Human Factor XII Binding to the Glycoprotein Ib-IX-V Complex Inhibits Thrombin-induced Platelet Aggregation*

Factor XII deficiency has been postulated to be a risk factor for thrombosis suggesting that factor XII is an antithrombotic protein. The biochemical mechanism leading to this clinical observation is unknown. We have previously reported high molecular weight kininogen (HK) inhibition of thrombin-induced platelet aggregation by binding to the platelet glycoprotein (GP) Ib-IX-V complex. Although factor XII will bind to the intact platelet through GP Ibα (glycocalicin) without activation, we now report that factor XIIa (0.37 μm), but not factor XII zymogen, is required for the inhibition of thrombin-induced platelet aggregation. Factor XIIa had no significant effect on SFLLRN-induced platelet aggregation. Moreover, an antibody to the thrombin site on protease-activated receptor-1 failed to block factor XII binding to platelets. Inhibition of thrombin-induced platelet aggregation was demonstrated with factor XIIa but not with factor XII zymogen or factor XIIf, indicating that the conformational exposure of the heavy chain following proteolytic activation is required for inhibition. However, inactivation of the catalytic activity of factor XIIa did not affect the inhibition of thrombin-induced platelet aggregation. Factor XII showed displacement of biotin-labeled HK (30 nm) binding to gel-filtered platelets and, at concentrations of 50 nm, was able to block 50% of the HK binding, suggesting involvement of the GP Ib complex. Antibodies to GP Ib and GP IX, which inhibited HK binding to platelets, did not block factor XII binding. However, using a biosensor, which monitors protein-protein interactions, both HK and factor XII bind to GP Ibα. Factor XII may serve to regulate thrombin binding to the GP Ib receptor by co-localizing with HK, to control the extent of platelet aggregation in vivo.

Recently, the functions of the contact system have been reevaluated in view of our increasing knowledge of the interactions of factor XII, prekallikrein, and kininogens with platelets, neutrophils, monocytes, and endothelial cells. Prolongation of the activated partial thromboplastin time found with individuals, having deficiencies of the contact proteins, originally suggested a coagulation defect. However, no hemorrhagic diathesis has ever been demonstrated. Recently, anticoagulant (1) and profibrinolytic (2) activities of these proteins have now been documented in vitro and in some cases in vivo (3).
The locus of the anticoagulant action of the contact system appears to be the platelet. Puri et al. (4) showed that HK 1 inhibited thrombin-induced platelet aggregation by inhibiting the binding of thrombin to platelets. Domain 3 of HK is responsible (5). Recently, the binding site on platelets, which mediates this effect, was shown to be GP Ib-IX complex (1). This observation helps to explain the fact that patients with Bernard-Soulier disease require 10-fold the concentration of thrombin needed to stimulate normal platelets (6). Kininogens also shift the dose-response curve for thrombin activation of platelets about 10-fold (1,4). These studies were confirmed by Joseph et al. (7), who demonstrated that a molecule of 74 kDa, isolated by ligand affinity chromatography utilizing HK, was a fragment of the GP Ib␣ chain. Furthermore, they demonstrated that biotinylated factor XII bound to the same Western-blotted GP Ib␣ fragment in a Zn 2ϩ -dependent interaction.
Platelet-fibrin thrombi, which result from the action of thrombin on fibrinogen and platelets, appear to be important in such complications as reocclusion after angioplasty or thrombolysis (8). At least two receptors are involved in the events triggered by the binding of thrombin receptors to platelets. The first is the high affinity, low capacity (40 sites/cell) receptor present on the 45-kDa amino terminus of glycoprotein Ib (9) expressed in the GP Ib-IX-V complex. The second is PAR1, which is coupled directly to G proteins and activated by the new amino terminus of the tethered peptide (10) resulting from thrombin cleavage, and is a moderate affinity receptor. Specific targeting of either receptor should modulate the effect of thrombin on platelets without inhibiting thrombin-induced proteolysis or the effects of other agonists for platelets including ADP, collagen, and thromboxane A2. Greco et al. (11) have presented evidence that both GP Ib and PAR1 are required to ensure optimal rate and extent of platelet responses, such as an increase in intracellular Ca 2ϩ , at a range of thrombin concentrations of 0.3-10 nM. Recently a third human protein, PAR4 (12), has been shown to be involved in thrombin activation of platelets at high thrombin concentrations, e.g. 30 nM (13).
Factor XII is a plasma protein present in a concentration of 370 nM. Factor XII is a glycoprotein of ␤-globulin mobility composed of 596 amino acids, 16.8% carbohydrate, with a molecular mass of 78 -80 kDa. Activation of zymogen factor XII is initiated by a single cleavage resulting in a two-chain molecule held together by a single disulfide bond designated factor XIIa.
The disulfide-linked enzyme is composed of a light chain, 28 kDa, containing the catalytic site of the enzyme, and a heavy chain, 52 kDa, containing the surface binding region of the molecule. In vitro, further proteolysis is known to transform human factor XIIa to a smaller enzyme, 30 kDa, lacking the heavy chain, designated factor XIIf (also known as ␤-factor XIIa). Factor XIIf retains its ability to rapidly cleave and activate some coagulation zymogens such as prekallikrein but not others such as factor XI. The entire primary amino acid sequence of factor XII has been reported and has revealed that the light chain and, more particularly, the heavy chain have structural features that, by homology, imply further physiological functions of the molecule (14 -16). The light chain of factor XIIa is homologous to plasmin, an enzyme that participates in fibrinolysis, and to tissue plasminogen activator, a protease that converts plasminogen to plasmin in the fibrinolytic pathway. The sequence homology is highest at the activation regions and near the active sites of these proteases. The heavy chain has extensive sequence homology with both fibronectin and tissue plasminogen activator. The heavy chain shares sequence homology with type II regions of fibronectin and is composed of approximately 60 residues including four halfcysteine residues. The type II homologies probably comprise the collagen-binding site in fibronectin and, by analogy, may be responsible for reported collagen-binding properties of factor XII. Two regions of factor XII are homologous to a sequence resembling epidermal growth factor that could represent regions of factor XII that interact with cells, such as neutrophils or monocytes. The heavy chain also contains a proline-rich region making up 33% of the sequence, the significance of which is unclear at this time, but it most likely functions as a connection domain between the light chain and heavy chains of factor XII.
HK and factor XII compete for the same binding site on endothelial cells (17). We tested the hypothesis that factor XII, like HK, would inhibit thrombin binding to the GP Ib-IX-V complex. We studied whether factor XII or its derivatives factor XIIa and XIIf were responsible for the modulation of thrombininduced platelet aggregation. We also determined the role of the other thrombin receptors, PAR1 and PAR4, for the inhibitory effect of factor XIIa on the activation of platelets by thrombin.

MATERIALS AND METHODS
Avidin-FITC, lot number 037H4839 (2.5 mg/ml), at a concentration of 36 M, was purchased from Sigma. All other reagents were of the best purity available.
Purified Proteins and Peptides-Human ␣-thrombin was a generous gift of Dr. John W. Fenton II (Division of Laboratories and Research, New York State Department of Health, Albany, NY) and had an activity of 2400 NIH units/mg and a purity of 99.24%. Human HK (1.6 mg/ml), human factor XII (1.01 mg/ml), and human factor XIIa (0.6 -0.75 mg/ ml) were purchased from Enzyme Research Laboratories, South Bend, IN. On SDS-PAGE, HK, a component of Ͼ95% migrating at a mass of 120 kDa, and factor XII, a single component of 80 kDa, both remained unchanged on reduction with mercaptoethanol. Factor XIIa was a single component of 80 kDa and resolved two components of 50 and 30 kDa on reduction. Factor XII was also purified from plasma by affinity chromatography using the monoclonal antibody to the heavy chain, B7C9, prepared in our laboratory as described previously (18), following prekallikrein removal with a monoclonal antibody 13G11 affinity column. Factor XII fragment was prepared by activation with plasma kallikrein and separated by anion exchange chromatography (19). Glycocalicin (1.1 mg/ml) was a generous gift of Dr. José A. López (Veterans Affairs Medical Center, Houston). The peptide SFLLRN was synthesized, purified, and characterized as published (18).
Inactivation of the Catalytic Activity of Factor XIIa-The buffer containing commercial human factor XIIa (0.6 mg/ml) was increased to 0.5 M NaCl, and the pH was adjusted to 7.5 with 0.5 N NaOH. The sample was then divided equally, and one portion brought to 1 mM AEBSF with the addition of 100 mM AEBSF solubilized in a neutral buffer. AEBSF is known to inhibit serine proteases with specificity similar to trypsin by irreversibly modifying the active site (20). The remaining portion had the identical volume of buffer added excluding the AEBSF inhibitor, and both samples were then dialyzed at 4°C for 4 h in HEPES/ Tyrode's buffer ϩ 0.5 M NaCl, pH 7.5, to remove inhibitor and exchange the buffer solution. Both samples were assayed for protein content, factor XII clotting activity, and amidolytic activity (19). AEBSF inactivation of factor XIIa reduced both coagulant and amidolytic activity by 92% compared with the control sample.
Platelet Isolation and Aggregation-Platelet isolation by gel filtration and platelet aggregation was performed as described previously (21). Briefly, an aliquot of platelets (Ͼ100,000/l) and added inhibitors in a 0.5-ml cuvette at 37°C was incubated with stirring for 2 min in the presence of 50 M free Zn 2ϩ . Aggregation was monitored for 5 min after the addition of 2 nM thrombin. The rate of aggregation was measured as the slope of the line in the presence of added inhibitors compared with the same sample measured by activation with thrombin alone.
Desensitization of the Platelet PAR1 Receptor-Desensitization of the human platelet PAR1 receptor was performed by a modification of the procedure (12) using platelets obtained from 100 ml of freshly drawn blood, collected into acid/citrate/dextrose. Briefly, platelets were washed by centrifugation, and after the addition of 1 mM EDTA and 1 M prostaglandin E 1 the resulting platelet pellet was incubated with 100 M SFLLRN peptide at room temperature for 5 min without stirring. Following incubation with the SFLLRN peptide, the resuspended platelet preparation was gel-filtered to remove any remaining inhibitors or agonists (21). The resulting platelet pool yielded concentrations between 2 and 6 ϫ 10 8 cells/ml and were further tested for response to SFLLRN-and thrombin-induced aggregation at 37°C.
Western Blotting Detection of Protein Complexes-Western blotting was performed according to the protocol of Towbin et al. (22). Briefly, 4 g each of two-chain HK, factor XII, and thrombin were electrophoresed on SDS-PAGE or in the absence of SDS (native gel) and then transferred to Immobilon-P (Millipore Corp., Bedford, MA). After blocking of the unbound protein surface sites, the transfer was probed with 1 g/ml of factor XII (all of the following conditions are performed in the presence of 50 M Zn 2ϩ ), washed, and detected with the polyclonal goat anti-factor XII antibody, FMR-43 (Federated Medical Research, raised in goats against whole factor XII antigen). After 2 h of incubation at room temperature, the transfer was washed and developed with antigoat alkaline phosphatase-conjugated IgG.
Antibodies-The monoclonal antibodies AK-1, which binds to GP IX close to its site of interaction with GP Ib, AK-2, which recognizes the 45-kDa ligand-binding peptide tail of GP Ib␣, and AK-3, which binds to the macroglycopeptide of GP Ib␣ (23), were provided by Dr. William J. Booth, University of Sydney, Westmead, Australia. AK-2 blocks thrombin binding to GP Ib␣ but does not inhibit thrombin-induced platelet aggregation, whereas AK-1 and AK-3 have no effect on thrombin binding. The monoclonal antibody to GP IX, FMC-25 (which does not block thrombin binding), was supplied by Dr. Manling Peng and developed by Berndt et al. (24). SZ2, which only blocks thrombin binding at low concentrations but can inhibit thrombin-induced platelet aggregation, was purchased from Biodesign International, Kennebunk, ME, and recognized GP Ib␣ (extracellular domain) (25). The biotinylated antibody TM-60 (260 g/ml), which binds to the amino-terminal 45-kDa peptide of GP Ib␣ and also inhibits thrombin-induced platelet aggregation (26), was a generous gift of Dr. Naomasa Yamamoto, The Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan. A rabbit polyclonal antibody, anti-STDR (11), raised to the synthetic peptide LLRN-PNDKYEPF was a gift from Dr. Nicolas J. Greco (Holland Laboratory, American Red Cross, Rockville, MD). This antibody blocks the binding of thrombin to PAR1.
Platelet Binding Assays-In the cell fluorescence assays (cytofluorescence) using biotinylated antibody or ligand, binding was detected, after washing, with FITC-labeled avidin (100 l of a 1-200 dilution of the stock reagent, resulting in a final FITC-avidin concentration of 180 nM) and incubated for 30 min at 23°C. Following washing and removal of the supernatant by vacuum, plates were analyzed for fluorescent antibody bound to the platelets at an excitation wavelength of 485 nm and detected at an emission wavelength of 530 nm in a Millipore fluorescence microtiter plate reader. Results were determined from the fluorescence detected in the presence of 50 M free Zn 2ϩ after subtraction of the fluorescence measured for the identical points in the absence of added Zn 2ϩ .
Comparison of the Effect of GP Ib-IX and Factor XII-specific Monoclonal Antibodies on Biotin-labeled Factor XII and HK Binding to Gel-filtered Platelets-Five monoclonal antibodies were tested using cytofluorescence for the ability to block zinc-dependent, biotin-labeled factor XII or HK binding to gel-filtered platelets. A monoclonal antibody to factor XII (B7C9) was used as a positive control. The monoclonal antibodies were all tested at 1 M. AK-2, TM-60 (both directed to the 45-kDa amino-terminal region of GP Ib␣), AK-3 (directed to the macroglycopeptide region of GP Ib␣), AK-1, and FMC-25 (both binding to GP-IX) were incubated with factor XII (370 nM).
Surface Plasmon Resonance Assays-Biomolecular interaction assays were performed using a Biacore 2000 (Biacore AB Uppsala, Sweden). To obtain a carboxyl-terminal linked ligand, glycocalicin was oxidized for 20 min in 10 mM sodium periodate, passed over a NAP-5 column (Amersham Pharmacia Biotech), and equilibrated in 10 mM sodium acetate, pH 4.0. The resulting protein (0.2 mg/ml) was coupled to flow cell 2 (flow cell 1 was chemically modified without coupled ligand as a background channel) on a CM-5 biosensor chip using the surface thiol method according to the BIAapplications handbook. Statistics-In the analyzed experiments, the experimental groups or points were compared with the control by paired Student's t test. Values were reported as significant when p Ͻ 0.05 where not otherwise indicated.

RESULTS
Factor XIIa Inhibition of Thrombin-induced Platelet Aggregation-The rate of platelet aggregation was determined in the presence of increasing concentrations of thrombin (0.2-2.0 nM) alone or with the addition of factor XIIa (370 nM). Factor XIIa significantly inhibited the observed maximal rate of aggregation by 60% at 1 nM thrombin and by 40% at 2 nM thrombin compared with the rates in the absence of factor XIIa (Fig. 1). Unlike the inhibition by HK of thrombin-induced platelet aggregation (1), the effect of factor XIIa is still evident at a thrombin concentration of 2 nM. In a separate experiment, thrombin-induced platelet aggregation (2 nM) was measured alone or with the additions of factor XII, XIIa, and XIIf at concentrations from 185 to 740 nM (50 -200% of the plasma concentration) (Fig. 2). The inhibition of aggregation was observed with increasing concentrations by factor XIIa only. Neither factor XII fragment (light chain, with the same enzymatic activity as factor XIIa) nor factor XII zymogen inhibited thrombin-induced platelet aggregation at any concentration tested suggesting that a conformational change in the heavy chain after proteolytic activation is required. To define further if the inhibition of thrombin-induced platelet aggregation by XIIa requires both the exposure of the factor XII heavy chain and a functional active site, we tested the inhibitory response of AEBSF-treated factor XIIa compared with a control XIIa sample prepared in parallel without the inactivation by AEBSF.

FIG. 2. Thrombin-induced platelet aggregation is inhibited by the heavy chain of factor XIIa.
Thrombin-induced platelet aggregation alone (2 nM) is compared with platelet aggregation in the presence of increasing amounts of factor XII zymogen (black bars), activated factor XII (light gray bars), and factor XII fragment (dark gray bars). The inhibition by each factor XII species of thrombin-induced platelet aggregation was expressed as a percentage of aggregation by thrombin alone. Factor XIIa containing the heavy chain, but not factor XII zymogen or XIIf, was a potent, dose-dependent inhibitor of thrombin-induced platelet aggregation. Factor XII fragment or factor XII zymogen could not block aggregation, indicating that this inhibition is due to the conformational exposure of the heavy chain of factor XIIa and that the catalytic domain is not sufficient.

FIG. 1. Factor XIIa inhibits thrombin-induced platelet aggregation.
Platelet aggregation, induced by thrombin concentrations from 0.2 to 2.0 nM, was measured at 37°C in the presence and absence of 370 nM factor XIIa. The rate of platelet aggregation was determined in three separate experiments (mean Ϯ S.E.) and plotted as a function of thrombin concentration. Gel-filtered platelets require more thrombin to stimulate aggregation (gray circles) in the presence of factor XIIa than in the absence of factor XIIa (black circles).
The inhibition profile of the inactivated factor XIIa that had only 8% of the catalytic activity of the parent factor XIIa was identical to the fully catalytically active factor XIIa (Fig. 3). These results indicate that the inhibitory potency of human factor XIIa resides in the heavy chain and that catalytic activity is not required.
No Complex Detected between Thrombin and Factor XII or Thrombin by Western Blot-Following SDS-PAGE or native (non-SDS) gel protein separation and transfer to Immobilon-P, HK, and thrombin, in the presence of 50 M Zn 2ϩ , were probed with factor XII, followed by a polyclonal antibody to factor XII (FMR-43) and an alkaline phosphatase-conjugated secondary antibody. No reaction with either HK or thrombin was detected by this ligand blot technique. In contrast, factor XII is readily detected with FMR-43 under both reduced and nonreduced conditions (data not shown). This experiment demonstrates that no detectable complex formation was observable between factor XII and either HK or thrombin.
Factor XIIa Effect on PAR1-Factor XIIa (370 nM) was incubated with gel-filtered platelets for 2 min, and then aggregation was induced with either thrombin (1 nM) or the TRAP, SFLLRN (10 M). In contrast to factor XIIa inhibition of throm-bin-induced aggregation, the rate of TRAP-induced aggregation was not significantly altered with the addition of factor XIIa (Fig. 4), indicating that factor XIIa did not inhibit the interaction of the new amino terminus of the tethered peptide resulting from the cleavage of PAR1 with its exosite on the same receptor. Platelets (1.7 ϫ 10 8 /ml) pretreated with murine IgG to block nonspecific Fc receptors were also incubated with 1 M anti-LLRNPNDKYEPF (rabbit polyclonal), an antibody that prevents initial binding of thrombin, or with 1 M rabbit IgG (nonspecific control), and then the binding of 250 nM biotinlabeled factor XII in the presence of 50 M Zn 2ϩ was detected using cytofluorescence. The antibody failed to block significantly the binding of biotin-labeled factor XII to gel-filtered platelets when compared with IgG or ligand alone (Fig. 5), indicating that factor XII does not bind to uncleaved PAR1 at the same site as thrombin. These experiments, taken together, demonstrate that factor XII does not inhibit thrombin action on platelets by binding to PAR1, and therefore its cleavage product, factor XIIa, acts on GP Ib complex.
Effect of XIIa on PAR4 -Unlike HK (1) the inhibitory effect of factor XIIa is not overcome completely by raising the concentration to 2 nM (Fig. 1). Nonetheless factor XIIa does not

FIG. 3. The active site of factor XIIa is not required to inhibit thrombininduced platelet aggregation. Factor
XIIa was treated with 1 mM AEBSF to modify covalently the active site serine and thus inactivate the catalytic site. Thrombin-induced platelet aggregation was determined in the presence of XIIa and AEBSF-inactivated XIIa. Inactivated factor XIIa (gray bars) showed no differences in concentration dependence (185-740 nM) when compared with untreated factor XIIa (black bars). The data are derived from three separate experiments expressed as mean Ϯ S.E. The results are reported as the percent of the rate of thrombin-induced aggregation with inhibitor compared with thrombin alone (2 nM). This experiment demonstrates that factor XIIa inhibition of platelet aggregation is due to the conformational exposure of the heavy chain.
inhibit the effect of TRAP-induced aggregation. Furthermore, an antibody to PAR1 (anti-STDR) did not block the binding of factor XII. These observations led us to consider the role of PAR4 in the inhibition of thrombin-induced platelet activation by factor XIIa. We studied the possible involvement of PAR4 by desensitizing the platelet response to PAR1 by incubation with SFLLRN (13) as described under "Materials and Methods." After desensitization, platelets did not aggregate in response to 100 M SFLLRN (data not shown). The concentration dependence of thrombin was markedly shifted. Instead of an observed aggregation response from 0.1 nM to reaching a maximum at 2 nM (see Ref. 1 and Fig. 1), no response to thrombin was found at 1 nM, and platelet aggregation determined at 2 nM was less than 10% the rate at 30 nM (Fig. 6A). These results show that for thrombin-induced aggregation by PAR4, the IC 50 was 4 nM, and at 5-10 nM, the rate of aggregation was 80 -90% that observed at 30 nM. When factor XIIa was tested as an inhibitor of thrombin-induced aggregation of PAR1-desensitized platelets, factor XIIa produced a modest (25%), but significant (p Ͻ 0.01), inhibition indicating that PAR4 contributes to the inhibitory response to factor XIIa at high thrombin concentrations (Fig. 6B). Under similar conditions HK (2 M) did not inhibit the effect of thrombin on platelets desensitized to PAR1 (results not shown).
Factor XII Displacement of Platelet-bound HK-Gel-filtered platelets were preincubated for 30 min with biotin-labeled HK (30 nM) and then increasing amounts of factor XII (5-100 nM) were added. Finally, the remaining platelet-bound biotin-labeled HK was detected by cytofluorescence. Factor XII was able to displace 50% of the bound HK at equimolar concentrations (30 nM), and excess factor XII (30 -100 nM) did not elicit any further displacement (Fig. 7). When excess free HK is present in this system, the factor XII displacement cannot be observed (data not shown). This experiment suggests that HK and factor XII may compete for a common binding site on the platelet surface. Since HK has been demonstrated to bind to GP Ib/IX/V on the platelet surface (1), it is likely that factor XII also binds to that receptor.
Failure of Anti-GP Ib and GP IX Antibodies to Block Factor XII Binding-The monoclonal antibodies used to map HK binding in Ref. 1 were used to identify the GP Ib complex locus for factor XII binding. The epitopes probed on GP Ib␣ (TM-60, AK-2), GP IX (AK-1, FMC-25), and the macroglycopeptide portion (AK-3) of the complex did not block factor XII binding to FIG. 6. PAR1-desensitized platelets respond to high concentrations of thrombin; factor XIIa modestly inhibits thrombin activation of the PAR4 receptor. Platelets desensitized to the PAR1 receptor by incubation with 100 M SFLLRN peptide were evaluated for the ability to aggregate with increasing concentrations of thrombin. Thrombin activation mediated through the PAR4 platelet receptor showed an IC 50 of 4 nM (A), and full activation was achieved at concentrations of thrombin above 10 nM. In a separate experiment, platelets desensitized to PAR1 were evaluated for the inhibition of thrombininduced platelet aggregation by factor XIIa. The rate of aggregation by thrombin alone (B, black bar) was compared with the rate observed in the presence of 370 nM factor XIIa (B, gray bar). Factor XIIa inhibited PAR4 thrombin activation by 25 Ϯ 5.5% S. E. (p Ͻ 0.01, n ϭ 5). These results show that human factor XIIa can modestly inhibit thrombininduced aggregation through the PAR4 receptor. gel-filtered platelets in the cytofluor system. In contrast, all but AK-3 blocked HK binding, as previously reported (1). The monoclonal anti-factor XII heavy chain antibody B7C9 was the only antibody able to block factor XII binding to platelets in these experiments, consistent with the previous finding that the factor XII heavy chain is required (Table I).
Factor XII Binds to Glycocalicin in a Zinc-dependent Manner-The carboxyl terminus of glycocalicin was coupled to the CM-5 sensor chip by a surface thiol. Surface plasmon resonance detected 12,000 RU of ligand bound, equal to 12 nmol of glycocalicin (1 pmol ϭ 1 RU). HK binding (200 nM) to the extracellular component of the GP Ib complex, glycocalicin was detected by surface plasmon resonance detection in the presence of 50 M Zn 2ϩ (Fig. 8A). Likewise, in the same experiment, factor XII (200 nM) bound at a level equal to HK (450 RU) but showed a 30% faster dissociation rate (Fig. 8B). No significant binding of HK or factor XII was observed in the absence of Zn 2ϩ . Both HK and factor XII demonstrate Zn 2ϩ -dependent and specific binding to immobilized glycocalicin (Fig. 8, C  and D).
In a separate experiment, factor XIIf was compared with factor XII for zinc-dependent binding to immobilized glycocalicin in the presence of 1 mM of the serine protease inhibitor, AEBSF, to ensure that the normally catalytically active XIIf did not cleave glycocalicin bound to the Biacore CM-5 chip. Factor XIIf did not bind to glycocalicin with or without the addition of Zn 2ϩ (Fig. 9, B and C), supporting the results observed in the platelet aggregation studies showing that XIIf did not block thrombin-induced platelet aggregation. The presence of 1 mM AEBSF did not negatively affect Zn 2ϩ -dependent binding of factor XII zymogen to glycocalicin. In fact, maximal binding was almost 50% higher (800 RU compared with 475 RU, Figs. 8 and 9) and, in the presence of AEBSF, serves to preserve factor XII from autolysis during the course of the experiment. These results confirmed that factor XIIf, in contrast to factor XII, does not bind to immobilized glycocalicin. DISCUSSION Our previous investigations have indicated that HK and certain peptides derived from domain 3 of HK can alter the concentration dependence of thrombin such that a 10-fold greater concentration is required to activate platelets in plasma (4) and in purified systems and to stimulate cells transfected with GP Ib-IX-V (1). Furthermore, by using monoclonal antibodies, we showed that HK can block thrombin binding by competing for sites on the GP Ib-IX-V complex. The present study demonstrates that factor XIIa also inhibits thrombin interaction with platelets in a mechanism also involving binding to the same receptor. Several lines of evidence lead to this conclusion. First, factor XIIa inhibits thrombin-induced platelet aggregation with up to 80% inhibition observed at low concentrations of thrombin (between 0.6 and 1.6 nM). This effect on aggregation at low concentrations, targeting the high affinity site, points to GP Ib-IX-V as the binding site. Second, factor XIIa fails to inhibit the action of the tethered peptide, SFLLRN, which stimulates the PAR1 receptor, and an antibody specific to the initial binding site for thrombin on PAR1 fails to inhibit the binding of factor XIIa. These observations make it very unlikely that factor XIIa alters thrombin activation by interacting directly with PAR1. Moreover, factor XII can inhibit the binding of biotinylated HK, which has been shown to bind to GP Ib-IX-V (1). Finally, HK and factor XII both directly bind to immobilized glycocalicin (Fig. 8), the extracellular subunit of GP Ib␣, in a Zn 2ϩ -dependent manner (Fig. 9) as detected by surface plasmon resonance.
Although the above experiments strongly imply that the mechanism of factor XII inhibition of aggregation involves binding to GP Ib-IX-V, several differences between the interactions of HK and factor XII should be noted. Factor XII does not inhibit HK binding more than 50%, even at concentrations of factor XII in excess of HK. This observation could be explained by the requirement for two noncontiguous sites present to bind to HK on the GP Ib-IX-V complex and only one site for factor XII. Support for this hypothesis is suggested by the finding (1) that antibodies to both GP Ib␣ (AK2 and TM60) and GP IX (FMC-25 and AK1) inhibit HK binding, whereas none of these antibodies inhibits factor XII binding (Table I). Alterna-  7. Factor XII directly blocks 50% of biotin-labeled HK binding to gel-filtered platelets. Gel-filtered platelets, in the presence of 50 M free Zn 2ϩ , are incubated with increasing concentrations of factor XII in a Millipore Multiscreen filtration plate followed by addition of 30 nM biotin-labeled HK. Following 30 additional min, the unbound reagents were removed, and the platelet-bound biotin-HK was detected with FITC-avidin on a Cytofluor 2300 system. The plateletbound HK is plotted as a function of factor XII concentration. The results are expressed as the mean Ϯ S.E. (n ϭ 4).
tively, a single site for HK binding may be more extensive involving the contributions of amino acids from the interface between GP Ib␣ and IX, and factor XII may only interact with amino acids from GP Ib␣ (glycocalicin). The lack of inhibition by the anti-macroglycopeptide is consistent with the recent localization of the thrombin-binding site Asp 272 -Glu 283 on GP Ib␣ (27) remote from the macroglycopeptide that lies close to the transmembrane portion of the receptor.
The number of high affinity thrombin-binding sites is about 40 -50 (9) compared with 25,000 copies of GP Ib-IX-V receptor complexes per platelet. In Bernard-Soulier disease 10 times as much thrombin is required to stimulate platelets in the absence of GP Ib-IX-V. HK blocks these sites and also results in a 10-fold shift of thrombin concentration requirements. There is no evidence that thrombin, HK, or factor XII binds to more than 50 sites on GP Ib. The structural basis for the difference in the receptor site is not known. Other investigators (28) have suggested differences in glycosylation or conformational changes. For HK there are 1,000 binding sites per platelet. However, the binding sites may not all be on GP Ib-IX because platelets express on their surface a receptor for the globular head of the complement component C1q (gC1q-R) that also binds kininogen. There have not been formal binding studies using 125 I-factor XII as a ligand for platelets so the number of sites is unknown. Moreover, factor XII also binds to gC1q-R. Further mapping is required to determine the exact sequences on GP Ib-IX-V complexes responsible for the binding of HK and factor XII and their relationships to the sequence responsible for thrombin binding. The present data suggest some overlap but do not indicate an exact identity.
Direct evidence that factor XII binds to GP Ib␣ has been obtained using surface plasmon resonance. First, both factor XII and HK but not factor XIIf bind to glycocalicin, confirming the results observed with factor XIIa inhibition of thrombininduced platelet aggregation (Fig. 2). Second, the binding of HK and factor XII is zinc-dependent (Fig. 8), in agreement with a previous report (7). In addition, a serine protease inhibitor, AEBSF, does not affect the binding, indicating that the catalytic region of factor XIIa is not required (Fig. 9); it actually facilitates binding, probably by preventing autoactivation and subsequent autolysis to factor XIIf, which would be inactive in preventing thrombin binding.
Thrombin alone is known to bind to several receptors on human platelets, including GP Ib IX-V complex (9), PAR1 (10), FIG. 8. Factor XII directly binds to platelet glycocalicin, the extracellular component of GP Ib␣. The direct binding of factor XII and HK was determined by surface plasmon resonance. The carboxyl terminus of glycocalicin was immobilized to a CM-5 sensor chip by surface thiol coupling. Analyte (200 nM) was observed for the ability to bind to glycocalicin in the presence or absence of 50 M Zn 2ϩ . As previously reported for HK (line A), factor XII (line B) demonstrates zincdependent binding to glycocalicin. Lines C and D show HK and XII binding, respectively, without zinc.
FIG. 9. Factor XII fragment does not bind to glycocalicin, indicating that the locus for factor XII binding resides in the heavy chain. The catalytic region of biotin-labeled factor XII (XIIf) was previously shown not to inhibit thrombin-induced aggregation of gel-filtered platelets. Factors XIIf and XII (200 nM), inactivated with 1 mM AEBSF to prevent proteolysis of the surface-bound ligand, were tested for the ability to bind to glycocalicin in the presence or absence of 50 M Zn 2ϩ . Zn 2ϩ -dependent binding of factor XII was unaffected by AEBSF (line A). Factor XIIf did not bind to glycocalicin with Zn 2ϩ (line B) or in 50 M EDTA (line C). The remaining lines show factor XII ϩ 50 M EDTA (line D), buffer ϩ Zn 2ϩ (line E), and buffer ϩ EDTA (line F). and PAR4 (13). (PAR-2 is a receptor for tryptase, not thrombin, and PAR-3 is found in reasonable copy number only in the mouse.) Thrombin binds to PAR1 and cleaves the peptide bond between Arg 41 and Ser 42 (10). The peptide LLRNPNDKYEPF represents amino acids 44 -55 of PAR1, and an antibody to this peptide, anti-STDR, inhibits thrombin binding to PAR1 (11) Anti-STDR failed to block factor XII binding (Fig. 5). Moreover, the new amino-terminal amino acid sequence, SFLLRN, which stimulates PAR1 to activate platelets, was not inhibited by either factor XIIa (Fig. 4) or HK (1); thus, the participation of PAR1 is unlikely. PAR4 is a recently cloned, G protein-coupled receptor (29) present on human platelets (30); its activation by thrombin requires very high concentrations (30 nM). When PAR1 is desensitized, the concentration dependence of thrombin shifts (Fig. 6A) so that low concentrations of thrombin (less than 2 nM) do not stimulate aggregation, but thrombin can stimulate maximally at Ͼ5 nM. The need for both PAR1 and PAR4 is consistent with recent observations (13, 31) that both receptors participate in thrombin signaling. Unlike HK, at 30 nM thrombin, factor XIIa inhibits platelet aggregation modestly about 25% (Fig. 6B) indicating that PAR4 is a target of the inhibition by XIIa, although its role is only at high concentrations of thrombin. These high concentrations can be present in the microcirculation during disseminated intravascular coagulation, the microenvironment of a thrombus, or on a cellular surface.
The structural requirements for inhibition of thrombin aggregation on platelets by factor XIIa display certain resemblance to factor XIIa interactions with leukocytes. Factor XIIa has been shown to stimulate neutrophil aggregation and degranulation (release of elastase) (32). Similar to the findings in this study, factor XIIf or zymogen factor XII does not stimulate neutrophils, implying that the heavy chain is required. However, in the case of neutrophils, the catalytic integrity of factor XIIa is required for activation, since unlike the effect of factor XIIa on thrombin-induced platelet activation, active site blocking agents inhibit stimulation. The effect of factor XIIa on platelets is more similar to the down-regulation of the Fc␥RI (immunoglobulin) receptors on monocytes. This effect requires the heavy chain but, in contrast to effects of factor XIIa on neutrophils and similar to its action on platelets, does not require the catalytic activity of the light chain (33). Preliminary evidence (34) suggests that the ability of factor XIIa to down-regulate Fc␥RI may lie within the amino-terminal 18 amino acids. This finding is consistent with the inhibition of factor XII binding to platelets by monoclonal antibody B7C9, whose noncontiguous epitopes have been mapped to amino acids 1-20 (35) and an icosapeptide in the "finger" region (19). These results indicate that the "surface-binding regions" of factor XII in the heavy chain may also be cell-binding sequences.
Our data show that, although factor XIIa is required for the inhibition of thrombin activation of platelets, factor XII will bind to the intact platelet through GP Ib␣ (glycocalicin) without activation. This result is similar to previous data from our laboratory (36) that show that changes in circular dichroism occur both on binding of factor XII to an anionic surface and following the autoactivation to factor XIIa. Thus, we suggest that following the binding of zymogen factor XII, autoactivation and/or proteolytic activation by plasma kallikrein produces factor XIIa bound to the platelet surface. The conformational changes allow factor XIIa to down-regulate thrombin activation of platelets. Further studies are necessary to test this hypothesis.
Peptides from these domains may be useful in modulating thrombin activation of platelets and may prevent thrombotic complications with fewer propensities for hemorrhagic complications. Peptides, peptidomimetic agents, or monoclonal anti-bodies targeting these cell-binding regions may be useful in modulating thrombin activation of platelets. Such an approach may diminish the hemorrhagic complications of direct thrombin inhibitors (hirudin), drugs that inhibit the synthesis of specific platelet agonists (aspirin) or global platelet function inhibitors (GP IIb/IIIa antagonists).
In conclusion, factor XIIa binds via the amino-terminal heavy chain region to platelet GP Ib␣ at a site that can inhibit the binding of thrombin. This effect requires the conformational alteration of the heavy chain that exists after proteolytic activation of factor XII and does not require the active site. PAR4 is also involved with factor XIIa inhibition of thrombin activation of platelets but only at high concentrations, whereas GP Ib is the target at concentrations of thrombin below 2 nM.