Controlled shedding of platelet glycoprotein (GP)VI and GPIb–IX–V by ADAM family metalloproteinases


Michael C. Berndt, Monash University, Department of Immunology, Alfred Medical Research and Education Precinct (AMREP), Commercial Road, Melbourne, Australia.
Tel.: 61 3 9903 0713; fax: 61 3 9903 0038; e-mail:


Background: Platelet glycoprotein (GP)VI that binds collagen, and GPIb–IX–V that binds von Willebrand factor, initiate thrombus formation.Objectives: In this study, we investigated the mechanisms of metalloproteinase-mediated ectodomain shedding that regulate the surface expression of GPVI, GPIbα (the major ligand-binding subunit) and GPV (that regulates thrombin-dependent activation via GPIbα).Methods and results: Immunoblotting human platelet lysates using affinity-purified antibodies against cytoplasmic domains of GPVI, GPIbα or GPV allowed simultaneous analysis of intact and cleaved receptor, and revealed (i) that a significant fraction of GPIbα, but not GPVI, exists in a cleaved state on platelets, even when isolated in the presence of metalloproteinase inhibitor (GM6001) or EDTA; (ii) the same-sized membrane-associated fragments of GPVI or GPIbα are generated by phorbol-ester (PMA), the mitochondrial-targeting reagent CCCP, the calmodulin inhibitor W7, or the thiol-modifying reagent, N-ethylmaleimide, that directly activates ADAM10/ADAM17; and (iii) GPV is shed by both metalloproteinase- and thrombin-dependent mechanisms, depending on the concentration of thrombin. Based on the predicted cleavage area defined by these studies, ADAM10, but not ADAM17, cleaved a GPVI-based synthetic peptide within the extracellular membrane-proximal sequence (PAR^Q243YY) as analyzed by MALDI-TOF-MS. In contrast, ADAM17, but not ADAM10, cleaved within the GPIbα-based peptide (LRG^V465LQ). Both ADAM10 and ADAM17 cleaved within a GPV-based peptide (AQP^V494TT). Metalloproteinase-mediated shedding of GPIbα from GPIb-IX-transfected or GPVI-transfected cells induced by W7 or N-ethylmaleimide was inhibited by mutagenesis of sequences identified from peptide analysis.Conclusions: These findings suggest surface levels of GPVI, GPIbα and GPV may be controlled by distinct mechanisms involving ADAM10 and/or ADAM17.


Glycoprotein (GP)VI that binds collagen [1] and the GPIb–IX–V complex that binds von Willebrand factor (VWF) and other ligands [2], form a unique adheso–signaling complex on the surface of human platelets [3], and act synergistically to initiate platelet adhesion, activation and integrin (αIIbβ3)-dependent aggregation in hemostasis and thrombosis [2,4,5]. GPVI is a member of the immunoglobulin (Ig) immunoreceptor family, and consists of two extracellular Ig domains, a mucin-like domain, transmembrane domain, and cytoplasmic tail [1]. It binds the major physiological ligand collagen, the snake toxin convulxin [1], and collagen-related peptide (CRP) [6]. GPIb–IX–V of the leucine-rich repeat family consists of four subunits: GPIbα disulfide-linked to GPIbβ and non-covalently associated with GPIX and GPV [2]. GPIbα, the major ligand-binding subunit of GPIb–IX–V, mediates platelet adhesion to subendothelial matrix (VWF or thrombospondin), activated endothelial cells (P-selectin or VWF/P-selectin), or leukocytes (αMβ2) [2]. Thrombus formation is significantly impaired in transgenic mice where platelets express a GPIb–IX–V complex lacking the GPIbα ligand-binding ectodomain, with a more pronounced phenotype than VWF-deficient mice [5]. GPIbα also plays a key role in regulating coagulation at the surface of activated platelets, by binding coagulation factors (F) XI and XII, high molecular weight kininogen, and α-thrombin [2]. Thrombin also signals directly via GPIbα under conditions where GPV is removed, suggesting GPV regulates thrombin-dependent platelet activation [7]. Recent evidence suggests surface expression levels of GPVI and GPIb–IX–V correlate with the thrombotic propensity of platelets, and may also play a role in platelet clearance and aging [8–12]. In recent clinical studies, platelet GPVI levels showed a significant correlation with acute coronary syndrome [13].

One important mechanism regulating expression levels of GPVI [14–16], GPIbα [8,12] and GPV [11] is metalloproteinase-mediated ectodomain shedding, whereby intact receptor is removed from the platelet surface and soluble fragment is released into the platelet supernatant. ADAM17 of the ADAM (a disintegrin and metalloproteinase) family of sheddases has been shown to mediate shedding of both GPIbα and GPV, as shedding occurs in platelets treated with recombinant purified ADAM17, and is defective in genetically modified mice where ADAM17 is catalytically inactive [11,16]. Although the platelet sheddase involved in GPVI shedding is unknown, there is circumstantial evidence suggesting that GPVI shedding occurs by a different mechanism to that of GPIbα or GPV. GPVI shedding is induced by GPVI ligands such as collagen, convulxin or CRP [14], or by anti-GPVI antibodies [17], and can be completely depleted from the platelet surface with little change in GPIbα levels.

ADAM17 and ADAM10 are both expressed on human platelets [11,16,18], and consist of a pro-domain, a catalytic domain containing a conserved HEXXHXXGXXH metal ion-coordination motif, a disintegrin domain, a transmembrane domain and cytoplasmic tail [19]. The pro-domain of ADAM10 and ADAM17 contains a cysteine-switch; that is, a free thiol that coordinates the active-site metal ion and inhibits metalloproteinase activity, but is released, activating the enzyme, following proteolytic removal of the pro-domain or on modification by thiol-modifying reagents such as N-ethylmaleimide (NEM) [19]. NEM induces shedding of GPVI from human platelets [3]. As originally shown for leukocyte L-selectin [20], ADAM-mediated shedding is also regulated by calmodulin, which binds to a juxtamembrane positively-charged/hydrophobic sequence in the cytoplasmic tail of the receptor. Disruption of this interaction by mutagenesis or treating L-selectin-expressing cells with calmodulin inhibitors induces shedding, suggesting calmodulin association protects against metalloproteinase-mediated ectodomain shedding. Calmodulin is directly associated with the cytoplasmic domains of GPVI and GPIb–IX–V (GPIbβ and GPV subunits), and is dissociated following platelet activation [21,22]. Like L-selectin, ectodomain shedding of GPVI and GPV is induced by calmodulin inhibitors such as W7 [11,14]. ADAMs may themselves be activated by calmodulin inhibitors as dissociation of calmodulin leads to activation of ADAM10 [19,23]. Switching on intracellular signaling pathways also activates ADAMs; for example, the phorbol ester, PMA, that activates protein kinase C (PKC), induces shedding of GPIbα, GPV and substrates on other cells [9,11,19,23]. Alternately, ligand-induced shedding of GPVI is blocked by inhibitors of GPVI-related signaling molecules, Src kinases (PP1/PP2), Syk (piceatannol) or PI3-kinase (wortmannin) [14], upstream of PKC activation.

The aim of the present study was to investigate the mechanisms of metalloproteinase-mediated ectodomain shedding of GPVI, GPIbα and GPV. Firstly, we used affinity-purified antibodies against the cytoplasmic domains of GPVI, GPIbα and GPV for immunoblot analysis of human platelet lysates (PL), allowing simultaneous detection of intact receptor and their membrane-associated proteolytic remnants. Secondly, based on this analysis, synthetic peptides corresponding to extracellular sequences of GPVI, GPIbα and GPV were treated with recombinant ADAM10 or ADAM17, and digests analyzed by mass spectrometry. Cleavage sites identified from the peptide analysis were mutated in recombinant GPVI and GPIbα to evaluate the effects on shedding in transfected cells. Together, these results suggest GPVI and GPIbα may be differentially shed by ADAM10 or ADAM17, respectively, and identify precise sequences susceptible to ADAM-dependent proteolysis.

Materials and methods


Thrombin was purchased from Sigma (St Louis, MO, USA). NEM, GM6001 (a broad-range hydroxamic acid-based metalloproteinase inhibitor), the calmodulin inhibitor W7 [N-(6-aminohexyl)-5-chloro-1-naphthalenesulfonamide] and PMA (phorbol 12-myristate-13-acetate) were purchased from Calbiochem (La Jolla, CA, USA). Recombinant human ADAM10 and ADAM17 were purchased from R&D Systems (Minneapolis, MN, USA). The protease inhibitor cocktail, CompleteTM, was from Roche Diagnostic (Mannheim, Germany). The GPVI agonist, convulxin [1], was a kind gift from Dr Kenneth Clemetson (Berne, Switzerland). Ser-Phe-Leu-Leu-Arg-Asn (SFLLRN) peptide (thrombin receptor agonist peptide, TRAP) was from Auspep (Parkville, Australia).


The anti-GPVI monoclonal antibody (mAb), hybridoma medium 6B-12, was used for immunoblotting at a dilution of 1:150 as previously described [14]. The murine anti-GPVI mAb, 1G5, was raised using a recombinant extracellular fragment of human GPVI [24], kindly provided by Dr Masaaki Moroi, Fukuoka, Japan. The anti-GPIbα mAb, WM23, and affinity-purified polyclonal antibodies to glycocalicin and recombinant GPVI cytoplasmic tail have been described elsewhere [25–27]. Rabbit polyclonal antipeptide antibodies were raised against synthetic peptides, CYSGHSL or CKIGQLFRKLIRERALG (containing an N-terminal Cys for conjugation) corresponding to cytoplasmic sequences of human GPIbα (605–610) or GPV (529–544), respectively, and affinity-purified as previously described [28]. A rabbit anti-ADAM10 antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, California, USA). Horseradish peroxidise (HRP)- or FITC-conjugated sheep antimouse or antirabbit IgGs were from Chemicon (Melbourne, Australia).

Immunoblot analysis of GPVI, GPIbα and GPV proteolysis on human platelets

Assays used for measuring shedding of GPVI were performed essentially as previously described [14]. Washed human platelets were resuspended at 5 × 108 mL−1 in Tyrode’s buffer with either vehicle only (DMSO) or in the presence of (all final concentrations) the metalloproteinase inhibitor GM6001 (100 μm) or EDTA (10 mm). For some experiments, GM6001 or EDTA were also included throughout the collection and washing steps as well as in the sample buffer. Samples were treated with either convulxin (0.5 μg mL−1), CRP (0.5 μg mL−1), NEM (2 mm) or W7 (150 μm) at room temperature for indicated times, then made 10 mm in EDTA, and platelets separated from supernatant by centrifugation. Alternately, some samples were treated with (all final concentrations) α-thrombin (0.1–10 U mL−1), TRAP (1–30 μm), PMA (10 μm) or CCCP (100 μm). The pellet was lysed on ice for 30 min in 0.01 m Tris–HCl, 0.15 m NaCl, pH 7.4 (TS buffer) containing 1% (w/v) Triton X-100 and Complete protease inhibitor. Samples were analyzed by electrophoresis on SDS-5–20%-polyacrylamide gels, and Western blotting with anti-GPVI mAb (6B-12), or anti-glycocalicin, anti-GPIbα cytoplasmic tail, anti-GPVI cytoplasmic tail, or anti-GPV cytoplasmic tail IgG. Blots were visualized using HRP-conjugated secondary antibodies and ECL (Amersham, Amersham, Buckinghamshire, UK).

ADAM10 and ADAM17 digestion of GPVI-, GPIbα- and GPV-based peptides

Synthetic peptides corresponding to membrane-proximal extracellular sequences of GPVI (Thr221-Gly248), GPIbα (Lys450-Phe483) and GPV (Ser485-Phe508) were purchased from Auspep, Melbourne, Australia, and HPLC-purified. Peptides (10 μg) were dissolved in 30 μL reaction buffer (20 mm imidazole, 5 mmβ-mercaptoethanol, 1 mm CaCl2, pH 7.0), in the absence or presence of either ADAM10 or ADAM17 at a final concentration of 1 μg mL−1. After 1 h at 37°C, MALDI-TOF analysis was performed with an Applied Biosystems Voyager-DE STR BioSpectrometry Workstation. Data were collected from 500 laser shots and the signal averaged and processed with Data Explorer software. Assignment of amino acid sequences was performed using the Compute Mw/pI tool (

Shedding of GPVI from GPVI-transfected cells

Rat basophilic leukemia (RBL) cells were cultured in Dulbecco’s modified Eagle’s medium, supplemented with 10% (v/v) fetal bovine serum (CSL, Melbourne, Australia), 0.75% (w/v) NaHCO3, 1 mm sodium pyruvate, as well as penicillin, streptomycin and antimycin. RBL cells were stably transfected with wild-type human GPVI cDNA in growth medium containing 1.0 mg mL−1 G418 as previously described [29]. A Gln243 to Lys (Q243K) substitution mutation was introduced into GPVI cDNA using the PCR method of quick change mutagenesis [30] and selected in G418-containing medium. Cells were sorted for expression of GPVI using a FACSvantage DiVa with anti-GPVI mAb (6B-12). RBL cells expressing wild-type or mutant GPVI were plated onto six-well plates 24 h prior to an experiment to achieve > 90% confluency. Culture medium was removed and cell monolayers rinsed three times with PBS buffer (0.01 m NaH2PO4, 0.15 m NaCl, pH 7.4) containing 1 mm Ca2+/Mg2+ then left untreated or treated with (final concentration) 150 μm W7 or 2 mm NEM at 37 °C. Cells were lysed in TS buffer containing 1% (v/v) Triton X-100 and CompleteTM protease inhibitor, cleared of nuclei and cell debris by centrifugation, and analyzed by Western blot as described above for platelets. For flow cytometry, cells were pelleted and resuspended in 0.1 mL TS buffer containing 0.1% (w/v) BSA and 10 μg mL−1 of anti-GPVI mAb, 1 G5. Tubes were placed on ice for 30 min, centrifuged and resuspended in 10 μg mL−1 FITC-conjugated anti-mouse IgG. Non-immune mouse IgG was used as a negative control. After 30 min on ice, bound antibody was measured in a FACStar flow cytometer and analysis performed using CellQuest software as previously described [30].

Shedding of GPIbα from GPIb-IX-transfected cells

Human embryronic kidney (HEK293) cells already stably transfected with GPIbβ and GPIX were used to express wild-type GPIbα or mutant GPIbα (Δ463–479) containing a deletion of residues 463–479, or (V465K) with a Val465/Lys substitution, as previously described for CHO cells [30]. Cells prepared as described above, were left untreated or treated with (final concentration) 150 μm W7 in the absence or presence of GM6001 (100 μm), and analyzed by Western blot with anti-glycocalicin IgG, or flow cytometry with anti-GPIbα mAb (WM23) and FITC-labeled anti-mouse IgG.


Immunoblot analysis of GPVI, GPIbα and GPV proteolysis on human platelets

In this study, we investigated the mechanisms for metalloproteinase-mediated ectodomain shedding that control surface expression of the platelet receptors, GPVI and GPIb–IX–V. To analyze shedding of GPIbα, GPVI and GPV on human platelets, we used antibodies against their cytoplasmic domains, allowing simultaneous detection of full-length receptor and remnant membrane-associated fragments. Firstly, blotting with anti-GPIbα cytoplasmic tail antibody showed that a significant fraction of total GPIbα on untreated resting platelets was cleaved (Fig. 1A). This was not because of cleavage during platelet isolation as including the metalloproteinase inhibitor GM6001 or EDTA throughout the isolation and washing steps did not alter the relative proportion of the ∼130-kDa species corresponding to intact GPIbα (under reducing conditions, GPIbβ is absent), and an ∼16-kDa fragment (Fig. 1Bcf. 1A). The antibody also detected a ∼95-kDa fragment of clipped GPIbα, presumably consisting of the cytoplasmic tail, transmembrane and sialomucin-like domain, but lacking the ∼45-kDa N-terminal domain [2], and this band was lost when shedding was induced. Like shedding of GPVI [14] and GPV [11], GPIbα shedding was induced by the calmodulin inhibitor W7, with almost complete loss of intact GPIbα (and clipped GPIbα, recognized by rabbit antiglycocalicin antibody, but not SZ2, consistent with GPIbα lacking the N-terminal domain, data not shown) in the platelet fraction, and a corresponding increase in soluble GPIbα (glycocalicin) as well as a fragment of GPIbα of lower molecular weight detected in the platelet supernatant (Fig. 1A, upper and lower panels, respectively). Unlike GPVI (see below), soluble glycocalicin is present even at zero time (not affected by EDTA), but EDTA strongly inhibits the W7-induced loss of intact GPIbα, and increase in glycocalicin (Fig. 1A). This suggests the release of GPIbα from the platelet surface is less tightly regulated than shedding of GPVI. Consistent with previously published results [12], loss of intact GPIbα on platelets and generation of the ∼16-kDa digestion fragment is induced by PMA (that activates PKC) and the mitochondrial-targeting reagent CCCP (that mimics platelet aging) (Fig. 2A).

Figure 1.

 Immunoblotting of platelets with anti-glycoprotien (GP)Ibα antibodies. (A) Platelet pellet (upper panel) or supernatant (lower panel) fractions blotted with anti-GPIbα cytoplasmic tail IgG (upper panel) or antiglycocalicin IgG (lower panel), of untreated platelets, or platelets treated with vehicle (DMSO), W7 (150 μm) or EDTA (100 μm) for the indicated times at room temperature. In the lower panel, a sample of platelet lysates is included for molecular weight comparison of full-length receptor. Results are representative of triplicate experiments with separate donors. (B) Platelet pellets of untreated platelets where the metalloproteinase inhibitor GM6001 (100 μm) or EDTA (10 mm) was included in the anticoagulant and washing buffers throughout the isolation procedure. NT, no addition. Samples were blotted with anti-GPIbα cytoplasmic tail IgG.

Figure 2.

 Shedding of glycoprotein (GP)Ibα and GPVI from platelets. (A) Platelet pellets blotted with anti-GPIbα cytoplasmic tail IgG, of untreated platelets, or platelets treated with PMA (10 μm) or CCCP (100 mm) for the indicated times at room temperature. (B) Platelet pellets blotted with anti-GPVI mAb, 6B-12 (upper panel) or anti-GPVI cytoplasmic tail IgG (lower panel) of untreated platelets, or platelets treated with PMA (10 μm) or CCCP (100 mm) for the indicated times. (C) Platelet pellets blotted with anti-GPVI cytoplasmic tail IgG of platelets either untreated, or treated with N-ethylmaleimide (2 mm), convulxin (0.5 μg mL–1), CRP (0.5 μg mL–1) or W7 (150 μm) for the indicated times.

Secondly, like GPIbα, PMA and CCCP also induced shedding of GPVI as shown by blotting platelets with anti-GPVI cytoplasmic tail IgG, resulting in loss of ∼62-kDa intact GPVI and the appearance of a ∼10-kDa remnant fragment that remained platelet-associated (Fig. 2B). However, the proportion of ∼10-kDa fragment relative to intact GPVI in untreated platelets was significantly lower than observed for GPIbα under the same conditions (Fig. 2Bcf. 2A). The same ∼10-kDa remnant fragment was induced by treating platelets with W7, NEM, convulxin or CRP, although even after 1 h there was minimal detectable ∼10-kDa fragment in untreated platelets (Fig. 2C).

Thirdly, blotting lysates of platelets that had been treated with convulxin, NEM, W7, thrombin or TRAP with anti-GPV cytoplasmic tail IgG (Fig. 3) showed that GPV is shed by both metalloproteinase- and thrombin-dependent mechanisms. The anti-GPV cytoplasmic tail IgG blotted intact GPV and a ∼5-kDa platelet-associated remnant when platelets were treated with convulxin or NEM, and cleavage was inhibited by EDTA (Fig. 3A), suggesting it was metalloproteinase-dependent, consistent with previously published findings [11]. The same ∼5-kDa fragment was also induced by the calmodulin inhibitor, W7 (Fig. 3B). Interestingly, high doses of thrombin (>1 U mL−1) or longer exposure (60 min) to 0.1 U mL−1 thrombin generated both the ∼5-kDa fragment and an additional ∼25-kDa platelet-associated fragment (not seen in NEM- or W7-treated platelets), revealing a separate cleavage site (Fig. 3B). The observed levels of the ∼25-kDa fragment varied from donor to donor (compare right and left hand panels of Fig. 3B). Treatment with the PAR-1 thrombin receptor-specific agonist, TRAP, however, generated only the ∼5-kDa fragment (Fig. 3B). This suggests that the ∼5-kDa fragment was derived by a metalloproteinase-dependent mechanism induced by TRAP and thrombin and inhibitable by GM6001, whereas the 25-kDa band was only induced by thrombin treatment and enhanced in the presence of GM6001 (Fig. 3B). This implies that thrombin induces GPV shedding by different mechanisms depending on the dose and length of incubation.

Figure 3.

 Immunoblotting of platelets with anti-glycoprotein (GP)V cytoplasmic tail immunoglobulin G (IgG). (A) Platelet pellets blotted with anti-GPV cytoplasmic tail IgG, of untreated platelets (NT), or platelets treated with convulxin (0.5 μg mL−1) or N-ethylmaleimide (2 mm), in the absence or presence of EDTA (10 mm), for the indicated times at room temperature. Results are representative of triplicate experiments with separate donors. (B) Platelet pellet lysates blotted with anti-GPV cytoplasmic tail IgG, of untreated platelets (NT), or platelets treated with thrombin (0.1–10 U mL−1) or TRAP (1–30 μm) in the presence or absence of 0.1 mm GM6001, or W7 (150 μm) for 10 min (thrombin and TRAP) or 1 h (W7) at room temperature. The right hand panel shows platelet pellet lysates from a second donor blotted with anti-GPV cytoplasmic tail IgG after treatment with 10 U mL−1 thrombin for up to 1 h.

ADAM10 and ADAM17 digestion of GPVI-, GPIbα- and GPV-based peptides

Previous evidence showing ADAM17 is involved in shedding of GPIbα and GPV [8,9,11], and that NEM, which directly activates ADAM10/ADAM17 [19], induces shedding of GPVI [3], led us to test whether ADAM10 and/or ADAM17 cleave synthetic peptides based on extracellular membrane-proximal sequences and the predicted area of cleavage based on the molecular size of the relevant membrane-bound receptor remnant. For GPVI and GPIbα, the molecular weight of the intact receptor and the platelet-associated remnant fragment recognized by the anti-cytoplasmic tail IgG suggests the sheddase cleavage site must reside within a limited sequence between the transmembrane domain and the mucin-like domain. Therefore, synthetic peptides based on this region of GPVI (Thr221-Gly248), GPIbα (Lys450-Phe483) and GPV (Ser485-Phe508) were digested with either ADAM10 or ADAM17, and analyzed by mass spectrometry. As summarized in Table 1, ADAM10, but not ADAM17, cleaved the GPVI-based peptide at Arg242/Gln243, but not a corresponding peptide where Gln243 was replaced by Lys. This mutation in the P1′ cleavage site would be expected to be disruptive for proteolysis based on the positive charges in both the lysine residue and the Zn2+ at the metalloproteinase active site. In contrast, ADAM17, but not ADAM10, cleaved the GPIbα-based peptide at Gly464/Val465, but not a corresponding peptide where Val465 was replaced by Lys. Surprisingly, the GPV-based peptide was cleaved by both ADAM10 and ADAM17 at the same site, Pro493/Val494, and both cleavages were blocked when Val494 was replaced by Lys (Table 1).

Table 1.   ADAM10 and ADAM17 digestion of glycoprotein (GP)VI-, GPIbα- and GPV-based peptides. Synthetic peptides corresponding to membrane-proximal extracellular sequences of GPVI (Thr221-Gly248), GPIbα (Lys450-Phe483) and GPV (Ser485-Phe508) were treated with ADAM10 or ADAM17 as described in Methods. Resultant peaks were analyzed by mass spectrometry in the m/z range of 1000–9000
PeptideSequenceCut by human rADAM10?Cut by human rADAM17?Site

Shedding of wild-type or mutant GPVI and GPIbα from transfected cells

The effect of mutating the ADAM10- or ADAM17-dependent cleavage sites in full-length GPVI and GPIbα, respectively, was tested in GPVI-expressing RBL cells or GPIbα-expressing HEK293 cells. HEK293 cells have previously been reported to express ADAM17 [31] and expression of ADAM10 on RBL cells was confirmed by flow cytometry using a polyclonal anti-ADAM10 antibody (data not shown). Initially, it was shown that wild-type GPIbα was shed from GPIb-IX-transfected HEK293 cells, as shown by blotting supernatants of W7- or NEM-treated cells with anti-glycocalicin IgG (Fig. 4A). Shedding was blocked by either EDTA or GM6001, confirming it was metalloproteinase-dependent. In contrast to the peptide data, mutant GPIbα containing a Val465/Lys substitution (which inhibited ADAM17-dependent cleavage of the GPIbα-based peptide) expressed on HEK293 cells was shed normally in response to W7, compared with wild-type GPIbα (data not shown), suggesting additional structural elements in the full-length receptor determine ADAM17-dependent cleavage. However, deleting the sequence 463–479 (encompassing the Gly464/Val465 cleavage site for ADAM17 identified in the peptide studies and downstream sequence) abolished W7-induced GPIbα shedding (Fig. 4B). This deletion was based on a similar deletion abolishing ADAM17-mediated shedding of L-selectin, where point mutations were also ineffective at blocking shedding [32]. Flow cytometry with the anti-GPIbα mAb (WM23) confirmed similar expression of wild-type and mutant GPIbα (Fig. 4B). Wild-type GPVI was also shed from GPVI-transfected RBL cells treated with W7 or NEM, but in this case, shedding was strongly inhibited in Gln243/Lys mutant GPVI (Q243K), disrupting the Arg242/Gln243 cleavage site for ADAM10 (Fig. 4C). Flow cytometry with anti-GPVI antibody (1G5) against the GPVI ectodomain confirmed equivalent expression of wild-type and mutant GPVI (Fig. 4C).

Figure 4.

 Shedding of glycoprotein (GP)Ibα and GPVI from transfected cells, and effect of mutating ADAM17 or ADAM10 cleavage sites. (A) Supernatant fractions blotted with anti-glycocalicin IgG of wild-type GPIbα-expressing HEK293 cells treated with TS buffer, W7 (150 μm) or N-ethylmaleimide (NEM) (2 mm) in the absence or presence of EDTA (10 mm) or GM6001 (100 μm), for the indicated times at 37 °C. A sample of platelet lysate (PL) is included for molecular weight comparison of full-length receptor. (B) Flow cytometry (upper panel) of wild-type GPIbα- or mutant GPIbα (Δ463–479)-expressing HEK293 cells with anti-GPIbα monoclonal antibody (mAb), WM23 (open histogram), or control IgG (filled histogram). Lower panel shows supernatant fractions of wild-type GPIbα- or mutant GPIbα (Δ463–479)-expressing HEK293 cells either untreated (NT) or treated with W7 (150 μm), in the absence or presence of GM6001 (100 μm), for 1 h at 37°C then blotted with antiglycocalicin IgG. (C) Flow cytometry (upper panel) of wild-type GPVI- or mutant GPVI (Q243K)-expressing rat basophilic leukemia (RBL) cells with anti-GPVI mAb, 1G5 (filled histogram) or control IgG (open histogram). Lower panel shows wild-type GPVI- or mutant GPVI (Q243K)-expressing RBL cell lysates blotted with anti-GPVI mAb, 6B12 after no treatment (NT) or treatment with W7 (150 μm) or NEM (2 mm) for 1 h at 37°C.


In this study, we used antibodies against cytoplasmic domains of the platelet collagen receptor, GPVI, and the ligand-binding GPIbα subunit of GPIb–IX–V, to compare the mechanisms involved in their metalloproteinase-mediated shedding, and synthetic peptides and mutant receptors to identify potential cleavage sites. Together, the results suggest that GPIbα and GPVI are cleaved by different mechanisms, involving ADAM17 and ADAM10, respectively. In addition, we show that GPV, which regulates thrombin-dependent platelet activation via GPIbα, may be regulated by both ADAM10 and ADAM17, and that it is shed by both metalloproteinase- and thrombin-dependent mechanisms, depending on the concentration of thrombin.

On resting platelets, blotting with anti-GPIbα tail and anti-GPVI tail antibodies shows that under conditions where a significant portion of GPIbα is cleaved, all of the detectable GPVI is intact. This shedding of GPIbα is not because of adventitious proteolysis of the receptor during isolation of platelets, as there was no effect of including metalloproteinase inhibitor (GM6001) or EDTA in the wash-up steps. Loss of intact GPVI, and generation of the same ∼8-kDa platelet-associated fragment, is induced in a time-dependent manner by GPVI ligands (collagen, convulxin or CRP), or by treating platelets with calmodulin inhibitor W7, PMA (that activates PKC) or CCCP (a mitochondrial-targeting reagent that mimics platelet aging). GPIbα shedding is constitutive, but is also increased by treating platelets with CCCP, PMA or W7 [8,9,12; this study]. Together, these combined results suggest GPVI shedding is much more tightly regulated than GPIbα, and that different shedding mechanisms may be involved. GPVI is shed by GPVI ligands such as collagen, convulxin and CRP [14] thrombin [15,16], or by anti-GPVI antibodies [17,27]; however, under conditions where GPVI is completely lost, GPIbα is still detected on platelets. This difference in regulation may also reflect the fact that GPVI directly binds calmodulin at a juxtamembrane sequence of the cytoplasmic domain [22], whereas the cytoplasmic domain of GPIbβ and not GPIbα binds calmodulin [21]. Cell-matrix adhesion mediated by proteolytic regulation of the adhesion receptor, CD44, is controlled by independent pathways involving ADAM10 and ADAM17, regulated by Ca2+/calmodulin and PKC-dependent pathways, respectively [23]. Related pathways may control GPVI/GPIb–IX–V-dependent adhesion in platelets.

To address the question of which particular sheddases may be involved in GPIbα and GPVI shedding, synthetic peptides based on extracellular membrane-proximal sequences of GPVI (Thr221-Gly248) or GPIbα (Lys450-Phe483) were digested with either ADAM10 or ADAM17, and analyzed by mass spectrometry. ADAM10, but not ADAM17, cleaved the GPVI-based peptide at Arg242/Gln243, whereas ADAM17, but not ADAM10, cleaved the GPIbα-based peptide at Gly464/Val465 (Table 1). Peptides where the residue at the P1′ position (Gln243 in the GPVI-based peptide, and Val465 in the GPIbα-based peptide) were substituted by Lys blocked ADAM-dependent cleavage. In GPVI and GPIbα expressed on RBL or HEK293 cells, respectively, mutating the residue at the P1′ position to Lys blocked W7-induced shedding of GPVI, but not GPIbα. However, deleting a sequence spanning the Gly464/Val465 cleavage site in GPIbα (Δ463–479) inhibited W7-induced shedding. The lack of effect on shedding by the Val465/Lys mutation is not unexpected based on mutagenesis studies on the ADAM17 substrate, L-selectin [32], and may further reflect the less stringent regulation of GPIbα shedding compared with GPVI. In this regard, differences in efficiency of ADAM-dependent cleavage of synthetic peptides and full-length proteins have been discussed elsewhere [19].

Metalloproteinase-dependent shedding of GPV, like GPIbα and GPVI, is also induced by PMA, W7 and NEM, generating an ∼5-kDa platelet-associated remnant [11; this study]. A GPV-based synthetic peptide (Ser485-Phe508) was cleaved by both ADAM10 and ADAM17 at Pro493/Val494 (Table 1). ADAM17 has previously been shown to cleave GPV [11]; however, cleavage by ADAM10 is unexpected. In this regard, both ADAM10 and ADAM17 may be activated following treatment with various agonists (including W7, NEM, PMA or convulxin); however, GPVI is cleaved predominantly by ADAM10 whereas GPV may be cleaved by either ADAM10 or ADAM17. Consistent with this, GPV is shed by GPVI agonists, convulxin (this study) or CRP [11]. GPV is also a known substrate for thrombin, and treating platelets with thrombin induces both metalloproteinase- and thrombin-dependent GPV shedding, generating ∼5-kDa and ∼25-kDa fragments respectively, depending upon length of exposure to thrombin. At low thrombin concentrations for 10 min, indirect shedding of GPV by activation of endogenous platelet metalloproteinases (EDTA or GM6001 inhibitable) generates predominantly the ∼5-kDa fragment (also induced by W7, NEM, PMA and TRAP). At higher thrombin concentrations or with longer incubation times, both ∼5-kDa and ∼25-kDa fragments are generated. In both cases, loss of the GPV ectodomain would presumably facilitate thrombin-dependent platelet activation via GPIbα [7].

Finally, these mechanisms are also likely to contribute to plasma levels of soluble GPVI and GPIbα [9]. Ectodomain fragments of GPVI [33], GPIbα (glycocalicin) [34] and GPV [35] are present in normal human plasma, and may act as potential thrombotic markers or modulators. Increasing biochemical, cellular and clinical evidence supports a critical role for platelet GPVI levels in thrombus formation [10,13], and relationships between shedding mechanisms of GPVI and other platelet receptors such as GPIbα and GPV warrant further investigation.


The authors thank A. Aprico, C. Llerena, L. Moore and C. Berndt for expert laboratory assistance, and I. Smith and S. Reeve for mass spectrometry analysis. This work was supported by the National Health and Medical Research Council of Australia.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.