Antibodies directed against GPVI are of interest, because some have the ability to block the interaction of GPVI with collagen [18,38], and others induce platelet GPVI depletion  (Table 1). The strengths of antibodies are (i) their specificity, and (ii) their capacity to easily saturate their target, owing to their high affinity and the relatively low number of GPVI copies at the platelet surface (4000 ± 1000).
GPVI-depleting antibodies As mentioned above, the first report of a GPVI-depleting antibody was made by Sugiyama et al.  in a GPVI-deficient patient. Interestingly, the antibody-induced GPVI deficiency observed in vivo could not be reproduced in vitro. Similarly, the GPVI downregulation induced by JAQ1 antibodies in mice occurs exclusively in vivo . Several cases of antibody-induced GPVI depletion have since been reported, with heterogeneous behavior: some induce GPVI depletion in vivo  and others in vivo and in vitro , and in some cases, but not all, the shedding of GPVI is shown by the presence of soluble GPVI in plasma [50,82]. The mechanism of GPVI downregulation appears to be dependent on activating signals and to involve internalization of the antigen–antibody complexes and/or shedding of the GPVI extracellular domain by metalloproteases . Recently, Takayama et al.  obtained recombinant antibodies mimicking the patient’s antibody. When injected into monkeys, the mouse mAb mF1232 and the chimeric antibody cF1232 induced loss of platelet surface GPVI by endocytosis without thrombocytopenia, and, interestingly, GPVI internalization was reproduced in vitro in the presence of cAMP-elevating agents. Antibody-induced GPVI deficiency is proposed as a novel therapeutic approach, with the potential disadvantage that the depletion is long-lasting.
Inhibitory antibodies A small number of inhibitory antibodies have been identified. It is important to consider that inhibitory antibodies should preferably be monovalent, to avoid clustering of GPVI or interaction with FcγRIIA. This is exemplified by the mAbs 9O12 and 1G5, which activate platelets in an FcγRIIA-dependent and FcγRIIA-independent manner, and which both induce activation-dependent shedding of GPVI in vitro [84,85].
Human single-chain antibodies have been obtained by phage display [18,22,86]. The 10B12 antibody binds to the D1 domain, and inhibits platelet interactions with CRP and collagen and thrombus formation. The 1C3 antibody maps a region of the D2 domain involved in the clustering of the receptor, and partly inhibits GPVI binding to its ligands. BLO8-1 is a human domain antibody that recognizes an epitope within the collagen-binding domain of GPVI, and inhibits CRP-induced platelet activation . The limit of natural antibodies is their relatively low affinity.
Several anti-human mAbs also have inhibitory properties. The 9O12, 204-11 and 5C4 antibodies were reported to limit the binding of GPVI to collagen [38,71,88]. The Fab fragments of 5C4, 1G5, OM2, OM4 and 9O12 have inhibitory effects on GPVI-mediated platelet activation in vitro and ex vivo or in vivo [38,46,85,89–93]. OM2 and OM4 were obtained by immunizing GPVI-deficient mice with human recombinant GPVI, and have good affinity for GPVI. The injection of OM2 Fab into cynomolgous monkeys resulted in inhibition of ex vivo collagen-induced platelet aggregation . In a model of cyclic flow reduction in the rat carotid artery, the number of complete occlusions was significantly reduced by intravenous administration of OM4 Fab, which cross-reacts with rat GPVI without prolonging the bleeding time .
Numerous studies have been conducted with 9O12, and their results support the assumption that targeting GPVI has a good antithrombotic efficacy with no bleeding side effects. The 9O12 antibody is a mAb with high affinity for human GPVI (10−9 M). In addition to blocking platelet adhesion and aggregation under static and arterial flow conditions, 9O12 Fab reduced platelet procoagulant activity by preventing collagen-induced phosphatidylserine exposure [38,94]; 9O12 Fab also impaired platelet adhesion and aggregation, and reduced phosphatidylserine exposure when blood was perfused at arterial shear rates on extracts of atherosclerotic plaques . The specificity of 9O12 is restricted to human and non-human primate GPVI. After administration of an intravenous bolus dose of 9O12 Fab to cynomolgous monkeys, neither the platelet count nor GPVI expression at the platelet surface was modified . Binding of 9O12 Fab to the platelet surface was rapid (peak at 30–150 min postinjection) and reversible, with a 75% decrease at day 1. There was no sign of bleeding. In contrast, ex vivo collagen-induced platelet aggregation was totally inhibited, platelet adhesion to collagen and thrombus growth at arterial shear rates were reduced as well as thrombin generation.
These data were promising, but there was still a need to obtain more evidence of efficacy and safety in arterial thrombosis models. Mice are obviously not entirely satisfactory for preclinical studies, as most of the blocking antibodies directed against human GPVI do not bind to mouse GPVI, and also because sequence variability in the intracellular and extracellular domains of GPVI could influence collagen-induced responses [14,18]. We thus constructed a transgenic mouse expressing human GPVI, to be subjected to various models of arterial thrombosis . These mice, obtained by knock-in, were viable and fertile, and platelets exhibited normal responses to all of the agonists tested. Surface-expressed human GPVI had a similar density to that observed on human platelets, and bound 9O12 Fab. After an intravenous bolus injection, platelet-bound 9O12 Fab peaked at 30 min. No modifications in the platelet count or GPVI expression were observed, and tail bleeding time and blood loss were not increased as compared with controls, in agreement with the data obtained in monkeys. However, ex vivo collagen-induced platelet aggregation and thrombus formation under flow conditions were profoundly impaired. The 9O12 Fab injection protected human GPVI-expressing mice against lethal thromboembolism induced by injection of a collagen/adrenaline mixture. In addition, 9O12 Fab provided protection against thrombosis after superficial laser injury of mesenteric arterioles, after deep laser injury under conditions where thrombin was blocked, and after mechanical injury to the aorta. The human GPVI-expressing mouse also allows study of the effects of GPVI antagonists on atherothrombosis by grafting of their bone marrow into irradiated ApoE−/− mice. The administration of 9O12 Fab proved to be highly effective in preventing arterial thrombosis triggered by the injury of carotid artery atherosclerotic plaques in these mice (Hechler B, Mangin P, Loyau S, Jandrot-Perrus M, Gachet C, manuscript in preparation).
Taken together, these data tell us that the human GPVI-expressing mouse is suitable for the preclinical in vivo evaluation of GPVI antagonists in terms of safety and efficacy.
Because the immunogenicity of murine antibody fragments is a major obstacle to their clinical development, it was necessary to carry out a humanization procedure. A humanized 9O12 scFv was obtained by grafting the complementarity-determining regions onto closely related human antibody variable domains and carrying out structural refinements . The humanized single-chain antibody (hscFv 9O12) retained the affinity, specificity and functionality of 9O12 Fab, including inhibition of collagen-induced platelet activation. Therefore, hscFv 9O12 represents the building block for the production of a monovalent humanized Fab with pharmacokinetic properties that are more suitable for therapeutic purposes than those of scFvs.