H. Ni, Canadian Blood Services and Department of Laboratory Medicine and Pathobiology, St Michael's Hospital, University of Toronto, 30 Bond Street, Room 2-006, Bond Wing, Toronto, Ontario, Canada M5B1W8. Tel.: +1 416 864 6060 × 6758; fax: +1 416 864 3060; e-mail: email@example.com
Summary. The role of vitronectin (Vn) in thrombosis is currently controversial; both inhibitory and supportive roles have been reported. To monitor directly the function of Vn in thrombotic events at the site of vascular injury, we studied Vn-deficient (Vn–/–) and wild-type (WT) control mice with two real-time intravital microscopy thrombosis models. In the mesenteric arteriole model, vessel injury was induced by ferric chloride. We observed unstable thrombi and a significantly greater number of emboli in Vn–/– mice. Vessel occlusion was also delayed and frequent vessel re-opening occurred. In the cremaster muscle arteriole model, vessel injury was induced by a nitrogen dye laser. We observed significantly fewer platelets, lower fibrin content, and unstable fibrin within the thrombi of Vn–/– mice. To define further the role of Vn in thrombus growth, we studied platelet aggregation in vitro. Consistent with our in vivo data, the second wave of thrombin-induced aggregation of gel-filtered platelets was abolished at a low concentration of thrombin in Vn–/– platelets. Interestingly, adenosine diphosphate (ADP)-induced platelet aggregation was significantly increased in Vn–/– platelet-rich plasma (PRP) and this effect was attenuated by adding purified plasma Vn. We also observed increased platelet aggregation induced by shear stress in Vn–/– whole blood. These data demonstrate that Vn is a thrombus stabilizer. However, in contrast to released platelet granule Vn which enhances platelet aggregation, plasma Vn inhibits platelet aggregation.
Platelet adhesion, aggregation, and formation of the polymerized fibrin matrix at the site of vascular injury are key events required to arrest bleeding. However, the same hemostatic processes also contribute to the generation of inopportune thrombi within atherosclerotic arteries, which is the leading cause of morbidity and mortality worldwide. It has been demonstrated that von Willebrand factor (VWF) and fibrinogen are essential for platelet adhesion and aggregation . Unexpectedly, we recently demonstrated that thrombus formation still occurs in mice lacking both VWF and fibrinogen , suggesting that other proteins such as fibronectin , thrombospondin , and vitronectin (Vn) [5–7] may still be able to mediate platelet adhesion and aggregation. Although Vn has the potential for both binding β3-integrin and stabilizing the fibrin matrix , its role in thrombosis and hemostasis remains nonetheless controversial. No alteration of bleeding time has been found in Vn–/– mice .
Vn is a major plasma protein (200–500 µg mL−1) as well as an abundant component of platelet α-granules and the extracellular matrix . Vn is a 75-kDa glycoprotein containing a single Arg-Gly-Asp (RGD) site with the potential for binding to integrins [5,10,11]. It also has a variety of binding sites for different proteins relevant to thrombosis such as plasminogen activator inhibitor-1 (PAI-1) [11,12], fibrin , heparin, collagen, and urokinase-type plasminogen actvator receptor , allowing for a possible role in regulating platelet adhesion, aggregation and fibrinolysis. However, most Vn in plasma is in an ‘inactive’ form, leaving most of its binding sites inaccessible to its ligands . Only 2% of plasma Vn is in an ‘active’ form, as defined by its ability to bind heparin . Conversely, in platelet α-granules and the extracellular matrix, Vn is present in an activated, multimeric form [15,16]. It is not clear whether the 2% active form of Vn in the blood is able to bind β3-integrins and compete with fibrinogen for platelet aggregation (i.e. attenuate platelet aggregation).
Previous studies using Vn–/– mice suggested that Vn inhibits the thrombotic response in vivo. Significantly faster vessel occlusion in FeCl3-injured carotid arteries and shorter thrombin time were found in Vn–/– mice compared with wild-type (WT) mice. A trend towards enhanced platelet aggregation was also observed in Vn–/– mice and significant inhibition of platelet aggregation by Vn was achieved by adding exogenous Vn to washed Vn–/– platelets . However, it has also been reported that Vn promoted thrombosis when carotid arteries were injured photochemically  or by FeCl3 with a minor modification of anesthesia ; the supportive role of Vn in thrombosis was explained by its role in maintaining PAI-1 function and thrombus stabilization. Recently, new evidence that Vn supports platelet adhesion and aggregation was demonstrated by use of a monoclonal anti-Vn antibody  supporting the results of an earlier antibody study . Hence, both inhibitory and supportive roles of Vn in thrombosis have been reported.
In this study, we investigated the role of Vn in thrombotic/hemostatic events by directly monitoring the process of thrombus growth in live mice lacking Vn using two models of intravital microcopy. In contrast to an earlier study , our data confirm that Vn promotes thrombus growth and stabilizes thrombi both before and after vessel occlusion. Interestingly, we demonstrated that while platelet Vn supports platelet aggregation, plasma Vn inhibits platelet aggregation.
Materials and methods
Vn–/– mice (backcrossed > 10 generations to C57BL/6J genetic background) were kindly provided by D. Ginsburg (University of Michigan, Ann Arbor, MI, USA). All experimental procedures were approved by the Animal Care Committee at St Michael's Hospital. C57BL/6J WT mice were purchased from Jackson Laboratories (Bar Harbor, ME, USA).
In vivo intravital microscopy thrombosis models
The mesentery arteriole thrombosis model Intravital microscopy was performed as we previously described [2,3,20,21]. Briefly, platelets were isolated from platelet-rich plasma (PRP) and fluorescently labeled with calcein acetoxymethyl ester (1 µg mL−1; Molecular Probes, Eugene, OR, USA). Labeled platelets (5 × 106 platelets g−1 of mouse) were injected into the tail vein of genotype-matching male mice. The mice were anesthetized and the mesentery was exteriorized. A single arteriole was chosen in each mouse based on size, shear rate and vessel exposure. The diameter of and shear rate in the arteriole in WT mice (99.2 ± 2.4 µm, 1537 ± 53.7 s−1) were similar to those in Vn–/– mice (95.2 ± 2.7 µm, P = 0.27, 1561 ± 64.2 s−1, P = 0.77). Injury was induced by FeCl3. The whole process of thrombus formation was monitored and recorded. A total of 26 WT mice and 23 Vn–/– mice were studied. The characteristics of thrombus formation were compared as before [2,3,20,21] by: (i) the number of fluorescently labeled platelets deposited on the vessel wall during the interval of 3–5 min after injury; (ii) the time required for the formation of a thrombus of diameter > 20 µm; (iii) the number of thrombi with diameter > 30 µm that embolized from the viewing field; (iv) the number of arterioles that reopened after occlusion; (v) vessel occlusion time.
The cremaster muscle arteriole thrombosis model To evaluate dynamic accumulation of platelets and fibrin within thrombi in vivo, we used a laser to induce an arteriole thrombus in the cremaster muscle . Briefly, male mice were anesthetized and a trachea tube inserted to facilitate breathing. Antibodies and anesthetic (Nembutal, 0.05 mg kg−1) were administered via a jugular vein cannulus. The cremaster muscle was exteriorized for intravital microscopy under an Olympus BX51WI microscope and was superfused throughout the experiment with preheated bicarbonate-buffered saline. Platelets were labeled by injecting a rat antimouse CD41 antibody (Leo.A1, Emfret, Germany, 0.1 µg g−1) secondarily labeled with Alexa-488-conjugated chicken antirat antibody (Molecular Probes, 0.5 µg g−1). Fibrin was labeled with a mouse antihuman fibrin antibody (T2G1, 2 µg g−1; Accurate Biochemicals, Westbury, NY, USA), which cross-reacts with mouse fibrin and does not recognize fibrinogen, coupled to Alexa-660 using the supplier's protocol (Molecular Probes). Multiple independent upstream injuries were performed with a pulsed nitrogen dye laser. A total of 21 arterioles in two Vn–/– mice and 16 arterioles in two WT mice were studied. Fluorescent images of thrombus formation were captured and analyzed on a fast PC using Slidebook (Intelligent Imaging Innovations Inc., Denver, CO, USA).
In vitro platelet aggregation
Platelet preparation WT and Vn–/– PRP was prepared from citrated whole blood (1 : 9 v/v) by centrifugation at 300 × g for 7 min at room temperature. Platelet-poor plasma (PPP) was prepared by centrifugation at 1500 × g for 25 min, then at 10 000 × g for 5 min to remove the remaining cells. Gel-filtered platelets were isolated from PRP using a Sepharose 2B chromatography column [2,3].
Platelet aggregation Platelet aggregometry was performed at 37 °C, 1000 r.p.m. using a dual-channel Payton aggregometer (Payton Associates Ltd., Scarborough, ON, Canada). Samples of 500 µL of PRP, adjusted to a concentration of 3 × 108 platelets mL−1 with autologous PPP, were aggregated by ADP (5–20 µm; Sigma, St Louis, MO, USA). We also performed similar experiments with WT and Vn–/– platelets diluted in PPP (1 : 6 volumes) from the opposite genotype and by adding purified murine multimeric plasma Vn (10 µg mL−1) (Molecular Innovations, Inc., Southfield, MI, USA). To study gel-filtered platelet aggregation, samples of 500 µL gel-filtered platelets, suspended in PIPES buffer (pH 7.0, 5 mm PIPES, 137 mm NaCl, 4 mm KCl, 0.1% glucose) to a concentration of 3 × 108 platelets mL−1, were aggregated by murine thrombin (0.5–2 U mL−1; Sigma). Some experiments were also repeated using BALB/C mice (Jackson Laboratories, Bar Harbor, ME, USA) as WT control or hirudin (50 µg mL−1) as an anticoagulant (kindly provided by J. Fareed, Loyola University, Chicago, IL, USA) on a computerized Chrono-log aggregometer (Chrono-Log Corp., Havertown, PA, USA) as indicated.
Shear-induced platelet aggregation
Whole blood samples from either Vn–/– mice (n = 8) or WT mice (n = 7) were subjected to different shear rates (0–26 600 s−1) with a cone-and-plate viscometer [23,24] (CAP-2000 viscometer; Brookfield Engineering Laboratories, Inc., Middleboro, MA, USA). Whole blood was collected from adult mice by heart puncture into 1/10th volume of 3.8% sodium citrate. Blood (67 µL) was subjected to the indicated shear rates for 30 s at 37 °C. Immediately after exposure to shear stress, 5 µL of tested blood were added to 20 µL of a 3 : 3 : 14 antibody mixture [R-phycoerythrin (R-PE)-conjugated hamster antimouse CD61 monoclonal antibody, fluorescein isothiocyanate (FITC)-conjugated rat antimouse CD62P monoclonal antibody (BD Biosciences Pharmingen, San Diego, CA, USA) and incubation buffer], fixed with freshly prepared paraformaldehyde (0.5%), and incubated in the dark for 30 min. Finally, samples were diluted with 450 µL of incubation buffer and acquired on a FACScan flow cytometer (Becton Dickinson, San Jose, CA, USA).
Data are presented as mean ± SEM. Statistical significance was assessed by unpaired Student's t-test or by χ2 test as indicated.
Mesenteric arteriole intravital microscopy model
Early platelet adhesion was not affected in Vn–/– mice Vn has been found in the vessel wall  and is involved in both fibrinolysis  and fibrin–Vn matrix formation [19,25]. Since Vn is a ligand of platelet β3-integrins, and both fibrin  and the fibrin–Vn matrix  are able to support platelet adhesion in vitro, we examined whether platelet adhesion was altered in Vn–/– mice. Using the mesenteric arteriole model, we observed that the number of single fluorescently labeled adherent platelets per minute, determined in the interval 3–5 min after the injury, was not significantly different in Vn–/– vs. WT mice (WT = 109.2 ± 8.7 min−1, Vn–/– = 98.1 ± 9.6 min−1, P = 0.39) (Fig. 1A).
Thrombus growth and stability were impaired in Vn–/– mice Several minutes after injury, platelets adhered more stably to the vessel wall and started to form visible platelet aggregates. For each mouse, we measured the time from injury to formation of a thrombus > 20 µm in diameter. Figure 1B shows that the mean time required for initial thrombus formation was increased by 1.8 min in Vn–/– mice, but statistical difference was not reached (P = 0.089). In WT mice, following initial thrombus formation, stable thrombi continuously grew and formed large stable thrombi along the injured vessel wall, which eventually led to complete vessel occlusion. While we observed little embolization in WT mice, thrombi in Vn–/– mice were more unstable and tended to dissociate and form frequent emboli during growth (number of emboli before occlusion: WT = 1.8 ± 0.5, Vn–/– = 5.4 ± 1.4, P = 0.011) (Fig. 1C).
Vessel occlusion was delayed and unstable in Vn–/– mice Vessel occlusion was defined as occurring when blood flow completely stopped for at least 10 s. As shown in Table 1, time to vessel occlusion in Vn–/– arterioles was 5.7 min longer than in wild-type arterioles (WT = 16.5 ± 0.7, Vn–/– = 22.2 ± 1.5 min, P = 0.007). There was also a tendency of downstream vessel occlusion caused by emboli in Vn–/– mice (3/23 in Vn–/– mice, 0/26 in WT mice), but no statistical difference was reached (χ2 = 3.507, P > 0.05). The impaired stability of Vn–/– thrombi was still observed during the 10 min after vessel occlusion. The occlusive thrombi in Vn–/– arterioles were fragile and easily dissociated into large emboli, resulting in partial reopening of occluded arterioles (Fig. 1D). The restoration of blood flow after vessel occlusion occurred more frequently in Vn–/– mice (13 of 23 injured arterioles) than in WT mice (six of 26 injured arterioles, χ2 = 5.77, P < 0.025) (Table 1).
Table 1. Comparison of vessel occlusion time and vessel reopening after occlusion in wild-type and Vn–/– mice.
Mean occlusion time (min)
No. of reopened vessels after occlusion
Wild type, n = 26
16.5 ± 0.7
Vn–/–, n = 23
22.2 ± 1.5
The cremaster muscle arteriole thrombosis model
Platelet accumulation was decreased within the thrombus of Vn–/– mice Since contradictory effects of Vn deficiency on thrombus formation have been observed in different animal models [9,17,18], and variations in thrombus formation can occur in different vascular beds [20,27], we also compared thrombus formation in these mice using the cremaster muscle arteriole thrombosis model. Immediately after injury, wide-field fluorescent and bright field images of the microcirculation showed platelet and fibrin deposition in the thrombus. Acquisition of the intensified thrombus images began approximately 10 s after the laser-induced arteriole injury. Fluorescent images showed rapid accumulation of platelets at the site of injury, amassing within the growing thrombus, and then diminishing in number.
Analysis of thrombi in WT mice showed platelet accumulation reached maximal deposition at approximately 50 s after arteriole injury, followed by a decrease in the amount of platelets. While the time required for maximal platelet deposition did not vary between Vn–/– and WT mice, the number of platelets that accumulated within the thrombus was significantly less in Vn–/– mice (Fig. 2A, P < 0.05); there was also a fluctuation of platelet content over time in Vn–/– mice (Fig. 2A), suggesting frequent embolization.
Fibrin content is lower and unstable within the thrombus of Vn–/– mice Fibrin appeared at the site of arteriole injury and expanded within thrombi shortly thereafter. In both Vn–/– and control mice, fibrin deposition increased with thrombus growth, and reached a maximum at approximately 50 s (Fig. 2B). However, fibrin levels were significantly lower in Vn–/– thrombi (P < 0.05). In each individual Vn–/– thrombus, fibrin content fluctuated during the time course (data not shown). These findings are consistent with the frequent premature embolization we observed in Vn–/– mice using both in vivo models.
Increased ADP-induced platelet aggregation in PRP in Vn–/– mice To define the role of Vn in thrombus growth, we studied platelet aggregation in vitro. Interestingly, Vn–/– PRP showed a 32% increase in aggregation compared with WT PRP (P < 0.05) (Fig. 3A). Similar results were also seen with lower doses of ADP (5 and 10 µm) (data not shown). This enhanced aggregation was abolished when Vn–/– platelets were aggregated in WT PPP but was restored when WT platelets were aggregated in Vn–/– PPP (Fig. 3C). These results were reproducible when using hirudin as an anticoagulant (Fig. 3B), and using BALB/c mice as WT control (Fig. 3D). Furthermore, when exogenous Vn (10 µg mL−1) was added to Vn–/– PRP, the enhancement of aggregation disappeared (Fig. 3E). These data indicated that plasma Vn, but not platelet Vn, is an inhibitor of platelet aggregation.
Decreased gel-filtered platelet aggregation induced by thrombin in Vn–/– mice Our in vitro ADP-induced platelet aggregation data contradict our in vivo thrombosis results. We hypothesized that this discrepancy may be due to different forms of Vn in plasma and platelets [15,16]. We therefore studied the contribution of endogenous platelet Vn in platelet aggregation by using gel-filtered platelets. After thrombin (0.5–1 U mL−1) stimulation, we observed a short first wave of aggregation (19.5 ± 3%), followed by a robust second wave reaching to 75.8 ± 4% of maximum percentage of aggregation in WT platelets. Interestingly, the same treatment did not induce a detectable second wave of aggregation in Vn–/– platelets even after prolonged recording (P < 0.0001, Fig. 4). Upon visual inspection, Vn–/– aggregates were smaller and more single non-aggregated platelets were present under the microscope (Fig. 4). At higher doses of thrombin (1–2 U mL−1), the second wave of Vn–/– platelet aggregation could be induced but was delayed (data not shown). These data suggest that granule-released Vn supports platelet aggregation.
Increased shear-induced platelet aggregation in Vn–/– mice Since high shear stress is able to activate platelets and induce platelet aggregation [23,24,28], we further examined shear-induced platelet aggregation (SIPA) in Vn–/– and control mice. As shown in Fig. 5, shear rates of 1000–8000 s−1 induced a higher percentage of SIPA in Vn–/– whole blood compared with controls (P < 0.05). However, no difference was found after exposure to higher shear rates (10 000–26 600 s−1), as can be found at the sites of arterial stenosis. Under these conditions (shear rates of 10 000–26 600 s−1), platelet granules were released, since P-selectin was significantly expressed on the platelet surface (data not shown). This result is consistent with our data regarding ADP-induced platelet aggregation in Vn–/– PRP, suggesting that plasma Vn, but not platelet Vn, may attenuate platelet aggregation.
The role of Vn in thrombosis and hemostasis is not fully understood and no clinical disorder is linked to Vn deficiency or mutation. In the present study, using two different intravital microscopy models, we demonstrated that Vn is a thrombus stabilizer and promotes thrombus growth at high shear stress. In addition to its role in fibrin stabilization in vivo, we identified an important role of platelet Vn in supporting platelet aggregation mediated by platelet granule protein(s) following thrombin stimulation. We also demonstrated a previously unidentified phenomenon that plasma Vn is able to diminish platelet aggregation.
Vn is an overall supportive factor in thrombus formation detected by two intravital microscopy thrombosis models
The role of Vn in mediating platelet adhesion at the site of vascular injury has not been studied, although the role of the fibrin–Vn matrix in supporting platelet adhesion was recently observed using a perfusion chamber . It has been demonstrated that VWF is essential for initiating platelet adhesion under high shear stress [1,2,29]. However, it is not clear whether Vn plays a role in platelet adhesion following tethering by VWF. In our intravital microscopy study in Vn–/– mice, we did not observe an influence of Vn deficiency on early platelet adhesion (Fig. 1A). This result indicates that Vn is not a major contributor for initiating platelet adhesion under high shear stress. Nevertheless, our result does not exclude the possible contribution of Vn or the fibrin–Vn matrix in supporting platelet adhesion at a later stage since, in our early studies, platelet adhesion did occur in mice lacking VWF , and fibrin indeed played a critical role in anchoring thrombus to the injured vessel wall .
We observed frequent thrombus embolization before and after vessel occlusion in Vn–/– mice (Fig. 1C,D), which is consistent with some earlier studies using carotid thrombosis models [17,18]. We previously demonstrated that thrombus stabilization is based on both fibrin formation and bridging bonds (such as fibrinogen, fibronectin, etc.) between platelets [3,20]. In this study, using the cremaster muscle arteriole thrombosis model, we provide direct evidence that fibrin deposition is impaired within thrombi in Vn–/– mice (Fig. 2B); this supports the prevailing view that Vn, by control of PAI-1, prevents fibrinolysis [17,18]. It is likely that bridging bonds between platelets are also impaired in Vn–/– mice, since we observed that platelet accumulation was decreased within thrombi (Fig. 2A) and there was a tendency to delay in the formation of the first thrombus (Fig. 1B). This impairment of bridge bonds may result from the deficiency of platelet granule Vn following thrombin stimulation. Impairment of both platelet accumulation and thrombus stabilization contributes to the delayed vessel occlusion in Vn–/– mice.
The contradictory reports in carotid thrombosis models in Vn–/– mice [9,17,18] might be accounted for by different protocols used in the individual experiments. Variation of results can also occur based on different vessels chosen [20,27]. However, in this study, although there are significant differences in these two intravital microscopy models, the results complement each other.
Impaired thrombin-induced platelet aggregation in Vn–/– mice
It has been reported that there was a trend towards enhanced thrombin-induced aggregation of washed Vn–/– platelets compared with WT platelets . This is in sharp contrast to our results (Fig. 4). This discrepancy may be due to the use of a traditional aggregometer vs. a microtiter well reader. Our results of thrombin-induced platelet aggregation are consistent with our in vivo data (Figs 1 and 2), in which vascular injuries induced by either FeCl3 or laser beam cause thrombin generation, which in turn induces platelet activation and granule release. We conclude that Vn is a supportive factor for thrombin-induced platelet aggregation.
The question of how Vn supports the second wave of thrombin-induced granule protein-mediated platelet aggregation is intriguing. A potential explanation is that Vn may enhance the formation of a macromolecular complex of β3-integrin ligands on the platelet surface. Contrary to the results found in fibrinogen-deficient platelets, in which fibronectin content was markedly increased [2,20], no alteration of fibrinogen, VWF, fibronectin, thrombospondin-1 (TSP-1), and P-selectin has been found in Vn–/– platelets (data not shown). However, this does not exclude the possibility that the promiscuous Vn, by its multiple binding sites, may directly or indirectly (e.g. via fibrin) form a macromolecular complex with these granule proteins, which increases the local avidity of these ligands for platelet β3-integrins. Highly adhesive ligand complexes may dramatically strengthen bonds between platelets, enhancing platelet aggregation and thrombus stability. Since no significant difference in thrombin-induced platelet aggregation has been found in mice lacking VWF (Ni and Wagner, data not shown), TSP-1 , and plasma fibronectin (approximately 80–90% platelet fibronectin decreased in these mice) [20,31], it seems that Vn plays a distinctive role (i.e. different from TSP-1, VWF, and fibronectin) in granular protein-mediated platelet aggregation.
Another explanation for Vn supporting the second wave of platelet aggregation is that Vn may enhance fibrin polymerization on the platelet surface after thrombin treatment. It has been reported that fibrin plays an important role in the second wave of platelet aggregation . Vn has two binding sites for fibrin  and it may have the potential to bridge fibrin monomers to polymers. During this study, we found that fibrin polymerization inhibitor peptide (GPRP) indeed abolished the second wave of platelet aggregation. There was also no difference in gel-filtered platelet aggregation when treated with thrombin receptor activating peptide (AYPGKF), which induces granule release without fibrin formation (data not shown). However, this question could be further addressed by using fibrinogen, and vitronectin/fibrinogen double-deficient mouse platelets.
Increased ADP- and shear-induced platelet aggregation in PRP and whole blood of Vn–/– mice
In our aggregometer assay, we clearly demonstrated that platelet aggregation in Vn–/– PRP was increased (Fig. 3A–D). This result was further supported by adding purified multimeric plasma Vn into Vn–/– PRP in which platelet aggregation was attenuated (Fig. 3E) and by the shear-induced platelet aggregation assay in whole blood using cone-and-plate viscometer (Fig. 5). These results are different from the antibody studies using human PRP [5,19] in which ADP-induced platelet aggregation may cause granule release. Vn-blocking antibodies therefore blocked the robust second phase of aggregation that masked the inhibitory role of plasma Vn.
In summary, this is the first study of thrombus formation in microvessels in Vn–/– mice using two distinct intravital microscopy thrombosis models. We found that Vn supports thrombus formation and stability after vascular injury. We also identified the different roles of plasma Vn (inhibition) and platelet Vn (enhancement) in platelet aggregation. Thus, at the sites or times of thrombus formation when weak platelet agonists such as ADP dominate, plasma Vn may prevent thrombus formation. However, at the sites of severe injury when thrombin is generated, the released platelet granule Vn may quench the inhibitory role of plasma Vn, and support platelet aggregation. We suggest that therapeutic strategies to target Vn, in particular antibodies against the platelet form of Vn, may have antithrombotic significance.
We thank D. Ginsburg for providing Vn–/– mice, and D. D. Wagner for advice and comments on the manuscript. We also thank C. M. Spring and W. Lalonde for assistance with preparation of the manuscript. This study was supported by Heart and Stroke Foundation of Canada (Ontario) Grant no. NA5252 and Start-up Funds from St Michael's Hospital and Canadian Blood Services, and a Connaught/University of Toronto New Staff Matching Fund Award.