Glycoprotein VI Is a Major Collagen Receptor for Platelet Activation: It Recognizes the Platelet-Activating Quaternary Structure of Collagen, Whereas CD36, Glycoprotein IIb/IIIa, and von Willebrand Factor Do Not

COLLAGENS OF THE subendothelium are major determinants of the thrombogenicity of the blood vessel wall.1 In particular, the fibrous collagens I and III, exposed as a consequence of injury to the vessel wall, are strong platelet agonists that induce shape change, release reaction, and aggregation. The mechanism of collagen-platelet interaction involves both the direct recognition of collagen receptors and indirect binding of collagen to the platelet surface via intermediary proteins.2,3 As an example of the latter, von Willebrand factor (vWF) plays an essential role in hemostasis by complexing to collagen(s) in the vessel wall and concomitantly binding to specific receptors on the platelet surface,4,5 the glycoprotein (GP) Ib/V/IX complex, and the activated GPIIb/IIIa complex,6thus forming a bridge between collagen(s) and platelets.

A number of platelet membrane proteins have been proposed as collagen receptors, including integrin α₂β₁(GPIa/IIa; as reviewed by Santoro and Zutter7), CD36 (GPIIIb, GPIV), and GPVI (p62). Their precise role in the overall process of interaction and the relationship between them remains unclear.

Nieuwenhuis et al8 and Kehrel et al9 have described patients with mild bleeding disorders attributable to deficient expression of platelet α₂β₁. The platelets from these patients had impaired collagen-induced aggregation but responded normally to all other platelet agonists. Further evidence for the involvement of α₂β₁ is that monoclonal antibodies (MoAbs) directed against the receptor are able to inhibit adhesion of platelets to collagen under either static or flow conditions.7,10-12 

CD36 binds to collagen type I fibers,13 and Fab fragments of polyclonal antibodies against CD36 have been reported to inhibit collagen-induced platelet aggregation.13,14 However, the role of CD36 as a platelet collagen receptor is not clear, because a significant proportion (1% to 4%) of Japanese and other East Asian populations lack this receptor on platelets but have no obvious bleeding disorder. Platelet adhesion to collagens I and III^15,16 and platelet aggregation and secretion induced by these collagens is normal in CD36-deficient platelets.17,18 

Other Japanese patients with mild bleeding disorders have been described whose platelets show reduced responses to collagen associated with functional deficiency in GPVI.19-21 One of these patients developed antibodies to GPVI, which, when added to normal platelets, caused their aggregation and whose Fab fragments were able to reduce platelet aggregation induced by collagen.21,22GPVI-deficient platelets showed defective second phase adhesion in flowing blood, a reaction that is attributable mainly to the platelet-platelet interaction.23 

The interaction of platelets with collagen involves firstly adhesion and, subsequently, activation leading to second phase adhesion, secretion, and ultimately aggregation. Different mechanisms and receptor populations may be involved in these two processes.24,25 Collagen-induced platelet activation requires the expression by collagen of both tertiary (triple-helical) and quaternary (polymeric) structure.26-28 Recently, Morton et al29 have described simple collagen-related peptides (CRPs), comprising a repeating Gly-Pro-Hyp motif, that mimic the collagen tertiary (triple-helical) structure. They have shown that, when cross-linked via either lysyl or cysteine residues to impart quaternary (polymeric) structure, these peptides (CRP-XL) are extremely active platelet agonists, being more reactive than collagen fibers. In contrast, non–cross-linked (nonpolymeric) CRP antagonizes platelet activation stimulated by either native collagen or CRP-XL.29 CRPs possess the same tertiary structure as collagen and, like collagen, this conformation is essential for activity. Like collagen, CRP molecules spontaneously assemble into an organized quaternary structure and, again like collagen, the polymeric structure is essential for activity. It is therefore most likely that the potent platelet reactivity of CRP-XL reflects the platelet reactivity to collagen.

The interaction of CRP-XL with platelets does not involve the integrin α₂β₁; thus, CRP supports static platelet adhesion that is not blocked by the α₂β₁-directed MoAbs, 6F1 and MoAb13.29 These same antibodies fully block α₂β₁-mediated adhesion to immobilized monomeric collagen. The MoAbs are also without effect on aggregation stimulated by CRP-XL.29 The recognition of α₂β₁ by collagen is essential for adhesion of platelets under flow. However, CRP-XL is unable to support adhesion under flow conditions.30 Furthermore, the purified I-domain of the α2 subunit, which mediates the adhesion of native collagens to α₂β₁, does not recognize CRP-XL,31 and affinity chromatography using CRP-XL does not extract α₂β₁ from platelet lysates (Barnes et al, unpublished data). All of this establishes CRP as an agonist that operates entirely independently from α₂β₁.

Despite the absence of recognition of the integrin α₂β₁, CRPs signal in platelets in precisely the same way as native collagen fibers.32,33 Specifically, they stimulate tyrosine phosphorylation of the Fc receptor γ chain,34 of p72^syk, and of phospholipase Cγ2 (PLCγ2)³⁵—all responses characteristic of platelet activation by collagen. CRPs are therefore valuable tools for the identification of the collagen receptor that recognizes the quaternary structure of collagen and is responsible for signaling leading to activation. To this end, we have evaluated the response to CRP-XL of platelets deficient in GPIIb/IIIa, CD36, vWF, and GPVI.

Ristocetin was obtained from Paesel (Frankfurt, Germany).

Human vWF was purified from Haemate HS (a kind gift of Centeon, Marburg, Germany), as previously described.36 

The vWF-directed MoAb, 4F9 (Immunotech, Marseilles, France), was conjugated to fluorescein isothiocyanate (FITC) using the antibody conjugation kit from Sigma (Deisenhofen, Germany) according to the manufacturer's instructions. The F/P ratio, calculated from the FITC absorbance at 495 nm and the IgG content, was 1.4. FITC-coupled MoAbs recognizing CD62 (P-selectin, GMP140, and PADGEM) or CD63 (GP53 and granulophysin), clone CLB-thromb/6 and CLB-gran/12, respectively, were obtained from Immunotech (Marseilles, France).

Highly purified human fibrinogen (Enzyme Research Labs, South Bend, IN) was conjugated with FITC using the FITC-celite (Calbiochem, Bad Soden, Germany) method according to Xia et al37 but using pH 7.8 buffer for 48 hours at 4°C, resulting in an F/P ratio of 5.0 to 5.2.

Collagen type I for use as a reference reagent has been described previously.18 

The triple-helical peptide, Gly-Lys-Hyp-[Gly-Pro-Hyp]₁₀-Gly-Lys-Hyp-Gly (CRP), was synthesized and cross-linked via its lysyl residues to yield CRP-XL as described in detail by Morton et al29 and Barnes et al.32 

All studies were performed with the patients or other blood donors giving informed consent. Samples from healthy controls were processed in parallel with all patient samples. None of the subjects had taken any medication affecting platelet functions for at least 2 weeks before the study.

The CD36-deficient platelets of a Japanese male blood donor Y.A. have been described before.18,38 Platelets from patient A.M. with Glanzmann's thrombasthenia type I were shown to contain less than 1% of normal amounts of GPIIb/IIIa.39 Patient P.R. had von Willebrand disease type 3 (according to the Sadler and Gralnick classification40). In addition, the patient suffered from an intestinal angiodysplasia, which is associated with life-threatening bleeding episodes, requiring frequent replacement therapy with vWF. The blood for the experiments was taken 2 weeks after the last treatment and directly before new medication. At that time the vWF:RCO was less than 1% of normal and vWF antigen was undetectable (Asserachrom vWF; Stago-Diagnostica, Asniers, France) in both the plasma and the platelets. The GPVI-deficient platelets of the female Japanese patient Y.A. have been described in detail.21,22,41 

Platelet aggregation studies were performed in an aggregometer (Chrono-Log, Haverton, PA) within 3 hours of venipuncture using platelet-rich plasma (PRP) containing 2 × 10⁸platelets/mL, as described by Born and Cross.42 Blood was anticoagulated with 0.1 vol of 3.8% Na-citrate and PRP was prepared by centrifugation at 200g for 10 minutes at room temperature.

PRP was diluted to 5 × 10⁷ platelets/mL with Tyrode's buffer, pH 7.4, to minimize the formation of platelet aggregates. Platelet suspension (200 μL) was added to 20 μL of ristocetin (at a range of concentrations in Tyrode's buffer, pH 7.4) without stirring to avoid aggregate formation. After 180 seconds, the platelets were fixed for 30 minutes with 1% formaldehyde (final concentration) in phosphate-buffered saline (PBS). Platelets were washed and resuspended in 200 μL of PBS, and FITC-conjugated anti-vWF MoAb was added at a saturating concentration (determined beforehand). After 1 hour of incubation at room temperature, 10⁴ single platelets were analyzed in a flow cytometer (FACScan; Becton Dickinson, Heidelberg, Germany). Background binding obtained from parallel samples with FITC-conjugated isotype-specific mouse IgG was subtracted from each test sample. Excitation was at a wavelength of 488 nm. The FACScan was used in a standard configuration with a 530 nm bandpass filter. Data were obtained from fluorescence channels in a logarithmic mode.

Diluted PRP (5 × 10⁷ platelets/mL) was preincubated for 3 minutes at room temperature with 150 μg/mL fibrinogen-FITC (saturating concentration). Two hundred microliters of platelet suspension was added to 20 μL of a solution containing a range of concentrations of CRP-XL in 0.01 mol/L acetic acid at room temperature. The reaction was stopped after 120 seconds by fixation with 1% formaldehyde as described above. Platelets were washed, resuspended in 200 μL of PBS, and analyzed in a flow cytometer as described above. The nonspecific background labeling was determined using platelets from different patients with thrombasthenia type I and control platelets treated with 10 mmol/L of the inhibitory peptide GRGDSP (Novabiochem, Bad Soden, Germany) to prevent specific fibrinogen binding.

Two hundred microliters of diluted PRP (5 × 10⁷platelets/mL) was added at room temperature to 20 μL 0.01 mol/L acetic acid containing CRP-XL at a range of concentrations. The reaction was stopped after 120 seconds with 1% formaldehyde, as described above. Platelets were washed and resuspended in PBS, and FITC-coupled MoAbs recognizing CD62 or CD63 were added and incubated for 1 hour at room temperature. The concentrations required for saturated binding of both antibodies were determined beforehand. After incubation, platelets were washed, resuspended in 200 μL of PBS, and analyzed as described above.

Platelets from 8 control donors were aggregated by CRP-XL at concentrations from 20 to 100 ng/mL. Aggregation curves from 2 such donors are shown in Fig 1A. The CD36-deficient platelets of donor Y.A. reacted as normal control platelets when treated with CRP-XL, which induced full aggregation at a concentration of 37 ng/mL (Fig 1B). Fibrinogen binding to CRP-XL-activated control and CD36-deficient platelets did not differ significantly (Fig 2). CD36-deficient platelets activated by CRP-XL expressed CD62 and CD63 to the same extent as control platelets (data not shown).

Because, as expected, specific fibrinogen binding after exposure to agonist was not observed in these GPIIb/IIIa-deficient platelets, activation was determined on the basis of the surface expression of CD62 (Fig 3A) and of CD63 (Fig 3B). Expression of both in GPIIb/IIIa-deficient platelets treated with CRP-XL was not lower than in controls.

Ristocetin, at concentrations up to 4 mg/mL, did not induce vWF-mediated platelet agglutination in PRP from patient P.R. (data not shown). The addition of human vWF (10 μg/mL) corrected the defect. No vWF binding to the surface of the patient's platelets was observed when these platelets were treated with ristocetin at concentrations ranging from 0.2 to 2 mg/mL, whereas 1.0 mg/mL induced maximum vWF binding to control platelets (data not shown). No vWF was detectable in samples of the patient's platelets and plasma taken at the time these platelets were activated with CRP-XL. CRP-XL induced normal aggregation of vWF-deficient platelets in vWF-deficient plasma (Fig 4A) and induced fibrinogen binding to a somewhat greater extent than in controls (Fig 4B). The surface expression of CD62 (Fig 5A) and of CD63 (Fig 5B) induced by CRP-XL in PRP of patient P.R. were identical in terms of both extent and concentration dependence to that occurring in control platelets.

Aggregation of GPVI-deficient platelets was totally absent even in the presence of a concentration of either collagen or CRP-XL 20-fold higher than was needed to aggregate normal platelets (Fig 6A and B). In contrast, aggregation of these platelets by ADP was normal (Fig 6C). Neither collagen nor CRP-XL induced any surface expression of CD62 (Fig7A and B) or CD63 (Fig 8A and B). There was no expression of fibrinogen binding in response to CRP-XL (Fig 9B), although collagen-induced fibrinogen binding was not absent but expressed to a decreased extent compared with normal platelets, especially at low concentrations of collagen (Fig 9A).

All collagens share a common tertiary structure, the collagen triple helix. The collagen molecule contains three α chains, each with a repeating Gly-X-Y sequence in which X and Y are frequently the amino acids, Pro and Hyp, respectively. The Gly-Pro-Hyp triplet represents approximately 12% of the primary sequence of type I collagen. Within the molecule each of these chains forms a left-handed polyproline II helix and the three helices are intertwined to form a right-handed superhelix, the collagen triple helix. Molecules then associate very specifically to yield the highly ordered quaternary structure of the collagen fiber.43 

The native triple helical structure of collagen is required for platelet secretion and aggregation. Furthermore, collagen molecules must also assemble into fibrous form for collagen to be an effective agonist.26-28 In contrast, monomeric collagen immobilized on plastic can serve as a very effective substrate for adhesion without any activation.44 This adhesion is mediated by the integrin α₂β₁ and can be blocked by antibodies to α₂β₁.7,10-12 

Although evidence has been presented suggesting that signaling in platelets by collagen is subsequent to recognition of α₂β₁,35,45-47 it is not clear how far α₂β₁ is directly responsible for signaling leading to platelet activation. Although collagen-induced phosphorylation events may under some circumstances be attenuated by blockade of α₂β₁ by antibodies, persistence of signaling has been reported in response to collagen or CRP-XL.29,34,35,47 Indeed, in platelets lacking GPVI, but in which α₂β₁ is expressed normally, collagen is able to activate the tyrosine kinase c-src, but not p72^syk, PLCγ2, or p125^fak.41 All of these are regarded as important signaling molecules in the activation of normal platelets by collagen. This indicates the possibility that GPVI (rather than α₂β₁) may be crucial as an activatory receptor for collagen in platelets.

A triple-helical synthetic analogue of collagen, composed simply of a repeating Gly-Pro-Hyp sequence, has been shown in polymeric form (CRP-XL) to elicit a full aggregatory response from human platelets without the involvement of the integrin α₂β₁.29 The triple-helical non–cross-linked peptide (CRP) is not activatory but inhibits platelet activation by collagen and by CRP-XL.29 This corroborates the existence of a signaling receptor on platelets other than, or in addition to, α₂β₁, which requires the tertiary and quaternary structures of collagen for platelet activation.

As shown here, CD36-deficient platelets bind fibrinogen and exhibit normal secretion and aggregation in response to CRP-XL (as is true for collagen15-18). CD36 cannot, therefore, be this collagen receptor.

Platelets bind indirectly to collagen, via vWF, through GPIb and GPIIb/IIIa complexes.48 This binding confers resistance to the shear forces generated by blood flowing along the vessel wall.49,50 However, platelets from a Glanzmann's thrombasthenia type I patient, totally deficient in GPIIb/IIIa, responded normally to CRP-XL, which induced normal secretion, measured as expression on the plasma membrane of CD62 (from the α-granules) and CD63 (from the dense bodies). These data exclude GPIIb/IIIa as the primary signaling receptor that recognizes the quaternary structure of collagen.

The same can be concluded for vWF. Despite the absence of vWF, both in the platelets and plasma of patient P.R., CRP-XL induced normal fibrinogen binding, normal secretion (CD62 and CD63 expression), and normal aggregation.

These results confirm the view that CD36, GPIIb/IIIa, and vWF are not essential for platelet activation by collagen.

In contrast, GPVI-deficient platelets were unresponsive to CRP-XL, showing that GPVI is a signaling receptor for the polymeric cross-linked triple-helical Gly-Pro-Hyp motif. Our findings are in line with the observation that antibody-induced cross-linking of GPVI in normal platelets results in activation of src and p72^syk in parallel with activation of PLCγ2²² and that CRP-XL stimulates tyrosine phosphorylation of PLCγ2 and p72^sykin normal platelets independent of the integrin α₂β₁.35 

In GPVI-deficient platelets CRP-XL did not induce the activation-dependent conformational changes in GPIIb/IIIa that are necessary for fibrinogen binding. However, it is interesting that fibrinogen binding was induced in these platelets by increasing collagen levels, although to a lower degree than in equivalent controls (Fig 9A). Collagen (but not CRP-XL) contains binding sites for α₂β₁, and GPVI-deficient platelets express α₂β₁, which may possibly be involved in signal transduction by collagen, sufficient for partial activation of GPIIb/IIIa. Nevertheless, collagen fibers provided insufficient signals to cause platelet aggregation in GPVI-deficient platelets, even at collagen levels 20-fold higher than are needed to aggregate normal platelets. Granule release was completely absent, as with CRP-XL (Figs7 and 8). Recently, the activation of c-src, but not of p72^syk or PLCγ2, has been shown in GPVI-deficient platelets stimulated with collagen,41 providing firm evidence for signaling, but not sufficient to cause aggregation, arising from a second population of collagen receptors as well as GPVI. It is plausible that GPVI may be linked more to positive feedback events, such as thromboxane production, a process essential for the aggregation of platelets by collagen, or granule release, providing further activation mediated by secreted ADP, so leading to full activation of GPIIb/IIIa. Results from these and other studies suggest that, in platelets, α₂β₁ has a major role in adhesion to collagen and a less-well defined role in platelet activation in support of GPVI. These ideas are consistent with the two-site two-step model of platelet activation by collagen.24,25,51 The major role for α₂β₁ may be to provide a stable interaction, especially under nonstatic conditions, to allow subsequent interaction of the Gly-Pro-Hyp motif with other, activatory receptors. This study establishes GPVI as an activatory receptor crucial for the full activation of platelets by collagen.

Un–cross-linked CRP, although not causing platelet activation, does inhibit aggregation by both CRP-XL and collagen fibers, which may suggest that GPVI can recognize the monomeric molecule. Lack of activation by monomeric CRP suggests, in turn, that clustering of GPVI molecules may be necessary for activation.

In view of the potency of the repeating Gly-Pro-Hyp sequence in triple-helical, polymeric form, we believe that Gly-Pro-Hyp in collagen may be a highly specific recognition sequence for GPVI. However, we cannot rule out the possibility, as we have recently conjectured,29 that GPVI recognizes the triple helix per se and that any assembly of Gly-X-Y amino acid triplets generating a triple-helical structure may be active. Further studies are in progress to affirm the crucial importance of the Gly-Pro-Hyp motif in the recognition of collagen by GPVI.

The authors thank all blood donors and patients involved in this study for their kind cooperation and Joachim Kardoeus for preparation of the figures and editorial assistance.