Mapping of the Catalytic Groove Preferences of Factor Xa Reveals an Inadequate Selectivity for Its Macromolecule Substrates*

Factor Xa (FXa) hydrolyzes two peptide bonds in prothrombin having (Glu/Asp)-Gly-Arg-(Thr/Ile) for P3-P2-P1-P1′ residues, but the exact preferences of its catalytic groove remain largely unknown. To investigate the specificity of FXa, we synthesized full sets of fluorescence-quenched substrates carrying all natural amino acids (except Cys) in P3, P2, P1′, P2′, and P3′ and determined the k cat /K m values of cleavage. Contrary to expectation, glycine was not the “best” P2 residue; peptide with phenylalanine was cleaved slightly faster. In fact, FXa had surprisingly limited preferences, barely more pronounced than trypsin; in P2, the ratio of thek cat /K m values for the most favorable side chain over the least was 289 (12 with trypsin), but in P1′, this ratio was only 30 (versus 80 with trypsin). This unexpected selectivity undoubtedly distinguished FXa from thrombin, which exhibited ratios higher than 19,000 in P2 and P1′. Thus, with respect to the catalytic groove, FXa resembles a low efficiency trypsin rather than the highly selective thrombin. The rates of cleavage of the peptidyl substrates were virtually identical whether or not FXa was in complex with factor Va, suggesting that the cofactor did not exert a direct allosteric control on the catalytic groove. We conclude that the remarkable efficacy of FXa within prothrombinase originates from exosite interaction(s) with factor Va and/or prothrombin rather than from the selectivity of its catalytic groove.

At the confluence of the formerly named intrinsic and extrinsic pathways, factor Xa (FXa) 1 is the midway protease of the blood clotting waterfall (1). FXa belongs to clan SA of the S1 family of serine peptidases along with thrombin and trypsin (2)(3)(4)(5). Without cofactors, activation of prothrombin by FXa is slow; it becomes efficient only when FXa complexes factor Va to form prothrombinase (6). Rapid inhibition of FXa by antithrombin also requires heparin as cofactor (7). However, tissue factor pathway inhibitor (TFPI) does not require any cofactor to rapidly neutralize FXa (8). FXa catalyzes a number of other reactions: activation of factor VII in a positive feedback within the tissue factor pathway (9), activation of factors V (10) and VIII (11), cleavage of protease-activated receptor 2 (12), and neutralization of protein S, albeit only in the presence of phospholipid and calcium (13). Thrombin (14) requires a cofactor for activation of protein C and factor XI, as well as for its inhibition by antithrombin and heparin cofactor II. In contrast to FXa, however, thrombin alone rapidly catalyzes a number of critical reactions in the cascade: cleavage of fibrinogen, activation of factors V and VIII, and activation of protease-activated receptor 1 (6,10,15). Trypsin, the archetypal endopeptidase of the digestive tract, does not require cofactors to rapidly hydrolyze (in appropriate conditions) most peptide bonds that follow an arginine or a lysine. The notable specificity of the blood coagulation peptidases result from at least four molecular mechanisms: (i) constraints built up by subsites 3 to 3Ј (S 3 to S 3 Ј) 2 of the catalytic groove, which accommodate the P 3 -P 3 Ј residues of the substrate or inhibitor; (ii) steric restrictions caused by surface loops surrounding the catalytic groove; (iii) exosites remote from the catalytic groove, which may anchor complementary motifs of the overall substrate protein; and (iv) cofactors that may overturn the specificity of the protease (4, 16 -21).
A comparison of the P 3 -P 3 Ј sequences of the known substrates and inhibitors of FXa suggests that glycine in P 2 and serine in P 1 Ј could favor catalysis. In this paper, we report a comprehensive study of FXa subsites preferences using fluorescence-quenched substrates. We also compared FXa preferences with those of thrombin and trypsin. Some of the subsite preferences of FXa were unexpected, but the main surprise came from the overall limited selectivity of its catalytic groove. Addition of factor Va, phospholipid, and calcium had no detectable influence on FXa preferences or on its catalytic efficiency. Thus, we conclude that the remarkable efficacy of FXa within prothrombinase must rely on exosite(s) interaction(s) rather than on a purely allosteric mechanism involving its catalytic groove.

EXPERIMENTAL PROCEDURES
Proteins and Reagents-Prothrombin was purified from human plasma and converted to thrombin as described previously (22). Bovine factor V/Va was from Kordia (Leiden, The Netherlands), and phospholipids vesicles were prepared by sonication of a 1 mg/ml mixture of phosphatidylcholine (80% w/w) with phosphatidylserine (20% w/w) (Sigma) as described previously (23). (24) was amplified by polymerase chain reaction using primers ACG-CGG-ATC-CGC-GAT-GGG-GCG-CCC-ACT-GCA and TCC-CCC-GGG-GGA-TCA-GTT-CAG-GTC-TTC-CTC-GCT-GAT-CAG-CTT-CTG-CTC-CTT-TAA-TGG-AGA-GGA-CGT-TA, which introduced a BamHI restriction site at the 5Ј end and an XmaI site at the 3Ј end, respectively. The 3Ј end primer also inserted in-frame with the C terminus of the FX cDNA a sequence coding for the epitope EEQKLISEEDLLGGY recognized by monoclonal antibody 9E10. The insert between the BamHI and XmaI sites of plasmid pNUT-hGH (22) was substituted for the product of amplification, and the full-length cDNA sequence was verified by dideoxy chain termination sequencing. Transfection and selection of BHK-21 cells were achieved as described previously (22,25). Production and purification of recombinant FX were also performed as described previously, except that c-Myc monoclonal antibody-agarose (CLON-TECH, Saint Quentin, France) was used for the purification. Recombinant FX was eluted by 100 mM glycine-HCl, pH 2.7. The pH was immediately neutralized by adding 30 l/ml of 2 M Tris, and FX was further purified (after dilution 1:4 in 50 mM Tris-HCl, pH 7.5, containing 5 mM EDTA) by anion exchange chromatography onto Q-Sepharose fast flow (Amersham Biosciences) developed with a linear gradient from 0.15 to 0.5 M NaCl in 50 mM Tris-HCl, pH 7.5.

Preparation of Recombinant FXa-Human FX cDNA
Recombinant FX was converted to FXa essentially according to Jesty and Nemerson (26) by incubation with the activator purified from the Russell's viper venom (RVV-X) purchased from Kordia. Briefly, RVV-X (5 mg/ml) in 200 mM NaHCO 3 , pH 8.3, containing 0.5 M NaCl was coupled to a 5-ml HiTrap N-hydroxysuccinimide-activated column (Amersham Biosciences) following the manufacturer's instructions. The column was sealed after loading with 4 mg of FX in 4 ml of kinetic buffer (50 mM Tris-HCl, pH 7.5, containing 0.15 M NaCl, 5 mM CaCl 2 , and 0.2% (w/v) poly(ethylene glycol); M r ϭ 6000), and the incubation was prolonged for 16 h at room temperature. The eluate, diluted 1:3 in 50 mM Tris-HCl, pH 7.5, containing 5 mM CaCl 2 , was loaded onto a 1-ml HiTrap heparin-Sepharose column (Amersham Biosciences) and eluted with 50 mM Tris-HCl, pH 7.5, containing 0.5 M NaCl and 5 mM CaCl 2 . Recombinant FXa appeared pure by SDS-polyacrylamide gel analysis and was indistinguishable from human FXa commercialized by Kordia with respect to its K m and k cat values for the hydrolysis of benzoyl-Ile-Glu(␥-OR)-Gly-Arg-pNA (S-2222, Biogenic, Maurin, France), Z-D-Arg-Gly-Arg-pNA (Biogenic), as well as the rate of prothrombin activation within prothrombinase. Recombinant and commercial FXa were nevertheless systematically compared in all experiments reported in this study and found virtually identical despite the C-terminal 9E10 epitope.
Titration of Trypsin, FXa, and Thrombin-The active site concentration of a stock trypsin solution (bovine, tosylphenylalanyl chloromethyl ketone-treated; Worthington, Lorne Laboratories, Twyford, UK) was determined by titration with p-nitrophenyl-pЈ-guanidinobenzoate. This titrated trypsin was used to determine the precise concentration of aliquots in 1 mM HCl of D-Phe-Phe-Arg-CH 2 Cl and D-Phe-Pro-Arg-CH 2 Cl (Calbiochem, Meudon, France). Briefly, 200 nM trypsin in kinetic buffer was incubated for 3 h at room temperature with various amount of each chloromethyl ketone (0.03-3 M). The reaction mixture was diluted 1:10 in kinetic buffer containing 100 M S-2222, and the remaining enzyme concentration was estimated from the rate of A 405 increase. The initial concentrations of chloromethyl ketone aliquots were deduced from the intercept to the x axis of a linear plot of the remaining activity versus the amount of inhibitor added. The active site concentrations of FXa (recombinant or human) and of thrombin were determined in the same buffer system and experimental conditions using the calibrated aliquots of chloromethyl ketone. For FXa titration, 1 M enzyme according to the A 280 (extinction coefficient of 1.25 ml mg Ϫ1 cm Ϫ1 ) was incubated with 0.1-5 M D-Phe-Phe-Arg-CH 2 Cl, and the remaining free enzyme concentration was measured with 100 M S-2222 as substrate. For thrombin titration, 20 nM enzyme according to the A 280 (extinction coefficient of 2.0 ml mg Ϫ1 cm Ϫ1 ) was incubated with 2-200 nM D-Phe-Pro-Arg-CH 2 Cl, and the remaining free enzyme concentration was measured with 100 M H-D-Phe-pipecolyl-Arg-pNA (Biogenic). Just prior to use, FXa and trypsin dilutions were systematically controlled by measuring the rate of S-2222 hydrolysis, and thrombin dilutions were controlled by measuring the rate of H-D-Phe-pip-Arg-pNA hydrolysi s.
Substrates and Kinetics Experiments-Fluorescence-quenched substrates were prepared by the solid phase method via a Fmoc (N-(9fluorenyl)methoxycarbonyl)/t-butyl strategy, purified by reverse phase chromatography on a C18 column, and their purity was checked by matrix-assisted laser desorption ionization-time-of-flight mass spectrometry (Tof-Spec-E, Micromass) as described previously (27). pNA substrates were prepared by the solution method following the procedure described in Alves et al. (28). Lyophilized substrates were resuspended in a minimum volume N,N-dimethylformamide, such that concentration in the stock solutions was about 5 mM, to ensure that during all kinetic experiments, the final amount of N,N-dimethylformamide never exceeded 0.5% (v/v). The concentration of the fluorescencequenched substrates was estimated from their A 360 , assuming an absorption coefficient of 10 4 M Ϫ1 cm Ϫ1 .
Determination of the k cat /K m values of hydrolysis was performed essentially as described previously (16,29). Assuming that the reaction The five series of fluorescence-quenched substrates had ABz-VQFRSLGDQ-EDDnp for a common framework. The values reported were obtained for the hydrolysis of the substrates by recombinant FXa and represent the weighted mean of at least three determinations (standard error was 18% or less). Cleavage always occurred between the arginine and the serine, even when a second arginine (or a lysine) was in the sequence. The substrate with proline in P 1 Ј (resistant to trypsin hydrolysis) was not determined (ND). The reactions were performed in pseudo-first order conditions, and the k cat /K m values were deduced from the progress curve, assuming a simple Michaelis-Menton mechanism with the encounter of enzyme and substrate limiting. The amino acid in P 3 , P 2 , P 1 Ј, P 2 Ј, or P 3 Ј that varies in each series is listed together with the estimated value of k cat /K m (in M Ϫ1 s Ϫ1 ). Within each series, amino acids are classified from the most favorable to the least favorable. Contrary to expectation, Gly was not the preferred P 2 amino acid; overall substrates with the P 2 aromatic side chain were also cleaved quite efficiently. All k cat /K m values were also estimated (twice) with FXa commercialized by Kordia giving similar, if not identical, results, in particular concerning the preference for P 2 Phe over Gly. The selectivity indexes for each subsite, which are calculated as the ratio between the most favorable side chain over the least, are as follows: P 3 , 6; P 2 , 289; P 1 Ј, 30; P 2 Ј, 19; and P 3 Ј, 6.
obeys a simple Michaelis-Menten mechanism with the encounter of enzyme and substrate limiting, when the reaction is performed at a substrate concentration below K m , the progress curve kinetic allows estimation of the k cat /K m of the reaction. Briefly, hydrolysis at 37°C was monitored by measuring the fluorescence at em ϭ 414 nm (slit, 4 nm) and ex ϭ 325 nm (slit, 10 nm) in a LS50B spectrofluorimeter (PerkinElmer Life Sciences) equipped with a thermostatted microplate reader. A microplate containing 180 l of substrate (1-8 M) in kinetic buffer containing 0.2% (v/v) Tween 20 was left in the thermostatted compartment until the temperature equilibrium was attained (10 -15 min), and the reaction was started by adding 20 l of enzyme (0.1-200 nM). Care was taken to keep the ionic strength constant, in particular the NaCl concentration (30 -32). The increase in fluorescence with time was monitored for up to 4 h (until at least 75% completion of the reaction). The pseudo-first order rate constant (k), the initial fluorescence (I o ), and the maximum fluorescence intensity (I max ) were estimated by nonlinear curve fitting analysis of the fluorescence intensity (INT) dependence on time (t) using the following equation.
where E is the enzyme concentration. The values of I o and I max estimated through the nonlinear curve fitting analysis were always consistent with the fluorescence intensities measured before and after 1 h of incubation of the substrate with 0.1 M trypsin (corresponding to the intact and fully cleaved peptide, respectively). The apparent rate constant will be pseudo-first order and will equal k cat /K m , provided that the initial substrate concentration is much less than its K m value for the enzyme. To assess whether this condition was met, progress curve kinetics were systematically performed at two initial concentrations of fluorescence-quenched substrates. the results indicated that with all of the substrates assayed k was, within the experimental error, identical at the two substrate concentrations. Thus, in this concentration range, the rate of hydrolysis was not dependent upon the amount of substrate, suggesting that k could be equated to the k cat /K m value. When substrates carried several potential cleavage sites, the preferred bond of hydrolysis was determined after purification of the products by reverse phase chromatography. Kinetics were performed in conditions identical to those used for the fluorescence studies except that the reactions were quenched at timed intervals by adding H 3 PO 4

TABLE II
Values of k cat /K m for the cleavage of fluorescence-quenched substrates by trypsin and thrombin The values listed for trypsin hydrolysis were obtained with the same five series of fluorescence-quenched substrates as for FXa (having ABz-VQFRSLGDQ-EDDnp for a common framework). These substrates specifically designed for the FXa study were poor substrates of thrombin, because of the P 2 phenylalanine. Therefore, three additional series of substrates were synthesized to fulfill the thrombin study; they had ABz-VGPRSFRDQ-EDDnp for a common framework in the P 2 and P 3 series and ABz-VGPRSFLDQ-EDDnp in the P 1 Ј series. Also, the values listed for thrombin concerning the P 2 Ј and P 3 Ј series are taken from Marque et al. (29); they were obtained with fluorescence-quenched substrates having ABz-VGPRSFLLK(DNP)D for a common framework. Amino acids in P 3 , P 2 , P 1 Ј, P 2 Ј, or P 3 Ј that varied within each series are listed together with the estimated value of k cat /K m (in M Ϫ1 s Ϫ1 ) representing the weighted mean of at least three determinations (final standard errors were 20% or less). All series of fluorescence-quenched substrates were assayed with trypsin to assess the possible influence of the framework on the results, but selectivity indexes and preferences were similar within matching series. For example, the values for trypsin listed in this study are in accord with those previously reported (28) for the hydrolysis of the series based on the ABz-VGPRSFLLK(DNP)D framework; selectivity indexes for S 2 are similar (31 versus 20), as well as the preferences. The results obtained with the P 3 Ј series are also consistent, even if the selectivity index estimated in the current study is higher than in our previous report (13 versus 2). The selectivity indexes are as follows: for trypsin, P 3 , 11; P 2 , 12; P 1 Ј, 80; P 2 Ј, 31; and P 3 Ј, 13; and for thrombin, P 3 , 51; P 2 , 19,000; P 1 Ј, 24,706; P 2 Ј, 636; and P 3 Ј, 52. The value reported is a lower limit because the substrate was cleaved at more than one site (Fig. 2). b ND, not determined.

Preferred P 2 Amino Acid for FXa Catalysis Was
Phenylalanine-To investigate the catalytic groove preferences of FXa, we synthesized five series of fluorescence-quenched substrates, having a common 10-amino acid-long framework (ABz-VQ-FRSLGDQ-EDDnp). The peptides carried a strongly fluorescent ABz group at the N-terminal end, but a C-terminal EDDnp quenched this fluorescence by resonance energy transfer. The k cat /K m value for the cleavage of each peptide by FXa was estimated by analysis of the increase of the fluorescence intensity upon hydrolysis. In each series of peptides, either the P 3 , P 2 , P 1 Ј, P 2 Ј, or P 3 Ј amino acid was varied, such that complete sets of peptides were prepared covering all natural amino acids (except cysteine). Proline in P 1 Ј was also avoided because it is known to prohibit trypsin cleavage.
Study of the P 2 preferences of FXa revealed that, contrary to expectation, glycine was not the most favorable amino acid; the peptide having a phenylalanine residue in P 2 was cleaved slightly faster than the one with a glycine (Table I). Overall, substrates with aromatic side chains in P 2 were cleaved efficiently, whereas the least favorable amino acid was aspartate. The FXa cleavage sites in prothrombin and factor VII have glycine in P 2 . In fact, a number of studies have established that FXa prefers Gly over Phe in P 2 position with chromogenic or fluorogenic substrates (30,(33)(34)(35)(36)(37). Furthermore, x-ray analysis suggests that Tyr 99 (in the chymotrypsinogen numbering system) 3 normally blocks S 2 of FXa (2, 3). Thus, it was quite surprising that bulky side chains (phenylalanine and tryptophan) could be as efficient as glycine in our study. The discrepancy could originate from the influence of the P 3 /P 4 apolar D-amino acid present in the substrates used in most previous studies on FXa specificity. Binding of a P 3 /P 4 D-amino acid is highly favorable and clearly improves the catalytic efficiency of FXa (30). It is conceivable that binding of a D-side chain in S 3 /S 4 distorts the neighboring S 2 , hence prohibiting the binding of a bulky P 2 side chain. On the other hand, as opposed to small P 4 -P 1 peptidyl substrates, our fluorescence-quenched substrates bind to subsites on both sides of the scissile bond. Thus, the observed specificity could also result from cooperative effects that would influence the P 2 preferences of FXa. To investigate such possible cooperative effects, we synthesized a series of chromogenic substrates (acetyl-VQXR-pNA, where X was F, T, G, or P). However, analysis of the progress curves of cleavage confirmed that, when constituted exclusively by L-amino acids, a chromogenic substrate having phenylalanine in P 2 is hydrolyzed slightly faster than its counterpart with glycine (4.1 Ϯ 0.2 versus 3.4 Ϯ 0.2 ϫ 10 4 M Ϫ1 s Ϫ1 ). Another possibility is that binding provides the necessary energy to move the aromatic ring system of FXa (Phe 174 , Tyr 99 , and Trp 215 ), perhaps by a simple rotation of Tyr 99 toward S 3 /S 4 , as is observed in kallikrein and factor IXa (2,3,38,39). Such movement would be prohibited when a D-amino acid side chain occupies S 3 /S 4 . In support of this hypothesis, FXa selects sequences with phenylalanine or tyrosine in P 2 (in addition to those having glycine) when offered a library of fusion proteins displayed on phages as substrates (40) or a library of combinatorial fluorogenic substrates (41). In addition, binding of TFPI evidently requires that Tyr 99 of FXa swing away from the "normal" S 2 (42). Finally, in the study of Castillo et al. (43) that also used fluorescence-quenched substrates, Gly as the "best" P 2 amino acid for FXa was only measured in comparison with substrates with Val, Ser, or Thr at this position.
From the P 2 series of fluorescence-quenched substrates, a value of 289 was calculated as the S 2 selectivity index of FXa. This selectivity index simply represented the ratio of the k cat /K m values between the best and the worst side chains in P 2 . Study of the P 1 Ј preferences of FXa revealed a surprisingly low specificity. The selectivity index was only 30 (i.e. 10 times less than for the P 2 residue). In general, FXa preferred small side chains in P 1 Ј (serine, threonine, and alanine), whereas amino acids with bulky or charged side chains were detrimental to catalysis (Table I). Preferences of FXa in P 2 Ј were even less pronounced with a selectivity index of 19. Nevertheless, FXa exhibited a slight preference for apolar or aromatic P 2 Ј residues. Selectivity indexes in S 3 and S 3 Ј were less than 6, meaning that it was difficult to discern preferences, although P 3 Ј hydrophilic side chains possibly favored catalysis. Overall, only the P 2 side chain of the fluorescence-quenched substrates allowed a marked selectivity to FXa.
The Catalytic Groove of FXa Behaves More as a Low Efficient Trypsin Than a Highly Selective Thrombin-In a previous study (29), we reported selectivity indexes of 636 and 20 in S 2 Ј for thrombin and trypsin, respectively. Accordingly, S 2 Ј of FXa would have a selectivity comparable with that of trypsin. To resolve this paradox, we completed the subsite mapping of thrombin and trypsin (Table II). As expected, trypsin was the most efficient and least selective protease. Within the P 2 series for instance, thrombin cleaved its best substrate with a k cat /K m value comparable with that of trypsin, whereas its worst substrate was hydrolyzed with a k cat /K m value 5500-fold lower. Thus, thrombin emerged as a very efficient and selective enzyme. The selectivity index of FXa was 24-fold higher than that of trypsin in P 2 , but for all other subsites, indexes of FXa were either comparable or even lower than for trypsin. Conversely, selectivity indexes of FXa were by far lower than for thrombin, including in P 2 (65-fold less). Fig. 1 represents a general comparison of the selectivity indexes for the three proteases. The most discriminating subsites were S 2 and S 1 Ј for thrombin, with selectivity indexes of 19,000 and 24,706, respectively, far higher than the corresponding indexes of FXa and trypsin. S 2 was the most selective subsite of FXa, whereas for trypsin it was S 1 Ј. Overall, the selectivity indexes suggested that the specificity of the catalytic groove of FXa resembled that of trypsin more than that of thrombin. The highest k cat /K m value obtained with FXa remained 177-fold lower than the k cat /K m value obtained with trypsin for the same substrate (3.0 10 4 versus 5.3 ϫ 10 6 M Ϫ1 s Ϫ1 ). Moreover, the peptide ABz-VQ-FRSLSDQ-EDDnp had the most favorable side chains from P 3 to P 3 Ј for FXa, whereas this sequence was quite distant from the optimum for trypsin (see below). Thus, in contrast to thrombin, FXa emerged not only as enzyme with low selectivity but also as a protease with relatively poor efficiency (30).
Assuming no cooperative effects, the optimal P 3 -P 3 Ј sequence for FXa hydrolysis would be QFR-SLS, that for thrombin would be MPR-SFR, and the theoretical optimal sequence for trypsin would be MRR-RVG. The later sequence is ambiguous, because it contains three potential cleavage sites. To resolve this uncertainty, we isolated and identified the products resulting from trypsin hydrolysis of the equivocal sequences. Trypsin cleavage of ABz-VQFRRLGDQ-EDDnp (Fig. 2) released ABz-VQFR and ABz-VQFRR with similar rate constants (9.5 and 9.2 ϫ 10 6 M Ϫ1 s Ϫ1 ). Cleavage of ABz-VQRRSLGDQ-EDDnp occurred with a slight preference after the second arginine (2.8 versus 2.3 ϫ 10 7 M Ϫ1 s Ϫ1 ). Cleavage of ABz-VQFRKLGDQ-EDDnp released ABz-VQFR with a rate constant twice that of ABz-VQFRK (8.2 versus 4.5 ϫ 10 5 M Ϫ1 s Ϫ1 ), and ABz-VQKRSL-GDQ-EDDnp was cleaved exclusively after the arginine in fifth position. From these data, it can be concluded that (i) trypsin slightly prefers Arg over Phe in P 2 and/or Arg over Leu in P 1 Ј, (ii) trypsin slightly prefers Arg over Gln in P 2 and/or Arg over Ser in P 1 Ј, and (iii) trypsin prefers Arg over Phe in P 2 and/or Lys over Leu in P 1 Ј. On the other hand, the unambiguous sequences (cleaved only once) demonstrate that (i) when lysine is in P 2 , trypsin cleaves exclusively after the arginine in P 1 , (ii) trypsin prefers Gln over Phe in P 2 , and (iii) trypsin prefers Ser over Leu in P 1 Ј. Therefore, the preferred P 2 and P 1 Ј amino acids must both be arginine, and consequently, the hypothetical optimal sequence for trypsin would be MRR-RVG with preferential cleavage after the second arginine. Interestingly, the above preferences of bovine trypsin differ slightly from that of rat trypsin (44 -47).

FIG. 2. Analysis of the products of ABz-VQFRRLGDQ-EDDnp hydrolysis by trypsin.
The P 3 -P 3 Ј residues within this substrate are ambiguous because two potential cleavage sites coexist; in fact, substrate was cleaved at both positions by trypsin. Aliquots of the reaction mixture (100 l in 1% H 3 PO 4 ) were loaded on a C18 reverse phase column developed with a 10 -80% linear gradient acetonitrile in 0.5% H 3 PO 4 . The curves represent the chromatograms of aliquots taken after (from front to back) 0, 1,2,4,6,8,10,15,20,25,30, and 40 min of incubation of 10 M ABz-VQFRRLGDQ-EDDnp with 5 nM trypsin. The peaks were detected by monitoring the A 360 , such that only peptides containing the EDDnp group are visible. The peptides were quantified by peak surface integration analysis and identified by N-terminal sequencing. The insets represent progress curves of the amount of each product and of the remaining intact substrate (expressed in percentages of the maximum). Considering that two cleavages were performed simultaneously by trypsin, the k cat /K m value reported in Table II constitutes a lower limit because the alternate site must have competed for cleavage of the preferred site. FXa and thrombin cleaved all of the fluorescence-quenched substrates used in this study exclusively after the generic arginine. Factor Va Influences Neither the Intrinsic Specificity nor the Efficiency of the Catalytic Groove of FXa-In complex with factor Va within the prothrombinase complex, FXa activates prothrombin very rapidly and selectively. Obviously the cofactor converts FXa from a poor trypsin analogue to a highly specific and efficient prothrombin activator, raising the question as to whether factor Va improves the catalysis and/or the selectivity of the catalytic groove of FXa. We compared the k cat /K m values for the hydrolysis by FXa of representative fluorescence-quenched substrates in the presence and absence of saturating amounts of factor Va. With each peptidyl substrate assayed, the k cat /K m value obtained in presence of factor Va (250 nM), calcium (5 mM), and phospholipid (35 M) was virtually identical to that obtained without cofactors (Fig. 3). Whether or not the peptidyl substrate presented a favorable sequence for FXa catalysis, factor Va had no detectable influence. Thus, factor Va did not improve the catalytic machinery of FXa for fluorescence-quenched substrate cleavage. The fact that factor Va was unable to alter FXa catalysis also implied that factor Va had no detectable influence on the selectivity of the catalytic groove. Consequently, rather than inducing a conformational change in the catalytic groove of FXa, factor Va likely provides a secondary binding site to the substrate (prothrombin) and/or modifies the conformation of prothrombin, allowing a better alignment of the scissile bond. DISCUSSION By using a collection of fluorescence-quenched substrates, our study provides a precise and complete mapping of FXa subsites. The most striking inference is that FXa emerges as a protease with loose selectivity and relatively low catalytic efficiency. The addition of factor Va influenced neither the selec-tivity nor the efficacy of FXa toward small, unconstrained peptides. Therefore, the high selectivity and efficiency of FXa within prothrombinase likely originates from secondary binding site interactions rather than from a remodeling of its catalytic groove by the cofactor.
A number of cleavage reactions have been attributed to FXa (prothrombin, factors V, VII, and VIII, and PAR-2 activation, as well as protein S inactivation), although the physiological relevance of all of these reactions has not been fully demonstrated. Nevertheless, the vast majority of the potential cleavage sequences found in proteins associated with blood coagulation resist FXa hydrolysis. It is therefore surprising that the overall specificity of FXa suggests that it could cleave (at least slowly) most, if not all, of the corresponding peptidyl substrates. Our results allow us to compute a theoretical k cat /K m for cleavage by FXa of the P 3 -P 3 Ј sequences of the main blood clotting reactions (Fig. 4). The most favorable sequence would be the reactive site loop of antithrombin. The calculated k cat /K m value of cleavage is 1.4 ϫ 10 4 M Ϫ1 s Ϫ1 , remarkably close to the association rate constant of the macromolecules. Interestingly, the least favorable reaction would be the release of fibrinopeptide A from fibrinogen. Several of the cleavages that are known to be performed by FXa are among the relatively favorable sequences. In addition to that of antithrombin, sequences that would be cleaved with k cat /K m values higher than 10 3 M Ϫ1 s Ϫ1 include PAR-2, the (nonactivating) cleavage of prothrombin, the activating site of factors VII, and several of the activating cleavages in factor V and VIII. Conversely, besides the release of fibrinopeptide A, sequences that would be cleaved with k cat /K m values less than 10 2 M Ϫ1 s Ϫ1 include a number of reactions that FXa is unable to fulfill: its own activation, that of The values were calculated assuming complete independence of the subsites. The outer circle represents k cat /K m value of 10 8 M Ϫ1 s Ϫ1 , the second circle represents 10 6 M Ϫ1 s Ϫ1 , the third circle represents 10 4 M Ϫ1 s Ϫ1 . All P 3 -P 3 Ј sequences (given in parentheses) are from the human protein.
Selectivity of the catalytic groove of FXa (open squares) appears inadequate to select its natural substrates and inhibitors. On the contrary, the selectivity of thrombin allows rationalization in part for its role in blood clotting; the most favorable sequences correspond to reactions catalyzed, and the least favorable ones correspond to reactions resisting thrombin. The protein C sequence is restrictive to thrombin, but activation requires thrombomodulin as a cofactor; the fibrinogen sequence is also restrictive, but thrombin uses exosite interactions for its cleavage. F, factor; Prot. S, protein S; PAR-1 and PAR-2, protease-activated receptors 1 and 2; ATIII, antithrombin; FVi, factor V inactivating site; tPA, tissue type plasminogen activator; Prot. C, protein C; Fg-A, A-chain of fibrinogen; Fg-B, B-chain of fibrinogen. The asterisks with protein S and TFPI indicate that Cys was taken as Ala to calculate the k cat /K m value. protein C, and the inactivating sites in factor V and VIII. Inconsistent with this picture, however, activation of plasminogen would be favorable, whereas the activation site of prothrombin and the reactive site loop of TFPI would not. In this regard, the requirement for a cofactor is amply documented for prothrombin activation only. Overall, the difference between the calculated k cat /K m values is rather limited (636-fold at the most). Furthermore, the same evaluation with thrombin renders the above reasoning relatively pointless. The most favorable sequence for thrombin within the clotting cascade (one of the activating sites in factor VIII) would be cleaved with a k cat /K m value 10 6 times higher than the least favorable sequence (activation of plasminogen). It is remarkable that reactions catalyzed by thrombin exhibit k cat /K m values higher than 10 4 M Ϫ1 s Ϫ1 , whereas reactions that thrombin is unable to catalyze mostly exhibit lower values. Exceptions include protein C and factor XI activation (both reactions require a cofactor) and fibrinopeptide A cleavage, which involves secondary sites of thrombin remote from the catalytic groove (4,16,17,48,49). TFPI also has a quite favorable sequence, but it is likely that, as for the basic pancreatic trypsin inhibitor (25), a steric hindrance prohibits binding. In agreement with the above rationale, chimeras of antithrombin carrying one of the prothrombin cleavage sites in place of the normal reactive site loop are slightly less effective inhibitors of FXa but are unable to inhibit thrombin (50), and mutating the P 3 , P 2 , P 2 Ј, or P 3 Ј residues in the reactive site loop of antithrombin has little influence on its reactivity toward FXa (51). Clearly, the subsites of thrombin control in part its specificity, whereas FXa cannot rely on its catalytic groove to select substrates.
X-ray data and NMR studies are consistent with the concept that FXa is a "weak trypsin" rather than an enzyme efficient and selective as thrombin. First, the catalytic groove of FXa is not hindered, quite similarly to the widely opened groove of trypsin, whereas that of thrombin is much narrower (2,4,52). Clearly, part of the specificity of thrombin originates from an exclusion by steric hindrance of numerous P 3 -P 3 Ј sequences (14,25,53). Lack of insertions leaves the catalytic triad exposed in FXa and trypsin, certainly explaining in part their low selectivity toward peptidyl substrates. FXa was not only poorly selective, it was also a low efficiency enzyme. Indeed, thrombin exhibited k cat /K m values in the 10 7 M Ϫ1 s Ϫ1 range for its best fluorescence-quenched substrates, which compared favorably with the values obtained with trypsin. In contrast, none of the peptidyl substrates were cleaved by FXa with k cat /K m values higher than 3 ϫ 10 4 M Ϫ1 s Ϫ1 , even when the P 3 -P 3 Ј sequence was nearly optimal. Thus, the selectivity of FXa was barely higher than that of trypsin, but its catalytic efficiency was a thousand times lower on average.
Within prothrombinase, the dramatic increase in the catalytic efficiency of prothrombin activation by FXa results from a 100-fold decrease of K m and a 3000-fold increase of k cat (54). Furthermore, factor Va exclusively enhances prothrombin activation, rendering FXa extremely selective as well as efficient. Cofactors may act by several mechanisms that are not mutually exclusive. The simplest mechanism would be a conformational change in the active site of the enzyme induced by cofactor binding. Structural alterations would improve catalysis by rearranging the geometry of the charge stabilizing system and/or that of the S 1 pocket. In this model, all reaction rates improve upon binding of the cofactor, simply because of the more efficient catalytic machinery. Such a mechanism does not apply to the FXa/factor Va interaction; cleavage rates of the fluorescence-quenched substrates were virtually identical whether or not factor Va was bound to FXa. Alternatively, still in a purely allosteric mechanism, the cofactors could ensure remodeling of one or several subsites, allowing a better binding and/or alignment of the scissile bond. Such a mechanism also does not apply to the FXa/factor Va interaction, at least with unconstrained peptidyl substrates; irrespective of the sequences of the fluorescence-quenched substrates, cleavage rates with or without factor Va were virtually identical. The same is true with peptidyl chloromethyl ketone inhibitors as reported by Walker and Krishnaswamy (54). Whether the probably constrained activation site in prothrombin requires a particular geometry to be cleaved by FXa remains to be explored. Besides possible remodeling of the catalytic groove, cofactors could induce formation of a secondary binding site for the substrate (55)(56)(57). The subsequent site could be either on the enzyme (resulting from a remodeling induced by the cofactor) and/or carried by the cofactor itself (20,21). In either case influence of cofactor binding on the hydrolysis of unconstrained peptidyl substrates would not be detectable. Providing a supplementary site for the substrate suggests at first that the K m value rather than the k cat value of the reaction would improve. However, it is conceivable that anchoring also drives the scissile bond of the substrate into a conformation allowing a better alignment with respect to the catalytic groove, thus improving both the K m and k cat values of the reaction. Our results establish that factor Va has no detectable influence on the catalytic efficiency and selectivity of the active site cleft of FXa toward small and unconstrained peptides. The inference is that factor Va improves the binding of prothrombin and/or allows a better alignment of the scissile bonds through the creation of secondary binding sites rather than by a purely allosteric mechanism exclusively involving the catalytic groove of FXa.