Contribution of basic residues of the autolysis loop to the substrate and inhibitor specificity of factor IXa.

The autolysis loop (residues 143-154 in chymotrypsinogen numbering) plays a pivotal role in determining the macromolecular substrate and inhibitor specificity of coagulation proteases. This loop in factor IXa (FIXa) has 3 basic residues (Arg143, Lys147, and Arg150) whose contribution to the protease specificity of factor IXa has not been studied. Here, we substituted these residues individually with Ala in Gla-domainless forms of recombinant factor IX expressed in mammalian cells. All mutants exhibited normal amidolytic activities toward a FIXa-specific chromogenic substrate. However, Arg143 and Lys147 mutants showed a approximately 3- to 6-fold impairment in FX activation, whereas the Arg150 mutant activated factor X normally both in the absence and presence of factor VIIIa. By contrast, Arg143 and Lys147 mutants reacted normally with antithrombin (AT) in both the absence and presence of the cofactor, heparin. However, the reactivity of the Arg150 mutant with AT was impaired 6.6-fold in the absence of heparin and 33- to 70-fold in the presence of pentasaccharide and full-length heparins. These results suggest that Arg143 and Lys147 of the autolysis loop are recognition sites for FX independent of factor VIIIa, and Arg150 is a specific recognition site for AT that can effectively interact with AT only if the serpin is in the heparin-activated conformation.

Factor IXa (FIXa) 1 is a vitamin K-dependent trypsin-like serine protease in plasma that binds to factor VIIIa (FVIIIa) on membrane surfaces in the presence of Ca 2ϩ (intrinsic Xase) to activate factor X (FX) to factor Xa (FXa) during the blood coagulation process (1)(2)(3)(4)(5). FIXa plays a key role in hemostasis, and its deficiency is associated with the life-threatening disease, hemophilia B (6). FIX circulates in plasma as a singlechain inactive zymogen with 415 amino acids (6). Proteolytic cleavage at two sites (Arg 145 -Ala, and Arg 180 -Val) by either one of two physiological activators, the factor VIIa-tissue factor complex or factor XIa, releases an activation peptide that converts the zymogen to an active enzyme consisting of a light and a heavy chain held together by a single disulfide bond (7)(8)(9). The N-terminal light chain of FIXa contains the non-catalytic Gla and two epidermal growth factor (EGF)-like domains, whereas the C-terminal heavy chain contains the trypsin-like catalytic domain of the molecule (10). The Gla and the Nterminal EGF domains are involved in the Ca 2ϩ -dependent assembly of the protease with FVIIIa on membrane surfaces (10,11). It is believed that the protease domain of FIXa also interacts with FVIIIa (12). The role of the C-terminal EGF domain of FIXa is not known. Crystallographic and mutagenesis data suggest that it makes extensive interactions with the catalytic domain and thus may be critical for the structural integrity and proper orientation of the catalytic domain for interaction with either substrate and/or cofactor on membrane surfaces (13,14). In the absence of a cofactor, FIXa exhibits very low catalytic activity toward both small synthetic and natural macromolecular substrates and inhibitors. Although FVIIIa enhances the proteolytic activity of FIXa toward its natural substrate FX by four or five orders of magnitude, the cofactor has no effect on the catalytic activity of FIXa toward small synthetic substrates (1,11,15). The exact mechanism by which FVIIIa improves the catalytic competency of FIXa toward FX is not known. Recent structural and mutagenesis data have indicated that the "99-loop" (per nomenclature of Ref. 25) of FIXa traps the protease in an inactive conformation by blocking the entrance of FX into the catalytic pocket. Thus, one mechanism by which FVIIIa may activate FIXa involves conformational rearrangement of the 99-loop (16).
Similar to its reaction with FX, the reactivity of FIXa with the target serpin inhibitor antithrombin (AT) is very poor in plasma (17,18). In this case, heparin functions as a cofactor to accelerate the rate of FIXa inhibition by the serpin by several orders of magnitude (18,19). Heparin has more recently been shown to accelerate the inhibition of FIXa and FXa by AT by similar mechanisms (20 -22). For both enzymes, the rate-accelerating effect of heparin is believed to arise from 1) the ability of heparin to change the conformation of the serpin to facilitate its optimal recognition by the enzyme (allosteric activation mechanism) and 2) the ability of a full-length heparin to bridge the inhibitor and enzyme in a ternary complex, thereby promoting the initial interaction between the two proteins (template mechanism). With respect to the latter template mechanism, recent mutagenesis data have indicated that both FIXa (20) and FXa (21) have a cluster of basic residues in their catalytic domains that can interact with heparin when physiological levels of Ca 2ϩ are present. Thus, the simultaneous binding of a full-length heparin to the basic site of either protease in the presence of Ca 2ϩ and to a basic site of AT improves the K D for the interaction of the two proteins, leading to a 100-to 1000-fold acceleration of the reaction (20 -22). In the case of the allosteric activation mechanism, the binding of a distinct pentasaccharide sequence of the full-length heparin to AT allosterically changes the conformation of the serpin, leading to an additional 200-to 500-fold improvement in reactivity of the serpin with both FIXa and FXa (20,22,23). Allosteric activation of AT by heparin has minimal effect on its reactivity with thrombin (23). It is not known how allosteric activation of AT by heparin specifically improves the reactivity of the serpin with the coagulation proteases, FIXa and FXa.
Structural data suggest that the catalytic domain of FIXa, similar to other coagulation proteases, has several solventexposed surface loops that surround the substrate-binding pocket of the enzyme (13,24). Although these surface loops are generally conserved at similar three-dimensional locations on all coagulation proteases, the length and amino acid residues forming the loops are not conserved in the members of the family. One of these surface loops consisting of residues 143-154 (chymotrypsinogen numbering (25)), referred to as the autolysis loop, has been demonstrated to play a crucial role in determining the substrate and inhibitor specificity of coagulation proteases (26 -29). In the case of FXa, we recently demonstrated that the basic residues of the autolysis loop minimally contribute to recognition of the substrate prothrombin in either the absence or presence of factor Va on negatively charged phospholipid vesicles (30). In its reaction with AT, however, we discovered that basic residues of this loop play distinct roles in FXa recognition of AT. Thus, neutralization of charges of both Arg 143 and Lys 147 by their substitution with Ala improved the reactivity of the mutant proteases with AT in the absence of heparin, whereas the substitution of Arg 150 with Ala led to an order of magnitude impairment in the reactivity of the mutant with AT specifically in the presence of heparin pentasaccharide (30). These results suggested that Arg 150 of factor Xa makes a specific interaction with AT when the serpin is in an activated conformation, presumably involving a recently proposed serpin exosite (31). Interestingly, similar to FXa and unlike thrombin, the autolysis loop of FIXa contains three basic residues that are conserved at the same three-dimensional locations (13,32). To understand the contribution of these basic residues to determining the substrate and inhibitor specificity of FIXa, we prepared three Gla-domainless mutants of FIX (GD-FIX) in which these residues were substituted with Ala. The mutants, GD-FIX R143A, K147A, and R150A together with the wild-type GD-FIX, were expressed in mammalian cells and purified to homogeneity (20). Following activation by Russell's viper venom protease, the catalytic properties of the mutants were characterized with respect to their ability to hydrolyze a FIXaspecific chromogenic substrate, activate the natural substrate FX in both the absence and presence of FVIIIa, and react with AT in either the absence or presence of pentasaccharide and full-length heparins. The results indicate that the basic residues Arg 143 and Lys 147 are critical for interaction of FIXa with FX independent of FVIIIa. Moreover, Arg 150 is a recognition site on FIXa for AT that can effectively interact with AT only if the serpin is in the heparin-activated conformation.

Mutagenesis and Expression of Recombinant
Proteins-The wildtype and FIX mutants, Arg 143 3 Ala (R143A), Lys 147 3 Ala (K147A), and Arg 150 3 Ala (R150A) (chymotrypsinogen numbering (25)) corresponding to residues Arg 312 , Lys 316 , and Arg 318 in the FIX cDNA numbering (6), were generated in the RSV-PL4 expression/purification vec-tor system in Gla-domainless forms (GD-FIX) by standard PCR mutagenesis methods as described before (20). After confirmation of the accuracy of the mutagenesis by DNA sequencing, the constructs were introduced into HEK293 cells, and the mutant proteins were isolated from 20 liters of cell culture supernatants by an immunoaffinity chromatography using the HPC4 monoclonal antibody as described previously (20).
Human plasma FIXa, FX, FXa, and FX activating enzyme from Russell's viper venom (RVV-X) were purchased from Hematologic Technologies Inc. (Essex Junction, VT). Human recombinant AT was prepared as described previously (33). Phospholipid vesicles containing 80% phosphatidylcholine and 20% phosphatidylserine (PC/PS) were prepared as described previously (34). Human recombinant FVIIIa was a generous gift from Dr. Philip Fay (University of Rochester, Rochester, NY). Unfractionated heparin (heparin sodium injection, USP, 10,000 units/ml) from beef lung and the active AT-binding pentasaccharide fragment of heparin (fondaparinux sodium) were purchased from Quintiles Clinical Supplies (Mt. Laurel, NJ). The concentration of pentasaccharide, used in molar excess over AT in all reactions, was calculated based on its weight assuming a molecular mass of 1728 Da (C 31 H 43 N 3 Na 10 O 49 S 8 ). The concentration of the unfractionated heparin was based on antithrombin binding sites and was determined by stoichiometric titrations of antithrombin with the polysaccharide, with monitoring of the interaction by changes in protein fluorescence as described previously (35). Heparin-Sepharose was purchased from Amersham Biosciences (Piscataway, NJ), the chromogenic substrates, Spectrozyme FXa (SpFXa) and CH 3 SO 2 -D-Leu-Gly-Arg-pNA (LGR-pNA) were purchased from American Diagnostica (Greenwich, CT) and Midwest Bio-Tech Inc. (Fishers, IN), respectively. Polybrene and paminobenzamidine were purchased from Sigma (St. Louis, MO).
Activation of FIX Mutants-FIX derivatives (ϳ0.5 mg) in 0.02 M Tris-HCl (pH 7.5), 0.1 M NaCl (TBS buffer) containing 5 mM Ca 2ϩ were converted to their active forms with 50 g of RVV-X for 3 h at 37°C as described previously (20). FIXa derivatives were separated from the snake venom by applying the activation mixtures on a 2-ml heparin-Sepharose column pre-equilibrated with the same buffer as described previously (20). The column was extensively washed with the same buffer containing 0.15 M NaCl followed by elution with 0.5 M NaCl. The active fractions were identified by an amidolytic activity assay, pooled, and active site-titrated with known concentrations of AT in the presence of pentasaccharide as described (20).
Cleavage of Chromogenic Substrates-The steady-state kinetics of hydrolysis of LGR-pNA (0.04 -5 mM) by FIXa derivatives (20 nM) were studied in TBS containing 0.1 mg/ml bovine serum albumin (BSA), 33% ethylene glycol, and 5 mM Ca 2ϩ . In some experiments ethylene glycol was omitted from the buffer. The initial rate of hydrolysis was measured at 405 nm at room temperature in 96-well plates by a V max Kinetic Microplate Reader (Molecular Devices, Menlo Park, CA) as described (20). The apparent K m and k cat values for substrate hydrolysis were obtained by non-linear regression fits of data to the Michaelis-Menten equation, and the catalytic efficiencies were expressed as the ratio of k cat /K m .
Factor X Activation by Factor IXa Mutants-The initial rate of FX activation by the wild-type and mutant GD-FIXa derivatives was measured in both the absence and presence of FVIIIa. In the absence of the cofactor, the time course of the initial rate of activation was measured by incubating each FIXa derivative (50 nM) with human FX (2 M) on PC/PS vesicles (50 M) in TBS containing 0.1 mg/ml BSA, 0.1% polyethylene glycol (PEG) 8000 and 5 mM Ca 2ϩ (TBS/Ca 2ϩ ) at room temperature. At different time intervals, small aliquots of the activation reactions were transferred into a 96-well assay plate containing Spectrozyme FXa (200 M in TBS containing 20 mM EDTA, and in a 100-l final volume). The initial rate of activation was determined by measuring the increase in amidolytic activity of the samples toward SpFXa as monitored at 405 nm by a V max Kinetic Microplate Reader. The concentrations of FXa generated in the activation reactions were determined from a standard curve prepared from the cleavage rate of SpFXa by known concentrations of plasma FXa under exactly the same conditions (20).
In the presence of factor VIIIa, the activation of factor X by the FIXa derivatives on phospholipid vesicles was evaluated by incubating each FIXa (20 nM) with varying concentrations of human FVIIIa (7.8 -500 nM) on 50 M PC/PS vesicles in TBS/Ca 2ϩ at room temperature. The activation reactions were initiated by addition of 1 M human FX and allowed to proceed for 60 min, following which they were terminated by addition of EDTA to a final concentration of 20 mM. The rate of FXa generation was measured from the cleavage rate of SpFXa as described above. The apparent K d for the factor VIIIa-FIXa interaction and max-imal FX activation rate for wild-type and variant FIXas were determined by non-linear regression fits of data by a hyperbolic kinetic equation (30).
Reaction with Antithrombin-The rate of inactivation of FIXa derivatives by AT in both the absence and presence of unfractionated fulllength heparin and the pentasaccharide fragment of heparin were measured under pseudo-first-order conditions by a discontinuous assay as described previously (20). In the absence of a cofactor, each FIXa derivative (10 nM) was incubated with AT (0.5-2.0 M) in TBS containing 0.1 mg/ml BSA, 0.1% PEG 8000, and 5 mM CaCl 2 for 30 -120 min. All reactions were carried out in 50-l volumes in 96-well polystyrene assay plates at room temperature. In the presence of pentasaccharide, the reaction conditions were the same except that FIXa derivatives were incubated with AT (50 -400 nM) in complex with a saturating concentration of pentasaccharide (1 M) for 0.5-120 min. In the presence of full-length heparin, each FIXa derivative (10 nM) was incubated with 200 nM AT in the presence of catalytic levels of heparin (0.6 -10 nM) for 10 s to 50 min in the same TBS buffer system. All reactions were stopped by addition of 50 l of LGR-pNA (final concentration of 1.0 mM) in TBS containing 1 mg/ml Polybrene. The remaining enzyme activity was measured with a V max Kinetics Microplate Reader at 405 nm as described above. The observed pseudo-first-order inactivation rate constants (k obs ) were determined by fitting the time-dependent change of the protease activity to a first-order rate equation. Second-order rate constants (k 2 ) were obtained from the slopes of linear plots of k obs versus the concentration of AT or AT-heparin complex as described (20). The AT-heparin complex concentrations were calculated from the dissociation constant for the AT-heparin interaction and total concentrations of AT and heparin using the quadratic equation as described (35).
Inhibition by p-Aminobenzamidine-The K i value for p-aminobenzamidine interaction with the active-site pocket of wild-type and mutant GD-IXa derivatives was evaluated. In all cases, GD-FIXa (20 nM) was incubated with increasing concentrations of the inhibitor (0 -320 M) in the presence of different fixed concentrations of LGR-pNA (312-2500 M) in TBS containing 0.1 mg/ml BSA, 33% ethylene glycol, and 5 mM Ca 2ϩ . The enzyme activity was measured from the initial cleavage rate of the chromogenic substrate as described above and the K i values were determined by global fitting of data to a kinetic equation for competitive inhibition as described (36).

Expression, Purification, and Activation of Recombinant
Proteins-Wild-type and Ala substitution mutants of FIX in Gladomainless forms were expressed in HEK293 cells and purified by immunoaffinity chromatography as in previous studies (20). SDS-PAGE analysis of the purified proteins under non-reducing conditions indicated that the isolated proteins are relatively pure and that they all migrate with similar expected molecular masses of ϳ60 kDa suggesting that 293 cells properly processed the recombinant proteins (Fig. 1). All GD-FIX derivatives were activated by RVV-X and, following their purification on a heparin-Sepharose column as described (20), were found to be fully active based on active-site titrations with AT in the presence of pentasaccharide.
Amidolytic Activity-Kinetic parameters for the hydrolysis of the chromogenic substrate LGR-pNA by FIXa derivatives are presented in Table I. Wild-type and mutant GD-FIXa derivatives exhibited similar or slightly improved k cat and K m values toward LGR-pNA in both the absence (data not shown) and presence of the activator, ethylene glycol (Table I). The mu-tagenesis of basic residues of the autolysis loop thus has no apparent adverse effect on the S3-S1 substrate binding pockets of FIXa mutants. In agreement with this finding, K i values for the interaction of all mutants with the S1 site-specific competitive inhibitor of serine proteases, p-aminobenzamidine, were comparable to or slightly improved relative to the wild-type enzyme (Table I). Previous mutagenesis studies of basic residues of the autolysis loop of FXa yielded similar results (30). Because the residues under study are solvent-exposed in the crystal structure of human FIXa (13), their mutagenesis is not expected to affect the catalytic pocket of FIXa.
Factor X Activation-The catalytic properties of FIXa derivatives toward activation of the natural substrate FX was studied in both the absence and presence of FVIIIa and Ca 2ϩ on PC/PS vesicles. The time course of FX activation in the absence of the cofactor indicated that GD-FIXa can convert FX to FXa at a very slow rate. Under the experimental conditions described under "Experimental Procedures" (2 M FX and 50 nM FIXa), similar initial rates of ϳ0.03 nM FXa/min were found for the wild-type and the R150A mutant of GD-FIXa. However, under the same conditions, the initial rate of FX generation by both R143A and K147A mutants (ϳ0.005 nM/min) was impaired ϳ6-fold. The initial rates of FX activation by all GD-FIXa derivatives were 7-to 8-fold slower in the absence than in the presence of PC/PS vesicles. The ability of GD-FIXa derivatives to activate FX in the presence of FVIIIa was also evaluated. To determine the effect of deleting the Gla domain on the high affinity interaction of FIXa with FVIIIa, the rate of FX activation by plasma FIXa and recombinant GD-FIXa derivatives were compared in the presence of increasing concentrations of FVIIIa on PC/PS vesicles. Interestingly, relative to full-length plasma FIXa, the apparent K d for the interaction of wild-type GD-FIXa with FVIIIa (82.7 nM) was weakened only 3.5-fold ( Fig. 2 and Table II). However, the rate of FXa generation by GD-FIXa was dramatically impaired (Fig. 2). In the presence of a saturating concentration of FVIIIa, the maximum rate of activation with the plasma-derived FIXa (211.8 nM/min/ nM) was 20,000-fold higher than the corresponding value with wild-type GD-FIXa (10.6 pM/min/nM, Table II). The R150A mutant exhibited a similar maximal rate of FXa generation and K d (app) for FVIIIa interaction, suggesting that Arg 150 does not play a role in the macromolecular recognition of factor X by the FIXa-FVIIIa complex. In the case of the R143A and K147A mutants, the maximal FX activation rates were impaired 3-to 4-fold and the K d (app) for interaction with FVIIIa was unaffected for R143A, but impaired ϳ3-fold for the K147A mutant (Table II).
Reaction with Antithrombin-We previously demonstrated that removal of the Gla domain from FIXa does not affect its reactivity with AT in either the absence or presence of heparin cofactors when physiological levels of Ca 2ϩ are present (20). Thus, the reactivities of only GD-FIXa derivatives with AT were evaluated under these conditions. Relative to the wildtype enzyme, there was minimal change in the reactivity of either R143A or K147A mutant enzymes with AT both in the absence and presence of pentasaccharide or full-length heparins (Table III). On the other hand, the R150A mutant reacted with AT with 6.6-, 70-, and 33-fold slower second-order association rate constants (k 2 ) than the wild-type enzyme when no heparin, pentasaccharide heparin, or full-length heparin was present, respectively ( Fig. 3 and Table III). The rate enhancement due to conformational activation of AT is given by the ratio of k 2 values in the presence and absence of pentasaccharide (Table III). The conformational activation rate enhancement was minimally affected by the R143A and K147A mutations, whereas it was reduced 10-fold by the R150A mutation. The rate enhancement due to the template effect of the fulllength heparin can be obtained from the ratio of k 2 values in the presence of heparin to those in the presence of the pentasaccharide (Table III). None of the mutants were found to be impaired with respect to the template effect of heparin. DISCUSSION It has become apparent in recent years that residues on exposed surface loops, near or remote from the catalytic pocket, play key roles in determining the macromolecular substrate and inhibitor specificity of coagulation proteases. It is believed that these loops either directly interact with target substrate or inhibitor molecules or provide recognition sites for interaction with specific cofactors of coagulation proteases on membrane surfaces. In this study, we demonstrated that the basic residues of the autolysis loop (Fig. 4) play specific roles in the catalytic function of FIXa in its reaction with the physiological substrate, FX, and the serpin inhibitor, AT. Thus, the two residues Arg 143 and Lys 147 of the loop are likely cofactor-independent recognition sites for FX based on our observation that the rate of FX activation was impaired to similar extents by mutating either residue in both the absence and presence of FVIIIa. On the other hand, Arg 150 does not play a significant role in the protease interaction with the substrate, because mutation of this residue was found to minimally alter the rate of FX activation. Conversely, both Arg 143 and Lys 147 were dispensable for FIXa recognition of AT in either the absence or presence of heparin cofactors, whereas Arg 150 was discovered to be an essential residue for the protease interaction with the serpin. The observation that the reactivity of the R150A mutant with AT is impaired in both the absence and presence of pentasaccharide suggests that Arg 150 interacts with both the native and activated conformations of the serpin. Interestingly, however, the impairment was 10-fold greater in the presence than in the absence of pentasaccharide, indicating that this residue interacts effectively with AT only when the serpin is in its activated conformation.
Similar to FIXa, the autolysis loop of FXa is basic with four residues Arg 143 , Lys 147 , Arg 150 , and Arg 154 conserved at the same three-dimensional locations (32,37). We recently demonstrated that Arg 150 of FXa plays a similar role in the protease reaction with AT. This was evidenced by the observation that the reactivity of the Arg 150 mutant of FXa with AT was impaired by an order of magnitude specifically in the presence of the pentasaccharide (30). Unlike FIXa and FXa, Arg 150 is not conserved in the autolysis loop of thrombin, which does not differentiate between the native and activated conformations of AT (23,24). These results suggest that structural differences in the autolysis loops of coagulation proteases are partly responsible for their differential reactivity with the native and the heparin-activated conformations of AT. The mechanism by which heparin activates AT is not completely understood. It was initially believed that the residues of the reactive-site loop of AT may not have a proper conformation to fit the catalytic pocket of coagulation proteases and that activation by heparin   allosterically changes the conformation of the reactive-site loop of the serpin, leading to its effective docking into the active-site pocket of certain target proteases (23,31,38,39). Recent mutagenesis studies, however, suggest that the determinants of the heparin-enhanced reactivity of AT with coagulation proteases do not reside in the reactive loop. This was demonstrated by the observation that mutagenesis of the P6 -P3Ј residues of the reactive site loop of AT did not lead to a noticeable difference in the pentasaccharide enhancement of the FXa inhibition rate by the mutant serpins (31). Based on such results, it has been hypothesized that allosteric activation of AT may expose a secondary binding site or "exosite" on the serpin, outside the P6 -P3Ј reactive site loop sequence, that interacts with target proteases (31). Our previous results with FXa (30) and current results with FIXa suggest that Arg 150 of the autolysis loop in both proteases interacts with a site on AT. However, this site on the native serpin is not appropriately positioned or sufficiently accessible for optimal interaction with either protease. Rather, conformational activation is necessary to make the putative recognition site on AT available for interaction with the autolysis loop of these proteases. Arg 150 of factors IXa and Xa thus appears to specifically interact with the exosite, which is made accessible in heparin-activated AT. However, the finding of only a partial ablation of the pentasaccharide enhancement of AT inhibition of the R150A mutants of both FIXa and FXa (30) suggests that the interaction of the autolysis loop with the putative exosite of AT may not account for all of the allosteric activation effect of the serpin. Another previous mutagenesis study (40) proposed a role for Gln 61 of the 60-loop in specifically interacting with the activated conformation of AT. Further mutagenesis studies, incorporating both mutations in one construct, are in progress to determine whether the interaction of both the autolysis and 60-loops of FIXa and FXa with AT can account for all the accelerating effect of heparin due to the allosteric mechanism.
For efficient substrate activation, the vitamin K-dependent coagulation proteases must have an intact Gla domain to assemble with their cofactors on negatively charged membrane surfaces in the presence of calcium. However, previous results have indicated that soluble tissue factor can bind to Gla-domainless factor VIIa (41) and factor Va can bind to GD-FXa (42) to enhance the catalytic efficiency of these proteases to activate their target physiological substrates but with impaired rate constants. Such results have indicated that these cofactors The inactivation rate constants in the absence of cofactor (k 2.uncat ) were determined from the residual activities of the FIXa derivative (20 nM) after incubation at room temperature with AT (0.5-2.0 M) for 30-120 min in TBS buffer, containing 5 mM CaCl 2 , 0.1 mg/ml BSA, and 0.1% PEG 8000, as described under "Experimental Procedures." The k 2,penta values were determined by the same procedures except that 10 nM enzyme was incubated with 50-400 nM AT for 0.5-120 min in the presence of 1 M pentasaccharide. The k 2,H values were determined by the same procedures except that the FIXa derivatives were incubated with AT (200 nM) in the presence of heparin (0.6-10 nM) for 10 s to 50 min. k 2,penta and k 2,H values were determined from the slopes of linear plots of k obs versus the concentration of AT-heparin complexes shown in Fig. 3  have both membrane and Gla-independent recognition sites for interaction with the coagulation proteases. In the case of FIXa, basic residues of the C-terminal 162-helix in the catalytic domain have been proposed to constitute a recognition site for interaction with FVIIIa (12). The same basic helix of FXa has also been demonstrated to be critical for interaction with factor Va and/or prothrombin in the prothrombinase complex (21). Both EGF-like domains of FIXa have also been implicated in binding to FVIIIa (43). However, it is not known if the Gla domain of FIXa interacts with the cofactor. The observation that FVIIIa functioned as a cofactor to enhance the rate of FX activation by GD-FIXa supports the proposal that the EGF and or/protease domains of FIXa have one ore more binding sites for the cofactor. Moreover, the finding that the apparent K d for the interaction of GD-FIXa with FVIIIa was only impaired 3.5-fold relative to that with full-length FIXa suggests that the Gla domain of FIXa may not significantly contribute to the high affinity of the protease-cofactor interaction. However, the observation that the catalytic efficiency of GD-FIXa toward FX at a saturating concentration of FVIIIa was impaired more than four orders of magnitude suggests that the Gla domain of FIXa plays a pivotal role in the cofactor-dependent activation of FX by the activation complex. The binding of the Gla domain to membrane surfaces may be required for an appropriate topographical arrangement of the protease domain in the activation complex for proper interaction with the substrate. Although GD-FIXa cannot interact with the membrane surface, its ability to activate FX was stimulated 7-to 8-fold by PC/PS vesicles. Thus, the binding of FX to the phospholipid surface through the Gla domain may be associated with a conformational change in the structure of the substrate that facilitates its optimal recognition by the protease. The Gla domain may also play a similar role in the function of FIXa. It is worth noting that previous results have indicated that the apparent K d for the interaction of Gla-domainless FXa with factor Va is three order of magnitude worse than that of wild-type FXa as determined in a prothrombinase assay (42). However, as indicated above, there is only small difference between the apparent K d values of FIXa and GD-FIXa with FVIIIa as determined by the FXa generation assay. These results appear to suggest that the two procoagulant proteases interact differently with their respective cofactors and that in the case of FIXa, the interaction of FVIIIa with one or more sites outside of the Gla domain primarily accounts for the binding energy of the protease-cofactor complex. A recent study that substituted the Gla and the Nterminal EGF domains of FXa with corresponding regions of FIXa reported that neither one of these domains of FXa interacts with factor Va as determined in a prothrombinase assay (44). However, another earlier study that employed an essentially identical FIXa/FXa chimeric molecule showed that the FXa mutant interacts with factor Va with 15-fold lower affinity (45). The reason for the discrepancy between these previous studies is not known. The possibility that the Gla domains of the two proteases, at least with respect to their interaction with cofactors on PC/PS vesicles, are functionally interchangeable needs to be investigated.
The observation that the rate of FXa generation by both R143A and K147A mutants was impaired to a similar extent in both the absence and presence of the cofactor suggests that none of the basic residues of the autolysis loop may interact with FVIIIa. The molecular basis for the impaired K147A-FVIIIa interaction may be explained by the observation that the 126-helix and 162-helix of FIXa contain binding sites for FVIIIa (43). Both of these helices are connected to the autolysis loop by a single ␤-strand, suggesting that mutation of Lys 147 may be associated with a conformational change in either one of these FVIIIa-binding sites that leads to slightly weaker interaction of the FIXa mutant with the cofactor. In support of this, it has been shown that proteolysis at the Lys 147 -Gly 148 bond of the FIXa autolysis loop also leads to an impaired interaction with FVIIIa (43). A previous mutagenesis study, which substituted Lys 147 of full-length FIXa with an Ala, essentially yielded the same results we obtained with the K147A mutant (29). Similar to the GD-FIXa mutant, the full-length FIXa mutant activated FX at a 4-to 10-fold reduced rate in both the absence and presence of FVIIIa on negatively charged membrane surfaces (29). Moreover, consistent with our results, the previous study showed that Lys 147 contributes less than 2-fold to the reactivity of FIXa with AT (29).
In summary, we have shown in this study that the three basic residues of the autolysis loop play distinct roles in determining the natural substrate and inhibitor specificity of FIXa. Although both Arg 143 and Lys 147 are cofactor-independent recognition sites for FX, neither residue makes a significant contribution to the interaction of FIXa with AT. On the other hand, Arg 150 has a minimal role in the interaction of FIXa with FX with or without cofactor, but it is essential for FIXa recognition of both the native and heparin-activated conformations of AT. Moreover, our results suggest that the putative site on AT that interacts with Arg 150 of FIXa is cryptic. Thus, heparin activation is required to expose a specific recognition exosite on AT for Arg 150 in the autolysis loop of both FIXa and FXa. These results provide a plausible explanation for the inability of thrombin and possibly other trypsin-like coagulation proteases, which lack an Arg at position 150, to differentiate between the native and the heparin-activated conformations of AT.