Active Site Modification of Factor VIIa Affects Interactions of the Protease Domain with Tissue Factor*

In the initiation of coagulation, tissue factor (TF) allosterically activates the serine protease factor VIIa (VIIa) through specific interactions with protease domain residues. These interactions, and consequently affinity for TF, may be influenced by conformational changes in the protease domain that result from zymogen-enzyme transition or occupancy of the active site by tight binding inhibitors. In functional competition and direct binding analysis, we determined affinities for zymogen and enzyme species of wild-type VII and of mutants at protease domain residues that contact TF. We demonstrate that TF binding is not influenced by zymogen activation, indicating that the protease domain of zymogen and enzyme dock similarly with TF. In contrast, active site occupancy enhanced the affinity for TF by predominantly decreasing the dissociation rate of the TF·VIIa complex. Of the three interface residues studied, only Met306 played a major role in the inhibitor-induced increase in affinity. Met306 is also important for transmitting the allosteric changes from TF to the active site, resulting in enhanced catalysis. This study thus provides evidence for a bidirectional conformational interdependence of the interface residue Met306 and the active site of VIIa.

In the initiation of coagulation, tissue factor (TF) allosterically activates the serine protease factor VIIa (VIIa) through specific interactions with protease domain residues. These interactions, and consequently affinity for TF, may be influenced by conformational changes in the protease domain that result from zymogen-enzyme transition or occupancy of the active site by tight binding inhibitors. In functional competition and direct binding analysis, we determined affinities for zymogen and enzyme species of wild-type VII and of mutants at protease domain residues that contact TF. We demonstrate that TF binding is not influenced by zymogen activation, indicating that the protease domain of zymogen and enzyme dock similarly with TF. In contrast, active site occupancy enhanced the affinity for TF by predominantly decreasing the dissociation rate of the TF⅐VIIa complex. Of the three interface residues studied, only Met 306 played a major role in the inhibitorinduced increase in affinity. Met 306 is also important for transmitting the allosteric changes from TF to the active site, resulting in enhanced catalysis. This study thus provides evidence for a bidirectional conformational interdependence of the interface residue Met 306 and the active site of VIIa.
Factor VII (VII) 1 binding to tissue factor (TF) at the cell surface is the major initiating step in the blood coagulation cascade (for a review, see Ref. 1). Besides its function in cell surface localization and acceleration of the activation of VII, TF is the cofactor for VIIa, enhancing proteolytic and amidolytic activity of the bound protease (2). Recently, the crystal structure of a complex of soluble TF and VIIa inhibited with Phe-Phe-Arg chloromethylketone (FFR-VIIa) demonstrated that each domain of VIIa makes contact with TF (3). The ␥-carboxyglutamic acid-rich domain of VIIa docks with the carboxylterminal fibronectin type III domain of TF, the first epidermal growth factor-like domain of VIIa contacts both domains of TF, and the second epidermal growth factor-like domain and pro-tease domain form a continuous surface that sits upon the amino-terminal domain of TF. TF contacts with the protease domain are distant from the active site and also from the amino-terminal insertion by which residue Ile 153 (c16) 2 forms a salt-bridge with Asp 343 (c194) that is critical for enzymatic activity of trypsin-like serine proteases. Because of the solvent accessibility of the insertion site, proteolytic activation of zymogen VII may proceed even if VII is docked to TF in a manner identical to the enzyme VIIa.
However, some studies reported a difference in the affinity for TF when zymogen VII was compared with the active protease (4 -6). Higashi et al. (7) furthermore proposed that binding epitopes in bovine zymogen VII are not accessible for binding to TF, unless VII is activated. We hypothesized that conformational consequences of zymogen activation likely affect interactions of the protease domain of VIIa with TF. By alanine scanning mutagenesis, we had previously identified residues in the VIIa protease domain that are involved in binding TF (8). By comparing the binding characteristics of zymogen and enzyme forms of mutants at residue positions that contact TF, we here address the question of whether specific residue side chains in the interface with TF play different roles in the interaction of zymogen as compared with enzyme.
The binding residues in the interface with TF also transmit the allosteric changes that result in the catalytic activation of VIIa (8). Moreover, a reverse effect of active site occupancy and TF binding exists, as demonstrated by the increased TF affinity upon modification of the active site of VIIa with tight binding inhibitors (7) or upon binding of substrates (9,10). We hypothesized that these conformational effects transmit to specific binding residues in the interface with TF. By characterizing the effect of active site modification of replacement mutants in the interface with TF, we identify one residue side chain that is influenced by inhibitor occupancy of the active site of VIIa. This analysis thus provides insight into the molecular details of the conformational changes in the protease domain of VIIa that affect TF interactions.

MATERIALS AND METHODS
Proteins-Full-length recombinant human TF from insect cells was reconstituted in 30% phosphatidylserine, 70% phosphatidylcholine as described (11). Plasma factor X was purified according to Fair et al. (12) followed by an immunoaffinity step to remove residual VII (8). The Kunitz-type inhibitor 5L15, which was selected from bovine pancreatic trypsin inhibitor by random mutagenesis and phage display for affinity for VIIa (13), was kindly provided by Dr. George Vlasuk (Corvas International, San Diego, CA).
Recombinant VII-The generation of VII mutants in the interface of the protease domain with TF has been described (8). Protein was purified from serum-free (Excel 301 supplemented with vitamin K 3 ) * This work was supported by National Institutes of Health Grant R01 HL-48752 and was performed during the tenure of an Established Investigator Award of the American Heart Association (to W.R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "adver- culture supernatant by antibody affinity on a Ca 2ϩ -dependent antibody to the ␥-carboxyglutamic acid-rich domain, as described (11). The protein obtained from this purification was predominantly zymogen VII, and part of the protein preparation was set aside for characterization of the zymogen forms. The remaining protein was dialyzed against 50 mM NaCl, 0.25 mM CaCl 2 , 20 mM Tris/HCl, pH 8.5, for further purification on Mono Q employing a CaCl 2 gradient elution (0.25-100 mM) in the same buffer. This latter step resulted in Ͼ80% activation to the active enzyme. The degree of ␥-carboxylation in the final preparation was determined by amino acid analysis after alkaline hydrolysis (14). These analyses were performed by Dr. Robert Harris (Commonwealth Biotechnologies, Richmond, VA). M306A VIIa had 10.8 ␥-carboxylated Glu residues, R277A VIIa 9.3, D309A VIIa 9.9, and wild-type VIIa 9.9.
Using oligonucleotide-directed mutagenesis with selection against a unique restriction site (15), we further generated a mutant with increased stability as zymogen by replacing the P1 (16) Arg with Gln (R152Q VII) and a mutant that lacked catalytic activity by Ala replacement of the active site Ser (S344A VII). The latter mutant also incorporated two replacements that precluded N-linked glycosylation (N145A and N322D). Elimination of N-linked glycosylation by these mutations in wild-type VIIa affected neither proteolytic function nor TF binding (data not shown). S344A VII was converted to S344A VIIa by incubation with factor IXa at a 1:10 enzyme:substrate ratio in 50 mM NaCl, 0.25 mM CaCl 2 , 20 mM Tris/HCl, pH 8.5, overnight at 37°C followed by purification on Mono Q employing a NaCl (20 -500 mM) gradient elution in 20 mM Tris/HCl, pH 8.5.
To test the effect of active site occupancy, mutants and wild-type VIIa were reacted with 0.5 mM dansyl-Glu-Gly-Arg chloromethylketone or D-Phe-L-Phe-Arg chloromethylketone (Calbiochem) in HEPES-buffered saline (HBS), 5 mM CaCl 2 for 2-4 h. Subsequently, protein samples were extensively dialyzed against HBS with six buffer exchanges over a 24-h period to remove unreacted peptide inhibitor.
Competitive Functional Assay to Determine Affinity for TF-In a linked functional assay (8), we measured the ability of catalytically inactive VIIa derivatives or nonactivable VII mutant to compete with active VIIa for TF binding. In an initial incubation of 20 -60 min at 37°C, 20 pM wild-type VIIa and varying concentrations of competitor were allowed to bind to a limiting concentration of 10 pM phospholipidreconstituted TF in HBS (150 mM NaCl, 15 mM HEPES, pH 7.4), 5 mM CaCl 2 , 0.2% bovine serum albumin. Factor X was added to 100 nM to measure catalytically active TF⅐VIIa complexes for 10 min at 37°C, followed by quenching with 50 mM EDTA and determination of factor Xa formation by chromogenic substrate (Spectrozyme FXa, American Diagnostica, Greenwich, CT) hydrolysis at ambient temperature. Data were fitted by a nonlinear least squares algorithm to the equilibrium binding equation assuming competition of VIIa and derivatives to the same binding site. The K D of active VIIa was 5 pM, as determined from a dose titration in the functional assay (8).
Competition was also analyzed in an amidolytic assay with recombinant soluble TF-(1-218) expressed in Escherichia coli (17). Dansyl-Glu-Gly-Arg chloromethylketone-inactivated wild-type or mutant VIIa at varying concentrations was incubated with 10 nM VIIa and 10 nM TF-(1-218) in HBS, 5 mM CaCl 2 , 0.2% bovine serum albumin for 1 h at 37°C to allow for complex formation. Chromozyme tPA (Boehringer Mannheim) was added to 1 mM, and initial rates of substrate hydrolysis were recorded in a kinetic microplate reader. Data were fitted to a competitive model of ligand binding, as above, assuming a K D (app) of 2.8 nM for active VIIa binding to soluble TF-(1-218) (8).
Surface Plasmon Resonance Analysis-Rate and equilibrium binding constants of VII and VIIa binding to TF were obtained by surface plasmon resonance using a BIAcore 2000, as described previously (8). A noninhibitory anti-TF monoclonal antibody (TF9 -10H10) was immobilized by amine coupling to the N-hydroxysuccinimide ester surface to give approximately 1000 response units. The standard flow buffer was HBS, 5 mM CaCl 2 , 1 mM CHAPS, and all proteins were diluted in this buffer. Full-length recombinant TF-(1-263) was injected to saturate the monoclonal antibody on the sensor surface. Association data were collected from 40-l injections of varying concentrations of VII or VIIa at a flow rate of 10 l/min. Dissociation data were collected for 250 s after return to buffer flow. Injections of 100 mM EDTA served to regenerate the chip surface by destabilizing the TF/VIIa interaction. Association rate constants (k on ) were determined from a linear fit of the concentration dependence of k S , while the dissociation rate constants (k off ) were determined from the early portion of the dissociation phase following injection of 1 M VII or VIIa. Binding kinetics for wild-type and mutant VIIa in the presence of the Kunitz-type inhibitor 5L15 were determined by premixing VIIa with 10 M 5L15. At this concentration, the inhibitor blocked amidolytic function of free VIIa, indi-cating complex formation in the absence of TF.
Western Blot Analysis of Proteins Recovered from the Sensor Chip-Because zymogen VII is readily activated to VIIa when bound to TF (18), we wanted to characterize the protein that was bound specifically to the immobilized TF. After allowing binding and return to buffer flow, 50 mM EDTA was injected to recover the specifically bound VII/VIIa in a single fraction. The recovered protein was reduced, separated by SDS-polyacrylamide gel electrophoresis, and transferred to Immobilon P (Millipore Corp.) for Western blotting. The VIIa light chain was reacted with monoclonal antibody F5-13B12, and bound primary antibody was detected by horseradish peroxidase-conjugated secondary antibody and chemiluminescence.

Comparison of VII and VIIa in a Competitive Functional
Assay-Recent data indicated that bovine zymogen VII bound with reduced affinity to TF as compared with active enzyme (7). Since the described difference may be attributable to the different assay formats used for determining TF binding by enzyme VIIa (direct amidolytic assay) versus zymogen VII (competitive assay), we produced a catalytically inactive mutant VIIa by active site Ser 344 3 Ala replacement. S344A VIIa was generated by activation of the zymogen form by factor IXa followed by repurification using ion exchange chromatography. We also generated a zymogen VII (R152Q VII) with increased stability by replacing the P1 residue Arg 152 with Gln to essentially abolish activation during purification and in functional assays. These mutants allowed us to compare zymogen and enzyme forms of human VII in a competitive functional assay. Zymogen and enzyme forms showed virtually identical dose titrations in competition with wild-type VIIa (Fig. 1). The calculated affinities for zymogen and enzyme were indistinguishable, providing evidence that the affinity of VII was not appreciably altered following zymogen activation to enzyme. The affinities also matched the functionally determined K D app (5 pM) (8) for VIIa binding to TF in the same assay format. In contrast, wild-type VIIa inactivated with FFR-VIIa showed a markedly enhanced inhibition and thus an increased affinity for TF, in accordance with previous studies analyzing active site-modified VIIa (7).
Comparison of Zymogen and Enzyme Forms by Surface Plasmon Resonance Binding Analysis-A limitation of the competitive assay approach is the presence of substrate that could differentially affect TF binding by the VII variants. We therefore used surface plasmon resonance analysis to directly compare TF binding by VII and VIIa in the absence of substrate. TF-(1-263) solubilized with detergent was captured to the chip surface by a noninhibitory monoclonal antibody. The saturation curves and data analysis for a typical experiment with zymogen VII and enzyme VIIa are shown in Fig. 2, demonstrating excellent data fit. Similar binding parameters were obtained for zymogen forms S344A VII and R152Q VII, providing evidence that the Arg 152 mutation does not influence TF interaction (Table I). Furthermore, similar kinetic parameters were obtained for wild-type VIIa and activated S344A VIIa mutant (Table I), indicating that replacement of the catalytic triad Ser is without effect on cofactor affinity. From these data we conclude that zymogen activation has little, if any, effect on the kinetics of VII binding to TF and that overall affinity is not influenced by zymogen activation.
Although overall binding of zymogen and enzyme appears to be similar, specific side chains in the protease domain interface with TF may have different roles in providing affinity for TF when zymogen and enzyme are compared. We had identified three residue side chains with significant contributions to TF binding when analyzed as enzyme VIIa (8). Comparison of the binding kinetics of enzyme and zymogen forms of mutants at Arg 277 (c134), Met 306 (c164), and Asp 309 (c167) revealed no difference in the binding parameters (Table I). To ensure that the zymogen forms were not activated during the surface plasmon resonance analysis, samples were recovered after binding to TF on the sensor chip. Western blot analysis (Fig. 3) of these samples demonstrates a degree of activation comparable with the zymogen:enzyme ratios in the preparation prior to injection onto the sensor chip. We thus exclude the possibility that zymogen is appreciably activated during analysis and conclude that the surface plasmon resonance analysis truly measures affinity of VIIa or VII, respectively. Thus, the affinity provided by the three most significant binding contacts of the protease domain with TF was not altered by zymogen activation, indicating that zymogen and enzyme dock similarly with TF.
Effect of Active Site Occupancy on the Affinity for TF-Previous studies (7,9,10) show that active site occupancy of VIIa influences cofactor affinity. By surface plasmon resonance, we analyzed TF binding of wild-type and mutant VIIa modified covalently with either dansyl-Glu-Gly-Arg chloromethylketone (EGR-VIIa) or FFR-VIIa. We also analyzed the effect of 5L15, a tight binding noncovalent inhibitor derived by phage maturation from bovine pancreatic trypsin inhibitor (13). When mutant or wild-type VIIa was compared before and after active site modification, the association rates were reduced 2-fold for EGR-VIIa, increased 2-3-fold for FFR-VIIa, and unchanged in the presence of the 5L15 (Table II). There was no difference in the association rates when comparing mutant with wild-type VIIa either free or with each of the specific modifications. These experiments did not identify a molecular mechanism of how inhibitor binding influences the association rate of VIIa with TF.
The binding defect resulting from mutations of protease domain residues at the TF-binding interface is attributable to an increased dissociation rate of the mutant enzymes relative to wild-type VIIa (Table II). A lower dissociation rate was a common characteristic of active site modification by inhibitors. Inhibitor binding to wild-type VIIa slowed dissociation 6-fold for FFR-VIIa, 4-fold for EGR-VIIa, and 3-fold with 5L15 present. Inhibitor binding to R277A VIIa and D309A VIIa decreased the rate of dissociation to an extent that was similar to wild-type VIIa (Table II). In contrast, the dissociation rates of M306A VIIa were not significantly affected by binding of 5L15 or by active site modification with dansyl-Glu-Gly-Arg chloromethylketone. M306A FFR-VIIa dissociation from TF decreased 3.5-fold as compared with 6-fold for FFR-VIIa, demonstrating a partial response to binding of this inhibitor. Diminished response of M306A VIIa to inhibitor binding is also demonstrated by changes in the calculated free energy of binding relative to free enzyme (Table II). With each inhibitor, modification of M306A VIIa failed to produce a ϳ0.5 kcal/mol decrease in the free energy of binding as compared with inhibited wild-type VIIa, R277A VIIa, or D309A VIIa. These data link Met 306 to conformational changes in the active site of VIIa.
Competition of M306A EGR-VIIa with Wild-type VIIa-The loss of TF binding affinity of active M306A VIIa and R277A VIIa were similar, as measured by amidolytic assay (8), linked proteolytic assay (8), or surface plasmon resonance (Table I). Both mutants exhibited a similar 0.5-0.6 kcal/mol loss in free energy of binding relative to wild-type VIIa (Table I). Following active site modification, the difference in free energy of binding was unchanged for R277A EGR-VIIa (0.6 kcal/mol) as compared with wild-type EGR-VIIa, whereas M306A EGR-VIIa showed a 1.4 kcal/mol difference in free energy of binding relative to wild-type EGR-VIIa. Since M306A did not respond to active site modification with dansyl-Glu-Gly-Arg chloromethylketone as measured by surface plasmon resonance (Table  II), we predicted that M306A EGR-VIIa should be a less potent competitor for VIIa binding to TF compared with R277A EGR-VIIa. The inhibition of active VIIa binding to soluble TF-(1-218) by M306A EGR-VIIa required significantly higher concentrations than the inhibition by R277A EGR-VIIa (Fig. 4), reflected in a 3-fold higher K i for the mutant at Met 306 . The calculated difference in the free energy of binding between mutant and wild-type EGR-VIIa are in good agreement with the data from the surface plasmon resonance experiments, demonstrating a 1.2 kcal/mol difference for M306A EGR-VIIa as compared with 0.5 kcal/mol for R277A EGR-VIIa. The competition experiment thus provides independent evidence for a prominent role of the Met 306 side chain in increasing the affinity for TF following active site inhibition. DISCUSSION This study analyzes the effect of conformational changes in the VII protease domain on binding to cofactor TF. VIIa is among the few trypsin-like serine proteases that require cofactor binding for full catalytic function. In part, the low catalytic activity of free VIIa may result from a labile activating insertion of the newly formed amino terminus after zymogen activation (6). Based on alanine scanning mutagenesis of the VIIa protease domain surface, we had found evidence that TF may function through a more extended stabilization of the "activation domain" (19), which would enhance catalysis also by influencing Asp 338 (c189), which forms the bottom of the P1 (16) specificity pocket. The latter model would suggest that VIIa may present itself in a zymogen-like state to TF and that activating conformational changes occur subsequently to docking of the protease domain with TF.
The affinity of zymogen VII and enzyme VIIa binding to TF was determined by a functional competition assay and direct binding analysis using surface plasmon resonance detection. Both experimental approaches clearly demonstrate similar binding affinities of VII and VIIa for TF. Studies performed in the bovine system by Higashi et al. (7) suggest that zymogen VII binds with lower affinity than VIIa, indicating speciesspecific properties of the bovine VII. Our results are consistent with previous results demonstrating little difference between human zymogen and enzyme binding to TF (20 -22).
The finding that VII and VIIa bind to TF with similar affinity does not exclude the possibility that TF contacts are different  for the protease domain of VIIa as opposed to VII. Three major energetically important contact residues in the protease domain contribute to TF⅐VIIa complex formation (8). We determined binding kinetics for zymogen and enzyme form of each of the mutants at position Arg 277 (c134), Met 306 (c164), and Asp 309 (c167). Since binding affinities for each mutant zymogen and enzyme form were indistinguishable, it is reasonable to conclude that the protease domain of VII or VIIa docks similarly with TF. Based on this finding, we speculate that the overall structure of the TF⅐VII complex is quite similar to the TF⅐VIIa structure (3) with the exception of the uncleaved amino terminus of the protease domain, which would be surface-exposed. In the suggested position, the scissile bond is easily accessible to activating proteases, such as TF⅐VIIa (23,24) and factor Xa (18,25), providing a rationale for the preferred activation of TF-bound zymogen VII. Consistent with earlier studies (7), we found that active site occupancy of VIIa resulted in tighter binding to TF. Each inhibitor had different effects on the association rate of VIIa. Covalent modifications with dansyl-Glu-Gly-Arg chloromethylketone decreased the association rate, whereas Phe-Phe-Arg chloromethylketone had the opposite effect. The Kunitz-type inhibitor 5L15 influenced the association rate insignificantly. The effect on the on-rate was not influenced by mutations of the three major contact residues of the protease domain with TF, and we could not identify the contacts that are responsible for these changes.
All three inhibitors also decreased the dissociation rate to a variable degree. Whereas R277A VIIa and D309A VIIa showed similar reductions in the dissociation rates, no changes or only partial changes were observed for M306A VIIa. For each of the inhibitors analyzed, Met 306 (c164) seemed to be responsible for ϳ0.5 kcal/mol of the decrease in the free energy of binding that results from inhibitor binding. These data provide support for the hypothesis of a conformational interdependence of the active site of VIIa and the interface of the protease domain with TF.
In the structure of FFR-VIIa in complex with TF, Met 306 packs in a hydrophobic pocket formed by TF residues Phe 76 , Tyr 94 , Pro 92 , and Arg 74 (3). Studies from our laboratory found a major importance (⌬⌬G 1.2 kcal/mol) for Phe 76 but little energetic contributions of Pro 92 or Tyr 94 when tested in a functional assay (26). In contrast, Kelley et al. (27) documented a major contribution (Ͼ1 kcal/mol) for Tyr 94 , when binding kinetics were determined with VIIa captured on a chip through active site modification with Glu-Gly-Arg chloromethylketone. The discrepancy in the results may be attributable to the covalent active site modification of VIIa, suggesting that Met 306 in modified VIIa interacts with Tyr 94 and that this interaction is responsible for the increased affinity of TF⅐VIIa complex formation.
We had previously shown that proteolytic and amidolytic function of M306A VIIa were concordantly reduced in comparison with wild-type VIIa, suggesting that activation of the primary specificity site is mediated by cofactor interactions at Met 306 (8). In contrast, the functional defect resulting from mutation of another interface residue, Arg 277 , was largely attributable to a defect in extended substrate recognition. The results here demonstrate a function of Met 306 in mediating an inhibitor-dependent increased affinity for TF and thus emphasize a bidirectional conformational interdependence of the interface residue Met 306 with the primary specificity site of VIIa.