Nematode Anticoagulant Protein c2 Reveals a Site on Factor Xa That Is Important for Macromolecular Substrate Binding to Human Prothrombinase*

The binding of recombinant nematode anticoagulant proteinc2(NAPc2)toeitherfactorXorXaisarequisitestepinthepathwayforthepotentinhibitionofVIIatissuefac-tor.WehaveusedNAPc2asatightbindingprobeofhumanXatoinvestigateproteinsubstraterecognitionbythehu-manprothrombinasecomplex.NAPc2bindswithhighaf-finity( K d (cid:1) 1 n M ) to both X and Xa in a way that does not require or occlude the active site of the enzyme. In contrast, NAPc2 is a tight binding, competitive inhibitor of protein substrate cleavage by human Xa incorporated into prothrombinase with saturating concentrations of membranes and Va. By fluorescence binding studies we show that NAPc2 does not interfere with the assembly of human prothrombinase. These are properties expected of an inhibitor that blocks protein substrate recognition by targeting extended macromolecular recognition sites (exosites) on the enzyme complex. A weaker interaction ( K d (cid:1) 260–500 n M ) observed between NAPc2 and bovine X was restored to a high affinity one in a recombinant chimeric bovine X derivative containing 25 residues from the COOH terminus of the proteinase domain of human X. This region 20%) 1-cm Poros HQ/M equilibrated same eluted ml/min) of

The proteolytic activation of prothrombin is catalyzed by the prothrombinase complex of coagulation (2)(3)(4)(5). Prothrombinase assembles through membrane-dependent interactions between the serine proteinase, factor Xa, and the protein cofactor, factor Va (2,3). Although solution-phase Xa is a competent enzyme, its incorporation into prothrombinase yields a profound increase in the rate of thrombin formation (2,3).
Prothrombin is the only known protein substrate cleaved efficiently by prothrombinase (2,6). Such stringent selectivity is not evident in the action of factor Xa on oligopeptidyl substrates, nor is the rate of peptidyl substrate hydrolysis significantly enhanced upon assembly of factor Xa into prothrombinase (7,8). Thus, the narrow and defined specificity of prothrombinase toward its protein substrate is unlikely to be solely explained by the specific recognition of residues surrounding the scissile bond by the active site of factor Xa within the enzyme complex.
Mechanistic studies of bovine prothrombinase function have borne out this suggestion (9 -11). A series of studies indicate that recognition of the biological substrate is achieved through stepwise interactions of the protein substrate at an extended macromolecular recognition site (exosite) in prothrombinase removed from the catalytic site of Xa, followed by docking of elements surrounding the scissile bond with the active site of the enzyme and subsequent cleavage (9 -11). The initial exosite-binding step dominates the affinity and binding specificity of the enzyme for the protein substrate. This interaction serves to tether the substrate to the enzyme, thereby directing ordered active-site docking and cleavage at two spatially distinct sites in the zymogen (11). Thus, surfaces on the enzyme complex, removed from the active site of factor Xa, contribute in a dominant way to the productive recognition of the protein substrate.
Although comparable mechanistic information is lacking in the human system, studies with human prothrombin derivatives have implied a principal role for an interaction between (pro)thrombin and factor Va, mediated by the (pro)fibrinogenbinding site in the substrate, in human prothrombinase function (12)(13)(14). However, findings with bovine proteins are not in complete agreement with these conclusions. Proteolytic derivatives of prothrombin lacking the fibrinogen-binding site apparently retain the ability to participate in exosite interactions with prothrombinase (10). A monoclonal antibody (␣BFX-2b) 1 directed against a region of the proteinase domain of factor Xa, spatially distinct from the active site, appears to block exosite binding of the protein substrate to bovine prothrombinase without affecting the active site (15). These points along with the necessary caveats associated with the findings highlight uncertainties in the field and cast doubt on the validity of directly extrapolating findings in the bovine system to explain the mechanism of action of human prothrombinase on its protein substrate.
Recombinant nematode anticoagulant protein c2 (NAPc2) is an 85-residue polypeptide originally isolated from the hematophagous hookworm, Ancyclostoma caninum (16). NAPc2 is a potent inhibitor (K i * ϳ10 pM) of factor X activation by the extrinsic Xase complex composed of VIIa and tissue factor (TF) (16). A critical component reaction in the inhibitory pathway is the initial interaction between NAPc2 and either factors X or Xa which facilitates delivery and greatly enhances the affinity of NAPc2 for the VIIa-TF complex (16,17). As NAPc2 binds tightly to human Xa in a way that is apparently independent of the active site of the enzyme (17), we have further characterized these interactions and used NAPc2 as a small, tight binding probe to investigate the function of human prothrombinase.
Plasma-derived Proteins-Plasma for the isolation of human proteins was obtained as a gift from the Plasmapheresis Unit of the Hospital of the University of Pennsylvania. Human factor X and prothrombin were isolated as before (22). Human factor X was further purified in a final step by immunoaffinity chromatography using the Ca 2ϩ -dependent monoclonal antibody 4G3 (23,24). Preparative activation of factor X with the purified activator from Russell's viper venom followed by affinity chromatography on benzamidine-Sepharose (25) yielded an approximate equimolar mixture of the ␣ and ␤ forms of factor Xa. Active site titration of several preparations with p-nitrophenol pЈ-guanidinobenzoate (26), yielded 0.88 -1.21 mol of active sites/mol of Xa. Factor Xa was inactivated with APMSF to yield Xa i and purified as described (27) or inactivated with acetothioacetyl-EGR-CH 2 Cl, reacted with Oregon Green 488 maleimide following thioester hydrolysis, and purified to yield OG 488 -hXa exactly following procedures described for bovine Xa (9). Human factor V was isolated from plasmapheresis plasma by modifications to the immunoaffinity approach described (28). Factor V containing fractions obtained by this approach were pooled, treated with 5 M D-Phe-L-Pro-L-Arg CH 2 Cl (Calbiochem), dialyzed into 20 mM Hepes, 0.15 M NaCl, 5 mM Ca 2ϩ , pH 7.4, and applied to a 1 ϫ 10-cm POROS HQ/M column (Applied Biosystems, Framingham, MA) equilibrated in the same buffer. Bound protein was eluted with a gradient of increasing NaCl (0.15 to 0.6 M) developed in the same buffer (7 min, 5 ml/min). Fractions containing factor V were pooled, concentrated by centrifugal ultrafiltration (Centricon 30, Amicon, Danvers, MA), exchanged into 20 mM Hepes, 0.15 M NaCl, 5 mM Ca 2ϩ , pH 7.4. by repeated concentration, and stored at Ϫ80°C. Factor V (3.5 M, 4.7 ml) was activated by incubation with 50 nM bovine thrombin at 37°C for 15 min. The reaction mixture was treated with 63 M APMSF, diluted with two volumes of 20 mM MES, 5 mM Ca 2ϩ , pH 6.0, and applied to a POROS HS/M column (0.46 ϫ 10 cm, Applied Biosystems) equilibrated in 20 mM MES, 0.1 M NaCl, 5 mM Ca 2ϩ , pH 6.0. Bound protein was eluted with a gradient of increasing NaCl (0.1 to 0.6 M) developed in the same buffer (5 min, 5 ml/min). Factor Va, eluting in a sharp peak at ϳ0.35 M NaCl, well resolved from thrombin and the activation fragments, was concentrated and exchanged into 20 mM Hepes, 0.15 M NaCl, 5.0 mM Ca 2ϩ , pH 7.4, by centrifugal ultrafiltration (Centricon-30), aliquoted, and stored at Ϫ80°C as a ϳ10 M solution. An important feature of this method is that factors V and Va were never exposed to (NH 4 ) 2 SO 4 or glycerol during purification or storage. Human prethrombin 2 and human thrombin were prepared by proteolysis of prothrombin and purified by established procedures (29). Methods for the preparation and purification of X, Xa, Va, and prethrombin 2 of bovine origin, used in some experiments, have been described (9,20,30).
Recombinant Proteins-NAPc2 was expressed in Pichia pastoris and purified as described (16). Integrity of the resulting product was established by mass spectrometry, amino acid composition, and NH 2 -terminal sequence analysis. NAPc2 (265 M) in 0.1 M NaP i , 0.15 M NaCl, pH 7.4, was reacted with 5.3 mM sulfo-N-hydroxysuccinimidyl-biotin (Pierce) for 2 h at 4°C, followed by dialysis against the same buffer to remove unreacted reagent. The product (NAPc2-biotin) contained 2.6 mol of biotin/mol of NAPc2 and was established to be heterogeneously modified by mass spectrometry but retained full factor X-dependent inhibitory activity toward VIIa-TF (17). Recombinant derivatives of factor X were expressed in HEK293 cells using the pCMV4 vector bearing cDNA encoding the human prothrombin signal sequence and propeptide followed by the sequence for the factor X derivative (pCMV4ss-pro-II-fX) as described for human factor X (23). The equivalent vector bearing the sequence encoding bovine X was prepared by PCR using the technique of splicing by overlap extension (31). The cDNA sequence encoding bovine X was recovered by reverse transcriptase-PCR from bovine liver poly(A) mRNA (Clontech, Palo Alto, CA). Following first strand synthesis, bovine factor X cDNA was amplified using primers A (5Ј-CAGCGGGTCCGGCGAGCCAAGTCATTCTTG-3Ј) and B (5Ј-GC-CGCGCCCGGG(T) 18 -3Ј) to yield a product encoding bovine factor X flanked by a 5Ј-extension containing the last 15 bases of the cDNA encoding the human prothrombin propeptide and a 3Ј-extension bearing a XmaI restriction site. In a parallel PCR, human prothrombin cDNA (32) was used as a template along with primers C (5Ј-GATC-CCAAGCTTGGGATGGCGCACGTCCGA-3Ј) and D (5Ј-CAAGAATG AGTTGGCTCGCCGGACCCGCTG-3Ј) to yield a product encoding the propeptide and signal sequence of human prothrombin flanked by a 5Ј-extension containing a HindIII restriction site and a 3Ј-extension complementary to the cDNA for bovine factor X. A final round of PCR used the two PCR products generated in the first round as template and primers B and C. The resulting product (ss-pro-II-X Bov ) was digested using XmaI and HindIII and ligated into the pCMV4 vector previously digested with the same enzymes to yield pCMV4 ss-pro-II-X Bov . A cDNA construct encoding a chimeric derivative of bovine X containing proteinase domain residues 240 -264 2 from human X was prepared by a similar strategy. In the sequential numbering system (33), the X Bov/Hum derivative is composed of bovine X residues 1-423 and human X residues 424 -448. The cDNA encoding the human prothrombin signal sequence and propeptide followed by residues 1-423 of bovine X was obtained using pCMV4 ss-pro-II-X Bov as a template, primer C (above), and primer E (5Ј-GGTTTTCATGGACCTGTCGATCCACTTGAG-3Ј). A parallel reaction with pCMV4 ss-pro-II-X Hum as template, primers F (5Ј-CTCAAGTGGATCGACAGGTCCATGAAAACC-3Ј), and G (5Ј-ATC-GCTTGAACCCAGGAG-3Ј) yielded a product encoding residues 424 -448 of human factor X flanked by appropriate extensions. The final product obtained by PCR using primers C and G and the initial two products as templates was digested and ligated into the cut vector as before to yield pCMV4 ss-pro-II-X Bov/Hum . The identity of all inserts was established by DNA sequence analysis. HEK293 cells were transfected with each of the vector constructs using LipofectAMINE 2000 (Invitrogen) along with plasmids bearing the neomycin resistance gene and paired basic amino acid-converting enzyme-furin as described (23). The production, purification, and detailed characterization of wild-type human factor X has been described (23). For the recombinant bovine derivatives, stable clones selected with G418 were identified by a functional assay in which conditioned medium was treated with the purified factor X activator from Russell's viper venom followed by measurements of the initial rate of SpXa hydrolysis. Selected clones were estimated to produce 5-20 g of functional protein/24 h in a confluent T-25 flask with 10 ml of medium. Two stable cell lines producing each derivative of bovine factor X were separately expanded into cell factories (Nalge-Nunc, Naperville, IL) for long term protein expression in serum-free medium supplemented with vitamin K (Abbott) (23).
Conditioned medium (20 liters) containing either bovine X derivative was applied to a column (4.6 ϫ 6 cm) of XAD-2 (Bio-Rad) to adsorb the indicator dye. A 50ϫ stock solution was used to adjust the colorless effluent to 20 mM Tris, 1 mM benzamidine, 5 mM EDTA, pH 7.4, and factor X was captured by application (24 ml/min) to a column (4.6 ϫ 10 cm) of Q-Sepharose (Amersham Biosciences) equilibrated in 20 mM Tris, 0.15 M NaCl, 1 mM benzamidine, pH 7.4. Bound protein was eluted (10 ml/min, 90 min) with a gradient of increasing NaCl (0.15 to 0.6 M) prepared in the same buffer. Fractions containing factor X, identified by a functional assay, were pooled and dialyzed into 20 mM MES, 0.15 M NaCl, 1 mM EDTA, 1 mM benzamidine, pH 6.0. A fraction (ϳ20%) of the resulting material was applied to a 10 ϫ 1-cm Poros HQ/M column (Applied Biosystems) equilibrated in the same buffer. Bound protein was eluted (2 ml/min) with a gradient of increasing Ca 2ϩ developed with buffer B: 20 mM MES, 0.15 M NaCl, 100 mM CaCl 2 , 1 mM benzamidine, pH 6.0 (0 -100% buffer B, 40 min). Fractions containing factor X functional activity were pooled from repeated runs, precipitated by the addition of (NH 4 ) 2 SO 4 (80% saturation), collected by centrifugation (56,000 ϫ g, 20 min), and dissolved in a small volume of 20 mM Hepes, 0.15 M NaCl, 1 mM benzamidine, pH 7.5. The sample was further purified using a HiPrep S-200 column (2.6 ϫ 60 cm, Amersham Biosciences) equilibrated in the same buffer and developed at 0.2 ml/min. Fractions containing factor X were pooled, precipitated with (NH 4 ) 2 SO 4 (80% saturation), collected by centrifugation (56,000 ϫ g, 20 min), dissolved in 50% (v/v) glycerol, and stored at Ϫ20°C. A representative yield of ϳ34 mg of purified product was obtained from 20 liters of conditioned medium.
Isothermal Titration Calorimetry-Measurements were performed using a MCS-ITC instrument (Microcal, Cambridge, MA) stabilized at 25°C. Factor X derivatives and NAPc2 were dialyzed into 20 mM Hepes, 0.15 M NaCl, pH 7.4, either with or without 2 mM Ca 2ϩ , and the dialysate was used for subsequent dilutions and control experiments. The cell (1.3298 ml) contained factor X or its derivatives at concentrations ranging from 5 to 9 M, and the injection syringe contained 120 M NAPc2. Heat flow was measured with continuous stirring, following a series of injections of NAPc2 spaced at 210-s intervals. In a typical experiment, the first injection was 3 l followed by 29 injections of 6 l each. When the results of two experiments were combined, the injection sequence for the second run was adjusted to yield offset values for the independent variable. Trivial heat flow due to ligand dilution and buffer mismatch was determined from an identical series of injections of NAPc2 into dialysate. SDS-PAGE analysis of samples taken before and after each experiment ruled out proteolysis during the experiment.
Kinetics of Peptidyl Substrate Cleavage in the Presence of NAPc2-Steady state kinetic constants for the cleavage of SpXa were determined either in the absence or presence of 40 nM NAPc2 (ϳ10 -80 ϫ K d ) in reaction mixtures (200 l) prepared in wells of a 96-well plate (Corning Glass), containing increasing concentrations of SpXa (24 values, 0 -500 M) and 1 nM Xa, 1 nM Xa plus 50 M PCPS, or 1 nM Xa plus 50 M PCPS and 25 nM Va. Initial velocities were determined by continuously monitoring the linear increase in absorbance at 405 nm with time at 25°C, measured in a kinetic plate reader (SpectraMax 250, Molecular Devices, Menlo Park, CA). NAPc2-dependent perturbations in initial velocity were independent of the initial incubation time, and equivalent results were obtained when reaction mixtures were initiated with enzyme without prior incubation with NAPc2. Therefore, the system was considered to be in rapid equilibrium, and the reported concentrations correspond to those present in the final reaction mixtures.
Equilibrium dissociation constants for the interaction between NAPc2 and factor Xa were determined by a local analysis of initial velocity measurements in reaction mixtures containing 0.5 nM Xa, 45 M SpXa, or 1 nM Xa plus 50 M PCPS, 100 M SpXa, and increasing concentrations of NAPc2. For the determination of all kinetic constants by global analysis, initial velocities were determined in reaction mixtures containing 1 nM Xa plus 50 M PCPS or 1 nM Xa plus 50 M PCPS and 25 nM Va and increasing concentrations of SpXa using NAPc2 fixed at 0, 0.5, 1, 2, 5, 10, and 20 nM.
Kinetics of Prethrombin 2 Cleavage-Initial velocities for prethrombin 2 cleavage by prothrombinase were determined discontinuously at 25°C exactly as described in the bovine system (9,20). Reaction mixtures (190 l) containing 50 M PCPS, 30 nM Va, and the indicated concentrations of prethrombin 2 and NAPc2 in assay buffer were initiated by the addition of 5 nM Xa (10 l, 100 nM). Aliquots (10 l) withdrawn either before the addition of Xa or serially at t ϭ 30, 60, 90, 120, and 180 s following initiation were quenched and further diluted, and the initial velocity of S2238 hydrolysis was determined by continuously monitoring the absorbance at 405 nm following the addition of 200 M S2238 as described previously (9,20). The concentration of thrombin formed as a function of time was determined by interpolation from the linear dependence of the rate of S2238 hydrolysis on known concentrations of human thrombin determined in each experiment. The initial steady state rate of thrombin formation in the initial reaction mixture was determined from the linear appearance of thrombin with time.
Fluorescence Anisotropy-Steady state fluorescence anisotropy was measured using a customized fluorescence spectrophotometer (RSM-1000, OLIS, Bogart, GA) operated in T-format with Glan-Thompson polarizers. Samples (2.0 ml) were maintained at 25°C in 1-cm 2 stirred quartz cuvettes, and anisotropy was measured using EX ϭ 490 and EM ϭ 520 nm with long pass filters (KV500, Schott, Duryea, PA) in the emission beams. With the exception of the instrumentation, all measurements including control experiments and scattering corrections were performed exactly as described in detail for similar measurements with bovine prothrombinase (10).
Assessment of Competition between ␣BFX-2b and NAPc2-Reaction mixtures (150 l), prepared in assay buffer, contained 1.1 nM Xa, 67 M PCPS, and increasing concentrations of NAPc2 in the presence of different fixed concentrations of ␣BFX-2b. Following prolonged incubation for 30 min at ambient temperature, initial rates were determined following the addition of 100 M SpXa. For consistency with the other experiments, final reactant concentrations were reported even though a final equilibrium was unlikely to have been established during the brief initial velocity determination.
Affinity Chromatography on a Heparin Matrix-High performance heparin affinity chromatography was performed using a 0.46 ϫ 10-cm POROS HE/M column (Applied Biosystems) equilibrated in 20 mM Hepes, 5 mM Ca 2ϩ , pH 6.5. Samples (1.8 ml) prepared in the same buffer containing either 5 M factor X or 5 M factor X plus 7.5 M NAPc2 were applied at 2 ml/min. Following washing (8 ml), bound protein was eluted with a gradient of increasing NaCl (0 -1.0 M, 10 min) prepared in the same buffer.
Ligand Blotting-Protein samples (5 g each) were subject to SDS-PAGE without disulfide bond reduction and electrophoretically transferred to Immobilon PSQ membranes (Millipore, Milford, MA) with a semi-dry apparatus (Hoefer, San Francisco, CA) using 20 mM Tris, 20 mM glycine, 10% (v/v) MeOH, pH 8.3, at 100 mA for 60 min. The membrane was blocked with blot buffer (20 mM Hepes, 0.15 M NaCl, 0.1% (v/v) Tween 20, pH 7.5) containing 0.1% (w/v) BSA for 30 min, washed 3 times with blot buffer, and incubated with 82 nM NAPc2biotin in blot buffer for 1 h. The membrane was washed 3 times with blot buffer and incubated (30 min) with avidin-biotinylated horseradish peroxidase complex (ABC Elite reagent, Vector Laboratories, Burlingame, CA) prepared according to the manufacturer's directions. Excess reagent was removed by 3 washes using blot buffer followed by 2 washes with blot buffer lacking Tween 20. Reactive bands were visualized following the addition of ECL reagent (Amersham Biosciences), and chemiluminescence was detected by exposure to XB-1 film (Eastman Kodak).
Data Analysis-In kinetic studies using factor Xa either in the presence of PCPS or in the presence of factor Va plus PCPS, reactant concentrations were chosen to ensure that the concentration of factor Xa was limiting, and the concentrations of PCPS and Va were saturating and well above the K d values for the individual interactions (see below) (40,41). In each case, the concentration of catalyst (E) was considered equivalent to the limiting concentration of factor Xa incorporated saturably into the Xa-PCPS binary complex or the Xa-Va-PCPS ternary complex (41).
All constants were obtained by fitting data to the indicated explicit expressions using nonlinear least squares regression analysis (42). The fitted constants are presented Ϯ95% confidence limits.
Heat flow traces obtained by ITC measurements were integrated, corrected for background, and analyzed using Origin Scripts provided by Microcal (Northampton, MA) using the expressions for the binding of titrant to identical and non-interacting sites (43)(44)(45). When necessary, non-linear least squares routines accommodated the simultaneous analysis of two or more data sets obtained with a staggered injection sequence. Analysis by either approach, using 1/Y weighting, yielded fitted values for ⌬H, K d , and stoichiometry. ⌬G and ⌬S were calculated using ⌬G ϭ RT lnK d and ⌬G ϭ ⌬H Ϫ T⌬S.
Steady state kinetic constants for the hydrolysis of SpXa either in the absence or in the presence of a saturating concentration of NAPc2 were determined by nonlinear least squares regression analysis according to the Henri-Michaelis-Menten equation.
Analysis according to Scheme I was done by combining nonlinear error minimization with numerical solution of rate and equilibrium expressions using the program Dynafit obtained as a gift from Petr Kuzmic (BioKin, Pullman, WA) (46). This approach used the rapid equilibrium assumption and assumed a stoichiometry (moles of NAPc2/ mol of E) of 1. Changes in initial velocity with increasing concentrations of NAPc2 at a single concentration of SpXa were analyzed by the "local" method in which the independently measured kinetic constants in the absence or presence of NAPc2 were taken as known values for K E,S , k cat(ES) , K EN,S , and k cat(ENS) . Local analysis yielded fitted values for K E,N and K ES,N . Data sets obtained using increasing concentrations of SpXa at different fixed concentrations of NAPc2 were analyzed "globally" to yield fitted values for K E,S , k cat(ES) , k cat(EN,S) , K E,N , and K ES,N . K EN,S was calculated from Equation 1, The errors reported for the numerical analyses correspond to linear approximations of the 95% confidence limits. Uncertainty in K EN,S was established by error propagation (42). Kinetic constants describing tight binding competitive inhibition of prethrombin 2 cleavage by prothrombinase were extracted by analysis of the rate data according to Equations 2 and 3, where v, V max , and K m have the usual meaning. S, E, and I refer to total concentrations of substrate, enzyme, and inhibitor, and I f refers to the free concentrations of I. Inhibition results from the interaction of I with E with an equilibrium dissociation constant of K i and stoichiometry n. Analysis according to Equations 2 and 3 yielded fitted values for V max , K m , n, and K i .  (47) solved iteratively by the Newton-Raphson method (42). Nonlinear least squares analysis yielded fitted constants for the indicator interaction (K d , n, ⌬ ϱ r ) as well as the equilibrium dissociation constant for the membrane-mediated interaction between Xa i and Va (K dComp ) and moles of Va bound per mol of Xa i at saturation (n Comp ).

RESULTS
Binding of NAPc2 to Factors X and Xa-Thermodynamic parameters describing the interaction between NAPc2 and factor X derivatives were established by ITC. Incremental additions of NAPc2 to a solution of factor X yielded significant heat flow that decreased to base-line levels at concentrations of NAPc2 in excess of 1 eq (Fig. 1). The combined analysis of two injection sequences, to enhance resolution of the sharp transition, yielded n Х 1 mol of NAPc2/mol of human X at saturation, ⌬H ϭ Ϫ13.8 Ϯ 0.05 kcal⅐mol Ϫ1 , and K d ϭ 0.87 Ϯ 0.11 nM. In line with previous findings (17), the data establish, in an unambiguous way, the ability of NAPc2 to bind tightly to the zymogen. Similar results were obtained with factor Xa i and with factor X in the absence of Ca 2ϩ (Table I). The thermodynamic constants indicate that NAPc2 binds tightly and with comparable affinity to both factors X and Xa i in an enthalpically driven reaction. Tight binding by NAPc2 is independent of Ca 2ϩ and neither requires zymogen activation nor a fully functional active site. ITC experiments with active Xa or with a recombinant Ser 195 3 Ala mutant yielded biphasic titrations, implying a more complex interaction (not shown). Analysis by sedimentation velocity revealed that whereas Xa i was monomeric, active Xa and Xa S195A were aggregated and polydisperse at micromolar concentrations (not shown). The complex titrations observed with these derivatives, possibly related to aggregation phenomena, precluded an unambiguous interpretation of ITC measurements with uninhibited Xa.
Modulation of Xa Active Site Function by NAPc2-As noted previously (16), NAPc2 was found to modestly perturb the initial rate of peptidyl substrate hydrolysis by Xa in the absence or presence of PCPS and Va. NAPc2-dependent changes in initial velocity varied with the nature and concentration of the peptidyl substrate but were independent of prior incubation of enzyme with NAPc2 (not shown), implying the rapid achievement of equilibrium between enzyme, substrate, and NAPc2. In all cases, changes in reaction rate saturated with increasing concentrations of NAPc2 over a concentration range, consistent with the tight binding parameters identified by ITC.
Steady state kinetic constants determined in the absence and presence of a vast excess of NAPc2 documented the kinetic basis for the ability of this probe to alter the initial velocity for  Table I. SpXa hydrolysis catalyzed by Xa, Xa saturated with PCPS (Xa-L), or Xa saturated with both PCPS and Va (Xa-Va-L) ( Table II, local analysis). In each case, saturating concentrations of NAPc2 modestly decreased the K m value for SpXa. An associated decrease in k cat was observed for Xa-L and Xa-Va-L. The findings suggest that NAPc2 binds to factor Xa regardless of the presence of other prothrombinase constituents and modestly perturbs active site function. The slightly enhanced affinity for SpXa in the presence of NAPc2 implies that this probe does not occlude the active site of the enzyme.
The findings are consistent with a kinetic model (Scheme I) in which NAPc2 (N) acts as a modulator of SpXa (S) hydrolysis catalyzed by Xa, Xa-L, or Xa-Va-L (E). N may bind to E or the ES complex and modestly alter steady state kinetic constants for the cleavage of S.
Binding of NAPc2 to Factor Xa and Prothrombinase-Equilibrium dissociation constants for the binding of NAPc2 to Xa, Xa-L, or Xa-Va-L were inferred from steady state kinetic studies and analysis according to Scheme I. When feasible, K E,N and K ES,N were "locally" determined from the change in initial velocity observed with increasing concentrations of NAPc2 (e.g. Fig. 2A), using the independently determined steady state kinetic constants in the absence and at saturating NAPc2 as known parameters. Alternatively, initial velocities measured at increasing concentrations of SpXa at different fixed concentrations of NAPc2 were used to globally extract all kinetic and equilibrium constants illustrated in Scheme I (e.g. Fig. 2B). Both approaches yielded a series of internally consistent constants (Table II). The inferred equilibrium dissociation constants (K E,N ) for the binding of NAPc2 to either solution-phase Xa or Xa saturated with PCPS were equivalent to each other and in tolerable agreement with constants derived from ITC measurements (Table I). However, K E,N was increased by a factor of ϳ10 when Xa was incorporated into prothrombinase with saturating concentrations of PCPS and Va (Table II). In each case, equilibrium dissociation constants inferred for the binding of NAPc2 to enzyme saturated with substrate (K E,S,N ) were modestly decreased. Taken together with the thermodynamic measurements, these data provide quantitative support for the conclusion that NAPc2 binds with comparable affinity to factors X and Xa, in a manner that neither requires a functional active site nor one that leads to the occlusion of the active site of the proteinase. Assembly of factor Xa into prothrombinase leads to a modulation in affinity for NAPc2, evident as 8 -10-fold increase in the equilibrium dissociation constant.
Inhibition of Protein Substrate Cleavage by Prothrombinase-In contrast to the small effects of NAPc2 on catalytic function assessed with SpXa, initial experiments revealed that this probe was a complete inhibitor of prothrombin activation by prothrombinase, in agreement with previous observations (16). Because of the known difficulties in the meaningful interpretation of the kinetics of prothrombin activation (20,30), mechanistic information was further sought using prethrombin 2, established as an appropriate substrate analog for the first half-reaction of prothrombin activation (20,30).
Initial velocity studies established NAPc2 as a tight binding inhibitor of prethrombin 2 cleavage by prothrombinase (Fig. 3). Inhibition was independent of prior incubation of NAPc2 with the enzyme (not shown) but depended on the concentration of SCHEME I b In the local method of analysis, steady state kinetic constants for the cleavage of SpXa were independently determined in the absence of NAPc2 (K E,S and k cat(ES) ) or in the presence of 40 nM NAPc2 (ϳ10 -80 ϫ K d ) to yield estimates of K EN,S and k cat(ENS) ⅐K E,N and K ES,N were determined separately by numerical analysis of velocity changes as a function of increasing concentrations of NAPc2 (e.g. Fig. 2A) according to Scheme I, using the determined steady state kinetic constants. In the global method, data sets obtained with increasing concentrations of SpXa in the presence of different fixed concentrations of NAPc2 (e.g. Fig. 2B) were globally analyzed according to Scheme I to yield all listed constants. c ND, not determined. substrate (Fig. 3). The data could be described by the rate expression for tight binding complete, competitive inhibition (Fig. 3) yielding a stoichiometry of ϳ1 and a K i in agreement with the equilibrium dissociation constant determined from studies of SpXa cleavage by prothrombinase (Table II). The fitted steady state kinetic constants for prethrombin 2 cleavage were also in agreement with independently determined values of K m ϭ 7.8 Ϯ 0.6 M and V max /E T ϭ 2.62 Ϯ 0.03 s Ϫ1 . Alternative inhibition mechanisms could be excluded on the basis of the series of criteria described previously (15,48). NAPc2 and prethrombin 2 therefore bind in a mutually exclusive manner to prothrombinase. As NAPc2 binding to Xa within prothrombinase does not occlude the active site of the proteinase, it follows that competitive inhibition of protein substrate cleavage is achieved by interfering with interactions at sites removed from the active site. These are properties expected of an exosite-directed inhibitor of prothrombinase (9). Influence of NAPc2 on the Assembly of Prothrombinase-Deleterious effects of NAPc2 on the equilibrium dissociation constant for the interaction between Xa and Va could provide an explanation for the findings. Kinetic studies in the presence of different fixed and saturating concentrations of Va failed to provide obvious evidence for the destabilization of prothrombinase by NAPc2 (not shown). Nevertheless, this possibility was further evaluated by binding studies.
Bovine Xa modified with dansyl-EGR-CH 2 Cl has proved a powerful fluorescent probe for thermodynamic and kinetic studies of the assembly of bovine prothrombinase (49,50). Because an equivalent approach with human proteins fails to yield a reliable change in probe intensity, anisotropy, or excited state lifetime (not shown), comparable methodology and physical information pertaining to the assembly of human prothrombinase is not available. Human factor Xa, containing Oregon Green 488 , covalently incorporated at the active site with a peptidyl chloromethyl ketone tether (OG 488 -hXa) produced a useful change in anisotropy following its assembly into prothrombinase. The anisotropy of OG 488 -hXa was minimally changed following the addition of saturating concentrations of PCPS despite the fact that Xa binds to these membranes with good affinity (40). Subsequent titration with increasing concentrations of factor Va yielded a saturable increase in anisotropy that was dependent on the fixed concentration of OG 488 -hXa (Fig. 4A). No change in anisotropy was observed in buffer containing EDTA in place of Ca 2ϩ (Fig. 4A). Titration of reaction mixtures containing OG 488 -hXa, Va, and saturating concentrations of PCPS with increasing concentrations of a nonfluorescent, inactivated Xa derivative (Xa i ) reduced the anisotropy (Fig. 4A, inset). Based on the model previously developed for the analysis of binding interactions in the assembly of bovine prothrombinase (41), the titration data yielded K d ϭ 4.34 Ϯ 0.56 nM for the interaction between OG 488 -hXa and human Va on the membrane surface and a stoichiometry of 1.17 Ϯ 0.04 mol of Va bound per mol of OG 488 -hXa at saturation. Analysis of the competition data (Fig. 4A, inset) yielded comparable equilibrium parameters for the binding of either OG 488 -hXa or Xa i to human Va, thereby documenting the reversibility and authenticity of the binding interactions monitored using OG 488 -hXa. In keeping with extensive work performed with bovine prothrombinase (10,41), these points outline the basic and quantitative features of the assembly of human prothrombinase and establish OG 488 -hXa as an appro-  (Table II). B, initial velocities were determined using reaction mixtures 1 nM Xa plus 50 M PCPS and increasing concentrations of SpXa and NAPc2 fixed at 0 (q), 0.5 (E), 1 (OE), 2 (‚), 5 (), 10 (ƒ), and 20 nM (f). The lines are drawn following global analysis according to Scheme I with the fitted parameters listed in Table II. priate reporter group for the assembly process.
NAPc2 was found to decrease the maximum anisotropy change observed following the assembly of OG 488 -hXa into prothrombinase. Despite the decreased amplitude of the signal, approximately equivalent equilibrium dissociation constants and stoichiometries for the assembly of prothrombinase were obtained from titration curves in the absence and presence of a vast excess of NAPc2 (Fig. 4B). Therefore, the decreased anisotropy change in the presence of NAPc2 does not result from a disruption of prothrombinase but rather an altered signal associated with the assembly of OG 488 -hXa into prothrombinase in the presence of NAPc2. The results support the interpretation that NAPc2 does not interfere, in an obvious way, with the membrane-dependent interaction between Xa and Va. Thus, selective inhibition of protein substrate cleavage by NAPc2 does not arise from a destabilization of the interaction between Xa and Va. The 10-fold reduction in the affinity of NAPc2 following the incorporation of Xa into prothrombinase cannot be explained by a matched decrease in the interaction between Xa and Va from the simple consideration of two linked binding interactions. The decreased affinity of NAPc2 for prothrombinase could arise from a Va-dependent modulation of surfaces in factor Xa that follows the interaction between OG 488 -hXa and Va detected by the anisotropy change.
Modulation of NAPc2 Binding by Ligands Targeting the Proteinase Domain-The functional effects of NAPc2 on Xa and prothrombinase function parallel those previously described with the monoclonal antibody ␣BFX-2b (15). Prolonged incubation of Xa with ␣BFX-2b in the presence of NAPc2 progressively reduced the ability of NAPc2 to modulate the initial velocity of SpXa (Fig. 5A). The slow, tight binding interaction between ␣BFX-2b and Xa (15) precluded a quantitative analysis of the potentially competitive equilibria established in this reaction system. However, simulations yielded the preliminary conclusion that the interaction between ␣BFX-2b and Xa or NAPc2 and Xa reduces the affinity for the second ligand by at least a factor of 40 (Fig. 5A). The data therefore suggest that ␣BFX-2b somehow interferes with the binding of NAPc2 to Xa.
Similar observations with high concentrations of low molecular weight heparin suggested that the low affinity interaction between heparin and Xa (51) could also affect NAPc2 binding. This was qualitatively documented by affinity chromatography using a heparin matrix. Factor X was near-quantitatively retained by the resin and could be recovered following elution with increasing concentrations of NaCl (Fig. 5B). In contrast, premixing of factor X with 1.5 molar eq of NAPc2 blocked binding of factor X to the resin (Fig. 5B). Although the quantitative basis for this effect is presently unclear, NAPc2 appears to significantly affect the binding of both heparin and ␣BFX-2b to factor X. Because both heparin and ␣BFX-2b are known to bind to sites in the heavy chain of factor Xa (15,51), the observations implicate a role for structures present in the proteinase domain in binding NAPc2.
Structural Studies of the Interaction of NAPc2 with Xa-A biotinylated derivative of NAPc2 (NAPc2-biotin) was found to reveal bands corresponding to human X in ligand blotting studies following SDS-PAGE without disulfide bond reduction. SDS-PAGE analysis of a series of plasma-derived and recombinant factor X derivatives along with human prethrombin 2 and human factor Va (Fig. 6, A and B) is compared with the corresponding ligand blot obtained with NAPc2-biotin without disulfide bond reduction (Fig. 6C). NAPc2-biotin could detect human factor X, both ␣ and ␤ forms of human Xa, as well as recombinant human factor X but not Va, prethrombin 2, and either plasma derived or recombinant bovine factor X (Fig. 6C). The data document the apparent specificity of NAPc2 for human factors X and Xa and confirm the previous suggestion (52) that NAPc2 binds weakly, if at all, to bovine factor X or even bovine Xa (not shown in Fig. 6). The sequences of human and bovine factor Xa are strikingly dissimilar in the 19 -25 residues at the COOH terminus of the proteinase domain following residue 240 (53). A recombinant chimeric derivative of bovine factor X, in which heavy chain residues 240 onward were replaced with the sequence present in human factor X, bound NAPc2-biotin in the ligand blotting approach (Fig. 6C). Thus, the COOH terminus of the heavy chain of human factor X somehow contributes to the high affinity interaction with NAPc2, and differences in this region are at least partly responsible for the surprising species selectivity of NAPc2. The quantitative bases for these findings were established by ITC (Table III). Recombinant human factor X yielded thermodynamic constants that were in agreement with those obtained in a parallel experiment with factor X isolated from human plasma. Both recombinant and naturally occurring forms of bovine factor X bound NAPc2 with 300 -500-fold lower affinity. In contrast, the recombinant chimeric form of bovine factor X bearing the human X sequence at the COOH terminus produced an isotherm with a sharp transition and an equilibrium dissociation constant comparable with that observed with human factor X (Table III). The findings imply that the COOHterminal residues of the proteinase domain of factors X and Xa somehow play a role in the high affinity interaction with NAPc2. Despite the clear-cut findings from ligand blotting, the measured thermodynamic parameters (Table III) suggest caution in pursuing the simplistic interpretation that residues in the COOH terminus of human factor X, not present in the bovine protein, provide the primary binding site for NAPc2.
Significance of the Selectivity of NAPc2 for Human Xa-The data establish exosite-mediated recognition of the protein substrate as an important feature of the function of human prothrombinase. As NAPc2 acts as a competitive inhibitor of protein substrate cleavage by human prothrombinase, it is possible that structural elements of human Xa that participate in binding NAPc2 also play a role in exosite-dependent macromolecular substrate recognition. Because both human and bovine prothrombinase likely employ similar mechanisms for substrate recognition, the apparent selectivity of NAPc2 for human Xa raises concerns regarding the significance of such conclusions.
In agreement with the modest affinity observed for the interaction between NAPc2 and bovine X, no evidence was obtained for an interaction between bovine Xa in the absence or presence of saturating concentrations of PCPS and nanomolar concentrations of NAPc2 (not shown). However, NAPc2 was a potent inhibitor of prethrombin 2 cleavage catalyzed by bovine Xa assembled into prothrombinase with saturating concentrations of PCPS and bovine Va (Fig. 7). The data illustrate that NAPc2 acts as a competitive inhibitor of protein substrate cleavage by bovine prothrombinase with an affinity (K i ϭ 1.4 Ϯ 0.09 nM) slightly superior to that observed for the human enzyme. Therefore, the modest affinity of NAPc2 for bovine X and Xa is greatly enhanced following the assembly of bovine Xa into prothrombinase. Initial velocity measurements of SpXa hydrolysis indicate that NAPc2 has no obvious effect on active site function of bovine Xa within prothrombinase (Fig. 7). The data indicate that regardless of differences at the COOH terminus of the proteinase, NAPc2 binds with comparable affinity to both human and bovine Xa within prothrombinase at a site distinct from the active site. This binding interaction selectively blocks exosite-dependent protein substrate recognition by either enzyme and leads to equivalent functional consequences.
The large increase in affinity for NAPc2 observed upon assembly of bovine Xa into prothrombinase provides further evidence to support the initial suggestion that the interaction between Xa and Va modulates the proteinase at sites distinct from the active site, leading to changes in the affinity for NAPc2. This modulation could either occur at the COOH  Table II with the simulated estimate that K E,N is increased by a factor of 40 following antibody binding. B, samples (1.8 ml) containing either 5 M human X (light chromatogram) or 5 M human X plus 7.5 M NAPc2 (dark chromatogram) in 20 mM Hepes, 5 mM Ca 2ϩ , pH 6.5, were subject to high performance affinity chromatography on a heparin matrix. Bound material was eluted with a gradient of increasing NaCl initiated at the arrow. Chromatograms were obtained by continuously monitoring absorbance at 280 nm (left axis) and conductivity (right axis, dotted line).
terminus of the proteinase domain itself or at additional sites in the proteinase that play a role in binding NAPc2. DISCUSSION Extensive work has suggested a dominant role for extended macromolecular interactions in determining protein substrate binding specificity for bovine prothrombinase (9 -11). One hallmark of this strategy is that ligands that block the exositemediated bimolecular interaction between enzyme and substrate are expected to act as competitive inhibitors of protein substrate cleavage without occluding the active site of the proteinase within the enzyme complex or interfering with the assembly of prothrombinase (10). By using NAPc2 as a tight binding probe, we now establish these formal criteria to apply to the function of human prothrombinase. Therefore, extended macromolecular interactions play a key role in substrate recognition by the human enzyme complex as well. The data further support the conclusion that occlusion of surfaces on factor Xa within prothrombinase, at sites distinct from the active site, is sufficient to interfere with exosite-dependent binding of the protein substrate.
NAPc2 was initially identified as a potent factor X-or Xadependent inhibitor of VIIa-TF (16). In agreement with previous work (17), our results support the idea that NAPc2 binds to human factors X and Xa equivalently and in a way that does not require or occlude the active site of the proteinase. Whereas these interactions appear essential for the delivery of NAPc2 to VIIa-TF (17), they also lead to tight binding inhibition of the action of prothrombinase on its protein substrate. Thus, NAPc2 targets two enzyme complexes of blood coagulation with unique strategies. However, inhibition of VIIa-TF by NAPc2 bound to either factors X or Xa occurs with far greater affinity than the inhibition of prothrombinase (17). The plasma concentrations of NAPc2 achieved in clinical studies suggest that the efficacy of NAPc2 as a therapeutic antithrombotic in humans (54,55) probably derives from inhibition of VIIa-TF.
The monoclonal antibody ␣BFX-2b, directed against the proteinase domain of Xa, has been established as a prototypic exosite-directed inhibitor in previous work (15) with bovine prothrombinase. Equivalent observations, now described in a completely independent approach with a substantially smaller 85-residue polypeptide probe, greatly increase the confidence in interpretations from such work. It now appears less likely that the distinctive features of such exosite-inhibitors can be explained by generalized steric phenomena associated with the binding of a very large probe to factor Xa within the enzyme complex. The fact that ␣BFX-2b is able to interfere with the binding of NAPc2 to human Xa further indicates that the comparable properties of the two probes arise from related binding interactions with the proteinase.
The high affinity interaction between NAPc2 and the chimeric X Bov/Hum derivative implicates the COOH terminus of the proteinase domain of human factor X in NAPc2 binding. Autoproteolytic cleavages following Lys 251 and Lys 249 at the COOH terminus of the proteinase domain lead to the conversion of ␣ human Xa to the ␤ form (56), with the probable release of the COOH-terminal peptide(s) (22). The ability of NAPc2 to bind both these forms of Xa evident in ligand blotting studies and the absence of heterogeneity in ITC experiments with approximate equimolar mixtures of ␣ and ␤ Xa i implies that NAPc2 binds both species equivalently. Together, the data suggest that side chains present in the 240 -251-residue region somehow contribute to NAPc2 binding.
X-ray structures of Xa are variably truncated or poorly resolved at the COOH terminus of the proteinase domain. Residues 240 -245 in a human Xa derivative (57) are located on a face of Xa clearly removed from the active site of the enzyme but immediately adjacent to the discontinuous epitope delineated for the binding of ␣BFX-2b (15). Residue 240 has been implicated in binding heparin (51). These points establish structural correlates for the ability of heparin or ␣BFX-2b to interfere with the interaction between NAPc2 and human Xa and provide a plausible physical explanation for the equivalent way in which both ␣BFX-2b and NAPc2 affect enzymic function.
Altered binding of NAPc2 to Xa following its assembly into prothrombinase replicates observations made with ␣BFX-2b (15). Modulation in binding was evident as a modest decrease in affinity for NAPc2 following the assembly of human prothrombinase and as a large increase in affinity following the assembly of the bovine enzyme complex. Without ascribing significance to magnitude or direction of the change, the data support the previous proposal (15) that the interaction between Xa and Va on the membrane surface detectably perturbs struc- a The derivatives of human (X Hum ) or bovine factor X (X Bov ) are prefixed with p or r to denote the plasma derived or recombinant species. r-X Bov/Hum denotes the chimeric recombinant derivative of bovine X containing heavy chain residues 240 -264 from human factor X. tures in the proteinase domain removed from the active site. Such perturbations could potentially be realized as a large enhancement in exosite-mediated binding affinity of the enzyme complex for the protein substrate.
The equivalent behavior of NAPc2 with either human or bovine prothrombinase allays concerns regarding the significance of a potentially species-specific probe when both the human and bovine enzyme complexes likely function equivalently. However, the potential role of the COOH-terminal region of the proteinase domain in binding interactions with NAPc2 is less certain. The thermodynamic measurements indicate that NAPc2 indeed binds bovine X with modest affinity and illustrate the inevitable dangers in the interpretation of negative results from blotting techniques. Recognizing that ITC measurements yield net free energies, the higher affinity of NAPc2 for binding human X/Xa can be explained by the added independent energetic contribution of just one salt bridge or one H-bond (58). Thus, the apparently large differences in the affinity of NAPc2 for human relative to bovine X/Xa could arise from relatively minor differences in protein-protein contacts.
Although there are many alternative possibilities, the data can be adequately described by a model in which the COOH-terminal region and/or neighboring structures of the proteinase domain contribute to the interaction with NAPc2. Structural differences in this region could account for the large difference in affinity observed for the binding of NAPc2 to human versus bovine X/Xa, and Va-dependent modulation of these structures could explain the altered affinity for NAPc2 following the assembly of prothrombinase. Alternatively, modulation by Va could occur at additional sites in factor Xa that participate in binding NAPc2. Based on the compact dimensions of NAPc2 determined by NMR (59), we speculate that such sites are unlikely to be found in a completely distant region of the proteinase. Binding of NAPc2 to some or all such sites specifically interferes with protein substrate binding without occluding the active site.
The previous discussion highlights the difficulty in distinguishing the contributions of alternate models in which Va is directly responsible for exosite docking of the substrate to prothrombinase (12) or promotes this interaction indirectly by binding Xa and modulating extended surfaces in Xa, which in turn bind the protein substrate (15). Despite a litany of valid arguments that can be offered, we are unable to rule out unequivocally the possibility that the effects of NAPc2 arise from direct or indirect effects on substrate-binding sites present on Va. However, the preponderance of evidence from independent approaches now points to a substantive role for extended surfaces on the proteinase domain of Xa within prothrombinase to protein substrate recognition (15). Although additional contributions from substrate-cofactor interactions may play a role, the data indicate that the ligation of surfaces in Xa, distinct from the active site, is sufficient to block exosite-dependent recognition of the protein substrate by prothrombinase. The binding footprint of probes such as NAPc2 and ␣BFX-2b, on a face of the proteinase domain of factor Xa, clearly removed from the active site, could define a region of the enzyme that plays a role in determining the binding specificity of prothrombinase for its protein substrate.