Protease and EGF1 Domains of Factor IXa Play Distinct Roles in Binding to Factor VIIIa

Previous studies revealed that cleavage at Arg-318–Ser-319 in the protease domain autolysis loop of factor IXa results in its diminished binding to factor VIIIa. Now, we have investigated the importance of adjacent surface-exposed helix 330–338 (162–170 in chymotrypsin numbering) of IXa in its interaction with VIIIa. IXWT, eight point mutants mostly based on hemophilia B patients, and a replacement mutant (IXhelixVIIin which helix 330–338 is replaced by that of factor VII) were expressed, purified, and characterized. Each mutant was activated normally by VIIa-tissue factor-Ca2+ or XIa-Ca2+. However, in both the presence and absence of phospholipid, interaction of each activated mutant with VIIIa was impaired. The role of IXa EGF1 domain in binding to VIIIa was also examined. Two mutants (IXQ50P and IXPCEGF1, in which EGF1 domain is replaced by that of protein C) were used. Strikingly, interactions of the activated EGF1 mutants with VIIIa were impaired only in the presence of phospholipid. We conclude that helix 330 in IXa provides a critical binding site for VIIIa and that the EGF1 domain in this context primarily serves to correctly position the protease domain above the phospholipid surface for optimal interaction with VIIIa.

A common cause of abnormal bleeding is either the deficiency of factor VIII (hemophilia A) or factor IX (hemophilia B). Factor IX is a vitamin K-dependent protein, and during physiologic clotting it is activated to a two-chain, disulfide-linked IXa molecule by VIIa/Ca 2ϩ /tissue factor (TF) 1 and by XIa/Ca 2ϩ (1). The domain organization of factor IXa is similar to those of the other two enzymes (factors VIIa and Xa) involved in the TFinduced coagulation and to that of an anticoagulant enzyme termed activated protein C (APC) (1). The light chain of IXa consists of an amino-terminal ␥-carboxyglutamic acid (Gla) domain (residues 1-40 out of which 12 are Gla residues), a short hydrophobic segment (residues [41][42][43][44][45][46], and two epidermal growth factor (EGF)-like domains (EGF1 residues 47-85 and EGF2 residues 86 -127), whereas the heavy chain contains the carboxyl-terminal serine protease domain with trypsin-like specificity (1,2). Activation peptide of residues 145-180, which is released upon conversion of factor IX to IXa, is rich in carbohydrate and is the least conserved region in IX from different species (3). Factor IXa hence formed converts factor X to Xa in the coagulation cascade; for a biologically significant rate, this reaction requires Ca 2ϩ , phospholipid (PL), and factor VIIIa (1). Thus, to this extent essential to physiologic hemostasis is the interaction of Ca 2ϩ -bound factor IXa (activated hemophilia B protein) with factor VIIIa (activated hemophilia A protein) on the PL surface provided by the platelets.
Based upon the crystal structure of the Gla domain of factor VIIa (4) and the Ca 2ϩ -binding properties of factor X (5), it would appear that this domain in IXa possesses several low to intermediate affinity Ca 2ϩ -binding sites. In addition, the EGF1 and the protease domain each possess one high affinity Ca 2ϩbinding site (6,7). The Ca 2ϩ -loaded conformer of the Gla domain binds to PL vesicles (8) and the EGF1 domain of IX is required for its activation by VIIa/Ca 2ϩ /TF (9). Further, Ca 2ϩ binding to the EGF1 domain has been reported to promote enzyme activity and factor VIIIa binding (10). The role of EGF2 domain is not clear but may be involved in binding to platelets and in factor X activation (11). Finally, the protease domain is thought to play a primary role in binding to factor VIIIa (12)(13)(14).
Recently, we have shown that mutations in the protease domain Ca 2ϩ -binding ligands decrease the affinity of factor IXa for factor VIIIa by ϳ15-fold and that proteolysis at Arg-318 -Ser-319 (150 -151) 2 in the autolysis loop results in a further decrease in this interaction by ϳ8-fold (16). Since residues in the protease domain Ca 2ϩ -binding loop as well as those in the autolysis loop may not directly participate in binding to factor VIIIa (2,17), we postulated that Ca 2ϩ binding to the protease domain and integrity of the autolysis loop stabilize yet another region in this domain of factor IXa that directly interacts with factor VIIIa (16). Based upon this, we opted to examine the role of IXa helix 330 -338 (162-170), which is adjacent to the autol-ysis loop in its interaction with VIIIa. All known sequences from different species in this surface-exposed helix of IXa are identical (15) and point mutations in eight of the nine residues in this region cause hemophilia B (18). Herein, we provide extensive data which indicate that this helix (residues 330 -338) in IXa constitutes a critical binding site for VIIIa and that the reduced biologic activity of helix 330 (162) mutants in hemophilia B patients stems from their substantially reduced affinity for factor VIIIa. We also present data which indicate that a role of the EGF1 domain in this context is to correctly position the protease domain above the PL surface for optimal interaction of factor IXa with factor VIIIa. Initial accounts of this work have been presented in abstract form (19,20).

EXPERIMENTAL PROCEDURES
Proteins and Reagents-Benzoyl-Ile-Glu-Gly-Arg-p-nitroanilide (S-2222) was purchased from Helena Laboratories. Dansyl-Glu-Gly-Argchloromethyl ketone (DEGR-ck) was obtained from Calbiochem. Phosphatidylcholine, phosphatidylserine, recombinant hirudin, and fatty acid-free bovine serum albumin (BSA) were obtained from Sigma. Factor IX-and factor VIII-deficient plasmas were purchased from George King Biomedicals, and activated partial thromboplastin time reagent was obtained from Diagnostica Stago. Normal human plasma factor IX (IX NP ) and factor X were isolated as described (15), and factor Xa was prepared as outlined (21). Purified human factor XIa, protein C, APC, and ␣-thrombin (IIa) were purchased from Enzyme Research Laboratories (South Bend, IN). Recombinant human TF of amino acids 1-243 containing the transmembrane domain was generously provided by Genentech Inc. (South San Francisco, CA) and reconstituted as described (22). Phosphatidylcholine-phosphatidylserine vesicles (75% phosphatidylcholine, 25% phosphatidylserine) were prepared by the method of Husten et al. (23) as outlined (22). Recombinant human factor VIIa was a generous gift of Novo-Nordisk (Copenhagen, Denmark). Purified human factor VIII was obtained from Dr. Leon Hoyer (American Red Cross, Rockville, MD). The preparation was free of all other coagulation factors and contained human albumin as a stabilizing agent. Purification of a mouse monoclonal antibody (mAb) that inhibits the interaction of factor IXa with factor VIIIa has been described (12).
Coagulation Assay of Factor IX and Factor IXa-Factor IX and factor IXa activities were measured in a one-stage assay with automated activated partial thromboplastin time reagent as described (24).
SDS-Gel Electrophoresis-SDS-gel electrophoresis was performed using the Laemmli buffer system (25). The acrylamide concentration was 12%, and the gels were stained with Coomassie Brilliant Blue dye.
Amino Acid Sequencing and Gla Analysis-Gla and amino acid sequence analysis were performed by Commonwealth Biotechnologies, Inc., Richmond, VA. Automated Edman degradation of each factor IX protein (ϳ0.5 nmol) was performed using an Applied Biosystems gas phase sequencer. Gla analysis of each sample was performed by alkaline hydrolysis followed by high pressure liquid chromatography analysis. The amount of Gla was quantitated based upon the 46 residues of Asp and Asn present per mole of factor IX (26).
Construction, Expression, and Purification of Recombinant Factor IX Proteins-The pRc/CMV vector (Invitrogen) was used for expression of wild-type and each mutant factor IX. In each case HindIII and XbaI sites in the multiple cloning sites of the vector were used for ligation of the DNA. Construction of the wild-type factor IX (IX WT ) and that of IX PCEGF1 , in which residues 52-85 in the EGF1 domain of IX WT have been replaced by the residues 51-92 from the corresponding domain of protein C, have been described (27). Point mutations in helix 330 (162) of protease domain of factor IX were introduced using the fragment elongation method of Nelson and Long (28), as described previously for IX Q50P (9). In each case the mutant primer was based upon factor IX gene sequence (26) and corresponded to six codons (18 bases) with a mutant base at the desired position involving the third codon. The base substitution for each point mutant was: L330I (CTT 3 ATT), V331A (GTT 3 GCT), D332Y (GAC 3 TAC), R333L (CGA 3 CTA), R333Q (CGA 3 CAA), T335A (ACA 3 GCA), L337I (CTT 3 ATT), and R338Q (CGA 3 CAA). Factor IX helix VII , in which 330 -338 (162-170) residues of factor IX were replaced by the corresponding residues of factor VII, was constructed using a 63-base primer. The first 18 bases of this primer corresponded to factor IX gene sequence coding for residues 324 -329 (156 -161), followed by 27 bases from the factor VII gene sequence (29) coding for residues (162-170), and the final 18 bases corresponded to the factor IX gene sequence coding for residues 339 -344 (171-177). The PCR was performed in the same fashion as de-scribed previously for point mutants (9,28). All inserts were sequenced (30) to confirm the mutations and to rule out any PCR errors. Expression of each factor IX recombinant protein and its purification was achieved exactly as described (9).
Molecular Modeling-A modeled structure of human factor IXa was obtained using a homology model building approach described previously (15). The starting template used was the structure of porcine factor IXa (Ref. 2; Protein Data Bank code 1PFX). Inasmuch as the model structure of human factor IXa, for all purposes as it relates to this paper, was the same as porcine factor IXa and the residues involved at the mutational sites are identical between the two proteins (15), we have used the x-ray structure of porcine factor IXa for analysis of our data. The model of IXa PCEGFI protein was constructed by replacing the EGFI domain of factor IXa with that of APC (Ref. 31; code 1AUT) and is described elsewhere (32).
Preparation of DEGR-ck Inhibited Various Factor IXa Proteins-IX NP , IX WT , and each mutant factor IX at 200 g/ml was activated by factor XIa (2 g/ml) for 90 min. The buffer used was TBS, pH 7.4 (0.05 M Tris, 0.15 M NaCl, pH 7.4), containing 5 mM Ca 2ϩ . SDS-gel electrophoretic analysis revealed full activation to factor IXa␤ without further degradation. DEGR-IXa NP , DEGR-IXa WT , and various DEGR-IXa mutant proteins were prepared as described previously (16), and free DEGR-ck was removed as described previously (22,33). DEGR-APC was prepared similarly.
Activation of Factor VIII by IIa-Activation of factor VIII was performed as outlined previously (16). Briefly, 28 nM (40 units/ml) of factor VIII was incubated with 0.2 nM IIa in TBS/BSA, pH 7.4 (TBS containing 1 mg/ml BSA), containing 5 mM CaCl 2 . In initial experiments, we established that the maximal increase (ϳ10-fold) in procoagulant activity occurred at 2 min, after which it steadily declined (16). Thus, at 2 min we routinely added recombinant hirudin (2.2 nM, final concentration) to inhibit IIa and the preparation was used immediately.
Measurements of Factor X Activation-Factor VIIIa potentiation of factor X activation by IXa NP , IXa WT , and each activated mutant was measured both in the presence and absence of PL. The buffer used was TBS/BSA, pH 7.4, and the reaction volume in each case was 50 l. The activation was carried out at 37°C for different time periods, at which point each reaction mixture received 1 l of 0.5 M EDTA to stop further generation of factor Xa. A 40-l aliquot was then added to a 0.1-ml quartz cuvette containing S-2222 in 75 l of TBS/BSA, pH 7.4. The final concentration of S-2222 was 100 M. The p-nitroaniline release was measured continuously (⌬A 405 /min) for up to 20 min (5,22). Factor Xa generated was calculated from a standard curve constructed using factor Xa prepared by insolubilized Russell's viper venom.
Factor X activation measurements were made under four sets of different conditions. 1) Ca 2ϩ and PL were present. In this system, activation was carried out for 5-15 min and the concentration of each IXa protein was 20 nM, PL was 25 M, and factor X was 100 nM. 2) Ca 2ϩ , PL, and factor VIIIa were present. In this system, activation was carried out for 15-120 s and the concentration of each IXa protein was 0.5 nM, PL was 10 M, VIIIa was 0.07 nM, and factor X was 15 nM. 3) Only Ca 2ϩ was present. In this system, activation was carried out for 2-20 min, and concentration of each IXa protein was 400 nM and factor X was 1 M. 4) Ca 2ϩ and VIIIa were present. In this system, activation was carried out for 15-120 s and the concentration of each IXa protein was 2 M, VIIIa was 14 nM, and factor X was 400 nM.
In all of the above factor X activation experiments, the incubation times chosen are those in which factor Xa generated was always less than 10 nM. This precautionary measure was taken to prevent activation of factor X by the generated factor Xa (16,34). Further, in those incubation mixtures that contained factor VIIIa, we also performed control experiments in which factor VIIIa was omitted. The rates of factor X activation in these control experiments were Ͻ10% of the experimental values in the presence of factor VIIIa and were subtracted prior to analysis of the data. For reactions done in the presence of Ca 2ϩ and PL, optimal concentrations of PL were determined. Rates of factor X activation versus PL showed a bell-shaped curve with a broad optima between 20 and 40 M PL. 3 The rate increased linearly from 0 -20 M, and after 40 M it showed a gradual decrease. Therefore, PL concentration in the absence of factor VIIIa was fixed at 25 M. In the presence of factor VIIIa, 10 M PL was used based upon previous observations (16,35). The concentrations of factor X selected for each set of reaction conditions are those which are below or at the K m values (35,36). 3 In this region of the Michaelis-Menten curve, the rate of formation of factor Xa is proportional to the substrate factor X and therefore to the affinity of the factor IXa enzyme for factor X.
Determinations of K d (app). Values for the Interaction of Factor VIIIa with Various Active Site-blocked Factor IXa Proteins-Apparent K d values of factor VIIIa binding to various active site-blocked factor IXa proteins were determined from inhibition experiments as detailed previously (16). Briefly, reaction mixtures (50 l) in the presence of PL contained 0.2 nM factor IXa WT , 0.48 M factor X, 0.07 nM factor VIIIa, 5 mM Ca 2ϩ , 10 M PL, and varying concentrations of DEGR-IXa proteins. Reaction mixtures in the absence of PL contained 0.1 M factor IXa WT , 2 M factor X, 14 nM factor VIIIa, 5 mM Ca 2ϩ , and varying concentrations of DEGR-IXa proteins. Reactions were carried out for 2 min, and the rates of factor X activation were measured as outlined above.
In these experimental conditions, both in the presence and absence of PL, factor IXa WT concentrations are below the EC 50 values at the factor X concentrations used. Further, in each case Ͻ10% of factor IXa is bound to factor VIIIa in the absence of the competitor and no measurable rates of activation of factor X were observed under these conditions in the absence of added factor VIIIa. For obtaining IC 50 values (concentration of DEGR-IXa yielding 50% inhibition), the data were fitted to the IC 50 four-parameter logistic equation of Halfman (37) given below.
y is the rate of Xa formation in the presence of a given concentration of DEGR-IXa protein represented by x, a is the maximum rate of Xa formation in the absence of DEGR-IXa, and s is the slope factor. Each point was weighted equally, and the data were fitted to Equation 1 using the nonlinear regression analysis program obtained from Erithacus Software (GraFit). The values of the slope factors were 0.9 Ϯ 0.1 in all experiments, indicating competition for a single binding site. The background value represented Ͻ5% of the maximum rate of Xa formation in the absence of DEGR-IXa. To obtain the K d (app) values for the interaction of DEGR-IXa proteins with factor VIIIa, we used the following equation as described by Cheng and Prusoff (38) and further discussed by Craig (39).
A is the concentration of IXa WT , and EC 50 is the concentration of factor IXa WT that gives a 50% maximum response in the absence of the competitor at a specified concentration of factor X used in the experiment.

RESULTS
Purification, ␥-Carboxyglutamic Acid Content, NH 2 -terminal Sequence, and Activity of Protease Domain Factor IX Mutants-SDS-electrophoretic analysis of factor IX proteins using the Laemmli system (25) is shown in Fig. 1A. Each protein is effectively homogenous in this system. Further, plasma factor IX and each recombinant protein had 11.5-12.5 Gla residues/ mol. The NH 2 -terminal sequence of each protein was also determined. All recombinant proteins revealed a major and a minor sequence. The major sequence in each case was Tyr-Asn-Ser-Gly-Lys, and the minor sequence in each case was Thr-Val-Phe. The major sequence corresponds to the sequence of mature protein in plasma, and the minor sequence corresponds to the protein in which the prosequence has not been cleaved (26). The minor sequence was not detected in plasma factor IX, and it amounted to less than 5% in each recombinant protein. The relative coagulant activity of each protein was: IX NP , 100% (180 Ϯ 10 units/mg); IX WT , ϳ90%; IX L330I , ϳ8%; IX V331A , ϳ6%; IX D332Y , ϳ2%; IX R333L , ϳ0.3%; IX R333Q , ϳ0.5%; IX T335A , ϳ40%; IX L337I , ϳ1%; IX R338Q , ϳ65%; and IX helixVII , not measurable.
Neutralization of an Anti-factor IX mAb by the Mutant Proteins-The ability of the IX mutant proteins to bind to a mAb (12), which interferes with the interaction of IXa and VIIIa, was studied using a coagulant based assay as described previously (40). For mutants that possessed Յ8% coagulant activity, a competition-based assay in which the mutant protein competed with the normal IX in binding to the mAb was used (40). Each mutant protein bound to the mAb with an equal affinity (K d ϳ15 Ϯ 8 nM).
Activation of Protease Domain Factor IX Mutants by VIIa/TF and by Factor Xla/Ca 2ϩ -The rates of activation of each protease domain mutant performed under conditions described previously (16) either by VIIa/TF/Ca 2ϩ or by factor Xla/Ca 2ϩ were similar to that of IX NP as analyzed by SDS-gel electrophoresis. The 90-min activation sample of IX WT and each mutant is shown in Fig. 1B. As compared with IXa NP , coagulant activity of Xla-activated IXa WT was ϳ95%, of IXa L330I was ϳ7%, of IXa V331A was ϳ6%, of IXa D332Y was ϳ2%, of IXa R333L was ϳ0.4%, of IXa R333Q was ϳ0.6%, of IXa T335A was ϳ35%, of IXa L337I was 0.8%, of IXa R338Q was ϳ80%, and of IXa helixVII was not measurable.
Activation of Factor X by Various IXa Protease Domain Mutants-In this section, we examined the ability of each activated mutant to activate factor X in the presence and absence of PL and factor VIIIa. The data obtained in the presence of Ca 2ϩ and PL are shown in Fig. 2A. One should note that the concentration of factor X used in this system was 100 nM, which is slightly less than the K m value (36) under these conditions. In this system, factor IXa WT and each mutant including IXa helixVII activated factor X at similar rates. The rate of factor Xa generation in each case was 0.1 nM/min, which is very close to an expected rate (0.12 nM/min) at an enzyme concentration of 20 nM used in our system. One may infer from these data that, in the IXa-Ca 2ϩ -PL system, each mutant interacts with factor X normally and that the active site of each mutant is not impaired.
Next, we studied the activation of factor X in a complete intrinsic Tenase system (IXa, Ca 2ϩ , PL, VIIIa). When a limiting concentration of factor VIIIa (70 pM) at 0.5 nM IXa and 15 nM factor X was employed, the rate of activation by each IXa mutant with the exception of IXa R338Q was significantly reduced. These data are provided in Fig. 2B and summarized in Table I. Considering an EC 50 (functional K d of IXa:VIIIa interaction) value of 1.2 nM (16), a k cat value of 300/min and a K m of 25 nM under these conditions (36), 3 an expected rate of factor X activation would be 2.3 nM/min; this rate is close to the rate of 2.6 Ϯ 0.2 nM /min observed in our case. Under these conditions, the rate of factor X activation by IXa L330I was ϳ0.95 nM/min, by IXa V331A was ϳ0.55 nM/min, and by IXa T335A was ϳ1.9 nM/min. For IXa R338Q , it was the same as IXa WT , and for other mutants (see Table I), it could not be measured. However, when factor VIIIa concentration was increased from 70 pM to 14 nM, the following rates of factor X activation were obtained: IXa D332Y ϳ4.1 nM/min, IXa R333L ϳ2.7 nM/min, IXa R333Q ϳ2.4 nM/min, and IXa L337I ϳ3.2 nM/min; for IXa helixVII , it was still not measurable. Note that a calculated rate for IXa WT (or IXa NP ) at 14 nM VIIIa would be 0.85 nM Xa generated/s. Thus, each of our protease domain mutants except IXa R338Q is impaired in its interaction with VIIIa in the IXa-Ca 2ϩ -PL-VIIIa system.
Mutations in factor IXa can affect VIIIa binding by at least two mechanisms: 1) by perturbation of factor IXa binding site for factor VIIIa, and 2) by altering the spacing above the PL surface of the IXa interactive site. To distinguish between these two possibilities, we studied the effect of VIIIa on the potentiation of factor X activation in the absence of PL. For these studies, we first measured the rates of factor X activation by the IXa mutants in the presence of Ca 2ϩ only. These data are presented in Fig. 3A. As predicted from the Ca 2ϩ /PL system, each IXa mutant in the presence of Ca 2ϩ only activated factor X at a rate (0.46 Ϯ 0.05 nM/min) comparable to IXa WT . However, as was the case with the Ca 2ϩ /PL/VIIIa system, the IXa mutants also activated factor X in the Ca 2ϩ /VIIIa system at rates that were slower than those obtained with IXa WT (or IXa NP ). These data are presented in Fig. 3B and summarized in Table I. Considering a k cat value of 1.1/min, a K m of 380 nM 3 , and an EC 50 (functional K d of IXa:VIIIa interaction) value of 2.2 M (16) at the 400 nM factor X concentration used in our system, the expected rate of Xa formation using 14 nM VIIIa and 2 M IXa would be 3.3 nM/min. The experimental rate obtained with IXa WT (or IXa NP ) was ϳ2.75 nM/min, a value close to the expected value. The rate of activation by IXa L330I was ϳ0.94 nM/min, by IXa V331A was ϳ0.54 nM/min, by IXa T335A was ϳ1.83 nM/min, and by IXa R338Q was ϳ2.22 nM/min. Under these conditions, i.e. in the presence of limiting concentrations of VIIIa, the rates of factor X activation by other mutants could not be measured. Cumulatively, our data indicate that the mutations in helix 330 (162) residues lead to an impaired interaction of IXa with VIIIa, which does not appear to be due to spatial misalignment of the IXa contact site above the PL surface. Our data also indicate that IXa residue Arg-338 (170) does not play a significant role in binding to factor X or to factor VIIIa.
Measurement of K d (app) Values for the Interaction of Protease Domain IXa Mutants with Factor VIIIa in the Tenase Complex-In the previous section, we established that helix 330 mutants of IXa are impaired in their interactions with factor VIIIa. To further these observations, we evaluated the K d of interaction of these mutants with VIIIa. These data are presented in Fig. 4. The steady-state inhibition curves (41) obtained in the presence (Fig. 4A) and absence (Fig. 4B) of PL were analyzed as outlined under "Experimental Procedures." The K d (app) values for the interaction of active-site blocked mutants with factor VIIIa are listed in Table I. The binding of factor VIIIa to each active-site blocked mutant (except for IXa R338Q ) was considerably weaker both in the presence and absence of PL. Compared with DEGR-IXa WT , L330I and T335A mutants had similarly (ϳ4-and ϳ2.5-fold, respectively) reduced affinity for factor VIIIa in the presence or absence of PL. However, other mutants (Table I)  Conceivably, in these mutants, in addition to the perturbation of factor VIIIa binding site, a further reduced affinity in the presence of PL could in part be due to the misalignment of the factor VIIIa contact site. It is noteworthy that IXa VIIhelix mutant failed to bind to factor VIIIa at concentrations ϳ140,000fold greater than the K d (app) for IXa WT in the presence of PL and at concentrations ϳ150-fold greater than the K d (app) in the absence of PL. Thus, both the kinetic and the binding data provide strong evidence that helix 330 in the protease domain of factor IXa provides a critical binding site for factor VIIIa.
Role of EGF1 Domain of Factor IXa in Binding to Factor VIIIa-In the previous section, we provide compelling evidence that the IXa helix 330 constitutes a critical binding site for factor VIIIa. However, data also exist to support the theory that the EGF1 domain is involved in binding to factor VIIIa (10,42,43). Notably, essentially all of the studies related to the EGF1 domain have been conducted in the presence of PL. Thus, one cannot differentiate whether the EGF1 domain is directly involved in binding to factor VIIIa or whether alterations in this region result in a misalignment above the PL surface of the region(s) in the protease domain, which is the direct contact site in interacting with factor VIIIa. To address this, we conducted studies both in the presence and absence of PL using two EGF1 mutants generated in our laboratory (9). The two mutants employed for these studies were factor IX Q50P , which lacks the EGF1 domain Ca 2ϩ binding site, and IX PCEGF1 , in which the EGF1 domain of factor IX has been replaced by that of protein C (9). Both mutants had the same NH 2 -terminal sequence and Gla content as normal IX and could be readily activated to IXa-like molecules by factor Xla (9). In the present study, we have found that these mutants also bind to the mAb (that interferes with the IXa:VIIIa interaction) with the same affinity (k d ϳ15 nM) as normal IX.
In the Ca 2ϩ /PL system (same conditions as in Fig. 2A, i.e. 20 nM IXa and 100 nM X), IXa Q50P activated factor X at 0.07 nM/min and IXa PCEGF1 at 0.06 nM/min. These rates are slightly slower than the rate (0.1 nM/min) obtained with IXa WT . These results could be attributed to a slight shift in the active site of the variant proteins relative to the EGF1 domain (32). When limiting concentrations of factor VIIIa in the complete Tenase system (VIIIa, 70 pM; IXa, 0.5 nM; and X, 15 nM) were employed, both the EGF1 domain mutants failed to activate factor X at measurable rates (Fig. 5A and Table I). As expected, both EGF1 domain mutants activated factor X in the presence of only Ca 2ϩ at rates similar to that of IXa WT . Thus, at 1 M X and 400 nM IXa, the rates of activation by IXa WT , IXa Q50P , and IXa PCEGF1 were ϳ0.46 nM/min, ϳ0.39 nM/min, and ϳ0.48 nM/min, respectively. Surprisingly, however, both mutants also activated factor X in the Ca 2ϩ /VIIIa system at rates similar to that of IXa WT (Fig. 5B and Table I). Consistent with these observations, K d (app) of IXa:VIIIa interaction was only impaired in the presence of PL (Fig. 6A) and not in its absence (Fig. 6B). K d (app) values both in the presence and absence of PL are provided in Table I. In control experiments, neither protein C nor DEGR-  3. Factor VIIIa-mediated potentiation of factor X activation by each protease domain IXa mutant in a system containing Ca 2؉ only. A, the time course of activation of factor X in IXa-Ca 2ϩ system. Activation mixture contained 1 M factor X, 5 mM Ca 2ϩ , and 400 nM various IXa proteins. B, the time course of activation of factor X in IXa-Ca 2ϩ -VIIIa system. Activation mixture contained 400 nM factor X, 5 mM Ca 2ϩ , 14 nM factor VIIIa, and 2 M various factor IXa proteins. Each reaction was stopped at a given time by the addition of EDTA and Xa generated was measured by S-2222 hydrolysis. Note that the effective concentrations of enzyme and substrate are different in A and B. Further, the data plotted in B represent the data which were obtained after subtraction of the background rates (Յ10%) in the absence of VIIIa (see "Experimental Procedures" for details). APC competed with IXa in binding to VIIIa (Fig. 6, A and B). From these data, we conclude that the EGF1 domain of factor IXa in the absence of PL does not play a significant role in its interaction with VIIIa and that in the presence of PL it may primarily function to correctly position the protease domain for optimal binding to VIIIa.

Role of Factor IXa Protease Domain in Binding to Factor
VIIIa-Initial evidence that led to the proposal that the protease domain of factor IXa is involved in binding to factor VIIIa came from the observations that a mAb to the protease domain inhibited factor IXa:factor VIIIa interaction (12). Additional biochemical studies supported this concept (13,14). Further studies mapped this antibody to residues 180 -310 of protease domain (44), and a part of the epitope was found to be located in the calcium binding loop (7). Moreover, a hemophilia B patient in which Glu-245 in factor IX, a ligand for Ca 2ϩ -binding in the protease domain, was replaced by valine has been identified (45). Based upon these studies, it was thought that the protease domain Ca 2ϩ -binding loop may constitute a part of the factor VIIIa binding site (45). However, site-specific mutations adjoining this loop did not lead to impairment in the clotting activity of factor IX, indicating that the Ca 2ϩ -binding loop does not directly contribute to factor VIIIa binding (17). Currently, it is believed that the binding of Ca 2ϩ to the protease domain indirectly affects binding of factor VIIIa to this domain (16).
In order to identify the region in the protease domain for IXa:VIIIa interaction, we studied the role of surface-exposed helix 330. The position of this helix is shown in Fig. 7A. The sequence in this helix is identical in factor IX from all species (15) and is different from all other homologous blood coagulation serine proteases (Table II). Further, helix 330 is located 12 residues away from the autolysis loop cleavage site to which it is connected via a single ␤-strand, marked 2 in Fig. 7A. Moreover, point mutations in eight of the nine residues in this helix are reported to cause hemophilia B (18).
The nature and orientation of the side chains along with the point mutants constructed in the present study are shown in Fig. 7B. All point mutants (except for R338Q) constructed in helix 330 residues primarily based upon hemophilia B patients had reduced affinity for factor VIIIa both in the presence and absence of PL (Table I). Residue Leu-330 (162) is located in a hydrophobic pocket (Fig. 8A) and makes van der Waals contacts with residues Leu-300 (131) and Phe-349 (181). In the mutant L330I, the hydrophobic contacts with Phe-349 (181) may be weakened (32), which could destabilize the helix resulting in a reduced affinity for factor VIIIa. Residues Val-331 (163), Asp-332 (164), and Thr-335 (167) are on the same side of the helix. The van der Waals and hydrogen bond interactions involving these residues are shown in Fig. 8B. When Val-331 (163) is changed to Ala-331 (163), the hydrophobic contacts involving this residue with Thr-335 (167), His-354 (185), and Tyr-395 (225) will be weakened. This again will destabilize the helix resulting in an impaired interaction with factor VIIIa (Table I). A change in Asp-332 (164) to Tyr-332 (164) will disrupt essentially all contacts in this pocket since tyrosine is a much bulkier residue than aspartic acid, which is expected to shift the helix away from the this pocket. Finally, a change of Thr-335 (167) to Ala-335 will result in disruption of its contacts with Asp-332, Val-331, and His-354. Residue Arg-333 (165) is solvent-exposed, and abolishing the positive charge by mutational change may abolish the direct interaction of this amino acid with factor VIIIa. As depicted in Fig. 8C, changing Leu-337 (169) to Ile may result in the disruption of another hydrophobic pocket and a slight shift of the helix (32). Notably, changing Arg-338 (170) to Gln (170) resulted in a minimal loss of biologic activity and affinity for factor VIIIa (Table I). This is consistent with a slightly increased activity of R338A mutant without an impairment in factor VIIIa binding described by Chang and co-workers (46). Note that all of the helix 330 point mutants have impaired interaction with factor VIIIa both in the presence and absence of PL. Furthermore, IXa helixVII mutant failed to interact with factor VIIIa even at very high concentrations, both in the presence or absence of PL. These data strongly support the conclusion that helix 330 in IXa represents a critical binding site for VIIIa.
Hemophilia B mutants (18) not expressed in the present study in helix 330 may have impaired interactions with factor VIIIa due to the following reasons. Changing Leu-330 (162) to proline may result in a turn accompanying a subtle directional change in the propagation of the helix, and changing Val-331 (163) to aspartic acid is expected to disrupt the hydrophobic interactions depicted in Fig. 8B. Replacement of Arg-333 (165) by glycine, and of Ala-334 (166) by aspartic acid or threonine could disrupt the direct binding of factor IXa to factor VIIIa. A change of Cys-336 to any other residue will disrupt the disulfide bond and a possible change in the local tertiary structure. Replacement of Leu-337 (168) to phenylalanine or proline is expected to disrupt the putative hydrophobic pocket (Fig. 8C), and, in the case of the proline mutant, it may also change the direction of the polypeptide. Similarly, replacement of Arg-338 by proline could result in the change of the direction of the polypeptide and disruption of the local tertiary structure.

Role of Factor IXa EGF1 Domain in Binding to Factor
VIIIa-The first EGF-like domain of factor IXa has also been implicated in binding to factor VIIIa (10,42,43). In the present study, two EGF1 mutants were examined for their abilities to bind to factor VIIIa both in the presence and absence of PL. In a system containing PL, IXa Q50P and IXa PCEGF1 interacted with factor VIIIa with ϳ20and ϳ100-fold reduced affinity, respectively. However, in the absence of PL, both activated mutants interacted with factor VIIIa with K d (app) indistinguishable from that of normal IXa (Table I). Since alterations of the EGF1 domain affect VIIIa binding only in the presence of PL but not in its absence, whereas mutations in helix 330 affect VIIIa binding in both the presence and absence of PL, it can be concluded that helix 330 in IXa provides a crucial binding site for VIIIa and that the EGF1 domain in this context primarily serves to correctly position the protease domain above the PL surface for optimal interaction with VIIIa. In a preliminary report (47), it has also been concluded that the residues 330 -338 in IXa are involved in binding to factor X. However, data presented in Figs. 2A and 3A do not support this concept. Further studies will be required to fully resolve this issue.
Modeling studies support the concept that helix residues 330 -338 may be shifted in the IXa PCEGF1 mutant (32). In this model, in which the membrane binding Gla domain coordinates are unchanged from normal IXa, the EGF2/protease domain coordinates including helix 330 are shifted ϳ1.5 Å. Further, the eight-residue insertion in the EGF1 domain lies on the same side as helix 330. As a consequence of these two deviations from normal IXa, it is possible that the interactions of helix 330 residues with factor VIIIa are weakened. In other studies in which EGF1 domain of factor IX was replaced by that of factor X (48) or factor VII (49), the resulting molecule either had normal activity (IXa XEGF1 ) or 4-fold increased activity (IXa VIIEGF1 ). The activity in the IXa VIIEGF1 was attributed to the increased affinity of this mutant for factor VIIIa (49). When EGF1 domain of factor VII was modeled into the factor IX molecule, the EGF2/protease domain coordinates including those of helix 330 were only shifted ϳ0.5 Å (32). Further, this shift in IXa VIIEGF1 was in the opposite direction to that observed in the IXa PCEGF1 molecule. Since subtle changes can lead to increase in the bond strengths, it is possible that helix 330 residues in IXa VIIEGF1 molecule are positioned more favorably to interact with factor VIIIa binding site. Furthermore, it is quite feasible that in other EGF1 mutants (10,24,42,43,50,51) and in a Gla domain mutant (52), the alignment of helix 330 above the PL surface is altered, which results in its impaired interactions with factor VIIIa. These conclusions are consistent with the observation of Lenting et al. (10) and of Christophe et al. (51), in which the light chain mutants of IXa, in the absence of PL, have minimal alterations (2-4-fold) in direct binding to factor VIIIa, whereas in the presence of PL interactions of these IXa variants with VIIIa were impaired 50 -200-fold. As is the case with IXa Q50P and IXa PCEGF1 (Table I), we anticipate that interactions of the above mutants with factor VIIIa in the absence of PL may be minimally impaired.
Concluding Remarks-Essential to hemostasis is the interaction of factor IXa with factor VIIIa. The present studies indicate that helix 330 in the protease domain of factor IXa provides a critical binding site for factor VIIIa. Although weaker interactions (10,51) and/or exposure of neoepitopes in IXa or VIIIa cannot be ruled out, a role of the EGF1 domain of factor IXa in this context may be to primarily serve as a spacer in properly positioning the IXa protease domain above the PL surface for optimal interaction with factor VIIIa. However, whether or not Gla and EGF2 domains are involved in direct binding to factor VIIIa is not known. Since the anti-factor IX mAb binds normally to helix 330 mutants including IXa helixVII , it probably interferes with IXa:VIIIa interaction primarily by steric hindrance or by locking the conformation of IXa heavy chain unfavorable for binding to factor VIIIa. Ca 2ϩ binding (loop 1 in Fig. 7A) to the protease domain is known to enhance IXa:VIIIa interaction (16). It most likely occurs via interactions through the autolysis loop (loop 2 in Fig. 7A) which directly connects to helix 330 (162). Furthermore, helix 330 (162) interacts with another surface-exposed helix, 293-301 (126 -132) (Fig. 8A), which is also conserved in factor IX from all species sequenced thus far (15); this possibly could represent another binding site for factor VIIIa; studies are under way to examine this concept. The region in factor VIIIa that interacts with the protease domain of IXa is quite possibly located in the A2 domain (36). Binding of A2 domain to IXa has been reported to increase the kcat of factor X hydrolysis (36). This could occur by modulation and/or stabilization of helix 330, which links di-rectly to the extended substrate binding site residues Ile-397 (227) (53) and Trp-385 (215) through hydrophobic contacts (Fig.  8C). Further, it should be noted (Table II) that the corresponding helix residues in factor VIIa bind to TF (4) and in factor Xa, they are involved in binding to factor Va. 4 Thus, a common function of this helix (162 in chymotrypsin numbering) in several blood coagulation proteases may be to serve as a binding site for the respective cofactor. Function of this helix, if any, in other serine proteases including chymotrypsin, trypsin, and APC remains to be elucidated. and one anticoagulant serine proteases The sequence of helix 330 is identical in factor IX from human, bovine, porcine, canine, rabbit, sheep, guinea pig, mouse, and rat (15). For comparison, the residue number for each protein corresponding to residue 162 in chymotrypsin is given in parentheses. A dash indicates the same residue as in factor IX. All sequences are taken from reference 15.
Protein Sequence