Platelet glycoprotein Ib alpha binds to thrombin anion-binding exosite II inducing allosteric changes in the activity of thrombin.

The glycoprotein (GP) Ib-IX complex is a platelet surface receptor that binds thrombin as one of its ligands, although the biological significance of thrombin interaction remains unclear. In this study we have used several approaches to investigate the GPIb alpha-thrombin interaction in more detail and to study its effect on the thrombin-induced elaboration of fibrin. We found that both glycocalicin and the amino-terminal fragment of GPIb alpha reduced the release of fibrinopeptide A from fibrinogen by about 50% by a noncompetitive allosteric mechanism. Similarly, GPIb alpha caused in thrombin an allosteric reduction in the rate of turnover of the small peptide substrate d-Phe-Pro-Arg-pNA. The K(d) for the glycocalicin-thrombin interaction was 1 microm at physiological ionic strength but was highly salt-dependent, decreasing to 0.19 microm at 100 mm NaCl (Gamma(salt) = -4.2). The salt dependence was characteristic of other thrombin ligands that bind to exosite II of this enzyme, and we confirmed this as the GPIb alpha-binding site on thrombin by using thrombin mutants and by competition binding studies. R68E or R70E mutations in exosite I of thrombin had little effect on its interaction with GPIb alpha. Both the allosteric inhibition of fibrinogen turnover caused by GPIb alpha binding to these mutants, and the K(d) values for their interactions with GPIb alpha were similar to those of wild-type thrombin. In contrast, R89E and K248E mutations in exosite II of thrombin markedly increased the K(d) values for the interactions of these thrombin mutants with GPIb alpha by 10- and 25-fold, respectively. Finally, we demonstrated that low molecular weight heparin (which binds to thrombin exosite II) but not hirugen (residues 54-65 of hirudin, which binds to exosite I of thrombin) inhibited thrombin binding to GPIb alpha. These data demonstrate that GPIb alpha binds to thrombin exosite II and in so doing causes a conformational change in the active site of thrombin by an allosteric mechanism that alters the accessibility of both its natural substrate, fibrinogen, and the small peptidyl substrate d-Phe-Pro-Arg-pNA.

Thrombin is one of the most potent physiological agonists of platelets, inducing activation responses such as cytokinesis, aggregation, secretion, and associated metabolic changes (1,2). The first receptor for thrombin identified on the platelet surface was the glycoprotein (GP) 1 Ib-IX complex (3)(4)(5)(6), although the biological significance of this interaction remains unknown 20 years later. In the interim, other thrombin receptors have also been identified on the platelet surface including three members of the seven transmembrane domain receptor superfamily known as proteolytically activated receptor-1 (PAR-1), PAR-3, and PAR-4 (7-10). PAR-1 was shown to mediate platelet activation (7) and took much attention away from GPIb-IX as a thrombin receptor until it was found that PAR-1 knockout mice had no demonstrable bleeding disorder and that their platelets retained normal responses to ␣-thrombin (11). Furthermore, synthetic peptides of PAR-3 and PAR-4 that mimic the sequences of the potentially activating tethered ligands, either failed to show (PAR-3) or showed very low (PAR-4) activation effects on platelets (9,10). Therefore, attention has again turned to the GPIb-IX complex to understand, for example, why the platelets from patients with Bernard-Soulier syndrome, which lack or have dysfunctional GPIb-IX complexes on their surfaces fail to respond to low doses of thrombin (12,13).
Glycoprotein Ib-IX consists of three polypeptides (GPIb␣, GPIb␤, and GPIX) that are each transmembrane proteins that span the platelet membrane once. They are also each members of the leucine-rich repeat superfamily (14). Glycoprotein Ib␣ and GPIb␤ are linked together via a disulfide bond that is situated in the extracellular space very close to the platelet membrane, and GPIX is noncovalently associated with these two. On the platelet surface, a fourth leucine-rich repeat protein, GPV, appears to link two GPIb-IX trimers together and may be responsible for even more complex aggregation states (14 -16). Of this multimeric complex it is the extracellular portion of GPIb␣, known as glycocalicin, that forms the site of ligand interaction. Glycocalicin can be released from the surface of platelets by proteolysis near the platelet membrane and consists of two subregions known as the macroglycopeptide and the amino-terminal domain (14). It is the amino-terminal domain, which consists of about 300 amino acids, that provides the sites within glycocalicin for ligand interaction (3), and within this the thrombin-binding site has been localized between residues 269 and 287 (17). The site on thrombin, however, where GPIb␣ binds is more controversial. Within the structure of thrombin four prominent regions have been identified: its active site, a sodium ion-binding site (18,19), and two surface electropositive patches known as anion binding exosites I and II (20) that have both been implicated in thrombin binding. Although some studies have indicated that GPIb␣ binds to anion-binding exosite I (also known as the fibrinogen recognition site) of thrombin (21)(22)(23)(24), the studies of De Cristofaro and colleagues (25)(26)(27) have implicated anion-binding exosite II (also known as the heparin binding site) of thrombin as the site of GPIb␣ interaction.
Thrombin is an allosteric serine protease (28) with changes in the conformation of its active site being induced by the binding of ligands at the other sites of the enzyme. For example, the binding of ligands at exosite I changes the rate of turnover of small synthetic peptidyl and natural substrates (29,30), whereas the binding of heparin at exosite II alters the kinetics of the inhibition of thrombin by hirudin (31) consistent with allosteric linkage between exosites I and II of thrombin (32). We wondered whether the binding of GPIb␣ to thrombin also produced an allosteric response in the enzyme. Furthermore, we sought to clarify which of the exosites of thrombin was the GPIb␣-binding site in the expectation that such knowledge would lead to the formulation of new hypotheses about the biological role of GPIb␣-thrombin interaction, because the exosite involved should be directly related to the consequence of this binding.
In this study, we have used four approaches to investigate the binding interaction of GPIb␣ with thrombin: 1) an HPLC method measuring the effect of ligands on thrombin-induced fibrinogen turnover that has been well characterized in the study of thrombin-fibrinogen and thrombin-thrombomodulin interactions (33,34), 2) mutant thrombins with single amino acid substitutions in either exosite I or II (35)(36)(37), 3) a resinbased assay for studying the direct binding of thrombin to GPIb␣, and 4) competition binding assays using small ligands of known exosite interaction as competitors of the GPIb␣thrombin interaction. Our results clearly indicated that the binding site for GPIb␣ is located in thrombin anion-binding exosite II and that the binding of GPIb␣ to thrombin induces conformational changes at the active site of thrombin by an allosteric mechanism that alters the activity of thrombin toward both physiological and small substrates.

Preparation of Glycocalicin and the Amino-terminal Fragment of GPIb␣ from Human Platelets
Glycocalicin was isolated from outdated human platelets using a modification of methods previously reported (38,39). Briefly, after glycocalicin was cleaved from the surface of platelets by calpain released as a result of their sonication, it was isolated by wheat germ agglutinin Sepharose 4B and Q-Sepharose anion exchange chromatography. Purified glycocalicin was dialyzed into 20 mM Tris-HCl, pH 8.0, and concentrated to 1 mg/ml, and aliquots were snap frozen in liquid nitrogen and stored at Ϫ80°C until use. The amino-terminal fragment of GPIb␣ was generated from purified glycocalicin by cleavage with porcine pancreatic elastase at an enzyme to substrate ratio of 1:250 (w/w). Porcine pancreatic elastase cleaves glycocalicin after residue Val 289 yielding the amino-terminal fragment and a macroglycopeptide, which are well separated by Q-Sepharose. Both fragments were concentrated, dialyzed, and stored as described for glycocalicin above.

Recombinant Expression of the Amino-terminal Fragment of GPIb␣
The amino-terminal region of GPIb␣ was expressed from baculovirus-infected insect cells (40) and purified as described previously (41). Briefly, the cDNA coding for the signal peptide and amino-terminal domain of GPIb␣ (residues Ϫ16 to 289) followed by the calmodulin (CaM) gene was inserted into the baculovirus expression vector pAcSG2 (Pharmingen, San Diego, CA). The resultant plasmid, pWIb␣wt, was then recombined with BacVector-3000 triple cut virus DNA (Novagen, Madison, WI) to produce infective virus. A high titer baculovirus stock was obtained from a single plaque by repeated infection of Sf9 insect cells. Recombinant virus was then used to infect High Five insect cells for expression of the GPIb␣-CaM fusion protein. The calmodulin moiety of the fusion protein was then exploited in the first stage of recombinant protein isolation since it binds to N-(6-aminohexyl)-5-chloro-1-naphthalenesulfonamide (W-7)-agarose (prepared as described previously (42)). Calmodulin could be released from the GPIb␣ (1-289) fragment by digestion with porcine pancretaic elastase (1:100 w/w E:S ratio), and the GPIb␣ (1-289) could be retrieved in pure form after passage over Q-Sepharose anion exchange resin.

Recombinant Prothrombin Expression
Recombinant mutant prothrombins R68E, R70E (exosite I mutants), R89E, and K248E (exosite II mutants) were expressed and characterized as described previously (36,37). Amino acid residues are numbered from the first residue of the human thrombin B chain. Purified prothrombins were activated with Echis carinatus snake venom, and the thrombin products were further purified to homogeneity by ion exchange chromatography on Amberlite CG-50. The purity of each thrombin was confirmed by SDS-polyacrylamide gel electrophoresis and silver staining. The activities of recombinant mutant and wild-type thrombins were tested as described (43,44).

Characterization of the Thrombin-GPIb␣ Interaction
Effect of Glycocalicin on the Thrombin-Fibrinogen Interaction-The binding of glycocalicin to thrombin was determined by modifying an HPLC-based assay that has been well characterized for the study of fibrinogen binding to thrombin (33). First, thrombin was incubated with fibrinogen under the desired solution conditions (5 mM Tris-HCl, 0.1% PEG-8000, pH 8.0 at 25°C) and the progress curves for fibrinopeptide A (FpA) release were measured to determine the values of the specificity constant k cat /K m . The fibrinogen concentration was 0.2 M, thrombin concentrations varied according to their specific activities in the range from 0.08 -2 nM, and the ionic strength was kept constant with NaCl. The reaction was initiated by addition of thrombin to the fibrinogen solution (or to the fibrinogen and glycocalicin solution) and quenched at different time intervals with 3 M perchloric acid. The sample was then centrifuged, and the amount of FpA in the supernatant was determined by HPLC, as described previously (33). Next, glycocalicin at varying concentrations was added to the assay system, and the equilibrium dissociation constants (K d ) for glycocalicin binding to thrombin were determined by analysis of the inhibition of FpA release as a function of glycocalicin concentration according to the following equation.
and x is the GPIb␣ concentration, e T is the active thrombin concentration, (k cat /K m ) 0 and (k cat /K m ) 1 are the specificity constants for FpA release in the absence of and at saturating concentration of GPIb␣, respectively, and t c ϭ K m /e T k cat corresponds to the point in the progress curve that gives the greatest change in the amount of FpA released as a function of the inhibitor concentration. The exosite mutant thrombins were substituted for wild-type thrombin in this assay to investigate the effect that GPIb␣ had on their cleavage reactions of fibrinogen. Before each assay the concentration of thrombin was adjusted to ensure the same thrombin activity was present in each reaction. The same assay was used for investigating the salt dependence of the thrombin-glycocalicin interaction in the range of 100 -150 mM NaCl. The data plotted as log K d versus log [salt] were fitted to a straight line according to the following expression (34).
The slope of this line yields ⌫ salt , which represents a thermodynamic measure of the effect of salt concentration on binding equilibria (28,34). Direct Thrombin Binding Assay-The direct binding of thrombin to GPIb␣ was investigated using the recombinant GPIb␣-CaM fusion protein expressed in insect cells. 50 l of W-7-agarose solution containing 25 l of packed resin was first added to a 500-l Eppendorf tube. The resin beads were then washed with washing buffer (100 mM NaCl, 50 mM Tris-HCl, pH 7.4) three times. 1 nmol of recombinant GPIb␣-CaM fusion protein in 100 l of binding solution (100 mM NaCl, 50 mM Tris-HCl, pH 7.4, 20 mM CaCl 2 ) was added to the 25 l of washed W-7-agarose beads and incubated for 2 h at 25°C. The beads were then washed once with binding solution and blocked with 3% bovine serum albumin, 0.1% PEG-8000 in binding solution for another 2 h. After blocking, duplicates of increasing quantities (2.5, 5.0, 7.5, 10, or 12.5 pmol) of thrombin (wild type or mutants) were added to separate tubes in binding solution in a total volume of 70 l and incubated for 2 h at 25°C. The total volume of each reaction, including W-7-agarose, was therefore 95 l containing final concentrations of thrombin of 26, 52, 79, 105, and 131 nM, respectively. As a negative control, W-7 beads were incubated for the indicated time in the absence of GPIb␣-CaM. The beads were kept suspended during each incubation step by inverting the tubes every 10 min. At the end of the incubation, the beads were pelleted by centrifugation at 600 ϫ g for 1 min. 50 l of each supernatant was transferred to a 96-well microtiter plate and incubated with 50 l of 2.5 mM thrombin substrate D-Phe-Pro-Arg-pNA. The release of p-nitroaniline was monitored by spectrophotometry at 405 nm using a THERMOmax microplate reader (Molecular Devices Sunnyvale, CA) at 5-min intervals from 5 to 60 min.
Competition of GPIb␣-Thrombin Binding Using Known Exosite Ligands-These assays were performed essentially as for the direct thrombin binding assay described above but with the addition of either hirugen (residues 54 -65 of hirudin with Tyr 63 sulfated; Sigma) or low molecular weight (LMW) heparin (Rhône-Poulenc Rorer Pharmaceuticals Inc., Collegeville, PA) added as inhibitors of exosites I and II, respectively. 50-l aliquots of W-7-agarose solution containing 25 l of packed beads were added to tubes and washed as above. 1 nmol of recombinant GPIb␣-CaM fusion protein in binding solution was added to the beads and incubated for 2 h at 25°C in constant suspension. After blocking for 2 h as above, 12.5 pmol of thrombin and various quantities (0, 50, 100, 250, 500, 750, 1000, 1500, and 2000 pmol) of hirugen or LMW heparin were added to duplicate tubes in a volume of 70 l making final concentrations of 0, 0.53, 1.05, 2.63, 5.26, 7.89, 10.53, 15.79, and 21.05 M, respectively, each in a total volume of 95 l. After 2 h incubation at 25°C, 50 l of each supernatant was transferred to a 96-well microtiter plate and incubated with 50 l of 2.5 mM thrombin substrate D-Phe-Pro-Arg-pNA. The release of p-nitroaniline was monitored spectrophotometrically at 405 nm using a THERMOmax microplate reader, as above.

Allosteric Effect of Thrombin Exosite Ligands on the
Amidolytic Activity of Thrombin 12.5 pmol of human ␣-thrombin was preincubated for 10 min in a total volume of 70 l as above with various concentrations (between 0 and 14.29 M) of glycocalicin, hirugen (Sigma), or LMW heparin (Rhône-Poulenc Rorer Pharmaceuticals Inc., Collegeville, PA). After incubation, 50 l of each solution was transferred to a 96-well microtiter plate and incubated with 50 l of 2.5 mM thrombin substrate D-Phe-Pro-Arg-pNA. The rate of hydrolysis of the chromogenic substrate was determined spectrophotometrically at 405 nm by a THER-MOmax microplate reader as above.

RESULTS
Noncompetitive Inhibition of Fibrinopeptide A Release-We began our studies by a detailed examination of the influence that glycocalicin had on the fibrinogen-thrombin interaction. Fibrinogen interacts with thrombin exosite I and is thereafter cleaved in its A␣ chain at Arg 16 and in its B␤ chain at Arg 14 , releasing FpA and FpB, respectively. We utilized a well characterized HPLC-based assay that quantitatively measures the release of FpA after thrombin-catalyzed hydrolysis of fibrinogen (33). Various concentrations of glycocalicin were added to solutions containing both thrombin and fibrinogen to determine whether the glycocalicin could inhibit the thrombin-mediated cleavage of fibrinogen. As shown in Fig. 1, the addition of glycocalicin decreased the amount of FpA released from the fibrinogen. The FpA release could not be totally inhibited, even by a large glycocalicin excess, however, indicating that glycocalicin was not acting in a competitive manner and was not binding to thrombin exosite I. The curve in Fig. 1 was best fit by Equation 1, which describes a noncompetitive mode of inhibition, and because thrombin is an allosteric enzyme (28), we concluded that the inhibitory effect of glycocalicin occurred by an allosteric mechanism. The equilibrium dissociation constant (K d ) for the thrombin-glycocalicin interaction derived from this equation was 1.04 Ϯ 0.008 M at 150 mM NaCl. Identical results were obtained when the amino-terminal domain of GPIb␣ (residues 1-289), whether derived from human platelets or recombinantly expressed, was substituted for glycocalicin (data not shown).
The Effect of Salt on the Glycocalicin-Thrombin Interaction- The binding affinities of two other ligands that interact with thrombin exosite II, heparin (45) and the chondroitin sulfate moiety of thrombomodulin (34) have been shown to be very sensitive to salt concentration, whereas the affinities of ligands binding to thrombin exosite I are not (33,34,46). Fig. 2 shows that the influence of glycocalicin on the thrombin-induced FpA release from fibrinogen was salt-dependent, with the derived equilibrium dissociation constants for the glycocalicin-thrombin interaction varying nearly an order of magnitude between the NaCl concentrations of 100 and 150 mm (K 100 mM NaCl ϭ 0.187 Ϯ 0.009 M; K d 125 mM NaCl ϭ 0.41 Ϯ 0.03 M; K d 150 mM NaCl ϭ 1.04 Ϯ 0.008 M). The thrombin-fibrinogen interaction is minimally affected within this range of salt concentration (33). At higher salt concentrations than reported here, the affinity of glycocalicin for thrombin dropped considerably, whereas at lower concentrations the K d could not be determined accurately because of changes in the kinetic mechanism for the release of fibrinopeptides.
The effect of salt on the interaction of glycocalicin with thrombin can be further quantified by plotting the data as log K d verses log [salt] on a straight line according to Equation 2 (33,34). When the log of the equilibrium dissociation constants derived for the glycocalicin-thrombin interaction at different concentrations of NaCl was plotted against the log of their respective [Na ϩ ], the straight line in Fig. 3 was obtained. The value of ⌫ salt for this interaction was calculated from the slope of the line as Ϫ4.2. For comparison, the equivalent plots reported by other investigators for the thrombin-heparin (45) and thrombin-hirugen (34) interactions are also shown in Fig. 3 as representatives of thrombin exosite II and exosite I interactions, respectively. The value of ⌫ salt for the glycocalicin-thrombin interaction is compared in Table I with the ⌫ salt values obtained for thrombin and other exosite I and II ligands obtained from the literature. As can be seen, thrombin ligands that bind at its exosite I and show little salt dependence (fibrinogen and hirudin) have values of ⌫ salt around 1.0, whereas known exosite II-binding ligands (heparin and the chondroitin sulfate moiety of thrombomodulin) are characterized by ⌫ salt values around 4 -5. In this respect, glycocalicin is behaving as a thrombin exosite II-binding ligand.
Interaction of Thrombin Exosite Mutants with Glycocalicin-To address more directly the exosite on thrombin that mediates GPIb␣ interaction, four thrombin exosite mutants were employed in binding studies. Previous investigators showed that mutations in the anion-binding exosite II of thrombin (R89E, R245E, K248E, and K252E) greatly reduced its affinity for heparin (35,36) and that mutations in exosite I (R68E and R70E) reduced its affinity for fibrinogen (37). We first determined the equilibrium dissociation constants for the interaction between each of these four mutants and glycocalicin in the fibrinogen assay utilized above. The K d values derived from the curves in Fig. 4 for the interaction of glycocalicin with the exosite I mutants were comparable with that of wild-type thrombin (K d R68E ϭ 0.1 M; K d R70E ϭ 0.034 M at 100 mM NaCl, although because thrombin exosite I mutants interfere with the thrombin-fibrinogen interaction, the concentration of exosite I mutants used in the assay was 4 -40-fold more than that of wild-type thrombin). The exosite II mutants, however, interacted very poorly with glycocalicin having 10-and 25-fold higher K d values than wild-type thrombin (K d R89E ϭ 1.4 M; K d K248E ϭ 4.9 M at 100 mM NaCl) and showed almost no inhibitory effect on FpA release.
Direct Binding of Thrombin Exosite Mutants to Recombinant GPIb␣-This assay utilized the properties of CaM to attach the recombinant GPIb␣-CaM fusion protein to W-7-agarose beads. The thrombin-binding properties of this GPIb␣-CaM fusion protein were indistinguishable from either glycocalicin or the amino-terminal domain of GPIb␣ (data not shown). The thrombin that bound to the immobilized GPIb␣ pelleted with the W-7-agarose beads and was quantified from the difference between the total amount of thrombin activity added to the tube and that remaining in the supernatant. In this way it was found that the thrombin exosite I mutants interacted with the GPIb␣ (1-289) attached to the resin in an identical way as did wild-type thrombin (Fig. 5). In contrast, the thrombin exosite II mutants showed minimal binding activity toward the resinassociated GPIb␣ (1-289) and had similar binding curves to the negative control in which no recombinant GPIb␣ (1-289) was attached to the resin (Fig. 5). Taken together, these studies also support the concept that the GPIb␣ binding-site on thrombin is within anion-binding exosite II.
Thrombin Exosite I and II Ligands as Inhibitors of Glyco-

FIG. 3. Comparison of the NaCl concentration dependence of the GPIb␣-thrombin interaction with those of the heparinthrombin and hirugen-thrombin interactions.
The dissociation constants for the glycocalicin-thrombin interaction (q) derived from the curves in Fig. 2 were plotted as a function of the respective Na ϩ ion concentration on a log-log scale and fitted by linear regression according to Equation 2. For comparison, the equivalent data previously reported for the thrombin exosite II-heparin (45) (f) and thrombin exosite I-hirugen (34) (OE) interactions are shown. The slope of the line for the glycocalicin-thrombin interaction, ⌫ salt , is listed in Table I along    calicin Binding-We next investigated the abilities of hirugen (a thrombin exosite I ligand) and LMW heparin (a thrombin exosite II ligand) to inhibit thrombin from binding to the GPIb␣ attached to the W-7-agarose beads. Again we derived the difference between the total thrombin activity added to controls with no GPIb␣-CaM fusion protein attached to the W-7 beads, and the residual thrombin activity left in the supernatant after pelleting the thrombin that had bound to the GPIb␣ (1-289) attached to the resin particles. During these experiments we observed that the activity of the unbound thrombin toward the peptidyl substrate D-Phe-Pro-Arg-pNA was itself influenced by the binding of hirugen and LMW heparin. This is shown in Fig.  6 together with the allosteric effect of glycocalicin on the amidolytic activity of thrombin toward this small synthetic tripeptidyl substrate. Whereas the binding of hirugen resulted in allosteric enhancement of the amidolytic activity of thrombin, the binding of LMW heparin and glycocalicin resulted in allosteric inhibition of the amidolytic activity of thrombin.
It can be seen from Fig. 7A that at a concentration of 21 M, LMW heparin completely inhibited thrombin binding to the GPIb␣ (1-289) on the resin, because all the available thrombin activity remained unbound in the supernatant. In contrast, hirugen had no effect on the thrombin-GPIb␣ (1-289) interaction, as determined by the absence of convergence of the two curves in Fig. 7B. In the presence of all concentrations of hirugen, most of the added thrombin binds to the GPIb␣ (1-289) attached to the resin and is removed from solution resulting in a marked and constant reduction in the unbound thrombin activity in the supernatants in each tube. These results indicate that LMW heparin, which binds to thrombin exosite II, can fully inhibit GPIb␣ binding to thrombin, whereas the exosite I ligand, hirugen, is not an inhibitor of this interaction. DISCUSSION Evidence has been accumulating for some time that thrombin undergoes changes in the conformation of its active site as an allosteric response to ligands binding at other sites of the enzyme. For example, the binding of proteins or peptides corresponding to segments of natural inhibitors and substrates that bind to exosite I (30) and the binding of the chondroitin sulfate moiety of thrombomodulin to exosite II (34) have been shown to alter the amidolytic activity of thrombin toward small synthetic substrates. Thrombomodulin binding to exosite I of thrombin enhances the cleavage of protein C at least 500-fold. Most of this effect has been attributed to the influence of thrombomodulin on the conformation of protein C in the thrombin-thrombomodulin-protein C ternary complex. However, an approximately 15-fold rate enhancement appears to be due to the influence of thrombomodulin on the active site architecture of thrombin as measured by turnover of small substrates that mimic the sequence cut by thrombin in protein C (28). Another notable example of allostery involves the sodium-binding site the occupancy of which triggers the transition of thrombin between the slow and the fast forms (28) that are primarily associated with the anticoagulant and procoagulant functions of thrombin, respectively (47).
Our present studies provide an additional example of the allosteric nature of thrombin. This became evident when we investigated the detailed effect of GPIb␣ on the thrombinfibrinogen interaction. The results showed that glycocalicin inhibited the cleavage of fibrinogen by thrombin, confirming the findings of previous investigators (23,24,48). The inhibition was not, however, of a competitive nature because it was not possible to completely inhibit FpA release by increasing the concentration of glycocalicin. Instead, the equation that best fit the inhibition curve of Fig. 1 described a noncompetitive mode of inhibition that would be consistent with the operation of an allosteric mechanism. Therefore, unlike the previous investigators (23,24,48), we concluded that glycocalicin did not bind to thrombin exosite I. Although our conclusions differed, the experimental findings reported here are similar to those of Jandrot-Perrus et al. (23,24) and De Marco et al. (48). The key to understanding the different interpretations of the findings resides in consideration of the magnitude of the inhibition caused by GPIb␣ on the thrombin-fibrinogen interaction. In no study did GPIb␣ completely inhibit thrombin-fibrinogen interaction. Our present data now indicate that this is because GPIb␣ does not compete for the same binding site on thrombin as fibrinogen but rather induces a conformational change at the active site through an allosterical mechanism that slows the rate of fibrinogen cleavage. As shown in Fig. 6, the alteration in the architecture of the active site of thrombin induced and K248E (f), were added. Thrombin that bound to the immobilized GPIb␣ (1-289) was determined from the difference between the total amidolytic activity of thrombin added to each tube and that remaining in the supernatant after pelleting the beads, as determined by its hydrolysis of substrate D-Phe-Pro-Arg-pNA. The negative control (E) had no recombinant GPIb␣-CaM fusion protein attached to the agarose beads to provide a measure of the total amount of thrombin that was available to bind to the GPIb␣. allosterically by the binding of GPIb␣ to its exosite II also decreases the amidolytic activity of thrombin toward the small synthetic tripeptidyl substrate D-Phe-Pro-Arg-pNA.
The binding of glycocalicin to thrombin was highly salt-dependent. This finding was reminiscent of ligands that bind to thrombin exosite II such as heparin and the chondroitin sulfate moiety attached to thrombomodulin and implicated GPIb␣ as a thrombin exosite II ligand. This would make GPIb␣ the first protein to directly bind to thrombin exosite II through proteinprotein interactions. In terms of negative charge density, however, the thrombin-binding site on GPIb␣ (residues 269 -287 (17)) could be said to resemble that of heparin. Of the 19 residues in this region of GPIb␣, 13 are negatively charged, including three sulfated tyrosines (residues 276, 278, and 279 (49,50)), which themselves give this region of GPIb␣ even more resemblance to the polysulfated glycosaminoglycan, heparin. The value of ⌫ salt is a quantitative measure of the salt dependence of a binding interaction and for the GPIb␣-thrombin interaction was determined to be Ϫ4. 2 (51), in close agreement with that recently reported by others (52) (Table I). This value of ⌫ salt far exceeds the values reported for fibrinogen and hirudin binding (Table I) and signals a much larger electrostatic contribution to the binding of GPIb␣ to thrombin. Consistent with this is the similarity of ⌫ salt for GPIb␣-thrombin with that of the heparin-thrombin interaction that has been characterized as a predominantly nonspecific electrostatic interaction (45). If it is assumed that the salt-dependent interactions between thrombin and its exosite II ligands are solely due to electrostatic association, then the values of ⌫ salt will also indicate the minimum number of ionic bonds involved in the binding (45). Thus, it might be expected that a minimum of four ionic bonds contribute to the association of GPIb␣ with thrombin.
More direct evidence that GPIb␣ binds to exosite II of thrombin was derived from our studies using mutant thrombins with single amino acid substitutions in either exosite I or II. The thrombin mutants with either R89E or K248E substitutions in exosite II both displayed dramatically decreased interactions with GPIb␣ (Figs. 4 and 5). In contrast, the exosite I mutations of R68E and R70E had little effect on GPIb␣ binding (Figs. 4 and 5). The mutant thrombins were employed here in two different assays. The first was in the same HPLC assay used above to investigate the effect that glycocalicin had on fibrinogen turnover by thrombin. Because, however, the thrombins with exosite I mutations themselves had reduced interactions with fibrinogen, potentially complicating the interpretation of the results of this assay, we also utilized an assay we had developed to study the direct binding of thrombin to a recombinant portion of GPIb␣ containing the thrombin-binding site (51). The results from both assays indicated that exosite II mutations, but not exosite I mutations, drastically reduced the binding of thrombin to GPIb␣.
The final way in which we investigated the GPIb␣ interaction site on thrombin was to employ hirugen and LMW heparin as inhibitors of binding to exosites I and II, respectively. As seen in Fig. 6 and by the dashed curves in Fig. 7, both of these small molecules caused allosteric conformational changes in the active site of thrombin as determined by changes in the amidolytic activity of thrombin toward the peptidyl substrate D-Phe-Pro-Arg-pNA. The solid lines in panels A and B of Fig. 7 represent how much thrombin was inhibited from binding to GPIb␣ immobilized on W-7-agarose beads by LMW heparin and hirugen, respectively. The convergence of the two curves in Fig. 7A indicate that at a concentration of 21 M, LMW heparin had inhibited all of the thrombin from binding to the immobilized GPIb␣ because all of the available thrombin amidolytic activity remained in the supernatant. Conversely, hirugen caused no detectable increase in the activity of unbound thrombin in the supernatant indicating that the binding of thrombin to the immobilized GPIb␣ was not being inhibited by hirugen. This again shows that GPIb␣ is interacting with thrombin exosite II.
In the present work we have determined the affinity for thrombin binding to glycocalicin as an inhibitor of the thrombin-fibrinogen interaction in an assay system that has been well characterized in the study of other inhibitors of this reaction (33,34). Under these conditions, glycocalicin behaved as a classical noncompetitive inhibitor of the thrombin-fibrinogen show the allosteric effects on the amidolytic activity of thrombin toward the peptidyl substrate D-Phe-Pro-Arg-pNA caused by adding just LMW heparin or hirugen, respectively, to thrombin in a solution of W-7-agarose with no GPIb␣ attached to the beads. The solid lines in each panel indicate the amount of thrombin that had bound to the GPIb␣ immobilized on the W-7-agarose in the presence of increasing concentrations of either LMW heparin (A) or hirugen (B) as competitors. Bound thrombin that had been removed from the solution by pelleting the W-7-agarose beads was calculated from the difference between the total thrombin activity added to the reaction tubes minus that remaining in the supernatant after pelleting the beads (unbound thrombin activity). The convergence of the solid curve with the dotted curve in A indicates that LMW heparin inhibited the binding of thrombin to the immobilized GPIb␣ so that all of the thrombin added to the reaction remained unbound by the addition of 20 M LMW heparin. Conversely, the addition of hirugen did not inhibit thrombin from binding to the immobilized GPIb␣.
interaction with a K d value of 1 M for binding to thrombin in the presence of 150 mM NaCl. Thus, thrombin binds more avidly to glycocalicin than to the extracellular amino-terminal fragment of the classical thrombin receptor, PAR-1, for which the K m is 15-30 M, also at 150 mM NaCl (53). If thrombin binding to the platelet surface involved the formation of a GPIb␣-thrombin-PAR-1 ternary complex, then thermodynamic considerations dictate that the 1 M thrombin-glycocalicin interaction added to that of the 15-30 M thrombin-PAR-1 interaction could reduce the K d of thrombin binding to as low as 15-30 ϫ 10 Ϫ12 M, which would be compatible with the K d values of 10 Ϫ8 -10 Ϫ10 M reported for thrombin binding to platelets (48,54). The finding that glycocalicin binds to thrombin exosite II, whereas PAR-1 binds at thrombin exosite I, would allow such a ternary complex. Alternatively, GPIb␣ might act as a "ligand-passing" receptor, initially trapping thrombin at the platelet surface to make it available to the PAR receptors, in a manner analogous to that proposed for the passing of tumor necrosis factor from tumor necrosis factor receptor-2 to tumor necrosis factor receptor-1 (55). In this role it would be advantageous for the association and dissociation of thrombin to and from GPIb␣ to be rapid, and a micromolar dissociation constant for their interaction would be consistent with this.
A further hypothesis, again involving a ternary complex mechanism, would describe a functional role for GPIb␣ binding to thrombin to retain and localize the enzyme at sites where fibrin generation is needed for the maturation and stabilization of blood clots (56). The platelet surface provides a major site for thrombin generation (57) through the clotting sequence. Like other factors involved in this pathway, the precursor of thrombin, prothrombin, is anchored to the platelet phosopholipid membrane by ␥-carboxylated glutamate residues (58 -61), which do not form part of the active ␣-thrombin enzyme when it is released from prothrombin by proteolysis. Perhaps at this time thrombin binds via its anion binding exosite II to GPIb␣ to be retained in the locality where fibrin generation is required. Fibrinogen would subsequently bind to thrombin exosite I and be cleaved to the products utilized for cross-linking into the insoluble fibrin matrix found in mature thrombi. Binding to GPIb␣ through exosite II would also prevent the inhibition of thrombin by antithrombin, as recently shown (52), because the inhibitory mechanism requires heparin binding (62) to thrombin exosite II. Obviously, it would be desirable to only temporarily prevent thrombin inhibition by antithrombin until the fibrin clot became large enough to stop further blood loss but not so large as to totally occlude blood flow through the vessel. The relatively weak affinity between thrombin and GPIb␣ described by a micromolar dissociation rate would be consistent with this, facilitating the ready release of thrombin from GPIb␣ for its subsequent inhibition by antithrombin. Furthermore, the 50% allosteric reduction in the rate of fibrinogen turnover, caused by the binding of GPIb␣ to thrombin at physiological ionic strength (Fig. 1), may be an additional mechanism to regulate blood coagulation in vivo, similar to the allosteric "switch" mechanism described recently for exosite inhibitors of factor VIIa (63). The localization of thrombin-GPIb complexes on the platelet surface might also recruit and activate additional platelets during the thrombus formation.
Although the precise role for thrombin binding to GPIb␣ is not known, several observations suggest that this interaction is physiologically important. For example, selective inhibition of thrombin binding to GPIb␣ shifts the dose-response curve of platelets induced by low doses of thrombin (21), and the thrombin-GPIb␣ interaction is necessary for thrombin-induced platelet procoagulant activity (64). Therefore, the thrombin-GPIb interaction may contribute to the initiation and maintenance of platelet responses during hemostasis.