The N-terminal epidermal growth factor-like domain in factor IX and factor X represents an important recognition motif for binding to tissue factor.

Factors VII, IX, and X play key roles in blood coagulation. Each protein contains an N-terminal gamma-carboxyglutamic acid domain, followed by EGF1 and EGF2 domains, and the C-terminal serine protease domain. Protein C has similar domain structure and functions as an anticoagulant. During physiologic clotting, the factor VIIa-tissue factor (FVIIa*TF) complex activates both factor IX (FIX) and factor X (FX). FVIIa represents the enzyme, and TF represents the membrane-bound cofactor for this reaction. The substrates FIX and FX may utilize multiple domains in binding to the FVIIa*TF complex. To investigate the role of the EGF1 domain in this context, we expressed wild type FIX (FIX(WT)), FIX(Q50P), FIX(PCEGF1) (EGF1 domain replaced with that of protein C), FIX(DeltaEGF1) (EGF1 domain deleted), FX(WT), and FX(PCEGF1). Complexes of FVIIa with TF as well as with soluble TF (sTF) lacking the transmembrane region were prepared, and activations of WT and mutant proteins were monitored by SDS-PAGE and by enzyme assays. FVIIa*TF or FVIIa*sTF activated each mutant significantly more slowly than the FIX(WT) or FX(WT). Importantly, in ligand blot assays, FIX(WT) and FX(WT) bound to sTF, whereas mutants did not; however, all mutants and WT proteins bound to FVIIa. Further experiments revealed that the affinity of the mutants for sTF was reduced 3-10-fold and that the synthetic EGF1 domain (of FIX) inhibited FIX binding to sTF with K(i) of approximately 60 microm. Notably, each FIXa or FXa mutant activated FVII and bound to antithrombin, normally indicating correct folding of each protein. In additional experiments, FIXa with or without FVIIIa activated FX(WT) and FX(PCEGF1) normally, which is interpreted to mean that the EGF1 domain of FX does not play a significant role in its interaction with FVIIIa. Cumulatively, our data reveal that substrates FIX and FX in addition to interacting with FVIIa (enzyme) interact with TF (cofactor) using, in part, the EGF1 domain.

Human factor IX (FIX) 1 and factor X (FX) are vitamin K-de-pendent glycoproteins with M r of 57,000 and 58,800, respectively (1,2). Factor VIIa-tissue factor (FVIIa⅐TF) complex activates FIX to FIXa and FX to FXa by cleaving Arg 145 -Ala 146 and Arg 180 -Val 181 peptide bonds in FIX (3) and the Arg 194 -Ile 195 peptide bond in FX (2). The resulting FIXa or FXa molecule consists of an N-terminal light chain and a C-terminal heavy chain linked by a disulfide bond. The light chain in each case contains a ␥-carboxyglutamic acid (Gla) domain and two epidermal growth factor-like domains (EGF1 and EGF2), whereas the heavy chain contains the serine protease domain. In the blood coagulation cascade, FIXa also activates FX to FXa in a reaction that requires factor VIIIa (FVIIIa), phospholipid (PL), and calcium. FXa formed by either pathway then activates prothrombin to thrombin in a reaction that requires factor Va, PL, and calcium (4). In addition, both FIXa and FXa activate FVII to FVIIa (5)(6)(7) and are inhibited by antithrombin (AT) (8,9).
The conversion of single chain zymogen FVII to enzyme FVIIa involves the cleavage of a single peptide bond between Arg 152 and Ile 153 . The FVIIa formed consists of a light chain of 152 amino acids and a heavy chain of 254 amino acids held together by a disulfide bond (10). Like FIXa and FXa, the N-terminal light chain of FVIIa contains the Gla domain and two EGF-like domains, whereas the heavy chain contains the serine protease domain (10). TF, the cellular cofactor for FVIIa, is composed of two fibronectin type III ␤-sandwich domains (11,12). Recently, high resolution x-ray structure of the complex of soluble tissue factor (sTF) and FVIIa has been reported (13). In this structure, the Gla and EGF1 domains make contact with the C-terminal domain of TF and the EGF2 and the protease domains make contact with the N-terminal domain of TF (13). Thus, FVIIa uses all of its four domains in binding to the Nand C-terminal domains of TF (13).
Efforts have been directed to understanding the regions in FVIIa⅐TF that interact with the substrates FIX and FX. By studying the effect of mutations in the C-terminal domain of TF, it has been proposed that this domain may interact with the Gla domains of FIX and FX (14). Similarly, by mutagenesis and docking experiments, it has been proposed that the Gla * This work was supported by National Institutes of Health Grant HL36365 and American Heart Association Grant 9950228N. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
A preliminary account of this work has been presented in abstract form (19).
domain of FVIIa interacts with the Gla domain of FX (15). Further, we reported earlier that the EGF1 domain of FIX is required for its activation by the FVIIa⅐TF complex (16). However, the role of the EGF1 domain of FX in this context has not been investigated. Moreover, it is not known whether FVIIa or TF in the FVIIa⅐TF complex interacts with the EGF1 domains of FIX and FX. Thus, the precise function of EGF1 domain of FIX or FX in its interaction with the FVIIa⅐TF complex is not known. Protein C is a serine protease with an anticoagulant function whose domain organization is similar to that of FVIIa, FIXa, or FXa (17,18). Further, activated protein C is not involved in the TF-induced coagulation, and its EGF1 domain near the N terminus has an eight-residue insertion (18). Therefore, substituting the EGF1 domain of FIX (or FX) with the EGF1 domain of protein C should replace the unique determinants present in the EGF1 domain of FIX (or FX) that provides specificity for its interaction with the FVIIa⅐TF complex. In this report, in addition to the above two replacement mutants (FIX PCEGF1 and FX PCEGF1 ), we used a point mutant (FIX Q50P ) and an EGF1 deletion mutant (FIX ⌬EGF1 ) of FIX to understand the function of this domain in TF-induced coagulation. Data are provided, which strongly indicate that TF interacts with the EGF1 domain in FIX and FX. Our findings represent the first report that assigns a specific function to the EGF1 domain in these proteins.

EXPERIMENTAL PROCEDURES
Reagents-Carrier-free Na 125 I was obtained from ICN Biomedicals, Inc. Benzoyl-Ile-Glu-Gly-Arg-p-nitroanilide (S-2222) was obtained from Diapharma Inc. Biotinylated Glu-Gly-Arg-chloromethylketone (biotin-EGR-CK) was purchased from Hematologic Technologies, Inc. Nitrocellulose membrane, polyethylene glycol 8000 (PEG), p-nitrophenyl phosphate, bovine serum albumin (BSA), bovine brain phosphatidylcholine, and phosphatidylserine were purchased from Sigma. Horseradish peroxidase-goat anti-mouse IgG and enhanced chemiluminescence (ECL) detection reagents were purchased from Amersham Biosciences. FVIIdepleted plasma and Neoplastin were obtained from Amersham Biosciences. Normal plasma FIX (FIX NP ), plasma FX (FX NP ), FXIa, Russell's viper venom (RVV), and AT were obtained from Enzyme Research Laboratory. Low molecular weight heparin was purchased from Rhône-Poulenc Rorer Pharmaceuticals Inc. A monoclonal antibody-purified human FVIII concentrate was obtained from Dr. Leon Hoyer (American Red Cross, Rockville, MD). It was activated with 1 nM thrombin in the presence of 0.1% BSA and 5 mM CaCl 2 in Tris/NaCl at 37°C for 2 min as described earlier (20). The formed FVIIIa was diluted and used immediately in the activation of FX by FIXa⅐FVIIIa⅐PL.
For ligand blot experiments, a Ca 2ϩ -dependent FIX monoclonal antibody (mAb) cell line was provided by Dr. Shirly Miekka of the American Red Cross, and the IgG was purified as described (21). A Ca 2ϩ -dependent mAb to the heavy chain of FX used for the ligand blot experiments was purchased from American Diagnostics, Inc. PL vesicles (75% phosphatidylcholine, 25% phosphatidylserine) were prepared by the method of Husten et al. (22). TF containing the transmembrane region (residues 1-243) was a gift from Genetech, Inc. The relipidation of the TF was performed as described (16). sTF that lacks the transmembrane region (residues 1-219) was a gift from  (24). SDS-Gel Electrophoresis-SDS-gel electrophoresis was performed using the Laemmli buffer system (24). The acrylamide concentration was 12%, and the gels were stained with Coomassie Brilliant Blue dye. All proteins used in the present study were ϳ98% pure.
Amino Acid Sequencing and Gla Analysis-Gla and amino acid sequence analysis were performed by Commonwealth Biotechnologies, Inc. (Richmond, VA). Automated degradation of each protein (ϳ0.5 nmol) was performed using an Applied Biosystems gas phase sequencer. Gla analysis of each protein was performed by alkaline hydrolysis followed by HPLC analysis. The amount of Gla was quantitated based upon Asp and Asn present per mol of each protein.
Expression and Purification of Recombinant Factors IX and VII-Recombinant FIX WT , FIX ⌬EGF1 , FIX PCEGF1 , and FIX Q50P were expressed in human embryonic kidney 293 cells and purified by using the IX A-7 mAb column as described (16,25). Each FIX protein had ϳ12 Gla residues/mol (16). To express FVII WT , the restriction sites AflII and XhoI were introduced at the 5Ј-and 3Ј-ends of VII cDNA for ligation into the pMon3360b expression vector (26) that was modified to contain AflII and XhoI sites. A stable cell line that expressed FVII WT was established as described in detail by Hippenmeyer and Highkin (26). Medium was collected in the presence of vitamin K as outlined earlier for FIX (16,25). FVII WT was purified by using a Ca 2ϩ -dependent mAb as described (27). It contained 9 -10 Gla residues/mol and had ANAFL as the N-terminal sequence. FVIIa was obtained as earlier, except insoluble FXa (Sepharose-FXa) was used instead of the soluble FXa as the activator (6). The resin was removed by centrifugation, and the supernatant was passed over a small Chelex-100 column to remove Ca 2ϩ . Aliquots were kept frozen at Ϫ80°C until used.
Expression and Purification of Recombinant FX-An expression vector for FX WT was constructed in which the prepro-leader sequence of FX WT was replaced with that of prothrombin as described by Camire et al. (28). The prepro-leader sequence of prothrombin was amplified by PCR using primers A and B (Table I) and a human liver cDNA library. The prepro-leader sequence of prothrombin was then linked to the FX cDNA sequence by the overlap extension method using primers A and C (25). The resulting chimeric DNA, containing the prepro-leader sequence of prothrombin followed by the FX sequence, was digested with AflII and XhoI and ligated into pMon3360b expression vector. A stable cell line that expressed FX WT was established as described (26). Medium was collected in the presence of vitamin K as outlined earlier for FIX (16,25). FX WT was purified using a Ca 2ϩ -dependent mAb to the Gla domain of FX (25) followed by FPLC Mono Q column. The conditions for the FPLC Mono Q column were the same as described previously for FIX purification (16,25). Construction of FX PCEGF1 was performed by the overlap extension method as described (25), and primers D and E (Table I) a Primer A contains a AflII site, and primer C contains an XhoI site. The restriction site sequences are underlined. The sequence in parenthesis for primers A and B correspond to the prepro-leader sequence of human prothrombin. The sequences in parenthesis is primer C correspond to the C-terminal six amino acid residues of FX as well as the stop codon. Primers D and E were used to construct FX PCEGF1 . These are hybrid primers containing FX and protein C DNA sequences. The sequences in parenthesis correspond to the human protein C sequences. each biotin-EGR-IXa sample was further diluted 50-fold in Tris/NaCl containing 1% BSA and 0.1% Tween 20. 100 l of the diluted sample was added to each well, and the plate was incubated at 37°C for 2 h for capture of the biotin-EGR-IXa by the FIX mAb. The plate was washed three times with Tris/NaCl containing 0.1% Tween 20 and 5 mM CaCl 2 . Each well then received 100 l of alkaline phosphatase-streptavidin in Tris/NaCl containing 1% BSA, 0.1% Tween 20, and 5 mM CaCl 2 . The plate was incubated at 37°C for 1 h. After three washings, each well received 100 l of substrate p-nitrophenyl phosphate (4 mg/ml) in the alkaline phosphatase buffer (100 mM NaCl, 5 mM MgCl 2 , 100 mM Tris, pH 9.5). The amount of p-nitrophenol generated was measured in a microtiter plate reader at 405 nm (Bio-Rad). FIXa concentration was then calculated from a standard curve generated starting with known amounts of preformed FIXa and formation of the biotin-EGR-IXa using the above protocol.
For activation of FIX with FVIIa⅐sTF, 4 M FIX was activated with 0.16 M FVIIa⅐sTF in the absence of PL. All other assay conditions were the same as those for the activation of FIX with FVIIa⅐TF⅐PL outlined above. For SDS-PAGE, 12 l of the reaction mixture was removed at different times and added to 2 l of 0.5 M EDTA and 5 l of 5-fold concentrated SDS-reducing buffer. Samples were placed in boiling water for 5 min and analyzed by SDS-PAGE (24). Measurements of Rates of Activation of FX by FIXa with or without FVIIIa-These experiments were performed exactly as described in detail earlier (30). The concentration of each reagent used is given in the legend to Fig. 6.

Measurements of Rates of Activation of FX by FVIIa-For
Measurements of Rates of FVII Activation by FIXa or FXa-A 50 nM concentration of each FIXa or FXa was incubated at 37°C with 1 M FVII in the presence of 35 M PL and 5 mM CaCl 2 in Tris/NaCl. 2-l aliquots were removed at different times and added to 100 l of 0.1% BSA in Tris/NaCl containing 10 mM EDTA. The aliquots were further diluted in 0.1% BSA in Tris/NaCl without EDTA and analyzed for FVII/FVIIa clotting activity in a one-stage assay (31). For this assay, 50 l of FVII-depleted plasma was incubated with 50 l of Neoplastin for 3 min at 37°C. Then 25 l of test sample and 50 l of prewarmed (37°C) 35 mM CaCl 2 were added and the clotting time was noted. Citrated pooled normal human plasma was used as a standard (1 unit/ml FVII). For SDS-PAGE analysis, samples were removed after a 2-h incubation period.
Binding of FIXa and FXa to AT-For these experiments, the final reaction mixtures contained the following: 2 M FIXa or FXa, 2 M AT, 5 mM CaCl 2 , and 10 units/ml low molecular weight heparin in Tris/ NaCl, pH 7.4. The total reaction mixture in each case was 150 l, and 20-l aliquots were removed at 0.15, 0.5, 1.5, 4, 12, and 30 min and added to 5 l of 5-fold concentrated SDS-reducing buffer and analyzed by SDS-PAGE. The protein bands were visualized by Coomassie Blue staining and quantitated by densitometry. The rate of complex formation of heavy chain of FIXa or FXa with AT was then calculated from the decrease in the intensity of band corresponding to the heavy chain of each enzyme and increase in intensity of the AT-heavy chain complex.
Ligand Blotting for Binding of sTF or FVIIa to FIX and FX-First sTF and FVIIa were electrophoresed on SDS-PAGE. The proteins were then transferred to a 0.2-m nitrocellulose membrane. The protocol used was that outlined by Sambrook et al. (32), using the Bio-Rad Mini-transfer apparatus. The membrane was blocked with 5% fat-free milk, 0.05% Tween 20 in Tris/NaCl at room temperature for 1 h. Each membrane was then incubated with various FIX or FX proteins at 5 g/ml in Tris/NaCl, 1% milk, 0.05% Tween 20, and 5 mM CaCl 2 at 4°C overnight. After three washes with 0.05% Tween 20 in Tris/NaCl and 5 mM CaCl 2 , the membrane was incubated with FIX mAb or FX mAb (1 g/ml) at room temperature for 2 h. A second antibody (horseradish peroxidase-goat anti-mouse IgG) and the ECL Western blotting detection kit were used to detect the primary mAb.
Synthesis and Folding of the EGF1 Domain of FIX-The EGF1 domain of FIX corresponding to amino acid residues 45-87 (NH 2 -YVDGQCESNPCLNGGSCKDDINSYECWCPFGFEGKNCELDVT-CONH 2 ) was synthesized on an ABI model 433A peptide synthesizer by Biomolecules Midwest, Inc. (Waterloo, IL). The C-terminal amino acid was coupled to Fmoc-Rink Amide MBHA resin using standard ABI protocols. Amino acid activation was performed using HBTU. The ␣amino group of the amino acid was Fmoc-protected, and the side chain groups were protected by t-butyl (Tyr and Ser), t-butyl ester (Asp and Glu), t-butoxycarbonyl (Lys), 2,2,4,6,7-pentamethyldihydrobenzofuran-5-sulfonyl (Arg), and trityl (His, Cys, Gln, and Asn). Fmoc deprotection was performed using 20% piperidine in 1-methyl-2-pyrrolidinone, and the peptide was simultaneously deprotected and cleaved from the resin using trifluoroacetic acid/phenol/anisole/water/methyl sulfide (85/5/4/ 4/2; v/v/v/v/v) for 2 h. The crude peptide was purified by reverse phase HPLC on a Vydac C-18 column using standard trifluoroacetic acid/ acetonitrile conditions (33). The purified reduced peptide gave a molecular mass of 4755 Da (expected mass 4755.2 Da) as determined by mass spectrometry analysis using a Finnigan LCQ Iontrap Electrospray mass spectrometer.
The synthetic FIX-EGF1 domain peptide was folded using an oxidoshuffling system (34,35). The lyophilized peptide was dissolved at a concentration of ϳ0.4 mg/ml in a solution containing 0.1 M Tris-HCl, pH 8.3, 50 mM CaCl 2 , 3 mM L-cysteine, and 0.3 mM L-cystine. The mixture was allowed to sit for 40 h at room temperature, at which time 10% trifluoroacetic acid was added to adjust the pH to 2.5. The refolded peptide was then purified by reverse phase HPLC and lyophilized, and its concentration was determined using the molar extinction coefficient of 2748 at 294.4 nm for a single tyrosine or tryptophan residue in 0.1 M NaOH (36). Two tyrosines and one tryptophan contained in our peptide were taken into account in calculating its concentration.
Ca 2ϩ Binding to the Folded IX-EGF1 Domain-Calcium ion activity was determined by using a Ca 2ϩ -specific electrode and a model 601A digital Ionlyzer (Orion Research). Titrations of the folded synthetic EGF1 domain at 400 M in 4 ml of Tris/NaCl, pH 7.4, were performed by adding 1-l increments of 400 mM CaCl 2 at room temperature. In these titrations, bound Ca 2ϩ was taken as the difference between the measured free Ca 2ϩ concentration and the total added.
Iodination of FIX WT and FX WT -FIX and FX were labeled with 125 I using IODO-GEN-precoated iodination tubes obtained from Pierce. The standard Tris buffer was used, and the procedure followed was that outlined by the manufacturer. The iodinated proteins were precipitated by the slow addition of solid ammonium sulfate to 80% saturation. Each suspension was centrifuged at 10,000 rpm in Eppendorf tubes for 10 min at 4°C. The pellets were dissolved in Tris/NaCl, pH 7.4, and dialyzed against the same buffer until the cpm in the dialysate were close to background. The radiospecific activity of 125 I-FIX was 8.5 ϫ 10 6 cpm/g, and that of 125 I-FX was 3.6 ϫ 10 6 cpm/g. SDS-PAGE radioactivity profiles of the proteins revealed a single molecular species characteristic of each protein (37). The radiolabeled proteins retained ϳ90% clotting activity as compared with the nonlabeled controls.
Determination of the Apparent K d of Binding of sTF to FIX and FX Proteins and to FIX-EGF1 Domain-Apparent K d values of 125 I-FIX and 125 I-FX binding to sTF were determined by direct binding assays using Costar enzyme immunoassay/radioimmunoassay, high binding, type I strip plates. Each well was coated with 100 l of either 10 g/ml sTF or BSA in 0.1 M NaHCO 3 , pH 8.5, overnight at 4°C. The wells were washed three times with Tris/NaCl, pH 7.4, containing 0.01% Tween 20. The wells were then blocked with 3% BSA in Tris/NaCl, pH 7.4, containing 0.05% Tween 20 for 4 h at room temperature. The wells were washed with Tris/NaCl, pH 7.4, containing 1 mM benzamidine, 1 mg/ml BSA, and 5 mM CaCl 2 (binding buffer). Each well (sTF or BSA control) then received 100 l of 125 I-FIX (3.7-366 nM) or of 125 I-FX (25-800 nM) in binding buffer. The wells were then allowed to sit at room temperature for 4 h, after which time they were washed three times with the binding buffer. The wells were separated and counted for 125 I radioactivity in a Packard Cobra II ␥-counter. Each BSA control cpm was subtracted from the corresponding experimental cpm to obtain the specific cpm due to binding of FIX or FX to sTF. The nonspecific cpm ranged from 22 to 33% of the experimental cpm. The specific cpm data were plotted against 125 I-FIX or 125 I-FX to obtain approximate K d values for the binding of FIX or FX to sTF. The data were fitted to a hyperbolic curve using the program GraFit from Erithacus Software.
In further experiments, the affinity of each protein (WT, mutants, FIX-EGF1 domain) was determined by equilibrium competition assays. In these experiments, a fixed final concentration of 125 I-FIX (10 nM) or 125 I-FX (30 nM) was used with varying concentrations of the competitor. A 100-l aliquot of each mixture was added to the sTF or BSA wells to generate the specific binding data as outlined above. The data were analyzed using the nonlinear regression analysis program (GraFit) to obtain the IC 50 values (concentration of the competitor yielding 50% inhibition) using the IC 50 four-parameter logistic equation of Halfman (38). To obtain the K d (app) values for the interaction of competitor proteins with sTF, the following equation described by Craig was used (39).
where A is the concentration of 125 I-FIX or 125 I-FX, and K d (WT) represents the dissociation constant of WT protein for sTF. To obtain fully ␥-carboxylated FX, we replaced the preproleader sequence of FX with that of prothrombin as described by Camire et al. (28). FX WT and FX PCEGF1 were expressed in BHK cells and purified using a Ca 2ϩ -dependent mAb whose epitope is in the Gla domain of FX (29). The purified proteins had normal Gla content (11 Ϯ 1) and revealed two N-terminal sequences in equimolar amounts: one corresponding to the heavy chain (SVAQA) and a second corresponding to the light chain (ANSFL). This indicates that the prepro-leader sequences of FX WT and FX PCEGF1 were completely removed. The specific activity was 92 Ϯ 10 units/mg for FX WT , which is similar to that of plasma FX (85 Ϯ 10 units/mg), and it was 12 Ϯ 2 units/mg for FX PCEGF1 in the two-stage clotting assay (31). SDS-PAGE analysis is presented in subsequent figures.

Expression of Recombinant FVII and FX-Purified
Activation of FIX WT , FIX Q50P , FIX PCEGF1 , and FIX ⌬EGF1 by FVIIa⅐TF⅐PL-The activation rates of FIX mutants by FVIIa⅐TF⅐PL obtained by measuring the amount of biotin-EGR-IXa formed are presented in Fig. 1A. Examination of the data in Fig. 1A reveals that the initial rates of activation of FIX-PCEGF1 and FIX Q50P are ϳ7 and ϳ30% of FIX WT , respectively. Under these conditions, FIX ⌬EGF1 was activated only minimally, and FIX WT was activated at a rate similar to that obtained with FIX NP (data not shown).
Activation of FX WT and FX PCEGF1 with FVIIa⅐TF⅐PL-The rates of activation of FX WT and FX PCEGF1 by FVIIa⅐TF⅐PL are presented in Fig. 1B. The initial rate of activation of FX PCEGF1 was ϳ5% of FX WT . It should be noted that the rate of activation of FX WT was similar to the rate obtained with FX NP (data not shown).
Activation of FIX and FX Proteins by FVIIa⅐sTF Complex-Studies presented above reveal that the activation rates of FIX and FX EGF1 domain mutants with FVIIa⅐TF⅐PL are impaired. Gla domains of FIX and FX bind to the PL vesicles, and the membrane surface provides a platform for their assembly for maximal activation. Thus, it is possible that the mutations in the EGF1 domain alter the distance between the cleavage site of FIX or FX and the PL surface, which results in the misalignment of the substrate and the enzyme FVIIa. This possibility was tested by using sTF (transmembrane region deleted TF) under reaction conditions that do not require assembly of FVIIa⅐TF and substrates on the PL surface.
The activation of each FIX mutant by FVIIa⅐sTF in this system where PL is absent is presented in Fig. 2A. Similar to the PL-containing system (Fig. 1A), activations of FIX PCEGF1 and FIX Q50P were impaired. However, in contrast to the PLcontaining system, the rate of activation of FIX ⌬EGF1 in the absence of PL was similar to that of FIX PCEGF1 . The initial activation rate was ϳ33% for FIX Q50P , ϳ12% for FIX PCEGF1 , and ϳ9% for FIX ⌬EGF1 when compared with the rate obtained with FIX WT (Fig. 2A). Again, it should be noted that FIX WT activated at a rate similar to FIX NP in this system (data not shown). Activation rates of FX WT and FX PCEGF1 with FVIIa⅐sTF were also determined. As is the case with the FIX PCEGF1 mutant, the FX PCEGF1 with FVIIa⅐sTF was activated at an initial rate that is ϳ10% of FX WT (Fig. 2B).

SDS-PAGE Analysis of FIX and FX Activation
Products-Next, we performed SDS-PAGE of various FIX and FX proteins activated by FVIIa⅐sTF. The data obtained for FIX WT are presented in Fig. 3A; data for FIX Q50P are in Fig. 3B; data for FIX PCEGF1 are in Fig. 3C; data for FIX ⌬EGF1 are in Fig. 3D; data for FX WT are in Fig. 3E; and data for FX PCEGF1 are in Fig.  3F. As is the case with the enzyme assays, the rate of activation in decreasing order for each FIX protein was FIX NP (not shown) Ϸ FIX WT Ͼ FIX Q50P Ͼ FIX PCEGF1 Ϸ FIX ⌬EGF1 . Further, the rate of activation of FX PCEGF1 as analyzed by SDS-PAGE was also considerable slower than the FX WT .

Rates of Activation of FX Proteins by FVIIa⅐PL in the Absence of TF-The activation of FIX or FX mutants by FVIIa⅐TF or
FVIIa⅐sTF with or without PL is impaired (Figs. 1-3). Next we measured the activation of FX mutants by FVIIa⅐PL without TF (or sTF) as outlined earlier (57,58). These results are depicted in Fig. 4. The activation rate was 0.19 Ϯ 0.02 nM/min for FX WT and 0.18 Ϯ 0.02 nM/min for FX PCEGF1 . Thus, the interaction between FX PCEGF1 and FVIIa⅐PL is normal in the absence of TF (or sTF). We were unable to measure the rates of activation of FIX under these conditions. Rate of FVII Activation by FIXa and FXa-FIXa and FXa are the enzymes that convert FVII to FVIIa (5-7). This activation requires only Ca 2ϩ and PL without the need for protein cofactors. Thus, this analysis should provide important data regarding whether or not FIXa and FXa mutants can function as enzymes in activating FVII to FVIIa. Such experiments are presented in Fig. 5. An examination of the results reveals that the activation rate of FVII by each FIXa or FXa protein is similar. The activation of FVII was also confirmed by SDS-PAGE (Fig. 5, inset). These data provide evidence that the mutant proteins are folded correctly and that the interaction of mutant enzymes (FIXa mutants and FXa PCEGF1 ) with their substrate FVII is normal. It also indicates that the EGF1 domain in FIXa or FXa does not play an important role in the activation of FVII.
Binding of FIXa and FXa Proteins to AT-AT binds to the heavy chain of FIXa or FXa. It forms a covalent bond with the active site serine of FIXa as well as that of FXa (8,9). Whether or not the mutations in the EGF1 domain in FIXa or FXa influence AT binding was examined by studying the tight complex formation of FIXa or FXa with AT. The rate of tight complex formation between AT and FIXa or FXa mutants appeared to be similar to that of FIXa WT and FXa WT as analyzed by SDS-PAGE. In each case, the FIXa⅐AT or FXa⅐AT tight complex formed at 0.15 min was ϳ20%, at 0.5 min it was ϳ50%, at 1.5 min it was ϳ80%, and at 4 min it was essentially complete. These initial data indicate that the EGF1 domains of FIXa and FXa are most likely not involved in AT binding and imply that the protease domains of the mutant FIXa and FXa are correctly folded to bind to their natural inhibitor. These data are consistent with the previous observation that the protease domain in these enzymes plays an important role in binding to AT (8,9).
Activation of FX WT and FX PCEGF1 by FIXa⅐PL and FIXa⅐PL⅐FVIIIa-In the coagulation cascade, FX can be activated either by the extrinsic pathway or by the intrinsic pathway. Studies presented above reveal that the activation of FX PCEGF1 by the extrinsic pathway is impaired. Here, we investigated whether or not the activation of FX PCEGF1 by the intrinsic pathway is normal. When the activation was carried out in absence of FVIIIa, the initial rate of activation for FX WT was 0.096 nM/min, and for FX PCEGF1 it was 0.073 nM/min. These data are presented in Fig. 6A. When the activation was carried out in the presence of FVIIIa, the rates of activation for FX WT and FX PCEGF1 were the same (2.2 nM/min). These data are presented in Fig. 6B. It should be noted that FX NP was activated by FIXa⅐PL with or without FVIIIa at the same rate as obtained with FX WT . These results indicate that the EGF1 domain of FX plays virtually no role or a very minor role in the activation of FX by the intrinsic pathway.
Ligand Blotting of FIX or FX Proteins to sTF or FVIIa on Nitrocellulose Membrane-FIXa and FXa mutants function normally as enzymes in activating FVII and in binding to the serpin inhibitor AT. Thus, our results indicate that the defect in FIX and FX mutants is in their binding either to FVIIa or to TF (or both) in the FVIIa⅐TF complex. To distinguish between these possibilities, we performed SDS-PAGE and transferred the FVIIa and sTF to the nitrocellulose membrane. Each FIX mutant was then used as a ligand to probe FVIIa or sTF on the membrane. When we probed the membrane with FIX WT , it bound to FVIIa as well as to sTF (Fig. 7A). In these experiments, FIX Q50P , FIX PCEGF1 , and FIX ⌬EGF1 also bound to FVIIa but not to sTF (Fig. 7, B-D). These results indicate that the EGF1 domain of FIX is important for its binding to TF. We also performed similar ligand blotting experiments using FX as a ligand. Both FX WT and FX PCEGF1 bound to FVIIa, whereas only FX WT bound to the sTF (Fig. 7, F and G). These results indicate that the EGF1 domain of FX, like that of FIX, is important in its binding to TF.
Affinity of sTF for Various Proteins-The Western ligand blot assays described above provide qualitative data for binding of sTF to various FIX and FX mutants. To obtain quantitative information, we studied binding of 125 I-FIX and 125 I-FX to immobilized sTF. A direct binding plot for FIX is presented in Fig. 8A and the K d (app) obtained from this plot for binding of 125 I-FIX to sTF was 130 Ϯ 20 nM. The affinity of WT and of each mutant FIX was then obtained by its ability to compete with 125 I-FIX in binding to sTF. These data are presented in Fig. 8B. Analysis of these data reveal that sTF interacts with FIX WT with K d (app) of ϳ150 nM, FIX Q50P with K d (app) of ϳ500 nM, FIX PCEGF1 with K d (app) of ϳ1500 nM, and FIX ⌬EGF1 with K d (app) of ϳ1600 nM. Thus, as compared with FIX WT , the point mutant (FIX Q50P ) has ϳ3-fold reduced affinity, whereas the replacement (FIX PCEGF1 ) or the deletion (FIX ⌬EGF1 ) mutant has ϳ10-fold reduced affinity for binding to sTF.
A direct binding plot for binding of 125 I-FX to sTF is presented in Fig. 9A and the K d (app) obtained from this plot for binding of FX to sTF was 500 Ϯ 100 nM. The affinity of WT and PCEGF1 mutant was then determined by competition with 125 I-FX for binding to sTF. These data are shown in Fig. 9B. The K d (app) for WT was ϳ900 nM, and the K d (app) for FX PCEGF1 was ϳ6 M in binding to sTF. Thus, compared with WT, the FX PCEGF1 has significantly reduced affinity for sTF.
Studies with FIX-EGF1 Synthetic Domain-Data presented thus far indicate that mutations in the EGF1 domain of FIX and FX impair their abilities to be activated by the FVIIa⅐sTF complex, and this property may be related to diminished binding of the mutants to sTF in the FVIIa⅐sTF. To examine whether isolated EGF1 domain binds to sTF, we synthesized the 45-87-residue segment of FIX representing its EGF1 domain. The fully reduced peptide had the correct mass (4755 Da) indicating no error in synthetic steps. The peptide was then oxidized and purified as outlined under "Experimental Procedures." The mass spectrometric analysis of the oxidized purified peptide is shown in Fig. 10. A peak corresponding to the correct mass of 4749 Da represented Ͼ90% of the total molecular species. This analysis also indicated that six cysteines had been oxidized to yield three disulfide pairs. Since only properly disulfide-paired and correctly folded EGF1 domain binds Ca 2ϩ (35), we studied the binding of Ca 2ϩ to the folded peptide. These data are presented in Fig. 10 (inset) using a Ca 2ϩ -specific electrode. This peptide contained a single Ca 2ϩ binding site (0.95/mol) with a K d of ϳ0.6 mM, which is consistent with correct folding of the domain. Importantly, the folded peptide inhibited the binding of FIX to sTF with a K i of ϳ60 M (Fig.  8B). Moreover, reduced and carboxymethylated peptide prepared essentially by the method of Sodetz and Castellino (40) neither bound Ca 2ϩ (data not shown) nor inhibited FIX binding to sTF (Fig. 8B). DISCUSSION The FVIIa⅐TF complex has high specificity and affinity for its substrates FIX and FX, which quite possibly involve regions that are remote from the cleavage sites. The peptides that do not bind to the active site of FVIIa specifically inhibit the activation of FX by FVIIa⅐TF as well as the TF-initiated clotting (41). Further, when the active site of FVIIa is blocked with peptidyl substrates, the resulting FVIIa⅐TF complex has nearly the same affinity toward its substrate FX as the uninhibited FVIIa⅐TF (42). Thus, exosites exist in FX that are responsible for this specific recognition of FX by the FVIIa⅐TF complex. Moreover, a FVIIa molecule with a point mutation in the Gla domain (Arg 36 3 Ala) activates FX at a reduced rate in the presence but not in the absence of TF (15). These results indicate that TF facilitates an optimal interaction between the Gla domains of FVIIa and FX (15). Further, TF residues Lys 165 and Lys 166 that are exposed in the crystal structure of the FVIIa⅐TF complex (13) are thought to interact with the Gla domain of FX or FIX (43,44). These TF mutants were deficient in supporting FVIIa activation of normal FX (or FIX) but not of the Gla domainless FX (43). In support of this, when Kirchhofer et al. (14) made a panel of additional TF mutants involving surface-exposed residues 157-185, they found that the mutant TF molecules had reduced affinity for FIX and FX. Based upon the topology of FVIIa⅐sTF complex (13), these residues in the C-terminal domain of TF would appear to interact with the Gla domain of substrates FIX and FX.
The EGF1 domain of FIX (or FX) also appears to play an important role in its activation by FVIIa⅐TF. In an earlier study when FVIIa⅐TF⅐PL was used as the activator, it appeared that in this system FIX ⌬EGF1 could not be activated. Further, FIX PCEGF1 was activated at a slower rate, whereas FIX Q50P was activated at a nearly normal rate in this system (16). The data in these studies were analyzed by SDS-PAGE of the reaction mixtures drawn at different times and thus were only qualitative. Further, the failure of activation of FIX ⌬EGF1 in this system could stem from the altered spatial alignment of domains in this mutant. To circumvent this problem, we have now developed an assay to quantitatively measure the formation of FIXa and have conducted studies using the FVIIa⅐sTF system, where anchoring of the Gla domain on the PL surface is not required. Additionally, we have performed studies using FX PCEGF1 and compared its activation properties with FX WT . Our data indicate that the activation rates of FIX Q50P , FIX PCEGF1 , FIX ⌬EGF1 , and FX PCEGF1 are impaired both by the FVIIa⅐TF (Fig. 1) and by the FVIIa⅐sTF (Figs. 2 and 3). A recent report revealed that mutations at residue 48 in FIX result in delayed activation by FVIIa⅐TF (45). This observation is consistent with the extensive kinetic data presented in this paper.
In ligand blot assays, all EGF1 FIX and FX mutants bound to FVIIa but not to TF (Fig. 7). However, results obtained by Western ligand blot assays are at best qualitative in nature. Further, such data do not allow estimation of the binding energy involved in complex formation. For these reasons, we measured the affinity of WT and mutant FIX and FX proteins for binding to immobilized sTF. These data indicate that FIX WT binds to sTF with K d (app) of ϳ150 nM. The affinity of the point mutant (FIX Q50P ) for sTF was reduced 3-fold, whereas the affinity of the replacement (FIX PCEGF1 ) or the deletion mutant (FIX ⌬EGF1 ) was reduced ϳ10-fold (Fig. 8). Similarly, the affin-   2 g), FIX (20 ng), or FX (20 ng) was electrophoresed on a 12% SDS-PAGE gel. The proteins were then transferred to nitrocellulose membranes. Each protein that was electrophoresed and blotted onto the nitrocellulose membrane is labeled at the top of each panel. The FIX proteins that were used as ligands to probe the membranes were FIX WT (A), FIX Q50P (B), FIX PCEGF1 (C), and FIX ⌬EGF1 (D). The FX proteins that were used as ligands to probe the membranes were FX WT (F) and FX PCEGF1 (G). In panels E and H, each labeled as buffer control, 1% milk solution was used to probe the membranes instead of the ligand FIX or FX, respectively. The FIX (or FX) bound to sTF and FVIIa was detected by FIX mAb (or FX mAb) as outlined under "Experimental Procedures." ity for sTF of FX PCEGF1 mutant was reduced ϳ6 -10-fold as compared with FX WT (Fig. 9). Of interest is the observation that FIX WT bound to sTF with ϳ3-fold higher affinity than FX WT . The reason(s) for this observation is not known. Exten-sive additional data are needed to understand whether there are real differences in the affinities of FIX WT and FX WT in binding to sTF or if they simply reflect experimental difficulties inherent to the technique employed. However, the conclusions drawn from this paper do not depend upon resolution of this issue.
The studies conducted with the EGF1 domain mutants yield data pertaining to loss of function. Direct evidence that the EGF1 domain is involved in binding to sTF comes from the ability of synthetic FIX-EGF1 domain to inhibit the interaction of FIX WT with sTF with a K i ϳ60 M. An interesting point emerges from such studies. FIX protein lacking the EGF1 domain or having the EGF1 domain of protein C only has a 10-fold reduced affinity for sTF (Fig. 8B). This information coupled with the basic thermodynamic principles involving equilibrium constants predicts that the isolated FIX-EGF1 domain should bind to sTF with a K d in the mM range. However, such is not the case, and the isolated properly folded FIX-EGF1 domain binds sTF with 60 M K d . Two explanations may be forwarded to explain this phenomenon. First, the binding of FIX to sTF involves two regions (Gla and EGF1 domains) and conformational strain occurs upon binding to the second site either in FIX or sTF. This would yield lower K d values for the isolated two fragments (FIX lacking the EGF1 domain and the EGF1 domain alone) due to the absence of steric constraints inherent in the full-length molecule. An alternative explanation might be that the synthetic EGF1 domain we have used is devoid of glycosylation, which might interfere with binding to sTF. Further studies are needed to address this issue. However, we currently prefer the first explanation.
These impaired interactions of mutants with FVIIa⅐TF (or sTF) are not due to defective folding of the proteins. All activated mutants (FIX by FXIa, and FX by RVV) bound AT and activated FVII to FVIIa normally. Furthermore, FX PCEGF1 could be activated by the FIXa⅐FVIIIa⅐PL complex at the same rate as FX WT (Fig. 6). These data are consistent with the observation of Lapan and Fay (46), who found that the protease domain of FX interacts with FVIIIa in the FIXa⅐FVIIIa complex. Importantly, our data indicate that the EGF1 domain of FIX or FX contains an exosite(s) that appears to directly interact with TF.
The EGF1 domain in FIXa or FXa also contributes to the assembly of FIXa⅐FVIIIa or FXa⅐FVa on the PL surface. Data exist that suggest that the EGF1 domain in FIXa or FXa may FIG. 9. Binding of FX WT and FX PCEGF1 to sTF. The data in A represent direct specific binding of 125 I-FX to immobilized sTF. An approximate K d (app) value for FX WT and sTF interaction was calculated to be ϳ500 nM. The data in B depict the abilities of FX WT and FX PCEGF1 to inhibit the binding of 125 I-FX to sTF. Competitors were FX WT (open circles) and FX PCEGF1 (closed circles). The concentration of 125 I-FX used was 30 nM with increasing amounts of the competitors. The curves represent best fit to the IC 50 four-parameter logistic equation of Halfman (38) .   FIG. 10. Mass spectrometry and Ca 2؉ binding analyses of the folded FIX-EGF1 domain. The deconvoluted spectrum of the peptide obtained using a Finnigan LCQ Iontrap Electrospray mass spectrometer is depicted. The inset shows binding of Ca 2ϩ to the folded peptide as determined by a Ca 2ϩ -specific electrode. The concentration of the peptide used was 400 M, and the free Ca 2ϩ concentration is plotted against r (mol of Ca 2ϩ bound/ mol of peptide). directly interact with its cofactor FVIIIa or FVa, respectively (47,48). However, data also exist that indicate that the EGF1 domain primarily serves as a spacer to properly align the protease domain in FIXa or FXa for optimal interaction with the cofactor (30,49). In support of this, a recombinant FIX molecule in which the EGF1 domain was replaced with that of FX had normal clotting activity (50), and another recombinant FIX molecule in which the EGF1 domain was replaced with that of FVII had 2-fold increased clotting activity (51). These data suggest that there are no unique determinants in the EGF1 domain of FIX that cannot be replaced by that of FX or FVII. Moreover, to study the functions of EGF domains in protein C, a replacement mutant in which both EGF domains were from FIX was constructed. This mutant was activated by thrombomodulin-thrombin complex at ϳ70% of the rate obtained with the wild type protein C (52). The activated protein C mutant was also defective in inactivating FVa and FVIII/ FVIIIa in a PL-containing system (52). The decreased activity of activated protein C mutant was attributed to direct proteinprotein interaction and/or to misalignment of domains/recognition sites with its physiological substrates. Thus, the EGF1 domain in each vitamin K-dependent protease may be involved in direct binding as well as in specific alignments of recognition motifs with other proteins involved in the assembly. We are currently examining the role of EGF1 domain of FXa in the activation of prothrombin in the presence and absence of FVa and PL using our FX PCEGF1 mutant.
Last, we have made attempts to define an exosite in the EGF1 domain of FIX or FX that may interact with TF. Three important considerations were given in search for such an exosite. First, since FVIIa⅐TF can activate FVII⅐TF efficiently (53), we opted to select regions that are not involved in binding of FVIIa EGF1 domain to TF. This excluded the interface region involving residues Gln 64 , Ile 69 , Phe 71 , Glu 77 , and Arg 79 of FVIIa (54) and by inference of FIX and FX. Further, this interface is not structurally conserved in FIXa and FXa. Second, we excluded side chains of those residues (Asn 47 , Gln 50 , and Asp 64 ) that are involved in binding to Ca 2ϩ (54). Third, our data indicate that FIX Q50P mutant in which the Ca 2ϩ -binding site is impaired is defective in binding to TF. Thus, we examined the region surrounding the Ca 2ϩ site that might be perturbed and therefore a likely candidate for binding to TF. Using these three criteria, we postulate an extended exosite region in the EGF1 domain of FIX as well as in FX that could be involved in binding to TF. This is shown in Fig. 11. Residues Asp 49 , Glu 52 , Ser 53 , Asn 58 , Phe 77 , Asn 81 , and Glu 83 are located on the surface and may be the key determinants for binding to TF. Numerous point mutations in these residues cause hemophilia B (55). Studies are in progress to mutate these residues to examine if they are indeed involved in TF binding.