Integrin α2-Deficient Mice Develop Normally, Are Fertile, but Display Partially Defective Platelet Interaction with Collagen*

The integrin α2-subunit was ablated in mice by targeted deletion of the ITGA2 gene. α2-Deficient animals develop normally, are fertile, and reproduce. Surprisingly, no obvious anatomical or histological differences were observed in mutant mice. Besides its significance in tissue morphogenesis, integrin α2β1 has been reported to play a major role in hemostasis by mediating platelet adhesion and activation on subendothelial collagen. To define its role in hemostasis, α2-deficient platelets were analyzed for their capacity to adhere to and aggregate in response to fibrillar or soluble collagen type I. We show that aggregation of α2-deficient platelets to fibrillar collagen is delayed but not reduced, whereas aggregation to enzymatically digested soluble collagen is abolished. Furthermore, α2-deficient platelets normally adhere to fibrillar collagen. However, in the presence of an antibody against GPVI (activating platelet collagen receptor), adhesion of α2-deficient but not wild type platelets is abrogated. These results demonstrate that integrin α2β1 significantly contributes to platelet adhesion to (fibrillar) collagen, which is further confirmed by the abolished adhesion of α2-deficient platelets to soluble collagen. Thus, α2β1 plays a supportive rather than an essential role in platelet-collagen interactions. These results are in agreement with the observation that α2β1-deficient animals suffer no bleeding anomalies.

Integrins are a large family of heterodimeric transmembrane receptors composed of noncovalently associated ␣and ␤-subunits that function as receptors for extracellular matrix components and also bind to counter receptors on other cells (1)(2)(3). Integrin receptors modulate critical cellular processes, including adhesion and spreading, migration, survival, gene expression, and differentiation. These processes are physiologically relevant to growth and development, angiogenesis, and hemostasis but may also be significant in pathological conditions such as tumor metastasis and thrombosis (4 -6).
␣ 2 ␤ 1 integrin (VLA-2, platelet GPIaIIa) was thought to play a pivotal role in development, differentiation, and tissue morphogenesis. It is widely expressed, especially on cell types entering the final stages of differentiation (13,14). ␣ 2 ␤ 1 receptors bind with high affinity to collagen I (15) and also to collagens II-V (16,17), and they mediate adhesion to laminins-1 and -5 (18,19). Contact with collagen of ␣ 2 ␤ 1 on fibroblasts and epithelial cells induces synthesis and activation of several matrix metalloproteinases (20 -22) and is therefore thought to play an essential role in connective tissue remodeling and resurfacing of wounds.
The ␣ 2 protein was initially isolated from platelets, where it is involved in the adhesion to subendothelial collagen at sites of vascular injury and thereby contributes to the formation of a hemostatic plug (23). However, the interaction between platelets and collagen is complex and can either occur indirectly via immobilized von Willebrand factor binding to platelet receptors glycoprotein (GP) 1 Ib-V-IX and/or activated ␣IIb␤ 3 integrin (24) or by direct interaction of collagen with specific receptors, including the Ig-like receptor GPVI (25,26) and ␣ 2 ␤ 1 integrin. GPVI is essential for this process as it mediates the activation of ␤ 1 and ␤ 3 integrins, which is a prerequisite for firm adhesion and thrombus growth (27). In contrast, the role of ␣ 2 ␤ 1 in both platelet adhesion and activation on collagen has been controversially debated. Although previous studies (28,29) emphasized an essential role of this integrin in platelet-collagen interactions and hemostasis, we have recently shown that mice lacking ␤ 1 integrins on their platelets display no major hemostatic defect. In vitro, however, ␤ 1 -deficient platelets failed to interact with enzymatically digested collagen and displayed partial defects in their activation by and adhesion to native fibrillar collagen (27). In these studies, however, it could not be clarified definitively whether these defects were based on the absence of ␣ 2 ␤ 1 alone, because ␤ 1 -deficient platelets also lack ␣ 5 ␤ 1 and ␣ 6 ␤ 1 .
To assess the function of the ␣ 2 ␤ 1 receptor in vivo, particularly in hemostasis, we generated ␣ 2 integrin-deficient mice. In contrast to previous reports that suggested that homozygous deletion of the ITGA2 gene results in embryonic lethality (5,30,31), these mice develop normally and reproduce. Strikingly, platelet counts and bleeding times are normal in ␣ 2 integrindeficient mice. Although the interaction of ␣ 2 -deficient platelets with soluble collagen is abrogated, they display only subtle defects in response to native fibrillar collagen.

EXPERIMENTAL PROCEDURES
Chemicals and Antibodies-Fibrillar type I collagen from equine tendon (Horm, Nycomed, Munich, Germany), high molecular weight heparin, and soluble, non-fibrillar type I collagen from rat tail were from Sigma.
Production of ITGA2 (Ϫ/Ϫ) Mice-A 439-bp cDNA fragment of mouse ␣ 2 integrin corresponding to exons 1-3 was used to screen a FIX II genomic library (Stratagene) of the 129SVJ mouse strain. A targeting vector was generated in pBluescript KS II (Stratagene) containing 3 kb of the promoter region, the first exon including the translation start, and 4 kb of the first intron. A HindIII and loxP site was placed 1 kb upstream of the first exon in a single StuI site, and a phosphoglycerate kinase-driven neomycin resistance (neo r ) cassette, flanked by loxP sites, was inserted 0.4 kb downstream of the exon into a single SacII site.
Embryonic stem (ES) cells of the E14 subclone IB-10 were grown under standard conditions. ES cells were electroporated (Bio-Rad Gene Pulser II) with the SmaI-linearized targeting vector. G418-resistant clones, cut with HindIII, were analyzed by RFLP analysis using the external probe Ex for hybridization. Homologous recombination reduced the 19-kb wild type fragment to 7.5 kb, indicating cointegration of the single loxP site. Single copy integration was confirmed by reprobing with a neo r probe (Int), resulting in a single 7.5-kb band. Deletion of the neo r cassette was achieved by electroporating correctly targeted clones with the plasmid pIC Cre expressing Cre recombinase. Loss of the neo r cassette reduced the 7.5-kb mutated fragment to 6 kb. Three individual clones then generated germ line transmitting chimeras as described (36).
Mice homozygous for the loxP-flanked exon were bred to mice expressing Cre recombinase in the zygote, 2 giving rise to offspring with a heterozygous deletion of the first exon of the ITGA2 gene. ITGA2 (ϩ/Ϫ) mice were intercrossed to generate mice with a null mutation of the ␣ 2 integrin.
Analysis of Integrin Expression-Platelets and tissue samples from ITGA2 (Ϫ/Ϫ) and wild type mice were homogenized in standard lysis buffer, and proteins were separated by SDS-PAGE under reducing conditions and transferred to nitrocellulose membranes. Following blocking, membranes were incubated with polyclonal rabbit antibodies against mouse ␣ 2 integrin. Bound antibodies were detected by horseradish peroxidase-conjugated anti-rabbit IgG and ECL (Amersham Biosciences).
Determination of Weight and Bleeding Times-Individual animals were weighed on postnatal day 23. For determination of bleeding times, mice were anesthetized, and 5 mm of tail tip was amputated with a scalpel. Tails were then blotted with filter paper every 15 s until the paper was no longer bloodstained (37).
Platelet Preparation and Counting-Mice were bled under ether anesthesia from the retro-orbital plexus. Blood was collected in a tube containing 10% (v/v) 7.5 units/ml heparin, and platelet-rich plasma (PRP) was obtained by centrifugation at 300 ϫ g for 10 min at room temperature. Before the start of the experiments, platelets were allowed to rest for 30 min in the presence of 0.02 units/ml of the ADP scavenger apyrase (adenosine 5Ј-triphosphate diphosphohydrolase), a concentration sufficient to prevent desensitization of platelet ADP receptors during storage. Platelets were kept at 37°C throughout all experiments. For determination of platelet counts, blood (20 l) was diluted 1:100 in Unopette kits (Becton Dickinson), and samples were allowed to settle for 20 min in an Improved Neubauer hemocytometer (Superior, Bad Mergentheim, Germany). Platelets were counted under a phase contrast microscope at ϫ400 magnification.
Aggregometry-To determine platelet aggregation, light transmission was measured using PRP adjusted to a platelet concentration of 3 ϫ 10 5 platelets/l with Tyrode's buffer (detailed above) containing 0.35% bovine serum albumin. Transmission was recorded on a Fibrintimer 4-channel aggregometer (APACT Laborgerä te und Analysensysteme, Hamburg, Germany) over 10 min and expressed as arbitrary units relative to 100% transmission (adjusted with plasma).
Platelet Adhesion-Fibrillar or soluble collagen (2 g in 100 l PBS per well) was immobilized on F96-MaxiSorp plates (Nunc, Wiesbaden, Germany) at 4°C overnight. The plates were then blocked with 1 mg/ml bovine serum albumin in PBS for 3 h at 37°C and washed with PBS. Washed platelets (10 7 /well) in Tyrode's/albumin buffer containing 1 mM CaCl 2 and 1 mM MgCl 2 were incubated in the wells for up to 60 min. After three washing steps, adherent platelets were quantitated fluorimetrically as described (35).

RESULTS AND DISCUSSION
Generation of Integrin ␣ 2 -Deficient Mice-A targeting vector was constructed in which the first exon of the ITGA2 gene along with the translation start were flanked by loxP sites to generate a mouse line that enables conditional inactivation of the ITGA2 gene (Fig. 1A). The vector was used to produce ES cell clones with a single homologous recombination event, as confirmed by Southern blot analysis (Fig. 1B). Deletion of the neo r cassette was achieved by transiently transfecting these ES cell clones with a vector expressing Cre recombinase. Southern blot analysis of resulting G418-sensitive clones revealed 7 clones lacking the neo r cassette and harboring the loxP-flanked first exon (Fig. 1B, Flox).
Three individual clones were used to generate germ line chimeras. Germ line transmission in progeny of the chimeras was confirmed by Southern blot analysis of HindIII-digested tail DNA, and mice heterozygous for the mutation in the ITGA2 gene were intercrossed to produce mice homozygous for the loxP-flanked first exon (Fig. 1C, fl/fl). These animals (ITGA2flox) appeared normal, and Western blot analysis of several mouse organs confirmed that insertion of the loxP sites did not interfere with integrin ␣ 2 expression (Fig. 1D).
To produce mice with a heterozygous ablation of the ITGA2 gene, ITGA2flox animals were bred to mice expressing Cre recombinase in the zygote, 2 leading to deletion of loxP-flanked regions. Deletion of the first exon was confirmed by Southern blot analysis, and heterozygous animals were intercrossed to produce ␣ 2 (Ϫ/Ϫ) mice (Fig. 1E).
Integrin ␣ 2 -Deficient Mice Develop Normally, Are Fertile, and Display No Obvious Anatomical Defects-Homozygous ␣ 2 -deficient mice are viable and show no striking phenotypical difference when compared with their heterozygous and wild type littermates. Complete loss of the ␣ 2 -subunit was confirmed by Western blot analysis of proteins extracted from several mouse organs and from platelets (Fig. 1F).
Litter sizes from breedings of heterozygotes are comparable with those of wild type animals, and genotyping of the viable offspring revealed normal Mendelian ratios, demonstrating that the loss of the integrin ␣ 2 -subunit does not result in embryonic lethality. There was no significant difference in size or weight at birth nor at 3 weeks of postnatal life between ␣ 2 (Ϫ/Ϫ) mice (females, 13.1 Ϯ 0.7 g; males, 15.1 Ϯ 1.6 g) and wild type animals (females, 12.8 Ϯ 1.8 g; males, 15.3 Ϯ 1.7 g). Integrin ␣ 2 (Ϫ/Ϫ) mice are fertile, and intercrossing these mice produced normal litter sizes. Notably, the progeny of ␣ 2 (Ϫ/Ϫ) mice also developed normally, indicating that the ␣ 2 (Ϫ/Ϫ) females have no severe defects in placenta formation or lactation. No morphological or histological changes were obvious in these mice.
That mice lacking ␣ 2 ␤ 1 receptors were viable and fertile was surprising, given in vitro data suggesting that tissue morphogenesis could be impaired due to lack of proper adhesion, spreading, and migration. Further work will demonstrate whether subtle temporal alterations are present in these mice.
Functional compensation by other collagen or laminin receptors, e.g. ␣ 1 ␤ 1 or the discoidin domain receptors, may be an explanation for this subtle phenotype. Interestingly, ablation of both collagen receptors, ␣ 1 ␤ 1 (38) and ␣ 2 ␤ 1 , present mice with only subtle phenotypes, differing thereby from other integrindeficient mice, most of which display severe defects (39,40).
Normal Platelet Counts and Bleeding Time in ␣ 2 -Deficient Mice-While integrin ␣ 2 ␤ 1 has long been recognized as a platelet collagen receptor, its exact role in hemostasis has been controversial (41). To address this question, we analyzed platelets from ␣ 2 -deficient mice. First, peripheral platelet counts were determined to assess platelet production in mutant mice. As shown in Fig. 2A, platelet counts were similar in control, ␣ 2 (ϩ/Ϫ), and ␣ 2 (Ϫ/Ϫ) mice. Flow cytometric analysis confirmed the absence of integrin ␣ 2 -subunits on homozygous mutant platelets, whereas the expression levels in heterozygous platelets were reduced by ϳ50% when compared with wild type (Table I). Interestingly, the levels of integrin ␤ 1 were reduced by ϳ30% in ␣ 2 (Ϫ/Ϫ) and ϳ15% in ␣ 2 (ϩ/Ϫ) platelets as compared with controls, whereas the expression of ␣ 5 and ␣ 6 was significantly increased in mutant platelets (Table I). In contrast to ␤ 1 integrins, the expression levels of other membrane glycoproteins, such as integrin ␤ 3 , GPVI, or the GPIb-V-IX complex, were not altered in mutant platelets (Table I). Western analyses of platelet lysates confirmed the absence of ␣ 2 (Fig. 1F) and normal expression of GPVI (not shown) in platelets from ␣ 2 (Ϫ/Ϫ) mice. These findings demonstrate that integrin ␣ 2 is not essential for megakaryocyte development and platelet production. However, its absence significantly alters the expression levels of other subunits of the ␤ 1 integrin family, suggesting that different ␣-subunits compete for association with ␤ 1 in platelets.
Previous reports (28,29) showed markedly increased bleeding in patients with reduced expression of ␣ 2 ␤ 1 integrin on platelets, suggesting a pivotal role of the integrin in hemostasis. In contrast to this hypothesis, we have shown recently (27) that bone marrow-chimeric mice with ␤ 1 integrin-deficient platelets display no increased bleeding tendency. To test directly these contrasting findings in a defined system, bleeding times were determined in ␣ 2 (Ϫ/Ϫ) mice. Strikingly, bleeding times were found comparable for ␣ 2 (Ϫ/Ϫ) and control mice (Fig. 2B), demonstrating that the lack of ␣ 2 ␤ 1 integrin on platelets, and also on other cells of the cardiovascular system, has no major effect on normal hemostasis in mice. This finding confirms and extends the observations made in ␤ 1 integrin mutant mice but stands in sharp contrast to the reported severe bleeding in patients with reduced ␣ 2 ␤ 1 levels on their platelets. The most likely explanation for this discrepancy is that these very few patients had additional defects in their platelets, although species-specific differences cannot be excluded. External probe (Ex) and internal probe (Int) used in Southern blot analysis are indicated. B, Southern blot analysis of HindIII-digested genomic DNA from ES cells. Homologous recombination resulted in a 21-kb wild type fragment and an additional 7.5-kb fragment specific for the mutated allele (HR) upon hybridization with probe Ex. Deletion of the neo cassette reduced the 7.5-kb fragment to 6 kb (Flox), and complete deletion of the neo cassette and the first exon reduced it to a 4-kb fragment (⌬). Hybridization with probe Int confirmed single integration of the targeting vector and deletion of the neo cassette. C, Southern blot analysis of HindIII-digested mouse tail DNA from ITGA2flox and wild type animals with probe Ex. A single 21-kb fragment was detected in wild type DNA (ϩ/ϩ), whereas in DNA from ITGA2 (ϩ/fl) mice the additional 6-kb fragment appeared. In ITGA2 (fl/fl) mice, only the 6-kb fragment was detected. D, Western blot analysis of organ lysates from wild type (ϩ/ϩ) and ITGA2 (fl/fl) animals. Introduction of loxP sites did not interfere with the expression of the 160-kDa integrin ␣ 2 -subunit in ITGA2flox animals. E, Southern blot analysis of HindIII-digested tail DNA from ITGA2⌬ mice with probe Ex. In ITGA2 (ϩ/ϩ) animals, a single 21-kb wild type fragment was detected, and in ITGA2 (ϩ/Ϫ) the additional 4-kb fragment appeared. Intercrossing of ITGA2 (ϩ/Ϫ) animals resulted in animals homozygous for integrin ␣ 2 gene ablation (Ϫ/Ϫ), illustrated by the presence of only the 4-kb fragment. F, Western blot analysis of platelet and organ lysates from wild type (ϩ/ϩ) and ITGA2 (Ϫ/Ϫ) mice. The integrin ␣ 2 -subunit was undetectable in protein extracts from (Ϫ/Ϫ) mice. Delayed Aggregation of ␣ 2 -Deficient Platelets in Response to Fibrillar Collagen-Several reports (28,29,42) suggested a central role of ␣ 2 ␤ 1 integrin during collagen-induced platelet aggregation. In contrast to these findings, the analysis of integrin ␤ 1 (Ϫ/Ϫ) platelets demonstrated a supportive rather than an essential role of ␤ 1 integrins in platelet-collagen interactions (27). To define unequivocally the role of ␣ 2 ␤ 1 integrin in this process, we induced aggregation of control, ␣ 2 (ϩ/Ϫ), and ␣ 2 (Ϫ/Ϫ) platelets using fibrillar type I collagen. Dose-response and maximum aggregation of mutant platelets did not differ from normal platelets (Fig. 3C). However, onset of aggregation was significantly delayed in ␣ 2 (Ϫ/Ϫ) platelets, and this was particularly evident at low collagen concentrations. Interestingly, no significant delay was observed in ␣ 2 (ϩ/Ϫ) platelets (Fig. 3B).
It is established that platelet activation by collagen strictly depends on functional GPVI (27,35,43). To test GPVI function, activation of control and mutant platelets was induced by the GPVI-specific agonist collagen-related peptide (CRP) (44). As shown in Fig. 3D, CRP-induced aggregation occurred with the same dose-response characteristics in control and mutant platelets. Further studies showed that ADP-and thrombininduced activation was not altered significantly in mutant platelets (not shown).
These results clearly demonstrate that integrin ␣ 2 is not essential for platelet activation by collagen, although the process is slightly delayed in the absence of the integrin. A similar delay in collagen-induced aggregation has been observed on human platelets in the presence of ␣ 2 ␤ 1 -blocking antibodies (45), in integrin ␤ 1 -deficient mouse platelets (27), or in mouse platelets lacking GPV (46). These defects most likely reflect a reduced stability of the initial platelet-collagen interaction due to the lack of collagen-binding sites on the cells.
Defective Activation of ␣ 2 -Deficient Platelets by Soluble Collagen-In vivo, secreted procollagen is proteolytically converted into collagen and assembled into insoluble, cross-striated fibrils (47). In vitro, collagen fibrils can be partly digested by pepsin, which cleaves the molecule in the non-triple helical region, thereby releasing "soluble" collagen (48). In numerous studies, such preparations of soluble collagen have been used to characterize the interaction of individual platelet receptors with collagen (48 -50). We have shown recently that aggregation in response to soluble collagen is abrogated in integrin ␤ 1 (Ϫ/Ϫ) platelets, but it remained unclear whether this defect was entirely based on the absence of ␣ 2 ␤ 1 or, possibly, of other ␤ 1 integrins (27). To test this directly, we induced aggregation of control and ␣ 2 (Ϫ/Ϫ) platelets with increasing concentrations of soluble collagen. No aggregation of ␣ 2 (Ϫ/Ϫ) platelets occurred at concentrations of up to 500 g/ml soluble collagen, whereas robust aggregation of control platelets was already seen at 5 g/ml (Fig. 4A). The critical role of ␣ 2 ␤ 1 integrin in this process was further confirmed by an ϳ5-fold right shift of the dose-response curve of the ␣ 2 (ϩ/Ϫ) platelets when compared with the control (Fig. 4B). Because platelet activation by   collagen strictly depends on GPVI (25,34,43), this finding suggests that integrin ␣ 2 ␤ 1 mediates the stabilization of GPVIcollagen interactions. This stabilization may be particularly important in the case of digested collagen which, in contrast to fibrillar collagen, lacks highly repetitive GPVI recognition sites.
Integrin ␣ 2 -Deficient Platelets Do Not Adhere to Soluble Collagen-Platelet attachment to collagen is a complex process involving the synergistic action of different receptors and signaling pathways (41). We have shown recently that GPVI is essential in this process as GPVI-deficient platelets do not adhere to fibrillar or soluble collagen. In contrast, platelets lacking ␤ 1 integrins display partial adhesion defects, but it was not clear whether these defects are due to the absence of ␣ 2 ␤ 1 alone, as integrins ␣ 5 ␤ 1 and ␣ 6 ␤ 1 are also not expressed on these platelets (27). To address this issue, adhesion of control, ␣ 2 (ϩ/Ϫ), and ␣ 2 (Ϫ/Ϫ) platelets to fibrillar as well as soluble collagen was tested in a static assay.
Fibrillar collagen induced comparable adhesion of mutant and control platelets in a time-dependent manner (Fig. 5A), confirming that ␣ 2 ␤ 1 is not essential for this process. However, blocking the major collagen-binding site on GPVI by JAQ1-Fab fragments (20 g/ml (51)) abolished adhesion of ␣ 2 (Ϫ/Ϫ) platelets, whereas control platelets adhered to the same extent. Adhesion of ␣ 2 (ϩ/Ϫ) platelets was markedly reduced and delayed in the presence of JAQ1 Fab fragments. These results demonstrate the existence of an ␣ 2 -dependent adhesion mechanism on fibrillar collagen that becomes dominant when the major collagen-binding site on GPVI is blocked (Fig. 5A).
A different picture emerged using soluble collagen, to which control platelets adhered in the absence, but not in the presence of JAQ1-Fab fragments, confirming the critical role of GPVI in this process. Strikingly, however, ␣ 2 (Ϫ/Ϫ) platelets did not adhere to soluble collagen in the presence or absence of JAQ1-Fab fragments, also demonstrating an essential role for ␣ 2 ␤ 1 integrin in this interaction. The importance of ␣ 2 ␤ 1 for adhesion to soluble collagen was further confirmed by the reduced and delayed adhesion of ␣ 2 (ϩ/Ϫ) platelets (Fig. 5B).
These results confirm the crucial role of GPVI in the inter-action of platelets with collagen and demonstrate that ␣ 2 ␤ 1 is essential for platelet adhesion to soluble collagen, whereas adhesion to fibrillar collagen is only dependent on the integrin when the major collagen-binding site on GPVI is blocked. Although in vivo collagens can be degraded in certain pathological situations, the majority of collagens is deposited in fibrillar form in normal vessel walls. Therefore, GPVI interaction with collagen at sites of vascular injury should not be dependent on integrin ␣ 2 ␤ 1 . Once the platelets are activated through GPVI, other adhesive receptors, most importantly integrin ␣IIb␤ 3 , can mediate firm attachment and thrombus growth (24,27).
In conclusion, we show that mice lacking ␣ 2 ␤ 1 integrin receptors develop normally, are fertile, and exhibit surprisingly subtle alterations. More sophisticated analyses will be required to illustrate whether other subtle defects are present and which other receptor(s) may compensate for loss of ␣ 2 ␤ 1 function. Possibly ␣ 2 ␤ 1 integrins are not essential for development but may be needed for tissue repair, host defense, or other challenges that the adult organism has to meet. The analysis of ␣ 2 -deficient platelets revealed a subtle rather than a major defect which is in line with recent studies on ␤ 1 -deficient platelets. The mice described here will allow detailed studies on the involvement of integrin ␣ 2 in thrombotic diseases where it has been proposed to play a major role (23).