(Received for publication, October 25, 1996, and in revised form, March 31, 1997)
From the Departments of Immunology and Vascular Biology, The Scripps Research Institute, La Jolla, California 92037
Epidermal growth factor-like (EGF) domain
Ca2+ binding sites in the homologous coagulation
factors VII, IX, and X stabilize the structural orientation of the
-carboxyglutamic acid-rich (Gla) domain relative to EGF-1.
Site-directed mutagenesis was employed here to analyze the functional
importance of Ca2+ binding to EGF-1 in factor VIIa (VIIa),
which initiates coagulation in complex with its cofactor, tissue factor
(TF). Ala replacements for Asp63 or Gln49
resulted in reduced TF affinity concordant with the number of eliminated Ca2+-coordinating oxygen atoms in the respective
side chains. Ca2+ binding to EGF-1 had no major direct
effect on contacts with TF residue Gln110 or on
interactions of VIIa residues Arg79 and Phe40,
suggesting that the stabilized Gla-EGF-1 orientation affects overall
docking. Gly, Ala, and Glu replacements at Asp46, which is
a Ca2+-coordinating residue at the Gla aromatic stack
carboxyl terminus, are consistent with the notion that an increased
flexibility of the Gla domain relative to EGF-1 contributes
significantly to loss of function. Certain mutants in the EGF-1
Ca2+ site had reduced proteolytic function, suggesting the
importance of the high affinity Ca2+ binding site for
macromolecular substrate interaction.
EGF1 modules function in
Ca2+-dependent extracellular protein-protein
interactions mediating cell adhesion and activation of protease
cascades. The homologous coagulation factors VII (VII), IX, and X are
characterized by one Ca2+-binding epidermal growth factor
module (EGF-1) which follows the amino-terminal -carboxyglutamic
acid (Gla)-rich domain. Functional importance of the EGF-1
Ca2+ binding site in factor IX was demonstrated by specific
mutations that cause hemophilia B (1) and by mutagenesis studies (2). Structure determinations of EGF modules of factors X (3), IX (4), and
VIIa (5) reveal Ca2+ binding through two backbone and four
side chain oxygen ligands in an octahedral coordination with one free
valence. In VIIa, Gly47 and Gln64 provide
backbone coordination, whereas Gln49 and Asp46
contribute one and Asp63 two side chain oxygens to the
coordination of Ca2+ (5). The Asp63 position of
EGF-1 is frequently modified by
-hydroxylation, which does not
affect Ca2+ affinity (6); however, plasma-derived and
recombinant VIIa do not contain erythro-
-aspartic acid at this
position (7). Ca2+ affinity for the EGF-1 modules range
from 30 to 250 µM (1, 8), but high affinity binding of
the factor X EGF-1 appears to be partially dependent on the carboxyl
terminus of the preceding Gla module (8). Structural analysis further
demonstrated that the orientation of the Gla versus the
EGF-1 module is dependent on Ca2+ binding to EGF-1 (9).
By equilibrium dialysis, two high affinity (~150 µM) Ca2+ sites were found in VIIa (10). These binding sites are also found in VIIa deleted of the amino-terminal Gla domain (des-1-38 VIIa), consistent with binding of Ca2+ to EGF-1 and to the protease domain, which has a Ca2+ binding motif (11) analogous to the trypsin catalytic domain Ca2+ site (12). Whereas fluorescence quenching indicated a ~30 µM affinity for the EGF-1 site, terbium phosphorescence measurements suggested a Ca2+ affinity of ~2 mM for the catalytic domain site (11, 13). This higher estimate for the catalytic domain site likely results from the experimental conditions and is also inconsistent with the Ca2+-dependent changes in the amidolytic function of VIIa which are attributable to saturation of the protease domain Ca2+ site with a midpoint of ~50-250 µM Ca2+ (11, 14, 15).
The interaction of VIIa with TF is Ca2+-dependent. In the absence of divalent metals, the KD of VIIa for TF is ~1.5-3 µM (10, 16). Ca2+ titration displays two transitions in the affinity of VIIa for TF: affinities are ~50-100 nM at 50-200 µM Ca2+ and ~5 nM at 1-5 mM Ca2+ (17). Ca2+ saturation of the Gla domain is likely responsible for the latter increase in affinity, since deletion of Gla in des-1-38 VIIa results in a similar loss of affinity for TF (16, 17). In part, Ca2+ saturation of the Gla domain may stabilize energetically important (14) hydrophobic contacts of Gla residues with TF (5). However, the increase in affinity for TF binding at µM Ca2+ cannot readily be explained from the structure of the TF·VIIa complex (5). The catalytic domain Ca2+ binding site is distant from the interface of the protease domain with TF. Ca2+ binding to this site may affect TF interaction only indirectly through long range conformational changes. In contrast, the EGF-1 Ca2+ site is near the VIIa light chain interface with the carboxyl-terminal module of TF, but neither the Ca2+ ion nor side chains involved in Ca2+ coordination are in contact with TF. In this study we used site-directed mutagenesis to analyze the contributions of the EGF-1 Ca2+ binding site to the interactions of VIIa with TF. We demonstrate that Ca2+ binding to this site is responsible for an increased affinity of VIIa for TF and provide evidence in support of the hypothesis that Ca2+ coordination in EGF-1 may play a functional role through the stabilization of the orientation of the Gla domain relative to EGF-1.
Soluble TF extracellular domain (TF1-218) was expressed in Escherichia coli and refolded in vitro (18). Full-length recombinant human TF from insect cells was reconstituted into 30% phosphatidylserine/70% phosphatidylcholine as described in detail (19). Plasma-derived factor X was purified according to Fair et al. (20) followed by immunoaffinity chromatography on monoclonal antibody f21-4.2 to reduce the levels of contamination by VII. After absorption to f21-4.2 and washes with 1 M NaCl, 10 mM EDTA, pH 8.0, factor X was eluted with 2 M guanidine HCl and immediately dialyzed against Tris-buffered saline (10 mM Tris, 150 mM NaCl, pH 7.4). The preparation contained < 1 pM VII/100 nM factor X.
Mutagenesis, Expression, and Purification of VIIThe VII
coding sequence in pED4 (19) was used for oligonucleotide-directed
mutagenesis using the proofreading polymerase T4 (21). We typically
sequenced across the mutated sequence including approximately 100 base
pairs upstream and downstream of the mutation to detect potential
synthesis errors in the mutagenic oligonucleotide. Mutated proteins
were transiently expressed by transfecting Chinese hamster ovary K1
cells using LipofectAMINE (Life Technologies, Inc.). Cells were
maintained in serum-free Excel 301 (JRH Scientific) supplemented with
vitamin K3 for 48 h to collect VII containing
supernatant. The supernatant was concentrated with Centricon 30 if
higher concentrations were needed to achieve saturation in the
functional assay. Stable expression and purification of VII by a
two-step procedure were as described (19). This involved affinity
purification on the Ca2+-dependent anti-VII
monoclonal antibody F4-2.1B followed by buffer exchange to the column
starting buffer (20 mM Tris/HCl, pH 8.5, 50 mM
NaCl, 0.25 mM CaCl2) for application on a Mono
Q ion exchange column. Protein was eluted with a linear
CaCl2 gradient (up to 100 mM) in loading buffer
to select further for fully -carboxylated VII. VIIAla49,
VIIAla220, and wild-type VII converted >90% to the active
enzyme during purification. Samples that were predominantly (>90%)
VII after the monoclonal affinity column were set aside for certain
experiments to characterize zymogen VII. Activation of
VIIAla63 after the two-step purification was <30%, and
this mutant was dialyzed against the column starting buffer for
activation with purified factor IXa at a 1:10 enzyme:substrate ratio
overnight at 37 °C. Activation of >95% was achieved for
VIIAla63, as judged by sodium dodecyl
sulfate-polyacrylamide gel electrophoresis under reducing conditions.
Gla content was obtained from amino acid analysis after alkaline
hydrolysis (22), and these analyses were performed by Dr. Robert Harris
(Commonwealth Biotechnologies, Richmond, VA).
In transient
transfection experiments, the concentration of wild-type or mutant VII
in the serum-free supernatant was determined by a monoclonal
antibody-based enzyme-linked immunoassay. Concentrations of purified
proteins were determined by BCA assay (Pierce). Mutants were tested for
proteolytic function and binding to TF in a previously described
functional assay (23), in which full-length TF reconstituted into mixed
phospholipid vesicles (70% phosphatidylcholine, 30% phosphatidylserine, w/w) was incubated with varying concentrations of
mutant or wild-type VII followed by the addition of factor X (100 nM) to determine complex formation at 37 °C. Factor Xa formation was measured by hydrolysis of chromogenic substrate, Spectrozyme FXa (American Diagnostica). Rates of factor Xa generation in dependence of the VII concentration were fitted to the single site
binding equation, as described by Krishnaswamy (24). This calculation
yielded apparent dissociation constants
(KD,app) as well as the
maximum rate of Xa generation as measures of proteolytic function of
the mutant VIIa in complex with TF. The effect of mutations in TF on
the interaction with wild-type VIIa or mutant VIIaAla63 was
analyzed with purified VIIa and TF from transient expression in
mammalian cells, as described previously in detail (23). Briefly,
full-length TF content in the transfected Chinese hamster ovary K1
cells was determined by immunoassay. For the functional assay as
described above, cells were lysed in 15 mM octyl--D-glucopyranoside (Calbiochem) in 20 mM HEPES, 130 mM NaCl, pH 7.4, and incubated
for 15 min at 37 °C, followed by dilution of the cell lysate to
yield a final concentration of 5 pM TF in the assay.
The amidolytic function of wild-type or mutant VIIa (5 nM) was analyzed in Tris-buffered saline, 5 mM CaCl2, 0.1% bovine serum albumin, pH 7.4, in the presence of increasing concentrations of soluble TF1-218 (0.00025-15 µM) at ambient temperature. TF·VIIa complexes were allowed to form for 5 min and were analyzed for amidolytic activity toward 1 mM chromogenic substrate Chromozym tPA (Boehringer Mannheim). Apparent dissociation constants were determined based on the saturation binding curves as described above.
Determination of Kinetic Parameters Km,app and kcatKm,app and kcat for protein substrate factor X hydrolysis by TF· VIIa were determined using relipidated full-length TF (50 pM) which was incubated with excess wild-type or mutant VIIa (5 nM) in Tris-buffered saline, 5 mM CaCl2, 0.1% bovine serum albumin, pH 7.4, for 10 min to allow for complex assembly. Increasing concentrations of factor X (0.015-3 µM) were added for 1-5 min at 37 °C, and the reaction was stopped by the addition of 100 mM EDTA. Factor Xa generation was assessed by hydrolysis of chromogenic substrate, and data were fitted to the Michaelis-Menten equation using the program Enzfitter (Elsevier Biosoft). For determination of proteolytic function in the absence of TF, 2 µM factor X was incubated with 250 nM wild-type or mutant VIIa at 5 mM Ca2+ and 37 °C. Rates of factor Xa generation were linear for 1 h, the maximum incubation time used for the calculation of rates of product formation determined in three independent experiments.
Characterization of TF Binding by Surface Plasmon Resonance MeasurementsBinding of wild-type VIIa, VIIaAla49, and VIIaAla63 to TF1-218 was analyzed at varying calcium concentrations using surface plasmon resonance on a BIAcore 2000 instrument (Pharmacia Biosensor). The noninhibitory monoclonal antibody to TF, TF9-10H10 (17), was immobilized on a CM5 sensor chip surface by amine coupling according to the manufacturer's recommendations. Soluble TF1-218 at 0.05 mg/ml and a flow rate of 10 µl/min was captured to the TF9-10H10 surface. VIIa proteins were dialyzed in a metal-chelating buffer consisting of HEPES-buffered saline (150 mM NaCl, 15 mM HEPES, pH 7.4), 20 g of Chelex 100 resin/liter. Binding analysis of wild-type or mutant VIIa was performed in HEPES-buffered saline, 0.005% P20 surfactant with the addition of either 10 mM EDTA or 0.05, 0.2, 1, or 5 mM CaCl2. Association of VIIa was monitored during a 4-min injection of 0.012-10 µM VIIa across the TF-saturated sensor chip surface. Dissociation was monitored over a 5-10-min range after return to buffer flow. After each analysis, regeneration of the chip surface was achieved by pulse injection of 100 mM EDTA or 10 mM HCl followed by saturation of the antibody with a new injection of TF1-218. The kinetic binding constants (ka, kd, and KD) were determined by nonlinear regression analysis as described previously for the TF-VIIa interaction (25) using the evaluation software provided by the manufacturer. The calculations of the association rate constants (ka) were based on multiple association sensograms with at least five different VIIa concentrations. The dissociation rate constant was calculated from the initial dissociation phase of the binding curves, and the equilibrium dissociation constant, KD, equals the kd:ka ratio.
To reduce the
affinity for Ca2+ binding by EGF-1, we generated individual
Ala replacement mutants for Gln49 and Asp63
which provide one and two side chain oxygen atoms, respectively, for
the coordination of Ca2+ (Fig. 1). The
mutants were transiently expressed in mammalian cells and tested for TF
binding and proteolytic activity in a linked functional assay (23).
Elimination of a single coordination in VIIaAla49 had
little impact on VIIa proteolytic function and reduced TF affinity by
<2-fold. In contrast, the expression levels typically achieved in
transient transfection experiments were below concentrations needed to
achieve full saturation of TF by the Asp63 Ala mutant,
indicating a severe reduction in TF binding and possibly diminished
proteolytic function (Table I).
|
For further study of these mutants, recombinant proteins were purified
from serum-free culture supernatant of stably transfected cells using a
two-step procedure involving monoclonal affinity and ion exchange
chromatography employing Ca2+ gradient elution to select
for fully -carboxylated proteins. The final preparations had the
expected percentage of Glu residues modified by
-carboxylation.
VIIAla63 had 9.7, VIIAla49 had 10.0, and
wild-type recombinant VII had 9.9 Gla residues. Plasma-derived VII
purified by the same protocol was used as a control and showed the
expected 10.0 Gla residues. VIIaAla49 yielded activated
VIIa during the purification, as typically observed for wild-type VIIa, whereas VIIaAla63 only partially converted to VIIa, despite
elution from the ion exchange resin by Ca2+ at a
concentration identical to that of wild-type VIIa. The normal chromatographic profile suggested that the Gla domain of
VIIAla63 adopted a conformational transition upon
Ca2+ binding similar to wild-type VII.
Ca2+-dependent changes in the intrinsic protein
fluorescence did not reveal differences between mutant and wild-type
proteins, and we also found identical Ca2+ dependence for
binding of both mutants and wild-type VIIa to a monoclonal antibody
directed to a Ca2+-sensitive epitope in the Gla domain
(data not shown). We thus were unable to detect a defect in the
conformation of the Ca2+-saturated Gla domain of this
mutant. Activation of the TF-bound mutant zymogen by factor Xa occurred
with rates indistinguishable from wild-type VII (data not shown),
indicating that the mutation in the EGF-1 Ca2+ site
selectively affected the "autoactivation" of free
VIIAla63 (26) which occurs during the purification.
Activation of VIIAla63 was achieved by cleavage with factor IXa in fluid phase. Purified proteins were analyzed in the functional assay at 5 mM Ca2+ (Table I). Consistent with data from transient transfection experiments, elimination of the single Ca2+ coordination provided by Gln49 did not impair factor X activation and caused no detectable change in the KD,app. The affinity of TF binding was reduced >100-fold as a consequence of the Asp63 mutation, and at saturation of cofactor a 4-fold reduction in the rate of factor X activation was apparent (Table I). The loss of function was similar with the zymogen form and the activated enzyme, excluding the possibility that a defect in the conversion of VII to VIIa additionally contributes to the functional defect of VIIAla63. We conclude from these results that the EGF-1 Ca2+ site is important for optimal interaction of VIIa with macromolecular substrate. It is noteworthy that elimination of a Ca2+-coordinating residue side chain (Glu220) in the protease domain Ca2+ site reduced catalytic function without diminishing the affinity for TF (Table I). Thus, Ca2+ binding to the catalytic domain has little impact on the TF interaction, despite a structural role in maintaining a catalytically active conformation of the VIIa protease domain.
Effect of EGF-1 Mutations on Amidolytic and Proteolytic Function of VIIaMutant and wild-type VIIa was saturated with increasing concentrations of soluble TF1-218 at 5 mM Ca2+, and the amidolytic function of the complex was determined with chromogenic p-nitroanilide substrate. The calculated apparent dissociation constants for these experiments demonstrated a significant reduction in TF binding of VIIaAla63 compared with wild-type VIIa. Since these experiments were conducted with soluble TF in the absence of lipids, the results are consistent with the notion that mutation of Asp63 affected protein-protein interactions with TF. At saturation of the mutants with TF1-218, there was no difference in chromogenic substrate hydrolysis between either VIIaAla49 or VIIaAla63 and wild-type VIIa (Table II). Kinetic parameters for macromolecular substrate factor X activation demonstrated that VIIaAla49 had normal proteolytic function at 5 mM Ca2+. In contrast, VIIaAla63 was defective in factor X activation resulting predominantly from a 3-fold decrease in kcat with unchanged Km,app (Table II). In the absence of TF and phospholipid, VIIaAla63 activated factor X with a 4-fold reduced rate (14 ± 4 pM factor Xa generated/min) compared with wild-type VIIa (52 ± 11 pM factor Xa generated/min) when tested at the same enzyme concentration (250 nM). These results establish that VIIaAla63 is selectively defective in macromolecular substrate activation independent of cofactor interactions.
|
The Ca2+ dependence of VIIa binding to soluble TF1-218 was analyzed by surface plasmon resonance detection. Ca2+ saturation increased the affinity of wild-type VIIa for TF by 3 orders of magnitude, reflected in a decrease of the KD from 4.7 ± 2.1 µM in 10 mM EDTA to 5.8 ± 2.9 nM at 5 mM Ca2+ (Table III). The association rate constant (ka) increased only modestly by ~4-fold, whereas a >100-fold decrease in the dissociation rate constant (kd) was largely responsible for the tighter binding of VIIa to TF. Ca2+ binding to VIIa resulted in a ~4 kcal/mol change of the calculated free energy of TF binding, and 75% of the change was observed at 50-200 µM Ca2+ (Fig. 2). This Ca2+ dependence is characteristic of high affinity Ca2+ binding sites located outside the Gla domain, and thus the energetic contributions to TF binding which result from Ca2+ binding are only to a lesser extent attributable to Ca2+ saturation of the Gla domain. Indeed the approximately 1 kcal/mol increase upon addition of 1 mM Ca2+ is consistent with the modest loss of binding function following truncations that remove the Gla domain from VIIa (14, 17).
|
There was no difference in TF binding when VIIaAla49 was compared with wild-type VIIa in the presence of the divalent cation chelator EDTA, but a subtle 2-fold increase in the dissociation rate for VIIaAla63 was found (Table III). This subtle difference could indicate that the Asp63 side chain, in part, may contribute to TF binding independent of its role in Ca2+ coordination. Both mutants were severely defective compared with wild-type VIIa at 50 and 200 µM Ca2+, reflected in a shift of the saturation curves displayed in Fig. 2. Whereas a defect in the association rate constant was observed only at 50 µM Ca2+ for VIIaAla49, the association rate for VIIaAla63 did not change with increasing Ca2+, resulting in a 4-fold lower rate compared with wild-type TF at 5 mM Ca2+. The defect in the dissociation rate constant and consequently in the KD was most pronounced at 50 and 200 µM Ca2+ for VIIaAla49, and the mutant displayed close to normal binding of TF at 5 mM Ca2+. This indicates that the elimination of a single coordinating oxygen atom results in diminished affinity for Ca2+ rather than completely abolishing Ca2+ binding. In contrast, VIIaAla63 had >10-fold reduced TF affinity at 5 mM Ca2+, the highest concentration that could be tested without technical difficulties likely arising from aggregation of VIIa. The results here document a significant binding defect introduced by the elimination of two of the coordinating oxygens of the Asp63 side chain. This binding defect was not readily compensated for by higher Ca2+ concentrations, establishing a prominent role for the EGF-1 Ca2+ binding site in increasing the affinity of VIIa for TF.
Mutations of Other Residues Involved in Ca2+ CoordinationThe functional roles of additional residues in the EGF-1 Ca2+ site were analyzed in transient transfection experiments. Asp48 plays a local structural role in the EGF-1 Ca2+ site, but it does not directly coordinate Ca2+ (4). Ala exchange of this residue decreased the affinity for TF ~10-fold, a loss of function intermediate between the severe reduction observed with VIIaAla63 and the subtle defect of VIIaAla49. Asp46, being localized in the Gla domain at the carboxyl terminus of the helical aromatic stack region, provides the fourth Ca2+-coordinating side chain oxygen atom (5). Unlike the subtle defect resulting from the elimination of a side chain with a single coordination in the case of Gln49, the Asp46 mutation more severely reduced TF binding and TF·VIIa proteolytic function (Table IV). The more prominent defect of the Asp46 mutation points to a function of the EGF-1 Ca2+ site in orienting the Gla domain including the aromatic stack region relative to EGF-1. Replacing the Asp46 side chain with Glu resulted in mutant protein with a less severe defect in TF binding and proteolytic function. Our interpretation is that providing Ca2+ coordination alone is not sufficient to restore function fully. The Gly replacement for Asp46 displayed a binding defect similar to the Ala mutant and a more pronounced defect in proteolytic function, an indication that flexibility at this position is detrimental. Indeed, the Asp at this position may function as a bridge connecting two structures: the EGF-1 through calcium coordination and the Gla domain by virtue of the Asp46 position at the end of the helix which harbors critical residues that constitute the hydrophobic stack. However, parts of the aliphatic side chain of Asp46 may contribute directly to macromolecular substrate interaction, and the mutagenesis results would also be consistent with this interpretation.
|
The defect in TF binding upon mutations of Asp63 could be explained by local conformational effects on an important contact with TF. Gln64 in EGF-1 and Ser43 in the Gla domain (Fig. 1) share Gln110 as a common contact in TF (5), and loss of Ca2+ coordination may position Ser43 or Gln64 in an unfavorable orientation for interaction with TF. Mutation of TF residue Gln110 reduced the affinity for wild-type VIIa ~3-fold (Table IV), a loss of function consistent with the study by Kelley et al. (25), but less than the 1.4 kcal/mol decrease in the free energy of binding reported by Gibbs et al. (27). Binding of VIIaAla63 to TFAla110 occurred with a KD,app of 1,600 pM, significantly higher than the KD,app of 459 pM for binding to wild-type TF. The effect of the mutation in TF and VIIa thus appear to be additive, indicating that the consequence of reduced Ca2+ binding to EGF-1 is not a local perturbation of the contact with Gln110 in TF.
The conclusion that the TF binding defect does not arise from a local conformational change supports the view that the EGF-1 Ca2+ site functions by stabilizing the structural orientation of the Gla module relative to EGF-1. In this model, Ca2+ binding may influence multiple interactions in the Gla domain and possibly the EGF-1 interfaces with TF. Alternatively, Ca2+ coordination and the orientation provided by the stack residue Asp46 may influence a specific contact of VIIa with TF. We tested the latter hypothesis by generating double mutants in VIIa and analyzing whether the mutational effect was additive. Mutation of Arg79 in EGF-1 and Phe40 in the aromatic stack were considered as contact residues potentially influenced by the EGF-1 Ca2+ site because of a ~10-fold reduction in TF binding which was similar to the mutational effect at Asp46 in the EGF-1 Ca2+ site (Table IV). The double mutants of Asp46 with Arg79 as well as that of Asp46 with Phe40 displayed additive mutational effects compared with the individual residue replacements. We conclude that Ca2+ binding to EGF-1 does not selectively influence these two contact residues, and we rather propose an effect of the EGF-1 Ca2+ site on multiple interactions in the interfaces of VIIa with TF.
From the mutational analysis of the VIIa EGF-1 Ca2+ binding site, we arrive at the following conclusions. (i) Approximately 75% of the increase in TF affinity caused by Ca2+ binding to VIIa is observed at 50-200 µM Ca2+; this Ca2+ concentration range is consistent with estimates for the affinity of the EGF-1 site for Ca2+ (8, 10, 13, 28). (ii) Mutations of Ca2+-coordinating residues in EGF-1 result in significantly reduced affinity for TF, particularly at 50 and 200 µM Ca2+. (iii) Mutations of the aromatic stack residue Asp46 result in mutational defects that are larger than those at the Gln49 position, which in the crystal structure of the complex (5) also provides a single Ca2+-coordinating oxygen atom. (iv) Ca2+ binding to EGF-1 does not affect the VIIa interaction with TF residue Gln110. (v) Certain mutants had defects in proteolytic function, suggesting that the EGF-1 Ca2+ site functions to support macromolecular substrate interaction with TF·VIIa.
Whereas Ca2+ has little impact on the conformation of the
TF extracellular domain (17), Ca2+ binding to the Gla
domain is important for high affinity binding to TF. Several studies
have provided convergent evidence that the Ca2+-saturated
conformation of the Gla domain is responsible for an approximately 1 kcal/mol increase in the free energy of binding for TF-VIIa
protein-protein interactions (14, 16, 17). The contribution of the Gla
domain, however, does not fully account for the effect of
Ca2+ on VIIa binding to TF. Indirect evidence from
competition experiments with synthetic peptides analogous to VIIa
protease domain sequences (10) or with a Glu220 Ala
mutant of VII (11) was interpreted to indicate that the protease domain
Ca2+ binding site contributes to TF interactions. These
studies did not provide a direct binding analysis of the mutants in the
catalytic domain binding site. The functional
KD,app for
VIIaAla220 determined in our experiments did not reveal
defects in TF binding, although the amidolytic and proteolytic function
of our mutant preparation was severely impaired, consistent with
previous analysis of this mutant (11). The decreased proteolytic
function (Table I) indicates incomplete Ca2+ saturation of
the site under our experimental conditions. We thus interpret the
unchanged KD,app as
evidence that the protease domain Ca2+ site, consistent
with its location distant from the TF interface (5), has little
importance for TF binding and that Ca2+ binding to EGF-1 is
largely responsible for the demonstrated Ca2+ dependence of
the TF-VIIa interaction (10, 15, 17). Notably, even at the Gla domain
saturating concentration of 5 mM Ca2+, a
significant binding defect remained for VIIaAla63 and
VIIaAla46, indicating that cooperative effects of
Ca2+ binding to the Gla domain do not substitute for the
requirement of Ca2+ binding to EGF-1.
Since the crystal structure of the TF·VIIa complex reveals no direct role of the EGF-1 Ca2+ site in TF binding, the binding defects resulting from the mutation of Ca2+-coordinating residues must be explained by an indirect effect on the VIIa light chain conformation. These indirect effects could be predominantly on EGF-1 or on Gla domain interactions with TF. In the first case, Gln64 is the most likely residue to be affected by loss of Ca2+ binding to EGF-1 because the carbonyl oxygen of Gln64 participates in Ca2+ coordination (5), and loss of Ca2+ binding to EGF-1 may affect the contact of the Gln64 side chain with Gln110 in TF. However, the mutational effect of TF residue Gln110 was additive with the replacement of Asp63 in VIIa, arguing against a local perturbation of this contact upon mutation in the EGF-1 Ca2+ site. The finding that the binding defects of the Asp46 and Arg79 were additive in the double mutant further suggests that lack of Ca2+ binding does not cause a major structural defect in EGF-1. We take this as evidence in favor of an influence of the EGF-1 Ca2+ site mutations on TF binding by the Gla domain rather than the EGF-1.
The structure of the TF·VIIa complex indicates a potential mechanism
by which the Ca2+ site in EGF-1 could impair Gla domain
binding to TF. The Ca2+-coordinating residue
Asp46 forms a bridge between the EGF-1 and the Gla domain,
since it is located at the carboxyl terminus of the -helical
hydrophobic stack region of the Gla domain. Ala mutations of
Asp46 and Phe40 displayed additivity of the
binding defect, indicating that the Phe40 interaction with
TF is not influenced by Ca2+ binding to EGF-1. However, one
must consider the possibility that the Gla domain may dock with TF in a
different orientation when Ca2+ coordination in EGF-1 is
lost. In this case, the additivity of the mutational effect with a
residue in the Gla domain would not exclude that the EGF-1
Ca2+ site affects binding of the Gla domain to TF. Indeed,
additional mutations at the Asp46 position support the
notion that an important function of Ca2+ coordination is
to restrict flexibility and to orient the Gla domain relative to EGF-1.
We found a less severe loss of function for the Glu replacement for
Asp46 relative to Ala, likely a consequence of
Ca2+ coordination by the Glu side chain. Although function
of the Glu replacement was not completely normal, the improved function relative to the Ala or Gly replacements may reflect restricted flexibility in the Gla domain-EGF-1 linkage.
Several of the Ala replacement mutants which had a significant binding defect also displayed a defect in proteolytic function. It is possible that the proteolytic defect is unrelated to the loss of Ca2+ coordination and rather reflects a direct interaction of the respective side chains with macromolecular substrate. Interactions of the substrate light chain with the EGF-1 Ca2+ site can be envisioned assuming a docking of the substrate Gla domain with Lys165 and Lys166 in the carboxyl-terminal module of TF (29). However, the Gla domain of VIIa is also critically important for protein substrate recognition (17), and a stabilization of the VIIa Gla orientation may be required for proper substrate binding. In support of this hypothesis, mutations of TF residues in the Gla domain interface also affected proteolytic function of TF·VIIa (30). The additional stabilization provided by the EGF-1 Ca2+ site may thus provide for optimal substrate docking with the Gla domain of VIIa and the adjacent TF residues Lys165 and Lys166 (31). This may suggest that the predominant functional role of the EGF-1 Ca2+ is to stabilize an optimal orientation of the Gla domain relative to EGF-1 even after docking of the two domains with TF.
We are grateful for the excellent technical assistance of Justin Shobe, Cindi Biazak, Jennifer Royce, and David Revak and for preparation of the manuscript by Jenny Robertson. We thank Dr. T. S. Edgington for helpful discussion, Dr. D. Stuart and Dr. A. Tulinsky for coordinates, and Dr. S. Krishnaswamy for curve fitting software.