Expression of the Platelet Receptor GPVI Confers Signaling via the Fc Receptor g -Chain in Response to the Snake Venom Convulxin but Not to Collagen*

The mechanism of signal transduction underlying the activation of platelets by collagen has been actively in-vestigated for over 30 years, but the receptors involved remain incompletely understood. Studies of human platelets, which are unresponsive to collagen, mouse knockout models, and platelet biochemical studies support the hypothesis that the recently cloned platelet surface protein GPVI functions as a signaling receptor for collagen. To directly test this hypothesis, we have expressed wild-type and mutant forms of GPVI in RBL-2H3 cells, which express the Fc e receptor g -chain (Fc R g ), the putative signaling co-receptor for GPVI in platelets, but lack GPVI itself. Expression of GPVI in RBL-2H3 cells confers strong adhesive and signaling responses to convulxin (a snake venom protein that directly binds GPVI) and weak responsiveness to collagen-related peptide but no responsiveness to collagen. To elucidate the mechanism of GPVI intracellular signaling, mutations were introduced in the receptor’s transmembrane domain and C-terminal tail. Unlike reported studies of other Fc R g partners, these studies reveal that both the GPVI transmembrane arginine and intracellular C-tail are necessary for coupling to Fc R g and for signal transduction. To our knowledge, these studies are the first to demonstrate a direct signaling role for GPVI and

Platelet activation is essential for both normal hemostasis and arterial thrombosis that occurs in the setting of vascular diseases such as stroke and myocardial infarction. One of the earliest steps in arterial thrombosis is the adhesion of circulating platelets to areas of injured vessel wall and the activation of adherent platelets, which recruits additional platelets to form a hemostatic plug. Activation of platelets at sites of vas-cular injury occurs in response to exposed subendothelial matrix proteins, the most important of which is collagen.
Exposed collagen initiates two essential platelet functions: the adhesion of circulating platelets to the site of injury and the activation of platelet signaling, which stimulates thrombus growth. Platelet adhesion to collagen has been shown to occur both indirectly, via interaction of platelet GPIb with plasma von Willebrand's factor bound to exposed collagen (1), and directly, via collagen interaction with the platelet integrin ␣ 2 ␤ 1 (2). In contrast, although the activation of platelets by collagen has been observed for over 30 years (3), the receptors and signaling pathways that mediate platelet activation by collagen are only beginning to be fully understood. Indirect evidence suggests that both ␣ 2 ␤ 1 and GPIb can initiate signaling when bound to collagen (1,4,5). However, this signaling does not appear to be sufficient to account for the magnitude of the platelet response to collagen.
GPVI is a recently cloned 62-kDa surface protein (6, 7) first identified by iodination of platelet surface glycoproteins (8). GPVI was proposed as a signaling receptor for collagen following the description of individuals with bleeding disorders whose platelets could not be activated by collagen and lacked GPVI despite having normal levels of the platelet integrin ␣ 2 ␤ 1 (9,10). Significant evidence suggests that GPVI signaling is sufficient to activate platelets and that GPVI may mediate collagen signaling in platelets. Platelets are activated by crosslinked anti-GPVI antibodies (11) and by convulxin (CVX), 1 a multimeric snake venom protein isolated from a South American rattlesnake which is capable of desensitizing platelets specifically to collagen and which binds specifically to GPVI (12,13). In addition, collagen signaling and GPVI signaling in platelets both employ the immunoreceptor signaling pathway (14) and require Fc R␥ (15). These data have led to a model of collagen activation of platelets in which adhesive roles are played by the integrin ␣ 2 ␤ 1 as well as GPIb and signaling roles are played by GPVI-Fc R␥, ␣ 2 ␤ 1 , as well as perhaps GPIb (14). This model has not been adequately tested, however, because of the absence of systems in which the contribution of each receptor can be studied in isolation.
The recent cloning of human and mouse GPVI (6, 7) reveals that GPVI is a type I transmembrane protein whose deduced amino acid sequence identifies it as an Ig domain-containing receptor homologous to the Fc and killer Ig-like receptors, some of which are known to signal via Fc R␥ (16). Consistent with its putative role as an Fc R␥ partner, GPVI has a charged arginine residue in its transmembrane domain that may mediate interaction with the Fc R␥ transmembrane domain in a manner analogous to that of the known Fc R␥ partners Fc␣ RI and PIR-A (16 -18). Direct functional evidence demonstrating that GPVI is a collagen receptor and that GPVI signaling is mediated by Fc R␥, however, are lacking. Transient expression of GPVI has been demonstrated to confer a slight calcium signal to collagen in the DAMI megakaryocytic cell line (6), but these cells express endogenous GPVI (data not shown and Ref. 6), ␣ 2 ␤ 1 (19), and perhaps other collagen receptors; it is therefore not clear if that response is mediated directly by GPVI or if GPVI expression is sufficient to enhance signaling by ␣ 2 ␤ 1 or other identified (e.g. p65 (20)) or unidentified collagen receptors.
To address the functional role of GPVI, we have stably expressed the receptor in RBL-2H3 cells, a rat basophilic leukemia cell line that expresses Fc R␥ and reproduces the platelet collagen responses of intracellular calcium mobilization and degranulation but does not express endogenous GPVI or ␣ 2 ␤ 1 . Our studies reveal that GPVI cross-linking by the GPVI-specific ligand convulxin initiates intracellular signaling but that GPVI alone is incapable of mediating a signaling response to collagen. A small signaling response is elicited by collagenrelated peptides (CRPs), however, and static adhesion studies support interaction with convulxin and CRP but not collagen. Finally, site-directed mutagenesis of the GPVI transmembrane domain and intracellular C-tail demonstrates that both the GPVI transmembrane arginine and the receptor C-tail are necessary for Fc R␥ interaction and intracellular signaling. These results provide insight into GPVI signal transduction and suggest that GPVI-signaling in response to collagen requires coreceptors for both ligand binding and intracellular signal transduction.

EXPERIMENTAL PROCEDURES
Materials-Type I collagen derived from equine tendons was obtained from Chronolog (Havertown, PA) and used for all studies shown. Studies were confirmed with type I collagen derived from bovine tendon and type III collagen derived from calf skin (Sigma). Convulxin was obtained from Sigma and purified from the venom of the Crotalus durissus rattlesnake using gel filtration as previously described (13). CRPs were synthesized as previously described using cross-linked cysteine residues (21). All other reagents were obtained from Sigma.
Cloning and Epitope Tagging of GPVI-A GPVI cDNA was generated by PCR from human platelet cDNA using primers based on published 5Ј-and 3Ј-untranslated sequences (sense strand primer: 5Ј-TCAGGA-CAGGGCTGAGGAACC-3Ј; antisense strand primer: 5Ј-TTGGATAC-GACCGTGCCTGGG-3Ј). Three distinct amplified 1.1-kilobase pair products were sequenced to obtain a consensus sequence that exhibited several differences from the published cDNA (6) but agreed with a cDNA sequence deposited directly in GenBank TM (accession number AB035073). All GPVI receptor amino acids reported here correspond to the protein predicted by the open reading frame of this cDNA starting at nucleotide number 13. FLAG-tagged GPVI was generated by replacing the endogenous signal peptide with that of interleukin-1 and placing the FLAG epitope (DYKDDDDK) in frame with GPVI at amino acid number 21, the predicted site of signal peptide cleavage (SignalP VI.I). HA-tagged GPVI was generated in an identical manner. Wild-type and epitope-tagged GPVI were expressed using the mammalian expression vector pcDNA3.0 (Invitrogen).
Site-directed Mutagenesis of GPVI-Site-directed mutagenesis was performed using the QuikChange mutagenesis kit (Stratagene). The oligonucleotide used for the R272L mutation was 5Ј-GCAACCTGGTC-CGGATATGCCTCGGGGCTGTG-3Ј. The oligonucleotide used for the R295STOP mutation was 5Ј-GGCAGAGGACTGGCACAGCTAGAG-GAAGCGCCTGC-3Ј. All mutants were made as epitope-tagged receptors as described above.
Platelet Aggregation Studies-Blood was collected into citrate buffer and platelet-rich plasma obtained as previously described (22). All studies were performed using platelet-rich plasma at a platelet density of 2 ϫ 10 8 platelets/ml. Creation and Screening of RBL-2H3 Cells Stably Expressing GPVI-RBL-2H3 cells (ATCC, Manassas, VA) were electroporated with linearized expression plasmids and selected in G418 (1.0 mg/ml active concentration; Life Technologies, Inc.) as previously described (23). Wildtype GPVI-expressing clones were directly tested for signaling in response to convulxin (below) and epitope-tagged clones tested for receptor expression by flow cytometry using M2 anti-FLAG antibody (Sigma) or anti-HA antibody (Sigma) as primary antibodies and fluorescein isothiocyanate-conjugated goat anti-mouse secondary antibody.
Intracellular Calcium Studies-Increases in cytoplasmic calcium in response to convulxin, CRP, collagen, and thrombin were measured using the calcium-sensitive dye Fura-2 as previously described at a final cell concentration of 2 ϫ 10 6 cells/ml (24). The buffer used for these studies was RPMI 1640 (Life Technologies) with HEPES 25 mM and 1 mg/ml BSA.
Functional Rescue of Surface FLAG-Fc R␥ in HEK-293T Cells to Measure GPVI-Fc R␥ Interaction-A stable HEK-293T cell line that stably expressed an N-terminally FLAG-tagged Fc R␥ in a manner identical to that previously described for DAP-12 (25) was a kind gift of Steve Spusta (University of California at San Francisco). These cells expressed FLAG-Fc R␥ at a level such that no cell surface FLAG-Fc R␥ was detectable by fluorescence-activated cell sorting despite readily detectable intracellular FLAG-Fc R␥ by Western blot analysis (data not shown). HA-tagged wild-type and mutated GPVI receptors were transiently expressed in FLAG-Fc R␥-expressing HEK-293T cells (Fugene6 transfection reagent; Roche Molecular Biochemicals). Surface expression of GPVI was followed with anti-HA antibody, and surface expression of Fc R␥ was followed with anti-FLAG antibody using flow cytometry as previously described for DAP-12 (25). Rescue of surface expression of FLAG-Fc R␥ was quantitated by comparing surface FLAG expression in GPVI-transfected versus mock-transfected 293T-FLAG-Fc R␥ cells as previously described (25).
Measurement of RBL Cell Adhesion-96-Well polystyrene high binding plates (Costar 3590; Corning Glass) were coated with 50 l/well of 20 g/ml BSA, convulxin, collagen, and fibronectin or 15 g/ml CRP in PBS with 0.9 mM calcium and 0.4 mM magnesium overnight at 4°C. Prior to application, the pH of the diluted protein solution was adjusted to pH 7-8. The plates were blocked with 150 l of 10 mg/ml BSA for 2 h at room temperature and washed with PBS with 0.9 mM calcium and 0.4 mM magnesium twice. RBL-2H3 cells were detached from plates with 5 mM EDTA, washed with PBS or saline solution containing 3 mM Ca 2ϩ and 1.5 mM Mg 2ϩ , and suspended in the same solution at a cell concentration of 2 ϫ 10 6 /ml. 2 ϫ 10 5 cells were applied to each well. After a 1-h incubation at room temperature, the plates were washed with the same solution seven times. To read the number of cells bound to each well, the ␤-hexosaminidase activity assay was used as previ-ously described (26). Briefly, 20 l of 0.5% Triton X-100 in PBS were added to each well to lyse the bound cells. 80 l of 1 mM substrate (p-nitrophenol N-acetyl-␤-D-glucosaminide; Sigma catalog no. N9376) in 0.05 M citrate buffer, pH 4.5, were subsequently added. After 1 h of incubation at 37°C, 100 l of 0.1 M sodium carbonate, 0.1 M sodium bicarbonate were added per well. The A 405 was measured with an Emax precision microplate reader (Molecular Devices, Inc., Sunnyvale, CA).

GPVI-expressing RBL-2H3 Cells
Signal in Response to Convulxin and CRP but Not to Collagen-RBL-2H3 cells stably expressing wild-type GPVI (GPVI-RBL) and FLAG-tagged GPVI (FLAG GPVI-RBL) were identified using adhesion to convulxin and flow cytometry to detect surface FLAG epitope, respectively. The ability of the putative GPVI ligands collagen, CRP, and convulxin to initiate calcium signaling in RBL, GPVI-RBL, and FLAG GPVI-RBL was tested in parallel with aggregation studies of human platelets performed using the same reagents on the same day (Fig. 1, Table I, and data not shown). GPVI-RBL and FLAG GPVI-RBL, but not untransfected RBL cells, responded to convulxin with a threshold concentration only 1.6-fold greater than that found necessary to aggregate human platelets (0.3 nM). These results demonstrate functional GPVI signaling and close concordance between the two assays for GPVI dose response, suggesting similar GPVI receptor density in the two cell types. In contrast, collagen elicited no response even at concentrations 500 times greater than that necessary to aggregate human platelets (100 g/ml). Similar results were obtained using three distinct GPVI-RBL clones and three distinct FLAG-GPVI RBL clones. The data shown are representative of experiments performed in RPMI with Chronolog Type I collagen, which elicits the most robust platelet FIG. 1. GPVI signaling in RBL-2H3 cells. Wild type (a) and RBL cells stably expressing GPVI (b-e) were exposed to collagen (100 g/ml), convulxin (1 nM), CRP (50 g/ml), or thrombin (10 nM), and calcium signaling was measured using the calcium-sensitive dye Fura-2. a, wild-type RBL cells have no response to convulxin or CRP but signal in response to thrombin. b, RBL cells expressing GPVI signal in response to the GPVI ligand CVX. c, GPVI-expressing RBL cells do not respond to collagen. d and e, GPVI-expressing RBL cells signal weakly in response to CRP. Results are representative of eight experiments performed with three distinct GPVI-expressing lines of RBL cells.

TABLE I Comparison of signaling by putative GPVI ligands in GPVIexpressing RBL cells and in platelets
The threshold concentrations of the platelet agonists convulxin, CRP, and collagen necessary to induce platelet aggregation were determined using platelet aggregometry and compared with those necessary for detection of calcium signaling in GPVI-expressing RBL cells using fluorimetry. Note the close concordance between the two assays for the GPVI-specific ligand convulxin and the significant discrepancy for the putative GPVI ligands CRP and collagen. responses in our hands (data not shown). In addition, no signaling responses were observed using a second source of type I collagen, type III collagen, or in the presence of higher cation concentrations (1 mM calcium and 0.5 mM magnesium; data not shown). Interestingly, CRP was capable of eliciting a small calcium response but required a concentration 500 times greater than that necessary to induce platelet aggregation (50 g/ml). CRP signaling responses were also only seen with the two GPVI-expressing RBL-2H3 cell clones that had the greatest sensitivity to convulxin. These data demonstrate that GPVI is a functional receptor but that GPVI expression alone is insufficient to reproduce the collagen and CRP signaling responses observed in platelets.
Adhesion of RBL and GPVI-RBL to the Putative GPVI Ligands Convulxin, Collagen, and CRP-To determine whether expression of GPVI is sufficient to mediate adhesion but not signaling to collagen, static adhesion assays were performed with wild-type and GPVI-expressing RBL cells (GPVI RBL and FLAG-GPVI RBL generated identical results; data not shown). Expression of GPVI conferred strong adhesion to convulxin and moderate adhesion to CRP, but no adhesion to collagen was detected (Fig. 2). GPVI R272L-and GPVI R295STOP-expressing RBL clones, which express ϳ5-fold and 10 -50-fold more surface GPVI than FLAG-GPVI, respectively (Fig. 3), also failed to bind collagen despite binding both CRP and CVX. Performing the assay using normal saline with 1 mM calcium and 0.5 mM magnesium yielded identical results (data not shown). In addition, no adhesion of GPVI-expressing or wildtype RBL cells was observed to bovine type I collagen or to bovine type III collagen (data not shown). Wild-type RBL cells adhered only to fibronectin (Fig. 2). Unlike fibronectin binding, GPVI-mediated adhesion to CRP and convulxin was not disrupted by 5 mM EDTA (Fig. 2B), consistent with a nonintegrinmediated mechanism of adhesion. Thus, adhesion assays of GPVI-expressing RBL cells are consistent with signaling assays and show strong GPVI interaction with convulxin, weaker GPVI interaction with CRP, and no GPVI interaction with collagen.
The GPVI Transmembrane Arginine and Intracellular C-tail Are Both Necessary for GPVI Signaling-The signaling roles of the GPVI transmembrane (TM) domain arginine (R272) and intracellular C-tail were tested by generating RBL cell lines expressing FLAG-tagged receptors in which the TM arginine is replaced by leucine (R272L-RBL) and the C-tail is truncated shortly following the TM domain (R295STOP-RBL). Both mutant GPVI receptors were expressed on the surface of RBL cells at levels equal to or greater than clones expressing wild-type GPVI (Fig. 3B). Consistent with the results of analogous mutations in related receptors, R272L-RBL did not signal in response to convulxin, confirming a necessary role for the GPVI TM domain arginine for signal transduction. Surprisingly, unlike similar C-tail truncation mutants of the Fc R␥ partners Fc⑀ RI and Fc␥ RIII (see Fig. 6), RBL cells expressing the GPVI C-tail truncation mutant R295STOP also failed to signal to convulxin, demonstrating an unexpected necessary role for the GPVI C-tail (Fig. 3C).
Loss of the GPVI Transmembrane Domain Arginine and C-tail Results in Loss of Coupling to Fc R␥-To determine why the R272L and R295STOP mutants of GPVI no longer supported signaling by CVX, we compared the ability of wild-type and mutant GPVI receptors to interact with Fc R␥ with a functional assay in HEK-293T cells stably expressing FLAG Fc R␥ and by direct biochemical means using coprecipitation. As for HEK-293T cells engineered to express low levels of the homologous immunoreceptor signaling adaptor DAP-12(25), FLAG-Fc R␥ is not expressed on the cell surface of these cells in the absence of a coexpressed Fc R␥ partner (Fig. 4A). The ability of an expressed receptor to rescue surface expression of FLAG-Fc R␥ therefore measures functional association with Fc R␥. Wild-type GPVI expression rescued 10 times more surface FLAG-Fc R␥ than mock-transfected cells (Fig. 4A). In contrast, GPVI R272L expression failed to rescue any FLAG-Fc R␥, consistent with a complete loss of association with Fc R␥ (Fig.  4A). GPVI R295STOP expression also failed to rescue surface FLAG-Fc R␥, indicating a lack of Fc R␥ interaction (Fig. 4A). Surface staining for the HA epitope confirmed that wild-type and mutant GPVI receptors were expressed at equivalent levels (Fig. 4B) and demonstrates that, as for the related Fc R␥ partner PIR␣ (18), GPVI expression in HEK-293T cells does not require Fc R␥ interaction. Loss of Fc R␥ interaction in GPVIR272L and GPVIR295 STOP was confirmed biochemically using convulxin to precipitate GPVI and subsequently assaying for associated Fc R␥ by immunoblotting (Fig. 5). Convulxin precipitation of wild-type GPVI, but neither mutant receptor resulted in coprecipitation of Fc R␥ (Fig. 5). Interestingly, immunoblotting of GPVI-R295STOP protein with anti-FLAG antibody reveals the presence of mature protein at the predicted molecular mass of ϳ57 kDa (5 kDa smaller than the wild-type and R272L receptors) and the presence of a large amount of protein at 36 -38 kDa, the predicted size for unglycosylated, incompletely processed protein. It is possible that this lower molecular mass species represents protein that cannot reach the cell surface because it cannot couple to Fc R␥ partners and is therefore targeted for degradation. The fact that the abundantly expressed R272L mutant escapes this fate supports the role of the transmembrane arginine in targeting unpartnered receptors for degradation. These results show that the GPVI transmembrane arginine is necessary but not sufficient for functional association with the receptor's signaling coreceptor Fc R␥ and that loss of signaling following truncation of the GPVI C-tail is due to loss of Fc R␥ interaction rather than loss of an unidentified, distinct signaling function. DISCUSSION Recent studies of human platelets that are unresponsive to collagen (10), mouse knockouts (27), and platelet signaling (reviewed in Ref. 14) have generated the hypothesis that the platelet surface protein GPVI mediates collagen signaling and does so through its interactions with the immunoreceptor signaling adaptor Fc R␥. We have expressed GPVI in RBL-2H3 cells and studied GPVI signaling in a heterologous system to directly and formally address this hypothesis. RBL-2H3 cells express endogenous Fc R␥ and are a model cell line for studying Fc⑀ RI receptor signaling (28). Like platelets, the signaling end points achieved by Fc⑀ RI receptor cross-linking and Fc R␥ signaling in RBL-2H3 cells include mobilization of intracellular calcium and degranulation. Unlike the megakaryocytic cell lines DAMI, HEL, and MEG-01, however, RBL-2H3 cells express neither endogenous GPVI nor the integrin receptor for collagen ␣ 2 ␤ 1 (data not shown). Thus, RBL-2H3 cells express the appropriate signaling machinery to study GPVI signaling without the ambiguity of endogenous collagen receptor expression.
Expression of wild-type and FLAG-GPVI conferred robust calcium signaling to the snake venom protein convulxin at a threshold concentration equivalent to that necessary to activate human platelets but no detectable response to collagen at a concentration more than 500 times greater than that necessary to activate human platelets (Fig. 1). Unlike collagen, convulxin has been demonstrated to directly bind GPVI (13,29) and was used by Clemetson et al. (6) to purify the GPVI protein from platelets. Thus, GPVI is a functional receptor in RBL-2H3 cells and signaling in RBL-2H3 cells closely resembles that in human platelets, but GPVI alone is not sufficient for collagen signaling.
To address GPVI-ligand interaction independent of signal transduction, we tested the adhesion of GPVI-RBL to putative GPVI ligands. GPVI expression conferred strong binding to convulxin and weaker binding to CRP but no detectable binding to collagen. Thus, the adhesion to immobilized proteins conferred by GPVI expression parallels the signaling responses observed to soluble agonists. These results are in contrast to those recently reported by Jandrot-Perrus et al. (7), who detected a small amount of collagen binding in a monocytic cell line (U937) stably expressing human or mouse GPVI. This discrepancy could reflect a difference in GPVI receptor density between the stable cell lines used or differences in methodology. In our hands, however, even clones expressing very high levels of GPVI such as the R272L clones (whose extracellular domains are wild type) confer adhesion to both CRP and convulxin but not collagen (data not shown and Fig. 3).
Is GPVI a bona fide collagen receptor, and, if so, why is GPVI expression insufficient to confer collagen signaling? Inadequate receptor density on RBL-2H3 cells is not a likely explanation, since the dose-response to CVX is similar in platelets and in GPVI-expressing RBL cells (Table I). One potential explanation for these results is that GPVI does mediate collagen sig-naling but that another coreceptor is required. This coreceptor might facilitate direct GPVI-collagen binding, or GPVI might mediate collagen signaling indirectly by linking a ligand-binding coreceptor to the signaling adaptor Fc R␥. Of the reported platelet collagen receptors, including the integrin ␣ 2 ␤ 1 (2), glycoprotein IV (30), and p65 (20), the most likely candidate is the ␣ 2 ␤ 1 integrin, whose high affinity for collagen may bring collagen to the platelet surface in an apparent concentration and/or configuration necessary for GPVI binding and signal transduction, although the precise role of ␣ 2 ␤ 1 remains controversial (31). Alternatively, GPVI may not be involved in collagen signaling, and an as yet unrecognized Fc R␥ partner may be the true collagen receptor.
Several lines of evidence support the model that collagen is a GPVI ligand but that GPVI absolutely requires a coreceptor such as ␣ 2 ␤ 1 for productive collagen interaction. CRPs, which structurally closely resemble collagen but are more potent activators of platelets (21), initiate a small amount of intracellular signaling in GPVI-expressing but not wild-type RBL-2H3 cells (Fig. 1). The CRP signaling response in GPVI-expressing RBL-2H3 cells, however, requires 500 times the concentration necessary to activate platelets, suggesting that the lack of observed signaling to the related ligand collagen may reflect an extremely low affinity rather than a complete lack of direct interaction. A necessary role for ␣ 2 ␤ 1 as a coreceptor for collagen signal transduction through GPVI is also supported by the description of an individual with reduced levels of ␣ 2 ␤ 1 whose platelets failed to aggregate in response to collagen (32), but the lack of genetic and molecular characterization of this individual precludes exclusion of associated defects in the platelet expression of GPVI, Fc R␥, or other unidentified platelet receptors. Finally, platelets from an individual lacking GPVI also demonstrated a loss of Fc R␥ expression despite normal Fc R␥ expression in other cell types (33), suggesting that GPVI may be the only Fc R␥ partner expressed in platelets and therefore must play a role in collagen signaling. Lack of genetic and molecular characterization of this individual, however, limits interpretation of this observation, and a megakaryocyte-specific block in Fc R␥ expression cannot be excluded. Thus, the preponderance of data support a complex model of collagen signal transduction at the platelet surface with necessary roles played by no fewer than four transmembrane proteins, GPVI, ␣ 2 ␤ 1 , and Fc R␥.
Studies of Fc R␥-deficient mouse platelets have revealed that, like Fc R␥ partners expressed in immune cells, GPVI is not expressed in the absence of Fc R␥ (27). Amino acid analysis of human and mouse GPVI reveals the presence of an arginine in the receptor transmembrane domain in a position identical to that of related Fc R␥ partners such as the Fc␣ receptor and the NK receptors PIR␣ and NKp46 (6). As found for the Fc␣ and PIR␣ receptors (16,18), mutation of this arginine does not interfere with receptor expression but results in a complete loss of receptor signaling (Fig. 3) and loss of interaction with the Fc R␥ (Figs. 4 and 5). Thus, the GPVI transmembrane arginine is required for Fc R␥ interaction, and Fc R␥ interaction is required for GPVI signaling.
The human GPVI C-tail is ϳ50 amino acids long and can be divided into basic, proline-rich, and serine/threonine-rich domains (Fig. 6). The mouse GPVI C-tail is shorter and lacks the serine/threonine-rich region (data not shown, and see Ref. 7). Truncation of the GPVI C-tail does not interfere with receptor expression but results in complete loss of signaling in RBL cells and loss of Fc R␥ interaction despite the presence of the transmembrane arginine (Figs. [3][4][5]. Thus, both the transmembrane arginine and the receptor C-tail are necessary, but neither alone is sufficient for intracellular signaling. Interestingly, similar truncation mutants (within 5-10 amino acids of the TM domain) have been studied with two other Fc R␥ partners, the Fc⑀ RI (34) and the Fc␥ RIII (35) receptors, with no loss of Fc R␥ interaction or signaling. It is intriguing to note that while all of these receptors couple functionally to the Fc R␥, the transmembrane domains of Fc⑀ RI and Fc␥ RIII are homologous to each other but demonstrate little homology to those of GPVI or Fc␣ receptor and lack the signature arginine residue of that subfamily of Fc R␥ partners. In addition, Fc⑀ RI and Fc␥ RIII share a chromosomal locus on human chromosome 1 (36), while GPVI and the related receptors discussed share a locus on chromosome 19, consistent with the existence of distinct ancestral receptors from which two receptor families may have evolved. Our results suggest that GPVI couples to the signaling adaptor Fc R␥ in a manner distinct from that of the previously studied Fc R␥ partners Fc⑀ RI and Fc␥ RIII. Precisely how the GPVI FIG. 5. Coprecipitation of Fc R␥ with wild-type and mutant GPVI receptors expressed in RBL cells. The interaction of Fc R␥ with wild-type and mutant GPVI receptors in RBL cells was determined biochemically by precipitation of GPVI receptors with convulxin-coated beads. CVX-precipitated protein was probed for GPVI using anti-FLAG antibody (upper panel) and for Fc R␥ using anti-Fc R␥ antibody (lower panel). RBL, untransfected RBL cells; RBL-GPVI, RBL cells stably expressing wild-type GPVI; RBL-GPVIR272L, RBL cells stably expressing GPVIR272L; RBL-GPVIR295STOP, RBL cells stably expressing GPVI R295STOP; ϩ, precipitation with convulxin-coated beads; Ϫ, control precipitations with BSA-coated beads. The GPVI-expressing RBL cell lines used were the same as those analyzed for surface expression and signaling in Fig. 3.

FIG. 6. Amino acid alignment of the transmembrane and intracellular domains of GPVI and other Fc R␥ receptor partners.
Alignment of the deduced amino acid sequences of human GPVI (hGPVI), mouse GPVI (mGPVI), Fc␣, PIR␣, Fc⑀ RI, and Fc␥ RIII receptors was performed using the ClustalW program (Macvector). Amino acid identities shared among the GPVI, Fc␣, and PIR␣ receptors are shaded. Amino acid identities shared between the Fc⑀ RI and Fc␥ RIII receptors are boxed. hGPVI R272 is in boldface type. *, the site of C-tail truncation for GPVI R295STOP; **, the site of C-tail truncation for an Fc⑀ RI receptor mutant; ***, the site of C-tail truncation for an Fc␥ RIII receptor mutant; TM, transmembrane domain; basic, portion of GPVI C-tail containing a significant number of basic amino acids; proline, portion of GPVI C-tail containing a cluster of proline residues; S/T, portion of human GPVI C-tail with a significant number of serine and threonine residues. C-tail facilitates Fc R␥ interaction and whether GPVI-related receptors also require their C-tails for Fc R␥ coupling remains uncertain and awaits further mutational analysis.
These studies provide the first functional analysis of GPVI as a signaling receptor and raise several important questions regarding the role of GPVI in vivo. The inability of GPVI to respond directly to collagen may suggest the evolution of a receptor adapted to operate in a highly specialized cellular environment in cooperation with other collagen receptors such as the integrin ␣ 2 ␤ 1 (a hypothesis supported by the megakaryocytic-specific pattern of expression of the receptor's mRNA (data not shown, and see Ref. 27), but the possibility that GPVI does not mediate collagen signaling cannot yet be definitively excluded. Our results extend the proposed model of platelet collagen signaling to one requiring no fewer than four receptor subunits and establish a heterologous system in which to further dissect this signaling pathway. Identification of the receptors involved in collagen activation of platelets and the molecular basis for this response may provide novel targets for anti-platelet therapies, which act at a critical point in thrombogenesis, the activation of newly adherent platelets at sites of atherosclerotic rupture.