Platelets in atherothrombosis: lessons from mouse models


Bernhard Nieswandt, Vascular Biology, Rudolf Virchow Center for Experimental Biomedicine, University of Würzburg, Versbacher Str. 9, 97078 Würzburg, Germany.
Tel.: +49 (0)931 201 48996; fax: +49 (0)931 201 48123; e-mail:


Summary.  Platelets play a central role in hemostasis and thrombosis but also in the initiation of atherosclerosis, making platelet receptors and their intracellular signaling pathways important molecular targets for antithrombotic and anti-inflammatory therapy. Historically, much of the knowledge about hemostasis and thrombosis has been derived from patients suffering from bleeding and thrombotic disorders and the identification of the underlying molecular defects. In recent years, the availability of genetically modified mouse strains with defined defects in platelet function and the development of in vivo models to assess platelet-related physiologic and pathophysiologic processes have opened new ways to identify the individual roles and the interplay of platelet proteins in adhesion, activation, aggregation, secretion, and procoagulant activity in vitro and in vivo. This review will summarize key findings made by these approaches and discuss them in the context of human disease.


Cardiovascular and cerebrovascular diseases continue to be the leading cause of death in the developed world. Although each occurrence of thrombotic disease may reveal a somewhat unique etiology, the overall understanding of the pathomechanisms underlying thrombosis has not changed since Virchow's observation in 1856. The components of Virchow's triad, vessel injury, hypercoagulability of the blood, and obstruction of blood flow, remain the predominant factors in thrombotic episodes. It is now recognized that blood platelets play a central role in physiologic blood clotting (hemostasis) but also in both the acute and chronic phases of arterial disease, a combination of events that is best described by the term atherothrombosis [1].

Platelets adhere at sites of vessel wall injury where components of the extracellular matrix (ECM) are exposed to the flowing blood which triggers sudden platelet activation and platelet plug formation, followed by coagulant activity and the formation of fibrin-containing thrombi that occlude the site of injury. These events are crucial to prevent post-traumatic blood loss but they are also a major pathomechanism in arterial thrombosis. In addition, platelets contribute to the development and progression of atherosclerosis through proinflammatory mechanisms. Therefore, much effort has been spent during the last decades on the characterization of the platelet receptors and signaling pathways that orchestrate the processes of adhesion, activation, coagulant activity, aggregation, and granule release in order to identify potential targets for antiplatelet drugs that efficiently inhibit the thrombotic and/or proinflammatory activity of the cells while preserving their hemostatic function. In addition to the enormous interest in their (patho-)physiologic role in vivo, platelets are increasingly used as a convenient and physiologically relevant model for the cellular mechanisms underlying regulated adhesiveness.

Studies on human platelet function have several obvious limitations that have been overcome by the use of animal models, most importantly mice. Because of its small size, high fertility, and exceptional reproductive capacity, the common laboratory mouse (Mus musculus, M. domesticus) has become the most frequently used inbred animal species for biologic research purposes. The advent of genetic methods that allow targeted manipulations in the mouse genome has opened new ways to study protein function and has been exceptionally successful in unraveling signal transduction pathways in platelets both in vitro and in vivo. Mutations introduced into the germline can range from null mutations to subtle changes in coding or non-coding sequences of genes, chromosomal translocations, and spatially and temporally restricted gene deletions using the Cre/loxP system. In parallel to these developments, a large number of in vivo models for platelet-related physiologic and pathophysiologic processes, such as hemostasis, arterial thrombosis, atherosclerosis, wound healing, as well as local and systemic inflammatory processes have been established. The quality of these models has been further improved by the availability of mouse-specific reagents allowing the detection of individual cell–cell interactions or spatiotemporal activation processes at sites of injury in the living animal. Arterial lesions can be induced in different branches of the vascular system and by different methods (reviewed by Day et al. [2]). One of the most frequently used procedures to induce injury is direct application of ferric chloride (FeCl3) to the adventitial surface of an artery which typically results in the formation of occluding platelet-rich thrombi within 20–40 min [3]. The exact mechanism by which thrombus formation is triggered in this model is not clear but it has been shown that the morphology of the thrombi is similar to those found in humans [4]. Another chemical method to trigger thrombosis involves the i.v. injection of the photoreactive substance Rose Bengal which rapidly accumulates in the membranes of endothelial and other cells. The subsequent exposure of an arterial segment to green light (540 nm) locally triggers a photochemical reaction of Rose Bengal resulting in the formation of reactive oxygen species that damage the endothelium and induce the formation of occlusive thrombi within 30–40 min [5,6]. More recently, various groups have used lasers to induce endothelial injury in the microcirculation [7], which typically triggers the formation of non-occluding thrombi within a few minutes. With this method, lesions of different type and severity can be produced depending on the intensity and exposure time of the laser beam which may be useful to identify potentially different mechanisms of thrombus formation triggered by the exposure of different layers of the vessel wall. Both thermal and chemical injury methods are convenient to perform and yield reproducible results but one has to be aware that they potentially modify the chemical properties of the vessel wall by denaturing proteins or other macromolecules at the lesion site which may alter their function. Endothelial denudation can also be induced mechanically in large arteries (aorta, carotid artery, femoral artery) either by direct damage of the luminal side of the vessel using a guide-wire [8,9] or indirectly by compression or ligation of the vessel using a forceps or a filament [10,11]. All these procedures result in the exposure of the native ECM thereby triggering thrombus formation by collagen- and thrombin-dependent mechanisms which may perhaps best reflect naturally occurring vascular lesions. It is important to note, however, that in all described injury models thrombus formation is triggered in normal healthy vessels whereas arterial thrombosis in humans, e.g. in myocardial infarction, mostly occurs in diseased (atherosclerotic) areas of the vascular system. Despite this major limitation, the combination of these models with genetically modified mouse lines and pharmacologic tools has improved our understanding of the hemostatic system enormously.

For the extrapolation of data from mouse to human platelets, an accurate determination of the mouse hemostatic system with regard to coagulation and fibrinolytic systems, platelet structure, and platelet receptor/enzyme systems is required and differences between the two species must be acknowledged and appreciated in the interpretation of the data. Platelet counts in mice on average are four times those of humans and platelets are only approximately one-half the volume of human platelets. Receptors and signaling pathways in mouse platelets show striking similarities to the human system, with virtually every protein represented and with every cascade appearing to serve similar functions in both species. Nonetheless, differences exist in the expression of individual proteins [e.g. protease-activated receptors (PAR), FcγRIIa], at the molecular structural level (varying degrees of homology of proteins), and therefore potentially also at the functional level. A comprehensive analysis of all aspects of platelet function in atherothrombosis is far beyond the scope of this review. Rather, we have focused on key mechanisms in the hemostatic/thrombotic cascade that have been examined in the mouse system (Table 1). An excellent and comprehensive review of the murine hemostatic system has been provided by Tsakiris et al. [12], and important aspects of murine platelet receptor function have recently been reviewed by Ware [13].

Table 1.  Hemostatic phenotype of mice with defects in platelets or adhesive proteins
ModelBleeding timeThrombus formation in vivoReferences
  1. *fl/fl Cre+ indicates mice with part of the gene flanked by loxP sites that have been crossed with mice that express the recombinase Cre under control of the interferon-inducible Mx promoter (conditional knockout).

  2. Low tissue factor mice were constructed by a human transgene rescue (hTF+) of TF-null mice.

Membrane receptors
 α2−/−NormalUnchanged (or probably delayed) thrombus formationHoltkotter et al. [38]
Grüner et al. [40]
He et al. [41]
 β1fl/fl Cre+*NormalUnchanged adhesion and thrombus formationNieswandt et al. [33]
Grüner et al. [40]
 β3−/−Severely prolongedNo thrombus formationMassberg et al. [9]
 β3YFNormal, but tendency to rebleedLaw et al. [86]
 GPIbα−/−Severely prolongedWare et al. [16]
Blockage of GPIbαNo tethering, no adhesionMassberg et al. [9]
 GPIbβ−/−Severely prolongedKato et al. [17]
 GPV−/−Normal/decreased (?)Accelerated adhesion and thrombus formation, but decreased tendency to form occlusive thrombiKahn et al. [22]
Ramakrishnan et al. [21]
Ni et al. [23]
Moog et al. [24]
 GPVI−/−NormalKato et al. [34]
Depletion of GPVIImpaired adhesionMassberg et al. [9]
 FcRγ−/−Impaired adhesionGrüner et al. [11]
 P2Y1−/−Moderately prolongedProtected from thromboembolismFabre et al. [48]
Leon et al. [49]
 P2Y12−/−Severely prolongedDelayed formation of small, unstable thrombiAndre et al. [52]
 P2X1−/−NormalDecreased thrombus size; protected from thromboembolismHechler et al. [67]
 PAR-3−/−Severely prolongedDefective thrombus formation; protected from thromboembolismWeiss et al. [63]
 PAR-4−/−Severely prolongedDefective thrombus formation; protected from thromboembolismWeiss et al. [63]
 TP−/−Severely prolongedThomas et al. [66]
 α2A−/−Moderately prolongedFormation of unstable thrombiB. Nieswandt (unpublished observation)
 Mer−/−Normal, but tendency to rebleedThrombus size severely reduced; partially protected from thromboembolismChen et al. [100]
Angelillo-Scherrer et al. [101]
 Axl−/−Normal, but tendency to rebleedThrombus size severely reduced; protected from thromboembolismAngelillo-Scherrer et al. [101]
 Tyro3−/−Normal, but tendency to rebleedThrombus size severely reduced; protected from thromboembolismAngelillo-Scherrer et al. [101]
 Fibrinogen−/−Formation of unstable thrombiNi et al. [85]
 FgγΔ5ProlongedFormation of unstable thrombiHolmback et al. [89]
Ni et al. [90]
 VWF−/−Severely prolongedDelayed adhesion, defective thrombus formationDenis et al. [26]
 Compound fibrinogen−/−, VWF−/−Delayed thrombus initiation, formation of highly fragile thrombiNi et al. [85]
 Low plasma fibronectin (fl/fl Cre+*)NormalDelayed thrombus formationSakai et al. [92]
Ni et al. [28]
 Vitronectin−/−NormalEnhanced thrombus formation/formation of unstable thrombi?Fay et al. [93]
Konstantinides et al. [94]
 Thrombopondin-1−/−UnchangedNi et al. [28]
 Low tissue factor (mTF−/−, hTF+)NormalFormation of small thrombiParry et al. [114]
Chou et al. [102]
 CD40L−/−NormalFormation of unstable, loosely packed thrombiAndre et al. [96]
 Gas6−/−NormalFormation of small thrombi; protected from thromboembolismAngelillo-Scherrer et al. [99]
 Gq−/−Severely prolongedReduced adhesion, no thrombus formation; protected from thromboembolismOffermanns et al. [69]
B. Nieswandt (unpublished observation)
 Gα12−/−NormalUnchangedMoers et al. [71]
 Gα13fl/fl Cre+*Severely prolongedReduced adhesion, no thrombus formationMoers et al. [71]

Tethering and adhesion

The initial tethering of platelets at sites of vascular injury requires the action of a receptor that functions irrespective of cellular activation and thereby facilitates rapid interactions that resist shear forces acting on the cells. This process is mediated by glycoprotein (GP)Ib-V-IX, a structurally unique receptor complex exclusively expressed on platelets and megakaryocytes. The receptor is encoded by four different genes, the α- and β-subunits of GPIb, GPIX, and GPV [14,15]. In humans, lack or dysfunction of this receptor has been associated with the Bernard-Soulier syndrome (BSS), a congenital bleeding disorder characterized by macrothrombocytopenia, giant platelets, and inability of these platelets to aggregate in response to ristocetin. Targeted deletions of the mouse genes encoding the individual GPs delineated the role of each GP in the expression of the complex and their function during thrombosis and hemostasis. Mice deficient in GPIbα [16] or GPIbβ [17] lack the entire receptor complex and reflect human BSS, as they display a severe bleeding phenotype and macrothrombocytopenia. While increased bleeding is attributed to the lack of the extracellular domain of GPIbα and the loss of the binding ability of the receptor to several ligands including von Willebrand factor (VWF), thrombin, or Mac-1 [18] the occurrence of macrothrombocytopenia has been shown to be dependent on the expression of the cytoplasmic tail of GPIbα [19] and to be linked to a disordered membrane composition in megakaryocytes [20]. Mice expressing GPIb-IX-deficient platelets have not been studied in thrombosis models. However, inhibition of the VWF-binding site on the receptor with F′(ab) fragments of the antibody p0p/B in wild-type mice abrogates platelet tethering and adhesion at the injured arterial wall confirming the mandatory function of this receptor for the adhesion process in vivo [9].

In contrast to GPIb and GPIX, the lack of GPV does not cause a BSS-like phenotype in mice as GPV is dispensable for surface expression of the receptor complex [21,22]. These findings are in line with the fact that no human mutations within the GPV gene have been described in BSS patients. Functional data concerning GPV obtained from two independently generated mouse lines are somewhat contradictory but reflect its role as a thrombin substrate and collagen receptor. Ramakrishnan et al. observed slightly decreased tail-bleeding times in GPV-null animals, which could be explained by the increased sensitivity of GPV-deficient platelets to thrombin [21]. Consistent with these observations, Ni et al. reported significantly accelerated platelet adhesion and thrombus formation in FeCl3-injured mesenterial arteries of those mice [23]. Shortly after that, Moog et al. demonstrated that GPV serves as a low-affinity collagen receptor on the platelet surface and that GPV-deficient platelets display reduced adhesion and aggregation to collagen in vitro and a decreased tendency to form occlusive thrombi in injured arterioles in vivo [24]. These discrepancies on the in vivo significance of GPV deficiency may be explained by different experimental conditions favoring either collagen- or thrombin-dependent thrombus formation.

The absence of VWF as the major ligand for GPIb in all compartments in humans causes severe defects in primary hemostasis and coagulation [25]. Such VWF type 3 patients have strongly reduced factor VIII levels and suffer from spontaneous bleeding, a phenotype that is also found in VWF knockout mice. These mice display a massively prolonged bleeding time and defective thrombus formation in FeCl3-injured arterioles emphasizing the importance of VWF in hemostasis and thrombosis [26]. However, in contrast to GPIb, VWF appears not to be essential for thrombus formation, as much delayed adhesion still occurs even under arterial flow conditions indicating that GPIb can initiate adhesion by interacting with other ligands. One of the strongest candidates is thrombospondin-1 which has been shown to interact with GPIb under high shear flow conditions in vitro [27]. Thrombospondin-1-deficient mice show no obvious hemostatic defect [28] but it will be interesting to study platelet adhesion and thrombus formation in mice lacking both VWF and thrombospondin-1.

In contrast to VWF, subendothelial fibrillar collagen is a highly efficient substrate in supporting firm platelet adhesion and thrombus formation as it directly activates the cells and supports platelet adhesion by direct and indirect mechanisms. Besides GPIb and integrin αIIbβ3, which interact with collagen via VWF, several collagen receptors have been identified on the platelet surface, most notably integrin α2β1 and the immunoglobulin superfamily member GPVI [29]. GPVI is a platelet-/megakaryocyte-specific low-affinity collagen receptor that non-covalently associates with the FcRγ-chain, which bears an immunoreceptor tyrosine-based activation motif (ITAM) and serves as the signal-transducing subunit of the receptor. A few GPVI-deficient patients have been reported in the literature some of which had anti-GPVI antibodies in their blood which was difficult to explain [30,31]. A likely explanation for this observation came from studies in mice showing that injection of an anti-GPVI antibody (JAQ1) into mice results in a down-regulation of the receptor from the platelet surface and a GPVI knockout like phenotype for at least 2 weeks [32]. These GPVI-depleted platelets did not respond to collagen and failed to firmly adhere to the immobilized protein under high or low shear flow conditions because of defective activation of integrins α2β1 and αIIbβ3 [33]. Similar observations were also made with platelets from FcRγ-chain knockout mice which fail to express GPVI or with wild-type platelets in which the ligand-binding site of GPVI had been blocked by the addition of JAQ1 in vitro. The generation of GPVI germline knockout mice was reported in 2003 by Kato et al. and confirmed the essential role of this receptor for collagen-induced platelet activation and spreading on the matrix protein [34]. However, the authors observed no major defects in the initial adhesion of GPVI-deficient platelets to collagen in vitro. This discrepancy is presently difficult to explain but is likely based on different experimental conditions, as initial adhesion was also found normal with FcRγ-chain-deficient platelets in that study which stands in contrast to the other reports [31,33,35]. Mice lacking the GPVI/FcRγ-chain complex show no major bleeding phenotype but they are profoundly protected from arterial thrombosis. Konishi et al. found markedly reduced platelet attachment and subsequent neointimal hyperplasia at sites of guide-wire-induced arterial injury in FcRγ-chain-deficient mice [8]. Similarly, Massberg et al. observed dramatically reduced thrombus formation in FeCl3 or mechanically injured arteries of GPVI-depleted mice. As shown by in vivo fluorescence microscopy, platelet tethering/slow surface translocation at sites of arterial injury, as well as firm adhesion, is impaired in those animals [9]. This surprising observation indicated that GPIb and GPVI/FcRγ-chain could act in concert to recruit platelets to the exposed subendothelial matrix under high shear, a hypothesis that was further supported by Goto et al. who reported reduced adhesion of GPVI-deficient human platelets to immobilized VWF under high shear [36]. Studies on arterial thrombus formation in GPVI knockout mice have not been reported to date.

The role of integrin α2β1 in hemostasis and thrombosis has been controversial. Initially, the integrin was thought to be the principal collagen receptor on platelets that is essential for shear-resistant adhesion to the ECM but is not involved in the activation process. However, it is now recognized that α2β1 plays a significant, but not essential role for the adhesion process and that it contributes to the signaling process directly [37] and indirectly by reinforcing GPVI–collagen interactions [29,35]. Two α2-deficient mouse lines were independently reported in 2002 [38,39]. These mice have normal tail-bleeding times and display only minor defects in their adhesion and aggregation response to native fibrillar collagen [38]. In line with this, α2-deficient mice form occlusive thrombi in the injured carotid artery although in one study this was found to be delayed [40,41]. The only subtle defect in those mice can be explained by the fact that multiple integrin–ligand interactions contribute to platelet adhesion at sites of injury as demonstrated by Gruner et al., who found unaltered platelet adhesion and thrombus formation in the injured carotid artery even in mice with a Cre-/loxP-mediated loss of the integrin β1 subunit in platelets which besides α2β1 also lack α5β1 (fibronectin receptor) and α6β1 (laminin receptor). In those mice, platelet adhesion at the site of injury was mediated exclusively by integrin αIIbβ3 which interacts with various ligands, including VWF and fibronectin [40]. On the contrary, inhibition of αIIbβ3 in wild-type mice reduced platelet adhesion by approximately 60% suggesting that this integrin is the principal receptor not only for aggregation (see below) but also for adhesion to the ECM in vivo. However, recent studies indicate that αIIbβ3 may require strong activation signals, such as provided by GPVI, to mediate shear-resistant adhesion whereas α2β1 can arrest the cells also under conditions of low integrin activation [11]. This may explain why mice lacking both GPVI/FcRγ-chain and α2β1 display severely defective hemostasis while single deficiency in either receptor has no such effect. Besides mediating adhesion and aggregation, platelet integrins contribute to cellular activation and trigger a variety of important functions-like spreading, procoagulant activity, and clot retraction through ‘outside-in’ signaling. This is best documented for αIIbβ3 [42], but α2β1 appears to have similar functions. Interestingly, these integrins regulate a similar set of intracellular signaling molecules as GPVI including Syk, SLP-76, and phospholipase C (PLC)γ2. Mice lacking these signaling molecules in platelets display significant bleeding defects [43–46] whereas deficiency in GPVI or α2β1 has no such effect.

Platelet activation

After initial adhesion of platelets to the ECM, extension of the thrombus requires a rapid response of platelets to locally produced/released soluble agonists, including ADP, thrombin, epinephrine, and Thromboxane A2 (TXA2), which amplify and sustain the initial platelet responses and recruit circulating platelets from the flowing blood into a growing hemostatic plug [47]. Most of the agonists involved activate platelets through receptors that couple to heterotrimeric G-proteins. The essential aim throughout is activation of integrin αIIbβ3, which requires activation of phospholipases, increase in intracellular calcium, suppression of cAMP synthesis, and reorganization of the platelet cytoskeleton. Knockout technology has created potential instruments to study the relevance of single effectors as well as the interplay between several partners in order to reach a better understanding of how initial responses of platelets are finally converted into stable thrombus formation.

One well described positive feedback mediator in this process is ADP. It is released from internal stores of activated cells and potentiates many platelet responses, including integrin αIIbβ3 activation, dense granule secretion, and platelet procoagulant activity. Platelets express at least two different ADP receptors, P2Y1 and P2Y12, which couple to Gq and Gi, respectively. Mouse platelets deficient in the P2Y1 receptor fail to aggregate in response to usual concentrations of ADP and aggregation induced by other agonists is impaired confirming an amplificatory role of P2Y1 in these processes [48,49]. In vivo, P2Y1-null mice have moderately increased bleeding times and are resistant against collagen/epinephrine or ADP-induced thromboembolism, whereas transgenic mice overexpressing P2Y1 in their platelets have a shortened bleeding time and are more susceptible to ADP- and collagen-induced thromboembolism and to arterial thrombosis triggered by FeCl3 [50]. In contrast, P2Y12-null platelets do not aggregate normally in response to ADP and lack the ability to inhibit cAMP formation, but P2Y1-associated responses such as shape change and PLC activation are retained [51]. These animals have a prolonged bleeding time and are protected from arterial thrombosis as measured in the FeCl3 model [52]. In detail, thrombus formation was delayed and only small and unstable thrombi were formed, which might be explained by the fact that platelet activation in response to different agonists was clearly impaired. In addition to inhibiting cAMP formation, the P2Y12 receptor activates phosphoinositide 3-kinase (PI3K) isoforms [53]. In vivo, PI3K γ-null mice are protected from ADP-induced thromboembolism but have normal bleeding times [53]. Together with previous in vitro findings using P2Y12 inhibitors [54], the pronounced phenotype of these animals, especially compared with that of P2Y1 knockout mice, identifies the P2Y12 receptor as the major receptor to amplify and sustain platelet activation.

Thrombin is rapidly generated at sites of vascular injury and, besides mediating fibrin generation, represents the most potent platelet activator. Platelet activation by thrombin induces shape change, secretion, and aggregation. No other platelet agonist appears to be as efficiently coupled to PLCβ as thrombin, resulting in a very rapid and efficient increase of cytosolic calcium. Platelet responses to thrombin are largely mediated by members of the PAR family, with PAR-1 and -4 being expressed in human platelets and PAR-3 and -4 in mouse platelets [55]. Despite this important difference between the two species, much knowledge about the significance of PARs in thrombosis and hemostasis is derived from studies with knockout mice. First evidence for so far unknown platelet PAR(s) came from studies with PAR-1-deficient mice as platelet responses to thrombin were unaltered [56]. This led to the identification of PAR-3 in mice [57,58], which was shown to mediate effects of low thrombin concentrations [59–62]. Accordingly, absence of PAR-3 caused a markedly delayed and reduced, but not absent response to thrombin [60] and a protection against FeCl3-induced thrombosis of mesenteric arteries [63]. Interestingly, PAR-4-deficient platelets showed a similar degree of protection in comparable in vivo studies [63], but these platelets were completely resistant to thrombin [64]. Although in murine platelets PAR-3 serves solely to facilitate cleavage of PAR-4 by thrombin, but does not participate in the signaling process [62], PAR-3 and PAR-4-null mice display a similarly prolonged bleeding time suggesting that fully intact thrombin responses are required for normal hemostasis. In contrast, PAR-1 is the main receptor in humans that mediates the activation of platelets by thrombin together with PAR-4 [65]. Therefore, although there are major similarities between the two species, the transfer of knowledge from the mouse model to the human situation is limited in the case of PAR function in platelets. For a detailed review on PARs, the reader is also referred to the state-of-the-art lecture presented by Coughlin [55] elsewhere in this book.

Thromboxane A2 is produced by platelets from arachidonic acid via the cyclo-oxygenase pathway. Once formed, it can diffuse across the platelet membrane and activate other platelets. In platelets, the TXA2 receptor exists in two splice variants, TPα and TPβ that differ in their cytoplasmatic tail. TPα/β couple to Gq and G12/13 and their activation triggers shape change, secretion, hydrolysis of phosphoinositides, increment in cytosolic calcium, and protein phosphorylation, but does not influence cAMP synthesis. TP-null mice have prolonged bleeding times and do not respond to TXA2 [66]. Furthermore, these platelets display delayed aggregation in response to collagen, confirming the role of TXA2 as a ‘second wave’ mediator in this process.

In contrast to the above-mentioned agonists, epinephrine and ATP are not full platelet activators per se. Mice lacking the fast ATP-gated P2X1 cation-channel are partially protected from collagen-/epinephrine-induced thromboembolism and the size of thrombi after a laser-induced vessel wall injury in small arteries was decreased. This might be explained by impaired aggregation and secretion of these platelets in response to collagen and by a reduced ability to adhere to the matrix protein under high shear conditions [67]. Epinephrine is able to potentiate the effect of other platelet stimuli by activating the Gz-coupled adrenergic α2A receptor resulting in the inhibition of adenylyl cyclase. In Gαz-deficient platelets, the inhibitory effects of epinephrine on adenylyl cyclase as well as the potentiating effects of epinephrine are impaired, while responses of other platelet activators are not affected. In vivo, Gαz-deficient mice show resistance to lethal thromboembolism induced by collagen/epinephrine, but not collagen/ADP [68]. No such findings were reported from mice lacking Gαi2 and Gαi3, indicating that α2A-adrenergic receptors are coupled to Gαz to mediate inhibition of adenylyl cyclase [68]. Recent data show that mice lacking the α2A receptor have a moderate hemostatic defect and display enhanced embolus formation in the FeCl3 model clearly suggesting a role for epinephrine in thrombus stabilization (B. Nieswandt, unpublished observation).

The generation of agonist-binding receptor knockout mice confirmed many in vitro findings and clearly defined the relative contribution of each ligand in hemostasis and thrombosis. Beyond this, mice deficient in intracellular signaling molecules, such as G-proteins, helped to understand platelet physiology and identified a defined crosstalk of signaling pathways. Mice platelets that lack Gαq fail to aggregate and secrete in response to thrombin, ADP, and TXA2 because of a lack of agonist-induced PLCβ-activation. These severe defects in platelet signaling are best-mirrored in vivo where Gαq-deficient mice display massively prolonged bleeding times and protection from collagen-/epinephrine-induced thromboembolism [69] and arterial thrombosis in different models (B. Nieswandt, unpublished observations). Studies in Gαq-deficient platelets provided the first proof for a role of G12/13 in platelet activation. Surprisingly, TXA2, which mediates its effects solely through G12/G13 and Gq, could still induce an activation of the small GTPase RhoA as well as platelet shape change [70]. These findings were further substantiated in G12 and G13 mutant mouse lines. In Gα13-, but not Gα12-deficient platelets, the rapid platelet shape change induced by low concentrations of stimuli-like TXA2, thrombin, or collagen is completely abolished. Additionally, the potency of platelet activators to induce degranulation and aggregation is clearly reduced, accompanied by defective activation of RhoA and an inability to form stable platelet thrombi on a collagen substrate under conditions of high shear [71]. This is supported by findings showing that RhoA-mediated signaling appears to be required for platelet aggregation under high shear conditions as well as for the irreversible aggregation of platelets in suspension [72,73]. Taken together, Gα13 is essentially required for full platelet activation, whereas Gα12 appears to be dispensable for this process. According to these in vitro data, mice with platelets deficient in Gα13 showed markedly increased bleeding times and failed to form stable occlusive thrombi in the injured carotid artery [71].

Platelets deficient in Gα13 and Gαq do not respond to TXA2, ADP, or thrombin and do not aggregate in response to collagen. However, primary adhesion to collagen occurs normally in those mice [74] suggesting that adhesion occurs largely independent of G-protein-mediated signaling while thrombus growth requires activation of at least two different G-proteins. This hypothesis is supported by findings demonstrating that concomitant stimulation of Gi and G12/13 [75,76], or Gi and Gq [77,78] are sufficient to induce platelet aggregation.

Platelet activation is a dynamic process with different receptors and signaling pathways being dominant in different phases. It has to be kept in mind that ADP, TXA2, and thrombin do not act on their own, but stimulation of platelets with either agonist results in the release of another one. For example, mice deficient in Gαi2 show a decreased ADP-induced platelet aggregation, but effects of thrombin and TXA2 are also impaired although these agonists act primarily through Gq and G13 [79,80]. Obviously, they require in part the action of released ADP to gain full platelet activation.

The final common pathway of the different signaling pathways in platelets is the activation of integrin αIIbβ3, the principal receptor for adhesion and aggregation. Major advances have been made in recent years in understanding the complex mechanisms how αIIbβ3 affinity is regulated downstream of activatory receptors in platelets. In many cases, genetically modified mice were instrumental in the identification of new proteins involved in this process and or/assessment of their in vivo significance [81,82] but discussion of these developments is beyond the scope of this review. This field has been excellently summarized in a recent review by Shattil and Newman [83].


Platelet aggregation is essential for the formation of the hemostatic plug at sites of vascular injury. This process is primarily mediated by αIIbβ3, which, once it has been activated, immobilizes a number of different adhesive substrates to the membrane of the activated platelet. The immobilization step is a prerequisite for both thrombus growth and stabilization. Accordingly, mice lacking the β3 integrin resemble the phenotype of Glanzmann thrombasthenia with absent platelet aggregation, reduced clot retraction and greatly reduced fibrinogen uptake into platelets [84]. These mice have markedly prolonged tail-bleeding times and display spontaneous hemorrhage in all developmental stages. In intravital microscopy studies performed in mesenteric arterioles, β3-null mice do not form any thrombi [85]. However, αIIbβ3 is not simply an anchorage for adhesive substrates, but signals through the membrane once ligands have bound. The contribution of this outside-in signaling process to the adhesive strengthening and to irreversible platelet aggregation has been confirmed in mice expressing a mutated cytoplasmic tyrosine motif in the integrin β3 chain, as these mice rebleed from tail wounds [86]. Further details of αIIbβ3 signaling with a focus on the Src-Syk cascade are also discussed in the review by S. P. Watson elsewhere in this issue [87].

Several adhesive substrates are known to bind to αIIbβ3 in order to bridge and further activate platelets in a growing thrombus, in particular fibrinogen, VWF, and fibronectin. Fibrinogen is a dual player in the process of hemostasis in that it functions as a ligand of αIIbβ3, but also as the primary building block for the formation of fibrin following thrombin cleavage. The availability of gene-targeted mice either lacking fibrinogen [88] or expressing modified forms of fibrinogen [89] has provided the possibility to explore the roles of fibrin(ogen) in vivo. Comparable with the human phenotype of congenital afibrinogenemia, fibrinogen-null mice develop overt bleeding after birth and do neither show platelet aggregation nor blood clotting in vitro [88]. Mice expressing a truncated form of fibrinogen, FgγΔ5, which is unable to bind to αIIbβ3, but can still be converted into fibrin, display a generally normal hematologic profile, but have an extended bleeding time following surgical challenge [89]. Notably, intravital microscopy studies in FeCl3 injured mesenteric arterioles revealed that both gene-targeted mice are still able to form thrombi [85,90]. However, these thrombi were unstable and embolized frequently in both mouse models: in fibrinogen-null mice by being stripped from the interface of thrombi and the vessel wall, in FgγΔ5-mice by rupture through the central or upper portion of the thrombi. These data impressively support the idea that fibrin formation is essential for anchoring thrombi to the injured vessel wall, whereas fibrinogen itself plays an important role in thrombus stability by bridging αIIbβ3 between neighbored platelets. However, fibrinogen does not have an exclusive role in platelet aggregation as VWF has been shown to be necessary to help establish inter-platelet bridges at high shear rates in vitro [91]. Intravital microscopy in VWF-null mice was able to prove this idea: although stable thrombi did still form under conditions of arterial flow in the absence of VWF, an open channel remained within the thrombus in most animals [85]. It can be concluded that VWF is necessary for further thrombus growth once local shear rates become high, i.e. if the diameter of the vessel lumen is decreased below a critical value. However, the fibrinogen-null mouse has shown that VWF alone is not sufficient to achieve stable platelet aggregation, supporting the hypothesis that concurrent binding of VWF to αIIbβ3 and GPIbα allows initial inter-platelet contacts, while fibrinogen is necessary for a permanent linkage between activated αIIbβ3 on neighbored platelets to finally ensure stable aggregate formation. Paradoxically, double knockout-mice lacking both fibrinogen and VWF can still form thrombi [85], but these thrombi are very fragile and frequently release small emboli. Interestingly, the platelet content of fibronectin is increased threefold in these mice, making it an attractive alternative candidate for platelet cohesion. A conditional knockout mouse with reduced plasma fibronectin levels (<2% compared with wild type) displayed normal tail-bleeding time and in vitro parameters of coagulation [92]. However, intravital microscopy revealed that fibronectin deficiency delays thrombus formation in arterioles by a continuous shedding of small aggregates from stably anchored thrombi [28]. No vessels occluded at the site of injury, demonstrating that fibronectin is an important and formerly under-recognized mediator of platelet/platelet interaction that fosters the continued growth and stability in a forming thrombus.

The role of two other soluble adhesive proteins that bind to αIIbβ3, vitronectin, and thrombospondin, is not yet well understood. Whereas vitronectin-null mice where initially found to have an enhanced rate of thrombus formation suggesting an antithrombotic effect of vitronectin [93], subsequent experiments demonstrated thrombus instability as the predominant finding [94]. In addition, shortened times to thrombosis could not be confirmed in the same thrombosis model, leaving the question about its relevance open for the moment. Thrombospondin-null mice seem not to display a hemostatic phenotype [28].

Maximal morphologic, secretory, and procoagulant responses do not only require binding of the above-mentioned adhesive substrates, but also the formation of a hemostatic synapse between neighbored platelets to enable ‘contact-dependent signaling’. Occupancy of αIIbβ3 by one of its ligands and subsequent outside-in signaling is a well-understood part of this synapse [95]; only recently, CD40L (CD154), a member of the TNF family, was identified as a new ligand of αIIbβ3. Mice lacking CD40L can still develop large thrombi, but unexpectedly these thrombi frequently rupture and embolize [96], suggesting that outside-in signaling via CD40L/αIIbβ3 [97] interaction is critical for thrombus stabilization under conditions of arterial flow. Other ligand–receptor pairs have also been implicated in contact-dependent signaling, including ephrin/Eph receptor kinases and Gas6/Axl-Tyro3-Mer receptor kinases [98,99]. Gas6, a member of the vitamin K-dependent protein family, is stored in α-granules and becomes secreted following platelet activation. It binds to Gas6 receptors on the platelet surface, suggesting an autocrine stimulatory mechanism. Gas6-null mice did not suffer spontaneous bleeding and had normal tail-bleeding times, but formed significantly smaller thrombi under both arterial and venous conditions of flow [99]. In addition, Gas6-null mice were protected from platelet-dependent thromboembolism. Comparable, but less pronounced results were found in mice lacking the Gas6-receptor Mer when subjected to models of FeCl3-induced thrombosis and pulmonary thromboembolism [100]. Only recently, Angelillo-Scherrer et al. [101] were able to show that each of the Gas6 receptors (Tyro3, Axl, and Mer) is important in transmitting a plug-stabilizing effect of Gas6, probably because of Gas6 receptor crosstalk which affects outside-in signaling via the αIIbβ3 integrin: deficiency of any of the receptors protected mice against thrombosis. There is evidence that Gas6/Gas6 receptors represent a mechanism that amplifies the response of other agonists while not evoking any response by itself, and further amplifier systems can be expected to fill the hemostatic synapse in the future.

Thrombus formation is inseparably connected to the initiation of the coagulation cascade to finally produce fibrin. This was elegantly demonstrated in mice expressing low levels of tissue factor (TF), which form only very small thrombi that do not incorporate fibrin [102]. Subjecting bone marrow chimera to a model of laser injury in cremaster muscle arterioles, these authors provided evidence that blood-borne TF contributes to the formation of fibrin during thrombus formation, whereas vessel wall TF alone was unable to sustain thrombin generation and fibrin formation throughout the thrombus. There is additional in vivo evidence that blood-borne TF comes from leukocyte-derived microparticles [103]. These microparticles do express PSGL-1 and are recruited into the thrombus via PSGL-1/platelet P-selectin interaction: neither P-selectin- nor PSGL-1-null mice are able to incorporate sufficient amounts of TF into the thrombus. These results are in line with observations showing that elevation of procoagulant microparticles (a subset of which contained TF) restored hemostasis in hemophilic mice in a P-selectin-/PSGL-1-dependent manner [84]. However, contradictory results were obtained using models of carotid artery and vena cava ligation [104]. Analyzing bone marrow chimera, low blood-borne TF did not limit thrombus formation in a wild-type mouse, and high blood-borne TF did not augment thrombus formation in a low TF animal. These authors conclude that vessel wall-derived TF plays a pivotal role in thrombosis. This discrepancy might be explained by the different nature and severity of the vessel wall injury as well as the different vascular beds used in the models; further studies will be required to evaluate the relative contribution of blood-borne vs. vessel wall TF in thrombosis.

Platelets and atherosclerosis

Atherosclerosis is a multifactorial and complex inflammatory disease of the vessel wall with deposition of lipids and local infiltration of leukocytes considered of greatest relevance for its pathogenesis. Nevertheless, there is growing evidence that platelets are not simply key players in thromboembolic complications of advanced atherosclerotic lesions, but may also contribute to the initiation and progression of the atherosclerotic plaque. Early data demonstrating the contribution of platelets came from experiments performed on mice with somewhat vaguely defined genetic defects and abnormal dense granules; some of these mutants showed reduced lesion formation after consumption of an atherogenic diet, indicating that components of platelets may affect atherosclerosis [105]. In 2002, Massberg et al. [10] were the first to demonstrate that platelets adhere to the vascular endothelium of carotid arteries in apoE-deficient mice before manifest atherosclerotic lesions develop. An antibody against GPIbα was able to limit atherosclerotic lesion formation and adhesion of leukocytes, indicating that platelet adhesion to endothelium precedes further events in the atherosclerotic process. GPIbα/VWF interaction appears to be one molecular mechanism that enables platelets to roll on endothelial cells; in accordance with this, VWF-/apoE-double-deficient mice are protected from atherosclerosis at branch points of arteries, which are known to be highly prone to lesion development in apoE-single-deficient mice [106]. Besides GPIbα/VWF, which probably has a restricted relevance in areas of high shear, P-selectin has gained attention as a relevant molecule in the formation of atherosclerotic lesions. The absence of P-selectin has been shown to inhibit leukocyte infiltration of atherosclerotic lesions in a mouse model, a process initially thought to be driven by the absence of P-selectin from the endothelium [107]. However, Huo et al. [108] were able to demonstrate that activated platelets exacerbate atherosclerosis in apoE-null mice in a P-selectin-dependent manner. Wild type, but not P-selectin-null platelets, interacted with atherosclerotic arteries of apoE-deficient mice, induced the expression of RANTES and PF-4 on the carotid arterial endothelium, and led to substantial leukocyte adhesion. The contribution of endothelial vs. platelet P-selectin in the process of atherosclerosis was further dissected using a bone marrow chimera model where apoE-null, P-selectin-null mice and apoE-null, P-selectin-expressing mice were transplanted with bone marrow of either genotype [109]. Endothelial P-selectin was found to be crucial for the promotion of atherosclerotic lesions, but platelet P-selectin did actively promote lesion development as well. Reduced platelet P-selectin expression might also partially explain the recent findings by Kobayashi et al. [110] that apoE-null mice lacking the gene for the platelet activator TXA2 display a 70% reduction in plaque formation when compared with apoE-null mice, as effective platelet activation is a prerequisite for P-selectin expression. Thus, current data do now support the idea that platelet P-selectin is an important player in the process of lesion formation, probably whenever it becomes delivered to leukocytes, be it as platelet–leukocyte aggregate or as content in leukocyte-coating microparticles, augmenting the infiltration of leukocytes into the atherosclerotic lesion.

It has to be mentioned that not all ligand–receptor pairs known to be relevant for thrombus formation are necessarily important in the initiation and progression of atherosclerosis. ApoE-null mice lacking the collagen receptor α2β1 are not protected from atherosclerosis [111]. Integrin β3-null mice, lacking αIIbβ3, developed even more pronounced atherosclerosis with either an apoE-null or low density lipoprotein receptor (LDLR)-null background [112]. The observation has been explained by an up-regulation of proinflammatory molecules, including CD40 and CD40L, in those animals. Interestingly, CD40L-deficient mice with an apoE-null background have reduced atherosclerotic plaque areas, although initial lesion development remains unaffected [113]. However, the particular contribution of platelet-derived CD40L has not been studied in this model.

Data obtained so far support the idea that platelets are indeed relevant co-players in the complex processes that lead to the formation of atherosclerotic plaques. Future studies will add to the yet short list of relevant molecules and hopefully aid evaluation of new antiatherogenic targets.


Experimental animal models have substantially contributed to a better understanding of the mechanisms underlying hemostasis and thrombosis over the last years. Despite the unquestionable contribution of the knockout technology, creative and novel approaches in imaging and image analysis were equally important to meet the challenges of analyzing the complex mechanisms underlying platelet-dependent thrombotic and inflammatory processes in vivo. Both approaches together were not only helpful in finally proving long-standing hypotheses, but have rather been used to discover new and previously unexpected mechanisms and causal relationships. At present, the assignability of results obtained in such models to humans may in some cases remain limited. However, we believe that this is primarily because of the lack of comparative data from patients on which such a judgment could be based. We are convinced that parallels between data obtained in experimental mouse models and clinical settings will materialize over time. Thus, although more work needs to be invested to reach a better standardization of in vitro and in vivo assays on platelet function in order to assure better comparability of the data, murine models have gained a key position in hemostasis and thrombosis research.