In normal conditions, platelets adhere to the subendothelium, particularly to collagen, after vessel wall injury (Sakariassen et al, 1987). Nowadays, it is assumed that this adhesion first involves platelet GPIb–IX via VWF, especially when shear rates are high (Ruggeri, 1997), then subsequently platelet GPIaIIa and GPVI in a two-step mechanism (Barnes et al, 1998). Platelets then aggregate to form thrombi after fibrinogen binding to their αIIbβ3 integrin receptor (Nurden et al, 1999). In vitro platelet aggregation and in vivo thrombi formation are thus null in patients with type I GT, who completely lack αIIbβ3. Nevertheless, it is not known whether in vitro platelet aggregation is also absent in classical type II and type III (variant) disease; data are available on the in vivo thrombi formation capacity in these patients. To address this point, we used an ex vivo model of a cylindrical perfusion chamber, in which human type III collagen is exposed to non-anticoagulated blood as under physiological conditions. This chamber is circular, like a vessel wall. Owing to the small amount of blood required, it can be used in the clinical investigation of patients with bleeding disorders. In normal subjects, we observed many mushroom-shaped large thrombi with an average platelet adhesion of 66% and a moderate fibrin meshwork. VWF was present at the base of the thrombi and within the thrombi. In patients with type II and variant GT, platelet–collagen adhesion was maximal and we surprisingly observed the formation of medium-sized thrombi, together with a thick meshwork of fibrin around the platelet aggregates. A strong labelling for VWF was detected both inside thrombi and fibrin network. The heterozygous daughters of the patients with type II or variant GT had normal platelet adhesion and normal thrombus formation. In contrast, in patients with type I GT, we confirmed that there was neither thrombi formation nor fibrin meshwork as already reported (Sakariassen et al, 1986; Weiss et al, 1986, 1987, 1991; Nurden et al, 1999), although platelet collagen adhesion was maximal and underlined by a positive thin staining for VWF. Furthermore, we noticed that type I GT platelets showed no signs of spreading, possibly reflecting the lack of β3-mediated outside/inside platelet signalling, which in normal platelets induces cytoskeletal protein activation, filopodia emission and platelet spreading (for review see Fox, 2001).
Our data, and those of pharmacological studies using the same model (Andréet al, 1996), stress the crucial role of αIIbβ3 in platelet spreading and thrombi formation. Enhanced platelet adhesion, which was observed in all the GT patients, might be due to the failure of platelets to aggregate and the subsequent availability of the other glycoproteins, such as GPIb–IX, GPIa–IIa and GPVI, to adhere to collagen. Of note, the formation of thrombi of medium size with a thick fibrin meshwork in type II and variant GT was very surprising and has never been reported before. To date, most investigators used everted segments of de-endothelialized rabbit aorta or human umbilical artery, which were exposed in an annular perfusion chamber intended to recirculate anticoagulated blood (Baumgartner, 1973; Tschopp et al, 1975; Sakariassen et al, 1986; Weiss et al, 1993). In these previous studies, the thrombi formation was probably prevented by anticoagulation, which inhibits fibrin formation and can affect mechanisms involved in thrombotic pathways. Moreover, all the results obtained previously with citrated blood from patients with type I or type II GT were always pooled, thus masking the type II response (Weiss et al, 1986, 1991). Nevertheless, a slight increase in fibrin was reported in a perfusion chamber using non-anticoagulated blood from a pool of patients with type I or II GT (Weiss et al, 1986; Kirchhofer et al, 1995). Furthermore, to our knowledge, thrombogenesis has never been studied in variant GT which is very rare. In this cylindrical perfusion chamber, the fibrin-dependent thrombogenic responses from patients with type II or variant GT, whatever their mutations, were identical. Thus, classical type II GT might be due to a qualitative abnormality of the αIIbβ3 complex, leading to a very reduced expression of this complex, while in classical type III GT the qualitative abnormality of the complex is accompanied by a normal or subnormal amount. Recently, several patients with classical type II GT and a qualitative abnormality either on the αIIb gene (Wilcox et al, 1995; Grimaldi et al, 1998) or on β3 (Grimaldi et al, 1996; Jackson et al, 1998; Ward et al, 2000; Fullard et al, 2001; Morel-Kopp et al, 2001; Ruiz et al, 2001), as in patient 3 (Schlegel et al, 1999), have been reported. The original classification might be modified towards a new classification as we proposed recently (Bellucci & Caen, 2002). The new type 1 GT would correspond to the classical type I while the new type 2 GT would include the type 2a (classical type II) and the type 2b (classical type III). The unexpected presence of fibrin meshwork in the thrombi from our classical type II and variant patients could be due to the fact that fibrin formation is not negligible in this cylindrical perfusion chamber, perhaps produced at least partly from fibrinogen released from alpha granules; it might be less important in type I GT, as the failure of platelets to form thrombi may result in the maintenance of an efficient blood flow that might reduce fibrin formation, and also allow any fibrin and thrombin formed to be removed from the capillary tube. We noticed that thrombin generation in platelet-rich plasma from our variant or type II GT patients, according to the method of Reverter et al (1996), was normal or subnormal but not increased (data not shown), in agreement with the data of other GT patients (Reverter et al, 1996). Thus, the strong fibrin meshwork observed in these GT patients might not necessarily originate from an increased platelet procoagulant activity. But type II GT platelets with a much-reduced amount of αIIbβ3 exhibited reduced fibrinogen binding (Lee et al, 1981), which was nil in our variant GT patients (Caen et al, 1983; Nurden et al, 1987). Thus, in type II and variant GT, this defective fibrinogen binding might favour either an increase in platelet–fibrin interaction or the stabilization of bound fibrin. In agreement with this hypothesis, Niewiarowski et al (1981) had reported that polymerized fibrin was incorporated into both type II GT and normal platelets. Among the possible binding sites for fibrin, studies with mutant fibrinogen peptide ligands raised the possibility that fibrinogen and fibrin bind to different or overlapping sites on αIIbβ3 (Rooney et al, 1996); β3 integrin was then shown to bind fibrin by either the αIIbβ3 or αvβ3 complex (Katagiri et al, 1995), and subsequently β1 integrin was also shown to allow platelet adhesion to fibrin (Tanoue et al, 1999). In our patients, clot retraction was normal in patient 1 and present, although reduced, in our type II GT patients and patient 2, confirming a positive reaction of platelets with fibrin. Very recently, a new variant GT patient with a Leu262Pro mutation in the integrin β3 was reported (Ward et al, 2000). In this patient, by the expression of the mutant β3 integrin in human embryonal kidney 293 cells, it was possible to directly show that the abnormal αIIbβ3 complex was able to bind fibrin but not fibrinogen (Ward et al, 2000). On the other hand, the strong labelling of VWF detected in our model, together with the fibrin network, both inside and between the thrombi, argues for a role for VWF in the interaction of platelets with polymerized fibrin, as previously shown by Loscalzo et al (1986). This interaction, which involves platelet GPIb, was well demonstrated in vitro in platelet-rich plasma by Beguin et al (1999), and does amplify platelet procoagulant activity and thrombin generation. Finally, in the flow conditions studied in our model, platelet interaction of type II and variant patients with fibrin could first involve αIIbβ3 or αvβ3, as it has not been observed in type I GT patients lacking αIIbβ3, which could be amplified by the interaction of fibrin-bound VWF with platelet GPIb. The precise molecular mechanisms remain to be defined in further studies.