The low-frequency isoform of platelet glycoprotein VIb attenuates ligand-mediated signal transduction but not receptor expression or ligand binding

Platelet glycoprotein VI (GPVI) is a 63-kDa type I transmembrane platelet glycoprotein that plays an important role in the activation of platelets and signal transduction induced by collagens and the snake venom protein convulxin (CVX).^1,2  The 2 most common human isomers, GPVIa and GPVIb, are distinguished by 5 amino acid substitutions produced by 5 biallelic single nucleotide polymorphisms (SNPs) within the 2 most common haplotypes, GP6a and GP6b, respectively.^3,–5  These amino acid substitutions are: S219P, K237E, and T249A in the ectodomains, and Q317L and H322N in the cytoplasmic domain.

In an elegant study, Joutsi-Korhonen et al⁵  showed that homozygous GP6b/b donors exhibit a modestly decreased expression of GPVI, with a concomitant decrease in P-selectin expression, platelet aggregation, and prothrombin conversion in response to cross-linked collagen-related peptide (CRP). Subsequent studies by the same group have clarified that the GP6b haplotype accounts for only a fractional effect (≤ 16%) on GPVI expression. A molecular basis for qualitative differences in the function of GPVIa and GPVIb has not yet been established, although it has been tacitly assumed without evidence that the nonconservative substitution S219P is a likely factor because it would be predicted to influence at least the local secondary/tertiary structure of the receptor.

In this study, we provide definitive evidence using both soluble and membrane-expressed GPVIa and GPVIb that neither S219P nor the 2 remaining ectodomain substitutions have a significant effect on the binding of GPVI to the specific ligands, CRP, type I collagen, or CVX. In contrast, the cytoplasmic amino acid substitutions influence the binding in vitro to both calmodulin (CaM) and Fyn/Lyn. Furthermore, Dami cells stably transfected with GPVIa or GPVIb bind equivalently to CVX or type I collagen, but the latter exhibit a statistically significant decrease in tyrosine phosphorylation of Syk after stimulation with CVX. These findings argue that the GP6a and GP6b haplotypes confer a significant effect on GPVI-mediated signal transduction, but a statistically insignificant effect on GPVI expression levels or ligand-binding affinity.

This research study was approved by The Scripps Research Institute, and all human participants gave written informed consent in accordance with the Declaration of Helsinki. Additional details of the following methods are provided in the supplemental data (available on the Blood website; see the Supplemental Materials link at the top of the online article).

Dami cells were purchased from the ATCC. Soybean trypsin inhibitor, leupeptin, aprotonin, and pepstatin A were purchased from Sigma-Aldrich. Rabbit polyclonal anti–human FcRγ and murine monoclonal anti-CaM are from Upstate Biotechnology. Human collagen type I was purified from placentae,⁶  and purified CVX was prepared, as previously described.⁷  CRP is a gift from the late Dr Michael Barnes (Cambridge University). Murine monoclonal antibody 6F1 (anti–integrin α₂β₁) is a gift from Dr Barry Coller (Rockefeller University), and murine monoclonal antibody HY101, specific for human GPVI, is a gift from Dr Mark Kahn (University of Pennsylvania). Antibodies specific for phosphorylated tyrosine (4G10) and Syk were purchased from Upstate Biotechnology.

Platelet mRNA from homozygous GP6aa or GP6bb donors served as template to amplify cDNA and generate soluble, recombinant 3XFLAG-tagged GPVI dimers. HEK293T cells were transiently transfected with the pGPVIa-Cys-3XFLAG or pGPVIb-Cys-3XFLAG, and soluble GPVI was affinity purified from the media using Anti-FLAGM2 Affinity Gel (Sigma-Aldrich). Recombinant soluble GPVI (rsGPVI) dimers have an MW_App of 110 kDa under nonreduced conditions; monomers, an apparent molecular weight (MW_app) of 60 kDa under nonreduced conditions. Dimers were separated from monomers using a Superdex 200 10/300 GL column (Amersham Biosciences). The sequence of these constructs is depicted in supplemental Figure 1.

rsGPVIa or rsGPVIb (0.5, 1, 2, 5, or 10 μg/mL) was incubated for 90 minutes at 37°C in the wells of microtiter plates coated with type I collagen (20 μg/mL), CVX (5 μg/mL), or CRP (1 μg/mL). Bound rsGPVI was quantitated with specific antibody by a horseradish peroxidase (HRP)–based enzyme-linked immunosorbent assay (ELISA; optical density 490 nm).

Dami cells were transfected with full-length recombinant (r) GPVIa or rGPVIb cDNA subcloned into pcDNA3 (Invitrogen) using the Effectene transfection system (QIAGEN), and stable cell lines were selected by drug resistance and adhesion to type I collagen. The association of endogenous FcRγ with transfected rGPVIa or rGPVIb was monitored by coimmunoprecipitation with rabbit anti-FcRγ IgG and LJ6.5. The association of endogenous CaM with transfected rGPVIa or rGPVIb was assessed by quantitative immunoprecipitation of aliquots of cell lysates containing equivalent amount of rGPVI protein with anti-CaM. The surface expression of rGPVIa or rGPVIb was monitored in flow cytometry using HY101, a neutral monoclonal antibody that neither activates GPVI nor inhibits ligand binding to GPVI.

Four maltose-binding protein (MBP) fusion constructs were synthesized, following the protocol of Suzuki-Inoue et al,⁸  and are designated by the amino acid substitution at positions 317 and 322 as QH (representing GPVIa) and QN, LH, and LN (representing GPVIb). The amount of each MBP-GVI construct that binds to CaM from platelet lysates was determined by immunoprecipitation and Western blot with anti-CaM.

The binding of MBP-GPVI cytoplasmic domain fusion proteins (MBP-GPVI) to the Src homology 3 (SH3) domains of the Src kinases Fyn and Lyn was measured by ELISA, as described.⁸  Glutathione sulfonyl transferase (GST)–SH3 fusion proteins were generously provided by Dr Steve Watson (University of Birmingham).

This is a modification of the published method.⁹  Blood was anticoagulated in EDTA (ethylenediaminetetraacetic acid), and platelets were isolated, washed, and lysed in the presence of 1 mM EDTA. GPVI was detected in Western blots using biotin-LJ6.5. The relative amount of GPVI in a test platelet sample is expressed relative to a standardized platelet sample, as described,⁹  in which the amount of GPVI per 20 μg of platelet protein in the standard sample is set to 10 arbitrary units.

Quantification of GPVI in photographic films derived from Western blots was performed by optical densitometry using the 2(−ΔΔC_T [cycle threshold]) method, as described by Livak and Schmittgen.¹⁰ 

The level of surface GPVI was determined by the quantitative binding of biotin-conjugated HY101 in flow cytometry, as described.^11,12 

Dami cells stably transfected with empty plasmid, GPVIa, or GPVIb were stimulated with CVX (2, 5, or 10 μg/mL) for 30, 90, or 270 seconds at 37°C. Cells were then lysed by addition of an equal volume of ice-cold lysis buffer (20 mM Tris-HCl, pH 7.3, 300 mM NaCl, 2 mM EDTA, 2 mM EGTA [ethyleneglycoltetraacetic acid], 2% [vol/vol] Nonidet P-40, 2 mM Na₃PO₄, 1 mM phenylmethylsulfonyl fluoride, 10 mg/mL leupeptin, 10 mg/mL aprotinin, 1 mg/mL pepstatin A), centrifuged at 15 000g for 10 minutes to remove detergent-insoluble material, and precleared with 30 mL of a 50% suspension of protein A-Sepharose in Tris-buffered saline-T buffer (20 mM Tris-HCl, 137 mM NaCl, 0.1% [vol/vol] Tween 20, pH 7.3). Samples were then immunoprecipitated with monoclonal anti-Syk antibody and 30 mL of protein A-Sepharose for 2 hours at 4°C. The beads were washed once with lysis buffer and 3 times with Tris-buffered saline-T buffer, and then solubilized in sodium dodecyl sulfate (SDS) electrophoresis sample buffer and boiled for 10 minutes. Samples were electrophoresed on SDS-polyacrylamide gels under reducing conditions (1% 2-mercaptoethanol), electrotransferred to polyvinylidene difluoride (PVDF) membranes, and immunoblotted with the anti-phosphotyrosine antibody 4G10. Blots were visualized using horseradish peroxidase-conjugated secondary antibody and the enhanced chemiluminescence detection system (Amersham Biosciences). Membranes were then stripped and reprobed with anti-Syk antibody.

GP6 genotypes were determined using a customized Nanogen-based SNP analysis (Nanogen) and direct Sanger sequencing of polymerase chain reaction amplicons, as described (12 049).

Statistical calculations were performed using SigmaStat Version 3.01 (SPSS). For continuous variables, descriptive statistics were calculated and reported as mean and standard deviation, and comparisons were made using the Student t test and analysis of variance. All tests are 2-sided and considered significant at P values less than .05.

In this study, the anticoagulation of blood in EDTA, the subsequent platelet isolation and washing in the presence of EDTA, and the lysis of platelets by SDS in the presence of EDTA minimize platelet activation during processing and essentially eliminate a disintegrin and metalloproteinase (ADAM)–10 cleavage of GPVI.¹³  Total platelet GPVI levels in 132 white, non-Hispanic donors were analyzed by Western blot. In a subset of these donors (n = 22), we compared total GPVI level, determined by Western blot, and platelet surface GPVI expression, determined by the binding of HY101 in flow cytometry (supplemental Figure 2). These results established that there is a statistically significant correlation between total platelet GPVI level and surface GPVI expression (P < .01).

In Figure 1, the level of total platelet GPVI in each donor platelet sample, measured by Western blot, is plotted as a function of GP6 genotype. The descriptive statistics for each genotype are listed in supplemental Table 1. The difference in GPVI levels between any pair of genotypes is not statistically significant: aa versus ab, P equals .749; aa versus bb, P equals .586; and ab versus bb, P equals .812.

rsGPVI, representing residues 1 to 269, was generated bearing either set of haplotype-specific ectodomain substitutions (rsGPVIa and rsGPVIb), and dimer-enriched preparations were produced by size exclusion chromatography using a Superdex 200 10/300 GL column in the presence of N-ethylmaleimide (nonreduced; supplemental Figure 3).

The ability of dimer-enriched rsGPVIa or rsGPVIb to bind in a dose-dependent manner to immobilized GPVI ligands was then investigated (Figure 2). rsGPVIa and rsGPVIb bound with a statistically equivalent affinity to type I collagen, with a K_d of 1.2 (± 0.4) and 1.3 (± 0.6) μg/mL (mean ± SD), respectively (Figure 2A). In the case of CRP, the corresponding K_d calculations were 1.3 (± 0.5) and 1.5 (± 0.6) μg/mL (Figure 2B); for CVX, 0.8 (± 0.4) and 0.6 (± 0.5) μg/mL (Figure 2C).

To ascertain whether expression of membrane-bound, full-length rGPVIa or rGPVIb might introduce additional conformational or molecular constraints that otherwise influence ligand binding, we compared the adhesion of Dami cells stably transfected with full-length rGPVIa or rGPVIb to type I collagen, CRP, and CVX (Figure 3). For these comparisons, stable Dami cell clones were selected that exhibited comparable surface expression of each rGPVI molecule, based on the binding of Hy101 in flow cytometry, and 3 clones expressing each isoform were used in subsequent experiments (supplemental Table 2).

First, there is no appreciable difference in the association of either rGPVIa or rGPVIb with endogenous Dami cell FcRγ, as determined by coprecipitation experiments with LJ6.5 or polyclonal rabbit anti-FcRγ (Figure 3A).

Adhesion to type I collagen was performed in the presence of 1 mM EDTA or 60 μg/mL 6F1 to eliminate the possible involvement of endogenous integrin α₂β₁. In either case, equivalent levels of adhesion were mediated by rGPVIa or rGPVIb (Figure 3B). Likewise, comparable adhesion to CRP (Figure 3C) or CVX (Figure 3D) was evident. Minor differences in mean adhesion to any ligand were not statistically significant (P > .05 in each pairwise analysis).

The ability of CaM to directly bind to MBP-GPVI cytoplasmic domain fusion proteins in vitro was measured by an immunoprecipitation assay (Figure 4A). In each experiment, the relative amount of MBP-GPVI present in each anti-CaM immunoprecipitate was LN or LH > QN or QH. These results indicate that CaM binds better to MBP-GPVI-LN (GPVIb) than to MBP-GPVI-QH (GPVIa). Our results also suggest that it is the Q317L substitution that has the greater influence on the increased binding to CaM. The results depicted in Figure 4A are representative of 3 independent experiments.

To confirm that rGPVIb binds to endogenous Dami cell CaM to a greater extent than does rGPVIa, we measured by immunoprecipitation the amount of rGPVI in lysates of stable Dami cell transfectants that is coprecipitated with endogenous CaM (Figure 4B). Predetermined volumes of rGPVIa or rGPVIb lysates containing equivalent amounts of rGPVI were precleared with washed protein G agarose beads and incubated overnight with anti-CaM, and the resulting CaM/anti-CaM immune complexes were isolated by adsorption to fresh protein G agarose beads. Immune complexes were eluted, and protein constituents were separated by SDS–polyacrylamide gel electrophoresis (PAGE). The relative amount of rGPVI recovered was determined by Western blot using LJ6.5, and the amount of CaM present in each sample was determined by the binding of anti-CaM.

It is clear in Figure 4B that an identical amount of anti-CaM will immunoprecipitate essentially equal amounts of CaM from lysates of Dami cells stably transfected with rGPVIa or rGPVIb, and that the same amount of CaM is precipitated from nontransfected Dami cells (mock). However, the amount of rGPVIb (b) that is coprecipitated is consistently increased relative to that of rGPVIa (a). As expected, rGPVI is not coprecipitated from nontransfected cells (mock). These results are typical of the findings made in 3 equivalent experiments and are consistent with an increased affinity of the cytoplasmic domain of rGPVIb for CaM.

The Fyn/Lyn binding site within the proline-rich region at 307-316 may be indirectly affected by Q317L and/or H322N. To test this, we measured the binding of MBP-GPVI to GST-Fyn-SH3 (Figure 5A), GST-Lyn-SH3 (Figure 5B), and GST-Btk-SH3 as a negative control (Figure 5C) that was attached to preblocked glutathione-coated microtiter plates (Sigma-Aldrich). The findings with a second negative control, GST-PLCγ2-SH3, were comparable with that obtained with GST-Btk-SH3 (data not shown).

We observed a decreased binding of GPVIb (LN) relative to GPVIa (QH) to GST-Fyn-SH3 or GST-Lyn-SH3, and the decrease is more striking in the presence of calcium (≥ 0.03 μM). The binding of QH and LH remained largely unaffected, whereas the binding of QN and LN was significantly diminished. These results suggest that the substitution H322N significantly weakens the affinity of SH3 domains for the GPVI cytoplasmic domain, particularly in the presence of calcium.

A converse ELISA in which Immulon-2 plates were coated with MBP (negative control) or MBP-GPVI constructs and the binding of GST-SH3 constructs was measured generated the same relative findings (data not shown), but with a less than desirable decrease in the signal to noise ratio.

Differences in the Fyn/Lyn-binding properties of the GPVIa and GPVIb cytoplasmic domains might influence the rate and extent of GPVI-mediated signaling. To address this question, we compared the time course and extent of Syk phosphorylation after CVX treatment of Dami cells stably transfected with GPVIa or GPVIb. At a constant CVX concentration (10 μg/mL), maximal Syk phosphorylation was attained by 90 seconds after CVX treatment (Figure 6), confirming prior studies in Dami cells and other megakaryocytic cell lines.^14,–16  The results of 1 of 3 independent assays is depicted in Figure 6A, and the mean plus or minus SD of the results of the 3 independent assays is plotted in Figure 6B.

The dose dependence of CVX-induced Syk phosphorylation is depicted in Figure 7. At 90 seconds after treatment with either 5 or 10 μg/mL CVX, a statistically significant decrease in Syk phosphorylation was seen in rGPVIb-transfected Dami cells, relative to those transfected with rGPVIa. The results of 1 of 3 independent assays are depicted in Figure 7A, and the mean plus or minus SD of the results of the 3 independent assays are plotted in Figure 7B.

We first reported that the level of platelet GPVI among normal subjects varies 5-fold,⁹  and others have confirmed this observation, reporting ranges from 2- to 10-fold.^11,17  These differences are apparently relevant in vivo because 2 independent clinical studies have determined that an increased level of GPVI is a risk factor for negative outcomes in coronary artery disease.^12,18,19  There was an initial optimism that GPVI level would correlate with the haplotypes GP6a and GP6b,⁵  but a more careful evaluation has shown that they account for no more than 16% of the variation in expression level.²⁰  Our findings in 132 normal volunteers in this study confirm that GP6a and GP6b do not have a statistically significant effect on GPVI expression levels. This may explain why several studies analyzing the association of the GP6 SNP S219P with risk for negative outcomes in acute coronary disease^21,,–24  or ischemic stroke²⁵  have generated conflicting results.

In this study, we provide conclusive evidence that there is no difference in ligand-binding capacities between the ectodomains of GPVIa or GPVIb. Soluble ectodomains of GPVIa and GPVIb bind equally well to the major ligands of this receptor. As dimers, each binds to type I collagen, CRP, or CVX in static adhesion assays to an equivalent extent and with essentially identical affinities. These results argue that the ectodomain substitutions, including S219P, do not influence the binding of GPVI to these ligands, even though they must certainly contribute to localized alterations in molecular conformation. A basic reason that these 3 substitutions are relatively innocuous (with respect to ligand binding) may lie in the fact that all 3 are located in the stalk region of GPVI, a region rich in O-glycosylated serine and threonine, which has not been shown to be involved in the interaction of GPVI with convulxin or collagens.²⁶  At the same time, we also show that, when expressed by stably transfected Dami cells, the full-length, membrane-bound rGPVIa and rGPVIb bind similarly to CRP, type I collagen, or CVX. In addition, equal amounts of rGPVIa or rGPVIb are coprecipitated with endogenous FcRγ from these stably transfected cells, indicating that there is not a haplotype-specific difference in the association of GPVIa or GPVIb with the coreceptor FcRγ chains.

In contrast, our evidence in this study establishes a significant difference in GPVI function associated with the cytoplasmic amino acid substitutions. With respect to the binding site for CaM at residues 294 to 303, previously defined by Andrews et al²⁷  and Locke et al,²⁸  the L317 allele of GP6b confers an increased binding to CaM in vitro. In addition, an increased level of endogenous Dami cell CaM was copurified with rGPVIb in vivo, relative to that which associates with rGPVIa. The exact functional importance of CaM binding to GPVI is not yet understood, although there is substantial evidence that increased platelet calcium levels resulting from platelet activation can induce the dissociation of the GPVI-CaM complex, rendering the freed GPVI mobile and susceptible to proteolytic cleavage by the metalloproteinase ADAM-10.^13,29  Because CaM binding to GPVI 294-303 regulates metalloproteinase-mediated ectodomain shedding, it is possible that the increased affinity of GPVIb may attenuate the rate or extent of GPVIb proteolysis and thereby decrease the production of shed, soluble GPVI. Because soluble GPVI has been shown in some cases to be an efficient inhibitor of thrombus formation in vitro³⁰  and in vivo,³¹  this could be one explanation why GP6b was found to be associated with increased risk for thrombosis in middle-aged men²¹  or older people.²⁴ 

The interaction of Fyn or Lyn with the proline-rich region of the GPVI cytoplasmic tail is critical for maximal signaling by this receptor,⁸  and it is accepted that the binding of the SH3 domain of Fyn and Lyn to the GPVI tail increases the intrinsic activity of these Src kinases. Thus, any change in GPVI that decreases this interaction might lead to impaired GPVI-mediated signaling. We have observed in this study that the ability of the GPVIb cytoplasmic tail to bind in vitro to Fyn and Lyn SH3 domains is decreased relative to that of GPVIa. Moreover, this decrease becomes statistically significant at calcium levels expected to be produced within the activated platelet.

The results of our analyses of CVX-induced Syk phosphorylation in stably transfected Dami cells confirm that cytoplasmic differences in the GPVI isoforms do modulate signal transduction induced by the GPVI-specific ligand CVX. Both the rate and extent of the tyrosine kinase Syk phosphorylation are significantly attenuated in GPVIb-transfected cells relative to those transfected with GPVIa. Lyn is known to be expressed by Dami cells, in which it plays a critical role in signaling initiated by the engagement of surface receptors such as very late activation antigen 5.³²  In platelets and GPVI-transfected cell lines, the tyrosine kinase Syk is phosphorylated early after GPVI engagement, evident at 30 seconds and maximal by 90 seconds.³³  Any perturbation in the cytoplasmic tail of GPVI that would attenuate the binding of Fyn/Lyn could diminish the rate of Syk phosphorylation and effector functions of GPVI-ligand binding. In this study, we have shown that the GPVIb cytoplasmic tail, and in particular its H322N substitution located downstream from the proline-rich binding region, has a diminished ability to bind to Fyn/Lyn and a diminished capacity to mediate signaling required for Syk phosphorylation.

In summary, our results show that the molecular differences conferred by the 2 major haplotypes GP6a and GP6b do not significantly influence expression or ectodomain-ligand binding, but do significantly alter Fyn/Lyn and CaM binding to the cytoplasmic domains, resulting in a significant difference in CVX-induced signal transduction, as measured by Syk phosphorylation. These differences provide a molecular basis for the previously documented difference in function of GP6aa versus GP6bb platelets.^5,20  The range of GPVI levels between donors remains large (at least 5-fold), regardless of GP6a or GP6b genotype (see Figure 1). Moreover, we have shown previously that the activities of the 5′ regulatory and promoter regions of the GP6a and GP6b haplotypes are indistinguishable and cannot account for this variation.³  The molecular basis for these differences in GPVI levels on platelets remains to be determined, but may be the result of additional, as yet undescribed regulatory elements within or outside of GP6.

This work was supported by R01 grants HL075821 and HL086904 from the National Heart, Lung and Blood Institute awarded to T.J.K., and Children's Hospital of Orange County Foundation grants awarded to D.J.N. This is manuscript MEM-18622 from The Scripps Research Institute.

National Institutes of Health

Contribution: E.T. and Y.C. performed the bulk of laboratory experiments, participated in study design, and wrote this manuscript; S.A.W. and D.J.N. participated in and supervised study design and laboratory experimentation, and completed key aspects of the study, particularly donor genotyping; K.F. and F.M.P. contributed to study design and supervision; and T.J.K. conceived and designed the study, supervised data accumulation and management, performed statistical analyses, and edited the final draft of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Thomas J. Kunicki, Department of Molecular and Experimental Medicine, The Scripps Research Institute, 10550 North Torrey Pines Rd, MEM-150, La Jolla, CA 92037; e-mail: tomk@scripps.edu.