ 56], thus identifying the 62 kDa protein as GPVI and demonstrating its involvement in collagen-induced platelet activation. To date, 3 GPVI-deficient patients [ 56[57]–58], and two patients with an autoantibody against GPVI [ 55, 59], have been described. All patients exhibit a mild bleeding tendency and slightly prolonged bleeding times. While their platelets have an essentially normal response to other physiologic agonists, they show a defective aggregation in response to collagen despite normal expression of integrin α2β1. Although GPVI has been implicated in platelet adhesion to collagen under static conditions [ 58, 60, 61], this is probably a consequence of its ability to induce platelet activation. Under flow conditions GPVI involvement relates to second phase adhesion, a process which is secondary to integrin αIIbβ3 activation [ 62]. GPVI has been demonstrated to recognize both the tertiary (triple-helical) and quaternary (polymeric) structure of collagen and is considered to be the crucial receptor mediating platelet activation [ 63]. The intracellular signalling events induced by GPVI engagement will be discussed below.
CD36 (GPIV)
CD 36 [ 64], an 88 kDa GP functioning as a scavenger receptor and cell-adhesion molecule [ 65], is expressed on several cell types, including platelets, monocytes/macrophages, reticulocytes/erythrocytes, microvascular endothelial cells and melanoma cells, and has been implicated in a variety of pathophysiological situations ranging from haemostasis and thrombosis to malaria, inflammation, lipid metabolism and atherogenesis [ 65].
It has been estimated that there are about 20 000 CD36 molecules on the platelet surface [ 66]. This GP has been proposed as a collagen receptor based on the observation that antibodies against it could inhibit collagen induced platelet activation and aggregation [ 67, 68]. In addition, incubation of normal platelets with Fab fragments of a monospecific polyclonal anti-CD36 antibody inhibited the early stages of adhesion to collagen type I under static [ 67] and under flow conditions [ 69]. Diaz-Ricart et al. [ 69] using citrated reconstituted whole blood also showed that platelets from CD36-deficient individuals have a decreased early adhesion. However, 3% to 11% of healthy Japanese blood donors lack CD36 without any apparent bleeding disorder [ 70]. Moreover, collagen induced aggregation [ 71] and metabolic responses [ 72] in CD36-deficient platelets have been shown to be normal. The discrepancies between these observations and the previous studies might reside in the divalent cation conditions employed. Utilyzing heparinized blood, Saelman et al. demonstrated that CD36-deficient platelets adhere normally to collagen type I, III, and IV under both static and flow conditions [ 73]. Remarkably, while collagen type V is not adhesive during flow [ 40], under static conditions adhesion of both homozygous and heterozygous CD36 deficient platelets to this collagen type was strongly reduced [ 73]. The peculiar behaviour of collagen type V was confirmed by Kehrel et al. who showed that CD36-deficient platelets aggregate normally with collagen types I and III but not in response to collagen type V [ 74]. Indeed, CD36-deficient platelets appeared even more sensitive to types I and III collagens than normal platelets [ 74], suggesting an inhibitory co-operation between CD36 and other collagen receptor(s). Taken together these observations seem compatible with the hypothesis that CD36 might be involved in the very first adhesion of platelets to collagen, but it is essential only for interaction with collagen type V. This may be relevant to the development of thrombotic events because collagen type V is increased in atherosclerotic plaques [ 75].