The Role of Glu192 in the Allosteric Control of the S2′ and S3′ Subsites of Thrombin*

Thrombin is an allosteric protease controlled through exosites flanking the catalytic groove. Binding of a peptide derived from hirudin (Hir52–65) and/or of heparin to these opposing exosites alters catalysis. We have investigated the contribution of subsites S2′ and S3′ to this allosteric transition by comparing the hydrolysis of two sets of fluorescence-quenched substrates having all natural amino acids at positions P2′ and P3′. Regardless of the amino acids, Hir52–65 decreased, and heparin increased thek cat/K m value of hydrolysis by thrombin. Several lines of evidence have suggested that Glu192 participates in this modulation. We have examined the role of Glu192 by comparing the catalytic activity of thrombin and its E192Q mutant. Mutation substantially diminishes the selectivity of thrombin. The substrate with the “best” P2′ residue was cleaved with ak cat/K m value only 49 times higher than the one having the “least favorable” P2′ residue (versus 636-fold with thrombin). Mutant E192Q also lost the strong preference of thrombin for positively charged P3′ residues and its strong aversion for negatively charged P3′ residues. Furthermore, both Hir52–65 and heparin increased thek cat/K m value of substrate hydrolysis. We conclude that Glu192 is critical for the P2′ and P3′ specificities of thrombin and for the allostery mediated through exosite 1.

Thrombin is an allosteric protease controlled through exosites flanking the catalytic groove. Binding of a peptide derived from hirudin (Hir 52-65 ) and/or of heparin to these opposing exosites alters catalysis. We have investigated the contribution of subsites S 2 and S 3 to this allosteric transition by comparing the hydrolysis of two sets of fluorescence-quenched substrates having all natural amino acids at positions P 2 and P 3 . Regardless of the amino acids, Hir 52-65 decreased, and heparin increased the k cat /K m value of hydrolysis by thrombin. Several lines of evidence have suggested that Glu 192 participates in this modulation. We have examined the role of Glu 192 by comparing the catalytic activity of thrombin and its E192Q mutant. Mutation substantially diminishes the selectivity of thrombin. The substrate with the "best" P 2 residue was cleaved with a k cat /K m value only 49 times higher than the one having the "least favorable" P 2 residue (versus 636-fold with thrombin). Mutant E192Q also lost the strong preference of thrombin for positively charged P 3 residues and its strong aversion for negatively charged P 3 residues. Furthermore, both Hir 52-65 and heparin increased the k cat /K m value of substrate hydrolysis. We conclude that Glu 192 is critical for the P 2 and P 3 specificities of thrombin and for the allostery mediated through exosite 1.
Thrombin (1, 2) is a multifunction serine protease finely tuned through: 1) restrictions fulfilled by the nature of the P 3 to P 3 Ј residues of the substrate, 1 that the S 3 to S 3 Ј subsites of the protease must accommodate, 2) steric hindrance resulting from surface loops, that limit access to the active site, 3) secondary exosites (one apolar and two positively charged) that strengthen the binding of cognate macromolecules (e.g. substrates such as fibrinogen but also inhibitors such as hirudin), and 4) allosteric control conferred by various cofactors (3). Several lines of evidence suggest that Glu 192 is a major player in the specificity of thrombin (4). 2 Overall, hydrolysis by throm-bin of its "best" substrates is little affected by the E192Q mutation, but new activities emerge that are severely restrained in normal thrombin. The E192Q mutant activates bovine factor X (5) and is efficiently inactivated by the serpin ␣ 1 -antitrypsin (6) and by the bovine pancreatic trypsin inhibitor (BPTI) 3 (7,8). Glu 192 also seems to participate in the allosteric modulation of thrombin (4,9). E192Q activates the anticoagulant protein C faster than thrombin; however, in the presence of the cofactor thrombomodulin, this difference disappears (4). Notably, the side chain of Glu 192 changes its conformation upon binding of a ligand to exosite 1 (10 -12).
Among the effectors altering thrombin catalysis, molecules as diverse as inorganic ions, tryptamine analogs, polysaccharides, and proteins (or peptides derived from these proteins) have been identified. Sodium ions are allosteric modulators of thrombin, along with other serine proteases having a tyrosine in position 225 in the chymotrysin numbering system (13,14) and tryptamine analogs enhance the ability of thrombin to activate protein C in the absence of calcium (15). Binding of a ligand to the positively charged exosite 1 also induces an allosteric transition in the active site of thrombin (16 -23). Specifically, binding to exosite 1 of a peptide derived from residues 53-64 of hirudin variant 2 (hirugen; NGDFEEIPEEYL), 4 from residues 53-67 of heparin cofactor II (EGEEDDDYLDLEKIF), from residues 52-69 of the thrombin receptor (PAR-1; KYEPF-WEDEEKNE), or from residues 269 -287 of platelet glycoprotein-Ib␣ (DEGDTDLYDYYPEEDTEGD) alters hydrolysis of peptidyl substrates. In general, these acidic peptides decrease the K m of substrates having a proline in P 2 position, whereas they increase the K m of those having a P 2 glycine. Catalysis of thrombin is also altered by the binding to exosite 1 of thrombomodulin, its soluble derivative containing growth factor-like repeats 4 -6, fibrinogen fragment E, or the peptide corresponding to residues 410 -427 of the ␥Ј-chain (17, 24 -27). However, only thrombomodulin and its soluble derivative allow rapid activation of protein C, increase the rates of thrombin inhibition by BPTI or the protein C inhibitor, and increase the rate of activation of the thrombin-activable fibrinolysis inhibitor (23, 28 -31). Thrombomodulin induces multiple conformational changes in the active site of thrombin, but alterations within the S 3 , S 2 , and S 1 subsites are similar to those induced by the binding of hirugen and thus do not explain, alone, the effect of * This work was supported in part by INSERM of France, by the Fondation pour la Recherche Médicale of France, and by the Accord de Coopération Biomédicale Franco-Brésilienne (INSERM/FAPESP 1999/ 2000). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Binding of a ligand to the second positively charged exosite of thrombin also alters catalysis. In particular, heparin (a polydisperse mucopolysaccharide of M r 8,000 -30,000) greatly increases the rate of thrombin inactivation by antithrombin and heparin cofactor II (34 -36). Other ligands such as prothrombin fragment 2 or a synthetic peptide derived from its residues 63-116, as well as a monoclonal antibody recognizing an epitope that overlap with exosite 2, also induce conformational changes in the active site of thrombin (37)(38)(39)(40)(41)(42)(43).
The effect of the simultaneous addition of exosite 1 and 2 ligands on thrombin activity is not fully understood. Heparin acts as a noncompetitive inhibitor for the formation of the thrombin-hirudin complex, suggesting a link between exosites 1 and 2 (44). Also consistent with the hypothesis that binding of a ligand to either of the exosites influences binding to the other, the 63-116 derivative of prothrombin fragment 2 increases the apparent K d of hirugen (40). Hence, the monoclonal antibody recognizing part of the exosite 2 of thrombin can bind simultaneously with hirugen (42). Also arguing that ligands of exosite 1 and 2 may bind thrombin simultaneously, bothrojaracin uses both exosites to inhibit thrombin (45). Finally, binding of heparin and fibrin contributes additively to the changes in the active site environment of thrombin, and fibrin (but not thrombomodulin) protects thrombin from inactivation by heparinantithrombin (46 -49).
Using a complete set of fluorescence-quenched substrates, we have compared the specificity of the "primed" subsites of thrombin with that of its E192Q mutant. Our results reveal that the mutation dramatically modifies the P 2 Ј and P 3 Ј preferences, which become much less restricted in E192Q as compared with thrombin. In an attempt to correlate this alteration with those triggered by exosite binding, we have systematically determined the rates of cleavage in the presence of an analog of hirugen based on the sequence of hirudin variant 1 (Hir 52-65 ), of heparin, and of both ligands simultaneously. Our results established opposing modulation for thrombin and E192Q.

EXPERIMENTAL PROCEDURES
Materials-Fluorescence-quenched substrates and Hir 52-65 (peptide NDGDFEEIPEEYLQ) were prepared via Fmoc/t-butyl strategy, as described previously (3). Lyophilized substrates were resuspended in N,Ndimethylformamide, and their concentrations were determined spectrophotometrically by assuming an absorption coefficient of 10 4 M Ϫ1 cm Ϫ1 at 360 nm. Concentration of the stock solution (about 50 mM) was high enough that during the progress curve kinetic experiments, the final amount of N,N-dimethylformamide never exceeded 0.2%. Human thrombin and its recombinant E192Q mutant were prepared as described previously (4,50). Unfractionated porcine mucosal heparin was obtained from Grampian Enzymes (Aberdeen, UK), and the substrates pyroglutamyl-Pro-Arg-p-nitroanilide (S-2366) and H-D-Phe-pipecolyl-Arg-p-nitroanilide (S-2238) were supplied by Chromogenix (Mölndal, Sweden).
Determination of the k cat /K m Values by Progress Curve Kinetics-The active site concentrations of trypsin (tosylphenylalanyl chloromethyl ketone-treated; Worthington, Lorne Laboratories, Twyford, UK), thrombin, and its E192Q variant were determined by titration with D-Phe-Pro-Arg-CH 2 Cl (36), and the k cat /K m values for the hydrolysis of the various fluorescence-quenched substrates were estimated as described previously (3). Briefly, 0.5 ml of the substrate solution (0.5 to 5 M) in 50 mM Tris-HCl, pH 7.8, containing 0.15 M NaCl and 0.2% (w/v) poly(ethylene glycol) (M r 6000) was equilibrated at 37°C for 10 -15 min in the cell holder of a Perkin-Elmer spectrofluorimeter (model LS-50 B). The reaction was started by the addition of 10 l of enzyme (lowest final concentration, 0.1 nM; highest final concentration, 100 nM) and followed for up to 4 h (until at least 80% completion) by monitoring the increase in fluorescence with time at em ϭ 414 nm (slit, 4 nm) and ex ϭ 325 nm (slit, 10 nm). The pseudo-first order rate constants (k obs ), the initial fluorescence of the uncleaved peptide (I o ), and the maximum fluorescence intensity (I max ) were estimated by nonlinear curve fitting of the dependence of the fluorescence intensity (INT) with time (t) using the following equation.
where E represents the enzyme concentration. Initial estimates for the values of I o and I max (used in the nonlinear curve fitting analysis) were obtained independently by measuring the fluorescence in the absence of enzyme and after hydrolysis by 100 nM trypsin for 1 h, respectively. Two initial concentrations of substrate, varying by a factor of at least 2, were systematically essayed. With both concentrations, the values of k obs were within the experimental error (Ϯ10%), suggesting that k obs could be equated to the k cat /K m value of the reaction (3). [52][53][54][55][56][57][58][59][60][61][62][63][64][65] and Heparin for Thrombin and Its E192Q Mutant-In the presence of an allosteric effector ([L]), k obs will be a function of the concentrations of the enzyme free of ligand ([E free ]) and that of the enzyme in complex with its ligand ([E-L]) as shown in the following equation.

Determination of the K d Values of Hir
k cat may also be expressed as a function of the total amount of enzyme ([E total ]).
Combining Equations 3 and 4 results in the following equation. The

RESULTS
The E192Q Mutation Abolishes the Restrictions that Govern S 2 Ј and S 3 Ј Specificities of Thrombin-A number of studies suggest that residue 192 of serine proteases determines in part their specificity (4 -9, 51-54). Although it is well established that substituting Gln for Glu at position 192 in thrombin alters its P 3 , P 2 , and P 3 Ј preferences, little is known about the possible outcome of the mutation with respect to its P 2 Ј preferences. To fully characterize alterations in the S 2 Ј and S 3 Ј subsites, we have synthesized two series of fluorescence-quenched substrates, each having the sequence Abz-VGPRSXXLK(Dnp)D, where the sequence Pro-Arg-Ser is optimal for thrombin cleavage, and X represents any one of the natural amino acids (except cysteine). For each substrate, we have estimated the value of k obs for its hydrolysis by thrombin, its E192Q variant, or trypsin (Table I). We then deduced for subsites S 2 Ј and S 3 Ј of each protease a selectivity index, which is defined as the ratio of the k obs for the best substrate over that for the worst (higher indexes indicate greater selectivity). We reasoned that this approach might neglect possible cooperative effects between subsites but would at least disclose the importance of subsites S 2 Ј and S 3 Ј. Precisely, results revealed a dramatic difference between thrombin and its E192Q variant with respect to their S 2 Ј selectivity ( Fig. 1); indexes were 636 and 49, respectively (20 for trypsin). Thus the simple exchange of the P 2 Ј residue could transform a peptide from being an excellent to a mediocre substrate of thrombin, whereas all substitutions have much less effect with E192Q and very little with trypsin. Hence, the overall preferences of E192Q were similar to thrombin; the most favorable P 2 Ј residue remained phenylalanine, and the most detrimental was aspartate; only the magnitude changed. Rather than switching specificity, the E192Q mutation appeared to remove the strong restriction normally conferred by the S 2 Ј subsite of thrombin. The selectivity was lost primarily by lessening the strongest restrictions. The substrates with the most favorable residues (Phe and Tyr) were cleaved 2-fold faster by E192Q; those with the deleterious residues (Asp, Glu) were cleaved 15 and 26-fold faster, respectively. Nevertheless, the E192Q mutant no longer preferred arginine over aliphatic residues; from the fourth preferred residue with thrombin, arginine became the twelfth with E192Q. Thus, in thrombin, the negative charge of Glu 192 appeared both to prevent accommodation of negatively charged residues in P 2 Ј and to facilitate that of cationic side chains. The preferences of trypsin were not only weak, they were also slightly different. Thrombin and E192Q preferred bulky aromatic side chains, trypsin preferred aliphatic side chains. Negatively charged residues were, as with thrombin and E192Q, the least favorable amino acids. Surprisingly, trypsin exhibited a modest preference for argi-nine (as thrombin) despite having a glutamine in position 192 (as the E192Q variant).
The outcome of the E192Q mutation on the P 3 Ј preferences of thrombin conformed with the above observations concerning the S 2 Ј subsite. The E192Q mutation also decreased by an order of magnitude the selectivity index (Table I), and the largest effects were observed with the least favorable residues (again, Asp and Glu). In essence, the E192Q mutation removed the constraints of the S 3 Ј subsite; the selectivity index was only 3.4 with E192Q, a value comparable with that of trypsin (2.0), reflecting indifference toward the substrate P 3 Ј side chain. Overall, the E192Q mutation specifically destroyed the ability of thrombin to attract or repel charged side chains in its S 2 Ј and S 3 Ј subsites, even though arginine was still the preferred P 3 Ј residue and acidic side chains were still unfavorable. Consistent with this concept, the E192Q mutation also alters the P 2 and P 3 preferences of thrombin, but rather than switching its specificity, it appears to minimize the restrictions that thrombin normally applies at these positions (4,5,9).
Exosite 1 Modulation of the S 2 Ј and S 3 Ј Subsites-In a previous study (9), we challenged the hypothesis that the E192Q mutation mimicked the allosteric alteration induced by Hir 52-65 in the S 3 subsite of thrombin (4). The data were in favor of exosite binding, causing changes in the conformation of the S 2 and/or S 1 subsites of thrombin rather than mitigating unfavorable interactions between Glu 192 and acidic residues in P 3 position of the substrate. In the present study, we have extended our previous investigation to subsites S 2 Ј and S 3 Ј of thrombin. We first verified that affinities of Hir 52-65 for thrombin and E192Q were comparable by measuring the k obs of H-D-Phe-pipecolyl-Arg-p-nitroanilide hydrolysis by each enzyme in the presence of increasing concentrations of the effector. Likewise, similar K d values for Hir 52-65 with either enzyme were obtained when the affinity was estimated using fluorescence-quenched substrates instead of the peptidyl-p-nitroanilide (Fig. 2). The effect induced by Hir 52-65 , however, depended upon the substrate used to detect the binding. With thrombin and E192Q, Hir 52-65 increased the value of k obs for H-D-Phepipecolyl-Arg-p-nitroanilide hydrolysis. In contrast, with most Abz-VGPRSXXLK(Dnp)D substrates, Hir 52-65 roughly halved the k obs value with thrombin but still increased that with E192Q (Fig. 2). The changes in the value of k obs were indeed the result of the effector binding to thrombin or E192Q and not of an interaction between the effector and the substrate or other TABLE I Pseudo-first order rate constants (k obs ) values for the cleavage by thrombin, its E192Q mutant, or trypsin, of the P 2 Ј and P 3 Ј series of fluorescence-quenched substrates Only residues P 1 Ј-P 3 Ј of the substrate (Abz-VGPRSXXLK(Dnp)D) are indicated; the amino acid specific to each substrate is underlined. All estimates of the k obs values (in M Ϫ1 s Ϫ1 ) had standard errors of less than 7%, and values represent the weighted mean of at least two determinations. The selectivity index of each enzyme corresponds to the ratio, within a series, of the highest over the lowest k obs values.
Concerning the P 3 Ј series of substrates, the changes induced by Hir 52-65 on the catalysis of thrombin and E192Q matched those observed with the P 2 Ј series. The k obs value of substrate cleavage in the presence of Hir 52-65 was decreased with thrombin and increased with E192Q (Table II). Furthermore, Hir 52-65 had little or no influence on the cleavage by E192Q of the peptide carrying a glutamate or an aspartate in P 3 Ј, whereas it halved the k obs values of their cleavage by thrombin. Thus, the greatest difference between thrombin and E192Q was again on the hydrolysis of the substrate having an aspartate in P 3 Ј (k obs value 14-fold higher versus 6 without Hir 52-65 ). The fluorescence-quenched substrates used in this study carried a C-terminal aspartate (in position P 6 Ј) that had been added to improve their solubility. Given the critical role of the negative charges in the allosteric regulation of thrombin, we wondered whether formation of the Michaelis complex with a thrombin carrying Hir 52-65 was not hampered. A potential conflict existed between the C-terminal aspartate of the substrate and Hir 52-65 , which itself was heavily negatively charged. Such a conflict could explain the paradoxical allostery induced by Hir [52][53][54][55][56][57][58][59][60][61][62][63][64][65] . In contrast to our fluorescence-quenched substrates, the p-nitroanilide substrates having a proline in the P 2 position are cleaved with higher k obs values in the presence of Hir 52-65 (9,18). To test this hypothesis, fluores-cence-quenched substrates lacking the C-terminal aspartate were synthesized. Solubility of Abz-VGPRSFLLK(Dnp) and Abz-VGPRSWLLK(Dnp) allowed estimation of the k obs values in experimental conditions that were the same as those for the peptides having a C-terminal aspartate. The values of k obs were similar, whether or not substrate carried a C-terminal aspartate, suggesting that residue P 6 Ј of a substrate had little influence, if any, on thrombin catalysis. The allosteric modulation triggered by Hir 52-65 was also of similar magnitude and in the same direction for both types of substrates. Therefore, the allostery induced by Hir 52-65 must have resulted from an alteration of the ability of protease to accommodate its substrate in a way that did not implicate the P 6 Ј residue. Accordingly, the observation that Hir 52-65 induced opposing effects on thrombin and E192Q catalysis must reflect differences between these proteases. This allowed us to conclude that far from mimicking the effect of Hir 52-65 binding to exosite 1, the E192Q mutation had an adverse effect on the allosteric regulation of thrombin.

Residues
As for the exosite 1 study, we first verified that the E192Q mutation did not grossly modify the affinity of heparin. K d values of 0.21 and 0.36 unit/ml were obtained for thrombin and E192Q, respectively. Possible heterotropic effects between the substrate and heparin were also ruled out by comparing hydrolysis of the Abz-VGPRSXXLK(Dnp)D substrates by trypsin in the presence or absence of heparin (Table II). On average, the value of k obs for the hydrolysis of the substrates by thrombin was increased 1.4-fold by heparin, and that for the hydrolysis by E192Q was increased 1.7-fold. With both thrombin and E192Q, higher increases in the k obs values were obtained with the substrates having either a lysine or an arginine in P 2 Ј position (about 1.8-fold with thrombin and 2.3-fold with E192Q). Thus, the effect of heparin on E192Q catalysis was comparable with that on thrombin, albeit more pronounced. With the P 3 Ј series of fluorescence-quenched substrates, the changes induced by heparin on the catalysis of thrombin and E192Q mirrored the above observations concerning the P 2 Ј series; in the presence of heparin, substrates were cleaved by thrombin and E192Q with higher k obs values (1.2-and 1.5-fold on the average, respectively; Table II). Thus, in contrast to the allostery mediated through exosite 1, the E192Q mutation did not overturn the effect induced by heparin. The relatively uniform effect of heparin on thrombin and E192Q catalysis also suggested that subsites S 2 Ј and S 3 Ј were not directly involved in this allostery. Therefore, our results confirm that a linkage exits between exosite 2 and the catalytic groove of thrombin but imply that this linkage is not mediated through Glu 192 . Simultaneous Modulation of Thrombin and E192Q by Hir [52][53][54][55][56][57][58][59][60][61][62][63][64][65] and Heparin-The full cofactor activity of thrombomodulin involves a chondroitin sulfate carried by this integral membrane receptor, suggesting that simultaneous binding to exosites 1 and 2 may occur. Hence, a study has reported that rather than simply circling thrombin, thrombomodulin may also cross-link two molecules of thrombin, raising doubts on the bivalent binding hypothesis (55). In addition, a linkage exists between exosites 1 and 2, even if binding of heparin to exosite 2 does not exclude occupancy of exosite 1 (39,40,48). In an attempt to reconcile the puzzling consequences of the simultaneous presence of exosites 1 and 2 effectors on thrombin activity, we have determined the effect of a mixture of heparin and Hir 52-65 on the hydrolysis of the fluorescence-quenched substrates. With thrombin, the simultaneous presence of the two effectors resulted in values of k obs that were intermediate between those obtained in the presence of each effector (Table II). Interpretation was difficult because opposing effects were induced by each effector separately. There was enough heparin and Hir 52-65 to saturate thrombin with either one or both, such that the concentration of free thrombin would be negligible during the assay. In a purely ternary complex model (i.e. heparin and Hir 52-65 bound simultaneously to thrombin), the progress curve of hydrolysis would simply reflect the simultaneous allosteric contribution of each exosite to a single species catalyzing a single reaction, which should be adequately represented by Equation 1. In fact, it was possible to analyze all data obtained in the presence of both heparin and Hir 52-65 according to Equation 1. In a strictly mutually exclusive binding model (i.e. thrombin bound heparin or Hir 52-65 , but not both), the progress curve of hydrolysis would reflect the independent contribution of two enzyme complexes (thrombin-Hir 52-65 and thrombin-heparin) catalyzing the same reaction (in parallel).
In this equation, ␥ would be the percentage of the total thrombin concentration in complex with Hir 52-65 , the remaining (1 -␥) being in complex with heparin; parameters k obs(1) and k obs (2) would be the k obs values determined in the absence of the other effector. With only one exception, it was feasible to analyze the progress curve data according to this mutually exclusive model. The nonlinear curve fitting analysis consistently provided a value for ␥ comprised between 0.2 and 0.6 (i.e. 20 -60% of thrombin in complex with Hir 52-65 , the remaining with heparin). Thus, most data were compatible with both the ternary complex as well as the mutually exclusive models, and it was not possible to discriminate between them by the nonlinear curve fitting approach. However, the fluorescence-quenched substrate having a proline in P 2 Ј position was cleaved with a k obs value higher in the presence of either Hir 52-65 (1.3-fold) or heparin (also 1.3-fold) and was cleaved with a k obs value even higher when both effectors where present (1.6-fold). We searched for another peptidyl substrate of thrombin whose rate of hydrolysis would be influenced in the same direction by heparin and Hir 52-65 and found that this was true with pyroglutamyl-Pro-Arg-p-nitroanilide. The value of k obs was higher in the presence of either Hir 52-65 or heparin and was even higher when both effectors were added simultaneously (Fig. 3). Obviously, such data could no longer be analyzed according to the mutually exclusive model, whereas they were still compatible with the ternary complex model. With the E192Q mutant, either effector increased the k obs values of hydrolysis, and when both Hir 52-65 and heparin were added, the value of k obs was even higher. Thus, the results were undoubtedly in favor of a simultaneous binding of the two effectors to a single molecule of E192Q. Therefore, our data suggest that heparin and Hir [52][53][54][55][56][57][58][59][60][61][62][63][64][65] can form a ternary complex with thrombin or E192Q, and it is reasonable to believe that the k obs values obtained with thrombin for the hydrolysis of the fluorescence-quenched substrates in the presence of the two effectors reflected the activity of this ternary complex. Hence, the value of k obs obtained in the presence of the two effectors did not equate to the sum of those obtained with each effector separately, suggesting that a linkage exists between the two exosites (Fig. 4). This linkage did not involve a decrease of the affinity of the ligands for their exosite. In the presence of 2 units/ml heparin, the k obs values for the cleavage of the Abz-VGPRSFLLK(Dnp)D substrate were virtually identical whether Hir 52-65 concentration was 10, 30, or 100 M (3.6, 3.5, and 3.6 10 6 M Ϫ1 s Ϫ1 , respectively), implying that the K d of Hir 52-65 for the E192Q mutant did not increase upon binding of heparin to exosite 2.

DISCUSSION
Overall our data substantiate and extend to the substrate leaving group the concept that specificity of the catalytic groove of thrombin originates from an exclusion of the unfavorable P 3 -P 3 Ј side chains rather than from a positive selection of favorable ones. Thrombin cleaved its best fluorescencequenched substrates with a k cat /K m value comparable with that of trypsin and the other substrates with k cat /K m values up to 15000-fold lower than trypsin. Thus, with respect to the catalytic groove, thrombin is a restricted trypsin. Glu 192 of thrombin emerged as a major determinant in this ability to repel unfavorable P 2 Ј and P 3 Ј side chains and also appeared to be involved in the allostery mediated through binding of an effector to exosite 1. However, it is unclear whether the link between exosite 1 and residue 192 is direct, indirect, or fortuitous. In contrast, Glu 192 seemed to play a minor role in the allostery of thrombin mediated through exosite 2, and heparin exerted a similar effect on thrombin and E192Q catalysis.
A direct link between Glu 192 and exosite 1 would imply that upon binding of an allosteric effector, the glutamate changes its conformation. Consistent with this hypothesis, the side chain of Glu 192 adopts different orientations in the x-ray structures of thrombin (10 -12, 56, 57). In particular, the side chain of Glu 192 occupies in free bovine thrombin a different conformation than when in complex with an inhibitor, an exosite ligand, or both. Compared with the D-Phe-Pro-Arg-CH 2 Cl-thrombin structure, the binding of hirugen produces only minor alterations in the orientation of the side chains of the catalytic triad and of the exosite 1 residues, whereas changes occur along segment Ala 190 -Gly 197 , which reach a maximum near Glu 192 (10). Whether this conformational difference of residue 192 is a consequence of hirugen or of D-Phe-Pro-Arg-CH 2 Cl binding remains unclear, but hirugen does not seem capable of inducing a conformational change when the active site is occupied (11). Several of the catalytic properties of E192Q thrombin suggest that the mutation mimics in part the modulation of thrombin by exosite 1 effectors. Compared with thrombin, the E192Q mutant activates protein C faster (4) and is inhibited by BPTI with a K I value 3 orders of magnitude lower (7). Thrombomodulin dramatically increases the rate of protein C activation by thrombin and increases 10-fold the k on value of thrombin inhibition by BPTI (31). Nevertheless, ␣ 1 -antitrypsin neutralizes E192Q (6), and E192Q activates bovine factor X (5), two functions that thrombomodulin is unable to trigger in thrombin. Furthermore, thrombomodulin also modifies the catalytic properties of E192Q (4), and as shown in this study, Hir 52-65 has an effect opposite on thrombin and E192Q catalysis. Thus, even if some of the modifications introduced by the E192Q mutation mimic (in part) the allostery triggered by exosite 1 ligands, they cannot explain alone the catalytic switch of thrombin. Undoubtedly, the link between Glu 192 and exosite 1 is more elaborate.
The link between exosite 1 of thrombin and Glu 192 could be indirect; that the conformation of Glu 192 changes upon binding of hirugen does not preclude that other modifications occur simultaneously. If the allosteric effector induces one (or several) conformational changes that are shared upon mutation of Glu 192 in thrombin, then the E192Q mutant and the complex may share catalytic properties that differ from free thrombin. It is well established that the autolysis loop of thrombin, in the vicinity of the S 2 Ј subsite, adopts various conformations (1,58). Docking to exosite 1 of the thrombin receptor peptide or of Hir 52-65 , as well as covalent binding to the active site of D-Phe-Pro-Arg-CH 2 Cl, all induce a reorganization of the autolysis loop (11,59). In particular, inhibited thrombin is 95-fold less susceptible to chymotrypsin cleavage within this loop than free thrombin (60). Conversely, mutations in the autolysis loop of thrombin alter hydrolysis of p-nitroanilide substrates that are devoid of PЈ residues (61,62). Thus, there is a link in thrombin between the autolysis loop and the subsites of the catalytic groove. Whether the E192Q mutation induces a similar alteration of the autolysis loop remains to be determined, but our study demonstrates a crucial role of Glu 192 to the P 2 Ј specificity of thrombin, raising the possibility that the S 2 Ј subsite is altered by the mutation. Crystallographic data, however, argue against this hypothesis; the conformation of the catalytic groove moiety comprising subsites S 2 Ј and S 3 Ј is the same in thrombin and in the complex of E192Q with BPTI (8). The side chain of Glu 192 also appears too distant to interact directly with the main chain of the P 2 Ј residue of Hirulog 3 (63). Thus, the link between exosite 1 and residue Glu 192 of thrombin must involve an alternate mechanism. A gross structural rearrangement of the 60-loop insertion occurs upon BPTI binding to E192Q; crystallographic data also demonstrate a structural link between the expulsion of the 60-loop and a reorganization of the 39-loop in the vicinity of exosite 1. It is therefore conceivable that the reverse process occurs with strong binding to exosite 1 facilitating expulsion of the 60-loop. Consistent with this hypothesis, the 60-loop insertion of thrombin forms in part the S 2 and S 3 subsites, and the E192Q mutation alters the P 2 and P 3 specificities (5, 9, 50). The deletion of three residues from the 60-loop insertion of thrombin (mutation des-PPW) also results in a nanomolar K I value for BPTI binding (64). FIG. 4. Comparison between the theoretical and observed values of k obs for the hydrolysis by E192Q of the P 2 series of fluorescence-quenched substrates in the presence of Hir 52-65 and heparin. Shown is a graph of the percentage of activity (k obs value without effector taken as reference) versus the P 2 Ј amino acid of the substrate (shown in one-letter codes). The concentration of each effector in the mixture was ample enough to saturate the E192Q mutant. The experimental values of k obs obtained in the presence of both effectors (Ⅺ) were systematically higher than the k obs values with either Hir 52-65 (E) or heparin (f) yet were lower than their sum (q). This observation suggests that a ternary complex between the enzyme Hir 52-65 and heparin can form, but that exosite 1 and 2 of thrombin are linked.
However, mutations E192Q and des-PPW are not interchangeable; they are additive with respect to BPTI binding, resulting in picomolar affinity (7). Furthermore, binding of fibrinogen, antithrombin, thrombomodulin, and hirudin is preserved by the E192Q mutation, yet is dramatically weakened by the des-PPW mutation. Thus the E192Q mutation in itself is unlikely to account for a major conformational change in the 60-loop; the mutation would rather favor a destabilization of the molecule triggered by the binding of BPTI.
Finally, the link between exosite 1 and Glu 192 may be deceptive. The allosteric effector and the E192Q mutation may use unrelated alterations to change the k obs value of fluorescencequenched substrate hydrolysis. The k obs value for the hydrolysis of a substrate depends on the binding energy, its turnover by the enzyme, and/or its approaching orientation. Glu 192 of thrombin occupies a central position in the catalytic groove and could influence binding as well as turnover. An increase in the number of productive collisions will be obtained if the substrate is guided to its docking site by a suitable electrostatic potential (65). The very high k on value for the inhibition of thrombin by hirudin results from such complementary electrostatic fields (66,67). Thrombin carries two strong positive patches that sandwich the catalytic groove, itself constituting a strong negative patch. Undoubtedly, the E192Q mutation obliterates in part this negative patch. Small substrates would not develop a strong electrostatic field, but the changes observed were small. In fact, the variations of the k obs values were related in part to the charge of the substrate; those carrying a lysine or an arginine in P 3 Ј were cleaved with higher k obs values by thrombin than by E192Q, and those with an aspartate or a glutamate were cleaved with higher k obs values by E192Q than by thrombin. All substrates of the P 2 Ј series were cleaved more efficiently by E192Q, but the increase of the k obs value was more pronounced for the substrates having negative P 2 Ј side chain. Thus, it is conceivable that an altered electrostatic field in E192Q allows for a better access/orientation to the negatively charged substrates or, conversely, impedes access/orientation to the positively charged substrates. On the other hand, the binding of Hir 52-65 to exosite 1 of thrombin neutralizes in part the positive patch that follows subsites S 3 Ј of the catalytic groove. Again, there was a (small) link between the charge of the P 3 Ј residue of the substrate and the changes of the k obs value. Complexes of Hir 52-65 with thrombin cleaved all substrates less efficiently than the free enzyme, but the decrease of the k obs value was less pronounced for the substrates having positive P 3 Ј side chain and was more pronounced for those having negative side chains. It is conceivable that neutralization of exosite 1 worsens the initial orientation of substrates negatively charged at P 3 Ј. Conversely, even if complexes of E192Q with Hir 52-65 cleaved all substrates more efficiently than E192Q, the increase of the k obs value was more pronounced for the substrates having a positive P 3 Ј side chain and less pronounced for those having a negative side chain. These observations are consistent with the hypothesis that binding of Hir 52-65 and the E192Q mutation influence the catalysis independently.