Aggretin, a Heterodimeric C-type Lectin from Calloselasma rhodostoma (Malayan Pit Viper), Stimulates Platelets by Binding to a 2 b 1 Integrin and Glycoprotein Ib, Activating Syk and Phospholipase C g 2, but Does Not Involve the Glycoprotein VI/Fc Receptor g Chain Collagen Receptor*

Aggretin, a potent platelet activator, was isolated from Calloselasma rhodostoma venom, and 30-amino acid N-terminal sequences of both subunits were determined. Aggretin belongs to the heterodimeric snake C-type lectin family and is thought to activate platelets by binding to platelet glycoprotein a 2 b 1 . We now show that binding to glycoprotein (GP) Ib is also required. Aggre-tin-induced platelet activation was inhibited by a monoclonal antibody to GPIb as well as by antibodies to a 2 b 1 . Binding of both of these platelet receptors to aggretin was confirmed by affinity chromatography. No binding of other major platelet membrane glycoproteins, with formaldehyde in phosphate-buffered saline 30 Platelets were washed resuspended in m of phosphate-buffered and platelets were analyzed by cytometry (FACScan; Becton Dickinson, Heidelberg, Germany). Nonspecific background labeling was determined using control platelets treated with 10 m M GRGDPS (No- vabiochem, Bad Soden, Germany) to prevent specific fibrinogen binding. second antibody, and chemiluminescence.

Aggretin, a potent platelet activator, was isolated from Calloselasma rhodostoma venom, and 30-amino acid N-terminal sequences of both subunits were determined. Aggretin belongs to the heterodimeric snake Ctype lectin family and is thought to activate platelets by binding to platelet glycoprotein ␣ 2 ␤ 1 . We now show that binding to glycoprotein (GP) Ib is also required. Aggretin-induced platelet activation was inhibited by a monoclonal antibody to GPIb as well as by antibodies to ␣ 2 ␤ 1 . Binding of both of these platelet receptors to aggretin was confirmed by affinity chromatography. No binding of other major platelet membrane glycoproteins, in particular GPVI, to aggretin was detected. Aggretin also activates platelets from Fc receptor ␥ chain (Fc␥)-deficient mice to a greater extent than those from normal control mice, showing that it does not use the GPVI/Fc␥ pathway. Platelets from Fc␥-deficient mice expressed fibrinogen receptors normally in response to collagen, although they did not aggregate, indicating that these platelets may partly compensate via other receptors including ␣ 2 ␤ 1 or GPIb for the lack of the Fc␥ pathway. Signaling by aggretin involves a dose-dependent lag phase followed by rapid tyrosine phosphorylation of a number of proteins. Among these are p72 SYK , p125 FAK , and PLC␥2, whereas, in comparison with collagen and convulxin, the Fc␥ subunit neither is phosphorylated nor coprecipitates with p72 SYK . This supports an independent, GPIb-and integrin-based pathway for activation of p72 SYK not involving the Fc␥ receptor.
Platelet-collagen interactions are integral to primary hemostasis (1,2). Resting platelets using several receptors adhering to subendothelium of damaged blood vessels are activated and spread to provide finally a new nonthrombogenic surface until vasculature regeneration occurs. Reversible binding between GPIb-V-IX 1 and von Willebrand factor, associated with collagen, is crucial to slow down the platelet (especially under high shear) so that it can bind more firmly via other receptors (3,4). This mechanism strongly parallels that of the selectins in leukocyte adhesion (5). Another important receptor is the ␣ 2 ␤ 1 integrin, which is essential for anchoring the platelet to collagen in the subendothelium (6) and for linking to the platelet cytoskeleton to prevent the receptor being torn from the membrane by the forces that it has to withstand. Activation induces the release of storage granules and the expression of new receptors on the platelet surface (7) as well as changes in other receptors such as the fibrinogen receptor, ␣ IIb ␤ 3 , which is critical for spreading. Although GPIb-V-IX and ␣ 2 ␤ 1 also participate in signaling to the platelet interior (8,9), recent studies, particularly in patients with platelet receptor deficiencies, have implicated GPVI/Fc␥ as a major collagen receptor for platelet activation (10 -12). Patients with platelets lacking any one of these receptors (GPIb-V-IX, ␣ 2 ␤ 1 , or GPVI/Fc␥) have increased bleeding times, and platelet adhesion to subendothelium or collagen is defective under flow conditions (13,14).
Because adhesion is rapid, the function of the individual receptor classes is difficult to assess. Inhibition of any of them can prevent platelet activation but does not indicate which synergies provide the final activating signal. In analyzing the mechanisms of each of these steps, reagents that bind to and activate platelets via individual receptor types are important tools. Convulxin, a C-type lectin from the venom of Crotalus durissus terrificus, the tropical rattlesnake, activates platelets by binding to and clustering the GPVI/Fc␥ receptor (15). Several other snake venom proteins are potent platelet activators and may act via collagen-like mechanisms. Some of these, including trimucytin from Trimeresurus mucrosquamatus (16) and aggretin from Calloselasma rhodostoma (17), have been reported to involve ␣ 2 ␤ 1 on platelets as receptor. We isolated a protein from C. rhodostoma venom that is a powerful platelet activator. This is most likely aggretin as described by Huang et al. (17), based on the molecular mass, sequence of subunits (18) and properties in activating platelets. Like aggretin it also belongs to the snake C-type lectin family. A C-type lectin from the same species, termed rhodocytin, was described later (19) with N-terminal sequences identical to those of aggretin, but it was reported to have some different properties. The mechanism of rhodocytin action on platelets was recently investigated (20,21), showing that rhodocytin interacts with the ␣ 2 -subunit of ␣ 2 ␤ 1 on the platelet surface. On the other hand, Eble et al. (22) have reported that rhodocytin does not bind to a recombinant ␣ 2 ␤ 1 complex. We now show that aggretin activates platelets via interaction with GPIb as well as ␣ 2 ␤ 1 ; however, GPVI/ Fc␥ is not required. This mechanism via GPIb and ␣ 2 ␤ 1 may also be relevant to activation of platelets via collagen.
Purification and Sequence Analysis of Aggretin Subunits-Dithiothreitol (1/20 volume of 650 mM) was added to 1 volume of 300 g/ml aggretin in 6 M guanidine HCl, 0.1 M Tris, pH 8.0, and incubated at 45°C for 30 min, followed by 1/80 volume of 4-vinylpyridine and incubation at room temperature for 1 h. 1/20 volume of 10% trifluoroacetic acid was added to the sample, and modified aggretin subunits were isolated by reverse phase HPLC on a wide pore C4 column (4.6 ϫ 250 mm; J. T. Baker) using an acetonitrile gradient (0.1% trifluoroacetic acid). N-terminal sequencing of S-pyridylethylated ␣and ␤-subunits of aggretin was done on an Applied Biosystem model 477 A pulsed liquid phase protein sequencer with model 120 A on-line phenylthiohydantoin amino acid analyzer.
Biotinylation of Aggretin-Purified aggretin was dialyzed against 10 mM sodium phosphate buffer, pH 8.0. Biotinamidocaproate N-hydroxysuccinimide ester in Me 2 SO (2 mg/ml) was added to aggretin at a molar ratio 2:1. The mixture was incubated at room temperature for 2 h. Biotin-aggretin conjugate was separated from free biotin by gel filtration on a Sephadex G-10 column.
Preparation of Washed Platelets, Platelet Aggregation, and Immuno-precipitations-Human platelets were isolated from buffy coats, less than 20 h after collection, obtained from the Central Laboratory of the Swiss Red Cross Blood Transfusion Service (15). Platelets were resuspended at 5 ϫ 10 8 platelets/ml in 20 mM Hepes, 140 mM NaCl, 4 mM KCI, 1 mM MgCl 2 ,1 mM CaCl 2 , 5.5 mM glucose, pH 7.4. For immunoprecipitation, aliquots (700 l, 5 ϫ 10 8 platelets/ml) of control, resting as well as activated platelets were solubilized in Hepes buffer containing 1.2% Triton X-100 with 1 mM phenylmethylsulfonyl fluoride, 2 mM EDTA, 2 mM N-ethylmaleimide, and 2 mM sodium orthovanadate. After centrifugation, platelet lysates precleared with protein A-Sepharose were stirred for 2 h with specific antibodies before adding 20 l of protein A-Sepharose followed by 6 h of incubation. Preparation of Triton X-100 Platelet Lysates, Wheat Germ Agglutinin Affinity Chromatography, and Biotinylated Aggretin-Avidin-Sepharose Affinity Chromatography-Human platelets were isolated from buffy coats as described above but in the presence of 10 M Iloprost. Washed platelets were diluted with phosphate-buffered saline to 5 ϫ 10 9 /ml and solubilized in phosphate-buffered saline containing 1.2% Triton X-100,1 mM phenylmethylsulfonyl fluoride, 100 M leupeptin, 2 mM N-ethylmaleimide, and 2 mM sodium orthovanadate with or without 5 mM EDTA. After centrifugation (40,000 ϫ g, 1 h, 4°C) the supernatant was applied to a column of wheat germ agglutinin-Sepharose 4B equilibrated with 130 mM NaCl, 10 mM Tris/HCl, pH 7.4 (buffer D). The column was washed thoroughly with buffer D containing 0.2% octanoyl-N-methylglucamide. The bound material was eluted with 2.5% N-acetylglucosamine in 10 mM Tris/HCl, 30 mM NaCl, pH 7.4 (buffer E) containing 0.2% octanoyl-N-methylglucamide. Biotinylated aggretin A was added to the pooled fractions containing eluted membrane glycoproteins, and after 2 h of incubation avidin-Sepharose was added to the mixture. After further incubation avidin-Sepharose was washed thoroughly with buffer D containing 0.2% octanoyl-N-methylglucamide. The avidin-Sepharose with bound biotinylated aggretin A and platelet proteins was boiled for 1 min with buffer E containing 1% SDS. Eluted proteins were separated by electrophoresis and transferred to the PVDF membrane.
FcR␥ Chain-deficient Mice-FcR␥ chain-deficient C57BL/6 (B6) mice have been previously described (25) and were kindly provided by Dr. Olle Korsgren (26). Normal B6 mice were used as controls. Whole blood was collected from normal and Fc␥-deficient mice anesthetized with phenobarbital by puncturing the inferior vena cava with heparinized syringes at a final concentration of 25 units of heparin/ml of blood.
Flow Cytometric Analysis of Fibrinogen-FITC Binding to Aggretin, Convulxin, and Collagen-activated Mouse Platelets-Mouse plateletrich plasma was diluted to 2.5 ϫ 10 7 platelets/ml with Tyrode's solution, buffered to pH 7.4, to minimize the formation of platelet aggregates and was preincubated for 3 min at room temperature with 150 g/ml fibrinogen-FITC (saturating concentration). Platelet suspension (100 l) was added to 20 l of the platelet agonist solution (aggretin, convulxin, or collagen). Platelet activation was stopped after 120 s by fixing with formaldehyde in phosphate-buffered saline for 30 min. Platelets were washed and resuspended in 100 l of phosphate-buffered saline, and 10 4 single platelets were analyzed by flow cytometry (FACScan; Becton Dickinson, Heidelberg, Germany). Nonspecific background labeling was determined using control platelets treated with 10 mM GRGDPS (Novabiochem, Bad Soden, Germany) to prevent specific fibrinogen binding.

RESULTS
Aggretin Is a Heterodimeric C-type Lectin-Aggretin was purified from lyophilized C. rhodostoma venom by gel filtration and ion-exchange chromatography. The final product gave 60-, 28-, and 13-14-kDa bands under nonreduced conditions and bands at 14 and 12 kDa under reduced conditions, by SDS-PAGE/silver staining analysis (Fig. 1). Gel filtration under nondenaturing conditions gave a peak at 60 kDa, suggesting that a tetrameric form (dimer of heterodimer) is the principle native form. Reverse phase HPLC gave two forms, one of which was more hydrophobic and eluted later from the HPLC column. Reduced and S-pyridylethylated aggretin was separated by HPLC into ␣and ␤-subunits (Fig. 1). Both the less and the more hydrophobic forms of aggretin gave similar subunits, but these forms also showed differences in hydrophobicity. The ␣-subunits and ␤-subunits, respectively, of the two forms had identical N-terminal sequences. The assignment to the ␣-subunit and the ␤-subunit was based upon the nomenclature used for the other heterodimeric snake C-type lectins where the ␣-subunit is defined as the larger. N-terminal sequencing gave the 30-amino acid sequences GLEDCDFGWSPYDQH-CYQAFNEQKTWDEAE for the ␣-subunit and DCPSGW-SSYEGHCYKPFNEPKNWADAERFC for the ␤-subunit. The sequences of both subunits shows full identity to the sequence of the cloned aggretin established by Chung et al. (18). Aggretin is a heterodimeric C-type lectin with strong sequence similarity to other venom C-type lectins interacting with platelet receptors (Fig. 2).
Aggretin Acts via GPIb-Aggretin is a powerful platelet agonist and induced maximal platelet aggregation at concentrations in the 40 -60 ng/ml range. Apart from convulxin, most venom C-type lectins investigated so far that affect platelet responses do so by binding to GPIb so that it was important to investigate this as receptor. Antibodies to GPIb (lb-23, SZ-2, and AP-1) did not inhibit aggretin-induced platelet activation. Echicetin, a snake venom C-type lectin that binds to GPIb and blocks platelet agonists acting via GPIb showed slight inhibitory effects in about 30% of the experiments. However, the MoAb VM16d, directed against the thrombin-binding site in GPIb, was able to inhibit aggretin-induced platelet aggregation completely in a dose-dependent way (Fig. 3A) that was reproducible with platelets from different donors. The protein-tyrosine phosphorylation occurring in platelets after activation with aggretin was also blocked by VM16d (Fig. 3B).
Aggretin Binds Both ␣ 2 ␤ 1 and GPIb from Platelet Lysates-No platelet receptors bound specifically to biotinylated aggretin/avidin-Sepharose or avidin-Sepharose as a control following affinity chromatography of platelet lysate prepared in the presence of EDTA. However, affinity chromatography of platelet lysate prepared in the absence of EDTA gave specific binding of ␣2␤ 1 and GPIb to biotinylated aggretin/avidin-Sepharose. Avidin-Sepharose 4B alone did not bind any membrane proteins from the platelet lysate (Fig. 4).
Aggretin Activates Mouse Platelets from the Fc␥-deficient Line and Therefore Does Not Act via GPVI/Fc␥-GPVI requires Fc␥ for platelet activation by collagen, and platelets from mice rendered deficient for Fc␥ were reported not to respond to collagen or collagen-like peptides (25). Aggretin was therefore tested on normal and Fc␥-deficient mouse platelets. A flow cytometric method was used to measure activation of GPIIb-IIIa (fibrinogen-binding sites) as a parameter of platelet stimulation using FITC-labeled fibrinogen. As expected the Fc␥-deficient mouse platelets did not respond to convulxin, the GPVI-specific C-type lectin, as agonist, whereas the control platelets were activated (Fig. 5A). The Fc␥-deficient mouse platelets gave a stronger response to aggretin than the control platelets (Fig. 5B) and also aggregated, whereas although both types of platelets gave a similar FITCfibrinogen binding response with collagen (Fig. 5C), the Fc␥deficient platelets did not give detectable aggregates. The data shown are the means of results from three different experiments.
Tyrosine Phosphorylation in Platelets Induced by Aggretin Compared with Collagen and Convulxin- Fig. 6 shows a tyrosine phosphorylation time range for platelets activated by 100 ng/ml aggretin compared with 1.5 g/ml collagen and 30 ng/ml convulxin. Comparison of the proteins phosphorylated on tyrosine showed several differences between platelets activated by aggretin and the others. A band at 90 kDa is persistently phosphorylated in aggretin, rapidly transiently phosphorylated in convulxin-activated, but not phosphorylated in collagen platelets. The bands at 36 -38 kDa, which are closely associated with signaling via GPVI/Fc␥ are very strongly phosphorylated in platelets activated by convulxin and more weakly in those activated by collagen as already described (15). In aggretinactivated platelets either the tyrosine-phosphorylated bands in the 36 -38-kDa range were very weakly phosphorylated or a different, slightly lower band was phosphorylated in response to this agonist. Unlike platelet responses to convulxin, neither the aggregation nor the tyrosine phosphorylation response to aggretin is virtually instantaneous. There is a clear dose-dependent lag phase during which the phosphorylation of p72 SYK increases slowly (Fig. 7).
Signal Transduction by Aggretin involves p72 SYK and PLC␥2 but Not Fc␥-Platelets activated by either aggretin, convulxin, or collagen as above were solubilized in Triton X-100 and centrifuged, and the supernatant was used for immunoprecipitation with antibodies to p72 SYK and PLC␥2 (Fig. 8, A and B). Both p72 SYK and PLC␥2 were activated and tyrosine-phosphorylated by all three agonists. However, tyrosine-phosphorylated Fc␥ was directly co-immunoprecipitated with p72 SYK platelets activated by collagen and convulxin but not in those activated by aggretin.
Aggretin Causes Strong Activation of p125 FAK , Even When GPIIb-IIIa Is Blocked-Platelets activated by aggretin, convulxin, or collagen in the presence or the absence of GPIIb-IIIa inhibitor (Ro44 -9883, 1 M) were solubilized in Triton X-100 and centrifuged, and the supernatant was used for immunoprecipitation with antibodies to p125 FAK (Fig. 8C). Convulxin with GPIIb-IIIa inhibitor gave little phosphorylation of p125 FAK , but collagen and aggretin induced phosphorylation of p125 FAK in the presence of the inhibitor, with aggretin having the strongest effect. Aggretin was reduced with dithiothreitol, then treated with vinylpyridine, and loaded on a C-4 HPLC column. Peak 1, aggretin ␤-subunit, hydrophilic form; peak 2, aggretin ␤-subunit, hydrophobic form; peak 3, aggretin ␣-subunit, hydrophilic form; peak 4, aggretin ␣-subunit, hydrophobic form. The two forms of each subunit had the same N-terminal sequence and may be due to differential glycosylation. The peaks in the 3-10-min interval originated from the reduction and S-pyridylethylation. mAu, milliabsorbance units at 216 nm; %B, percentage of 80% acetonitrile. The inset shows starting material and fractions 1-3 analyzed by SDS-PAGE/silver staining.

DISCUSSION
Several earlier studies reported proteins from the venom of C. rhodostoma that activate platelets, and it was suggested that one of these, called aggregoserpentin or aggretin, acts via ␣ 2 ␤ 1 (17). Aggretin was recently cloned, and the full amino acid sequence was established (18). We isolated a protein from C. rhodostoma venom that has N-terminal sequences of ␣and ␤-subunits identical to aggretin. This protein has a molecular mass of 28 kDa with subunits of 12 and 14 kDa on SDS-PAGE. Under nondenaturing conditions by gel filtration this protein has a molecular mass around 60 kDa. This implies that two subunits of the protein are linked together by disulfide bridges and that two heterodimers interact noncovalently to form tetramers. Aggretin aggregated platelets in a dose-dependent manner with a long lag phase as previously described (17).
In all previous reports (17,18) the authors showed that aggretin acts via the ␣ 2 ␤ 1 receptor. We confirmed that aggretin binds to ␣ 2 ␤ 1 but showed that it binds to GPIb as well. GPIb involvement in activation of platelets via aggretin was shown by affinity chromatography and inhibition of aggretin-induced platelet aggregation using an anti-GPIb monoclonal antibody. The monoclonal antibody VM16d completely inhibited aggretin-induced platelet aggregation as well as changes in proteintyrosine phosphorylation. Other monoclonal antibodies against GPIb (lb-23, SZ-2, and AP-1) did not inhibit aggretin-induced platelet activation. This indicates that the aggretin-binding site on GPIb␣ is limited to the same region as the epitope for VM16d, on the complex double-loop domain (27). Both GPIb and ␣ 2 ␤ 1 were detected in eluted material specifically bound to biotinylated aggretin-avidin-Sepharose. These receptors bound to aggretin only when the platelet lysate was prepared in the absence of EDTA or EGTA. When the platelet lysate was prepared with EDTA, proteins were not specifically bound to aggretin, showing that divalent cations are important for aggretin interactions with platelet receptors. It is not yet clear whether divalent cations are necessary for the active structure of aggretin or whether they are necessary for receptors to maintain a structure that is recognized by aggretin.

FIG. 2. A comparison of the N-terminal sequences of the subunits of aggretin with members of the snake venom C-type lectin family.
The sequences are highly similar, and aggretin clearly belongs to the same protein family. Assignment to ␣and ␤-subunits was on the basis of the subunit size. The positions with amino acids conserved in at least four of the six subunit sequences, respectively, are shown with bold letters.

FIG. 3. Antibodies against ␣ 2 ␤ 1 and GPIb inhibit aggretin-induced platelet aggregation and protein-tyrosine phosphorylation.
A, platelet aggregation was induced by 40 ng/ml of aggretin. Anti-␣ 2 ␤ 1 antibody (6F1, left panel) or anti-GPIb antibody (VM16d, right panel) at various concentrations were added to platelets 1 min before aggretin. The amounts of antibodies added are indicated. B, washed human platelets (700 l, 5 ϫ 10 8 platelets/ml) were stirred at 1000 rpm at 37°C. Aggretin at 40 ng/ml was added, and aliquots were withdrawn at the times indicated and dissolved in SDS buffer containing inhibitors. After separation by SDS-PAGE (7-17% acrylamide gradient) and transfer to PVDF membranes, the proteins were incubated with anti-phosphotyrosine antibody 4G10 before detection by peroxidase-linked second antibody and chemiluminescence. Anti-␣ 2 ␤ 1 antibody (6F1, 50 g/ml) or anti-GPIb antibody (VM16d, 20 g/ml) were added to the platelets 1 min before aggretin. The left panel shows changes in protein-tyrosine phosphorylation in platelets activated by aggretin without any antibodies; the middle panel shows changes in platelets pretreated with 6F1; and the right panel shows changes in platelets pretreated with VM16d.
Shin and Morita (19) isolated a C-type lectin that activates platelets from venom of C. rhodostoma that they called rhodocytin. They did not establish an activation mechanism but showed that it was not via GPIb. The N-terminal sequences that they determined are identical with the first 22 and 19 amino acids, respectively, of aggretin. Recently, the same group reported that the ␣ 2 subunit of ␣ 2 ␤ 1 is involved in the platelet interaction with rhodocytin and investigated some aspects of signal transduction occurring in platelets after activation with rhodocytin (20,21). The situation has been complicated by a recent report that rhodocytin does not bind to a recombinant ␣ 2 ␤ 1 complex, whereas another C. rhodostoma venom C-type lectin, rhodocetin, composed of noncovalently associated ␣and ␤-subunits with different sequences than those of aggretin or rhodocytin, bound strongly to the recombinant complex and blocked its binding to collagen (22).
In earlier studies, lack of involvement of GPIb as a receptor for aggretin or rhodocytin was based on the failure of monoclonal antibodies or echicetin to affect platelet activation. In fact, Shin and Morita (19) do show an increase in the lag phase response to rhodocytin in the presence of echicetin, and the slope of aggregation was also slightly inhibited. We also found that echicetin had a minor inhibitory effect on platelet responses to low concentrations of aggretin (data not shown). The monoclonal antibodies that had no effect on aggretin were AP-1 and 6D1 (17). Thus, the binding site for aggretin on GPIb must partly overlap with the echicetin-binding site but not with the epitopes of AP-1, 6D1, SZ-2, or Ib-23. On the other hand, VM16d shares the same binding site as aggretin. VM16d is a well characterized monoclonal antibody to GPIb with its epitope in the double-loop region (27) that inhibits thrombinbinding but not von Willebrand factor binding (28) and must therefore bind to one face of the GPIb␣ molecule. These results suggest that the previous studies did not detect aggretin (rhodocytin) binding to GPIb because the anti-GPIb reagents used did not bind to the same region (or face) of GPIb. FIG. 4. Platelet surface glycoproteins binding to biotinylated aggretin-avidin Sepharose 4B. Platelets were lysed with 1.2% Triton X-100 in buffer containing inhibitors (without EDTA), and glycoproteins were isolated by wheat germ agglutinin-affinity chromatography. Biotinylated aggretin was added to aliquots and incubated together for 2 h. Then avidin-Sepharose 4B was added, and the mixture was incubated for a further 4 h. As a negative control, avidin-Sepharose was incubated with an aliquot of platelet glycoproteins without adding biotinylated aggretin. After intensive washing both avidin-Sepharose samples were boiled in 1% SDS. Eluted proteins were separated by SDS-PAGE, transferred to a PVDF membrane, and treated with anti-␣ 2 (lanes 1-3) or anti-GPIb (lanes 4 -6) monoclonal antibodies before detection by peroxidase-linked second antibody and chemiluminescence. Lanes 1 and 4, platelet glycoproteins; lanes 2 and 5, eluate from avidin-Sepharose without biotinylated aggretin; lanes 3 and 6, eluate from biotinylated aggretin-avidin-Sepharose. Considerable evidence has accumulated that among the potential candidates for collagen receptors, at least ␣ 2 ␤ 1 and GPVI/Fc␥ are critical. It was suggested earlier that ␣ 2 ␤ 1 is essentially an adhesion receptor, whereas GPVI/Fc␥ is the main activation receptor for collagen, although this is controversial (11, 29 -31). Collagen-related peptides based on a repetitive GP*P sequence (where *P is hydroxyproline) signal via Fc␥ to activate p72 SYK and PLC␥2 (31)(32)(33).
Convulxin, a hexameric C-type lectin from C. durissus terrificus, clusters GPVI/Fc␥ to activate platelets via a tyrosine phosphorylation pathway (15). To establish whether or not aggretin requires the GPVI/Fc␥ receptor, the effects of aggretin on platelets from Fc␥-negative mice were compared with those on normal mouse platelets (Fig. 5). Platelet binding of fluorescein-labeled fibrinogen was used to monitor the activation of GPIIb-IIIa as a marker of platelet stimulation. Control experiments with convulxin, showed a clear difference between the two types of mouse platelets with a strong activation of GPIIb-IIIa on the control mouse platelets, whereas the Fc␥-negative platelets were not activated (Fig. 5A). Our results with convulxin on the Fc␥-deficient mouse platelets support those obtained by Gibbins et al. (34) as well as earlier studies with collagen-related peptides and GPVI-deficient human platelets (31) and confirm our earlier conclusions that convulxin acts via GPVI (15). In the case of aggretin, platelets from both types of mice responded.
Surprisingly, although the normal mouse platelets reacted strongly, the Fc␥-deficient platelets responded even more powerfully (Fig. 5B), suggesting either that more GPIb/␣ 2 ␤ 1 is expressed or that signaling via these receptors is up-regulated by the loss of the Fc␥ pathway. Both types of platelet also aggregated in response to aggretin. The GPIIb-IIIa responses of both the Fc␥-deficient platelets and the normal platelets to collagen were almost identical (Fig. 5C) (note that total GPIIb-IIIa activation after 2 min was measured here; therefore differences in rates were not studied). However, neither Fc␥deficient mouse nor GPVI-deficient human platelets (31) aggregate to collagen. Therefore, a signal from GPVI/Fc␥ is essential for the aggregation response to collagen. The clear results obtained with convulxin as well as Western blot studies with anti-Fc␥ antibodies (data not shown) confirmed the Fc␥deficient status of these platelets. Keely and Parise (35) reported that Fc␥RIIA played an essential role in activation of platelets via cross-linking with antibodies to ␣ 2 ␤ 1 . Because mouse platelets do not express Fc␥RIIA (36), it is unlikely that this receptor is involved in platelet activation by aggretin.
Earlier studies on Fc␥-deficient mouse platelets reported that they show only low responses to collagen (37) including a remnant low activation of p72 SYK and PLC␥2. Although the exposure of the fibrinogen-binding sites is not identical to aggregation, a close correlation might be expected between these two parameters. However, Poole et al. (37) did not study aggregation to collagen with the Fc␥-deficient mouse platelets. In the case of human platelets from a patient with GPVI deficiency, also lacking Fc␥ (38), the platelets also still responded to collagen but much more slowly than normal human platelets (31). It was suggested that the remnant response of GPVI-deficient human platelets to collagen could be due to its interactions with ␣ 2 ␤ 1 alone.
Suzuki-Inoue et al. (21) showed that rhodocytin induced aggregation of Fc␥-deficient mouse platelets and that ␣ 2 bound to a rhodocytin affinity column. Rhodocytin binding to liposomes containing recombinant ␣ 2 ␤ 1 was not inhibited by EDTA (21). Eble et al. (22), however, found no rhodocytin binding to recombinant ␣ 2 ␤ 1 whether in an enzyme-linked immunosorbent assay or in competition with ␣ 2 ␤ 1 binding to collagen. The question of whether aggretin and rhodocytin are identical or variants with the same N-terminal sequences is not yet clear.
In platelets stimulated by aggretin, p125 FAK was strongly activated. In contrast to platelets treated with convulxin, FIG. 6. Time dependence of tyrosine phosphorylation in proteins from platelets activated by aggretin, collagen, or convulxin. Washed platelets (700 l, 5 ϫ 10 8 platelets/ml) were stirred at 1000 rpm at 37°C. Agonist was added, and aliquots were withdrawn at the times indicated and dissolved in SDS buffer containing inhibitors. After separation by SDS-PAGE (7-17% acrylamide gradient) and transfer to PVDF membranes, the proteins were incubated with anti-phosphotyrosine antibody 4G10 before detection by peroxidase-linked second antibody and chemiluminescence.
FIG. 7. Dose dependence of lag phase of aggregation and tyrosine phosphorylation in proteins from platelets activated by aggretin. Upper panel, washed human platelets (500 l, 5 ϫ 10 8 platelets/ml) were stirred at 1,100 rpm at 37°C, and aggregation was induced by aggretin at 2 g/ml (line A), 0.4 g/ml (line B), or 0.08 g/ml (line C) (an arrow marks the point of addition). Lower panel, aliquots of aggretin-activated platelets were lysed with SDS at the times shown, and proteins were separated by SDS-PAGE, transferred to a PVDF membrane, and detected with anti-phosphotyrosine antibody (4G10), peroxidase-linked second antibody, and chemiluminescence.
where the activation of p125 FAK is a consequence of the secondary involvement of GPIIb-IIIa (15) in aggretin-treated platelets, the activation of p125 FAK was only slightly blocked by GPIIb-IIIa inhibitors and is thus a primary consequence of the interaction between aggretin and ␣ 2 ␤ 1 and GPIb.
It is still controversial whether ␣ 2 ␤ 1 is constitutively active or needs to be activated by inside-out signaling to bind collagen. This is difficult to test with collagen because of its numerous recognition sequences that can interact with different receptors. In the two-step, two-receptor model (39), which is the simplest version proposed for collagen-platelet interactions in primary hemostasis, it was suggested that the platelet binds through ␣ 2 ␤ 1 first and then is activated via interactions with GPVI/Fc␥. However, platelet activation via GPVI/Fc␥ may be necessary to modulate ␣ 2 ␤ 1 to allow it to bind to collagen. In adhesion of platelets to the subendothelium, the situation may be further complicated because under high shear stress conditions it may be the GPIb-V-IX receptor that causes platelet activation via von Willebrand factor bound to collagen to bring ␣ 2 ␤ 1 into action. Another piece of evidence supporting at least a basal level of constitutive activation comes from the results with the GPVI-deficient platelets from the Japanese patients where collagen is still able to induce a weak response; however, other collagen receptors, including GPIb/von Willebrand factor, may participate in this process.
The main platelet activation pathway by both collagen and convulxin was shown to involve phosphorylation of Fc␥, leading to activation of p72 SYK and hence of PLC␥2. A role for ␣ 2 ␤ 1 in activation of p72 SYK and of PLC␥2 by pathways requiring Fc␥RIIA was also suggested by studies using specific antibodies and inhibitors (35), and Polanowska-Grabowska et al. (40) showed that ␣ 2 ␤ 1 signals via dephosphorylation of a heat shock protein complex. However, these investigations are limited by the use of collagen as agonist where interactions with more than one receptor cannot be excluded or by the use of activating antibodies where a role of Fc␥RIIA was critical. It has been suggested that aggretin or rhodocytin may provide a good tool for investigating signaling via ␣ 2 ␤ 1 . However, because aggretin binds to both ␣ 2 ␤ 1 and GPIb the question about the role of other receptors in preactivation of ␣ 2 ␤ 1 before it can bind agonists such as collagen is still open. Because aggretin is again an agonist with more than one binding site for platelet receptors, this suggests that binding and clustering of ␣ 2 ␤ 1 may not be enough to activate platelets, and therefore GPIb or GPVI (in the case of collagen) is necessary to give adequate stimulation of signaling pathways. Platelet activation with aggretin-induced activation of p72 SYK and of PLC␥2 but not Fc␥ co-immunoprecipitated with p72 SYK . The absence of a direct, early, signaling role for Fc␥ in the platelet response to aggretin supports the results obtained using platelets from Fc␥-deficient mice where the GPIIb-IIIa activation response to aggretin was stronger than that obtained with platelets from normal, control mice. Aggretin, by activating and clustering ␣ 2 ␤ 1 and GPIb, gives a powerful signaling response in platelets. This strongly supports a major signaling role for ␣ 2 ␤ 1 integrin in collageninduced platelet activation as well. Results measuring "soluble collagen" binding to resting and activated platelets (41) suggest that, although there is a residual binding to resting platelets, this is considerably enhanced in activated platelets. Because the structure of soluble collagen is poorly defined, a role for GPVI/Fc␥ in resting platelets in binding this collagen can still not be excluded. In the platelet response to collagen, ␣ 2 ␤ 1 acts synergistically with GPVI/Fc␥ and modulates and controls the response to this receptor. This is clearly seen in the strength of the platelet response to convulxin compared with collagen. In this, ␣ 2 ␤ 1 shows a number of parallels to GPIIb-IIIa. Normally, GPIIb-IIIa needs to be activated by signaling from other receptors to change conformation and allow fibrinogen binding. However, platelets can adhere to and be activated by a fibrinogen-coated surface (42), so there must be a recognition potential even in the "nonactivated" conformation. Clustering of GPIIb-IIIa leads to activation of p72 SYK without obvious Fc␥ involvement; however, the mechanism for this is not yet known (43). Aggretin clustering of ␣ 2 ␤ 1 and GPIb appears to induce a similar signaling pathway that may be common to several integrins. The availability of aggretin as a specific reagent, activating platelets via ␣ 2 ␤ 1 and GPIb independently of GPVI/ Fc␥, should allow further analysis of the signaling pathways from these receptors. Together with reagents specific for GPVI/ Fc␥, such as convulxin (15) or collagen-related peptides (32), it should provide the tools to analyze how these major receptors work together in collagen-induced platelet activation. Control platelets or platelets activated by aggretin (AGN), collagen (COLL), or convulxin (CVX) were used. A, platelets were lysed in Triton X-100 buffer containing inhibitors. Aliquots were immunoprecipitated with anti-p72 SYK antibody. The immunoprecipitates were separated by SDS-PAGE. After transfer to PVDF membrane, the proteins were incubated with 4G10 anti-phosphotyrosine antibody and detected by peroxidase-linked second antibody and chemiluminescence. The membrane was stripped and reprobed for p72 SYK and for Fc␥. B, platelets were lysed in Triton X-100 and immunoprecipitated with anti-PLC␥2 antibodies. After SDS-PAGE and transfer to PVDF membranes, the proteins were detected with 4G10 anti-phosphotyrosine antibody. The membrane was stripped and treated with anti-PLC␥2 antibodies. C, platelets activated in the absence (Ϫ) or the presence (ϩ) of 1 M Ro-44 -9883, a GPIIb-IIIa inhibitor, were immunoprecipitated with anti-pp125 FAK antibodies. After SDS-PAGE separation and transfer to PVDF membrane, the immunoprecipitates were incubated with 4G10 anti-phosphotyrosine antibody. The membrane was stripped and reprobed for pp125 FAK .