Vitronectin Inhibits the Thrombotic Response to Arterial Injury in Mice

C57BL/6J mice were purchased from Jackson Labs (Bar Harbor, ME). The generation of VN-deficient mice by homologous recombination in embryonic stem cells has been reported previously.22PAI-1–deficient (PAI-1 −/−) mice were a gift from Dr P. Carmeliet (University of Leuven, Leuven, Belgium).23To eliminate potential effects of genetic differences among mouse strains on experimental results, consecutive generations of mice carrying the null VN allele were backcrossed to C57BL/6J mice. Only mice that were the product of ≥8 backcrosses were used in experiments comparing VN−/− mice with VN +/+ mice. PAI-1−/− mice used in experiments were the product of greater than 10 backcrosses to the C57BL/6J genetic background. Genotyping of mice was performed by polymerase chain reaction (PCR) analysis of tail DNA as described previously.22 24 All animal care and experimental procedures complied with the Guide for Care and Use of Laboratory Animals (Department of Health, Education, and Welfare Publication No. NIH 78-23) and were approved by the University of Michigan Committee on Use and Care of Animals.

Native human VN was purchased from Molecular Innovations Inc (Royal Oak, MI). Human α-thrombin was from CalBiochem (La Jolla, CA). Thrombin substrate (Spectrozyme TH) was from American Diagnostica (Greenwich, CT). Ferric chloride (FeCl₃) was from Mallinckrodt Chemical (Paris, KY). Human VN purified by heparin-affinity chromatography, reptilase (Atroxin), and human fibrinogen (plasminogen-free) were from Sigma (St Louis, MO).

A previously described carotid artery thrombosis protocol was used.25 26 Adult mice (6 to 8 weeks old; weight, ∼25 g) were anesthetized by intraperitoneal injection of pentobarbital (120 mg/kg). The left common carotid artery was surgically exposed and a miniature Doppler flowprobe (Model 0.5VB; Transonic Systems, Ithaca, NY) was placed on the surface of the artery. Sodium chloride solution (0.9%) was placed in the surgical wound to allow Doppler monitoring, and baseline blood flow was recorded using a Transonic Model T106 flowmeter. Thereafter, sodium chloride solution was removed from the wound and filter paper (0.5 × 1.0 mm) saturated with 10% FeCl₃ was applied to the adventitial surface of the carotid artery, immediately proximal to the flow probe. After 3.0 minutes, the filter paper was removed, saline solution was again placed in the wound, and carotid blood flow was monitored (ie, it was not possible to monitor carotid artery blood flow during the application of FeCl₃). Time to thrombotic occlusion after initiation of arterial injury was defined as the time required for blood flow to decline to 0 mL/min. If the carotid artery was observed to be thrombosed at the earliest time point that flow could be monitored after initiation of injury (ie, 3.0 minutes), time to occlusion was recorded as ≤3.0 minutes. The operator was blinded to mouse genotype while performing all experiments.

For some animals, the arterial vasculature was perfusion fixed immediately after completing the thrombosis protocol, as described previously.27 Injured arterial segments were excised, embedded in paraffin, sectioned, and subjected to hematoxylin and eosin staining.

Mice (6 to 8 weeks old) were anesthetized by intraperitoneal injection of phenobarbital (100 mg/kg) and placed in a restraining chamber from which the tail protruded. The distal 1 mm of the tail was amputated and the tail was immersed for 10 minutes in 1 mL of 0.9% NaCl warmed to 37°C. Blood loss was determined by measuring the absorbance of saline at 560 nm and comparing the result to a standard curve constructed from known volumes of mouse blood.

Platelet aggregation was studied using a previously described microtiter plate assay.28 Blood was collected into citrate anticoagulant from anesthetized mice by inferior vena cava puncture with a 25-gauge needle. Platelet-rich plasma (PRP) was prepared by centrifuging blood (120 g for 6 minutes) in 0.5 mL polypropylene tubes at room temperature in a swing-out rotor. After adjustment to a platelet count of 2.5 × 10⁸/mL by the addition of citrated platelet-poor plasma, 95 μL of count-adjusted PRP was placed in 96-sample microtiter plate wells and incubated at 37°C in a SpectraMax 340 microtiter plate reader (Molecular Devices, Sunnyvale, CA). ADP (12.5 μmol/L) was added, and the absorbance of wells at 595 nm was monitored at 20- to 30-second intervals. Plates were shaken automatically for 15 seconds between each reading. The percentage of aggregation was calculated as described.28 Gel-filtered platelets suspended in Tyrode’s buffer,29 prepared by Sepharose 2B chromatography (Sigma), were used to study thrombin-induced aggregation as described above.

Platelet-poor plasma was prepared by centrifuging citrated mouse blood for 8 minutes at 16,000g. Thrombin times, activated partial thromboplastin times (APTTs), and reptilase times were performed using a KC4A Micro apparatus (Amelung GmbH, Lemgo, Germany) according to the manufacturer’s instructions. Thrombin amidolytic activity was measured by incubating thrombin and Spectrozyme TH at room temperature in 0.01 mol/L Tris-HCl, 0.14 mol/L NaCl, pH 7.5, and monitoring the absorbance of reaction mixtures at 405 nm in a microtiter plate reader. Platelet counts and hematocrits of citrated whole blood were measured using a Model H-10 blood cell counter (Texas Instruments Laboratories, Houston, TX). Fibrinogen/fibrin degradation products (FDP) were measured with a Staphylococcal clumping factor assay (Catalog 850-ST; Sigma) according to the manufacturer’s instructions. Plasma fibrinogen concentrations were determined by the fibrin clot opacity method, as described.30 31 In this assay, the limit optical density of dilute plasma during prolonged incubation with thrombin is directly proportional to plasma fibrinogen concentration. Briefly, 20 μL of citrated plasma were added to a spectrophotometer cuvette containing 400 μL of 0.05 mol/L Tris-HCl (pH 7.5), 0.15 mol/L NaCl. Thrombin/calcium solution (20 μL) was added to yield final concentrations of 0.45 U/mL and 0.9 mmol/L, respectively. After 10 minutes of incubation at room temperature, absorbance at 340 nm was measured. Fibrinogen concentration in pooled plasma prepared from 4VN +/+ mice was defined as 100%. Fibrinogen levels in pooled (n = 4 mice) VN −/− plasma were determined by comparison to a standard curve constructed from dilutions of pooledVN +/+ plasma. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting of diluted human plasma was performed with the PhastSystem (Pharmacia, Uppsala, Sweden). Primary antibody was goat antimouse fibrinogen (Accurate Inc, Westbury, NY). Secondary antibody was peroxidase-conjugated rabbit antigoat IgG (Zymed Labs, South San Francisco, CA). Blots were developed by the chemiluminescence method (ECL reagent kit; Amersham, Little Chalfont, UK).

Data are presented as mean ± 1 standard error of the mean (SEM), unless otherwise indicated. The two-sample t-test or Mann-Whitney Rank Sum test were used to determine if significant differences existed between experimental groups.

Experimental groups consisted of 14 VN +/+ mice (5 male and 9 female) and 15 VN −/− mice (6 male and 9 female). Representative carotid artery blood flow tracings before and after vascular injury are shown in Fig 1. Thrombotic occlusion occurred ≤3.0 minutes after initiation of vascular injury in 8 of 15 VN −/− mice, but only in 1 of 14 VN +/+ mice (Table 1). Median occlusion times were 16.4 minutes for VN +/+ mice and ≤3.0 minutes for VN −/− mice. Mean occlusion times (calculated by using values of 3.0 minutes as occlusion times for mice whose arteries were already occluded when flow monitoring was resumed after vascular injury) were significantly shorter in VN−/− mice (6.0 ± 1.2 minutes) than in VN +/+ mice (17.8 ± 2.3 minutes, P < .001). Blood platelet counts and hematocrits did not differ significantly between VN+/+ and VN −/− mice (Table 2). APTTs, reptilase times, and bleeding in response to tail tip amputation were similar between groups. Plasma fibrinogen measured by the fibrin clot opacity method did not differ between groups (100% ± 10.8% and 93.8% ± 7.2% in VN +/+ mice and VN −/− mice, respectively; P > .7). Similarly, Western blot analysis of diluted plasma samples showed no apparent differences in plasma fibrinogen between VN +/+ mice and VN −/− mice (data not shown). Fibrinogen/fibrin degradation products were not detectable in pooled serum obtained from VN +/+ mice (n = 4) orVN −/− mice (n = 4). Control experiments showed that the FDP assay readily detected murine fibrin degradation products (data not shown). Mean carotid artery blood flow prior to injury did not differ between VN +/+ mice and VN −/− mice (0.78 ± 0.2 mL/min and 0.85 ± 0.2 mL/min, respectively,P > .2). No significant differences in gross or microscopic appearance of uninjured arteries were noted between VN +/+ andVN −/− mice (n = 3 each group, data not shown). Histologic analysis of thrombi (n = 4) recovered immediately after carotid injury showed that they consisted predominantly of platelets, with no discernible differences between genotypes (Fig2).

Given the platelet-rich composition of carotid artery thrombi generated in this model, platelet aggregation studies were performed to compare platelet function of VN +/+ and VN −/− mice. Pooled samples of platelet-rich plasma were prepared and in vitro aggregation was induced with ADP (12.5 μmol/L). As shown in Fig 3A, no significant differences were observed between VN +/+ and VN −/− mice. There was a trend towards enhanced thrombin-induced aggregation of washed VN −/− platelets compared with VN+/+ platelets (Fig 3B). Consistent with this observation, addition of heparin-affinity-purified (ie, denatured-renatured) human VN (200 μg/mL) to washed VN −/− platelets significantly inhibited thrombin-induced aggregation (P < .005; Fig 4). However, native human VN (350 μg/mL) had no detectable effect on thrombin-induced aggregation (data not shown).

Because thrombin is an important determinant of platelet-dependent arterial thrombosis,32 we performed experiments to test the hypothesis that VN produces its anticoagulant effect by inhibiting thrombin procoagulant activity. Thrombin times were performed by adding human α-thrombin (4.4 U/mL) to pooled samples (n = 4) of citrated plasma. Thrombin times of plasma from VN −/− mice (20.5 ± 2.1 seconds) were significantly shorter than those ofVN +/+ mice (34.2 ± 6.7 seconds, P < .01). Furthermore, addition of native human VN (350 μg/mL) to VN−/− mouse plasma prolonged the thrombin time into the normal range, whereas the addition of an equal volume of buffer containing bovine serum albumin (BSA; 350 μg/mL) had no effect (Table 3). VN also inhibited clotting of purified human fibrinogen (thrombin times, 32.9 ± 4.9 seconds and 42.9 ± 2.4 seconds in the absence and presence of VN [350 μg/mL], respectively; P < .05), but had no effect on thrombin amidolytic activity, measured by thrombin (25 nmol/L) hydrolysis of low molecular weight substrate (Spectrozyme TH; 150 μmol/L; data not shown).

VN accelerates the capacity of PAI-1 to inhibit thrombin by greater than 200-fold.19 20 If the antithrombotic effect of VN observed in our model were mediated by enhancement of the antithrombin function of PAI-1, then PAI-1 deficiency would be expected to mimic VN deficiency, resulting in accelerated thrombus formation. To test this hypothesis, we measured the time to occlusive thrombus formation after FeCl₃ carotid artery injury in PAI-1 +/+ mice (n = 6) and PAI-1 −/− mice (n = 6). However, mean occlusion times did not differ significantly between groups (15.5 ± 1.9 minutes and 13.8 ± 2.4 minutes forPAI-1 +/+ mice and PAI-1 −/− mice, respectively; P > .5).

In this study, we observed a significantly enhanced rate of thrombus formation after arterial injury in VN −/− mice compared with VN +/+ mice. We used ferric chloride injury to trigger thrombosis in our experiments. This method has been used in a variety of species and vascular sites to trigger platelet-dependent thrombosis.26,33,34 Iron ions enhance conversion of O₂⁻ and H₂O₂ to oxidizing species, such as hydroxyl radical, that injure endothelial cells and markedly increase tissue factor expression in vitro and in vivo.35-38 The markedly enhanced rate of thrombosis inVN −/− mice compared with VN +/+ mice suggests that VN plays a previously unsuspected role in protecting the injured arterial wall against thrombus formation.

Our in vitro studies suggest that the antithrombotic effect of VN may be mediated by inhibition of platelet-platelet and thrombin-substrate interactions. VN binds platelet glycoprotein IIb/IIIa,16,18providing a mechanism by which VN may modulate platelet function. Because VN contains a single RGD and mutagenesis of this sequence blocks platelet binding,3 monomeric VN would not be expected to support platelet aggregation. In fact, prior in vitro studies demonstrated that VN inhibits platelet aggregation and competes with fibrinogen and von Willebrand factor for binding to platelets, leading to the hypothesis that VN inhibits platelet-dependent thrombosis.18 Our experiments provide the first in vivo data to support this hypothesis. We observed that VN purified under denaturing conditions, which expose its cell attachment site,10 inhibited thrombin-induced platelet aggregation. However, native VN, which does not bind integrins,10 did not. A hypothesis that is consistent with our in vivo and in vitro data is that plasma VN undergoes a conformational change at sites of vascular injury, thereby exposing its integrin-binding site and inhibiting platelet-platelet interactions by competing with fibrinogen, von Willebrand factor, or other factors for binding to glycoprotein IIb/IIIa on activated platelets. Such a negative feedback mechanism could serve to prevent excessive platelet accumulation and vascular occlusion after injury to the blood vessel wall. The proportion of VN capable of binding heparin is less than 2% in plasma, but increases over threefold upon formation of serum,39 suggesting that activation of the coagulation cascade triggers conversion of native VN to a conformationally altered form capable of binding platelets. Several factors generated or released at sites of arterial injury bind VN and expose its RGD site, including thrombin-antithrombin complex, PAI-1, and C5b-7.7,11,12 This functional activation of VN by ligand binding also may explain the inhibition of platelet aggregation by anti-VN antibodies in experiments by Asch and Podack.17 It is possible that the antibodies used in these experiments induced a conformational change in VN that exposed its RGD site, thereby inhibiting platelet aggregation by enabling VN to compete with fibrinogen for binding to platelet glycoprotein IIb/IIIa. In addition to the plasma, VN is present in platelets and the blood vessel wall.40,41 Although platelet VN and blood vessel wall VN exist to a significant extent as multimeric forms capable of supporting platelet aggregation,13 we observed accelerated platelet thrombus formation in VN −/−mice. These results suggest that, in this model, the platelet or vessel wall pools of VN are not required for platelet accumulation at sites of arterial injury and that the antithrombotic properties of VN are dominant over its potential procoagulant function.

An additional mechanism by which VN may inhibit thrombosis is by downregulating thrombin function. We demonstrated that thrombin induces clot formation more rapidly in VN-deficient plasma than in normal plasma and that addition of purified VN to VN-deficient plasma prolongs the thrombin time. Similar VN effects were observed on thrombin clotting of purified fibrinogen. VN deficiency had no effect on reptilase clotting times, suggesting that VN does not affect polymerization of fibrin monomer. VN did not inhibit thrombin amidolytic activity. Together, these results suggest that VN inhibits thrombin-fibrinogen interactions by binding to thrombin at a site distinct from its active-site, as suggested previously by Naski et al.20 A direct antithrombin effect of VN could contribute to the enhanced thrombotic response observed in VN−/− mice. In addition to its direct effects, it has been proposed that VN may inhibit thrombin function indirectly by accelerating PAI-1–mediated thrombin inhibition greater than 200-fold.19 20 However, we did not observe a significantly accelerated thrombotic response in PAI-1 −/− mice, which appears to exclude VN-dependent thrombin inhibition by PAI-1 as a mechanism responsible for the longer thrombosis times observed inVN +/+ mice.

In summary, we have shown that VN-deficient mice form occlusive arterial thrombi at an accelerated rate compared with wild-type mice. We hypothesize that the antithrombotic properties of VN are mediated by its interaction with platelet glycoprotein IIb/IIIa and by its capacity to inhibit thrombin-fibrinogen interactions, although we cannot exclude that other mechanisms may be operative as well. These findings represent the first phenotypic abnormality observed in VN−/− mice and suggest an important role for VN in inhibiting platelet-dependent thrombosis at sites of arterial injury, a previously unrecognized function of this adhesive glycoprotein.

The authors thank Randal Westrick for assistance with mouse breeding and genotyping, Mary Ellen Wechter for assistance with phlebotomy, and Drs Alvin Schmaier and Benedict Lucchesi for sharing laboratory equipment.