The elusive role of the potential factor X cation-binding exosite-1 in substrate and inhibitor interactions.

A number of studies suggest that blood-clotting factor X (FX) uses secondary site(s) to interact (as a substrate) with its activators. Numerous pieces of evidence also imply that, within prothrombinase (as an enzyme), activated FX (FXa) uses exosite(s) for cofactor Va and/or prothrombin recognition. Similarly, FXa exosite(s) seem to govern interaction with inhibitors. An obvious difference between FXa and thrombin resides within a region called exosite-1: positively charged in thrombin and clearly of opposite polarity in FXa. To investigate the role of this potential cation-binding exosite, we prepared a series of mutants within loops 34-40 and 70-80 of FX. Overall, the mutations induced relatively subtle, non-synergistic modulation. The potential exosite was dispensable for FX activation and is unlikely to constitute a critical region for factor Va binding, albeit it is clearly important for prothrombin activation. Our data also implicate loop 34-40 of FXa in the interaction with the tissue factor pathway inhibitor, in prevention of plasminogen activator inhibitor-1 binding, and in tempering inhibition by heparin-activated antithrombin. Compared with FX, mutants with reduced electrostatic potential potentiated thrombin production in FX-depleted plasma, whereas mutants with inverted electrostatic potential impeded clotting. Despite the definite consequences observed, disruption of the potential cation-binding exosite of FX had rather weak effects, far from what would be expected if this region was as crucial as in thrombin.

Coagulation factor X (FX) 1 is a vitamin K-dependent zymogen activated into FXa by cleavage at a single bond (Arg 15 -Ile 16 in the chymotrypsin numbering system 2 ) hydrolyzed by either of two complexes: FVIIa bound to tissue factor (TF), which triggers coagulation after injury, or FIXa associated with FVIIIa, which amplifies thrombin production after initiation of coagulation. FXa is the only endogenous prothrombin activator, but substantial thrombin production occurs only within the prothrombinase complex, where FXa binds to FVa in the presence of calcium on an appropriate phospholipid surface. Two inhibitors control FXa activity, viz. the tissue factor pathway inhibitor (TFPI) and antithrombin.
Extended binding sites (exosites) play a crucial role in virtually any substrate, cofactor, or inhibitor recognition within the coagulation cascade. Perhaps best characterized is the case of thrombin taking advantage of two exosites for extended interactions in numerous functions (1). Exosite-1 is formed by a set of basic residues comprising loops 34 -40 and 70 -80, which neighbor each other in the folded structure (2); it interacts with fibrinogen, thrombomodulin, platelet activator receptor-1, FV, FVa, heparin cofactor II, and hirudin. Exosite-2, comprising loop 91-102 and helices 165-173 and 233-245 (3), is also formed by a set of basic residues that interact with heparin, with the chondroitin sulfate of thrombomodulin, and within prothrombin with kringle-2 (4). In FVII, the region corresponding to exosite-1 of thrombin is critical for FX activation (5,6); in FIXa, this region is also important for FX activation in the absence of FVIII and for its inhibition by antithrombin in the absence of heparin (7). In contrast to thrombin, FVIIa, FIXa, and FXa share a negatively charged patch within segment 70 -80, whereas FXa is unique in that both loops 34 -40 and 70 -80 are negatively charged. Thus, the region of FXa corresponding to exosite-1 of thrombin forms a negative cluster, raising the question as to whether it could constitute a functional exosite. FXa has a unique macromolecule substrate, but exosites may also participate in its inhibition and/or in the activation of its zymogen. In fact, a number of studies have established that exosite(s) must be critical in several FXa functions. Electrostatic interactions involving loop 34 -40 of FXa appear to play a significant role in binding to the second Kunitz domain of TFPI (8,9). FXa also appears to interact with a complementary surface on pentasaccharide-activated antithrombin (10), and the region of FXa topologically equivalent to exosite-2 of thrombin seems to be involved in the binding of heparin and FVa (11). Above all, prothrombin recognition by prothrombinase undoubtedly involves one or more exosites on FXa and/or FVa (12)(13)(14).
In a previous study (15), we reported that the catalytic groove of FXa is minimally selective with unconstrained peptides and that FVa has little effect, if any, on this selectivity. These two observations therefore also favor the hypothesis that FXa uses secondary binding sites to improve binding and/or to ensure proper alignment of the substrate scissile bond. In the present study, we have examined the role of the potential cation-binding exosite-1 of FX by expressing mutants of FX in which part of the electrostatic potential of FX was neutralized or reversed. We obtained mixed results in that essential functions of FX and FXa were unaffected by mutations within segment 70 -80, whereas prothrombin activation and TFPI inhibition were impaired by the loop 34 -40 mutations. Thus, despite a high conservation between species, the potential exosite-1 of FX appears to be less important than its thrombin counterpart.
Preparation of Human Recombinant FX and Derivatives-The shuttle vector pNUT-FX (15) was used as a template to prepare the vectors expressing FX and derivatives. Prior to mutagenesis within the potential exosite-1 of FX, two modifications were introduced in pNUT-FX. First, to facilitate purification of the expressed material, a sequence coding for the 9E10 epitope at the C-terminal end of the FX coding region was replaced with the sequence DQVDPRLIDGK, recognized by monoclonal antibody HPC-4 (Roche Applied Science, Meylan, France). Second, to improve ␥-carboxylation, the sequence coding for the propeptide of FX was replaced in a three-step mutagenesis procedure with the homologous sequence in human prothrombin (20,21). Following these modifications, the vector was used as a template to prepare vectors for variants E36Q/E37Q/E39Q, E36K/E37K/E39K, E74Q/E76Q/E77Q, and E74K/E76K/E77K of FX (Table I). Constructs were obtained through a single PCR derived from the circle mutagenesis method (22). Briefly, the reaction mixture (50 l) contained 2.5 units of Pfu DNA polymerase (Stratagene, Amsterdam Zuidoost, the Netherlands) in the buffer recommended by the manufacturer, 0.5 mM each dNTP, 125 ng of each sense/antisense primer pair (Table II), and 50 ng of template. Reactions comprised an initial denaturation at 95°C, followed by 16 cycles of 45 s at 95°C, 60 s at 55°C, and 26 min at 68°C. Matrix was degraded for 1 h at 37°C with 10 units of DpnI (Ozyme, St. Quentin en Yveline, France), and the mixture was used to transform DH5␣ competent cells by standard techniques. The sequences of FX derivatives were verified by dideoxy chain termination sequencing. Transfection, selection, and culture of BHK-21 cells were carried out as described previously (15,17,23). Production and purification of recombinant FX and derivatives in milligram quantities were performed by a combination of anion-exchange and affinity (using antibody HPC-4 coupled to agarose) chromatography as described for prothrombin derivatives (17,23). Final products appeared to be homogeneous by Coomassie Blue staining following SDS-PAGE. One hour prior to use as a zymogen, FX and derivatives were incubated with 1 mM phenylmethanesulfonyl fluoride to neutralize any trace of FXa. Control experiments indicated that, after 30 min in Tris-HCl buffer, phenylmethanesulfonyl fluoride was fully hydrolyzed and would not interfere with other reactions.
FX Activation Kinetics-Activation of FX and derivatives was studied at 25°C under pseudo first-order conditions in kinetic buffer (50 mM Tris-HCl, pH 7.4, containing 0.15 M NaCl, 5 mM CaCl 2 , 2 mg/ml bovine serum albumin, 0.2% (w/v) polyethylene glycol (M r 8000), and, when required, 16 M phospholipid vesicles). To ascertain that pseudo firstorder conditions were met, two zymogen concentrations were assayed, with all other experimental conditions the same. With RVV-X, the reaction was initiated by the addition of 50 nM activator to 0.5 M FX or variant (alternatively, 0.25 M). With FVIIa (100 pM) plus TF (1.3 nM) or with FIXa (50 pM) plus FVIIIa (8.6 nM), reactions were initiated by the addition of the mixture to 10 nM FX or variant (alternatively, 5 nM). At timed intervals, an aliquot was withdrawn, and the reaction was quenched by the addition of 50 mM EDTA. The progress of FXa (or activated variant) formation was monitored by the rate of S2222 hydrolysis. Pseudo first-order rate constants (k) were estimated by nonlinear regression analysis of the amount of activated FX (FXa t ) upon time (t) using Equation 1, in which FXa 0 represents the initial FXa concentration (normally zero), and FXa max is the amount of FXa after full activation. The k cat /K m values of FX activation were deduced from the ratio of k to the activator concentration (RVV-X, FVIIa, or FIXa). Large quantities of FXa and derivatives were prepared by passage through a 5-ml HiTrap N-hydroxysuccinimide-activated column (Amersham Biosciences, Orsay, France) coupled to RVV-X. Following activation, the active site concentration was determined by titration with H-D-Phe-Phe-Arg-CH 2 Cl as described previously (15). Before each kinetic determination, the concentration of FXa and derivatives was systematically controlled by measuring the rate of 100 M S2222 hydrolysis.
Prothrombin Activation Kinetics-Prothrombin activation by FXa or derivatives was examined by progress curve kinetics of a coupled reaction in which FXa or derivatives (with or without FVa) were added to a mixture of prothrombin (12 nM to 1.5 M) and S2238 (100 M). Reactions were performed at 25°C in kinetic buffer containing 16 M phospholipid vesicles and were started by the addition of 0.2 nM FXa alone or 0.1 nM FXa plus catalytic amounts of FVa. The rationale behind this approach is that because 100 M S2238 is amply saturating, the rate of the second reaction (pNA release from S2238) is a direct function of thrombin concentration. On the other hand, as long as the initial prothrombin concentration is Ͼ0.1-fold its K m for the activator, activation is steadystate, and thrombin increases linearly with time. Thus, if the thrombin concentration steadily increases, the rate of pNA release steadily increases (i.e. accelerates). As long as Ͻ10% of both substrates are consumed, acceleration remains constant and is proportional to the rate of the first reaction. Practically, the increase in pNA with time (t) was monitored, and acceleration coefficients (␣) were estimated by fitting data corresponding to Ͻ10% prothrombin and S2238 consumption to Equation 2 (24,25), in which pNA 0 represents the initial concentration of pNA (normally zero), ␤ is the rate of pNA release independent of thrombin (also normally zero), and ␣ is a function of the K m and k cat of S2238 for thrombin and of prothrombin for its activator. For any given concentration of S2238 and prothrombin, ␣ is directly proportional to the activator concentration (FXa alone or in complex with FVa). Under the conditions used, the rate of prothrombin formation by FXa alone represented Ͻ1% of that catalyzed by FVa⅐FXa and was neglected in the presence of FVa. The FVa⅐FXa concentration was estimated using Equation 3 (19), PAI-1 were estimated as described previously (16,18,26) by titration against thrombin and the E36K/E37K/E39K variant of FXa, respectively. Care was taken to verify that full inhibition was achieved by comparing residual activities after 3 and 6 h of incubation. Inhibition of FXa and derivatives by antithrombin (without polysaccharide) or by PAI-1 was examined under pseudo first-order conditions and monitored by progress curve kinetics. Estimates of the overall association rate constants (k a ) were obtained essentially as described previously (18,26). Antithrombin (1 M) and PAI-1 (10 -20 nM) concentrations were such that significant inhibition occurred over the time period of the experiment (Ͻ1 h). For inhibition by antithrombin, the reaction was performed in the presence of 5 g/ml Polybrene (Sigma) to neutralize any trace of heparin. Reactions were performed at 25°C in kinetic buffer and initiated by the addition of 250 -500 pM FXa (or derivative) such that Ͻ10% of the substrate (S2222, 100 M) was hydrolyzed over the time period of the experiment. The observed rate constants of inhibition (k obs ) were estimated by fitting the data to Equation 4 for slow binding inhibition (27), where A 405(0) and A 405(t) represent A 405 initially (before the addition of FXa) and at time t, respectively, and v i and v s are the initial and final (normally zero with serpins) rates of S2222 hydrolysis, respectively. The k a was deduced from k obs using Equation 5 to account for the competition introduced by the substrate, where S2222 0 is the initial S2222 concentration, K m is its Michaelis constant for FXa (or derivatives), and I 0 is the active concentration of added serpin. The k a of TFPI for FXa (or derivatives) was also estimated by progress curve kinetics, except that kinetic buffer contained 5 mM EDTA instead of CaCl 2 . The reaction was initiated by the addition of 200 pM FXa or derivatives to a mixture of 2-150 nM TFPI and 100 M S2222. The k a values were estimated using Equations 4 and 5 in which the initial concentration of TFPI was substituted for I 0 .
The polysaccharide-catalyzed association rate constant (k a(cat) ) of activated antithrombin for FXa (or derivatives) was also estimated by analysis of the data from progress curve kinetics completed under pseudo first-order conditions. Reactions were started by the addition of 250 pM FXa to a mixture of antithrombin (2.5-10 nM), 100 M S2765 (used instead of S2222 to slow down the reaction by competition), and a limiting amount of heparin (5-25 nM) or a saturating amount of pentasaccharide (150 nM). With a saturating amount of pentasaccharide, k a(cat) was estimated using Equations 4 and 5 in which k a(cat) was substituted for k a , and the total (saturated) antithrombin concentration was substituted for I 0 . In the presence of saturating heparin, the reaction was too fast to allow accurate analysis even with recordings as frequent as every second. Thus, inhibition reactions were performed in the presence of a catalytic amount of heparin, and k a(cat) values were estimated using Equation 6 (28), where AT free and AT⅐PS represent the concentrations of free and polysaccharide-bound antithrombin, respectively, and k a(uncat) is the k a in the absence of polysaccharide estimated as described above. The antithrombin⅐polysaccharide concentration was estimated using Equation 3 in which antithrombin⅐polysaccharide was substituted for FVa⅐FXa, and the concentration of heparin was substituted for FVa 0 and that of antithrombin for FXa 0 . The K D of the anti-thrombin⅐polysaccharide complex was determined by intrinsic fluorescence measurement as described by Olson et al. (28).

Plasma (ex Vivo) Studies of FXa and Derivatives-
The half-lives of FXa and derivatives in plasma were estimated by the addition of 20% (v/v) kinetic buffer containing 40 mM CaCl 2 , 400 units/ml hirudin, and 50 nM FXa (or derivative) to citrated human plasma. At timed intervals, aliquots were withdrawn, and residual FXa activity was estimated from the rate of hydrolysis of 100 M S2765. Rates of inhibition (k) were estimated by nonlinear regression analysis of residual FXa versus time (t) using Equation 7, where FXa max represents the amount of FXa initially added, and FXa 0 is the final residual FXa activity (to account for the final rate of S2765 hydrolysis, either independent of FXa or due to FXa protection from inhibition). The half-life of FXa was calculated by dividing ln(2) by the observed rate of inhibition. The procoagulant potential of FX (and derivatives) was evaluated by its ability to induce thrombin formation in FX-depleted plasma. Clotting was triggered at 37°C by the addition of 50% (v/v) kinetic buffer containing 40 mM CaCl 2 , 160 M phospholipid vesicles, 0.5 nM TF, and 160 nM FX (or derivatives) to plasma containing 16 mM Gly-Pro-Arg-Pro-amide to prevent fibrin polymerization (29,30). Particular care was taken to verify that FXa or activated derivatives were undetectable prior to the addition of their zymogens to plasma. At timed intervals, aliquots were withdrawn, and reactions were quenched by diluting samples (1:10) in kinetic buffer without calcium and with 10 mM EDTA and 100 mM benzamidine. The amount of thrombin was monitored from the rate of 100 M S2238 hydrolysis after further dilution (1:20) in kinetic buffer.

Potential Exosite-1 of FX and Zymogen Activation-Human
is critical for most functions, we wondered if the negative cluster of FX constitutes a secondary binding site (Fig. 1). To investigate this, we prepared a series of mutants in which sets of three glutamates were replaced with glutamine or lysine, viz. E36Q/E37Q/E39Q, E36K/E37K/E39K, E74Q/E76Q/E77Q, and E74K/E76K/E77K (Table I) To be functional, an exosite evidently requires a complementary motif to exist on the macromolecular substrate, inhibitor, or cofactor (32,33). Because FX is initially a zymogen, it constitutes the substrate during its activation to FXa and thus may display complementary motif(s) to activator(s). In our search for a role for the potential exosite-1 of FX, we first determined the susceptibility of FX mutants to activation by RVV-X, TF⅐FVIIa, and FVIIIa⅐FIXa (Fig. 2). FX and derivatives were activated by RVV-X with comparable k cat /K m values (Table III), thus excluding the potential exosite-1 of FX as a player in this reaction. Undoubtedly, secondary binding site(s) are implicated in FX activation by TF⅐FVIIa because uncleavable FX (with glutamine substituted for the arginine of the scissile bond) has unchanged affinity for TF⅐FVIIa (34). The impact of the FX mutations on activation by TF⅐FVIIa was, however,   Clearly, our results suggest that the potential exosite-1 of FX is nonessential for its activation by RVV-X, TF⅐FVIIa, or FVIIIa⅐FIXa. The contribution of loops 34 -40 and 70 -80 of FIX to its activation by TF⅐FVIIa has not been studied in detail, but its calcium-binding site (within segment 70 -80) is not involved in the reaction (35). In contrast, positive charges within proexosite-1 of prothrombin are essential for its activation by prothrombinase (36,37), whereas E39K prothrombin is normally activated (17). In addition, proexosite-1-specific ligands such as the sulfated C-terminal tail of hirudin block prothrombin activation by prothrombinase (13,38).
Role of Loop 34 -40 of FXa in Prothrombin Activation-We next considered whether the potential exosite-1 of FX contributes to enzyme catalysis. FX variants were activated in large quantities by RVV-X; their active site content was assessed by H-D-Phe-Phe-Arg-CH 2 Cl titration; and their amidolytic activity was characterized. The k cat and K m values of S2222 and S2765 were comparable for FXa, E36Q/E37Q/E39Q, and E36K/E37K/ E39K, but were slightly altered (Ͻ2.5-fold) for E74Q/E76Q/ E77Q and E74K/E76K/E77K (Table IV). Because the pNA substrates do not have a C-terminal extension that could directly interact with the potential exosite-1 of FX, this was perhaps not surprising. The small changes observed could reflect an allosteric coupling with the active site of FXa, could result from an alteration of the electrostatic potential of FXa, and/or could originate from a newly created, nonproductive, low affinity binding site for pNA substrates such as the one described in thrombin (39).
A number of studies have established that the specific cleavage of the two peptide bonds in prothrombin by prothrombinase is governed by interactions remote from the catalytic groove (13,40). To explore if the potential cation-binding exosite-1 of FXa is important in prothrombin activation, we recorded progress curves of pNA release and evaluated the acceleration of S2238 hydrolysis due to increasing thrombin concentrations (Fig. 3). Initial velocities of thrombin production at various

TABLE V Kinetics of prothrombin activation by FXa or derivatives
The kinetic constants (k cat and K m ) of prothrombin activation were estimated by nonlinear regression analysis of the dependence of the acceleration of S2238 hydrolysis by increasing thrombin concentrations (estimated as described in the legend to Fig. 3 a Prothrombin activation by prothrombinase was performed with a subsaturating concentration of FVa (250 pM); prothrombinase concentration was calculated according to the estimated K D of each FXa variant for FVa (Fig. 4).

TABLE VI Kinetics of fluorescence-quenched substrate cleavage by FXa and derivatives
The kinetic constants (k cat /K m ) for substrate hydrolysis were estimated by nonlinear regression analysis of the increase in fluorescence upon time using a pseudo first-order equation as described previously (15) and are given together with their S.E. The amino acid specific to each substrate is shown in boldface. ABz, ortho-aminobenzoyl; EDDnp, N- (2,4-dinitrophenyl)ethylenediamine.  (Table V).
Overall, the K m of prothrombin was little affected (changed Ͻ2-fold) whether or not FVa was present, ruling out the possibility that FXa utilized its potential exosite-1 to capture prothrombin. The loop 70 -80 mutations possibly increased the k cat of the reaction, whereas rates were severely impaired by the loop 34 -40 substitutions: with or without FVa, the k cat values decreased by ϳ10-fold for E36Q/E37Q/E39Q and E36K/ E37K/E39K (Table V). A profound alteration of the FXa catalytic groove is unlikely to be the cause of the generalized defect in prothrombin activation because the amidolytic activities of the loop 34 -40 mutants were normal. To explore the possible influence of the mutations on the prime side of the scissile bond, we measured the hydrolysis rate of fluorescencequenched substrates with representative P 1 Ј, P 2 Ј, and P 3 Ј residues (15). The catalysis of these substrates was very little influenced by the mutations: compared with FXa, the rate of cleavage of the fluorescence-quenched peptides by the variants was increased or decreased by 2-fold at most (Table VI). In the presence of FVa, the defect in prothrombin activation by the loop 34 -40 mutations could originate from a decreased affinity for the cofactor. We estimated the K D values of the FXa variants for FVa by measuring the rate of thrombin production with increasing FVa concentrations, but the mutations had only limited impact: values of 0.3 Ϯ 0.1, 0.6 Ϯ 0.2, and 0.7 Ϯ 0.4 nM were obtained for FXa, E36Q/E37Q/E39Q, and E36K/E37K/ E39K, respectively (Fig. 4).
Overall, the defect in prothrombin activation of the loop 34 -40 mutants resulted in deficient hydrolysis of prothrombin that was essentially independent of FVa and prothrombin binding and did not involve a major remodeling of the catalytic groove of FXa. Thus, the defect would result from a conformation that prevents the scissile bond from obtaining the optimal geometry for turnover. Consistent with this picture, mutations D70K and E80K in FXa moderately affect prothrombin activation (41); peptide 58 -97 of FXa (which binds CaCl 2 with high affinity) does not compete with FXa for FVa binding (31); and a monoclonal antibody directed against nearby residues 82-91 and 102-116 blocks prothrombin activation without affecting prothrombinase assembly (14). In addition, FVa itself does not seem to induce major allosteric remodeling of the catalytic groove of FXa (15,42). The core FVa-binding region of FXa is localized within the positively charged region of FXa, comprising Arg 125 , Arg 165 , Lys 169 , and Lys 230 within the region topologically equivalent to exosite-2 of thrombin, diametrically op-posed to loop 34 -40 (11,(43)(44)(45)(46). In this model, the potential cation-binding exosite-1 of FXa would remain, within prothrombinase, exposed to the solvent and therefore accessible, conceivably to interact with the anion-binding proexosite-1 of prothrombin.
Potential Exosite-1 of FXa and Interaction with Unactivated Antithrombin-Despite being its main inhibitor in plasma, the reactivity of antithrombin for FXa depends only weakly on the sequence of its reactive site loop, suggesting that one or more exosite(s) govern the interaction (10). Without polysaccharide, FXa and E36Q/E37Q/E39Q were inhibited by antithrombin with comparable k a values; E36K/E37K/E39K, E74Q/E76Q/ E77Q, and E74K/E76K/E77K had k a values 2-3-fold lower (Table VII). Thus, the potential exosite-1 of FXa appears to be only moderately involved in the interaction with antithrombin alone. The same is true for thrombin: mutations in segment 70 -80 do not impede inhibition by antithrombin (47), and the E39K mutant is inhibited with a k a value comparable with that with thrombin (17). A direct participation of segment 70 -80 is also unlikely because it is positively charged in the anticoagulant activated protein C (as in thrombin), which is resistant to antithrombin inhibition (contrary to thrombin). Conversely, direct participation of the charged loop 34 -40 is unlikely because it is either neutral or negatively charged in thrombin, FXa, and FIXa (three targets of antithrombin). As a consequence, our observation renders it improbable that the positively charged loop 34 -40 of activated protein C is one of the locks preventing its inhibition by antithrombin.
Role of Loop 34 -40 of FXa in the Interaction with Heparinactivated Antithrombin-Heparin enhances the rate of FXa inhibition by antithrombin by ϳ1000-fold (28). Structural data suggest that residues of the reactive site loop of antithrombin have a non-optimal conformation for interacting with FXa, but that binding of heparin to antithrombin alters this conformation and/or frees the loop, allowing it to fit into the catalytic groove of the target (48). In addition, heparin chains that are long enough enhance interaction with thrombin and FXa through a template mechanism (49). Binding of a specific pentasaccharide to antithrombin effectively accelerates FXa inhibition, but is ineffective in enhancing thrombin inhibition. The pentasaccharide therefore allows a distinction between effects due to antithrombin remodeling compared with heparin bridging. Kinetic data also suggest that a region outside the reactive site loop (i.e. an exosite) may primarily cause the enhanced potency of pentasaccharide-activated antithrombin for FXa (50). That the pentasaccharide enhancement is detectable with FXa and not with thrombin implies that a complementary exosite exists on FXa that is absent on thrombin. Specifically, a Inhibition by heparin-activated antithrombin was analyzed using Equation 6 to estimate k a(cat) ; free and bound antithrombin concentrations were estimated using Equation 3.
b Inhibition by pentasaccharide-activated antithrombin (with a saturating amount of pentasaccharide) was estimated using Equations 4 and 5 in which k a(cat) was substituted for k a , and v s was set to zero.
c Inhibition by TFPI was analyzed using Equations 4 and 5.
a striking difference between thrombin and FXa is exosite-1: positively charged in thrombin and negatively charged in FXa. Compared with FXa, the k a(cat) for inhibition of the FXa derivatives by pentasaccharide-activated antithrombin decreased by Ͻ3-fold (Table VII). Thus, as with unactivated antithrombin, the potential exosite-1 of FXa appears to be only moderately involved, if at all, in direct or indirect interaction with the serpin. In the presence of full-length heparin and CaCl 2 , the loop 70 -80 mutants and E36Q/E37Q/E39Q were also inhibited by antithrombin, with k a(cat) values slightly lower than for FXa, but E36K/E37K/E39K was neutralized more rapidly: the k a(cat) was nearly 5-fold higher (Table VII). Enhanced interaction could originate from more favorable orientation of the approaching molecules caused by alteration of the electrostatic field. Interestingly, E36K/E37K/E39K eluted from a Resource-Q column at higher ionic strength compared with FX or derivatives. Alternatively, enhanced interaction could result from a higher affinity for heparin. Consistent with this hypothesis, E36K/E37K/E39K eluted at higher ionic strength compared with FX or variants from a heparin-5PW TSK column (Fig. 5). The heparin-binding site of FXa (Arg 93 , Lys 96 , Arg 125 , Arg 165 , Lys 169 , and Lys 236 ) is at the same topological location as in FIXa and thrombin (11,51), on the opposite side of the potential exosite-1 (i.e. far from loops 34 -40 and 70 -80). Nevertheless, higher affinity for heparin could result from the auxiliary contribution of loop 34 -40 to binding and/or through a linkage with the heparin-binding site.
Overall, the striking difference between exosite-1 of FXa and thrombin appears to be unrelated to the mechanisms of inhibition by antithrombin: the potential exosite-1 of FXa is somewhat dispensable. Consistent with this conclusion, mutations D70K and E80K in FXa have little consequence on the potency of heparin-activated antithrombin (41), whereas E37Q or E39Q affects it only moderately (52). In FIXa, loop 34 -40 seems to be more important: a FIXa variant carrying loop 34 -40 of FX is virtually insensitive to antithrombin inhibition, although heparin restores, in part, the ability of the serpin to neutralize the chimera (7).
Role of Loop 34 -40 of FXa in TFPI Binding and in Preventing Inhibition by PAI-1-TFPI is a more effective inhibitor of FXa compared with unactivated antithrombin, but a low FIG. 5. Affinity for heparin of FX and derivatives. FX or derivatives were loaded onto a 7.5 ϫ 75-mm heparin-5PW TSK column equilibrated in 50 mM Tris-HCl, pH 8.0, containing 50 mM NaCl and 5 mM calcium, and the column was developed by increasing the salt concentration (dashed line, right scale). The graph represents the elution profiles of wild-type FX (q), E36Q/E37Q/E39Q (E), and E36K/ E37K/E39K (OE). FX variants E74Q/E76Q/E77Q and E74K/E76K/E77K eluted essentially as did FX. Compared with their zymogens, all activated FX and derivatives eluted from the heparin-5PW TSK column at ϳ50 and 80 mM higher NaCl concentrations in two overlapping peaks that probably corresponded to the ␣ and ␤ forms, respectively. to yield an observed rate constant, allowing calculation of the half-life. FXa activity plateaued to ϳ25% of the initial activity, and this inhibition-resistant activity remained stable for at least 1 h (without added FXa, S2222 hydrolysis was undetectable).
The serpin PAI-1 is unable to inhibit FXa, its natural target being tissue-type plasminogen activator (t-PA). In contrast to FXa, loop 34 -40 of t-PA comprises three basic residues, viz. Arg 37A , Arg 37B , and Arg 39 , which interact with a negatively charged region in the distal part of the reactive site loop of PAI-1 (54,55). It was therefore of interest to examine whether PAI-1 could interact with E36K/E37K/E39K (Fig. 6). Indeed, PAI-1 neutralized E36K/E37K/E39K with a k a in excess of 10 5 M Ϫ1 s Ϫ1 (i.e. a value ϳ25-fold lower than that of t-PA inhibition by PAI-1). Under the same conditions, E36Q/E37Q/ E39Q, E74Q/E76Q/E77Q, E74K/E76K/E77K, and FXa were all resistant to PAI-1. Thus, simple exchange of the negative residues within loop 34 -40 of FXa with lysine is sufficient to transform FXa into a PAI-1-sensitive enzyme. Consistent with this observation, replacement of loop 34 -40 of thrombin with that of t-PA allows PAI-1 to neutralize thrombin (56,57), and substitution of loop 34 -40 of t-PA with that of a plasminogen activator isolated from a snake venom also allows its neutralization by PAI-1 (18).
Potential Exosite-1 of FX and Thrombin Formation in Plasma-Results with purified material suggested that mutations within loop 34 -40 of FX decreased prothrombinase efficiency by ϳ10-fold, but made FXa more resistant to TFPI and antithrombin, albeit PAI-1 neutralized it. Thus, mutations that promoted clotting on the one hand prevented it on the other. The combination of these opposing effects made it difficult to predict the overall effect on clotting: would clot formation be delayed due to lower prothrombinase efficacy, or would it be earlier due to lower sensitivity to plasma inhibitors? To investigate the overall effect of the FXa mutations, we compared the half-lives of FXa and its derivatives in plasma (Fig. 7). The half-lives of E74Q/E76Q/E77Q and E74K/E76K/E77K were slightly longer than normal, in accord with their lower sensitivity to antithrombin (77 Ϯ 12 and 112 Ϯ 10 s versus 59 Ϯ 11 s), but the half-life of E36K/E37K/E39K (52 Ϯ 9 s) was comparable with that of FXa, ruling out the possibility that it could be a longer lasting enzyme in plasma. We also compared the ability of the FX derivatives to promote thrombin formation in FX-depleted plasma (Fig. 8). The reaction was started by the addition of a mixture of TF, phospholipid, calcium, and FX (or variant) to plasma containing Gly-Pro-Arg-Pro-amide to prevent fibrin polymerization (29,30). Compared with FX supplementation, the addition of E36Q/E37Q/E39Q or E74Q/E76Q/ E77Q allowed quicker formation of more thrombin, whereas the addition of E36K/E37K/E39K or E74K/E76K/E77K triggered later formation of less thrombin.
In summary, partial neutralization of the negative charges within the potential exosite-1 of FX enhanced thrombin formation, whereas partial inversion decreased its procoagulant potential. Overall, the potential exosite-1 of FX played only a limited role in the balance of coagulation.
General Conclusion-Mutations within loops 34 -40 and 70 -80 of FX induced subtle modulations of FXa activity, with possible opposing effects. Our results suggest that the potential exosite-1 of FX is unlikely to constitute a critical exosite for FX activation. Nevertheless, our study implicates loop 34 -40 of FXa in prothrombin activation, albeit excluding it from FVa binding; loop 34 -40 also seems to be important in the interaction with TFPI, in preventing PAI-1 binding, and possibly in tempering inhibition by antithrombin. Finally, the results suggest that the electrostatic potential of FX may influence its procoagulant properties. However, mutations within the potential exosite-1 of FX do not result in the major outcome that would be expected for a crucial exosite such as exosite-1 of thrombin, in which a single amino acid change can result in almost complete loss of function (2).