Analysis of the interaction of platelet collagen receptor glycoprotein VI (GPVI) with collagen. A dimeric form of GPVI, but not the monomeric form, shows affinity to fibrous collagen.

Glycoprotein VI (GPVI) is a platelet-specific glycoprotein that has been indicated to react with collagen and activate platelets. Its structure was recently identified by cDNA cloning (Clemetson, J. M., Polgar, J., Magnenat, E., Wells, T. N., and Clemetson, K. J. (1999) J. Biol. Chem. 274, 29019-29024). However, the mechanism of the interaction between collagen and GPVI has not been analyzed in detail because both collagen and GPVI are insoluble molecules. In this study, we expressed the extracellular domain of GPVI as soluble forms as follows: the monomeric form (GPVIex) and the dimeric form of GPVI fused with the human immunoglobulin Fc domain (GPVI-Fc(2)). Purified GPVIex strongly inhibited convulxin (Cvx)-induced platelet aggregation but only weakly inhibited that induced by collagen-related peptide. However, only GPVI-Fc(2), and not GPVIex, inhibited collagen-induced platelet aggregation. The dimeric form of GPVI exhibits high affinity for collagen, as concluded from measurements of GPVI binding to immobilized collagen by both the enzyme-linked immunosorbent assay and surface plasmon resonance methods. GPVI-Fc(2) bound to the surface of immobilized collagen with a dissociation constant (K(D)) of 5.76 x 10(-7) m, but the binding of GPVIex was too weak to allow estimation of this parameter. Cvx did not inhibit the binding of dimeric GPVI to collagen, indicating that the binding site of GPVI to collagen was different from that to Cvx. Taken together, our data indicate that the high affinity binding site for collagen is composed from two chains of GPVI. Furthermore, they suggest that the binding sites for Cvx are different from the collagen-binding sites and do not need to be formed by two GPVI molecules. Because dimeric GPVI is the only form that shows high affinity to fibrous collagen, our results indicate that GPVI would be present as a dimeric form on the platelet. Moreover, surface plasmon resonance indicated that there is no detectable interaction between soluble collagen and GPVI, supporting our previous observation that GPVI only reacts with fibrous collagen.

and are activated on the exposed collagen surface, leading to thrombus formation. Many proteins on the platelet surface were reported to be putative collagen receptors, but among them, only two glycoproteins have properties consistent with them being relevant collagen receptors under normal physiological conditions: one is glycoprotein (GP) 1 VI and the other is integrin ␣ 2 ␤ 1 (GPIa/IIa). Platelets deficient in either integrin ␣ 2 ␤ 1 (1,2) or GPVI (3,4) show loss of reactivity toward collagen, and antibodies against integrin ␣ 2 ␤ 1 , such as 6F1 (5) and P1E6 (6), and the Fab fragment of an anti-GPVI antibody (7) inhibited collagen-induced platelet aggregation. Snake venom convulxin (Cvx) (8) and collagen-related peptide (CRP), which mimic the collagen triple helix (9), can each activate platelets by binding specifically to GPVI, so both are useful tools for analyzing the function of GPVI. A number of studies showed that the binding of collagen, CRP, and Cvx to GPVI induced platelet activation through tyrosine phosphorylation of the Fc receptor ␥-chain, Syk, phospholipase C, and many other proteins (10), thus indicating that GPVI is a key receptor for collagen-induced platelet activation.
The cDNA cloning of GPVI revealed that GPVI belongs to the immunoglobulin superfamily that contains two C2 immunoglobulin-like domains and an Arg residue in the transmembrane region that makes a salt bridge with the Asp residue of the Fc receptor ␥-chain (11)(12)(13). Although its structure was identified, the mechanism for the reaction of GPVI with collagen remains unclear. Our previous results obtained by analyzing platelet binding to fibrous collagen suggested that GPVI is reactive only with fibrous collagen and not reactive with soluble collagen (4,18). In the present study, to facilitate the analysis of these complex reactions, we used a simplified model system in which we assessed the collagen binding ability of two forms of the extracellular domain of GPVI. We expressed the extracellular domain of GPVI as the soluble monomeric protein (GPVIex) and the soluble dimer consisting of two molecules of the fusion protein GPVI-Fc domain (from human immunoglobulin) (GPVI-Fc 2 ). The dimeric form, and not the monomeric form, shows high affinity to collagen, and only GPVI-Fc 2 inhibited collagen-induced platelet aggregation. These results indicated that the specific conformation with high affinity for fibrous collagen is constructed from two GPVI molecules and thus suggested that GPVI would be present as a dimeric form on the platelet surface, with two GPVI molecules being connected by the Fc receptor ␥-chain.

EXPERIMENTAL PROCEDURES
Expression and Purification of Soluble GPVIs-The cDNA of GPVI containing the extracellular two Ig domains of human GPVI without the signal sequence (642 bp, residues 21-234 (12)) was obtained by PCR using the GPVI cDNA as the template and the oligonucleotide TTA-AGCTTCAGAGTGGACCGCTCCCCAAGC (with the HindIII site underlined) and AATCTAGAGGAATGAGACGGTCAGTTCAGCG (with the XbaI site underlined) as the forward and reverse primer, respectively. Pfu turbo DNA polymerase (Stratagene, La Jolla, CA) was used. The PCR product was purified by using a QIAquick gel Extraction Kit (Qiagen K.K., Tokyo, Japan), digested with HindIII and XbaI, purified again, and ligated to the pSecTag vector (Invitrogen). The pSecTag vector contains c-Myc and His tag sequences at the ligation site, and the obtained fusion protein contains these epitopes at its COOH terminus. The ligation mixture was transformed into Escherichia coli XL-10 (Stratagene). The obtained construct was verified by restriction enzyme digestion and DNA sequencing. The linker sequence between the Ig signal sequence and GPVI was deleted by using the QuickChange Site-directed Mutagenesis Kit (Stratagene) according to the manufacturer's instructions. The sequence of this plasmid, pGPVI mono, was checked by dideoxy DNA sequencing of the entire coding portion. To prepare a dimeric GPVI, DNA of the human immunoglobulin Fc domain was amplified from pBOS-Fc (14) using AATCTAGAGAGCCCAAATC-TTGTGA and AGGGCCCCGGCCGTCGCACTCAT as primers and inserted to the XbaI-ApaI site of pGPVI-mono. This recombinant was designated as pGPVI-Fc.
Human embryonic kidney (HEK) 293 cells were transfected with the pGPVI mono or pGPVI-Fc vector using the calcium phosphate precipitation method (15). Stable cell lines expressing recombinant protein monomeric GPVI (GPVIex) and dimeric GPVI-Fc (GPVI-Fc 2 ) were selected in medium containing 100 g/ml Zeocin. For the purification of GPVI-Fc protein, the culture medium was centrifuged, and the obtained supernatant was applied to a column of protein A-Sepharose (Amersham Biosciences). After extensive washing with phosphate-buffered saline (PBS), GPVI-Fc 2 protein was eluted by ImmunoPure Elution Buffer (Pierce). The eluted fractions were concentrated by using an ULTRAFREE-15 (Millipore, Marlborough, MA) centrifugal filters and applied to a column of Superdex 200 (Amersham Biosciences), equilibrated with PBS (10 mM sodium phosphate, 154 mM NaCl, pH 7.4). GPVIex was purified using a HiTrap Ni 2ϩ -chelating column (Amersham Biosciences) and a Superdex 200 column. The NH 2 -terminal sequence of the recombinant GPVIex protein was confirmed to be DQS-GPLPKP by an amino acid sequencer (model 377A, Applied Biosystems). For the assay to determine inhibitory effect against platelet aggregation, samples were dialyzed against HEPES-Tyrode buffer (136 mM NaCl, 2.7 mM KCl, 0.42 mM NaH 2 PO 4 , 12 mM NaHCO 3 , 5.5 mM glucose, and 5 mM HEPES, pH 7.4).
Enzyme-linked Immunosorbent Assay (ELISA)-Microtiter wells (ImmunoMax; Nalge Nunc. International K.K., Tokyo, Japan) were reacted with 50 l of bovine acid-soluble type I collagen (20 g/ml in 0.9% NaCl) overnight at 4°C. CRP was immobilized to a React-Bind maleic anhydride-activated polystyrene plate (Pierce) using similar conditions, reacting each well with 50 l of 20 g/ml CRP in PBS overnight at 4°C. The wells were washed once with PBS and incubated with blocking buffer (PBS containing 0.2% bovine serum albumin and 0.05% Tween 20) for 1 h at room temperature. For the ELISA, 50 l of GPVI-Fc 2 solutions in blocking buffer were added to the wells and incubated for 2 h. After washing extensively with PBS containing 0.05% Tween 20, 50 l of 5000-fold diluted horseradish peroxidase-conjugated anti-human Fc antibody (American Qualex Antibodies, San Clemente, CA) was added to the wells and incubated at room temperature for 1 h. The wells were washed 4 times with PBS containing 0.05% Tween 20, and then color was developed using an o-phenylenediamine ELISA kit (Nakalai Tesque, Kyoto, Japan) for 20 -30 min at room temperature according to the manufacturer's instructions. Experiments were performed in triplicate and also repeated with different GPVI preparations. Binding to bovine serum albumin-coated wells was measured as the background value, and each of the binding values was subtracted by this background level.
To determine the effect of dimerization of monomeric GPVIex on binding to collagen, GPVIex (final concentration of 1 M) was preincubated with various concentrations of anti-Myc antibody 9E10 for 30 min and then reacted with immobilized collagen. The binding of cross-linked GPVIex to the collagen surface was measured by assessing the amount of biotinylated Cvx bound to collagen-bound GPVIex. After incubating the reactants with the immobilized collagen for 30 min and washing the wells with PBS containing 0.05% Tween 20, the wells were incubated with 50 l of biotinylated Cvx (5 g/ml) for 1 h, washed again, and the bound biotinylated Cvx was detected with a VECTASTAIN Elite kit (Vector Laboratories, Burlingame, CA). The developed color was measured as indicated above.
For the GPVIex binding assay, the amount of bound GPVIex was estimated using the anti-Myc monoclonal antibody 9E10 by the abovedescribed method with the following modifications: the amounts of GPVIex bound to immobilized collagen or CRP were estimated by incubating each well with 50 l of anti-Myc monoclonal antibody 9E10 (10 ng/ml) for 1 h, washing again, incubating with 50 l of rabbit anti-mouse IgG (HϩL)-horseradish peroxidase conjugate (Bio-Rad) for 1 h, and color developing as described above.
To determine the effects of CRP and Cvx on GPVI-Fc 2 binding to collagen, GPVI-Fc 2 (20 g/ml) was preincubated with various concentrations of CRP or Cvx, and the binding to collagen was measured by the above-described ELISA method. Data were expressed as percentages of the control value and presented as the mean value Ϯ S.E. from three independent experiments.
Surface Plasmon Resonance Spectroscopy-Analyses were carried out at 25°C with the BIAcore 2000 system (BIAcore AB, Uppsala, Sweden) using HES buffer (10 mM HEPES, 150 mM NaCl, 2 mM EDTA, pH 7.4). Bovine type I collagen in 10 mM sodium acetate buffer, pH 5.0, was covalently coupled to a CM5 chip (Biacore) using an Amine Coupling Kit (Biacore) according to the manufacturer's instructions. GPVI-Fc 2 in 10 mM sodium acetate buffer, pH 5.5, was also coupled to a sensor chip by the same procedure. Regeneration of the collagen surface was achieved by running 15 l of 10 mM HCl through the flow cell at 30 l/min two times. A control surface was reacted with the amine coupling reagent in the absence of ligand and then blocked with ethanolamine.
GPVIex and GPVI-Fc 2 solutions of several concentrations were perfused over the immobilized collagen at a flow rate of 20 l/min, and the resonance changes were recorded. The sensorgram of the immobilizedcollagen surface was subtracted by that of the control surface, and the data thus obtained were analyzed by nonlinear curve fitting of the Langmuir binding isotherm with BIAevaluation software (Biacore). The binding of soluble collagen was measured by injecting soluble bovine type III collagen (50 g/ml in HES buffer containing 0.05% Tween 20) over the surface of immobilized GPVI-Fc 2 at a flow rate of 10 l/min. The preparation of soluble bovine type III collagen was described in detail previously (16).
Platelet Preparation-Whole blood was drawn from the cubital vein of healthy volunteers into 0.1 volume of 3.8% sodium citrate. The platelet-rich plasma (PRP) was obtained, added with sodium prostaglandin I 2 (Funakoshi, Tokyo, Japan) at the final concentration of 0.1 g/ml, and then centrifuged at 900 ϫ g for 12 min to isolate the platelets. The obtained platelets were washed once with 6.85 mM citrate, 130 mM NaCl, 4 mM KCl, and 5.5 mM glucose, pH 6.5. Then the washed platelets were finally suspended with HEPES-Tyrode buffer at concentrations of 2-4 ϫ 10 8 platelets/ml.
Platelet Aggregation Assay-Platelet aggregation was monitored by a whole blood aggregometer (Chrono-Log Corp., Haverton, PA) with stirring at 37°C using PRP or washed platelets in HEPES-Tyrode buffer. Platelets were activated with CRP (0.2 g/ml), equine type I collagen (1-2 g/ml, Chrono-Log), or Cvx (23.8 ng/ml); and the change of transmission was recorded.
Other Materials-CRP was synthesized by the method of Morton et al. (9) as previously described (18). Cross-linked CRP has a platelet activating activity at concentrations of 0.2 g/ml or less. Cvx was purified from Crotalus durissus terrificus venom (Miami Serpentarium Laboratories, Punta Gorda, FL) according to the method described previously (8). Cvx was also biotinylated by using sulfo-NHS-biotin (Pierce) as described before (12). Bovine type I and type III collagens were obtained from Koken Co., Ltd. (Tokyo, Japan). The anti-Myc monoclonal antibody 9E10 was purchased from Santa Cruz Biotechnology (Santa Cruz, CA) or Immunotech International (Marseilles, France).

Production and Isolation of Soluble Monomeric and Dimeric
GPVIs-The extracellular domain of GPVI (residues 21-234) was expressed as the monomeric (Myc-His 6 ) tag fusion protein and as the dimeric human immunoglobulin Fc domain fusion protein. These GPVIs were expressed as secreted soluble proteins using a eukaryotic cell line, HEK 293 cell, which prevented misfolding and any non-glycosylation of the expressed proteins, as occasionally happened when using E. coli. About 2 mg/liter of the recombinant GPVIs were obtained from the culture mediums for both forms of GPVI. The molecular mass of GPVIex is 41 and 42 kDa under non-reduced and reduced conditions in SDS-PAGE, respectively (Fig. 1, lanes 3 and 4). GPVI-Fc 2 has a molecular mass of 150 and 68 kDa under the non-reduced and reduced conditions in SDS-PAGE, respectively ( Fig. 1, lanes 1 and 2). The molecular mass of the purified GPVI-Fc 2 was estimated to be ϳ180 kDa from gel filtration (data not shown), confirming that the purified GPVI-Fc 2 is present as a dimer. Both biotinylated Cvx (12) and anti-human GPVI monoclonal antibody 204-11 2 recognize native GPVI and recombinant GPVIex and GPVI-Fc 2 only under the non-reduced condition, suggesting that these recombinant GPVIs have conformations similar to that of native GPVI (data not shown).
Effect of Recombinant GPVIs on Platelet Aggregation-Next, we tested the activities of the monomeric and the dimeric forms of GPVI against platelet aggregation. As shown in Fig. 2, GPVIex did not inhibit collagen-induced platelet aggregation even at a high concentration (100 g/ml), but it did inhibit cross-linked CRP-induced platelet aggregation, although very weakly (Fig. 3). In contrast, GPVI-Fc 2 at 10 g/ml strongly inhibited cross-linked CRP-induced platelet aggregation. Cvxinduced platelet aggregation was markedly inhibited by both GPVIex and GPVI-Fc 2 at concentrations of 0.5 and 0.56 M, respectively (data not shown).
GPVIex did not inhibit collagen-induced platelet aggregation even at a concentration of 100 g/ml, but GPVI-Fc 2 abrogated collagen-induced platelet aggregation at 20 g/ml without preincubation (Fig. 2B). These results indicate that the two forms of GPVI show different binding affinities to collagen itself and the collagen-mimetic CRP.
Binding of Monomeric and Dimeric GPVIs to CRP and Collagen-The binding of monomeric and dimeric GPVIs to immobilized ligands was analyzed by the ELISA method (Fig. 4). Both proteins showed dose-dependent and saturable binding to the immobilized-CRP surface (Fig. 4A). EDTA did not affect the binding of either of the recombinant GPVIs to CRP and collagen (data not shown). Although it is impossible to compare quantitatively the extent of bindings of GPVIex and GPVI-Fc 2 because we used different antibodies to detect these proteins, we could not detect any strong binding of GPVIex to type I and type III collagens (Fig. 4B). Even when the sensitivity was increased by using a longer color-developing time, only a small increase in color development was observed at a high concentration of GPVIex, 400 g/ml (data not shown). In contrast to GPVIex, the bindings of GPVI-Fc 2 to bovine type I and type III collagens were each dose-dependent, becoming saturated at 400 g/ml (Fig. 4C). These results also indicated that collagen preferably binds to dimeric GPVI-Fc 2 .
The Effect of Dimerization of Monomeric GPVIex on Binding to Bovine Type I Collagen-The above results suggested that the dimerization of GPVI may be necessary to induce a specific conformation with a high affinity for collagen. To test this hypothesis, we cross-linked the monomeric GPVIex with the anti-Myc monoclonal antibody 9E10, which enabled us to make dimeric GPVIex because each chain of the antibody could bind to one GPVIex molecule through the Myc tag at its COOHterminal. Fig. 5 illustrates the effect of dimerization on the affinity of GPVIex for collagen. GPVIex preincubated with 9E10 exhibited enhanced binding to collagen, with the binding enhancement depending on the molar ratio of 9E10 to GPVIex. At the molar ratio of 1:2 (9E10:GPVIex), the binding enhancement was maximum, with a 3.4-fold higher amount of binding than that of GPVIex alone. When the binding of cross-linked GPVIex was compared with that of GPVI-Fc 2 , from the developed color using the same measuring method, the binding of cross-linked GPVIex corresponded to the amount of binding obtained at about 160 nM GPVI-Fc 2 , indicating that the binding of the cross-linked GPVIex was about 10 -20% of that of GPVI-Fc 2 (data not shown). These curves also indicate that an excess amount of 9E10 rather decreases the binding of GPVIex to collagen, which would be explained by a decrease in the amount of dimerized GPVIex at the higher ratio of the antibody to GPVIex, where there is a higher chance for antibody reacting with only one GPVIex molecule, and thus not forming a dimer. These results support the hypothesis that the dimerization of GPVI would be responsible for the high affinity binding of GPVI-Fc 2 to collagen and negate the possibility that the high affinity is due to GPVI-Fc 2 having a conformation different from that of GPVIex.
The Effect of CRP and Cvx on Dimeric GPVI-Fc 2 Binding to Bovine Type I Collagen-Because Cvx and CRP were reported to be specific ligands for GPVI, we determined their ability to compete with collagen for the binding to GPVI-Fc 2 by the ELISA method. CRP inhibited the binding of GPVI-Fc 2 to immobilized collagen in a dose-dependent manner (Fig. 6A). The concentration required for 50% inhibition of binding (IC 50 ) was calculated to be 4.3 Ϯ 0.4 g/ml. However, Cvx did not inhibit the binding of GPVI-Fc 2 to collagen but instead enhanced the apparent binding to collagen at low concentrations (Fig. 6B). A possible explanation for the enhancing effect of Cvx can be proposed on the basis of its multiple subunit structure (17). Cvx has multiple binding sites for GPVI, and GPVI binds with Cvx at sites different from the collagen-binding ones. As a result, Cvx would be able to bind multiple molecules of GPVI and thereby help to accumulate more GPVI at the collagen surface. Furthermore, Cvx did not inhibit GPVI-Fc 2 binding to collagen, providing further evidence that the Cvx-binding site of GPVI is different from its collagen-binding site. Because the binding to type I collagen was almost completely inhibited by CRP, the collagen-binding site of GPVI may be the same or shared with that for CRP.
Binding Analysis of GPVI-Fc 2 and Collagen Using Surface Plasmon Resonance-Kinetic analysis of the binding of collagen and the GPVIs was performed by the SPR method. Collagen and CRP were immobilized on sensor chips under acidic conditions, and the interactions between flowing GPVI-Fc 2 or GPVIex and immobilized collagen or CRP were measured under physiological conditions. The sensorgrams at different ligand concentrations were obtained and normalized by subtracting the background signals from the collagen (Fig. 7A) or CRP (Fig. 7B) responses. The kinetic data obtained from 3 independent experiments are summarized in Table I  table, dissociation constants were obtained by two calculation methods, from k on and k off and from the equilibrium binding equation using resonance units calculated at the equilibrium. Curve fitting indicated that our data fit better to a two-state model than a one-state one, suggesting that there may be a conformational change of the complex after the initial association of GPVI-Fc 2 with collagen (Table II). However, even in the two-state model, the apparent K D values are not substantially different from those of the one-state model, and the transition rate is very small, suggesting a rather small contribution of the transition (conformational change) of the complex to GPVI binding. The K D value of GPVI-Fc 2 indicated that dimeric GPVI has at least 10-fold higher affinity to collagen than CRP. The binding of GPVIex to the collagen surface was so weak that  reliable kinetic parameters could not be calculated from the obtained experimental data (Fig. 8A). In contrast to collagen, the K D value of GPVIex for CRP was measurable, being 8.5 Ϯ 0.1 ϫ 10 Ϫ5 M (Fig. 8B).
In the above experiments, the interaction of GPVI and immobilized collagen was analyzed. These immobilized collagens would contain some fibrous collagen because the SPR experiment was performed under physiological conditions, after the collagen was first immobilized to the chips under acidic conditions. To analyze the interaction of GPVI with only the soluble form of collagen, we prepared soluble type III collagen as described under the "Experimental Procedures," and we analyzed the interaction by the SPR method. As shown in Fig. 9, the sensorgram indicates that there is no significant interaction of the immobilized GPVI-Fc 2 with soluble collagen, although the higher concentrations of soluble collagen could not be tested because only a low concentration of soluble collagen could be obtained. This result clearly indicates that GPVI does not have any significant affinity for soluble collagen, and the interaction between immobilized collagen and GPVI is solely attributable to the interaction with fibrous collagen only. DISCUSSION GPVI is a platelet-specific membrane protein whose structure was recently identified from cDNA cloning (11)(12)(13). Serving as a physiological collagen receptor on platelets, its function is to bind to collagen and activate platelets, as concluded from studies on GPVI-deficient platelets from patients (3,4,18). It has been hypothesized that the high affinity interaction of platelets with collagen through integrin ␣ 2 ␤ 1 , another collagen receptor, functions in platelet adhesion, and the lower affinity interaction between collagen and GPVI mainly serves to induce activation pathways in platelets (10,19). Although quantitative studies on the interaction between integrin ␣ 2 ␤ 1 and collagen have been performed (16,20,21), no quantitative analyses of the interaction between GPVI and collagen have been reported. In this study, we prepared soluble forms of recombinant GPVI, and we analyzed their interaction with collagen.
To obtain a sufficient amount of the recombinant proteins, we inserted the Ig signal sequence instead of the original sequence in the pSecTag vector expression system. The Ig signal sequence significantly increased the secretion of the recombinant GPVIs, which could be ascribed to the short hy-drophobic core in the signal sequence of GPVI (data not shown). We expressed the extracellular domain of GPVI conjugated with Myc and His tags at the COOH-terminal end (GPVIex) and the fused form of this extracellular domain with the IgG Fc domain (GPVI-Fc 2 ). GPVIex is a monomeric form, and GPVI-Fc 2 is a dimeric form, in which two GPVI-Fc molecules are cross-linked by disulfide bonds formed from the Cys in the Fc domain of each molecule (Fig. 1).
GPVI has been indicated to form a complex with Fc receptor ␥-chain through the ionic bonds between Arg of GPVI and Asp of the FcR ␥-chain in the transmembrane domains (12,22). Because FcR ␥-chain is present as a dimer cross-linked by a disulfide bond, GPVI should also be present as a dimer form on the platelet surface. Jandrot-Perrus et al. (13) reported that the GPVI extracellular domain-Fc domain fusion protein inhibited collagen-induced platelet aggregation when collagen was preincubated with it. Our data (Fig. 2) indicating that only GPVI-Fc 2 , and not GPVIex, inhibited collagen-induced platelet aggregation is consistent with the observations of Jandrot-Perrus et al. (13); however, preincubation with collagen was unnecessary for the inhibitory effect of our dimeric protein. In contrast, both the monomer and the dimer inhibited Cvx-induced platelet aggregation at a similar molar concentration. The inhibitory effect of the GPVI proteins on cross-linked CRP-induced aggregation was intermediate between its effects on collagen-and Cvx-induced platelet aggregations. The GPVIex weakly inhibits the cross-linked CRP-induced platelet aggregation. These results suggested that GPVI-Fc 2 has a binding affinity to collagen, whereas GPVIex, the monomeric form, does not.
It is possible that a difference in conformation may explain the different reactivity of GPVIex and GPVI-FC 2 , but several of our observations indicate that this is unlikely. In Fig. 5, we cross-linked GPVIex by reacting the anti-Myc antibody 9E10 with the COOH-terminal Myc tag of GPVIex, and we showed that the cross-linked GPVIex also exhibited the ability to bind to immobilized collagen, although its binding activity was not the same as that of GPVI-Fc 2 . The lower binding efficiency of the cross-linked GPVIex compared with that of GPVI-Fc 2 can be explained from the loosely cross-linked GPVIex. More importantly, we observed that the efficiency of binding becomes lower at higher antibody to GPVIex ratios, suggesting the importance of dimerization by the antibody. These results support the hypothesis that the amount of the dimerized form of GPVI is mainly correlated to the ability to bind collagen, thus dismissing the possibility that the collagen binding portion of GPVI-Fc 2 has a conformation intrinsically different from that of GPVIex.
ELISA and SPR were used to analyze the GPVI-collagenbinding reaction. Both methods indicated that GPVI-Fc 2 binds to collagen, but GPVIex has essentially no affinity for collagen. SPR experiments indicated that GPVI-Fc 2 has a K D of 5.76 Ϯ 0.64 ϫ 10 Ϫ7 M (Table I). This value is markedly higher (lower affinity) than the K D value that we obtained for the soluble collagen-integrin ␣ 2 ␤ 1 interaction, 6 -86 ϫ 10 Ϫ9 M (16,20). This indicates that the activated integrin ␣ 2 ␤ 1 is mainly related to the tight binding of platelets to collagen. GPVIex did not show any significant interaction with collagen by ELISA and SPR methods. Even very high concentrations of GPVIex could only bring about slight inhibitions of platelet aggregation or collagen binding (Fig. 3B). On the other hand, GPVIex could bind CRP in a dose-dependent manner (Fig. 4A) and inhibit Cvxinduced platelet aggregation. Thus, the CRP and Cvx binding abilities of GPVIex are similar to those of the dimeric GPVI Fc 2 , and it is only the affinity toward collagen that differs. In addition, dimerization of monomeric GPVIex with monoclonal antibody increases its binding to immobilized collagen. Taken together, our results indicate that the high affinity to collagen is attributable to the dimeric structure of GPVI. Such enhancement of affinity induced by receptor dimerization has been reported for several Ig superfamily receptors (23).
In one of our previous papers (16), we showed that platelet binding to soluble collagen is characteristically very different from platelet binding to fibrous collagen. Platelet binding to soluble collagen is strongly dependent on divalent cations and    strongly inhibited by anti-integrin ␣ 2 ␤ 1 antibodies, whereas platelet binding to fibrous collagen is independent of the former and only slight affected by the latter. The binding of GPVIdeficient platelets to fibrous collagen is decreased in the presence of EDTA (4), which suggests that GPVI does not bind to soluble collagen and specifically binds to fibrous collagen. To test this hypothesis, we measured the binding of soluble collagen to the immobilized GPVI-Fc 2 by the SPR method, and we found that there is no interaction between GPVI-Fc 2 and soluble collagen (Fig. 9). This also suggested that the GPVI binding to immobilized collagen measured by the ELISA or SPR method is not due to monomeric collagen but is solely attributable to the binding to fibrous collagen that is formed when the wells or SPR chip surfaces are washed with physiological buffer, after the initial acidic conditions employed to coat the surfaces.
GPVI-Fc 2 also has affinity to CRP. It inhibited the crosslinked CRP-induced platelet aggregation and bound to immobilized CRP. The K D of the interaction between GPVI-Fc 2 and CRP was 5.26 Ϯ 5.89 ϫ 10 Ϫ6 M. The larger K D value compared with the K D value for the interaction with collagen would be due to the larger dissociation rate, which suggested that some portion of the collagen structure other than the regular Gly-Pro-Hyp repeats would be involved in the dissociation of the bound collagen from GPVI. Although both the ELISA and SPR experiments showed the similar binding of dimeric and monomeric GPVI to CRP (Fig. 4A and Fig. 8B), GPVIex was far less inhibitory against cross-linked CRP-induced platelet aggregation than GPVI-Fc 2 (Fig. 3). This discrepancy would come from the difference between the immobilized CRP and cross-linked CRP. GPVI-Fc 2 is suggested to bind more strongly to crosslinked CRP than GPVIex.
GPVIex and GPVI-Fc 2 strongly inhibited Cvx-induced platelet aggregation at similar molar concentrations (data not shown). Also, Cvx did not inhibit GPVI-Fc 2 binding to immobilized collagen, but instead rather enhanced it (Fig. 6B). These results indicated that the Cvx-binding sites of GPVI are completely different from the collagen-binding sites of GPVI. This is further supported by the results of Nieswandt et al. (24) who found that the anti-GPVI antibody JAQ-1 inhibits the collageninduced aggregation but not the Cvx-induced one.
Zheng et al. (22) reported that GPVI-expressing RBL cells did not react with collagen, although the cells react with Cvx with a similar concentration dependence to that of platelets. These GPVI-expressing cells show weak reactivity to CRP. The reactivities of these GPVI-expressing RBL cells to collagen receptor agonists are very similar to the reactivities of our monomeric GPVIex to these agonists. GPVIex does not bind to collagen but reacts with Cvx at a concentration similar to that reactive with GPVI-Fc 2 . GPVIex also reacts weakly with CRP. These results suggest that the GPVI expressed on the RBL cells would be present as a monomeric form complexed with the FcR ␥-chain. Recently, the same group advanced their studies and showed that the RBL cells expressing GPVI on the cell surface at a similar density to that of native platelets can react with collagen and induce intracellular Ca 2ϩ release (25). RBL cells expressing about 50% the amount of GPVI in platelets showed much weaker reactivity to collagen. If we assume that 100% density would indicate 100% formation of GPVI dimer, a 50% density would suggest that the most of the GPVI-FcR ␥-chain would be in a 1-to-1 with monomeric GPVI. These results also support our hypothesis that the dimeric form of GPVI would be the active form of GPVI, and thus GPVI must be expressed at a high density to obtain an active GPVI in the cultured cells. However, platelets from the parents of a previously reported GPVI-deficient patient (4) and platelets from heterozygous FcR ␥-chain-deficient mouse (25) expressed only about one-half the normal amount of GPVI but had completely normal reactivity to collagen. Although further studies are required to prove conclusively the presence of the dimeric form of GPVI on the platelet surface, our present results would suggest the presence of a biosynthesis pathway for the GPVI-FcR ␥-chain complex that specifically produces a dimeric complex in platelets and megakaryocytes.
In this study, we prepared two forms of GPVI, the monomeric form GPVIex and the dimeric form GPVI-Fc 2 , and our data presented herein indicate that the dimeric GPVI-Fc 2 has high affinity to fibrous collagen. These results, along with other data from our laboratory and by other groups, suggested that GPVI would be present as a dimeric form in platelets. Our studies also show that the dimeric form of GPVI, GPVI-Fc 2 , could be useful for analyzing the interaction of GPVI with insoluble collagen.