Residues 88-109 of factor IXa are important for assembly of the factor X activating complex.

Activated platelets and phospholipid vesicles promote assembly of the intrinsic factor X (FX) activating complex by presenting high-affinity binding sites for blood coagulation FIXa, FVIIIa, and FX. Previous reports suggest that the second epidermal growth factor (EGF)-like domain of FIXa mediates assembly of the FX activating complex (Ahmad, S. S., Rawala, R., Cheung, W. F., Stafford, D. W., and Walsh, P. N. (1995) Biochem. J. 310, 427-431; Wong, M. Y., Gurr, J. A., and Walsh, P. N. (1999) Biochemistry 38, 8948-8960). To identify important residues, we prepared several chimeric FIXa proteins using homologous sequences from FVII: FIXa(FVIIEGF2) (FIX Delta 88-124,inverted Delta FVII91-127), FIXa(loop1) (FIX Delta 88-99,inverted Delta FVII91-102), FIXa(loop2) (FIX Delta 95-109,inverted Delta FVII98-112), FIXa(loop3) (FIX Delta 111-124,inverted Delta FVII114-127), and point mutants (FIXaR94D and FIXa(loop1)G94R). In the presence and absence of FVIIIa, a 2- to 10-fold reduced V(max) of FX activation (nm FXa min(-1)) was observed for FIXa(FVIIEGF2), FIXa(loop1), FIXa(loop2), and FIXa(loop1)G94R, whereas FIXa(loop3) and FIXaR94D were normal. For all of the FIXa proteins, K(m)((app)) values were normal as were EC(50) values for interactions with FVIIIa. However, K(d)((app)) (in nm) for the FX activating complex assembled on phospholipid vesicles was increased for FIXa(FVIIEGF2) (43.3 +/- 2.70), FIXa(loop1)(10.9 +/- 2.8), FIXa(loop2) (70.5 +/- 1.60), and FIXa(loop1)G94R (17.1 +/- 2.90) relative to FIXa(N) (3.9 +/- 0.11), FIXa(WT) (4.6 +/- 0.17), FIXa(loop3) (4.5 +/- 0.20), and FIXaR94D (2.2 +/- 0.09) suggesting that reduced V(max) is a result of impaired complex assembly. These data indicate that residues 88-109 (but not Arg(94)) are important for normal assembly of the FX activating complex on phospholipid vesicles.

Factor IX (FIX) 1 is a plasma glycoprotein required for normal hemostasis (1,2). Individuals with a deficiency or a dysfunction of FIX exhibit a bleeding tendency-hemophilia B (3). The mature form of FIX contains 415 amino acids and ϳ18% carbohydrate by mass. FIX circulates as a single chain zymogen until activation by either FXIa or FVIIa/tissue factor (FVIIa/TF) (4,5). FIX is activated by sequential cleavage at Arg 145 and Arg 180 to generate FIXa. The light chain of FIXa contains the ␥-carboxyglutamate-rich domain (Gla domain) and two consecutive domains with homology to epidermal growth factor (EGF1 and EGF2), whereas the heavy chain is the trypsin-like serine protease domain. Residues 146 -180 comprise the carbohydrate-rich activation peptide (6,7).
FIXa is a serine protease that participates in the intrinsic pathway of blood coagulation. FIXa activates FX as part of the intrinsic FX activating complex. The FX activating complex consists of FIXa, FVIIIa (a nonenzymatic cofactor), and FX (the normal macromolecular substrate) assembled on a procoagulant surface (8,9). Under physiological conditions, thrombinactivated platelets or endothelial cells can provide this surface (10 -13). Additionally, synthetic phospholipid vesicles containing phosphatidylserine can also support assembly of the FX activating complex and activation of FX (8,14). Assembly of the surface-bound FX activating complex results in a dramatic increase (ϳ20 million-fold) in the catalytic efficiency (k cat /K m ) of FX activation versus that of FIXa alone in solution (9,15). The surface localization of FIXa is a requisite step in the catalytic enhancement of FIXa when assembled in the FX activating complex (9,14).
Recent investigations have focused on the contributions of the EGF-like domains to FIXa biochemistry. Analysis of patient data suggests that residues contained within EGF1 and EGF2 are important for FIX(a) functions. Several point mutations within the EGF-like domains result in dysfunctional FIX(a) activity and a bleeding tendency (3,16). Investigations of the EGF-like domains have found several important functions for these domains. EGF1 is important for interactions with FVIIa/TF (17), FVIIIa (18), and FX (19,20). EGF1 does not appear to be important for binding to surfaces such as phospholipids (21), platelets, (22), or endothelial cells (23). While EGF2 has not been extensively characterized, experiments have indicated its importance for activation of FX. A chimeric FIXa protein in which both EGF1 and EGF2 domains were replaced by those of FX possessed about 4% clotting activity (24). In contrast, a chimeric protein with the FX EGF1 domain but the wild type FIX EGF2 domain was entirely normal (24), suggesting an essential function for EGF2. More recent experiments from our laboratory have indicated that EGF2 is likely involved in mediating surface binding to both platelets and phospholipids (25,26).
To further define the contribution of the EGF2 domain to surface binding, we have prepared several chimeric FIXa proteins (see Fig. 1) using homologous sequences from FVII. These FIXa proteins include FIXa FVIIEGF2 (FIX⌬88 -124,ٌFVII91-127), 2 FIXa loop1 (FIX⌬88 -99,ٌFVII91-102), FIXa loop2 (FIX⌬95-109,ٌFVII98 -112), and FIXa loop3 (FIX⌬111-124,ٌFVII114 -127). The point mutants (FIXaR94D and FIXa loop1 G94R) have been prepared to address the functional importance of Arg 94. These chimeras and mutant FIXa proteins were tested for their ability to activate FX as part of the FX activating complex assembled on phospholipid vesicles. From these studies we have determined that residues 88 -109 (but not Arg 94 ) in the EGF2 domain are important for phospholipid surface assembly of the FX activating complex.

MATERIALS AND METHODS
The mammalian expression vector pCMV5 was obtained as a generous gift from Dr. David Russel from the University of Texas Southwestern Medical Center (Dallas, TX). The cDNA encoding human FIX and the cDNA encoding the human vitamin K-dependent ␥-glutamyl carboxylase (pCMV.hgc) were obtained as generous gifts from Dr. Darrel Stafford from the University of North Carolina (Chapel Hill, NC). The cDNA encoding human FVII (in the vector pUC19-FVII) was supplied from American Type Culture Collection (59790, Manassas, VA). Cloned DNA polymerase from Pyrococcus furiosis and reaction buffer were purchased from Stratagene (La Jolla, CA). Oligonucleotides were purchased from Invitrogen. Materials for plasmid purification, PCR product purification, and transfection of mammalian cells were purchased from either Promega (Madison, WI) or Qiagen (Hilden, Germany).
Transformed human embryonic kidney cells (HEK293) were obtained from American Type Culture Collection (CRL1573). Dulbecco's modification of Eagle's medium was purchased from Mediatech (Herndon, VA). Vitamin K 1 (2-methyl-3-phytyl-1,4-naphthoquinone) was purchased from Abbott Laboratories (Chicago, IL). Other tissue culture reagents were purchased from either Sigma or Invitrogen. Antibodies for FIX enzyme-linked immunosorbent assay were purchased from Enzyme Research Laboratories (South Bend, IN). Q-Sepharose anion exchange resin was purchased from Sigma. Centri-prep 3 concentration units were purchased from Amicon Division, W. R. Grace and Co. (Danvers, MA).
Human FIX, human FX, human antithrombin III (ATIII), and the FX activator from Russell's Viper (Vipera russeli) venom were purchased from Enzyme Research Laboratories (South Bend, IN). The FX preparation, obtained as a lyophilized powder, was dissolved in sterile water and dialyzed against HEPES-Tyrodes buffer (HT) before being used in experiments. Human FXIa was purchased from Haematologic Technologies (Essex Junction, VT). High-purity recombinant human FVIII was obtained as a generous gift from Baxter Healthcare Corp. (Duarte, CA). Thrombin was purchased from Sigma. The chromogenic substrate S-2765 (N-␣-benzyloxy-carbonyl-D-arginyl-glycyl-L-arginine-para-nitroanalide-dihydrochloride) was purchased from Dia Pharma Group (Stockholm, Sweden).
Expression and Purification of FIX Proteins-Stable cell lines expressing the various FIX chimeras and mutants were prepared by transfection of HEK293 cells using the calcium phosphate precipitation technique with commercially available reagents (Promega, Madison, WI) and the manufacturer's protocol. Transfection experiments included the appropriate FIX expression vector, pCMV.hgc, and a neomycin resistance vector (pSV2Neo). G418-resistant cell lines were subcultured with the aid of sterile polyester swabs (Hardwood Products Company, Guilford, ME). Expression of FIX in the medium was determined by enzyme-linked immunosorbent assay as previously described (26). Cell lines exhibiting maximal expression of FIX proteins were expanded for preparative-scale protein expression. FIX proteins were purified from serum-free conditioned medium using Q-Sepharose chromatography as described previously (26,30).
Q-Sepharose was used as a "pseudoaffinity" strategy for the purification of recombinant FIX proteins from serum-free conditioned medium. Serum-free conditioned medium was adjusted to contain 5 mM benzamidine and 5 mM EDTA, centrifuged (6000 ϫ g), and filtered through cellulose acetate membranes (0.22-m pore size) to remove cell debris. Q-Sepharose was pre-equilibrated with TBS supplemented with 2 mM benzamidine and 2 mM EDTA. The medium was chromatographed, and the resin was washed with the equilibration buffer. EDTA was removed from the resin by washing with TBS supplemented with 2 mM benzamidine. FIX proteins were eluted from the column with TBS supplemented with 2 mM benzamidine and 5 mM CaCl 2 . The FIX proteins recovered from the Q-Sepharose column were dialyzed against HT and concentrated using Centri-prep 3 concentrators. FIX proteins were stored in small aliquots under liquid nitrogen.
␥-Carboxyglutamate Analysis-␥-Carboxyglutamate analysis was performed as described by Przysiecki (31). It was generously performed by Dr. Rodney Camire in the laboratory of Dr. Katherine High (Children's Hospital of Philadelphia, Philadelphia, PA).
Preparation of Phosphatidylserine/Phosphatidylcholine (PSPC) Vesicles-Extruded phospholipid vesicles composed of (mol/mol) 25% phosphatidylserine and 75% phosphatidylcholine were prepared according to the following protocol: dioleoyl-phosphatidylcholine and porcine brain phosphatidylserine were mixed in a 3:1 molar ratio and dried under nitrogen. The lipids were redissolved in benzene, frozen on dry ice, and lyophilized until dry. The lipids were resuspended in HT and incubated on ice for 30 min with intermittent mixing. The lipids were subjected to a freeze-thaw cycle (dry ice, 37°C) five times and extruded by passage through three stacked polycarbonate membrane filters (100-nm pore size) a total of six times. The final lipid concentration was ϳ2 mM.
Activation of FIX Proteins by FXIa-FIX proteins were activated to their FIXa forms as follows: FIX proteins were diluted in HT supplemented with 10 mM CaCl 2 . FXIa was added to a 1/200 molar ratio, and the reactions were incubated at 37°C for 90 min. Complete activation was judged by SDS-PAGE/silver staining and by active site titration with ATIII. FIXa proteins were stored in small aliquots at Ϫ80°C Western Immunoblotting-Activated FIX proteins were separated on a 8-16% polyacrylamide gradient gel. The gel was placed in transfer buffer (48 mM Tris, 30 mM glycine, 0.037% w/v SDS, 20% v/v methanol) for 20 min with gentle agitation. Immobilon-P PVDF membrane (Millipore) and filter paper were cut to fit the gel. Filter paper was wetted with transfer buffer. PVDF was wetted with methanol then transfer buffer. The transfer of proteins from the gel to the PVDF was performed in a Transblot SD semi-dry transfer cell (Bio-Rad) at constant current (0.8 mA per cm 2 ) for 20 min. The PVDF was agitated in blocking buffer (10% Irish cream in TBS) overnight while shaking. The PVDF was incubated for 1 h with 1:1000 dilution (0.5 mg/ml stock) of affinitypurified goat anti-human-FIX polyclonal antibody conjugated to horse-radish peroxidase in TBS-Tween (TBS supplemented with 0.1% v/v Tween 20) at 4°C.
After incubation with the horseradish peroxidase-conjugated antibody, the PVDF was washed two times for 5 min each in TBS-Tween. A commercially available substrate (Kirkegaard and Perry Laboratories, Gaithersburg, MD) was used for visualization according to the following protocol: 3,3Ј,5,5Ј-tetramethylbenzidine (0.4 g/L) solution was mixed with an equivolume of hydrogen peroxide (0.02% in citric acid buffer) and 0.1 volume of membrane-enhancing reagent. The PVDF was incubated at room temperature with the substrate solution until the desired intensity of the bands was achieved. The reaction was terminated by rinsing the membrane with distilled water, and the PVDF membrane was allowed to dry at room temperature.
Active Site Titration of FIXa Proteins-FIXa was active site-titrated by ATIII according to the following method (32): ATIII dilutions (0 -100 nM) and FIXa proteins (ϳ100 nM) were prepared in HT supplemented with BSA (0.5 mg/ml), heparin (20 g/ml), and 5 mM CaCl 2 . FIXa dilution was added to a tube containing the ATIII dilution, incubated at 37°C for 20 min, then diluted 10-fold using HT supplemented with BSA (0.5 mg/ml). Residual FIXa activity was determined by assaying FXa generation activity in the following assay conditions: 5 mM CaCl 2 , 400 nM FX, FVIII (5 units/ml), and 20 M PSPC vesicles. The reactions were initiated by addition of thrombin to 1 unit/ml. After 2 min, the reactions were terminated by the addition of stopping buffer (50 mM HEPPS, pH 8.1, 175 mM NaCl, and 20 mM EDTA). FXa generated was determined by hydrolysis of the chromogenic substrate S-2765. S-2765 dissolved to 600 M in 50 mM HEPPS, pH 8.1, and 175 mM NaCl was mixed with the sample in micrometer wells, and change in absorbance at 405 nm was monitored immediately.
Assay of FXa-FXa was assayed by hydrolysis of the FXa-specific chromogenic substrate S-2765. FXa (50 l) was added to wells of a microtiter plate, and S-2765 (50 l) was added to a final concentration of 350 M. Change in absorbance at 405 nm was monitored immediately. Unknown FXa concentrations were determined by comparison to a standard curve prepared with known dilutions of FXa.
Determination of K m(app) and V max in the Absence of FVIIIa-FIXa proteins were added to 10 nM in reaction vessels containing 20 M extruded PSPC vesicles in HT supplemented with BSA (0.5 mg/ml) and 5 mM CaCl 2 . FIXa proteins were incubated with the PSPC vesicles at 37°C for 2 min. The reactions were initiated by addition of FX to the indicated final concentrations. After 20 min at 37°C, the reactions were stopped by addition of EDTA to 10 mM. FXa was determined as described above. Velocity of FXa generation (nM FXa/min) was plotted as a function of input FX concentration (nM FX). Values of K m(app) and V max were determined as described below.
Determination of K m(app) and V max in the Presence of FVIIIa-FIXa proteins were added to 15 nM in reaction vessels containing 1 M extruded PSPC vesicles in HT supplemented with BSA (0.5 mg/ml) and 5 mM CaCl 2 . FIXa proteins were incubated with the PSPC vesicles at 37°C for 2 min. FVIII was activated to FVIIIa (5 units/ml final) by 0.1 unit/ml thrombin for 1 min immediately before addition to the reaction vessels. After a 1-min incubation at 37°C, the reactions were initiated by addition of FX to the indicated final concentrations. After 2 min at 37°C, the reactions were stopped by addition of EDTA to 10 mM. FXa was determined as described above. Velocity of FXa generation (nM FXa/min) was plotted as a function of input FX concentration (nM FX). Values of K m(app) and V max were determined as described below.
Determination of EC 50 and Velocity in the Presence of Saturating FVIIIa (V max 8 ) for Stimulation of FXa Generation-FVIIIa was titrated into the FX activation reaction to determine the concentration that results in 50% maximal stimulation of rate of activation (EC 50 ) and the maximum velocity observed when FVIIIa is at saturating concentrations (V max 8 ). FIXa proteins were added to 5 nM in reaction vessels containing 500 nM extruded PSPC vesicles in HT supplemented with BSA (0.5 mg/ml) and 5 mM CaCl 2 . FVIII was added to the indicated concentration and activated by 0.1 units/ml thrombin for 1 min. The reactions were initiated by addition of FX to 250 nM. After 2 min at 37°C, the reactions were stopped by addition of EDTA to 10 mM. FXa was determined as described above. Velocity of FXa generation (nM FXa/min) was plotted as a function of input FX concentration (nM FX). Values for EC 50 and V max 8 were determined as described below. Determination of K d (app) and Velocity in the Presence of Saturating FIXa (V max 9 )-Various concentrations of FIXa proteins were incubated with 1 M PSPC vesicles in HT supplemented with BSA (0.5 mg/ml) and 5 mM CaCl 2 for 2 min at 37°C. FVIII was activated to FVIIIa (5 units/ml final) by thrombin (0.1 units/ml) for 1 min at 37°C immediately before addition to the reaction mixes. The reactions, initiated by addition of FX to 250 nM, continued for 2 min at 37°C and were stopped by addition of EDTA to 10 mM. FXa was determined as described above. Velocity of FXa generation (nM FXa/min) was plotted as a function of input FIXa concentration (nM FIXa). K d(app) was determined as described below.
FX Titration in the Absence of a Surface-FIXa proteins were added to 50 nM in reaction vessels containing HT supplemented with BSA (0.5 mg/ml) and 5 mM CaCl 2 . FIXa proteins were incubated at 37°C for 2 min. The reactions were initiated by addition of FX to the indicated final concentrations. After 1 h at 37°C the reactions were stopped by addition of EDTA to 10 mM. FXa was determined as described above. Velocity of FXa generation (nM FXa/min) was plotted as a function of input FX concentration (nM FX). Data were analyzed as described below.
Data Analysis-Rates of FXa generation from all reactions were fitted to a rectangular hyperbola curve using a non-linear least squares fit. Constants for K m , V max , EC 50 , V max 8 , K d(app) , and V max 9 were derived using the same Kalaidegraph Software. Between-group differences were tested for statistical significance using analysis of variance followed by pair-wise comparisons with the Bonferroni adjustment procedure for multiple comparisons maintaining an experiment-wise Type 1 error level of 0.05 (33). For the solution-phase activation of FX, data analysis was performed by least squares regression. A separate regression analysis was performed for each protein. Slopes and intercepts were tested for significance along with linearity of regression. Ninetyfive percent confidence limits were generated for each slope and intercept. Differences between slopes were tested using an independent t test with a Bonferroni adjustment to correct for multiple comparisons (34). Differences were considered significant if the probability of chance occurrence was less than or equal to 0.05. 9 Values-Predicted V max values (see Table II) are shown for comparison with experimentally determined V max values. The predicted V max values were determined from the following equation:

Prediction of V max Values Derived from Estimated Complex Assembly Determined from K d(app) and V max
is the velocity in the presence of saturating FIXa concentration and K d(app) is the apparent dissociation constant for the binding of FIXa to the factor X activating complex and the input FIXa concentration used in the experiment (15 nM).

RESULTS
All of the chimeric and mutant FIX proteins were expressed in HEK293 cells and purified from conditioned serum-free medium (30). The amino acid sequences of FIX and FVII within the EGF2 domain that were exchanged to prepare the chimeric proteins utilized in this study are shown in Fig. 1. All of the recombinant FIX proteins were found to be Ͼ95% pure as judged by SDS-PAGE and staining with Coomassie Brilliant Blue ( Fig. 2A). Each of the FIX proteins was found to comigrate with plasma-derived FIX (FIX N ) suggesting normal post-translational modifications including removal of the pre-and propeptides and addition of carbohydrate. Furthermore, each of the recombinant FIX proteins was found to have the expected amount of ␥-carboxyglutamate (Gla) modification (10 -12 mol Gla/mol protein) ( Table I).
Each of the FIX proteins was activated with FXIa. Both silver-stained gels and Western blotting of SDS-PAGE showed disappearance of the zymogen band and appearance of heavy and light chains of FIXa. All of the FIX proteins were entirely converted to the FIXa form as judged by SDS-PAGE and silver staining (Fig. 2B). A band corresponding to the molecular mass of the FIXa heavy chain (28 kDa) can be observed for all of the FIXa proteins. The FIXa light chain stains poorly as has been observed by others (17,35). All FIXa proteins were also examined by Western immunoblotting, which confirmed complete conversion to the FIXa form (data not shown). Bands corresponding to the molecular mass of the heavy chain (ϳ28 kDa), the light chain (ϳ18 kDa), and the activation peptide (ϳ10 kDa) were observed with equal intensity for each of the eight proteins. No bands at ϳ70 kDa were detectable either by silver staining or Western blotting indicating complete activation of the FIXa proteins. Complete activation of all eight FIX proteins was also supported by active site titration with ATIII ([FIXa] ϭ ϳ100 nM as estimated from A 280 , Fig. 3, Table I). Starting activities of the proteins at ATIII ϭ 0 reflect differences in their enzymatic activity in the second stage of this assay. Complete ATIII neutralization of each of their activities occurs at 90 -110 nM.
Since effective FX activation involves assembly of the enzymatic complex and flux of FX to FXa, the contribution of each of the loops of the EGF2 domain, as well as that of Arg 94 , to assembly of the FX activating complex was determined from FX titrations on phospholipid vesicles (Fig. 4A). Kinetic constants determined from the isotherms are reported in Table  II. When compared with FIXa N and FIXa WT , the proteins FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R were found to have a reduced V max (2-to 4-fold), whereas FIXa loop3 and FIXaR94D were entirely normal. These observations confirm previously published observations indicating the importance of the FIXa EGF2 domain for assembly of the FX activating complex and activation of FX (25,26).
Interestingly, the R94D mutant showed no reduction in V max . This observation was in contrast to our expectations given that Arg 94 has been implicated to be important for interaction with FVIIIa (36). Consistent with this observation, the FIXa loop1 G94R protein showed no restoration of activity versus the FIXa loop1 mutant. These data suggest that Arg 94 itself is not essential for assembly of the FX activating complex and activation of FX. A discussion of the contrast between our observations and those of other investigators (16,(35)(36)(37) is presented below. While Arg 94 is present within the linear sequence determined to be important for FX activation (residues 88 -109), it does not appear to be essential for this activity.
Since there is no increase in K m for FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R, the reduced V max is not likely to be a result of deficient interactions with FX. K m(app) values for each of these proteins (actually ϳ2to 3-fold lower that those of the FIXa N and FIXa WT ) show no evidence of impaired interaction with FX. Therefore, we considered the possibility that the reduced V max may be a result of impaired ability to interact with either FVIIIa or the phospholipid surface, which would result in a reduced k cat or a reduced effective enzyme concentration, respectively.
In the presence of activated platelets or phospholipids, FVIIIa is known to contribute to a several thousand-fold increase in k cat (9,14,15). The reduced V max values could result from a defective interaction with FVIIIa such that the full increase in k cat is not achieved by the mutants. This hypothesis was tested in two ways. FX titrations were carried out in the  presence of phospholipid vesicles but in the absence of FVIIIa as shown in Fig. 4B and kinetic constants determined from the isotherms are reported in Table II. The proteins FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R were again found to have a reduced V max (2-to 3-fold) in the absence of FVIIIa suggesting that an impaired interaction with FVIIIa cannot explain the reduced V max of FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R. The magnitude of the reduction in V max (2to 3-fold,  Table III). A reduced V max 8 (5-to 10-fold) for FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R was also observed in this assay (Fig. 5B). Deficient interactions with FVIIIa would have been expected to result in an increased EC 50 as well as a reduced V max 8 . We therefore conclude that a deficiency of functional interactions with FVIIIa does not account for the reduced V max of FXa generation for FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R.
Previous investigations suggested that the FIXa EGF2 domain promotes surface assembly of the FX activating complex (25,26). The reduced V max of FIXa FVIIEGF2 , FIXa loop1 , FIXa-loop2 , and FIXa loop1 G94R could be attributed to deficient surface complex assembly. It is predicted by this hypothesis that the activities of all FIXa proteins in the solution phase should be indistinguishable, although in the absence of a surface, an increased K m would be expected (9). When FX titrations were carried out in the absence of a surface (Fig. 6), these data suggested that the FX activating activities of all FIXa proteins are equivalent in the absence of a reaction surface and confirm that interactions with FX are normal.
Since FX activation by the FIXa proteins is nearly equivalent in the absence of phospholipids, the EGF2 domain may be important for complex assembly only in their presence. To determine the affinity of the FIXa proteins for the FX activating complex in the presence of phospholipid, FIXa titrations were carried out, and the K d(app) determined from the isotherms shown in Figs. 7, A and B  Clearly, FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R show a reduced affinity for the FX activating complex as manifested by an increased K d (app) .
Using these data, we were able to predict a V max (in addition to the experimentally determined V max ) for each of the FIXa proteins under the experimental conditions used for our FX titrations in the presence of FVIIIa (Fig. 4A, Table II ) since FX activating activity shows a strong dependence on FIXa complex association. The excellent agreement between the determined V max values and those predicted (Table II) is evidence that the reduction in V max is a consequence of reduced complex assembly on the part of FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R. DISCUSSION Previous investigations by our laboratory have strongly implicated the importance of the EGF2 domain of FIXa in medi-  (Figure 7 and Table IV).    ating surface FX activating complex assembly (25,26). Our objective in this study was to confirm the function of the EGF2 domain and to identify residues within the EGF2 domain that contribute to this interaction. Four chimeric proteins have been developed: FIXa FVIIEGF2 FIXa loop1, FIXa loop2 , and FIXa loop3 to determine the contribution of each of the loops of the EGF2 domain to assembly of the FX activating complex. The loops are defined by the conserved disulfide-pairing characteristic of EGF-like domains. Loop 1 comprises residues 88 -99; loop 2 includes residues 95-109; and loop 3 includes residues 111-124 (Fig. 1).
Structural analysis of FIXa (37) suggests that Arg 94 may form a salt bridge with Glu 78 in EGF1. The Glu 78 -Arg 94 salt bridge could potentially be important for stabilization of the tertiary structure of the FIXa light chain. Hemophilia B patient data (16) and functional studies (36) have suggested that both Glu 78 and Arg 94 may be important residues for activation of FX mainly by promoting interaction with FVIIIa. To determine the contribution of Arg 94 to assembly of the FX activating complex, two point mutations, FIXR94D and FIX loop1 G94R, were prepared.
Optimal rates of FXa generation are achieved only upon assembly of the FX activating complex, which consists of FIXa, FVIIIa, and the substrate, FX, bound to a reaction surface such as phospholipid vesicles or thrombin-activated platelets (2,9,14). Understanding of the interactions of FIXa with the reaction surface is key to understanding assembly of the FX activating complex.
The importance of the Gla domain for the binding of FIXa to phospholipid surfaces, thrombin-activated platelets, and endothelial cells has long been recognized (23, 38 -40). However several observations suggest that domains in addition to the Gla domain can mediate surface binding. A peptide corresponding to residues Gly 4 -Gln 11 of the Gla domain was relatively ineffective as an inhibitor of platelet-mediated FX activation (K i ϭ 165 M) while able to displace ϳ50% of the surface-bound FIXa molecules (K i ϭ 3 nM) (41). FIXa molecules in which the Gla domain was removed by proteolysis, or in which the Gla domain was chemically modified, retained both platelet binding activity (with reduced affinity, K d ϭ 5 nM) and FX activation activity (42). Moreover, chimeric molecules with FVII or FX residues inserted into the Gly 4 -Gln 11 sequence of FIX(a) bind to a reduced number of sites (250 sites/platelet) with decreased affinity (K d ϭ 10 -15 nM) and promote FX activation at a reduced rate, a defect which is entirely explained by the decreased binding affinity (40). It is therefore likely that domains other than the Gla domain can mediate surface binding as well.
The interaction of EGF with the EGF-receptor has been well characterized (43). By virtue of their structural homology with EGF, it was thought that the EGF-like domains may contribute to surface binding (37,44,45). EGF-like domains are thought to be mediators of protein-protein interactions by conservation of this function from EGF. A physiological contribution of the EGF-like domains is suggested by the finding that mutations occurring within these domains are associated with a bleeding tendency (3,16). There has been a recent interest in further characterizing these domains and determining their contribution to FIX(a) procoagulant functions.
The contribution of the EGF-like domains of FIXa to assembly of the FX activating complex is poorly understood. Several investigations have indicated that the EGF1 domain of FIXa does not contribute to surface-binding properties of FIXa. Chimeric proteins in which the EGF1 domain of FX was substituted for that of FX were entirely normal for all functional properties measured including rate of FX activation, platelet binding affinity, and platelet binding stoichiometry (22,24). A chimeric FIXa protein with the EGF1 domain of FVII was reported to have no functional impairments but actually showed an ϳ3-fold increase in catalytic activity (21). In contrast, a FIX chimera in which the EGF1 domain was substituted for that of protein C was found to be impaired in its activation by FVII/TF and impaired (as the enzyme) in its   (17,18). However, these functional deficiencies were found to result from structural abnormalities of the chimeric protein rather than from loss of EGF1 residues involved in functional contacts. Thus the primary contribution of the EGF1 domain to assembly of the FX activating complex appears to be a structural role. Recent investigations from our laboratory have indicated that the EGF2 domain may mediate binding to the reaction surface. Chimeric FIXa proteins in which the EGF2 domain of FX was substituted for that of FIX were found to have a reduced rate of FX activation (25). Further characterization of the FIXa FXEGF2 protein showed that it had a reduced platelet binding affinity and stoichiometry (26). The contribution of the FIXa EGF2 domain was further investigated in this study by preparing several FIXa EGF2 chimeras and mutant proteins. In this paper we have presented the purification, characterization, and evaluation of the kinetic properties of these FIXa proteins.
Our results reported here define a subset of residues within the EGF2 domain (88 -109, excluding Arg 94 ) that are primarily responsible for optimal binding of FIXa to the FX activating complex. However, residues 111-124 did not appear to be important since the FIXa loop3 protein (where residues 111-124 were converted to the FVII sequence) was entirely normal in all assays. It is possible that the FVII residues can effectively substitute for the FIX residues. The consequence of this phenomenon would be to mask any functional contribution of these residues that would have otherwise been detected. However, it is clear from our investigation that residues present within loop1 and loop2 of EGF2 (namely 88 -109, excluding Arg 94 ) are important for FX activating complex assembly.
The observation that Arg 94 is not essential for FX activation complex assembly is surprising. Mutation of either residue (E87K or R94S) has been linked to hemophilia B with near normal antigen concentration (16). Observations by Christophe, et al. (36) have indicated that a salt-bridge between Glu 78 of the EGF1 domain and Arg 94 of the EGF2 domain is important for normal interactions with FVIIIa. While an E78K point mutation resulted in an impaired interaction with FVIIIa, an R94D point mutation did not result in a functional deficiency (36). 3 Subsequent investigations have determined that the R94S substitution introduces a carbohydrate modification to this site and accounts for the functional impairment of the R94S substitution (35) in the hemophilia B patient. The normal behavior of FIXaR94D suggests a compensation mechanism that allows the R94D protein to function normally while the E78K protein is impaired. Collectively, these studies indicate that Arg 94 is not essential for any functional activity studied thus far.
Kinetic analysis of the FIXa proteins has identified a deficiency of FX activation for FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R characterized by a reduced V max of FX activation in the presence of phospholipids (Tables II and III ,. Our data support the conclusion that this kinetic defect results from impaired complex assembly in the presence of phospholipids and does not result from impaired interactions with either FVIIIa or FX. First, we excluded the possibility that these mutants were not completely activated or that there was a failure of some active sites to develop after cleavage of the FIX zymogen. Protein stains, Western immunoblot analysis and active site titration with ATIII of FXIa-activated FIX proteins indicates that all of the proteins were completely activated. Western immunoblot analysis shows complete disap-pearance of the ϳ70 kDa zymogen band and appearance of bands corresponding to the heavy chain (ϳ28 kDa), light chain (ϳ18 kDa), and activation peptide of FIXa (ϳ10 kDa). Active site titration with ATIII indicated that the expected number of active sites were formed as predicted from protein concentration ( Fig. 3 and Table I). Therefore, the reduced V max of FX activation for FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R is not due to a reduced concentration of active sites. Finally, fluid phase activation of FX was identical for all forms of FIXa, indicating a similar concentration of active sites.
An additional consequence of the mutations could be impaired interactions with FVIIIa, FX, or both. The physiological function of FVIIIa is a large (1000-fold) increase in k cat for FX activation (9,15). The reduced V max of FX activation for FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R could result from deficient interaction of these proteins with FVIIIa. However, two lines of evidence indicate that the interaction of all FIXa proteins with FVIIIa is normal. First, the V max defect of FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R is manifested in both the presence and absence of FVIIIa (Fig. 4, A and B, Table II). The reduced V max is of the same magnitude (2to 4-fold) in both the presence and absence of FVIIIa, and the relative increase in V max for all of the FIXa proteins attributed to the presence of FVIIIa was similar suggesting that functional consequences as a result of FVIIIa binding were entirely normal. EC 50 values for the stimulation of FXa generation by FVIIIa are entirely normal for all of the FIXa proteins (Fig. 5, Table III). The EC 50 values suggest that the affinity of each of the FIXa proteins for FVIIIa is normal. Therefore it is highly unlikely that reduced or deficient interaction of FIXa proteins with FVIIIa would result in the reduced V max of FX activation.
Finally, the reduced V max of FX activation cannot be shown to be a result of impaired interactions of the FIXa mutants with FX. K m(app) values for all of the FIXa proteins are normal both in the presence and absence of FVIIIa (Fig. 4, A and B, Table  II). While we have observed a V max deficiency, this property is observed only in the presence of a reaction surface (Figs. 4 -5). The solution-phase activation of FX by all of the FIXa proteins suggested that the rates of FX activation (k cat ) and thus the affinities for FX are normal (Fig. 6).
The hypothesis that the V max deficiencies of FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R are a result of reduced surface complex assembly is consistent with all of our experimental results and is confirmed by the FIXa titrations in the presence of phospholipid vesicles presented in Fig. 7, A and B and Table IV. The K d(app) is a measure of the affinity of the FIXa proteins for the surface and the proteins (FVIIIa and FX) complexed with it. The increased K d(app) values for FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R indicate a reduction in the affinities of these proteins for the surfacebound complex. In strong support of this argument is the agreement of determined V max values with those predicted based on the K d(app) of each of the FIXa proteins (Table II). The predicted V max values were determined by calculating the fraction of functional complexes based on the input FIXa concentration and the K d(app) as described in the experimental procedures. It has been shown that the concentration of functional active sites, and therefore the activity (V max ), show a strong correlation with surface occupancy of the FIXa protein (11). The reduced ability of these proteins to assemble into complexes on phospholipid vesicles would result in a reduced number of functional active sites. The reduced effective active site concentration would be expected to result in a reduced V max while other kinetic constants (K m(app) for FX, EC 50 for FVIIIa) would be unaffected.
Unexpectedly, we observed a reduction in V max 9 for FIXa FVIIEGF2 , FIXa loop1 , FIXa loop2 , and FIXa loop1 G94R. This observation could be interpreted as a reduction in k cat and hence as evidence that the catalytic properties of these proteins have been perturbed. However, data presented in our accompanying paper (46) offer an alternative explanation. V max 9 is dependent on both k cat as well as the number of available complexes for the FIXa proteins to bind. The reduced V max 9 that we observe could be a consequence of reduced complex stoichiometry (lower number of available binding sites). It is not immediately clear how to reconcile this observation with the current working model for assembly of the FX activating complex in the presence of phospholipids. These observations clearly indicate the need for further investigation.
The aim of this study has been to determine which residues in the FIXa EGF2 domain are important for assembly of the FX activating complex. We have shown here that residues 88 -109 (excluding Arg 94 ) are important for normal FIXa activation of FX. These residues appear to be primarily responsible for efficient assembly of the FX activating complex in the presence of phospholipids. Moreover, the failure of FIXa EGF2 mutations to promote normal clotting in hemophilia B patients may be explained by the reduced complex assembly properties of these proteins. Thus, an important function of the EGF2 domain of FIXa is to confer normal FX activating complex assembly in the presence of a reaction surface.