• Factor XII can contribute to thrombus formation in human and nonhuman primate blood.• An antibody that blocks factor XII activation (15H8) produces an antithrombotic effect in a primate thrombosis model.The plasma zymogens factor XII (fXII) and factor XI (fXI) contribute to thrombosis in a variety of mouse models. These proteins serve a limited role in hemostasis, suggesting that antithrombotic therapies targeting them may be associated with low bleeding risks.Although there is substantial epidemiologic evidence supporting a role for fXI in human thrombosis, the situation is not as clear for fXII. We generated monoclonal antibodies (9A2 and 15H8) against the human fXII heavy chain that interfere with fXII conversion to the protease factor XIIa (fXIIa). The anti-fXII antibodies were tested in models in which anti-fXI antibodies are known to have antithrombotic effects. Both anti-fXII antibodies reduced fibrin formation in human blood perfused through collagen-coated tubes. fXII-deficient mice are resistant to ferric chloride-induced arterial thrombosis, and this resistance can be reversed by infusion of human fXII. 9A2 partially blocks, and 15H8 completely blocks, the prothrombotic effect of fXII in this model. 15H8 prolonged the activated partial thromboplastin time of baboon and human plasmas. 15H8 reduced fibrin formation in collagen-coated vascular grafts inserted into arteriovenous shunts in baboons, and reduced fibrin and platelet accumulation downstream of the graft. These findings support a role for fXII in

1106 The plasma protease factor XIIa (FXIIa) contributes to vascular occlusion in murine thrombosis models, at least partly through activation of factor XI (FXI). While there is good correlation between plasma FXI levels and thrombotic events in humans, the situation is not as clear for FXII (the precursor of FXIIa), suggesting fundamental differences in thrombus formation in mice and humans. To facilitate studies on the effects of FXII/XIIa on thrombus formation, we developed novel inhibitory antibodies to human FXII, designated 9A2 and 15H8, by immunizing FXII-deficient mice with human FXII. Using recombinant human FXII molecules that lack various domains, and chimeras in which specific domains in FXII are replaced with those from the related protein hepatocyte growth factor activator, we determined that 9A2 and 15H8 bind to the FXII/XIIa non-catalytic heavy chain at different sites. 9A2 binds on or near the EGF2 domain, while 15H8 binds to the fibronectin type I and/or kringle domain. These areas have been implicated in FXII binding to polyanionic surfaces. Saturating concentrations of 9A2 or 15H8 reduced FXII activity by 50% and 90%, respectively, in an aPTT assay using normal plasma, while combining the antibodies resulted in >95% inhibition. However, in assays in which clot formation was triggered by adding FXIIa directly to plasma, preincubation of FXIIa with either antibody did not prolong the clotting time. Furthermore, neither antibody had a strong effect in a chromogenic assay of FXI activation by FXIIa, indicating the antibodies interfere with the aPTT assay primarily by inhibiting FXII activation. FXII activation in the aPTT assay is initiated by addition of a polyanion such as silica to the plasma to induce contact activation. In vivo, polymers of inorganic phosphate (polyP) may serve a similar function. Contact activation is triggered in plasma when FXII bound to the polyanion is activated, probably by trace amounts of FXIIa or another protease present in the plasma. Once formed, FXIIa converts the zymogens prekallikrein and FXI to the proteases kallikrein and FXIa, both of which can activate additional FXIIa to amplify the process. In the presence of 9A2 or 15H8, activation of pure FXII in the presence of either silica or polyP was significantly reduced. Interestingly, the antibodies actually potentiated FXII activation by kallikrein or FXIa in the absence of a polyanion. Taken as a whole, these results suggest that binding of 9A2 or 15H8 to FXII results in conformational changes that make FXII a better substrate for kallikrein and FXIa, possibly by mimicking the effect of FXII binding to a polyanion, but that prevent activation of FXII by FXIIa (autoactivation), blunting the overall rate of activation. We tested the effects of 9A2 and 15H8 in a mouse model in which thrombotic occlusion of the carotid artery is induced by exposing the vessel to a 3.5% solution of ferric chloride. Wild type C57Bl/6 mice develop arterial occlusion within 5 to 10 minutes, while FXII-deficient mice are resistant to arterial occlusion. Infusion of human FXII into FXII-deficient mice restores the wild type phenotype. 15H8 prevented thrombus formation in mice reconstituted with human FXII, while 9A2 reduced the rate of thrombotic occlusion by 50%. In an ex vivo flow model, perfusion of human blood through collagen-coated tubes at a shear rate of 300 sec−1 results in tube occlusion by platelet and fibrin rich clot in ∼15 minutes. 15H8 effectively blocked fibrin formation and reduced platelet accumulation, preventing tube occlusion. 9A2 was also effective at preventing clot formation, but there was evidence of some fibrin accumulation over time. In summary, the monoclonal anti-human FXII IgGs 9A2 and 15H8 prevent thrombus formation in whole blood in vivo and ex vivo by interfering with FXII activation. Our data support the hypothesis that pharmacologic inhibition of FXII activation may have therapeutic utility in disorders that are driven or aggravated by the blood contact system. Disclosures: No relevant conflicts of interest to declare.