Definition of a Factor Va Binding Site in Factor Xa*

We reported previously that residue 347 in activated fX (fXa) contributes to binding of the cofactor, factor Va (fVa) (Rudolph, A. E., Porche-Sorbet, R. and Miletich, J. P. (2000) Biochemistry 39, 2861–2867). Four additional residues that participate in fVa binding have now been identified by mutagenesis. All five resulting fX species, fXR306A, fXE310N, fXR347N, fXK351A, and fXK414A, are activated and inhibited normally. However, the rate of inhibition by antithrombin III in the presence of submaximal concentrations of heparin is reduced for all the enzymes. In the absence of fVa, all of the enzymes bind and activate prothrombin similarly except fXaE310N, which has a reduced apparent affinity (∼3-fold) for prothrombin compared with wild type fXa (fXaWT). In the absence of phospholipid, fVa enhances the catalytic activity of fXaWT significantly, but the response of the variant enzymes was greatly diminished. On addition of 100 nm PC:PS (3:1) vesicles, fVa enhanced fXaWT, fXaR306A, and fXaE310N similarly, whereas fXaR347N, fXaK351A, and fXaK414Ademonstrated near-normal catalytic activity but reduced apparent affinity for fVa under these conditions. All enzymes function similarly to fXaWT on activated platelets, which provide saturating fVa on an ideal surface. Loss of binding affinity for fVa as a result of the substitutions in residues Arg-347, Lys-351, and Lys-414 was verified by a competition binding assay. Thus, Arg-347, Lys-351, and Lys-414 are likely part of a core fVa binding site, whereas Arg-306 and Glu-310 serve a less critical role.

The activation peptide and the serine protease domain form the heavy chain of fX, which is bound to the light chain via a single disulfide bond.
Factor Xa interacts synergistically with cofactor (fVa), substrate (prothrombin), and a phospholipid surface to form the prothrombinase complex which supports maximally efficient prothrombin activation (11). Several of these interactions were independently evaluated for a variant recombinant fX, fXa R347N (residue 165 in chymotrypsin numbering), and it was found that substitution of Arg-347 selectively reduces fVa affinity (12). The current study describes further delineation of the fVa binding site of fXa through targeted mutagenesis of additional residues in the surface epitope that includes Arg-347.
Arg-306, Lys-351, and Lys-414 (125, 169, and 230 in chymotrypsin numbering, respectively) have been substituted by alanine. Glutamate 310 (129 in chymotrypsin numbering) was substituted by asparagine. The fVa affinity of the resulting enzymes was probed using functional and binding studies. It was determined that Arg-347, Lys-351, and Lys-414 compose a core cofactor binding epitope and Glu-310 and Arg-306 form an extended region of the epitope.
Immunoglobulins-Anti-fX murine monoclonal antibodies utilized in the study were developed by standard methods. 3698.1A8.10 reacts with the Gla domain in the presence of calcium, 3514.5H12.10 reacts with the Gla domain independent of calcium, and 3448.1D7.20 binds to EGF-2. Activation of fX was monitored by a two-site immunofluorescent assay utilizing 3448.1D7.20 and 3514.5H12.10 as described (13). Murine anti-fX monoclonal antibody directed against the heavy chain was purchased from Enzyme Research Laboratories, Inc. (South Bend, IN). For immunoblotting, an antibody mixture of 3448.1D7.20, 3514.5H12.10, and the anti-fX heavy chain monoclonal were utilized.
Activation-The assay buffer used was 10 mM HEPES, pH 7.0, 100 mM NaCl, 5 mM CaCl 2 , 1 mg/ml bovine serum albumin, 1 mg/ml polyethylene glycol 8000. All variant zymogens were activated as described previously for fX R347N (12). Briefly, variants or fX WT (100 nM) were activated with 2 nM RVV-X for end point activation or 50 pM RVV-X for initial rate measurements. Each fX was activated by fVIIa (40 pM) in the presence of 0.5 nM lipidated tissue factor, and by fIXa (2 nM) in the presence of fVIIIa (4 units/ml) on PC:PS vesicles (20 M). All concentrations were subsaturating for the activators and initial rates of activation (Ͻ10% of substrate utilized) were determined. Reactions were monitored over time by quenching aliquots of the reaction in EDTA buffer (10 mM HEPES, pH 7.0, 100 mM NaCl, 5 mM EDTA, 1 mg/ml bovine serum albumin, 1 mg/ml polyethylene glycol 8000) and measuring the rate of hydrolysis of Spectrozyme FXa (100 M).
Thrombin Formation-Assays were performed in assay buffer as described (12). Kinetic values were calculated from the least squares fit of the data to Equation 1, where v ϭ the observed initial rate of thrombin formation, V max ϭ the maximal initial rate of thrombin formation, [Z]ϭ the concentration of the component being varied in the experiment, and K m ϭ the apparent concentration of the variable component required to reach half-maximal thrombin formation under the conditions specified. Data are expressed as moles of thrombin (IIa)/s/mol of Xa. Samples were removed from each reaction at various times and quenched into EDTA buffer. Thrombin formed was quantitated by incubating quenched samples with 500 M Spectrozyme TH, a thrombin chromogenic substrate. The conditions for the experiments have been described in detail (12). In summary, 10 nM each fXa was incubated with prothrombin alone (0 -450 M). In the absence of phospholipid, varying concentrations of fVa (0 -250 nM) were incubated with 10 M prothrombin and fXa (0.5 nM). In the presence of PC:PS vesicles (3:1, 20 M), 20 pM fXa was incubated with fVa (0 -1 nM) and the reaction was initiated with 1 M prothrombin. Thrombinactivated platelets (10 8 /ml) were incubated with fXa (0 -1 nM), and thrombin formation was initiated by the addition of 1 M prothrombin. Competition Binding Assay-Reagents for the latex bead-based binding assay were prepared as described (12). Briefly, fX S379A was labeled with 125 I using Bolton-Hunter reagent (Amersham Pharmacia Biotech). Radioactivity not incorporated into protein was removed using a Bio-Spin 6 column (Bio-Rad). The specific activity of the labeled protein was typically 2000 cpm/ng. Labeled FX S379A was activated using RVV-X as described above.
Latex beads (1.0 m; Interfacial Dynamics Corp., Portland, OR) were coated with PC:PS (3:1) according to published methods (12). Beads coated with PC only were also prepared and used in pilot experiments to demonstrate specificity; i.e. no fVa-dependent binding without PS. Nonspecific binding accounted for Ͻ10% of binding, and ϳ50% of 125 I-fXa S379A binding was prevented by the addition of 1 nM unlabeled fXa WT .
The concentrations of all assay components were empirically determined as described (12). In the assay, 1 nM fVa was incubated with 0.2% (v/v) PC:PS beads and 0.8% (v/v) uncoated beads for 10 min. 125 I-fXa S379A (final concentration 1.0 nM) was mixed with various concentrations of either wild type or mutant fXa, added to the fVa/bead mixture, and incubated with mixing for an additional 10 min. 125 I-fXa S379A was utilized as the labeled fXa species for greater stability over time, i.e. to minimize fXa-dependent proteolysis. The beads were collected by centrifugation, and the pellets and supernatants were counted separately. Nonspecific binding was determined from reactions not containing fVa and subtracted from counts bound in the pellets. The percentage of bound fXa was quantitated relative to the amount bound in reactions with no unlabeled fXa. The apparent affinity (K d(app) ) represents the concentration of unlabeled fXa required to displace 50% of 125 I-fXa S379A from the fVa-bound beads and is calculated from the following equation.
In the equation, y ϭ % bound 125 I-fXa S379A , x ϭ concentration of added, unlabeled fXa, and s ϭ a slope factor. fXa Model of the fVa Binding Site-The model of fXa was generated from the published coordinates (21) using RIBBONS version 3.0, developed by M. Carson at the University of Alabama (Birmingham, AL). In the structure, the Gla domain is missing and EGF-1 is disordered.

RESULTS
Mutagenesis-We demonstrated previously that the employed expression system is efficacious for the production of fully functional recombinant wild-type and variant fX (12)(13)(14). Construction of the variant molecules in the current study was based on two mutational strategies. Residues Arg-306, Lys-351, and Lys-414 were substituted with alanine, whereas Glu-310 and Arg-347 were substituted with asparagine, resulting in the creation of a potential N-linked glycosylation site, NX(C/S/ T), (Table I). The apparent molecular weight of fX E310N was elevated, indicating the addition of a carbohydrate group at this residue (Fig. 1). However, as previously described, asparagine substitution of Arg-347 did not result in an added carbohydrate, as the electrophoretic mobility of fX R347N was not altered as compared with fX WT (Fig. 1; Ref. 12).
Activation-All variants hydrolyzed a peptide substrate, Spectrozyme FXa, normally (data not shown). Initial rates of activation by RVV-X and the extrinsic (fVIIa/TF) and intrinsic (fIXa/fVIIIa) activation complexes were quantitated by monitoring Spectrozyme FXa hydrolysis. Activation of all variant zymogens by all activators was very similar to the activation of fX WT .
Inhibition-The impact of the substitutions on the active site of fXa was probed by studying the interaction between the enzymes and the physiologic inhibitors, ATIII and TFPI. All variant enzymes were inhibited normally by these inhibitors as compared with fXa WT and second order rate constants for inhibition of fXa WT were consistent with reported values (Table  II; Ref. 22). Therefore, the active site of fXa was not compromised by these mutations. Inhibition of each enzyme by ATIII was also examined in the presence of heparin pentasaccharide and full-length heparin. The second order rate constants for ATIII inhibition in the presence of full-length heparin were reduced for all variants as compared fXa WT (Table II). Rate constants for inhibition of fXa R306A and fXa E310N were modestly reduced (66% and 64% of fXa WT , respectively), whereas those for fXa R347N (12.6% of fXa WT , 12), fXa K351A (20.1%), and fXa K414A (21.1%) were markedly reduced. In contrast to these effects with full-length heparin, inhibition of all variants in the presence of the pentasaccharide was equivalent to that of fXa WT (Table II). Thus, the basic residues at positions 347, 351, Prothrombin Activation in the Absence of fVa and Phospholipid-The catalytic function of the enzymes was evaluated using the physiological substrate, prothrombin (Fig. 2). Multiple concentrations of prothrombin were incubated with the mutant enzymes and the apparent affinity for prothrombin and catalytic turnover were compared with that of fXa WT  Prothrombin Activation in the Presence of fVa Ϯ Phospholipid-To examine the interaction between the variant enzymes and cofactor, thrombin formation was first quantitated in the presence of fVa, but in the absence of phospholipid. All enzymes demonstrated markedly reduced function as compared with fXa WT (Fig. 3). fVa enhanced the catalytic function of fXa WT , e.g. at 10 M prothrombin, fXa WT was catalytically more than 1000-fold faster in the presence of 250 nM fVa. The catalytic activity of the variants was also enhanced by the addition of cofactor, although not nearly to the same extent. In an effort to increase the concentration of cofactor in the presence of the enzymes, prothrombin turnover was evaluated on 100 nm PC:PS (3:1) vesicles which bind fXa and fVa and colocalize the enzyme with cofactor. Under these conditions, all variant enzymes displayed measurable function (Fig. 4). Here, the apparent fVa affinity for fXa R306A and fXa E310N were only slightly reduced relative to fXa WT (fXa WT K d(app) ϭ 106 pM, fXa R306A ϭ 203 pM, and fXa E310N ϭ 157 pM), whereas fXa K351A and fXa K414A demonstrated a greater reduction in cofactor affinity (fXa K351A K d(app) ϭ 366 pM, and fXa K414A ϭ 556 pM). As previously reported, the relative fVa affinity of fXa R347N was markedly reduced under these conditions (fXa R347N K d(app) ϭ 2299 pM). Maximal turnover of prothrombin by all variant enzymes was found to be within 2-fold that of fXa WT (range, 18.9 -28.9 s Ϫ1 ) in the presence of cofactor and phospholipid vesicles. These data are most easily explained by the hypothesis that the major effect of substitution of Arg-347, Lys-351, or Lys-414 is the reduction in fVa affinity in the presence of PC:PS vesicles with little or no impact on catalytic activity of the fVa-fXa complex.
If the mutations only impact cofactor affinity, the full catalytic potential should be realized under ideal conditions. This hypothesis was evaluated using activated platelets, which provide an ideal phospholipid surface and saturating fVa. Indeed, under these conditions, all fXa species are very similar to fXa WT (Fig. 5).
Competition Binding Assay-To confirm the results of the functional studies and more directly probe the role of these residues in fVa binding, we employed a previously described binding assay (12). In the assay, active-site modified, radiolabeled fXa ( 125 I-fXa S379A ) was mixed with varying concentra-   . Aliquots were removed from the reactions for initial rate measurements, diluted 10-fold in EDTA buffer, and incubated with 500 M Spectrozyme TH. The initial rate of hydrolysis was monitored at 405 nm. The concentration of thrombin was determined from a standard curve prepared from dilutions of maximally activated prothrombin. The rate of thrombin formation was calculated for the indicated prothrombin concentrations and expressed as moles of thrombin (IIa)/mole of Xa/s. The maximal rate of thrombin formation is expressed as k cat (s Ϫ1 ).
tions of unlabeled, mutant enzyme. Both enzymes were then allowed to compete for fVa binding on a PC:PS-coated latex bead. Factor Va binding is expressed for the variant enzymes in Fig. 6 as the percentage of 125 I-fXa S379A bound to the bead in the presence of increasing concentrations of added, unlabeled fXa. fXa R306A and fXa E310N demonstrated modest reductions in fVa affinity in the assay (fXa WT K d(app) ϭ 1.83 nM, fXa R306A ϭ 2.96 nM, fXa E310N ϭ 4.03 nM), whereas the relative fVa affinity of fXa K351A , fXa K414A , and fXa R347N was markedly reduced (K d(app) ϭ 8.81, 12.17, and 19.45 nM, respectively). Consistent with the functional studies on phospholipid vesicles, substitution of residues Lys-351, Lys-414, and Arg-347 resulted in a selective loss of fVa affinity in the absence of prothrombin, whereas mutation of Arg-306 and Glu-310 reduced fVa binding to a much lower extent.
fVa Binding Site of fXa-The juxtaposition of the targeted residues is apparent from the fXa structure ( Fig. 7; Ref. 21). Residues found to contribute to the fVa binding site are clustered in a defined region on the fXa surface in the serine protease domain. The impact of surrounding residues (Arg-273, Ile-357, Arg-406, Lys-406, and Lys-420) on fVa affinity was evaluated using the fVa-sensitive functional assays described above, i.e. in the presence and absence of PC:PS vesicles. Substitution of these residues had no impact on fVa affinity in these assays (data not shown), thereby delineating the fVa binding epitope. DISCUSSION The current study describes substitutions in the surface epitope of fX that includes Arg-306, Glu-310, Arg-347, Lys-351, and Lys-414. All mutants were synthesized and secreted in the expression system in a manner equivalent to that of fX WT . Activation of these zymogens by fVIIa/TF, fIXa/fVIIIa, and RVV-X, hydrolysis of a small, peptide substrate, and inhibition by TFPI and ATIII all occurred at a near-normal rates. Based on these studies, substitution of these residues does not compromise the active site structure of fXa.
Heparin can accelerate the rate of fXa inhibition by ATIII by mediating conformational changes in the inhibitor and by acting as a template that binds both inhibitor and enzyme (23). The individual contributions of structural alterations in ATIII and binding to fXa were evaluated using heparin pentasaccharide, which binds ATIII, but does not act as a template. All variant enzymes were inhibited normally by ATIII in the presence of heparin pentasaccharide. However, the rate of inhibition by ATIII in the presence of full-length heparin was reduced for all variants, indicating that these residues contribute to the heparin binding capacity of fXa. Substitution of the glutamic acid at position 310 results in the addition of a carbohydrate group, which could sterically hinder the binding of heparin to the fXa surface. Substitution of the basic residues at positions 347, 351, 414, and 306 likely attenuates ionic interactions between heparin and fXa. These data are corroborated by structural analysis of thrombin and fXa. Lys-351 and Lys-414 are adjacent to the analogous epitope defined in thrombin as the heparin binding exosite (24). Moreover, Padmanabhan and co-workers (25) have described the heparin binding domain of fX as a large basic region, which either includes or is adjacent to these residues. Delineation of the entire heparin binding domain of fXa will require additional mutagenesis.
All variant enzymes bound and activated prothrombin in a manner similar to fXa WT with one exception, fXa E310N . The apparent affinity of fXa E310N for prothrombin was 3-fold lower than fXa WT . It is conceivable that the added carbohydrate sterically hinders substrate binding. Alternatively, the added sugar may restrict conformational mobility of fXa and prevent the enzyme from adapting a favorable conformation for substrate binding in the absence of a lipid surface.
This study examined the independent interactions between fXa and other prothrombinase complex constituents. It was found that the sensitivity to aberrant fVa binding is heightened under suboptimal conditions for prothrombin turnover, i.e. in the presence of fVa Ϯ PC:PS vesicles, allowing clear delineation of the residues that compose the fVa binding site. As expected, substitutions of residues that comprise the core binding site result in a more severe phenotype as compared with mutations in the residues that form the extended regions of the site. The power of this methodology is exemplified by fXa R306A and fXa E310N , which displayed a reduced function in the absence of phospholipid. On addition of PC:PS vesicles, which increase the local concentration of prothrombinase complex components and maximize productive interactions, these enzymes bind fVa with a similar apparent affinity to fXa WT . In contrast, the phenotype of fXa R347N , fXa K351A , and fXa K414A was only overcome under the ideal conditions of activated platelets, suggesting a more critical role of these residues in fVa binding. Therefore, Arg-306 and Glu-310 may form part of the binding site that extends beyond the core epitope whereas Arg-347, Lys-351, and Lys-414 likely contribute to the core of the binding epitope.
It has been proposed that fVa mediates conformational changes in fXa that contribute to the cofactor-driven enhancement of prothrombin activation (11, 26 -28). In the absence of a phospholipid surface, cofactor enhanced the activity of fXa WT , but had little effect on the variant enzymes. There are at least two possible explanations for these data. Either the mutations compromised the functional capacity of fXa by attenuating these proposed cofactor-mediated structural rearrangements, or fVa binding is impaired by the mutations. If these structural rearrangements are attenuated by the substitutions, the aberrant phenotype would not likely be overcome even under optimal conditions for substrate turnover. It was demonstrated that mutant enzymes with less severe phenotypes (fXa E310N and fXa R306A ) can function nearly normally on a PC:PS vesicle surface in the presence of fVa and the function of all mutant enzymes was similar to the wild-type enzyme under the conditions provided by activated platelets. Moreover, the competition binding assay, which directly probes cofactor binding, demonstrated a more marked loss of fVa affinity for variant enzymes with more severe phenotypes in the functional assays. The data do not support the postulate that substitution of these , fXa K351A (f), or fXa K414A (q), and the reaction was incubated for an additional 10 min. The beads were collected by centrifugation, and the supernatant and pellet were counted separately. The fVa-specific binding (calcium-, phosphatidylserine-, and fVa-dependent) was quantitated as the percentage of bound 125 I-fXa S379A . The concentration of unlabeled fXa required to displace 50% of bound, 125 I-fXa S379A is represented as the apparent affinity (K d(app) ). Nonspecific binding has been subtracted from all values shown. residues prevented fVa-mediated changes in the fXa structure, but rather that the fVa affinity is selectively attenuated by the mutations.
Analysis of the tertiary structure of fXa provides compelling structural support for the existence of a fVa binding epitope in the serine protease domain of fXa. Using peptide mapping studies, Chattopadhyay et al. (30) identified three regions of fXa that were found to contribute to fVa binding: 211-222, 254 -269, and 263-274 (30). Each of these regions was evaluated by mutagenesis using functional assays and found not to contribute to fVa binding (data not shown). The discrepancy between the studies may reflect differences in methodologies. The region identified by the current study has also been described as a cofactor binding site for other coagulation enzymes. For example, the analogous surface loop of fIXa, which corresponds to residues 344 -352 in fXa, was recently identified to form a binding epitope for fVIIIa (31,32). Moreover, Banner et al. (29) identified multiple regions of the fVIIa surface that contribute to TF binding including one region analogous to fX residues 346 -349.
Targeted mutagenesis of the residues that surround the proposed cofactor binding epitope was also conducted as part of this study. It was found that residues Arg-273, Ile-357, Arg-406, Lys-406, and Lys-420 do not contribute to fVa binding, but rather outline the fVa binding epitope. Based on this survey of adjacent epitopes, structural examination, and comparisons to other cofactor binding epitopes, it is likely that Arg-347, Lys-351, and Lys-414 form the core of a fVa binding epitope of fXa, which is extended to include Arg-306 and Glu-310.