A novel viper venom metalloproteinase, alborhagin, is an agonist at the platelet collagen receptor GPVI.

The interaction of platelet membrane glycoprotein VI (GPVI) with collagen can initiate (patho)physiological thrombus formation. The viper venom C-type lectin family proteins convulxin and alboaggregin-A activate platelets by interacting with GPVI. In this study, we isolated from white-lipped tree viper (Trimeresurus albolabris) venom, alborhagin, which is functionally related to convulxin because it activates platelets but is structurally different and related to venom metalloproteinases. Alborhagin-induced platelet aggregation (EC50, <7.5 microg/ml) was inhibitable by an anti-alphaIIbbeta3 antibody, CRC64, and the Src family kinase inhibitor PP1, suggesting that alborhagin activates platelets, leading to alphaIIbbeta3-dependent aggregation. Additional evidence suggested that, like convulxin, alborhagin activated platelets by a mechanism involving GPVI. First, alborhagin- and convulxin-treated platelets showed a similar tyrosine phosphorylation pattern, including a similar level of phospholipase Cgamma2 phosphorylation. Second, alborhagin induced GPVI-dependent responses in GPVI-transfected K562 and Jurkat cells. Third, alborhagin-dependent aggregation of mouse platelets was inhibited by the anti-GPVI monoclonal antibody JAQ1. Alborhagin had minimal effect on convulxin binding to GPVI-expressing cells, indicating that these venom proteins may recognize distinct binding sites. Characterization of alborhagin as a GPVI agonist that is structurally distinct from convulxin demonstrates the versatility of snake venom toxins and provides a novel probe for GPVI-dependent platelet activation.

Viper venom proteins that activate platelets have played a significant role in elucidating mechanisms of platelet activation. Several groups have described a 50-kDa protein of the C-type lectin family, alboaggregin-A, from venom of the whitelipped tree viper, Trimeresurus albolabris, which binds to GPIb-IX-V (6 -9) and GPVI (10,11) and activates platelets by a mechanism that may involve one or both receptors. A related C-type lectin-like protein, aggretin from the Malayan pit viper, Calloselasma rhodostoma, has been reported to activate platelets by a mechanism that involves GPIb and GPIa-IIa but not GPVI (12). An analogous, ϳ85-kDa C-type lectin-like protein, convulxin, from the venom of the tropical rattlesnake, Crotalus durissus terrificus, activates platelets by a mechanism involving GPVI (5,(13)(14)(15).
In the present study, we isolated a protein termed alborhagin from T. albolabris venom that is functionally related to convulxin, because it is a potent agonist at GPVI, but that is a member of the metalloproteinase family. Evidence is presented to suggest that alborhagin binds to GPVI at a site distinct from that of convulxin. Alborhagin therefore represents a novel tool to further characterize GPVI.
Purification of Alborhagin-Throughout the purification procedure, fractions were assayed for their ability to induce platelet aggregation of citrated platelet-rich plasma and analyzed by SDS-polyacrylamide gel electrophoresis as previously described (7,21 7.4. Fractions containing an ϳ60-kDa protein (reduced and nonreduced) that eluted from the heparin-agarose column were dialyzed into 5 mM NaH 2 PO 4 , pH 6.8 and loaded at 25 ml/h onto a 10 ϫ 1-cm hydroxylapatite column equilibrated in the same buffer. Bound protein was eluted with a linear 5-200 mM NaH 2 PO 4 , pH 6.8 gradient. Fractions containing alborhagin were pooled, concentrated using an ultrafiltration device (Amicon, Danvers, MA) fitted with a YM30 membrane, and dialyzed into TS buffer. The concentration of purified protein was estimated using the BCA method with bovine serum albumin as standard according to the manufacturer's instructions (Pierce).
Sequence Analysis-N-terminal sequence analysis was performed as previously described (7,17). For internal sequence analysis, alborhagin (0.58 mg) was dialyzed into distilled water and digested with trypsin (0.1 mg of trypsin/mg of protein) overnight at 37°C. Tryptic fragments were separated by reverse-phase high pressure liquid chromatography, eluted by a linear 0 -60% (v/v) acetonitrile gradient, and sequenced as previously described (17).
Alborhagin-dependent Digestion of Fibrinogen-Digestion of human fibrinogen (Kabivitrum, Stockholm, Sweden) at a final concentration of ϳ100 g/ml in TS buffer by alborhagin (10 g/ml, final concentration) for 30 min at 22°C was carried out in the presence of EDTA (10 mM, final concentration) or CaCl 2 (10 mM, final concentration) according to a published method (22).
Platelet Aggregation-Platelet aggregation of human platelet-rich plasma (0.32% citrate, final concentration) was performed using a whole blood Lumiaggregometer (Chronolog, Havertown, PA) as previously described (7,17,21). Alborhagin was added to platelets at a final concentration of 2.5-25 g/ml. In some assays, the anti-␣IIb␤3 antibody CRC64 at 20 g/ml or mocarhagin at 10 g/ml was pre-incubated with the platelets for 5 or 6 min, respectively, at 37°C prior to the addition of alborhagin. Other assays included EDTA at a final concentration of 10 mM.
The effect of PP1 (10 M) on alborhagin-induced aggregation of washed human platelets, isolated as previously described (23), was determined at 37°C in the presence of indomethacin (10 M) and apyrase (2 units/ml). Alborhagin-dependent aggregation of mouse platelets (1 ϫ 10 8 /ml) in the absence or presence of the monoclonal antibody JAQ1 at a final concentration of 10 g/ml was carried out essentially as previously described (19).
For immunoprecipitation studies, platelets (2 ϫ 10 8 /ml) were preincubated with vehicle (0.1% Me 2 SO) or 10 M PP1 for 5 min at 37°C prior to the addition of 3 g/ml alborhagin or 1 g/ml convulxin for 90 or 30 s, respectively. Platelets were then lysed by adding an equivalent volume of ice-cold lysis buffer (20 mM Tris-HCl, pH 7.3, 300 mM NaCl, 2 mM EDTA, 2 mM EGTA, 2% (v/v) Nonidet P-40, 2 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml aprotinin, 1 g/ml pepstatin A), centrifuged at 15,000 ϫ g for 10 min to remove detergent-insoluble material, and pre-cleared with 30 l of a 50% suspension of protein A-Sepharose in TBS-T buffer (20 mM Tris-HCl, 137 mM NaCl, 0.1% (v/v) Tween 20, pH 7.3). Samples were then immunoprecipitated with monoclonal anti-PLC␥2 antibody and 30 l of protein A-Sepharose for 2 h at 4°C. The beads were washed once with lysis buffer and three times with TBS-T buffer, then solubilized in electrophoresis sample buffer, and boiled for 10 min. Immunoblotting with the anti-phosphotyrosine antibody 4G10 was carried out as described above. Membranes were then stripped and reprobed with anti-PLC␥2 antibody.
Alborhagin-dependent Ca 2ϩ Mobilization in GPVI-transfected Jurkat Cells-GPVI-transfected Jurkat cells were transfected with GPVI using methods similar to those described above for K562 cells (10,23). Jurkat cells are human T lymphocytic cells that do not express endogenous GPVI or FcR␥. Ca 2ϩ flux in response to alborhagin (10 g/ml) in untransfected or GPVI-transfected cells was measured as previously described (24). The effect of alborhagin (0.1-100 g/ml, final concentration) on binding of convulxin (1 g/ml) to GPVI-expressing Jurkat cells was assessed using the methods described above for K562 cells.

RESULTS
Purification and Characterization of Alborhagin and Its Autolytic Fragment-In the course of our purification of the platelet agonist alboaggregin-A from the venom of the white-lipped tree viper, T. albolabris (7), we identified a structurally distinct protein termed alborhagin that was also a potent platelet agonist. Alborhagin had an apparent molecular mass on SDSpolyacrylamide gels of ϳ60 kDa under reducing (Fig. 1A) and nonreducing (data not shown) conditions and bound strongly to heparin (eluted at Ͼ0.5 M sodium chloride). Several lines of evidence suggested that alborhagin was a member of the metalloproteinase-disintegrin family. First, it was immunoreactive toward an antibody against the cobra venom metalloproteinase-disintegrin, mocarhagin (Fig. 1B). This antibody is also cross-reactive with jararhagin (data not shown). Second, like other venom metalloproteinase-disintegrins (22,25), alborhagin showed metalloproteinase activity toward the fibrinogen ␣-chain in the presence of Ca 2ϩ , but there was no significant digestion in the presence of EDTA (Fig. 1C). Third, amino acid sequences of tryptic fragments corresponding to a total of 36 residues ( Fig. 2A) were analogous to sequences within the metalloproteinase domain of the venom metalloproteinase-disintegrins jararhagin (26), HR1B (27), and MT-d (28). Fourth, we obtained the sequence from the disintegrin domain by N- FIG. 1. Characterization of alborhagin. A, SDS-polyacrylamide (5-20%) gel of purified alborhagin electrophoresed under reducing conditions and stained with Coomassie Blue. B, Western blot of mocarhagin (moc) (5 g) and alborhagin (albo) (5 g) electrophoresed on SDSpolyacrylamide (5-20%) gels under reducing conditions, transferred to nitrocellulose, probed with rabbit anti-mocarhagin antibody, and visualized using a peroxidase-coupled goat anti-rabbit IgG and the ECL reagent. C, digestion of fibrinogen (100 g/ml) by 10 g/ml alborhagin (Alborh.) at 22°C for 80 min in the presence of either 10 mM EDTA or 10 mM Ca 2ϩ .
terminal sequencing of a 42-kDa autolytic fragment of alborhagin (Fig. 3A). Autolytic digestion of alborhagin was inhibited in the presence of EDTA (Fig. 3B), suggesting that it involved the metalloproteinase activity of alborhagin. The 42-kDa fragment did not bind heparin (data not shown) and was isolated in the flow-through of a heparin-agarose column to allow separation of any undigested alborhagin. The N-terminal sequence of the 42-kDa fragment (Fig. 2B) was similar to the region at the boundary between the metalloproteinase and disintegrin domains of jararhagin (26). This sequence was conserved in MT-d (Fig. 2B), which also undergoes autodigestion (28).
Alborhagin-dependent Platelet Aggregation-Alborhagin induced aggregation in platelet-rich plasma, with maximal activity at ϳ7.5 g/ml (Fig. 4A). Aggregation was strongly inhibited by the anti-␣IIb␤3 antibody CRC64, which blocks ligand binding to ␣IIb␤3 (18), and by EDTA (Fig. 4A), suggesting that aggregation was ␣IIb␤3-dependent and involved alborhagininduced platelet activation. In the presence of EDTA, alborhagin still induced a platelet shape change (Fig. 4A), suggesting that alborhagin could induce platelet activation independently of its metalloproteinase function. In contrast to intact alborhagin, however, the ϳ42-kDa alborhagin fragment corresponding to the disintegrin and C-terminal regions minus the metalloproteinase domain (see above) did not induce platelet aggregation at a final concentration of 10 g/ml in platelet-rich plasma (data not shown). Together, these results suggested that an intact metalloproteinase domain, but not proteolysis, is necessary for alborhagin-induced platelet activation.
Pre-treating platelets with the cobra venom metalloproteinase mocarhagin had no effect on alborhagin-dependent platelet aggregation (data not shown). Mocarhagin has previously been shown to cleave the GPIb␣ chain between Glu-282 and Asp-283 and abolish binding of vWF to the GPIb-IX-V complex on platelets and vWF-dependent platelet aggregation (17). The lack of effect of mocarhagin on alborhagin activity therefore suggests that alborhagin does not act through GPIb-IX-V on platelets.
Alborhagin-induced Signaling in Platelets-Treating [ 32 P]orthophosphate-loaded platelets with alborhagin at 10 g/ml resulted in phosphorylation of the previously characterized protein kinase C substrates p47/pleckstrin (30) and myosin light chain (31) on a time scale comparable with that induced by collagen (data not shown). Additionally, alborhagin stimulated the tyrosine phosphorylation of a range of proteins, measured using an anti-phosphotyrosine antibody (Fig. 5A). This platelet phosphorylation profile resembled that induced by collagen or convulxin (Fig. 5B), a venom protein of the C-type lectin family previously shown to activate platelets by an interaction with GPVI (13,14). For both alborhagin and convulxin stimulation, prominent phosphorylated bands were observed at ϳ135, ϳ72, ϳ36, and ϳ12 kDa. The time course for convulxin-induced phosphorylation peaked at 15 s before declining. In contrast, alborhagin stimulated a slower, sustained increase in tyrosine phosphorylation reminiscent of responses to collagen and collagen-related peptide. The transient nature of the increase in tyrosine phosphorylation induced by convulxin is thought to be related to its more powerful stimulatory action (32). PLC␥2 was identified as being specifically phosphorylated in response to both alborhagin and convulxin, a response that was blocked by PP1 (Fig. 6) along with the increase in whole cell phosphorylation (data not shown). Of the other phosphorylated bands, it is probable that the 12, 36, and 72 kDa bands phosphorylated both by alborhagin and convulxin correspond to the FcR␥, LAT, and Syk/SLP76, respectively, based on previous studies (13,14,32).
Interaction of Alborhagin with GPVI-expressing Cells-The similar pattern of tyrosine phosphorylation and sensitivity to PP1 in platelets treated with alborhagin implicated GPVI in the mechanism of alborhagin activity. We therefore examined the activity of alborhagin on two cell lines transfected with GPVI. In FcR␥-expressing K562 cells transfected with GPVI (10), alborhagin induced concentration-dependent GPVIdependent phosphorylation of Syk, whereas there was no response in mock-transfected cells (Fig. 7A). K562 cells contain no endogenous GPIb-IX-V or GPVI expression, as assessed by flow cytometry (data not shown), suggesting that GPVI was specifically required for alborhagin-induced Syk phosphorylation in this system. Alborhagin also induced elevation of cytosolic Ca 2ϩ , specifically in GPVI-transfected Jurkat cells but not in untransfected cells (Fig. 7B). Jurkat cells do not express GPVI or FcR␥, although the role of the latter is taken by the T cell receptor chain. 2 These results provide additional evidence that alborhagin targets GPVI.
Effect of Alborhagin on Convulxin Binding to GPVI-Binding of convulxin (1 g/ml) to GPVI-expressing K562 cells was partially inhibited by alborhagin in a dose-dependent manner. There was ϳ40% inhibition at a final concentration of 10 g/ml alborhagin, with no increased inhibition up to 100 g/ml (Fig.   8A). Similar experiments using GPVI-expressing Jurkat cells showed that there was only minimal displacement of convulxin binding by 10 g/ml alborhagin, with no further inhibition at 100 g/ml (Fig. 8A). Together, these data suggest that although alborhagin and convulxin may both bind GPVI, they may recognize separate binding sites. This is supported by studies on mouse platelets using a rat monoclonal antibody that binds GPVI, JAQ1. JAQ1 has previously been shown to specifically 2 D. Tulasne, T. Bori, and S. P. Watson, unpublished observation. inhibit aggregation of mouse platelets by collagen-related peptide and low concentrations of collagen but not by convulxin or high concentrations of collagen (19,33,34). JAQ1 inhibited alborhagin-induced shape change and aggregation of mouse platelets (Fig. 8B), suggesting that it binds to a site that is the same as, or located close to, that used by collagen-related peptide but that is distinct from that used by convulxin. Convulxin binding is not blocked by JAQ1 (34), implying that alborhagin binds to a distinct epitope. In this regard, the effect of JAQ1 is consistent with alborhagin binding to the collagenrelated peptide binding site, whereas convulxin may bind to the site used by the second binding site in collagen (33). DISCUSSION In the present study, we isolated a novel protein termed alborhagin from T. albolabris venom that is functionally related to alboaggregin-A and convulxin, because it potently activated platelets by an interaction with GPVI; but in contrast to these C-type lectin-like proteins, alborhagin was structurally different. Several lines of evidence suggested that alborhagin may be a member of the metalloproteinase-disintegrin family. First, the molecular mass of alborhagin on SDS-polyacrylamide gels (ϳ60 kDa; nonreduced and reduced) and its metalloproteinase activity (EDTA-inhibitable) toward fibrinogen ␣-chain were consistent with other metalloproteinase-disintegrins (17,22,25,26,35). Second, alborhagin was immunoreactive toward an antibody against the cobra venom metalloproteinase mocarhagin (17). Third, the sequences of four peptides throughout the length of alborhagin revealed similarity to jararhagin, HR1B, and MT-d of the metalloproteinase-disintegrin family (26 -28). This combined evidence suggests that alborhagin is very likely to be a metalloproteinase-disintegrin, although in the absence of the full sequence, we cannot exclude the possibility that it is a closely related protein. Like MT-d, a venom metalloproteinase-disintegrin from Agkistrodon halys brevicaudus (28), alborhagin underwent autolysis, yielding an ϳ42-kDa fragment. Autolysis was prevented by the presence of EDTA, suggesting that it involved the metalloproteinase activity of alborhagin. The N-terminal sequence of the 42-kDa fragment corresponded to sequences at the N terminus of two naturally occurring jararhagin-derived fragments, one-chain botrocetin (36) and jaracetin (37). This result supports the possibility that autodigestion might be a general mechanism for processing venom metalloproteinases.
Several lines of evidence demonstrated that alborhagin was a potent platelet agonist and that, like convulxin, it appeared to target GPVI. First, alborhagin induced platelet aggregation that was ␣IIb␤3-dependent, because it was inhibitable by a blocking anti-␣IIb␤3 monoclonal antibody (CRC64). Second, alborhagin activated protein kinase C, as shown by phosphorylation of p47/pleckstrin and myosin light chain (30,31). Third, alborhagin induced a tyrosine phosphorylation profile similar to that induced by convulxin (this study; 12-14, 38 -40). One of these tyrosine-phosphorylated proteins, PLC␥2, was identified as being phosphorylated in response to both convulxin and alborhagin. PLC␥2 phosphorylation induced by either agonist was inhibitable by the Src kinase family inhibitor PP1. Importantly, PP1 also blocked both alborhagin-and convulxin-dependent platelet aggregation.
Additional evidence that alborhagin could target GPVI was provided by studies showing that alborhagin, like convulxin, induced GPVI-dependent phosphorylation of Syk in FcR␥expressing K562 cells cotransfected with GPVI but not in untransfected cells. Alborhagin also induced elevation of cytosolic Ca 2ϩ in Jurkat cells expressing recombinant GPVI but had no effect in untransfected cells. Interestingly, although both alborhagin and convulxin appeared to target GPVI, alborhagin only partially inhibited (by ϳ40%) convulxin binding to recombinant GPVI expressed on K562 cells and had a negligible effect on binding to Jurkat cells, suggesting that the proteins may recognize separate sites. This was supported by the observation that alborhagin-dependent aggregation of mouse platelets was inhibited by an anti-mouse GPVI monoclonal antibody, JAQ1. This antibody also blocks aggregation to collagen-related peptide and low concentrations of collagen but not convulxin (34).
Finally, the observation that structurally distinct viper venom proteins, convulxin and alboaggregin-A of the C-type lectin family (10,11,13,14) and alborhagin (this study), may target distinct sites at the same receptor (GPVI) parallels the observation that other viper venom proteins target distinct sites on vWF. Two-chain botrocetin of the C-type lectin family and jararhagin of the metalloproteinase-disintegrin family interact with vWF at distinct sites, enabling it to bind GPIb-IX-V (35). These latter observations may be related to recent evidence suggesting that C-type lectin-like proteins and metalloproteinase-disintegrins might be derived from a common precursor (41).
In conclusion, we have identified a novel viper venom protein, alborhagin, that is a potent platelet agonist. Alborhagin is functionally related to convulxin, because it appears to target GPVI, but is structurally different and appears to target the collagen receptor at a distinct site, thereby making it a novel probe for analysis of GPVI-dependent platelet activation.