Dimers of the platelet collagen receptor glycoprotein VI bind specifically to fibrin fibers during clot formation, but not to intact fibrinogen

The platelet collagen receptor glycoprotein VI (GPVI) has an independent role as a receptor for fibrin produced via the coagulation cascade. However, various reports of GPVI binding to immobilized fibrin(ogen) are not consistent. As a collagen receptor, GPVI‐dimer is the functional form, but whether GPVI dimers or monomers bind to fibrin remains controversial. To resolve this, we analyzed GPVI binding to nascent fibrin clots, which more closely approximate physiological conditions.


| INTRODUC TI ON
To maintain hemostasis, thrombus formation is initiated when flowing blood contacts exposed subendothelial tissues of injured blood vessels, activating two reaction pathways. One pathway involves binding of the platelet-specific collagen receptor glycoprotein VI (GPVI) to subendothelial collagen exposed in injured vessels, initiating a signaling cascade leading to platelet activation, aggregation, and thrombus formation. [1][2][3][4] Constitutively present GPVI dimers comprise 29% of the total GPVI (monomers and dimers) in resting platelets. The dimers bind with high affinity to collagen fibers, but not soluble monomeric collagen, 2 and are the functional form of this receptor. 5,6 Platelet activation increases the number and clustering of dimers, bringing signaling molecules into closer proximity to enhance platelet activation. 7 The second pathway occurs simultaneously, driven by tissue factor release from damaged sub-endothelium that produces active thrombin through the coagulation cascade; thrombin converts fibrinogen in blood to fibrin monomers, which then polymerize to form a gel-like fibrin clot.
Clots are stabilized by activated factor XIII, factor XIIIa, which catalyzes fibrin crosslinking. Thrombin also activates nearby platelets, leading to phosphatidylserine expression and formation of a prothrombinase complex with coagulation factors V and VIII, enhancing thrombin formation, further driving fibrin formation and platelet activation. Many blood cells, including platelets are trapped within the fibrin clot and numerous fibrinolysis-and coagulation-related proteins are bound to the fibrin fibrils. 8 Many of these proteins, such as α2-plasmin inhibitor, plasminogen, von Willebrand factor (VWF), fibronectin, and albumin, [9][10][11] are covalently crosslinked to fibrin by factor XIII and affect fibrin clot stability. Intravital microscopy has detected fibrin fibers at the site of injury, and thrombi with tightly packed and activated platelet cores were formed in the intravascular space over the injured site. 12,13 Loosely adherent platelets formed a shell-like layer outside the platelet core. 12 Recently, GPVI was reported to bind both fibrin and fibrinogen, independent of its function as a collagen receptor, leading to platelet activation. 14,15 Thus, thrombus formation is a process combining platelet activation and blood coagulation, with GPVI involved in both systems.
Details about this role of GPVI in coagulation, whether there is direct binding to both fibrin and fibrinogen, and whether GPVI-monomer or -dimer can serve as the receptor remain controversial. Watson's group reported that GPVI monomer binds to fibrin and fibrinogen, whereas the dimer does not, using assays with immobilized fibrin or fibrinogen. 16,17 However, in our hands, these results could not be reproduced by a similar ELISA method using a fully glycosylated dimeric recombinant Fc-fusion protein of the GPVI-extracellular domain (GPVI-Fc 2 ) and indeed we found the opposite results. We found that GPVI-Fc 2 scarcely bound to fibrin, and not to fibrinogen at all, but did bind to a part of the fibrinogen molecule, the D-fragment, and to the crosslinked fibrin degradation product D-dimer, both with one-site binding kinetics. 18 We found that the corresponding GPVI-monomer did not bind well to any of these fibrinogen substrates in ELISA assays, 18 which was later confirmed by Zhang et al. 10 Using a different analysis system, Ebrahim et al 19 reported that GPVI-dimer does not bind to fibrin under flow conditions.
In the present studies, we developed a method to quantitate the binding of GPVI-dimer and -monomer to fibrin clots formed from fibrinogen by thrombin in the presence of GPVI. Our system avoids the shortcomings of ELISA assays that rely on immobilization of a low concentration of soluble fibrin. Interaction of GPVI with such immobilized fibrin or fibrinogen would be very different from the interaction with fibrin clots formed with near-physiological concentrations of fibrinogen; such clots comprise larger fibrin fibers that have a characteristic three-dimensional structure and form a complex network. Whilst retaining the simplicity of a biochemically defined system, by studying the GPVI−fibrin interaction during clot formation, our assay method is closer to pathophysiology.  20,21 Fibrin-bound factor XIII is also activated by thrombin and catalyzes both fibrin−fibrin and also plasma protein−fibrin crosslinks. Fiber formation is a very complex process, making it imperative that we understand the basic interaction between fibrin and each specific ligand, in this case GPVI, in a defined system as the first step in the absence of other proteins.
Our results indicate that GPVI-dimer binds specifically to fibrin clots but GPVI-monomer binds either nonspecifically or with very low affinity to the fibrin clots. We also suggest that GPVI contributes significantly to the binding of resting platelets to fibrin. We demonstrate here that the collagen-binding GPVI-dimer is the form that also binds specifically to fibrin clots.

| Materials
We used a fibrinogen that is a plasminogen-, VWF-, and fibronectindepleted human fibrinogen preparation (fibrinogen-3; Enzyme Research Laboratories). The fibrinogen was biotinylated using

Essentials
• Glycoprotein VI (GPVI) was reported to bind to fibrin (ogen) but its mechanism is controversial.
• Only GPVI-dimer binds to the clots like a classical receptor; the monomer does not.
• Platelet surface GPVI also binds to the clots, but integrin αIIbβ3 is the major fibrin receptor.
sulfo-NHS-SS-biotin; Thermo Scientific) according to the manufacturer's instructions. Biotin incorporation was 0.712 mol biotin per mol fibrinogen, and the derivative's clotability was the same as the unlabeled fibrinogen. Fibrinogen was also labeled with AlexaFluor-555 using a labeling kit (Invitrogen); 0.65 mol AlexaFluor-555 was incorporated per mol fibrinogen. Dimeric forms of GPVI, GPVI-Fc 2 and Revacept, were previously reported. GPVI-Fc 2 contains the extracel- Fluorescently labeled secondary antibodies were from Li-Cor Biosciences. The inhibitory GPVI dimer−specific antibody mFab-F has been reported previously. 3

| Washed platelet preparation
Citrate-anticoagulated blood from healthy volunteers was obtained with informed consent in accordance with the Treaty of Helsinki.
Washed platelets were prepared as described previously 23 and re-

| Quantitative analyses of GPVI binding to fibrin clots
The reaction mixture consisted of fibrinogen (1 mg/ml, final concentration) and equal concentrations of GPVI-Fc 2 and albumin (control protein). Thrombin (1 U/ml) was added, and the mixture was then incubated for 30 min at 37°C. Formed clots were pelleted by centrifugation at 14 000 rpm for 15 min in a microcentrifuge.
The clots were washed twice with 0.5 ml of phosphate-buffered saline and dissolved in 0.1 ml of 1% SDS/8M urea solution. Each sample was mixed with 1/3 volume of SDS-sample buffer and 10 μl of each subjected to nonreduced SDS-PAGE and immunoblotting.
The blots were stained with a mixture of 1G5 (mouse anti-pan GPVI) and anti-mouse albumin as primary antibodies and IRDye 800CW anti-mouse antibody (LI-COR) as secondary antibody; GPVI-Fc 2 and albumin bands were quantitated by the Odyssey image analyzer. Because GPVI-Fc 2 and the control protein were stained using different primary antibodies, we normalized the GPVI and control protein bands to a normalization standard (applied to the same gel), which was a sample containing 25 μg/ml each of GPVI-Fc 2 and control protein, as well as fibrinogen, with no added thrombin, which was then subjected to nonreduced SDS-PAGE/Western blotting along with the clot samples. The amounts of GPVI-Fc 2 or control protein in the fibrin clots were calculated as a percent of the respective bands in the normalization standard, and the specific binding is deduced from the difference in the percent of the GPVI-Fc 2 band and that of the control band at each GPVI-Fc 2 concentration.
To quantitate GPVI-Fc 2 binding to crosslinked fibrin clots, the reaction mixture contained fibrinogen (1 mg/ml final concentration), equal concentrations of GPVI-Fc 2 and human Fc (control protein), Ca 2+ (2 mM) and FXIII (2 μg/ml). Clot formation was induced by adding thrombin as described previously. The obtained crosslinked clots were each washed as described above and then dissolved with SDS/urea containing 2% 2-mercaptoethanol and subjected to reduced SDS-PAGE/western blotting. The blots were stained with IRDye 800CW anti-human antibody (LI-COR). GPVI-Fc and human Fc were quantitated as described previously and expressed as a percent of the normalization standard (25 μg/ml each of GPVI-Fc 2 and human Fc).
For quantitation of GPVI-His in either noncrosslinked or crosslinked clots, the western blots were stained with mouse anti-His antibody (Abcam, primary antibody) and IRDye 800 CW anti-mouse antibody (secondary antibody).

| Effects of inhibitors on GPVI-Fc 2 binding to fibrin clot
We determined how other GPVI-binding proteins: fibrinogen D-fragment, D-dimer, CRP-XL (crosslinked collagen-related peptide, a GPVI-specific agonist; synthesized by the Farndale lab), and mFab (anti-GPVI dimer) affected GPVI-Fc 2 binding to the clots. Each mixture of GPVI-Fc 2 , fibrinogen, and test proteins was preincubated for 10 min and then clotting initiated by adding thrombin. The clots were then processed as described above and analyzed for fibrin-bound GPVI-Fc 2 by SDS-PAGE/ immunoblotting.

| Confocal imaging of platelet binding to fibrin clots
Each clotting mixture contained 80% unlabeled fibrinogen and 20% Alexa Fluor 555-labeled fibrinogen (total fibrinogen concentration of 1 mg/ml). After adding thrombin (0.2 U/ml) to the mixture, it was immediately pipetted into a well of an 8-chamber Nunc Lab-Tek II Chamber slide (Thermo Scientific). The formed clots on the slides were stored at 4°C overnight and then treated with a 1/100 dilution of Protease Inhibitor Cocktail, washed twice with PBS, and blocked with 1% BSA/PBS for 1 h before use.

| Statistical analyses
The paired t-test was applied to compare the data using Prism version 8 software (GraphPad).

| GPVI-Fc 2 does not bind to fibrinogen
We previously reported, 18 assessed by ELISA, that GPVI-Fc 2 bound strongly to fibrinogen D-fragment, but only very weakly to fibrinogen. Because others 17 reported that GPVI monomers bound to fibrinogen, we verified that the lack of binding to fibrinogen was not due to steric hinderance, conformational change, or limited exposure of its binding site because of directly immobilizing it on polystyrene wells or other surface. 24 We examined GPVI-Fc 2 binding to fibrinogen by using biotinylated fibrinogen indirectly immobilized to streptavidin-coated wells; less than one biotin molecule was incorporated per fibrinogen molecule, so most of the fibrinogen would be immobilized by a single biotin-streptavidin bond. GPVI-Fc 2 bound to neither indirectly immobilized nor directly immobilized fibrinogen, but showed strong, saturable binding to fibrinogen fragment D ( Figure 1), verifying that it does not bind to intact fibrinogen.

| GPVI-Fc 2 binding to fibrin clots
We developed an assay to measure GPVI-Fc 2 binding to fibrin clots.
In this assay, the bound proteins on fibrin clots are separated from the unbound form in solution by centrifugation to form pellets and by thoroughly washing the pellets with buffer. The fibrin pellets are then dissolved in SDS/urea solution and the amounts of GPVI-Fc 2 are analyzed by immunoblotting.
The validity of our assay was shown by the following results: (1) very little albumin or human Fc, used to determine nonspecific binding, was bound to the clots, so the procedures were sufficient to remove the unbound proteins trapped in the gel. GPVI-Fc 2 binding to crosslinked clots was examined by including Ca 2+ (2 mM) and FXIII (2 μg/ml) in the clotting mixtures; these more-resistant clots were dissolved in 1% SDS/8M urea/2% mercaptoethanol and detected as GPVI-Fc (single chain) by antihuman IgG antibody because 1G5 does not react with reduced GPVI. GPVI-Fc 2 also binds specifically to the crosslinked clots, but its affinity was increased by about 2.4-fold; K D = 81.66 μg/ml for the

| Analysis of GPVI monomer binding to fibrin clot
We compared the binding of dimeric and monomeric GPVI to fibrin clots by using fully glycosylated proteins having the same amino acid In marked contrast, binding curves of GPVI-His are almost linear and nonsaturating ( Figure 5A), suggesting that the binding of GPVI monomer to the fibrin clots is nonspecific or of very low affinity.
Crosslinking fibrin did not affect GPVI-His binding.

| The effects of inhibitors on GPVI-Fc 2 binding to fibrin clots
We examined how proteins that interfere with the interaction between GPVI-Fc 2 and fibrinogen fragment-D or D-dimer affect dimer binding to the clot ( Figure 6). The inhibitory GPVI-dimer-specific antibody mFab-F partially but significantly inhibited the GPVI-Fc 2 binding to the clot (100 μg/ml, p = .070; 150 μg/ml, p = .0014; 200 μg/ml, p = .0019 n = 6), indicating a GPVI-dimer-specific interaction. Fragment-D and D-dimer weakly inhibited (p = .0129 and p = .0235, respectively), supporting the conclusion that GPVI-Fc 2 binds to these fragments. 18 Although the GPVI-specific agonist F I G U R E 1 ELISA analysis of GPVI-Fc 2 binding to immobilized fibrinogens. Biotinylated fibrinogen was immobilized to streptavidin-coated wells and fibrinogen, biotin-fibrinogen, and fibrinogen D fragment were directedly immobilized on the wells of the ELISA plate. After washing, fibrinogen-coated wells were reacted with different concentrations of GPVI-Fc 2 for 1 h and the bound GPVI-Fc 2 was detected by 1G5 (anti-GPVI primary antibody, 5 μg/ml)/IR 800CW anti-mouse IgG (secondary antibody). ○, biotinfibrinogen immobilized on streptavidin-coated surface; ▲, biotinfibrinogen directly immobilized on surface; ▲, nontreated fibrinogen directly immobilized; •, fibrinogen fragment D directly immobilized. GPVI-Fc 2 bound to fibrinogen fragment D in a dose-dependent and saturable manner. However, GPVI-Fc 2 did not bind to any of the immobilized intact fibrinogens, regardless of if they were directly or indirectly immobilized on the surface. GPVI, glycoprotein VI; GPVI-Fc 2 , GPVI-extracellular domain; IgG, immunoglobulin G F I G U R E 2 GPVI-Fc 2 binding to noncrosslinked fibrin clot. Details of the preparation of the fibrin clot and quantitation of the fibrinbound GPVI are described in the Methods. (A) A typical gel pattern for analyzing fibrin-bound GPVI-Fc 2 by our method. (B) The quantitative relationship between the amount of GPVI-Fc 2 and band strength; total clot-bound GPVI-Fc 2 and albumin (nonspecific binding) are expressed as a percent of the normalization standard (clotting mixture containing 25 μg/ml each of GPVI-Fc 2 and albumin, but no added thrombin). Specific binding of GPVI-Fc 2 (%, red circles) is calculated by subtracting nonspecific binding (albumin, %, black squares) from the total binding of GPVI-Fc 2 (%, black circles) and fit to a one-site binding model (red curve). The GPVI-Fc 2 binding to the clot is dose-dependent and saturable, exhibiting classical receptor binding kinetics. GPVI, glycoprotein VI; GPVI-Fc 2 , GPVI-extracellular domain [Color figure can be viewed at wileyonlinelibrary.com] CRP-XL strongly inhibited GPVI-Fc 2 binding to fragment-D and Ddimer, 18 it did not affect GPVI-Fc 2 binding to fibrin clots.

| Analysis of platelet receptor binding to fibrin clot
These data indicate that purified GPVI-Fc 2 binds to fibrin clots, so we formed clots from a mixture of platelets and fibrinogen to examine how the platelet-surface receptors, GPVI, GPIb, and CD61

| Confocal imaging of platelets bound to crosslinked and noncrosslinked clots
Fluorescently labelled fibrin clots were formed with or without crosslinking, and then washed platelets allowed to adhere to the clots, followed by staining for αIIbβ3. In the noncrosslinked clots, almost all the control platelets were round in shape with no protrusions, suggesting that they are not activated, with only a few well-spread platelets observed. Particularly evident was the "halo" of fibrin, colocalizing with integrin αIIbβ3 on the platelet periphery.
Dasatinib (tyrosine kinase inhibitor, which inhibits Src kinases) prevented all spreading and decreased association with fibrin, but the platelets still adhered. mFab-F did not affect the platelet binding in four of six preparations, but there is apparently less fibrin bound around the platelet perimeter.
F I G U R E 3 GPVI-Fc 2 binding to crosslinked fibrin clot. GPVI-Fc 2 binding to crosslinked fibrin clot (formed in the presence of Ca 2+ and factor XIII) was analyzed as described in Figure 2 and the Methods. (A) Typical gel pattern for analyzing the GPVI-Fc 2 binding to crosslinked fibrin clot. (B) Quantitative results. GPVI-Fc 2 was reduced, and thus detected as GPVI-Fc with anti-human IgG antibody that detected the Fc portion of GPVI-Fc and the control protein was human Fc. After normalization with the standard, total binding of GPVI-Fc 2 (black circles, %) was subtracted by the value for human Fc (black squares, %) to obtain specific binding of GPVI-Fc 2 (%, red circles); specific binding was fitted to a one-site model. GPVI-Fc 2 binding to the crosslinked clot is dose-dependent and saturable, but had higher affinity than its binding to the noncrosslinked clots. GPVI, glycoprotein VI; GPVI-Fc 2 , GPVI-extracellular domain [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 4 Relationship between band strength and applied amount of GPVI-Fc 2 . Amount of GPVI-Fc 2 as detected by our SDS-PAGE/Western blotting system nearly linearly increases with the amount of GPVI-Fc 2 applied to SDS-PAGE. This validates that our system for analyzing GPVI-Fc 2 binding to clots is quantitative in the concentration of GPVI-Fc 2 that we used. GPVI, glycoprotein VI; GPVI-Fc 2 , GPVI-extracellular domain Platelets adhered to crosslinked clots had notably different morphology. The control platelets were well-spread on crosslinked fibrin, and mFab-F did not affect this. Dasatinib appeared to decrease the number of well-spread platelets and the platelets had a spikey appearance. No sample had evident fibrin colocalized on the platelet periphery. These results suggest that resting platelets would be slightly activated on the fibrin fibers and the crosslinked fibrin would more strongly activate them, resulting in spreading.

| DISCUSS ION
Thrombus formation is a process combining platelet activation and blood coagulation, with platelet GPVI involved in both systems by functioning independently as a collagen receptor and fibrin receptor, respectively. The collagen-binding functional form of GPVI has been established to be a dimer composed of two GPVI molecules, but there are ongoing controversies about which form of GPVI, the monomer or dimer, is the functional fibrin receptor and about the GPVI binding to fibrinogen or fibrin. Our present study aimed to resolve these issues.
The discrepant observations by different laboratories may arise from using ELISA assays with immobilized fibrinogen or fibrin, a system far from the physiological environment of this interaction. ELISA cannot address the GPVI-fibrin-fiber interactions under physiological conditions, where platelets would interact with clots formed from a three-dimensional network of fibrin fibers that have been crosslinked via the transglutaminase activity of FXIIIa. Physiologically, fibrin forms a gel-like clot at higher fibrinogen concentrations by thrombin. Fibrin has specific affinity for numerous plasma proteins, including plasminogen, factor XIII, fibronectin, VWF, and prothrombin 25 ; and plasminogen binding to fibrin depends on fibrin structure. 21 Thus, the artificially formed and immobilized fibrin used in ELISA assays, typically formed with fibrinogen concentrations <100 μg/ml, is in the soluble form (monomeric or partially polymerized fibrin) and as such has properties F I G U R E 5 Analysis of Revacept and GPVI-His binding to fibrin clots. Revacept, which contains 2 chains of the extracellular domain of GPVI, was used as GPVI-dimer and GPVI-His, comprised of one chain of the same GPVI-extracellular domain as Revacept, was used as GPVI-monomer. Binding was measured by the same methods as described in Figures 2 and 3. Revacept and GPVI-His were detected by 1G5 or anti-human IgG and anti-His antibodies, respectively. (A) Typical curves for the binding of GPVI-His to noncrosslinked (left) and crosslinked fibrin (right) clots. The binding in the concentration range of 50 to 200 μg/ml is nearly linear, consistent with it being of very low affinity or nonspecific. Crosslinking the clots had no effect on GPVI-His binding. (B) Typical curves for the binding of Revacept to noncrosslinked fibrin (left) and to crosslinked fibrin (right). Revacept binding is dose-dependent and saturable and Revacept has a higher affinity for the crosslinked fibrin clot than the noncrosslinked one. □, total binding of Revacept or GPVI-His; •, specific binding; ◊, binding of control proteins not fully reflecting those of fibrin fibers in physiological clots. The inhibition of GPVI-dependent platelet activation by GPRP, an inhibitor of fibrin clot formation, [15][16][17] supports our hypothesis that GPVI binds only to fibrin fibers. Here we developed a system to quantitatively determine GPVI dimer and monomer binding to fibrin fibers in clots to reconcile the different findings. We determined GPVI binding to fibrin clots formed from thrombin cleavage of near-serum concentrations of fibrinogen, which mimics the clots formed from a three-dimensional network of fibrin fibers as seen in vivo. Furthermore, we have used fully glycosylated recombinant proteins of the GPVI extracellular domain (GPVI-Fc 2 and Revacept and GPVI-His, the one-chain form containing the same GPVI-domain as Revacept along with a His tag) to assess this interaction, also removing further uncertainty caused by using truncated or incompletely glycosylated recombinant proteins of the GPVI extracellular domain.
We verified that neither directly immobilized fibrinogen nor indirectly immobilized biotin-fibrinogen bound to GPVI-Fc 2 , but observed binding of fibrinogen D-fragment that was concentrationdependent and saturable (Figure 1), as reported previously. This suggests that the GPVI-dimer binding site in the D-domain of fibrinogen is not exposed to the surface in an intact fibrinogen molecule and its processing by the coagulation cascade to fibrin is required to make these sites available.
In our clot assay, GPVI-Fc 2 shows a typical one-site binding curve with a K D of 1.19 ± 0.26 μM for noncrosslinked fibrin and 0.50 ± 0.17 μM for crosslinked fibrin, but this was not a significant difference (p = .065). These K D values are similar to the K D (0.3-0.4 μM) that we reported for GPVI-Fc 2 binding to immobilized D fragment or D-dimer 18 by ELISA and similar to its K D (0.576 μM) for collagen as measured by surface plasmon resonance. 1 GPVI dimer, and not GPVI monomer, was indicated to bind to collagen fibers that are prepared from collagen gel but GPVI dimer does not bind to soluble (mostly monomeric) collagen. 23 The GPVI-dimer binding to fibrin clots is inhibited by the GPVI-dimer-specific antibody mFab-F ( Figure 6), and this is weakly inhibited by fragment D and D-dimer, a consistent set of observations. The modest effect of fragment D or D-dimer is reasonable because each is competing with a high concentration of fibrin. CRP-XL, however, was not inhibitory, although it strongly inhibited GPVI-Fc 2 binding to fragment-D or D-dimer, 18 suggesting that GPVI-Fc 2 binding sites for fibrin and fragment D would be slightly different and CRP-XL would bind closely to the latter site. The binding site of fibrinogen for GPVI dimer would be exposed after fibrin clot formation or proteolytic degradation by plasmin to fragment D.
The reported conformational change in fibrinogen upon its conversion to fibrin 20,21 may enable GPVI dimer to access its binding site in the D-domain.
Revacept, whose structure is similar to GPVI-Fc 2 , specifically bound to noncrosslinked (K D = 1.18 ± 1.32 μM) and crosslinked fibrin clots (K D = 0.617 ± 0.172 μM; Figure 5B). In marked contrast, monomeric GPVI-His exhibited nonsaturating binding to either type of clot, consistent with either very low affinity or nonspecific binding to the clots ( Figure 5A). The low affinity of GPVI-His could be deduced from its monovalency compared with bivalent Revacept.
Onselaer et al 16  GPVI-His might have a weak affinity for the fibrin clot and our assay method, which is more complex because the clot is a gel, may enhance the overall binding of GPVI-His. Previous papers 16,17 did not F I G U R E 6 Effects of inhibitors on GPVI-Fc 2 binding to fibrin clots. The effects of the anti-GPVI dimer-specific antibody mFab-F; collagen mimetic CRP-XL; fibrinogen fragment D, and D-dimer on GPVI-Fc 2 binding to noncrosslinked fibrin were analyzed. The concentrations of each antagonist in μg/ml are given in parentheses. mFab-F significantly inhibits the binding at concentrations greater than 100 μg/ml and fragment D and D-dimer weakly inhibited the binding at the highest concentration of 100 μg/ml. CRP-XL had no effect. *p < .05, **p < .01, compared with control (no added inhibitor). GPVI, glycoprotein VI; GPVI-Fc 2 , GPVIextracellular domain report the concentration dependence of GPVI monomer binding to fibrin but nonetheless concluded that it binds to fibrin, despite saturability of binding being ignored.
Interaction between fibrin and platelet-surface receptors αIIbβ3, GPVI, and GPIb were examined by our clot system. Fibrin clots were formed in the presence of platelets, the platelets were dissolved with platelet lysis buffer, and then proteins bound to fibrin were analyzed by immunoblotting. The very low amount of β-tubulin (used as the negative control) detected in fibrin (Figure 7) indicated that our lysis condition successfully dissolves platelets and there is no retention of nonspecifically bound proteins in the fibrin fraction.
Some platelet proteins other than GPVI and integrin β3 may also precipitate with the cytoskeleton under our lysis conditions, particularly GPIb. 26,27 To exclude this possibility, we analyzed the proteins precipitated in the absence of fibrinogen. Similar amounts of GPIb were found in the cytoskeletal precipitate and in the fibrin clot, so F I G U R E 7 Analysis of the contribution of GPVI on platelet binding to fibrin clot. Fibrin clots were formed in the presence of platelets and the following clot-bound proteins originating from platelets were analyzed: GPIb/Gc (glycocalicin), CD61 (integrin β3), GPVI, and tubulin were measured. The platelet cytoskeleton of thrombin-activated platelets was also precipitated in separate experiments to compare the difference in associated proteins. We determined the effect of inhibitors on the amount of these associated proteins. (A) An example of the gel pattern used to analyze the clot and cytoskeleton-associated proteins from platelets. The immunoblots were stained with appropriate primary antibodies followed by fluorescently labeled secondary (see Methods) and quantitated by a Licor Odyssey scanner. R (negative control; no thrombin treatment), C (control, no added inhibitor), hFab (human Fab only control), mFab (mFab-F, anti-GPVI-dimer), Int (Integrilin, antagonist of integrin αIIbβ3), Plts (platelets; original sample, no treatment). (B) Summary of the quantitation of platelet proteins associated with the clot and platelet cytoskeleton. The amounts of protein bands were calculated as percent of the original number of corresponding bands in platelets shown in plts in the figure. GPIb/Gc is found in both the cytoskeleton and clot fractions. CD61 is only associated with the clot and is significantly inhibited by integrilin (****p < .0001, compared with the control) and increased by mFab-F (**p ≤ .005, compared to the control Fab [hFab]). GPVI is only associated with the clot fraction and integrilin that inhibits αIIbβ3 increases the amount of clot-associated GPVI (*p ≤ .05). GPVI, glycoprotein VI; GPVI-Fc 2 , GPVI-extracellular domain [Color figure can be viewed at wileyonlinelibrary.com] the GPIb associated with the fibrin clot can mainly be ascribed to its association with the cytoskeleton. GPIb was reported to bind to VWF in platelet adhesion to fibrin under flow conditions, 28 but because our fibrinogen preparation is VWF-free, this interaction is not relevant to our assay. Both platelet-derived GPVI and integrin β3 were detected significantly in the clots, indicating that the fibrin clot contains the platelet proteins specifically binding to fibrin. Integrilin markedly decreased clot-bound β3, indicating that active αIIbβ3 contributes to platelet binding to fibrin. However, Integrilin increased the amount of clot-bound GPVI, suggesting that GPVI in platelets can bind to fibrin, in lieu of active αIIbβ3. This is consistent with our observation that αIIbβ3-deficient platelets from a Glanzmann's thrombasthenia patient retained low-level but significant binding to fibrin and fragment-D as assessed by both static and flow adhesion methods. 18 Zhang et al 10  reported that the addition of fibrin induced weak platelet aggregation and platelet spreading on a fibrin surface, indicating that fibrin activated the platelets. Because the clot assay requires the addition of thrombin, it would be difficult to resolve activation by fibrin from thrombin-induced activation. Thus, it would be reasonable to admit that most of the platelets would be activated by thrombin. Only activated platelets bind to fibrin, 29 so they would bind to fibrin mainly through activated αIIbβ3. Our results thus suggest that GPVI may significantly contribute to the binding of resting platelets to fibrin.
To analyze the interaction of resting platelets with fibrin clots, we allowed resting platelets to adhere to the fibrin clot and observed the morphology of the adhered platelets and fibrin (Figure 8).
Control platelets bound to noncrosslinked fibrin showed that some platelets are spread, whereas spreading is slightly decreased by F I G U R E 8 Confocal images comparing platelets bound to noncrosslinked and crosslinked clots. Clots were formed with a mixture of unlabeled and Alexa Fluor 555-labeled fibrinogen (1 mg/ml final concentration) without (noncrosslinked clots) and with Ca 2+ and factor XIII (crosslinked clots) by adding thrombin; details are described in the Methods. Platelets (control: no inhibitor, m-Fab-F: 100 μg/ml, Dasatinib: 50 nM) were allowed to bind to the clots for 30 min, unadhered platelets removed from the clots and then the clots were fixed for each condition. Noncrosslinked clots: only a few platelets were spread on the noncrosslinked clots, the rest were round in shape, with evident fibrin colocalizing with integrin αIIbβ3 on the periphery of the platelets in the control. Dasatinib-treated prevented all spreading and decreased association with fibrin, compared with the controls, but the platelets still adhered. mFab-F did not affect the platelet binding, but in four of six preparations, there is apparently less fibrin bound around the platelet perimeter. Crosslinked clots: Control platelets were well spread on the crosslinked fibrin, and mFab-F did not affect this. Dasatinib appeared to decrease the number of well-spread platelets. None of the samples had evident fibrin colocalized on the platelet periphery, compared with those in the noncrosslinked clot [Color figure can be viewed at wileyonlinelibrary.com] mFab-F and strongly decreased by dasatinib, a tyrosine kinase inhibitor that strongly inhibits GPVI-induced platelet activation but only weakly inhibits thrombin and ADP-induced platelet activation through the inhibition of outside-in activation by integrin αIIbβ3. 30,31 Furthermore, control platelets show strong colocalization with fibrin(ogen) and integrin αIIbβ3 and the colocalization is decreased by mFab-F and dasatinib. These results suggest that platelets are weakly activated on fibrin and the activation is decreased by inhibition of GPVI binding to fibrin and its activation pathway. Platelet activation by fibrin via the GPVI pathway has been reported in previous studies. [14][15][16] In contrast, platelet binding to crosslinked fibrin is very different. Most of the control platelets are spread on the fibrin, and mFab-F and dasatinib had little or no apparent effect, although dasatinib-treated platelets show filopodia and less spreading. In these crosslinked clots, we could not detect colocalization of integrin αIIbβ3 and fibrin (ogen), which might be because crosslinked fibrin forms tightly packed fibrin fibers and fibrin is unable to move to the platelet surface. These results suggest stronger activation on crosslinked fibrin than on noncrosslinked fibrin, a phenomenon that should be explored in future studies.
This study resolves the question of which form of GPVI acts as the fibrin receptor. The dimer is the form of GPVI specifically interacting with fibrin as a receptor and this is consistent with the expected higher avidity of dimeric receptors. From the analogy between collagen binding and fibrin binding by GPVI, we can make the following hypothesis: GPVI dimers bind to both collagen and fibrin fibers, but not to their monomeric forms (soluble collagen and fibrinogen, respectively). This means that a binding site specifically formed in the dimer comprising two GPVI monomers recognizes a complementary binding site in the tertiary structure of either type of fiber. It follows that GPVI monomers, which cannot form such a binding site, is unable to bind to either collagen or fibrin fibers.

ACK N OWLED G M ENTS
We thank AdvanceCOR GmbH (Martinsried, Germany) for providing us with Revacept and GPVI-His.