INTRODUCTION

TF,¹ an integral membrane protein, is the essential protein cofactor for the plasma serine protease, FVIIa. The cell surface complex of TF and FVIIa (TF·FVIIa) is the first enzyme in the clotting cascade and is responsible for initiating blood coagulation both in normal hemostasis and in thrombotic disease (reviewed in Ref. 1). The natural substrates for the TF·FVIIa complex are the serine protease zymogens, FVII, FIX, and FX. Being closely related members of the vitamin K-dependent family of serine proteases, all three substrates are homologous to each other; each having an amino-terminal Gla domain, two epidermal growth factor-like domains, and a serine protease domain. The two epidermal growth factor-like domains of FVIIa (most particularly the first) have been implicated as being critical for TF binding, while the Gla domain and protease domain may each contribute to TF binding as well (2, 3, 4), albeit to a lesser degree.

The most notable feature of the Gla domains of vitamin K-dependent proteins is that they confer the ability to bind to membranes containing phosphatidylserine or other negatively charged phospholipids (5, 6). While the affinity of GDFVIIa for TF is reduced when compared with intact FVIIa, GDFVIIa exhibits the same degree of allosteric activation by TF that intact FVIIa does when examined by amidolytic activity (2, 7). This indicates that at saturation the protein-protein components of the interaction between TF and GDFVIIa that are required for complete enhancement of enzymatic activity are established. Despite this, the TF·GDFVIIa complex is a much poorer activator of FX than is the native TF·FVIIa complex (2, 8). Thus, how the Gla domain of FVIIa contributes to catalysis of macromolecular substrate hydrolysis is as yet unclear.

From the perspective of substrate, removal of the Gla domain of FX results in slow rates of activation by the TF·FVIIa complex,² presumably due to the loss of membrane-binding potential. However, this effect is also observed in the absence of a phospholipid surface using a complex of FVIIa and soluble TF (9). These findings indicate that the Gla domain of FX also plays a separate, and as yet only poorly defined, role in the recognition of FX by the TF·FVIIa complex other than its known ability to bind to phospholipid membranes.

Although free FVIIa has limited enzymatic activity, binding to TF causes a large, reversible increase in its catalytic activity that can be detected with tripeptidyl-amide substrates as well as macromolecular substrates (10, 11, 12, 13, 14, 15). The nature of the allosteric activation of FVIIa by TF (or of any blood coagulation serine protease by its cognate protein cofactor) is not well understood. The ability of TF to enhance the rate of cleavage of small amide and ester substrates by FVIIa indicates that it brings about changes in or near the catalytic center of FVIIa. However, TF might also contribute to the formation of an extended substrate binding site for recognition of the much larger macromolecular substrates. Indeed, factor Va is proposed to provide this function for its cognate protease, FXa, whereas thrombomodulin provides a similar function for protein C activation by thrombin (reviewed in Ref. 16).

Evidence has recently been obtained in support of the idea that TF may interact directly with macromolecular substrates. An initial study reported that chemical derivatization or alanine mutagenesis of two lysine residues (Lys¹⁶⁵ and Lys¹⁶⁶) on TF greatly reduced the ability of the resulting TF·FVIIa complex to activate FX (17). Mutations at Lys¹⁶⁵ or Lys¹⁶⁶ had additive effects, with the double mutant Lys^165,166  Ala having a greater defect in FX activation than either mutation alone. Interestingly, the mutations were without effect on the affinity of FVIIa for TF or enhancement of FVIIa amidolytic activity by TF. Because the defect in FX activation of single mutations at either site could be partially overcome by increasing the PS content of the vesicles into which TF had been reconstituted, these investigators proposed that the two lysine residues may indirectly affect FX activation via interaction with negatively charged phospholipid head groups. In a subsequent study by another group, it was observed using TF expressed on cell surfaces and TF from crude cell extracts that mutating these two residues to alanine resulted in modest changes in the apparent K_m for FX activation as well as in the k_cat for this reaction (18). Furthermore, they showed that the defect in FX activation did not require interaction between TF and phospholipid head groups, since the phenomenon was observed even when complexes of FVIIa and soluble TF were used. It was therefore proposed that Lys¹⁶⁵ and Lys¹⁶⁶ may be involved in recognizing FX as a substrate, either directly by binding to FX or indirectly by binding to FVIIa and allosterically activating the enzyme. An alanine-scanning mutagenesis study of the region surrounding these residues in the linear sequence of TF (residues 157-167) also showed that this region of TF was important for macromolecular substrate recognition (19). Recently, it has been reported that these same two residues on TF play a role in the rate of binding of complexes of FXa and tissue factor pathway inhibitor to the TF·FVIIa complex (20).

Subsequent to the mutagenesis studies cited above, the x-ray crystal structure of the isolated extracellular region of TF was reported (21, 22). The protein is a member of the cytokine/growth factor receptor family and is composed of two fibronectin type III modules joined together at an angle of roughly 125°. The putative FVIIa binding site on TF identified by alanine-scanning mutagenesis comprises side chains located on both modules and the interface region joining the two (reviewed in Ref. 23). Residues Lys¹⁶⁵ and Lys¹⁶⁶ are exposed to solvent and located near the end of module 2, close to the expected phospholipid surface. This location was confirmed in the very recently published x-ray crystal structure of the complex of FVIIa and the TF extracellular domain (24).

Based on their location in the crystal structure of TF, we hypothesized that Lys¹⁶⁵ and Lys¹⁶⁶ are positioned to interact directly with the Gla domain of FX and that this interaction is important for efficient use of FX as a substrate by the TF·FVIIa complex. We therefore predicted that mutation of Lys¹⁶⁵ and Lys¹⁶⁶ in TF should have no effect on the activation of FX from which the Gla domain has been removed (Gla domainless FX). In the present study we report evidence that TF residues Lys¹⁶⁵ and Lys¹⁶⁶ contribute to activation of intact FX but not Gla domainless FX.

EXPERIMENTAL PROCEDURES

Chromozym t-PA (N-Methylsulfonyl-D-phenylalanyl-glycyl-arginyl-4-nitroanilide acetate) was from Boehringer Mannheim; Spectrozyme FXa (MeO-CO-D-cyclohexylglycyl-glycyl-arginyl-4-nitroanilide acetate) was from American Diagnostica (Greenwich, CT); and S-2222 (N-benzoyl-isoleucyl-glutamyl-glycyl-arginyl-4-nitroanilide hydrochloride) was from Chromogenix. Poly-L-lysine, Tris base, Lubrol-PX, EDTA, Antifoam C, and Tricine were from Sigma. Bovine serum albumin (fraction V, fatty acid-free), Mes, Hepes, and octyl--D-glucopyranoside were from Calbiochem. PC and PS were from Avanti Polar Lipids (Pelham, AL). Nonsterile and untreated 96-well microplates were from Corning (Corning, NY).

Recombinant human TF (wild-type and mutant) was expressed in Escherichia coli and purified as originally described for soluble TF (25) with modifications for the production of membrane-anchored TF as described previously (26). The form of membrane-anchored TF used in this study has had most of the cytoplasmic domain removed and consists of amino acids 1-244 according to the numbering system used previously (27). The cytoplasmic domain of TF is dispensable for activity and is commonly omitted from recombinant TF, since it interferes with production in, and purification from, E. coli (28). The purified mutant TF proteins migrated as a single band with M_r = 30,000, as did wild-type TF when subjected to polyacrylamide gel electrophoresis (data not shown). Grossly correct folding of wild-type and mutant TF proteins was verified by examining the support of FVIIa amidolytic activity as described (29) and the binding affinity for FVIIa as described (2). TF was relipidated by the octyl--D-glucopyranoside dialysis method (30) as originally applied to TF (31). Typically, TF was incorporated into phospholipid vesicles prepared from PC or PCPS, at a molar ratio of TF to total phospholipid of about 1:8000. Blank phospholipid vesicles were prepared in the same manner but in the absence of TF. The amount of available TF on the outside of the vesicles was quantified as described (2). Other proteins were purified as described: FX from human plasma (32), FVIIa from human plasma (33), recombinant GDFX (9), and FIX from human plasma (34).

The above referenced construct encoding membrane-anchored TF (26) was used for the production of TF mutants. Residues Lys¹⁶⁵ and Lys¹⁶⁶ were mutated via PCR-based mutagenesis (35) using appropriate flanking primers and the following pairs of mutagenic primers (mutations underlined): 5-TCA AGT TCA GGA ACA GCC AAA ACA-3 and 5- TCC TGA ACT TGA AGA TTT CC-3 (Ala¹⁶⁵-Ala¹⁶⁶ mutant); 5-TCA AGT TCA GGA ACA GCC AAA ACA-3 and 5-GT TCC TGA ACT TGA AGA TTT CC-3 (Glu¹⁶⁵-Glu¹⁶⁶ mutant); 5-TCA AGT TCA GGA ACA GCC AAA ACA-3 and 5-T TCC TGA ACT TGA AGA TTT CC-3 (Gln¹⁶⁵-Gln¹⁶⁶ mutant). PCRs were hot started with Ampliwax (Perkin-Elmer) and amplified for 25 cycles using 30 s at 95 °C, 1 min at 50 °C, and 30 s at 72 °C per cycle. The coding sequences of all TF cDNA constructs were verified by DNA sequencing.

The construct for expressing GDFVIIa was made from the wild-type FVII cDNA (ATCC 59790; American Type Culture Collection, Rockville, MD) by PCR-based deletion mutagenesis using the following primer pair to remove the 4-carboxyglutamate-rich domain: 5-ACG AGG CCT -3 and 5-CC AGC TAG CGA ATT CTA -3. The former (forward) primer codes for the N-terminal region of GDFVIIa starting at Phe⁴⁰ (underlined) in the wild-type FVII sequence. This primer also contains a StuI restriction site at the 5 end for cloning purposes. It should be noted that inclusion of this site introduces a Pro (boldface type) located between the antibody epitope tag of the construct (see below) and mature GDFVIIa. The latter (reverse) primer codes for the last 6 amino acids of FVII (underlined), a stop codon, and restriction sites (EcoRI and NheI) for cloning purposes. The PCR was hot started and amplified for 35 cycles using 1 min at 95 °C, 1 min at 50 °C, and 1.5 min at 72 °C. All ramp times were set to 0.5 min. The PCR product was digested with StuI and NheI and subsequently ligated between the StuI and XbaI sites of the RSV-PL4 expression vector (36), thus placing a 12-amino acid antibody epitope tag at the N-terminus of GDFVIIa for purification purposes.

The construct for GDFVIIa was confirmed by DNA sequencing before being transfected into human 293 cells (ATCC CRL 1573) by the lipofection technique (37) (Life Technologies, Inc.). Stable clones were selected using G418 (Geneticin, 400 µg/ml active drug; Life Technologies, Inc.), and high secretors were subsequently identified by sandwich enzyme-linked immunosorbent assay using the antiepitope tag monoclonal antibody and biotinylated rabbit anti-human FVII polyclonal antibody (prepared from Assera VII; Diagnostica Stago). Recombinant GDFVIIa was purified in the two-chain form from roller bottle cell supernatants by immunoaffinity chromatography using the Ca²⁺-dependent monoclonal antibody HPC4 (which recognizes the 12-amino acid epitope tag positioned at the N-terminus of GDFVIIa) as described previously for human FX variants (9). GDFVIIa showed normal amidolytic activity when compared with recombinant wild-type FVIIa (not shown).

Activation of FX or GDFX by the various TF·FVIIa complexes was quantified using a two-stage discontinuous chromogenic assay essentially as described previously (38) but with some modification. Briefly, in the first stage, TF in phospholipid vesicles was incubated with FVIIa or GDFVIIa in HBSA (20 mM Hepes-NaOH, pH 7.5, 100 mM NaCl, 5 mM CaCl₂, 0.1% bovine serum albumin) for 5 min at 37 °C. Prewarmed substrate solution (FX or GDFX) was then added to the mixture to initiate the reaction, after which 20-µl samples were removed at varying times into 25 µl of stop buffer (20 mM Mes-NaOH, pH 5.8, 20 mM EDTA, 8% Lubrol-PX, 0.01% Antifoam C) and maintained on ice. In the second stage, the pH of the quenched samples was adjusted to 8.3 by a 10-µl addition of 2 M Tricine, pH 8.4, and the FXa present in each sample was subsequently determined at ambient temperature by adding 50 µl of Spectrozyme FXa to yield a final substrate concentration of 0.48 mM. The change in absorbance at 405 nm was monitored over 10 min using a V_max microplate reader (Molecular Devices Corp., Menlo Park, CA), and initial rates were determined (mOD min¹). These values were then converted to nM FXa by comparison to a standard line prepared with purified human FXa. For assays in solution, rates of FX or GDFX activation were measured as above, except that TF was not reconstituted into phospholipid vesicles, and all solutions contained 0.1% Triton.

Activation of FIX by the TF·FVIIa complex was assayed as described (38) with the following modifications. In the first stage, 1 nM TF relipidated in 10-14 µM PCPS was incubated with 50 pM FVIIa in HBSA at 37 °C for 5 min, after which FIX was added to a final concentration of 150 nM. Timed samples (10 µl) were added to 40 µl of stop buffer (0.5 mM Tris-HCl, pH 7.0, 1 mM EDTA, 0.2% Lubrol PX, 0.1% bovine serum albumin, 0.01% Antifoam C) and maintained on ice. In the second stage, 10 µl of each quenched sample was added to a separate microplate well containing 40 µl FX (150 nM) in 20 mM Hepes-NaOH, pH 7.5, 57.5 mM NaCl, 5 mM CaCl₂, 0.1% bovine serum albumin. FXa generation and hydrolysis of chromogenic substrate by FXa were started simultaneously by adding 50 µl of a mixture of S-2222 and poly-L-lysine to final concentrations of 0.5 mM and 10 nM, respectively. The absorbance was monitored over 40 min at both 405 and 490 nm to correct for scatter due to the high level of additives in the reaction. The resulting absorbance difference profile was then used to calculate initial rates of FIXa generation as described previously (38). Control experiments indicated that FVIIa carried over from the first stage had negligible activity in the second stage of the assay.

Effective concentrations of enzyme-cofactor complexes given in the figure legends were calculated using published K_d values for the various complexes (2, 39, 40) and the known experimental reactant concentrations according to Equation 8 of Ref. 41.

RESULTS

It has previously been reported that mutation of two adjacent Lys residues to Ala at positions 165 and 166 of TF greatly decreased the ability of the resulting TF·FVIIa complexes to activate FX (17). However, these mutations did not decrease the affinity for FVIIa, and the mutants showed normal enhancement of FVIIa amidolytic activity. In the present study we chose to mutate these two residues to Ala or Glu in order to examine the effects of truncating these side chains or reversing their charge. In addition, Gln substitutions were included as an approximately isosteric control for the Glu mutations without charge reversal. Because previous studies indicated that the effects of mutating Lys¹⁶⁵ and Lys¹⁶⁶ separately were additive in the double mutant, we chose only to explore double mutants at positions 165 and 166 in the present study. As shown in Table I, all three TF mutants exhibited wild-type levels of cofactor function toward the amidolytic activity of FVIIa. The mutants also exhibited affinities for FVIIa that were at least as strong as the affinity of wild-type TF for FVIIa (Table I).

Table I. Amidolytic activity and FVIIa binding affinity of wild-type and mutant TF TF Amidolytic activitya Kd(app) mOD min1 % nM Lys165-Lys166 (wild type) 19.3  ± 0.8 100 0.15 Ala165-Ala166 20.2  ± 0.9 105  ± 5 0.15 Glu165-Glu166 19.2  ± 0.7 99  ± 4 0.076 Gln165-Gln166 19.6  ± 0.5 101  ± 3 0.093 a  Mean amidolytic activity (±S.E.; n = 3) expressed as mOD min1 and percentage of the mean wild-type activity.

As shown in Fig. 1, mutating Lys¹⁶⁵-Lys¹⁶⁶ to Ala¹⁶⁵-Ala¹⁶⁶ severely impaired FX activation by the TF·FVIIa complex. Furthermore, when contrasted to the Ala¹⁶⁵-Ala¹⁶⁶ mutations, the charge reversal mutations (Glu¹⁶⁵-Glu¹⁶⁶) impaired FX activation more severely when TF was relipidated in PCPS, whereas the relatively conservative mutations (Gln¹⁶⁵-Gln¹⁶⁶) impaired FX activation less severely. The impairments in FX activation were not dependent on PS in the vesicles, since the mutants showed even greater defects in FX activation when TF was relipidated in pure PC vesicles. In addition, the defects remained apparent even when the reactions were done in the complete absence of a phospholipid surface. Examination of the kinetic parameters for activation of FX by the FVIIa·TF complex (Table II) indicated that mutating TF residues Lys¹⁶⁵ and Lys¹⁶⁶ decreased k_cat by 2.0-6.6-fold, and increased K_m by 10-20-fold, yielding a reduction in catalytic efficiency of 21-94-fold.

Table II. Summary of kinetic constants for activation of FX Experiments were done using 1 nM FVIIa and 10 pM TF relipidated in 81-90 nM PCPS vesicles. The FX concentration was varied in the reaction mixtures from 0 to 1 µM for wild-type TF and from 0 to 10 µM for mutant TFs. Values of kcat and Km are means (± S.E.; n  2) and were obtained by nonlinear least squares fitting of initial rate versus substrate curves with the Michaelis-Menten equation. TF kcat Km kcat/Km min1 nM M1s1 Lys165-Lys166 (wild type) 139  ± 6 52  ± 11 4.5  × 107 Ala165-Ala166 34  ± 1 1056  ± 132 5.4  × 105 Glu165-Glu166 21  ± 1 735  ± 120 4.8  × 105 Gln165-Gln166 69  ± 2 544  ± 77 2.1  × 106

As with FX activation, mutating Lys¹⁶⁵-Lys¹⁶⁶ to Ala¹⁶⁵-Ala¹⁶⁶ impaired FIX activation (Fig. 2). The charge reversal mutations (Glu¹⁶⁵-Glu¹⁶⁶) also impaired FIX activation more severely, while the more conservative mutations (Gln¹⁶⁵-Gln¹⁶⁶) impaired FIX activation less severely. However, the relative defect of these TF mutants for supporting FIX activation (14-26% of wild type) was significantly smaller in magnitude than for supporting FX activation (3.9-17% of wild type).

We hypothesized that residues Lys¹⁶⁵ and Lys¹⁶⁶ of TF can interact directly with the Gla domain of FX. A strong prediction of this hypothesis is that mutating these residues should be without effect on the activation of GDFX by the TF·FVIIa complex. As seen in Fig. 3, the TF mutants supported essentially the same initial rates of GDFX activation as wild-type TF. This was true whether TF was reconstituted into phospholipid vesicles or solubilized in detergent.

To examine the potential role of the Gla domain of FVIIa in this scenario, FX activation was measured using GDFVIIa as the enzyme, in complex with either wild-type TF or mutant TF. The results (Fig. 4) demonstrated that all three mutants were defective in supporting FX activation under these conditions. In PCPS vesicles, the Ala¹⁶⁵-Ala¹⁶⁶ mutant had a somewhat higher activity (79% of wild type) than the other two TF mutants (50-52% of wild type). However, when TF was solubilized in detergent, the Ala¹⁶⁵-Ala¹⁶⁶ mutant showed activity that was more similar to the other mutants (38-57% of wild type). Overall, in the absence of the Gla domain of FVIIa, the TF mutants exhibited less severe defects in FX activation than when compared using intact FVIIa.

When GDFX was employed as the substrate for the TF·GDFVIIa complex (using TF reconstituted either in PCPS or PC vesicles), the three TF mutants actually supported higher activity than did wild-type TF (Fig. 5). Furthermore, under these conditions the Ala¹⁶⁵-Ala¹⁶⁶ and Glu¹⁶⁵-Glu¹⁶⁶ mutants supported higher activity than did the Gln¹⁶⁵-Gln¹⁶⁶ mutant. When TF was dissolved in detergent rather than reconstituted into phospholipid, however, the TF mutants supported activity levels similar to that of wild type. Interestingly, the TF mutants supported similar initial rates of GDFX activation whether relipidated or in solution (both were 1.4-1.5 mM/min/M), while wild-type TF supported 3-4-fold lower initial rates of GDFX activation in phospholipid vesicles (0.38-0.44 mM/min/M) compared with assays carried out with TF in solution (1.2 mM/min/M).

DISCUSSION

Mutation of Lys¹⁶⁵-Lys¹⁶⁶ in TF to Ala¹⁶⁵-Ala¹⁶⁶ substantially reduced the ability of the resulting TF·FVIIa complexes to activate FX while not affecting the affinity of TF for FVIIa or the enhancement of FVIIa amidolytic activity. This result confirms those of previous reports (17, 18, 19). In addition, these results were extended to show that charge reversal mutations of these residues (Glu¹⁶⁵-Glu¹⁶⁶) caused an even greater defect in FX activation using TF relipidated in PCPS vesicles, whereas more conservative substitutions (Gln¹⁶⁵-Gln¹⁶⁶) generally supported somewhat higher levels of FX activation than observed with the Ala¹⁶⁵-Ala¹⁶⁶ mutant. As with the Ala¹⁶⁵-Ala¹⁶⁶ mutant, the Glu¹⁶⁵-Glu¹⁶⁶ and Gln¹⁶⁵-Gln¹⁶⁶ mutants had normal affinity for FVIIa and showed normal enhancement of FVIIa amidolytic activity. The results also support the idea of a general defect in macromolecular substrate recognition when these residues are mutated, since FIX activation was also severely impaired by mutating Lys¹⁶⁵ and Lys¹⁶⁶. However, the defect in FIX activation exhibited by these mutants was notably less severe than that observed for FX activation. Therefore, residues Lys¹⁶⁵ and Lys¹⁶⁶ of TF appear to be more important in enhancing FX activation than FIX activation. Subtle, indirect effects of these mutations on TF structure that affect allosteric activation of FVIIa specifically for macromolecular substrates (but not small peptidyl substrates) cannot be ruled out but seem unlikely.

The effect on FX activation of mutating TF residues 165 and 166 was found to be largely due to an increased K_m for FX, although effects on k_cat were also observed. Whereas this supports the previous suggestion of a general effect of mutating Lys¹⁶⁵-Lys¹⁶⁶ to Ala¹⁶⁵-Ala¹⁶⁶ on FX activation (17, 18), the effects on K_m observed here with the Ala¹⁶⁵-Ala¹⁶⁶ mutations are more severe and result in a 10-fold larger overall effect on the catalytic efficiency of FX activation than previously observed: 83-fold in this study versus 8-fold (18). However, this discrepancy may reflect the use of crude detergent lysates of whole cells as the source of TF and the generally high standard errors (25-52%) previously reported for determination of the kinetic parameters of FX activation with this mutant (18). Examination of k_cat/K_m for the three TF mutants examined here revealed that the Glu¹⁶⁵-Glu¹⁶⁶ mutations produced the greatest overall defect (94-fold decrease in catalytic efficiency), whereas the Gln¹⁶⁵-Gln¹⁶⁶ mutations produced the smallest overall defect (21-fold decrease), consistent with the initial rate data.

The generally lower activation rates observed in this study with GDFX versus FX are consistent with our previous studies using GDFX (9). Normalized rates of activation of FX and GDFX by the various enzyme complexes using wild-type TF are summarized in Table III. In all but the two cases using FX and TF/PCPS, substrate levels were well below the estimated K_m (substrate/K_m  0.1). Thus, under these conditions, and for purposes of this discussion, these activation rates can be considered roughly equivalent to k_cat/K_m. The Gla domain of FXa has been proposed to interact with phospholipid as well as with factor Va (42). Thus, the reduced activation rates of GDFX using FVIIa and TF/PCPS are probably largely reflective of a reduced capacity of GDFX to bind to phospholipid. However, in the absence of phospholipid (reactions using TF/Triton), removal of the FX Gla domain still resulted in a roughly 30-fold reduction in the rate of activation by the TF·FVIIa complex. This is consistent with the previously suggested importance of the FX Gla domain for efficient utilization of FX as a substrate by the soluble TF·FVIIa complex (9). Because Lys¹⁶⁵ and Lys¹⁶⁶ are solvent-exposed in the x-ray crystal structure of TF and are located in a region predicted to lie near the phospholipid surface in intact TF (21, 22, 24), we hypothesized that these residues on TF may be ideally positioned to form direct contacts with the Gla domain of FX. Consistent with our prediction, all three TF double mutants of these residues supported levels of GDFX activation that were indistinguishable from wild-type TF.

Table III. Summary of activation rates obtained with various enzyme-substrate pairs using wild-type TF Enzyme Substrate Activation Ratea M/min/M FVIIa · TF/PCPS FX 1.2 (±0.1)  × 102 FVIIa · TF/PC FX 1.5 (±0.1)  × 101 FVIIa · TF/Triton FX 1.6 (±0.2)  × 101 FVIIa · TF/PCPS GDFX 1.6 (±0.1)  × 103 FVIIa · TF/Triton GDFX 4.8 (±0.4)  × 103 GDFVIIa · TF/PCPS FX 9.7 (±0.03)  × 102 GDFVIIa · TF/Triton FX 2.1 (±0.07)  × 103 GDFVIIa · TF/PCPS GDFX 4.4 (+0.9)  × 104 GDFVIIa · TF/PC GDFX 3.8 (±0.08)  × 104 GDFVIIa · TF/Triton GDFX 1.2 (±0.2)  × 103 a  Activation rates are normalized to concentration of enzyme complex to yield apparent first-order rates of activity (M/min/M). Individual reaction conditions are described in the corresponding figure legends.

The Gla domain of FVIIa was previously shown not to contribute to FVIIa amidolytic activity, either in the presence or absence of TF (2). Furthermore, removal of the Gla domain (and the aromatic stack region; Phe⁴⁰-Tyr⁴⁴) from FVIIa had only moderate effects on FX activation in the absence of TF (2). However, in that same study, TF·GDFVIIa complexes had a greatly reduced ability (740-fold) to activate FX as compared with TF·FVIIa complexes, consistent with earlier reports (8). This effect was also observed in the absence of phospholipid using the soluble extracellular domain of TF (2), albeit to a lesser degree (76-fold). Consistent with these earlier results, the present study showed a similar (roughly 80-fold) decrease in the rate of FX activation in solution (using TF/Triton) upon removal of the FVIIa Gla domain (but not the aromatic stack). This indicates that the Gla domain of FVIIa must be present in order for the full effect of TF cofactor function to be expressed. In addition, the effects of mutating residues 165 and 166 of TF on FX activation were much less severe when GDFVIIa was used as the enzyme, compared with intact FVIIa.

While this paper was being prepared for publication, the x-ray crystal structure of the complex of FVIIa and a cleaved form of the extracellular domain of TF was reported (24). In that structure, although the hydrophobic stack adjacent to the Gla domain of FVIIa forms hydrophobic contacts with several residues on TF, there are only minor contacts between the Gla domain itself and TF. Residues 163-165 of TF were poorly ordered, so the exact location of Lys¹⁶⁵ is not known with certainty. However, residue Lys¹⁶⁵ appears to lie near the FVIIa Gla domain and thus could conceivably form contacts with this part of FVIIa. The side chain of residue Lys¹⁶⁶ extends away from the side chain of Lys¹⁶⁵ by nearly 180° (21, 22) and therefore is located on the side of TF that is away from FVIIa (24). At a minimum, the side chain of Lys¹⁶⁶ (if not also Lys¹⁶⁵) should be readily accessible to substrate molecules such as FX.

There are several possible explanations for the results we obtained upon deletion of the FVIIa Gla domain. One intriguing possibility is that the Gla domain of FVIIa may help to align the residues 165 and 166 region of TF, thereby promoting interactions between Lys¹⁶⁶ (and possibly Lys¹⁶⁵) on TF and the Gla domain of FX. It is also possible that direct interactions between the Gla domains of FVIIa and FX are important in substrate recognition. If so, residue Lys¹⁶⁵ of TF might interact with the Gla domain of FVIIa in such a way as to promote Gla domain to Gla domain contacts by enzyme and substrate. In either scenario, removing the Gla domain of FVIIa would diminish the effects of mutating TF residues Lys¹⁶⁵ and Lys¹⁶⁶, and these effects would only be observed when the Gla domain of FX was present.

As expected based on the above results, when both GDFX and GDFVIIa were used in the activation assays, wild-type and mutant TF supported similar rates of GDFX activation when assays were carried out in solution. Surprisingly, however, the activities of the TF mutants were 2-3-fold higher than that of wild-type TF when reconstituted into phospholipid vesicles (composed either of PCPS or pure PC). The basis for the superior activity of the TF mutants under these conditions is not known. When taken together, the results presented here are consistent with the notion that Lys¹⁶⁶ (and possibly Lys¹⁶⁵) of TF interact directly with the Gla domain of FX. Furthermore, the Gla domain of FVIIa is important in this recognition process, perhaps by promoting the proper alignment of these binding regions with each other and/or by directly interacting with the FX Gla domain.

The importance of the FVIIa Gla domain in this interaction is also supported by the very similar activation rates (<2-fold difference) observed for GDFX versus FX using GDFVIIa and TF/Triton (Table III). Similarly, when GDFX is used as the substrate, removal of the FVIIa Gla domain has very little impact on activation rates (4-fold) either in the presence or absence of a phospholipid surface. Studies are currently under way to identify specific sites on the FX Gla domain that may be responsible for interaction with Lys¹⁶⁶ and/or Lys¹⁶⁵ of TF and to elucidate the molecular mechanism underlying the observed supporting role of the FVIIa Gla domain in this interaction.