The factor IX gamma-carboxyglutamic acid (Gla) domain is involved in interactions between factor IX and factor XIa.

During hemostasis, factor IX is activated to factor IXabeta by factor VIIa and factor XIa. The glutamic acid-rich gamma-carboxyglutamic acid (Gla) domain of factor IX is involved in phospholipid binding and is required for activation by factor VIIa. In contrast, activation by factor XIa is not phospholipid-dependent, raising questions about the importance of the Gla for this reaction. We examined binding of factors IX and IXabeta to factor XIa by surface plasmon resonance. Plasma factors IX and IXabeta bind to factor XIa with K(d) values of 120 +/- 11 nm and 110 +/- 8 nm, respectively. Recombinant factor IX bound to factor XIa with a K(d) of 107 nm, whereas factor IX with a factor VII Gla domain (rFIX/VII-Gla) and factor IX expressed in the presence of warfarin (rFIX-desgamma) did not bind. An anti-factor IX Gla monoclonal antibody was a potent inhibitor of factor IX binding to factor XIa (K(i) 34 nm) and activation by factor XIa (K(i) 33 nm). In activated partial thromboplastin time clotting assays, the specific activities of plasma and recombinant factor IX were comparable (200 and 150 units/mg), whereas rFIX/VII-Gla activity was low (<2 units/mg). In contrast, recombinant factor IXabeta and activated rFIX/VIIa-Gla had similar activities (80 and 60% of plasma factor IXabeta), indicating that both proteases activate factor X and that the poor activity of zymogen rFIX/VII-Gla was caused by a specific defect in activation by factor XIa. The data demonstrate that factor XIa binds with comparable affinity to factors IX and IXabeta and that the interactions are dependent on the factor IX Gla domain.

ciency state (4,5). FIX is a member of a family of proteases, including the hemostasis-related proteins prothrombin, factor VII, factor X, and protein C (the "vitamin-K dependent proteases"), which require specific post-translational modifications for normal activity (3,6). At the N terminus of the mature forms of these proteins is a region rich in glutamic acid called the ␥-carboxyglutamic acid or "Gla" domain. Glutamic acid residues in the Gla domain are modified by the addition of a carboxyl group to the ␥-carbons in a reaction catalyzed by the vitamin K-dependent enzyme ␥-glutamyl carboxylase (6,7). Proper ␥-carboxylation is required for Gla domain binding to calcium and phospholipid, two properties that are indispensable for proper protease activity during coagulation (8). In vivo, Gla domain-dependent protease-substrate interactions take place on the phospholipid membranes of damaged cells and activated platelets. Binding to phospholipid accelerates the conversion of substrate to product by decreasing the K m for the reactions several orders of magnitude (8 -10). The use of coumarin compounds as therapeutic anticoagulants is based on their ability to interfere with vitamin K metabolism, causing incomplete ␥-carboxylation of the Gla domains of prothrombin and factors VII, IX, and X (8). In addition to phospholipid, protease-substrate interactions involving vitamin K-dependent proteins require a protein cofactor, which further improves catalytic efficiency by increasing the k cat for the reactions (9).
FIX is a 56-kDa polypeptide comprised of C-terminal trypsinlike catalytic (heavy chain) domain and an N-terminal noncatalytic (light chain) region separated by an 11-kDa activation peptide (1,11). Two proteolytic cleavages are required to liberate the activation peptide from the remainder of the molecule to produce FIXa␤ (11,12). Activation may occur by two distinct mechanisms mediated by the plasma serine proteases factor VIIa (EC 3.4.21.21) and factor XIa (EC 3.4.21.27) (2,3,13,14). Activation of FIX by factor VIIa is a typical coagulation protease-substrate interaction requiring calcium, phospholipid, and a protein cofactor (the membrane protein tissue factor) (13,15). In this reaction the Gla domains of both FIX and factor VIIa form critical interactions with the phospholipid surface (16 -20). The importance of the FIX Gla domain to FIX activation by factor XIa is less certain because the reaction appears to involve a mechanism distinctly different from typical vitamin K-dependent protease-substrate interactions. Although calcium is required (1,21,22), phospholipid has little influence on the process (1,11,23). Indeed, factor XIa lacks a Gla domain, suggesting that it may interact poorly with phospholipid (3,24,25). Furthermore, a protein cofactor has not been identified for FIX activation by factor XIa, at least when the reaction occurs in liquid plasma. These observations raise the possibility that the FIX Gla domain is not required for activation by factor XIa.
This evidence not withstanding, there are some data to support a role for the Gla domain in FIX activation by factor XIa. A Gla domain mutation at amino acid 4 associated with hemo-philia B interferes with factor XIa-mediated activation of FIX (26), as does failure to remove the FIX propeptide from the N terminus (27). Monoclonal antibodies directed against FIX-Gla block activation of FIX by factor XIa (28,29). However, the antibodies interfere with other activities such as FIX activation by factor VIIa/tissue factor and factor X activation by FIXa␤. Therefore, nonspecific steric interference cannot be ruled out as the mechanism of action. In this report we describe studies on FIX binding to factor XIa. The work demonstrates that both FIX and FIXa␤ bind to factor XIa but not zymogen factor XI and that the FIX Gla domain plays a critical role in the interactions.

EXPERIMENTAL PROCEDURES
Plasma and Recombinant FIX Proteins-Plasma-derived FIX, FIXa␣, and FIXa␤ were purchased from Enzyme Research Laboratories (South Bend, IN). The cDNAs for human wild type FIX and chimera FIX/VII-Gla were gifts from Dr. Darrell Stafford, University of North Carolina, Chapel Hill (30). FIX/VII-Gla contains the factor VII signal peptide, propeptide, Gla domain, and aromatic stack region (factor VII amino acids 1-48) and the FIX epidermal growth factor-like (EGF) 1 and 2 domains, activation peptide, and catalytic domain (factor IX amino acids 50 -415). cDNAs were ligated into the mammalian expression vector pJVCMV, which contains a cytomegalovirus promoter (31). 50 million HEK293 cells (a human fetal kidney fibroblast line, ATCC CRL 1573) were cotransfected with 40 g of expression construct and 2 g of plasmid RSVneo, which contains a gene conferring resistance to neomycin. Transfection was by electroporation (Electrocell Manipulator 600 BTX, San Diego). Transfected cells were grown in Dulbecco's modified Eagle's medium with 5% fetal bovine serum containing the neomycin analog G418 at 500 g/ml. G418-resistant clones were transferred to 96-well culture plates, and supernatants were tested for protein expression by enzyme-linked immunosorbent assay using goat anti-human FIX antibodies (Affinity Biologicals, Hamilton, ON). Expressing clones were expanded in 175-cm 2 culture flasks, and the medium was changed to serum-free Cellgro Complete Medium (Mediatech, Herdon, VA) supplemented with 10 g/ml vitamin K 1 (phytonadione, Abbot). The medium was exchanged every 48 -72 h. Conditioned medium was supplemented with benzamidine to a final concentration of 5 mM and stored at Ϫ20°C pending purification. Recombinant proteins are designated by the prefix "r" to distinguish them from plasmaderived proteins (no prefix). To generate rFIX that is incompletely ␥-carboxylated (rFIX-des␥), transfected HEK293 cells expressing rFIX were grown in medium supplemented with 5 g/ml sodium warfarin (Sigma) instead of vitamin K.
Purification of Recombinant Proteins-Proteins were purified from conditioned medium by monoclonal antibody affinity chromatography. Antibodies were linked to 5 ml of Affi-Gel 10 (Bio-Rad) at 3 mg of IgG/ml of gel. For rFIX, the antibody used was humanized murine monoclonal IgG SB 249417 (GlaxoSmithKline, King of Prussia, PA), a calcium-dependent antibody that recognizes the properly ␥-carboxylated FIX Gla domain (32). One to two liters of conditioned medium were run across the column, followed by washing with 25 mM Tris-HCl, pH 7.5, 100 mM NaCl (TBS) containing 2.5 mM CaCl 2 . Elution was with TBS containing 25 mM EDTA. For rFIX-des␥ and rFIX/VII-Gla, a monoclonal murine IgG against the FIX catalytic domain (kindly provided by Dr. George Broze, Washington University, St. Louis, MO) was used. Washing was with TBS containing 2.5 mM CaCl 2 , and elution was with TBS containing 2.5 mM CaCl 2 and 2.0 M sodium thiocyanate. Protein containing fractions from elutions were pooled and concentrated in an Amicon concentrator, dialyzed against TBS, and stored at Ϫ80°C. Protein concentration was determined by dye binding assay (Bio-Rad).
Recombinant Factor XIa-Ala 557 -Preparation of recombinant factor XI has been described previously (31). Briefly, HEK293 cells were transfected with an expression construct consisting of the human factor XI cDNA in pJVCMV to generate stable expressing clones, and recombinant factor XI was purified from conditioned medium by affinity chromatography using monoclonal IgG 1G5.12 (31). This process was used to generate a recombinant variant of factor XI in which the active site serine residue at amino acid position 557 was changed to alanine (factor XI-Ala 557 ) by site-directed mutagenesis. Factor XI-Ala 557 was converted to the "active" form (factor XIa-Ala 557 ) by diluting to 300 g/ml in TBS containing 5 g/ml human factor XIIa (Enzyme Research Laboratory) and incubating at 37°C. Conversion of the 80-kDa zymogen to the 45-kDa heavy and 35-kDa light chains of the "activated" species was followed by reducing SDS-PAGE. Fac-tor XIa-Ala 557 was separated from factor XIIa by repurification over the 1G5.12 column.
Determination of Specific Activities of Zymogen and Active Recombinant Proteins in Plasma Clotting Assays-Activated partial thromboplastin time (aPTT) assays were performed as follows. The protein to be tested (plasma FIX, rFIX, rFIX-des␥, or rFIX/VII-Gla) was diluted to 5 g/ml in TBS containing 0.1% bovine serum albumin (TBSA). Serial dilutions of the 5 g/ml stocks were prepared in TBSA prior to testing in the aPTT assay. 65 l of FIX-deficient plasma (Diagnostica Stago, Asnieres-sur-Seine, France), 65 l of the protein to be tested, and 65 l of PTT-A reagent (Diagnostica Stago) were mixed in a fibrin cup and incubated at 37°C for 5 min. 65 l of 25 mM CaCl 2 was added, and the time to clot formation was determined on a fibrometer (DataClot II, Helena Laboratories, Beaumont, TX). Results were compared with serial dilutions of pooled normal plasma (George King, Overland Park, KS). By definition, undiluted normal plasma has a FIX activity of 1 unit/ml or 100%. For modified partial thromboplastin time (modified PTT) assays, proteases to be tested (plasma FIXa␤, rFIXa␤, or rFIXa␤/ VII-Gla) were diluted to 5 g/ml in TBSA, and serial dilutions were prepared as described above. 65 l of FIX-deficient plasma, 65 l of rabbit brain cephalin prepared by the method of Bell and Alton (33), and 65 l of the protein to be tested were mixed in a fibrin cup warmed to 37°C. 30 s later, 65 l of 25 mM CaCl 2 was added, and the time to clot formation was determined on the fibrometer. Results were compared with those for plasma FIXa␤, which was arbitrarily assigned an activity of 100%.
FIX and FIXa␤ Binding to Factor XI and Factor XIa Studied by Surface Plasmon Resonance (SPR)-SPR studies were performed on a dual flow cell Biacore X device (Biacore, Inc., Uppsala, Sweden). The zymogen or activated versions of plasma-derived factor XI or recombinant factor XI-Ala 557 were immobilized on a carboxymethyl dextran (CM5) surface using amine coupling chemistry. The surface of the flow cell was activated by injection of a mixture of N-hydroxysuccinimide and EDAC for 5 min at a rate of 10 l/min. Factor XI or XIa at 50 g/ml in sodium acetate buffer, pH 5.5, was manually injected onto the activated surface. Finally, the remaining active sites on the flow cell were blocked by injecting 1 M ethanolamine for 5 min. A flow cell to assess nonspecific background binding was prepared using plasma kallikrein instead of factor XIa. Plasma kallikrein is structurally highly similar to factor XIa but interacts poorly with FIX (25,31,34).
Analyte (FIX, FIXa␤, rFIX, rFIX-des␥, or rFIX/VII-Gla) was injected across the flow cells at varying concentrations (4 nM to 2 M) in HEPESbuffered saline (10 mM HEPES, pH 7.4, 150 mM NaCl, 0.005% polysorbate 20) containing 2.0 mM CaCl 2 at a flow rate of 35 l/min. A 2-min association time was determined to be adequate for plasma-derived FIX, FIXa␤, and wild type rFIX and was used for all subsequent experiments. Removal of bound analyte to regenerate the flow cell was accomplished by infusing HEPES-buffered saline containing 3 mM EDTA. Flow cells were equilibrated with HEPES-buffered saline containing 2.0 mM CaCl 2 prior to subsequent runs. A similar procedure was used for determining nonspecific binding of analyte to kallikrein-coated flow cells. Data obtained for analyte binding to immobilized factor XI/XIa were corrected for nonspecific binding by subtracting the kallikrein flow cell signal from the factor XI/XIa flow cell signal obtained with the same analyte. BIAevaluation software provided by the manufacturer was used for data analysis. The BIAcore equilibrium analysis method was used for all interactions. Briefly, the response at equilibrium (Ru eq ) for each concentration of analyte was determined as above. A nonlinear regression routine was used to determine the K d by fitting the data to the 1:1 interaction steady-state affinity model, where C represents the concentration of analyte and R max the maximal binding capacity. SPR experiments on the effect of the FIX Gla domain-specific antibody SB 249417 on the FIX-factor XIa interaction were performed similarly, with the following exceptions. The concentration of analyte (FIX) was fixed at 1 M, and various concentrations of SB 249417 (0.01-5.0 M) were mixed with the analyte and incubated for 5 min at room temperature prior to injecting into the flow cells. The data were fit to a single site competition model, and the inhibition constant (K i ) was calculated by nonlinear regression (GraphPad Prism, version 3.0, GraphPad, San Diego).
Effect of Monoclonal Antibody SB 249417 on Factor IX Activation by Factor XIa-Inhibition of FIX activation by factor XIa was assessed by a colorimetric assay. A reaction mixture containing 200 nM FIX, 1 nM factor XIa, and varying concentrations of SB 249417 (1 nM to 10 M) in reaction buffer (50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 2.5 mM CaCl 2 , and 0.05% CHAPS) was incubated at 37°C for 30 min. After incubation, FIXa␤ generation was determined by adding the reaction mixture to an equal volume of a detection mixture comprising reaction buffer and 66% ethylene glycol containing 2 mM FIXa p-nitroanilide-conjugated peptide substrate S299 (American Diagnostica, Greenwich, CT). After incubating for 10 min at 37°C, the release of p-nitroanilide as a measure of FIXa activity was determined at a wavelength of 405 nm using a spectrophotometric microplate reader. In the absence of FIX in the assay, no p-nitroanilide signal was detected. Data were fit to a single site competition model, and an inhibition constant (K i ) was calculated by nonlinear regression (GraphPad Prism).
Time Courses of FIX Activation followed by Western Immunoblot and SDS-PAGE-Plasma-derived or recombinant proteins were diluted to 100 nM in TBS containing 2.5 mM CaCl 2 , and the solution was warmed to 37°C in a water bath. Reactions were started by the addition of plasma-derived factor XIa to a final concentration of 1 nM. At time points between 0 and 60 min, 30-l volumes were removed and mixed with 10 l of 4ϫ nonreducing SDS-sample buffer (500 nM Tris-HCl, pH 6.8, 40% glycerol, 10% SDS). Samples were size fractionated on 12% polyacrylamide gels followed by transfer to nitrocellulose membranes. Blots were developed with goat anti-human FIX polyclonal IgG (Affinity Biologicals) using an ECL chemiluminescence Western blotting detection kit (Amersham Biosciences). The intensities of bands on autoradiographs for Western blots of recombinant protein activation by factor XIa were measured using a Bio-Rad Imaging Densitometer model GS-670. Measurements for bands representing the zymogen (56 kDa band) and active protease (45 kDa band) were determined for each lane. For each lane, the value for the active protease was divided by the sum of the signals for the zymogen and active protease for each time point to determine the percent zymogen converted to protease.
Cleavage of FIX by Russell's viper venom protease (RVVP) was assessed as follows. Plasma-derived or recombinant FIX proteins were diluted to 1.0 M in TBS containing 2.5 mM CaCl 2 , and the solution was warmed to 37°C in a water bath. Reactions were started by he addition of RVVP (Enzyme Research Laboratory) to a final concentration of 15 nM. At time points between 0 and 120 min, 30-l volumes were removed and mixed with 10 l of 4ϫ reducing SDS-sample buffer (500 nM Tris-HCl, pH 6.8, 40% glycerol, 10% SDS, 10% 2-mercaptoethanol). Samples were size fractionated on 12% polyacrylamide gels, and gels were stained with GelCode Blue stain reagent (Pierce).

Binding of FIX and FIXa␤ to Factor XIa Studied with SPR-
Initially, binding of plasma-derived FIX in solution to immobilized plasma-derived factor XIa was studied. FIX concentrations between 4 nM and 2 M were tested using 2-min association and 90-s dissociation times. FIX rapidly associates with, and dissociates from, factor XIa (Fig. 1A) in a process that is dependent on calcium (data not shown). Dissociation appears to be nearly complete, indicating that little nonspecific irreversible binding to components of the flow cell is occurring. A plot of plasma FIX bound to factor XIa as a function of FIX concentration is shown in Fig. 1B (open circles). The plot was derived from data in Fig. 1A corrected for nonspecific binding as determined by the simultaneous infusion of the identical concentrations of FIX across the reference flow cell containing immobilized plasma kallikrein. Certain characteristics of the FIX-factor XIa interaction (the rapid dissociation rate in particular) preclude a kinetic fit of the data. We used equilibrium binding analysis to determine the binding constant, where binding at steady state is plotted against the concentration of FIX, and K d is determined in the traditional manner as the concentration of analyte (FIX) occupying 50% of available binding sites. Using this method, a K d for the FIX-factor XIa binding interaction of 120 Ϯ 11 nM was obtained. This result is in reasonably good agreement with published values of K m for activation of FIX by factor XIa (160 -180 nM) determined by chromogenic substrate assay (31,35). Interestingly, FIXa␤ also bound to plasma factor XIa with a K d of 110 Ϯ 8 nM (Fig. 1B,  open squares). Apparently liberation of the activation peptide during FIX activation does not alter the affinity of the protein for factor XIa. Neither FIX nor FIXa␤ bound to zymogen factor XI (Fig. 1B, closed circles and closed squares), indicating that factor XI undergoes conformational changes upon conversion to factor XIa which exposes the FIX binding site.
A significant concern when using SPR to study the binding of a substrate to its enzyme is, of course, that the substrate may be converted to product by the enzyme on the surface of the flow cell. In the case under consideration, this could confound interpretation of results for binding of FIX to factor XIa. To address this issue, we prepared a recombinant version of factor XIa, factor XIa-Ala 557 , in which the active site serine of the catalytic domain was replaced with alanine. Wild type factor XIa expressed in HEK293 cells has been shown to have activity similar to that of plasma-derived factor XIa in plasma and purified protein based assays (31,36). As expected, factor XIa-Ala 557 lacks activity in plasma clotting assays and does not cleave the factor XIa chromogenic substrate S-2366 (data not shown). The binding of FIX and FIXa␤ to factor XIa-Ala 557 was studied by SPR in the same manner as binding to plasmaderived factor XIa (Fig. 1C, open circles and squares). The K d values for binding of FIX and FIXa␤ to factor XIa-Ala 557 (152 Ϯ 41 and 129 Ϯ 27 nM, respectively) are comparable with those obtained with plasma-derived factor XIa. Again, FIX and FIXa␤ do not bind to uncleaved "zymogen" factor XI-Ala 557 (Fig. 1C, closed circles and squares). The results indicate that conversion of FIX to FIXa␤ by immobilized factor XIa, if it , SPR studies of human plasma-derived FIX and FIXa␤ binding to immobilized plasma-derived factor XI and factor XIa. E, FIX binding to factor XIa; Ⅺ, FIXa␤ binding to factor XIa; q, FIX binding to factor XI; and f, FIXa␤ binding to factor XI. C, SPR studies of human plasmaderived FIX and FIXa␤ binding to recombinant factor XI/XIa-Ala 557 . E, FIX binding to factor XIa-Ala 557 ; Ⅺ, FIXa␤ binding to factor XIa-Ala 557 ; q, FIX binding to factor XI-Ala 557 ; and f, FIXa␤ binding to factor XI-Ala 557 . Data have been corrected for nonspecific binding and analyzed as described under "Experimental Procedures." The data shown in the figure are representative runs for each experiment. All experiments were performed in triplicate.
Factor IX-Gla Domain Binds to Factor XIa occurs, does not influence the results of the binding assays appreciably. The studies demonstrate that both zymogen and activated FIX bind to factor XIa with similar affinity and that a catalytically functional factor Xla molecule is not required for binding.
Recombinant FIX Proteins and Activity in Plasma Clotting Assays-To determine the importance of the FIX Gla domain in binding to, and activation by, factor XIa, recombinant versions of FIX with altered Gla domains were prepared ( Fig. 2A). Recombinant proteins were expressed in the human fibroblast cell line HEK293 because this line has been shown to ␥-carboxylate properly a high percentage of expressed recombinant vitamin K-dependent protein (37). In a standard aPTT assay, plasma-derived FIX demonstrated a specific activity of 200 units/mg (1 unit equaling the FIX activity in 1 ml of normal plasma). rFIX expressed in the presence of vitamin K had a specific activity of 75% (150 units/mg) of that of plasma FIX in the aPTT assay. FIX expressed in the presence of the vitamin K antagonist warfarin (rFIX-des␥) demonstrated significantly reduced activity (Ͻ1% normal activity or Ͻ2 units/mg of protein) when tested under the same conditions. As shown in Fig.  2B, the presence of warfarin in the cell culture results in a protein that is not recognized by a monoclonal antibody (SB 249417) that requires calcium and a properly ␥-carboxylated FIX Gla domain for protein recognition. This demonstrates that adding warfarin to the cell culture system effectively interferes with ␥-carboxylation of the Gla domain. rFIX in which the Gla domain has been replaced by the corresponding domain from factor VII (rFIX/VII-Gla; Fig. 2A) was expressed in the presence of vitamin K. This chimeric protein also demonstrated poor activity in the aPTT assay (Ͻ1% normal activity or Ͻ2 units/mg of protein).
The aPTT assay requires FIX to be activated by factor XIa and the FIXa␤ subsequently generated to activate factor X. Poor activity in this assay could, therefore, reflect a defect in one or both of these steps. To define further the abnormalities in rFIX-des␥ and rFIX/VII-Gla, we attempted to activate recombinant proteins by incubating them with a high concentration (100 nM) of factor XIa before testing them in clotting assays (modified PTT assay). This step removes the requirement for activation by factor XIa from the clotting process. rFIX-des␥ could not be activated to FIXa␤ by prolonged incubation with high concentrations of factor XIa. rFIX/VII-Gla was completely converted to the active form; however, it was noted that activation was considerably slower than with wild type rFIX (see below) or plasma FIX. In the modified PTT assay, rFIXa␤ demonstrated ϳ80% of the activity of plasma-derived FIXa␤, consistent with results from the conventional aPTT assay. Interestingly, rFIXa␤/VII-Gla also demonstrated significant activity in this assay (ϳ60% of the activity of plasma FIXa␤). This suggests that the factor VII Gla domain is a reasonably good substitute for the FIX Gla domain when FIXa␤ is incorporated into the factor Xase complex in plasma with zymogen factor X, factor VIIIa, calcium, and phospholipid. Furthermore, the finding strongly indicates that the poor performance of rFIX/VII-Gla in the aPTT assay is caused by a specific defect in activation by factor XIa rather than a global abnormality of protein structure which affects multiple FIX functions.
Recombinant FIX Protein Binding to Factor XIa Studied with SPR-Recombinant proteins were tested by SPR for their ability to bind to factor XIa (Fig. 2C). rFIX binds to plasma factor XIa with a K d of 107 nM, a value similar to those obtained for plasma-derived FIX and FIXa␤ (Fig. 1, B and C). In contrast, neither rFIX-des␥ nor rFIX/VII-Gla demonstrated binding above background (Fig. 2C). These data demonstrate that proper ␥-carboxylation of the Gla domain is required for FIX binding to factor XIa and that the factor VII Gla domain is not an adequate substitute in this interaction.
The humanized murine monoclonal antibody SB 249417 recognizes the FIX Gla domain in a calcium-dependent manner (32) 2 and is sensitive to conformational changes to the domain which accompany incomplete ␥-carboxylation (Fig. 2B). Furthermore, the antibody does not recognize rFIX/VII-Gla in Western blot assays or enzyme-linked immunosorbent assays (data not shown), indicating that binding involves epitopes specific to the FIX Gla domain. The effect of SB 249417 on FIX binding to factor XIa was examined using SPR. As shown in Fig. 3A, the antibody is a potent inhibitor of binding, with a K i of 34 nM (EC 50 320 nM).
FIX Activation by Factor XIa and RVVP-Initially, we examined the capacity of SB 249417 to inhibit FIX activation by factor XIa using a chromogenic substrate assay. Consistent with results from SPR studies (Fig. 3A), SB 249417 is a potent 2 J. Toomey and D. Gailani, unpublished observations. Factor IX-Gla Domain Binds to Factor XIa inhibitor of FIX activation by factor XIa, with a K i of 33 nM (Fig.  3B). The chromogenic substrate assay used to follow this reaction is relatively insensitive to FIXa␤ and, therefore, may not detect low levels of activation. For this reason, conversion of rFIX molecules to rFIXa␤ by a low concentration of factor XIa (1 nM) was examined by western immunoblot (Fig. 4). As can be seen in Fig. 4A, zymogen wild type rFIX undergoes nearly complete conversion to rFIXa␤ within 1 h under the conditions of the assay (Fig. 4A). No activation of rFIX-des␥ was observed (Fig. 4B). rFIX/VII-Gla appears to be slowly converted to FIXa␤, with a small increase in the activated form detectable at late time points (Fig. 4C). This finding is consistent with the earlier observation that prolonged incubation of rFIX/VII-Gla with high concentrations of factor XI will eventually result in complete conversion to the active protease. This indicates that the activation cleavage sites on rFIX/VII-Gla are accessible to factor XIa. Progress curves for activation of the recombinant proteins were constructed using densitometry measurements from Western blot autoradiographs (Fig. 4D). The initial slopes of the progress curves (first 5 min for rFIX, 60 min for rFIX-des␥ and rFIX/VII-Gla) were 4.1 nM/min for activation of rFIX, 0.2 nM/min for activation of rFIX/VII-Gla, and 0.0 nM/min for rFIX-des␥. Thus, the initial rate of activation of rFIX was ϳ20-fold greater than for rFIX/VII-Gla.
The venom of Dabois russelli (Russell's viper) contains a protease, RVVP, which cleaves FIX between Arg 180 and Val 181 at the C terminus of the activation peptide to produce an active FIX intermediate called FIXa␣ (38). As with the activation of FIX by factor XIa, this reaction is enhanced by calcium but not by phospholipid (39). rFIX proteins were incubated in the presence of purified RVVP and calcium (Fig. 5). rFIX and rFIX/VII-Gla are converted to FIXa␣ in a similar manner, consistent with this reaction not having a specific requirement for the FIX Gla domain. Furthermore, this experiment demonstrates that the conformation of rFIX/VII-Gla is sufficiently like FIX that it interacts properly with RVVP. In contrast, rFIX-des␥ is not cleaved by RVVP, indicating that poor ␥-carboxylation induces significant enough conformational changes in the protein to interfere with activation. Significant structural alteration induced by poor ␥-carboxylation could explain our inability to activate rFIX-des␥ with high concentrations of factor XIa (100 nM), even with prolonged periods of incubation (Ͼ12 h). DISCUSSION Activation of FIX, a key step in the formation and maintenance of a fibrin clot, is controlled by at least two distinct mechanisms. Initiation of fibrin formation involves binding of factor VIIa in plasma to the membrane protein tissue factor at a site of blood vessel injury (2,3). Factor VIIa/TF activates factor X and FIX in reactions that require calcium and a phospholipid surface (13,40,41). Activated factor X (factor Xa) then converts prothrombin to thrombin, which initiates fibrin formation. It is postulated that factor VIIa/TF is inhibited relatively early in the coagulation process by the TF pathway inhibitor (2,42). FIXa␤ and its cofactor, factor VIIIa, would be required for generation of factor Xa to sustain coagulation after inhibition of factor VIIa/TF. In this model, bleeding in hemophilia (congenital deficiency of factor VIII or FIX) is caused by poorly sustained factor X activation rather than a failure to initiate fibrin formation (2,43). Given the severity of the bleeding disorders associated with complete deficiencies of FIX and factor VII, it appears that FIX activation by factor VIIa is important for most hemostatic challenges. In contrast, factor XI deficiency is associated with bleeding that typically requires a more severe hemostatic insult (trauma or surgery) (44). FIX activation via factor XIa, therefore, is most likely required in certain situations to supplement FIXa␤ produced through factor VIIa/TF.
Factor XI has an unusual structure for a coagulation protease. The protein is a disulfide bond linked dimer of two iden- tical 80-kDa polypeptides (25,45) and lacks the Gla domain that is a characteristic feature of other coagulation proteases (24). The noncatalytic N-terminal "heavy chain" portion of the factor XI polypeptide comprised of four 90 -91-amino acid repeats called apple domains (A1-A4 from the N terminus), which mediate binding to proteins, platelets, and glycosaminoglycans (25,31,36,46,47). Using recombinant factor XI in which apple domains were replaced with corresponding domains from the functionally distinct protein plasma prekallikrein, it was determined that factor XI A3 likely contains a binding site for FIX (31,35). Studies based on peptide mimicry suggest that the C-terminal half of A2 is also involved in FIX binding (47). Zymogen factor XI, unlike factor XIa, does not cleave small chromogenic substrates. This indicates that conversion of zymogen to active protease involves a conformational change that gives small molecules, as well as portions of FIX, access to the catalytic active site. The SPR studies demonstrate that conversion to factor XIa is also required for binding of FIX and FIXa␤ to the protease. Binding sites for FIX are probably masked in the zymogen and become available for binding upon activation. Catalytic activity is not required for the binding interaction because factor XIa lacking an active site serine residue binds normally to FIX and FIXa␤. It is interesting that both FIX and FIXa␤ bind to factor XIa. This indicates that the activation peptide is not required for the binding interaction, which is consistent with a report showing that rFIX lacking an activation peptide is activated by factor XIa (48). Furthermore, it supports the notion that binding of FIX and factor XIa involves interactions remote from the activation cleavage sites. The importance of the dimeric structure of factor XIa to FIX binding and activation is not entirely clear. In solution, a monomeric variant of factor XIa activates FIX with kinetic parameters that are similar to those for activation by wild type factor XIa (31). However, recent work suggests that the physiologic environment for this reaction may be the surface of activated platelets (46,49,50). In this environment, the dimeric structure of factor XI is necessary for normal FIX activation, possibly because one polypeptide of the dimer is required for binding to the platelet, whereas the other interacts with FIX (46,49,50).
The structural elements on FIX required for binding to factor XIa have not been clearly defined. The noncatalytic portion of the molecule contains (from N terminus to C terminus) the Gla domain, a short aromatic stack region, two EGF-like domains, and the activation peptide (3). Some Gla domain mutations causing hemophilia B are associated with poor factor XIa mediated activation, as is the failure to remove the propeptide from the N terminus of the molecule (26,27). Similarly, single amino acid substitutions in the EGF-1 (factor IX New London) (51) and EGF-2 (52) domains associated with cross-reactive material-positive hemophilia B are activated poorly by factor XIa. However, all of these mutations interfere with other activities of FIX in addition to activation by factor XIa. A rFIX protein containing the factor VII EGF1 domain is activated normally by factor XIa (53,54), whereas a study employing alanine scanning mutagenesis on portions of the EGF2 domain identified only a single amino acid (position 89) that appeared to be necessary for normal activation by factor XIa (55).
Using SPR, plasma coagulation, antibody inhibition, and purified protein activation assays we have demonstrated that the FIX Gla domain either contains all or a portion of the factor XIa binding site or is required for the proper conformation of the binding site. The results with chimera rFIX/VII-Gla are particularly intriguing. In its active form, this molecule appears to activate factor X reasonably well in a plasma clotting assay and is activated by the snake venom protease RVVP in a manner similar to wild type FIX. This indicates that rFIX/VII-Gla is structurally and catalytically similar to FIX and strongly suggests that the failure of the protein to demonstrate activity in a standard aPTT assay is because of a specific defect in factor XIa-mediated activation. SPR confirmed that this protein has a profound defect in binding to factor XIa. Activation of factor X by FIXa␤ is a phospholipid-dependent reaction involving the Gla domains of both proteins (8). It would appear that the Gla domain of factor VII is similar enough to that of FIX to substitute for it in this reaction. In contrast, factor VII-Gla cannot substitute for FIX-Gla during FIX activation by factor XIa, a reaction not influenced by phospholipid. Liebman and co-workers (28) demonstrated that a Fab fragment of a monoclonal antibody that interacts with the factor IX Gla domain phospholipid-binding epitope inhibited activation of factor IX by factor XIa. This group postulated that factor XIa may have a surface domain with features similar to those of phospholipid vesicles which may mediate the interaction with factor IX. However, phospholipid, although not enhancing FIX activation by factor XIa, does not inhibit the reaction either (23). 3 This suggests that the FIX Gla domain may contain one or more epitopes for a protein-protein interaction with factor XIa and that the epitopes are distinct from the phospholipid-binding elements on the domain. This is not to say that FIX binding to phospholipid is not relevant to activation of the protein by factor XIa. As mentioned above, a physiologic site for the reaction may well be the surface of activated platelets. Although factor XIa binds to platelets largely through a protein-protein interaction involv-3 D. Gailani, unpublished observations. ing glycoprotein 1b (56), it is likely that FIX would bind to the platelet surface, at least in part, through a phospholipid-protein interaction involving the Gla domain. Work is under way to identify specific residues in the Gla domain involved in the interaction between FIX and factor XIa, to determine whether they are distinct from those required for phospholipid binding.
The notion that a Gla domain may be involved in a binding interaction not involving phospholipid has precedent. Amino acids 3-11 in the FIX Gla domain are critical for high affinity binding of the protein to aortic endothelial cells (57). It appears that this interaction is caused by binding of FIX to collagen IV and not phospholipid in the cell membrane (58). Using scanning force microscopy, Wolberg and colleagues (59) demonstrated that FIX binds specifically to collagen IV in the later molecule collagenous domain. Regan and co-workers (60) have shown that the protein C Gla domain is involved in a proteinprotein interaction with the protein C receptor on endothelial cells. A recombinant prothrombin molecule containing the protein C Gla domain bound specifically to cells expressing the protein C receptor. Along similar lines, Lockett and Mast (61) presented data suggesting that an interaction between the C terminus of TF pathway inhibitor and the factor Xa Gla domain is required for proper TF pathway inhibitor mediated-inhibition of factor Xa. Thus, Gla domain involvement in proteinprotein interactions may be common.