Platelet glycoprotein V spatio-temporally controls fibrin formation

The activation of platelets and coagulation at vascular injury sites is crucial for hemostasis but can promote thrombosis and inflammation in vascular pathologies. Here, we delineate an unexpected spatio-temporal control mechanism of thrombin activity that is platelet orchestrated and locally limits excessive fibrin formation after initial hemostatic platelet deposition. During platelet activation, the abundant platelet glycoprotein (GP)V is cleaved by thrombin. We demonstrate, with genetic and pharmacological approaches, that thrombin-mediated shedding of GPV does not primarily regulate platelet activation in thrombus formation but rather has a distinct function after platelet deposition and specifically limits thrombin-dependent generation of fibrin, a crucial mediator of vascular thrombo-inflammation. Genetic or pharmacologic defects in hemostatic platelet function are unexpectedly attenuated by specific blockade of GPV shedding, indicating that the spatio-temporal control of thrombin-dependent fibrin generation also represents a potential therapeutic target to improve hemostasis. Through genetic mouse models, pharmacological interventions and in vitro assays, Beck et al. show that thrombin-mediated platelet glycoprotein V (GPV) shedding does not affect platelet activation but prevents excessive thrombin-mediated fibrin deposition and thereby controls hemostasis, thrombosis and thrombo-inflammation. The GPV-mediated spatio-temporal control of fibrin formation on thrombogenic surfaces could be targeted to restrict thrombosis while preserving hemostasis.

The activation of platelets and coagulation at vascular injury sites is crucial for hemostasis but can promote thrombosis and inflammation in vascular pathologies. Here, we delineate an unexpected spatio-temporal control mechanism of thrombin activity that is platelet orchestrated and locally limits excessive fibrin formation after initial hemostatic platelet deposition. During platelet activation, the abundant platelet glycoprotein (GP)V is cleaved by thrombin. We demonstrate, with genetic and pharmacological approaches, that thrombin-mediated shedding of GPV does not primarily regulate platelet activation in thrombus formation but rather has a distinct function after platelet deposition and specifically limits thrombin-dependent generation of fibrin, a crucial mediator of vascular thrombo-inflammation. Genetic or pharmacologic defects in hemostatic platelet function are unexpectedly attenuated by specific blockade of GPV shedding, indicating that the spatio-temporal control of thrombindependent fibrin generation also represents a potential therapeutic target to improve hemostasis.
Hemostasis is the physiological mechanism that limits bleeding after blood vessel injury through intertwined activation of circulating platelets and the plasmatic coagulation cascade 1 . The adhesion of platelets to extracellular matrix proteins and von Willebrand factor (VWF) initiates the hemostatic response that is supported by exposure of subendothelial tissue factor (TF), which triggers coagulation and local thrombin generation 2 . This results in a fibrin network encasing platelets in a stable thrombus 3 . Thrombin generation requires feed-forward reactions that involve platelet activation by thrombin-mediated cleavage and activation of G protein-coupled protease-activated receptors (PARs) 4 and amplification of coagulation reactions on the surface of activated platelets 5 . Generated thrombin forms fibrin and thereby stabilizes thrombi through platelet receptor GPIIb/GPIIIa engagement and activates factor XIII (FXIII) to cross-link fibrin fibers 6 . These processes are regulated with high fidelity 7 to ensure efficient hemostasis while preventing thrombosis and thrombo-inflammatory diseases 8,9 . In addition, thrombin activity in the circulation is limited by specific plasmatic coagulation inhibitors and by thrombomodulin on endothelial cells, which captures thrombin to initiate the coagulation-regulatory and vascular-protective protein C pathway 10 . Article https://doi.org/10.1038/s44161-023-00254-6 The delineated pathway of enhanced in vitro thrombin signaling in Gp5 −/− platelets could not explain the similar prothrombotic phenotype of Gp5 −/− and Gp5 dThr mice in vivo, suggesting that shed GPV regulates thrombus formation by a mechanism unrelated to the regulation of platelet activation following vascular injury. We therefore next asked whether platelet procoagulant function might be regulated by GPV. Measurements of TF-initiated thrombin generation in PRP did not, however, uncover differences between Gp5 −/− , Gp5 dThr and WT platelets (Extended Data Fig. 2f-k), in line with previous results from GPV-deficient platelets 12 , excluding alterations in platelet membrane procoagulant activity.
We next evaluated whether the known collagen interaction of GPV might contribute to thrombus growth modulation by GPV. Platelet activation is triggered through two major signaling pathways. Specifically, soluble agonists, including thrombin and secondary mediators ADP and thromboxane A2, act through G protein-coupled receptors, whereas immobilized and/or multimeric ligands signal through immunoreceptor tyrosine-based activation motif (ITAM)-coupled receptors, C-type lectin-like receptor 2 (CLEC-2) and GPVI. Platelet GPVI is the major activating collagen receptor, and GPVI deficiency and antagonism protect from arterial thrombosis with more moderate effects on hemostasis 30 . We analyzed thrombus formation in the absence or presence of platelet GPVI to uncover potential collagen-binding functions of GPV. GPVI was immunodepleted from platelets by injection of the anti-GPVI antibody JAQ1 (ref. 31 ) 5 d before inducing FeCl 3 mesenteric arteriole injury (Extended Data Fig. 3a,b). As reported previously 31 , GPVI depletion markedly attenuated occlusive thrombus formation in WT mice in vivo. Surprisingly, loss of GPVI was without effect in the absence of GPV, and the shortened occlusion times of Gp5 −/− mice persisted even after GPVI depletion in two distinct vascular beds (Extended Data Fig. 3c-e).
In addition, GPV deficiency prevented the prolongation of the bleeding time associated with GPVI depletion in WT animals (Extended Data Fig. 3f). These data essentially excluded the possibility that GPV regulated GPVI-collagen interaction or contributed to collagendependent platelet activation under these experimental conditions. Rather, GPV deficiency over-ruled the hemostatic and thrombotic defects caused by the absence of GPVI and restored thrombus formation in vivo. It has previously been shown that functional defects related to GPVI-ITAM-mediated platelet activation can be attenuated by increased local thrombin generation in different vascular beds 32 , and mouse GPVI does not interact with mouse fibrinogen 33 . Thus, the demonstrated reversal of GPVI inhibition in Gp5 −/− mice suggested that soluble GPV (sGPV) regulated thrombin activity during thrombus formation.

sGPV binds to thrombin and localizes to fibrin
We therefore evaluated the role of GPV in thrombin-mediated fibrin formation on collagen-TF spots in recalcified whole blood under flow in vitro 34 . Time to fibrin formation was shortened and the overall amount of fibrin generated was increased in Gp5 −/− mice (Fig. 2a-c) and, importantly, also in Gp5 dThr mice ( Fig. 2d-f) compared to WT controls. Quantitative imaging of formed thrombi and generated fibrin 35 showed increased thrombus height, based on multilayer and contraction scores, as well as fibrin formation, based on fibrin surface coverage and fibrin score, in the blood of both mutant mouse lines (Fig. 2g). Of note, this ex vivo experimental setup produced results entirely in line with the in vivo findings that Gp5 −/− and Gp5 dThr mice concordantly displayed accelerated thrombus formation.
These data indicate that cleavage of GPV is a critical step in an autoregulatory limitation of fibrin generation. Thrombin binds to de novo generated fibrin via the regulatory thrombin exosites I and II and thereby becomes protected from coagulation inhibitors in the blood 36 . We hypothesized that sGPV directly or indirectly affected thrombinfibrin interactions. We first evaluated the direct interaction of sGPV The GPIb-GPIX complex mediates platelet binding to VWF and is crucial for hemostasis. Mutations in GP1BA, GP1BB or GP9 cause Bernard-Soulier syndrome, a rare bleeding disorder characterized by giant platelets 11,12 . GPV is associated with the GPIb-GPIX complex but is not required for GPIb expression or functional interactions 13 . GPV is an abundant 88-kDa platelet-and megakaryocyte-specific leucinerich repeat transmembrane protein 14 that interacts with collagen 15 and has minor importance for platelet function 16,17 . GPV is proteolytically cleaved by thrombin during thrombus formation 18,19 , but the physiological roles of the shed 69-kDa extracellular fragment in hemostasis and thrombosis has remained elusive.

Thrombus formation is accelerated in Gp5-mutant mice
We studied the role of GPV in thrombus formation by comparing wildtype (WT) and Gp5 −/− mice in FeCl 3 -induced thrombosis of mesenteric arterioles in vivo. In line with previous observations 20 , Gp5 −/− mice displayed faster onset of thrombus formation and shortened occlusion times without increased embolization, indicating a prothrombotic phenotype in the absence of GPV (Fig. 1a,b). Platelet hyper-reactivity to thrombin is the presumed but unproven mechanism for enhanced thrombosis in Gp5 −/− mice and is thought to be related to thrombinmediated cleavage of GPV. To directly study the relevance of thrombin-mediated GPV cleavage, we generated a mouse carrying a point mutation in the thrombin cleavage site of GPV (Gp5 dThr ; Extended Data Fig. 1a). Platelets of these mice showed unaltered surface expression levels of GPV compared to those of WT mice, and GPV was completely resistant to cleavage by thrombin (Extended Data Fig. 1b,c and Extended Data Table 1). By contrast, cleavage of the mutant GPV by endogenous a disintegrin and metalloproteinase (ADAM)17 (ref. 21 ) was not affected (Extended Data Fig. 1b-e), demonstrating the thrombin specificity of the Gp5 dThr mutation. Unexpectedly, Gp5 dThr mice displayed accelerated thrombus formation in the FeCl 3 arteriolar injury model and, in this respect, resembled Gp5 −/− mice (Fig. 1c,d).
We next tested the hypothesis that membrane-bound GPV might act as a regulator of thrombin-mediated PAR activation 25 , supported by the high-affinity binding of thrombin for GPIbα 26,27 . Blockade of the GPIbα-thrombin interaction on mouse platelets with Fab fragments of the anti-GPIbα antibody p0p/B 28,29 (Extended Data Fig. 2a) indeed diminished platelet activation, particularly at low thrombin concentrations (Extended Data Fig. 2c). Although human and mouse platelets are activated by thrombin through different PARs, these antibody inhibition data indicated that mouse platelets are similar to human platelets 25 in requiring GPIbα for thrombin-induced activation at threshold agonist concentrations. Remarkably, the anti-GPIbα antibody completely abolished the enhanced activation of Gp5 −/− platelets relative to WT platelets (Extended Data Fig. 2c), implying that loss of GPV sensitized to GPIbα-dependent thrombin signaling. By contrast, activation of Gp5 dThr platelets at threshold concentrations of thrombin was indistinguishable from that of WT platelets with or without anti-GPIbα p0p/B (Extended Data Fig. 2b,c). Thus, surface GPV regulates platelet responsiveness to thrombin primarily by interference with GPIbα-dependent PAR signaling in mouse platelets (Extended Data Fig. 2d,e). Article https://doi.org/10.1038/s44161-023-00254-6 and thrombin. We stimulated platelets with biotinylated thrombin and showed that sGPV coprecipitated in the thrombin pulldown using streptavidin-coated beads (Fig. 2h), consistent with direct interaction of thrombin with sGPV.
Because GPV release was required to attenuate fibrin formation of recalcified whole blood perfused over collagen-TF spots (Fig. 2a-f), we next quantified the colocalization of GPV with fibrin in this setting. Using confocal microscopy with super-resolution mode, we excluded  in the image analysis platelet-rich areas based on GPIX staining and subsequently quantified the colocalization of GPV with fibrin (Fig. 2i,j and Supplementary Fig. 1). Quantification of GPV intensities showed that GPV accumulated with fibrin in platelet-free areas of thrombi (Fig. 2j). Based on these data, we reasoned that, upon initiation of a hemostatic platelet response, thrombin-mediated cleavage of GPV formed sGPV-thrombin complexes, which limited thrombin diffusion and activity in the forming fibrin clot. To test this concept, we recombinantly expressed the ectodomain of human GPV using a construct that included the thrombin cleavage site (rhGPV) (Fig. 3a). Aggregation of rhGPV at high concentrations prevented us from performing experiments with full dose-response curves. However, thrombin-mediated platelet activation was only marginally inhibited by 290 nM (20 µg ml −1 ) rhGPV at threshold thrombin concentrations (Extended Data Fig. 4a,b), in line with the conclusion that platelet activation by thrombin is primarily regulated by membrane-bound GPV. In sharp contrast, rhGPV at the same concentration impaired fibrin formation in a static polymerization assay ( Fig. 3b and Extended Data Fig. 4c,d) triggered specifically by thrombin, whereas fibrin polymerization induced by another protease, batroxobin, was unaltered in the presence of rhGPV ( Fig. 3b and Extended Data Fig. 4c,d). Importantly, sGPV localized to fibrin polymers independent of the clot-inducing enzyme, indicating direct interactions of GPV with fibrin independent of thrombin-GPV complex formation.
These data also indicate that sGPV in the developing clot directly inhibits thrombin's activity to form fibrin. We therefore studied the interaction of rhGPV with thrombin in the absence of fibrinogen. Incubation of equimolar concentrations of thrombin with rhGPV containing the thrombin cleavage site resulted in time-dependent rhGPV proteolysis, demonstrating that GPV is a thrombin substrate independent of anchoring to the platelet surface (Extended Data Fig. 4e-g). However, formation of sGPV stopped after 5-10 min, when approximately 50% of the substrate was consumed. Measurements of thrombin activity toward a chromogenic substrate were unchanged during the reaction time, excluding instability of the enzyme (Extended Data Fig. 4g). Of note, at the end of the incubation time, sGPV still inhibited thrombin's activity to form fibrin (Extended Data Fig. 4h-j). To distinguish between substrate depletion and product inhibition, we performed the same reaction with a tenfold higher substrate concentration. This reaction yielded essentially the same amount of product, implying that the generated sGPV caused product inhibition and interacted with thrombin independent of fibrin binding. Attempts to chemically cross-link sGPV and thrombin under these conditions were unsuccessful, suggesting that the required amino groups were not in close enough proximity.

rhGPV reduces fibrin formation and protects from thrombosis
In addition, rhGPV impaired fibrin formation in human (Extended Data Fig. 5a-c) and mouse blood (Extended Data Fig. 5d-g) in the collagen-TF-induced thrombus-formation assay under flow, supporting a role for sGPV in limiting thrombin activity toward fibrin. Analysis of the formed fibrin fibrils by confocal microscopy revealed a fine, dense and branched network consisting of thin, clearly distinguishable fibers in control samples, whereas fibers were generally thicker but less frequent and structurally less defined in the presence of rhGPV (Extended Data Fig. 5f), confirming that rhGPV impedes fibrin formation. Of note, a Histagged fusion protein did not reduce fibrin formation, demonstrating specificity of the recombinant GPV.
We next measured thrombin activity in the outflow of the flow chamber and found that less thrombin activity was recovered in rhGPV-treated samples compared to controls (Extended Data Fig. 5h). Conversely, we found more thrombin in the outflow of the chambers perfused with Gp5 −/− and Gp5 dThr versus WT blood (Extended Data Fig. 8a), further supporting the conclusion that sGPV controls thrombin activity specifically in fibrin clots. We next imaged thrombin activity in flow chambers cleared of blood by perfusion with Tyrode's buffer and the thrombin substrate Z-GGR-AMC. We found reduced thrombin activity in clots formed in the presence of rhGPV (Extended Data Fig. 5i,j). In sum, these data support a role for GPV in retaining thrombin in fibrin clots and limiting thrombin's activity in fibrin formation.
We next tested whether rhGPV could modulate thrombus formation in vivo. Indeed, a single intravenous dose of 20 µg rhGPV before thrombosis induction reduced arterial thrombus formation in two different experimental models. In a model of mechanical injury to the abdominal aorta in which blood flow and occlusive thrombus formation were monitored by an ultrasonic flow probe ( Fig. 3c and Extended Data Fig. 5k), 14 of 15 mice did not form stable thrombi after rhGPV administration within the observation period of 30 min, whereas 18 of 18 arteries occluded in the control group. In FeCl 3 -induced mesenteric arteriole injury, time to occlusion was markedly prolonged in mice treated with rhGPV ( Fig. 3d,e).
In addition, rhGPV treatment provided protection from thromboinflammatory neurological damage and improved neurological outcome in the transient middle cerebral artery (MCA) occlusion (tMCAO) model of ischemic stroke ( Fig. 3f-h) in which the concerted action of platelets, the coagulation system and immune cells is known to drive post-ischemic cerebral infarct growth 37 . Of note, infarct volumes of Gp5 −/− and Gp5 dThr mice after tMCAO were comparable to those of WT mice ( Fig. 3f-g), suggesting that thrombin activity in WT mice is already above threshold values needed to fully promote infarct progression under these experimental conditions. Importantly, no large intracranial the experiment; every dot above it did not show any occlusion/fibrin formation within the observation period. g, Subtraction heatmap of parameters of thrombus (increased multilayer and contraction score) and fibrin formation in mutant mice compared to WT controls. Colors represent unchanged (black), decreased (green) or increased (red) parameters. h, Pulldown of sGPV from WT platelets using streptavidin beads after stimulation with biotinylated thrombin in the presence or absence of GM6001 (matrix metalloproteinase inhibitor). Pulldown of sGPV was not observed from Gp5 dThr platelets or in the presence of hirudin, which prevents GPV cleavage by thrombin. Eluates were analyzed by western blotting using GPV-specific antibodies and streptavidin-horseradish peroxidase (HRP). Representative results of five independent experiments are shown. M, marker; mGPV, murine GPV; SN, supernatant; arrow indicates pulldown of sGPV. i,j, Recalcified blood was perfused over collagen-TF microspots. Samples were stained for platelets (anti-GPIX antibody, yellow), fibrin(ogen) (cyan) and GPV (magenta) and analyzed with a Zeiss Airyscan microscope. i, Representative WT image and magnified view. Scale bar, 4 µm. j, Quantification of GPV intensities inside fibrin-rich but nonplatelet area. Background GPV intensity in Gp5 −/− images is displayed as a dashed line. Two-way ANOVA followed by Tukey's multiple-comparison test, n = 14 (WT), n = 12 (Gp5 dThr
Fibrin score (0-3) Color key   hemorrhages were observed in Gp5 −/− or Gp5 dThr and rhGPV-treated WT mice (Fig. 3g). Of note, MCA vessel diameters were similar in Gp5 −/− and WT mice (Extended Data Fig. 6). In addition, hemostatic function evaluated by the tail-bleeding time assay was also comparable between mice injected with rhGPV and vehicle-treated controls (Fig. 3i), indicating that this pathway might be targeted safely. Together, these data showed that sGPV specifically limited fibrin formation and pathological intravascular thrombus growth without impairing the initial platelet activation required for hemostasis.

Blocking GPV cleavage offsets defects in hemostatic platelet function
To further study thrombin interaction with GPV, we generated a panel of anti-GPV monoclonal antibodies (termed DOM monoclonal antibodies; Extended Data Fig. 7) and first evaluated their ability to inhibit thrombin-mediated GPV cleavage (Fig. 4a). Cleavage of substrates by thrombin involves binding and allosteric regulation by thrombin exosites I and II that flank the active site 38 . Blockage of exosite I with the thrombin-binding aptamer HD1 (ref. 39 ) was more efficient than blocking exosite II with HD22 (ref. 40 ), whereas a non-blocking aptamer HD23 had no effect on thrombin-mediated release of GPV from platelets (Extended Data Fig. 7g). Thus, thrombin interaction with fibrin and GPV occurred through overlapping sites 36 . With this screening assay, we identified the monoclonal antibody DOM/B that markedly reduced thrombin-mediated GPV cleavage and synergized with thrombin exosite-directed aptamers ( Fig. 4a and Extended Data Fig. 7g), whereas the monoclonal antibody DOM3 was non-inhibitory (Fig. 4a,f and Extended Data Fig. 7b). In PRP, we tested inhibitory activities of DOM/B in a thrombin-induced clotting assay, in which fibrin is formed independent of platelet activation. Whereas thrombin exosite II blockade with HD22 prolonged clotting times, clotting was unaffected by DOM/B treatment and was also indistinguishable between WT, Gp5 −/− and Gp5 dThr samples (Extended Data Fig. 7h). Thus, thrombin regulation by GPV specifically occurs under conditions of platelet activation under flow.
In addition, DOM/B had no effect on thrombin-induced platelet activation (Extended Data Fig. 7d-f), indicating that this monoclonal antibody did not sterically hinder the interaction of GPV with the GPIb-GPIX complex involved in GPIbα-thrombin-PAR platelet signaling. Remarkably, however, DOM/B significantly shortened time to fibrin formation and increased the amount of generated fibrin under flow conditions (Fig. 4b,c) as well as thrombin activity in the outflow of the flow chamber (Extended Data Fig. 8a), thereby reproducing the phenotypes seen with Gp5 −/− and Gp5 dThr mice. In line with reduced proteolytic release of GPV in the presence of DOM/B, we found less GPV colocalizing with fibrin than in controls (Extended Data Fig. 8b). The panel of anti-GPV monoclonal antibodies was also evaluated for interference with collagen-dependent platelet activation. Whereas DOM/B and DOM3 were non-inhibitory (Extended Data Fig. 7b,c), inhibition of platelet activation in this assay by DOM/C indicated that this monoclonal antibody was directed against the collagen-binding site of GPV (Extended Data Fig. 7a). DOM/C did not interfere with thrombinmediated GPV cleavage (Fig. 4a) and did not enhance fibrin formation under flow, consistent with the crucial and specific role of sGPV release in this context (Fig. 4d,e). These data suggested that the collagenbinding activity of (s)GPV is functionally not required for its ability to modulate fibrin formation under flow in vitro and in vivo.
We next evaluated the effect of blocking GPV-thrombin interaction with DOM/B on fibrin and thrombus formation in vivo. Of note, injection of DOM/B did not cause platelet depletion and the monoclonal antibody remained detectable on the surface of circulating platelets for up to 6 d (Extended Data Fig. 8c,d). In line with the observed increased fibrin formation under flow in vitro, DOM/B treatment caused accelerated thrombus formation in FeCl 3 -injured mesenteric arterioles in vivo, whereas neither blockade of the collagen-binding site on GPV with DOM/C nor the non-inhibitory DOM3 affected thrombus formation (Fig. 4f,g). These data showed that the release of sGPV acts as a safety valve to limit thrombus growth after initial platelet activation required for hemostasis. We have previously shown that the absence of the two major collagen receptors GPVI and integrin α 2 β 1 causes severe bleeding in mice 41 . In line with the reversal of bleeding defects caused by GPVI deficiency in Gp5 −/− mice, DOM/B treatment restored hemostasis and thrombus formation in the complete absence of the two major platelet collagen receptors GPVI and α 2 β 1 (Extended Data Fig. 8e-g, summarized in Extended Data Fig. 8h).
These results suggest that interference with GPV cleavage can be used to enhance fibrin formation in the context of defective hemostasis caused by diverse mechanisms. We therefore investigated whether GPV-cleavage blockade with DOM/B not only restored hemostasis in the case of defective (hem)ITAM signaling but also other genetic and pharmacological impairments of platelet function. Lack of the small GTPase RhoA causes macrothrombocytopenia and defective platelet activation, resulting in a bleeding defect 42 . Similarly, neurobeachin-like 2 (NBEAL2) deficiency leads to macrothrombocytopenia and lack of α-granules, resulting in severely impaired hemostasis 43 . Interestingly, DOM/B treatment improved and restored hemostasis in mice with platelet RhoA deficiency or lacking NBEAL2 (Fig. 4h,i).
Thrombocytopenia is a major clinical challenge occurring frequently in the context of a variety of pathologies or medical treatments that is associated with increased bleeding and often with the need for immediate therapeutic intervention 44,45 . To test a possible benefit of a GPV-cleavage blockade in this setting, we induced severe thrombocytopenia by reducing platelet counts to 5-10% of normal by injecting a platelet-depleting antibody 46,47 . While a resulting severe bleeding defect was observed in all nine platelet-depleted control mice, this was significantly attenuated by DOM/B treatment, and, remarkably, nine of 11 DOM/B-treated thrombocytopenic mice managed to stop bleeding within the observation period (Fig. 4j). The current clinical standard of care to reduce the risk of heart attack and ischemic stroke is the pharmacological inhibition of platelet function by purinergic P2Y 12 ADP receptor blockers alone or in combination with acetyl salicylic acid. As seen in humans, mice treated with the P2Y 12 blocker clopidogrel exhibited increased bleeding that was reversed by treatment with DOM/B, blocking thrombin-dependent GPV release from platelets (Fig. 4k). Thus, specific targeting of GPV with DOM/B prevented prolongation of the bleeding time caused by thrombocytopenia, genetic defects and anti-platelet therapy, indicating clinical potential of anti-GPV treatment to restore hemostasis by improving thrombin-dependent fibrin formation.

Blocking GPV cleavage increases fibrin formation in human blood
We therefore evaluated the relevance of this concept for human platelet function and screened a panel of newly generated anti-human GPV (hGPV) monoclonal antibodies (termed LUM monoclonal antibodies) for their ability to interfere with thrombin-mediated cleavage of GPV. Recapitulating the inhibitory properties of DOM/B in the mouse system, LUM/B prevented thrombin-mediated cleavage of GPV on human platelets, whereas other anti-hGPV monoclonal antibodies (LUM1-LUM5) were non-inhibitory ( Fig. 5a and Extended Data Fig. 9c). LUM/B per se did not activate human platelets (Extended Data Fig. 9a,b) or influence thrombin clotting (Extended Data Fig. 9e) or thrombin-induced platelet activation as shown by unaltered integrin α IIb β 3 activation or P-selectin exposure in the presence of the antibody (Fig. 5b and Extended Data Fig. 9d) significantly accelerated fibrin formation in recalcified whole blood (Fig. 5c-e), whereas the non-inhibitory LUM3 antibody was without effect (Extended Data Fig. 9f,g). Thus, spatio-temporal control of fibrin formation on thrombogenic surfaces by GPV is a species-conserved mechanism to restrict thrombosis while preserving hemostasis.

Discussion
We here delineate the function of platelet GPV that is proteolytically released by thrombin in the context of platelet activation at sites of vascular injury. Genetic blockade of thrombin-mediated shedding of GPV uncovered the crucial role of sGPV as a regulator of fibrin formation and thrombus growth. The cleavage of membrane-bound GPV occurs after platelet adhesion and initiation of thrombin generation and is thus temporally separated from the initial hemostatic response. We show that generated sGPV then interacts with generated fibrin and dampens thrombin activity toward fibrinogen. This spatially restricted action of sGPV after release from the platelet surface specifically limits thrombus growth in vitro and in vivo (summarized in Fig. 6). As demonstrated by pharmacological application of rhGPV, sGPV provides protection from thrombo-inflammatory neurological damage in an experimental model of ischemic stroke without causing hemostatic impairments. Conversely, specific blockade of thrombin-mediated GPV shedding can enhance local fibrin formation in a variety of contexts associated with severe defects in platelet function. This unique spatio-temporal control of thrombin activity by GPV can thus be harnessed to promote hemostasis. Prevention of thrombosis while preserving hemostasis has been a central goal of anti-thrombotic drug development. Despite the broader application and safety of target-selective oral anticoagulants, preventing bleeding complications remains an unmet clinical need. Although a recent study provides proof of principle that platelet-mediated thrombin generation can rescue bleeding defects due to increased fibrinolysis 47 , there is an unmet clinical need for more general and specific spatio-temporal control of fibrin formation. New hemostatic agents are approved or in development to bypass genetic or acquired deficiencies in the coagulation cascade 48 , but platelet transfusion remains the only therapeutic option to acutely restore defective platelet function or severe forms of thrombocytopenia to secure hemostasis. Akin to the strategies that 'inhibit the inhibitors' of coagulation, we here propose a therapeutic strategy of tailored activation of hemostatic fibrin plug formation in the spatio-temporal context of platelet deposition at sites of vessel wall injury. By increasing local thrombin bioavailability without compromising scavenging of thrombin by endothelial cell-expressed thrombomodulin, this approach has little risk to interfere with physiological anticoagulation in the body and vascular-protective and anti-inflammatory signaling of the protein C-PAR1 pathway 10 .

Animals
Mice were maintained under specific pathogen-free conditions (constant temperature of 20-24 °C and 45-65% humidity with a 12-h lightdark cycle, ad libitum water and food access), and experiments were performed in accordance with German law and the governmental bodies and with approval from the District of Lower Franconia. Gp5 −/− (ref. 12 ) and Itga2 −/− (ref. 49 ) mice were kindly provided by F. Lanza (Inserm-Université de Strasbourg, Strasbourg, France) and B. Eckes (Department of Dermatology, University of Cologne, Cologne, Germany), respectively. Gp5 dThr mice, which carry a point mutation in the sequence corresponding to the thrombin cleavage site of GPV, were generated and are described in Extended Data Fig. 1a. Gp5 dThr mice were intercrossed with Flip-positive mice to delete the Neo cassette and backcrossed to the C57Bl/6J background. Rhoa fl/fl (ref. 50 ) mice were kindly provided by C. Brakebusch (University of Copenhagen, Copenhagen, Denmark). To generate MK-or platelet-specific knockout mice, floxed mice were intercrossed with mice carrying the sequence for Cre recombinase under control of the Pf4 (platelet factor) promoter. Nbeal2 −/− (ref. 43 ) mice were described previously. All mice were kept on a C57Bl/6J background, and all animal experiments and analysis of the corresponding data were performed blinded.
Female Wistar rats (strain RjHan:WI, starting at 6 weeks of age) were immunized to generate monoclonal anti-GPV antibodies.

Human blood samples
For this study, blood samples were obtained from healthy volunteers, free from anticoagulant or anti-platelet therapy for at least 4 weeks, following written informed consent in accordance with the Declaration of Helsinki and after approval by the Institutional Review Boards of the University of Würzburg (votes 167/17 and 295/20).
New antibodies were generated by hybridoma technology following immunization of Gp5 −/− mice or Wistar rats with recombinant hGPV protein or GPV immunoprecipitated from mouse or human platelet lysates.

Expression and purification of recombinant hGPV
The gene fragment encoding the GP64 signal peptide (MVSAIV-LYVLLAAAAHSAFA), the human GPV extracellular domain (amino acids 17-518) and a decahistidine tag was amplified, inserted into the pFastBac dual vector and transformed into the DH10Bac Escherichia coli strain (Thermo Fisher Scientific). The resulting bacmid DNA was prepared and then transfected into Sf9 insect cells (Thermo Fisher Scientific, 11496015) using CellFectin II reagent (Thermo Fisher). The high-titer P2 baculovirus stock was prepared from scaled-up Sf9 cells in Sf-900 II serum-free medium (Thermo Fisher Scientific, 10902088) following the instructions for the Bac-to-Bac Baculovirus Expression System (Thermo Fisher Scientific) and used to induce hGPV expression in Sf9 cells (2 × 10 6 cells per ml, MOI of 2) for 72 h. hGPV was purified from insect cell medium by Ni affinity chromatography using a GE Healthcare Ni Sepharose excel column (elution buffer: 20 mM sodium phosphate, 500 mM NaCl, 500 mM imidazole, pH 7.4), followed by size-exclusion chromatography using a GE Healthcare HiLoad 16/600 Superdex 200pg column (elution buffer: PBS containing 0.1% Tween-20, pH 7.4). The purified protein was stored at −80 °C in PBS containing 0.1% Tween-20 and 20% glycerol. Notably, rhGPV precipitated at concentrations above 0.5 mg ml −1 , limiting its biophysical characterization.

Antibody treatment
For antibody treatment experiments, mice were separated into antibody and control treatment groups in a randomized manner using https://www.random.org/lists/. Next, 100 µg JAQ1 IgG was injected intraperitoneally at day 7 and day 5 before the experiment, resulting in a GPVI-knockout-like phenotype 31 . All other antibodies (each 100 µg) were injected i.v. or i.p. directly before the experiment.

Platelet depletion
Thrombocytopenia was induced by intravenous injection of the rat anti-GPIbα IgG antibody R300 (Emfret Analytics, 0.14-0.18 µg per g body weight). This low dose of platelet-depletion antibody reduced the platelet count to 5-10% of the initial platelet count 46 . Peripheral platelet count was determined by flow cytometry 16 h after platelet depletion (before the tail-bleeding time experiment).

Treatment with clopidogrel or rhGPV
Mice were fed orally with 3 mg per kg clopidogrel 48 h and 24 h before the experiment. A dose titration for rhGPV was performed using 10-80 µg rhGPV per mouse in an arterial thrombosis model. The dose of 10 µg rhGPV resulted in variable results, while no difference in the anti-thrombotic effect of rhGPV was noted for 20, 40 or 80 µg rhGPV per mouse. Consequently, 20 µg (as the lowest rhGPV concentration that yielded reproducible results) was chosen for all in vivo experiments, and 20 µg ml −1 was chosen for in vitro assays. Mice were injected intravenously with 20 µg rhGPV 5 min before the experiment.

Aggregation assay
Washed platelets (160 µl with 1.5 × 10 5 platelets per µl) and PRP (only used for ADP stimulation) were prepared as described. For aggregometry, washed platelets were analyzed in the absence (thrombin) or presence (all other agonists) of 70 µg ml −1 human fibrinogen. Antibodies (10 µg ml −1 ) were pre-incubated for 5 min at 37 °C before the experiment. Light transmission was recorded on a four-channel aggregometer (Fibrintimer, APACT) for 10 or 20 min (in the presence of LEN/B) and expressed in arbitrary units, with buffer representing 100% light transmission. Platelet aggregation was induced by addition of the indicated agonists.

Flow cytometry
To determine glycoprotein surface expression levels, whole blood was diluted 1:20 with Ca 2+ -free Tyrode's buffer or PBS and stained with saturating amounts of fluorophore-conjugated antibodies for 15 min at room temperature in the dark.
Washed platelets adjusted to 50,000 per µl in Tyrode's buffer with Ca 2+ were stimulated with the indicated agonists and incubated with saturating amounts of fluorophore-conjugated antibodies to determine platelet activation or thrombin-mediated cleavage of GPV. Human thrombin (Sigma, 10602400001) was used to stimulate platelets.
All samples were analyzed directly after addition of 500 µl PBS on a FACSCalibur (BD Biosciences) using CellQuest Pro (version 6.0) software. Data were analyzed using FlowJo (version 10.7 and version 10.8.1). An exemplified gating strategy based on FSC-SSC characteristics is provided in Supplementary Fig. 2.

Thrombin-and N-ethylmaleimide-induced cleavage of GPV
Washed platelets were adjusted to a concentration of 1 × 10 6 platelets per µl in Tyrode's buffer without Ca 2+ and diluted 1:1 with Tyrode's buffer without Ca 2+ (for resting and thrombin-stimulated samples (human thrombin (Sigma, 10602400001))) and with Tyrode's buffer with Ca 2+ (for N-ethylmaleimide (2 mM final concentration (f.c.))-incubated samples). Stimulation with 867 pM thrombin (for human thrombin, 0.1 U ml −1 is equivalent to 867 pM thrombin) for 30 min at 37 °C was performed in the presence of 40 µg ml −1 integrilin and 5 µM EGTA to prevent platelet aggregation. Afterwards, the platelet suspension was diluted, incubated with saturating amounts of FITC-conjugated platelet surface antibodies and directly analyzed on a FACSCalibur. The residual platelet suspension was pelleted, and the supernatant was analyzed in an ELISA for GPV.

GPV enzyme-linked immunosorbent assay
Ninety-six-well plates (Hartenstein, F-form) were coated with 50 µl per well DOM/C antibody (30 µg ml −1 ) in carbonate buffer overnight at 4 °C, blocked with 5% non-fat dried milk in PBS for 2 h at 37 °C and washed. Samples were applied to plates, incubated for 1 h at 37 °C and washed. Plates were incubated with HRP-labeled DOM/B antibody for 1 h and washed again three times, and samples were developed using TMB substrate. The reaction was stopped by addition of 0.5 M H 2 SO 4 . Optical density was measured on a Multiskan EX device (Thermo Electron). Absorbance was read at 450 nm; the 620-nm filter served as the reference wavelength. Plasma samples from Gp5 −/− mice served as the negative control; supernatant after platelet thrombin stimulation served as the positive control.

Clot retraction
Clot-retraction studies were performed at 37 °C in an aggregometer tube containing diluted PRP (3 × 10 5 platelets per µl), thrombin (34.68 nM) and CaCl 2 (20 mM). Clot retraction was recorded with a camera over a time span of 2 h after activation.

Thrombin time
To determine thrombin time, citrated PRP was diluted 1:1 in PBS and stimulated with 17.5 nM f.c. bovine thrombin (equivalent to 2 U ml −1 ). Thrombin time was analyzed with a four-channel mechanical ball coagulometer (Merlin Medical).

Thrombin generation
Thrombin generation was quantified in recalcified citrate anticoagulated PRP with platelet count adjusted to 1.5 × 10 5 platelets per µl. Platelets were resuspended in pooled plasma preparations from two to four mice with the same genotype. After platelet stimulation with the indicated agonists (15 min at 37 °C), samples in duplicates (four volumes) were transferred to a polystyrene Immulon 2HB 96-well plate. The wells either contained one volume thrombin calibrator or TF (1 pM f.c.). Coagulation was initiated by adding one volume of fluorescent thrombin substrate (2.5 mM Z-GGR-AMC). Thrombin generation was measured and analyzed using Thrombinoscope software (version 5.0.0.742, Thrombinoscope) 63,64 .

Static fibrin polymerization
Unlabeled fibrinogen (1.35 mg ml −1 f.c.) and AF A488-labeled fibrinogen (45 µg ml −1 f.c.) were mixed (30:1) in the absence or presence of rhGPV (20 µg ml −1 , stained with LUM/B AF647). Fibrin polymerization was initiated by addition of 867 pM thrombin or 1 U ml −1 batroxobin (Loxo) in the presence of 5 mM CaCl 2 . The mixture was immediately transferred to an uncoated eight-well 15µ-slide (ibidi), which was placed in a dark humidity chamber for 2 h at room temperature to allow fibrin polymerization. Images were obtained using a Leica SP8 inverted microscope with a ×63 oil-immersion lens. Optical z stacks (8 µm, step size of 0.1, Nyquist conform) were deconvolved (Huygens Essential software, version 21.04) and are shown as a maximum projection (ImageJ software). Fibrin fibers per visual field and surface coverage of hGPV staining were quantified using Fiji.

Coagulation flow chamber
Glass coverslips were coated with collagen type I (10 µl, 50 µg ml −1 ) and TF (10 µl, 100 pM or 10 pM for experiments with human blood or mouse blood, respectively) and blocked with 1% BSA-PBS. Citrated whole blood was recalcified by co-infusion with 6.3 mM CaCl 2 (f.c.) and 3.2 mM MgCl 2 (f.c.) and perfused over the collagen-TF spots for up to 6 min at a shear rate of 1,000 s −1 (ref. 65 ). Before each experiment, blood samples were prelabeled with AF488-conjugated fibrinogen, an anti-GPIX derivative (mouse) or an anti-GPIbβ derivative (human) to stain platelets. For human samples, control and antibody-treated samples of the same donor were always run in parallel.
Time series experiments. Before each experiment, blood samples were pre-incubated with AF488-conjugated fibrinogen, AF647-conjugated anti-GPIX derivative (mouse) or AF647-conjugated anti-GPIbβ derivative (human) platelets. For time series experiments, fluorescence microscopic images were captured at 30-s intervals to evaluate kinetics for up to 6 min (Leica DMI 6000 B, ×63 objective, Leica Application Suite (LAS) X software (version 1.9.013747)). Recorded images were further processed with the background-subtraction method Instant Computational Clearing) to remove out-of-focus blur on a Leica THUNDER microscope (LAS X software version 3.7). Next, the exported images were analyzed for surface area coverage of fibrin formation with self-written Python scripts 66 . In detail, in the first step, an entropy filter with a disk size of five pixels was applied, followed by a median Article https://doi.org/10.1038/s44161-023-00254-6 filter (disk size, ten pixels) and Otsu thresholding. To compensate for a nearly empty field of view during the first images in the time series, we introduced a 'scaling factor', sf, with which we multiplied the found Otsu threshold value and thus increased the threshold slightly for the first images (usually the first 60-150 s, with sf = 1.03-1.05). The thresholded area (as a fraction of the whole field of view) represents the area covered by fibrin.
For heatmap representation, mean values were univariate scaled from −4 to 4. Gene effect heatmaps were constructed by subtracting scaled average values of the control strain from those of the mutant strain. For details, see ref. 35 .
Three-dimensional confocal microscopy. Samples were stained as described for Time series experiments. To acquire z stacks of complete thrombi, samples were fixed with 4% paraformaldehyde (PFA)-PBS, mounted with ProLong Glass (Thermo Scientific) and further analyzed by confocal microscopy (Leica SP8 inverted microscope). z stacks of thrombi were acquired with an SP8 confocal microscope (Leica) in HyVolution mode (×63; z step, 0.1 µm). Images were deconvolved using Huygens Professional software (version 21.04) with a signal-to-noise ratio between 2 and 10 (identical for regions of interest in the same animal), an automatic background subtraction using the 'in/near object' option with a search radius of 2 µm and a maximum of 40 iterations. The deconvolved dataset was exported in Imaris file format and visualized with Fiji 67 (https://www.biovoxxel.de/development/).
To analyze GPV-fibrin localization outside the thrombus, blood samples were pre-incubated with AF488-conjugated fibrinogen, AF546-conjugated anti-GPV derivative (as direct labeling of rhGPV with a fluorophore abolished its biological activity) and AF405-conjugated anti-GPIX derivative. Single images from the bottom of a thrombus were acquired using a Zeiss LSM 980 Airyscan microscope (×63 objective) in super-resolution mode using the smart setup. Images were deconvolved with Zeiss ZEN software (version 3.2) and analyzed with Fiji. First, masks from the fibrin(ogen) and platelet (GPIX) channels were generated using Li thresholding. Masks were then used as 'positive' (pixel intensity of 1 inside the structure and 0 outside the structure) and 'negative' (pixel intensity of 0 inside the structure and 1 outside the structure) imprints and applied to the GPV channel. To analyze GPV intensities outside the thrombus but inside fibrin fibers, the positive fibrin mask was multiplied with the negative thrombus mask.
In detail, after Li thresholding, the intensity values in the binarized images were changed to 0 and 1 by dividing through 255. Next, this procedure for preparation of the positive and negative binary masks was repeated for the GPIX channel. Afterward, the prepared masks based on single-channel thresholds were combined by multiplication (leading to a value of 1 if positive in both masks). This resulted in the following regions: (1) inside fibrin but non-platelet area and (2) outside fibrin and non-platelet area → background count rate.
Finally, the intensity of GPV was determined in the area of fibrin by multiplication of the obtained masks with the GPV channel. Values outside the mask were set to 0, and pixels inside the mask have an intensity of 1, thus the original intensity of GPV was preserved. This calculation was performed for all masks, and surface coverage as well as raw integrated density were determined. Next, the average intensity per pixel was calculated inside the covered area, and the raw integrated density was divided by the number of pixels to obtain the average intensity per non-zero pixel.
For each step, a Fiji macro was recorded. All Fiji macros and Python scripts used for fluorescence image analysis can be downloaded from https://github.com/HeinzeLab/GPV-flowchamber.

Thrombin activity
The coagulation flow assay was performed as described above without staining for platelets and fibrin(ogen). The outflow was collected in 10 mM EDTA and 1.5 µM HD1. Thrombin activity was measured immediately using the fluorogenic thrombin substrate Pefafluor TH (Pentapharm) at 460 nm.
Thrombin activity in the formed thrombi was determined using the fluorogenic thrombin substrate Z-GGR-AMC.

Western blot after pulldown
Washed platelets were adjusted to 1 × 10 6 platelets per µl and either left unstimulated or were stimulated with biotinylated thrombin (433 pM) for 15 min at 37 °C. When indicated, hirudin (0.1 U ml −1 ) or GM6001 (100 µM f.c.) was added before platelet stimulation. Platelets were pelleted, and the supernatant was incubated with magnetic streptavidin beads to pull down biotinylated thrombin. After incubation, beads were collected and washed. The eluate was used for western blot analysis, and the samples were detected with an anti-GPV antibody (R&D).

GPV-thrombin interaction in solution
We performed cross-linking experiments with BS3 or DTSSP following incubation of thrombin with varying concentrations of rhGPV. However, conditions with efficient cross-linking could not be established. Thrombin interaction with rhGPV was furthermore analyzed in a solution (HEPES buffer), and generation of sGPV was followed by western blotting and detection with an anti-hGPV antibody (Santa Cruz, sc271662). A serial dilution of cleaved sGPV was used to quantify the reaction product. The stability of thrombin was analyzed under the same conditions by determining thrombin activity with the chromogenic substrate S2238 (Chromogenix).

Tail-bleeding time
Mice were anesthetized by intraperitoneal injection of triple anesthesia (Dormitor, 0.5 µg per g; midazolam, 5 µg per g; and fentanyl, 0.05 µg per g body weight), and a 1-mm segment of the tail tip was removed using a scalpel. Tail bleeding was monitored by gently absorbing blood on filter paper at 20-s intervals without directly contacting the wound site. When no blood was observed on the paper, bleeding was determined to have ceased. The experiment was manually stopped after 20 min by cauterization.

Light sheet fluorescence microscopy
Sample preparation. The vasculature was stained by intravenous injection of AF647-conjugated anti-CD105 (clone MJ7/19, purified in house, 0.4 µg per g body weight) and AF647-conjugated anti-CD31 (BioLegend, clone 390, 0.4 µg per g body weight) antibodies. After in vivo labeling (30 min), mice were anesthetized by intraperitoneal injection of medetomidine (0.5 µg per g body weight), midazolam (5 µg per g body weight) and fentanyl (0.05 µg per g body weight) and transcardially perfused with ice-cold PBS to wash out the blood and ice-cold 4% PFA (P6148, Sigma-Aldrich, pH 7.2). Brains were removed, washed with PBS and dehydrated in methanol solutions of increasing concentrations (50%, 70%, 95%, 100%) to fix the tissue. The methanol was replaced stepwise by a clearing solution consisting of one part benzyl alcohol to two parts benzyl benzoate (BABB; 305197 and B6630, Sigma-Aldrich). After incubation in the clearing solution for at least 2 h at room temperature, tissue specimens became optically transparent and were used for light sheet fluorescence microscopy (LSFM) imaging on the following day 68 .

Segmentation of brain LSFM images
Images acquired by LSFM were saved in TIFF format and converted to the Imaris file format (Imaris 9.9, Bitplane, Oxford) for further processing and segmentation. Using the built-in image-processing tools, first the background was subtracted from both channels, and secondly a 3 × 3 × 3-voxel median filter was applied to the vessel channel (AF647 fluorescence). Next, the median-filtered vessel channel was segmented using the surface tool. Here, four-voxel smoothing (10.4 µm) and local contrast intensity thresholding (10 µm in diameter) were applied. The intensity threshold was adjusted manually to ~50% of the automatically proposed value. Finally, objects smaller than 1,000 voxels were removed.

Determination of vessel diameter
To determine the diameter of the vessel at selected regions of interest, the generated vessel surface was masked onto the vessel fluorescence such that the intensity outside the surface was zero, while, inside the surface, the original, median-filtered intensity values were present.
To estimate the diameter, the 'measurement points' option in Imaris was used, which was placed directly on the border of the selected vessel regions of interest. Correct placing of the measurement points was ensured by 3D inspection of the images.

Transient middle cerebral artery occlusion
Focal cerebral ischemia was induced by tMCAO as previously described 70 . Briefly, a silicon-coated thread was advanced through the carotid artery up to the origin of the MCA, causing an MCA infarction. After an occlusion time of 60 min, the filament was removed, allowing reperfusion of the MCA territory. The extent of edema corrected brain infarction was quantitatively assessed 24 h after reperfusion on 2,3,5-triphenyltetrazolium chloride-stained consecutive brain sections. Neurological function was analyzed by calculating a neuroscore (score, 0-10) based on the direct sum of the Grip test (score, 0-5) and the inverted Bederson score (score, 0-5).

Posterior communicating artery scores
Posterior communicating artery (PcomA) scores were determined in brains from mice that were perfused with PBS followed by 3 ml black ink diluted in 4% PFA (1:5, vol/vol).

Mechanical injury of the abdominal aorta
To open the abdominal cavity of anesthetized mice (8-20 weeks old), a longitudinal midline incision was performed, and the abdominal aorta was exposed. A Doppler ultrasonic flow probe (0.5PSB699, Transonic Systems) was placed around the vessel, and thrombus formation was induced by a single firm compression (20 s) with a forceps upstream of the flow probe. Blood flow was monitored over 30 min or until complete occlusion occurred (blood flow stopped for >5 min). The abdominal aorta was excised and embedded in Tissue-Tek. Sections (5 µm) were fixed and stained according to the Carstairs method to distinguish platelets and fibrin 71 . Images of Carstairs stained sections were acquired on a Leica DMI400B using LAS software (version 2.6.0.R1).

Data analysis
The presented results are mean ± s.d. from three independent experiments per group, and lines represent mean values, if not stated otherwise. Normal distribution was tested using the Shapiro-Wilk normality test. If passed, P values were calculated using the two-tailed unpaired t-test (two groups); if the values were not normally distributed, differences between two groups were analyzed using the Mann-Whitney two-tailed test. For more than two groups, one-way ANOVA (Kruskal-Wallis test) followed by Dunn's test for multiple comparisons was performed using GraphPad Prism software (versions 7.03 and 7.05.). Two-tailed paired t-test (two groups, normally distributed), Wilcoxon matched-pairs signed-rank test (two groups, not normally distributed) and Friedman test followed by Dunn's test for multiple comparisons (more than two groups, not normally distributed) were used for paired comparisons. For statistical analysis of non-occluded versus occluded vessels, Fisher's exact t-test was used. P values <0.05 were considered statistically significant.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
All data supporting the analyses presented in this study are provided in the text and its associated files.