Molecular Characterization of an Extended Binding Site for Coagulation Factor Va in the Positive Exosite of Activated Protein C*

The anticoagulant human plasma serine protease, activated protein C (APC), inhibits blood coagulation by specific inactivation of the coagulation cofactors factor Va (FVa) and factor VIIIa. Site-directed mutagenesis of residues in three surface loops of a positive exosite located on APC was used to identify residues that play a significant role in binding to FVa. Eighteen different residues were mutated to alanine singly, in pairs, or in triple mutation combinations. Mutant APC proteins were purified and characterized for their inactivation of FVa. Three APC residues were identified that provide major contributions to FVa interactions: Lys193, Arg229, and Arg230. In addition, four residues made significant minor contributions to FVa interactions: Lys191, Lys192, Asp214, and Glu215. All of these residues primarily contribute to APC cleavage at Arg506 in FVa and play a small role in the interaction of APC with the Arg306 cleavage site. In conjunction with previously published work, these results define an extensive FVa binding site in the positive exosite of APC that is primarily involved in binding and cleaving at Arg506 on FVa.

Activated protein C (APC) 1 provides a critical feedback mechanism to inhibit the coagulation pathway. Protein C circulates in the blood as an inactive serine protease zymogen that is activated by thrombin when it is complexed with thrombomodulin on the surface of endothelial cells (1,2). APC then inactivates the procoagulant cofactors factor Va (FVa) and factor VIIIa via specific cleavages of these proteins (3)(4)(5).
Recent evidence suggests that a basic exosite located on APC is critical for recognition and inactivation of FVa. This exosite was first recognized in a molecular model of the serine-protease domain of APC and was proposed to be a likely candidate for protein-protein interactions (6). The presence of the exosite was confirmed in the x-ray crystallographic structure of the gladomainless APC (7). The exosite is generally located in the same position as anion binding exosite I of thrombin, an exosite that is involved in a variety of thrombin-protein interactions. Subsequently, studies have been done to characterize the in-teractions of this exosite with the APC substrate, FVa. Peptide inhibition studies clearly implicated the autolysis loop (residues 301-316, chymotrypsin (CHT) residues 142-153) in binding of APC to FVa (8). Alanine scanning of individual residues in this loop was used to identify residues Arg 306 , Lys 311 , Arg 312 , and Arg 314 as critical residues in the interaction of APC with FVa. Furthermore, these residues are primarily involved in interactions with the Arg 506 cleavage site on FVa, and mutation of them has little effect on Arg 306 cleavage (9). Another study determined that the naturally occurring type II APC mutation of Arg 352 to Trp (CHT 187) results in defective FVa inactivation as well (10). Residues in loop 37 (CHT 37-39) were also determined to be important for FVa Arg 506 cleavage, whereas one mutation in the calcium binding loop (CHT 70 -80) at Arg 229 (CHT 74) and mutations in loop 60 (CHT 60 -67) had little effect on FVa inactivation (11). Studies have also implicated this positive exosite in interactions with thrombomodulin and heparin (12,13).
As a continuation of our autolysis loop studies, we undertook a complete alanine-scanning mutagenesis of the remaining surface loops in the basic exosite of APC (Fig. 1). We inspected the crystal structure of gla-domainless APC to verify that candidate residues were on the surface of APC and mutated every surface-exposed residue in these loops to alanine, either singly or in pairs. Additionally we made double and triple mutation combinations of some individual mutations that had negative effects. In loop 37, which contains residues 190 -193 (CHT 36 -39), residues Ser 190 , Lys 191 , Lys 192 , and Lys 193 were mutated. In loop 60, which contains residues 214 -222 (CHT 60 -68), residues Asp 214 , Glu 215 , Lys 217 , Lys 218 , and Arg 222 were mutated. In the calcium-binding loop, which contains residues 225-235 (CHT 70 -80), residues Asp 227 , Arg 229 , Arg 230 , Trp 231 , Glu 232 , Lys 233 , and Trp 234 were mutated. Combined with previously published work (9, 10), analysis of these mutations allowed us to define an extensive binding site for FVa in the positive exosite of APC.
Expression and Purification of Recombinant Protein C-Mutant protein C expression vectors were constructed as described (17). Purified recombinant protein C was prepared as described using a fast flow Q-Sepharose column (9,17). Some preparations of recombinant protein C were purified further by chromatography on a calcium-dependent sheep polyclonal anti-protein C antibody-Sepharose column (18) as described (9). Final concentration of protein C was determined using the Asserachrom Protein C enzyme-linked immunosorbent assay (ELISA) assay from American Bioproducts (Parsippany, NJ).
Functional Assays-Protein C was activated by thrombin. Protein C in HBS (50 mM HEPES, 150 mM NaCl) with 2 mM EDTA and 0.5% BSA, pH 7.4, at a concentration of 600 g/ml was incubated for 2.5 h with 12 g/ml thrombin at 37°C followed by the addition of 1.1 units of hirudin per unit of thrombin to inactivate the thrombin. Controls were done in amidolytic assays, clotting assays, and FVa inactivation assays to verify that the thrombin and hirudin used had no effect on subsequent assays. All APC mutants were quantitated using an active site titration adapted from Chase and Shaw (19) using APC at ϳ8 M in HBS and p-nitrophenol-guanidino benzoate at 0.1 mM with an extinction coefficient for p-nitrophenol of 11,400 M Ϫ1 cm Ϫ1 calculated for pH 7.4. K m and k cat for the chromogenic substrate, Pefachrome PCa, were determined by varying substrate concentration from 1.43 to 0.0446 mM in HBS, 0.5% BSA, 5 mM CaCl 2 , pH 7.4 with APC at 5.7 nM. Michaelis constants were derived using Eadie-Hofstee plots. Alternatively, the 5 mM CaCl 2 was replaced with 5 mM EDTA for similar determinations. Color development was measured with an Optimax microplate reader (Molecular Devices, Sunnyvale, CA) (16).
Dilute prothrombin time clotting assays were performed according to the following procedure. Plasma (50 l) was incubated with 50 l of APC in HBS with 0.5% BSA at APC concentrations from 8 to 32 nM (2.7-11 nM final concentration) for 3 min at 37°C. Then clotting was initiated by adding 50 l of Innovin (Dade Behring Inc., Newark, DE) diluted 1:60 in HBS, 0.5% BSA, and 25 mM CaCl 2 . The clotting time was measured using an ST4 coagulometer (Diagnostica Stago, Asnieres, France). APC inactivation by serpins present in plasma was measured essentially according to the protocol of Heeb et al. (20).
Inactivation of FVa was measured according to the following procedure. A mixture of 1 nM FVa with 25 M phospholipid vesicles was made in 50 mM HEPES, pH 7.4, 100 mM NaCl, 0.5% BSA, 5 mM CaCl 2 , and 0.1 mM MnCl 2 . Inactivation was initiated by the addition of APC. Onemicroliter aliquots were removed at time points and added to 40 l containing 1.25 nM factor Xa (FXa) with 31 M phospholipid vesicles, followed by the addition of 10 l of 3 M prothrombin (final concentrations: 1 nM FXa, 20 pM FVa, 25 M phospholipid vesicles, and 0.6 M prothrombin). After 2.5 min a 15-l aliquot of this mixture was quenched by the addition to 55 l of HBS containing 10 mM EDTA, 0.5% BSA, pH 8.2. Chromogenic substrate CBS 34-47 was added, and the rate of thrombin formation was assessed by measuring the change in absorbance at 405 nm.
Curve fitting of the pseudo-first order time courses of FVa inactivation was done according to Nicolaes et al. (21) as described (9). k 506 is the apparent second-order rate constant for cleavage at Arg 506 in native FVa, k 306 is the apparent second-order rate constant for cleavage at Arg 306 in FVa int , and k 306 Ј is the apparent second-order rate constant for cleavage at Arg 306 in intact FVa. k 306 Ј was determined from the inactivation time courses of Gln 506 -FVa fit to a single exponential and was taken as a constant in determination of k 506 and k 306 . Data were fit to the equations using nonlinear least-squares regression analysis.

Production and Characterization of Protein C Mutants-
Wild type and mutant recombinant protein C were recovered at levels varying from 0.1 to 1.5 mg/liter of conditioned medium. After purification, recombinant protein concentrations were determined by ELISA. Silver-stained gels showed that the proteins appeared to be 75-95% pure (data not shown) and also showed no apparent changes in molecular mass (62 kDa) for the mutants with respect to wild type recombinant protein C.
Protein C was activated with thrombin as described under "Materials and Methods," and concentration of APC was determined by active site titration with p-nitrophenol guanidinobenzoate. Following active site titration, standard curves for activity with the amidolytic substrate Pefachrome PCa were performed, and this amidolytic assay was used to determine the activity of samples immediately prior to use in all subsequent assays. K m and k cat for the substrate Pefachrome PCa were derived from titrations of substrate in HBS with 0.5% BSA and 5 mM CaCl 2 , pH 7.4 (Table I). For APC containing mutants in the calcium-binding loop, K m and k cat were also determined in the same buffer except that the CaCl 2 was replaced with 5 mM EDTA. Negligible changes in K m and k cat for this small chromogenic substrate are taken to suggest that mutation did not result in a major structural defect in the protein that would complicate interpretation of any functional results.
Most of the mutant APCs did not significantly differ from wild type in their K m and k cat for Pefachrome PCa. Only four mutants had notable differences. First, the mutation W231A had an increased k cat both in the presence and the absence of CaCl 2 ; however, the K m was unchanged. The APC mutant K191A/K192A/K193A also had an increased k cat . The mutant R222A had a large change in both K m and k cat in the presence and absence of CaCl 2 . Finally, the mutant D227A had significantly altered kinetic parameters for cleavage of Pefachrome PCa, but in this case the difference in K m from wild type was smaller in the presence of CaCl 2 than it was in the presence of EDTA. This suggests that this mutation had some effect on calcium binding. In fact, the backbone carbonyl of Asp 227 is proposed to be a calcium ligand (6), so it is not unexpected that mutation of this residue could affect calcium binding. Alterations in the APC amidolytic activity caused by these two mutations (Arg 222 and Arg 227 ) suggest that the conformation of  the mutant is compromised in some way such that it is not possible to definitively ascribe changes in anticoagulant function to altered APC-FVa interactions. Therefore, we cannot draw firm conclusions about the interactions of these two residues with factor Va.
To further confirm that the mutations with effects on FVa inactivation rates did not have significant changes in APC active site functional activity, several mutants and wild type APC were assayed for their rate of inactivation in plasma, which primarily reflects inactivation of APC by the serpins, ␣-1-antitrypsin (AAT), and protein C inhibitor (22). Wild type APC activity had a half-life (t1 ⁄2 ) for inactivation in plasma of 21.4 min. The APC mutants D214E/A215A (t1 ⁄2 ϭ 24.1), R229A (t1 ⁄2 ϭ 18.8), R230A (t1 ⁄2 ϭ 23.6), R229A/R230A (t1 ⁄2 ϭ 27.5), and K191A/K192A/K193A (t1 ⁄2 ϭ 21.0) had half-lives in plasma that were very similar to that of wild type APC. In contrast, the two APC mutants that appeared to have perturbations of the active site, R222A (t1 ⁄2 ϭ 208 min) and D227A (t1 ⁄2 ϭ 83.0 min), were clearly defective in their inactivation by serpins, further confirming that these proteins have global conformational changes that affect active site interactions with various substrates.
Functional Activity of Mutant APCs-All of the mutant APCs were evaluated for anticoagulant function in a dilute prothrombin time assay as described under "Materials and Methods." This assay is highly sensitive to factor V activity in plasma, but controls were done to verify that it is insensitive to factor VIII activity as well as factor IX activity (data not shown), confirming that this assay reflects APC inactivation of FVa. Fig. 2A shows the anticoagulant activity of selected APC mutants. Mutation of Arg 229 or Arg 230 to Ala significantly decreased the anticoagulant activity with a somewhat greater effect for R229A. The double mutation of R229A/R230A had an even larger decrease in anticoagulant activity.
The anticoagulant activity of each APC mutant was estimated from Fig. 2A and from identical experiments for other APC mutants (data not shown), and anticoagulant activity of each mutant is shown in Fig. 3 as a percentage of wild type. Although, in the calcium-binding loop mutations at residues Arg 229 and Arg 230 markedly decreased APC activity, mutation of residues Trp 231 , Glu 232 , Lys 233 , and Trp 234 did not have any negative effect on anticoagulant activity. Therefore, it would appear that these residues do not have a significant role in FVa recognition by APC. As seen in Fig. 3, specific Lys residues in loop 37, especially Lys 193 , are essential for normal anticoagulant activity, whereas Ser 190 has no role in anticoagulant activity. In loop 60, residues Asp 214 and Glu 215 moderately contribute to anticoagulant activity, while Lys 217 and Lys 218 have no significant role in anticoagulant activity. Mutation of Arg 222 to Ala greatly reduced anticoagulant activity, but, as noted above, this mutation altered amidolytic activity, so it likely caused a global structural perturbation of APC. Although the reduction in anticoagulant activity observed for R222A may be because of loss of a specific binding interaction between APC and FVa, we cannot draw any conclusion because global structural perturbations could have the same effect. In the calciumbinding loop, mutation of Asp 227 also had a moderate effect on anticoagulant activity, but, as was the case for R222A, this mutation altered amidolytic activity and appeared to cause structural perturbations of APC. Furthermore, it is likely that this mutation affected calcium ion binding. Therefore, we cannot draw any conclusions about the role of Asp 227 in FVa inactivation.
To define effects of mutations on specific cleavages in FVa by APC, we studied FVa inactivation in a purified system. Fig. 2B shows results from experiments using normal plasma-derived FVa. Cleavage rate constants for Arg 506 and Arg 306 cleavages by each APC mutant were derived from inactivation time courses according to established equations (9,21). Wild type APC had a value for k 506 (Arg 506 cleavage) of 93 ϫ 10 6 M Ϫ1 sec Ϫ1 . Fig. 4A shows k 506 values for wild type and mutant APCs. Wild type APC had a value for k 306 (Arg 306 cleavage following Arg 506 cleavage) of 3.0 ϫ 10 6 M Ϫ1 sec Ϫ1 . Fig. 4B   also determined for all APC mutants by measuring cleavage of plasma-derived Gln 506 -FVa. The value of k 306 Ј for wild type APC was 1.7 ϫ 10 6 M Ϫ1 sec Ϫ1 . This value was about 60% of the value of k 306 , and all APC mutants gave similar values for k 306 Ј relative to their respective k 306 , values (data not shown). Consistent with previously published data (9,21), this suggests that cleavage at Arg 506 modestly enhances the rate of cleavage at Arg 306 .
From Fig. 2B and Fig. 4A it is clear that both the R229A and the R230A mutations had large effects in decreasing the FVa Arg 506 cleavage rate constants, and the double mutation R229A/R230A had an even larger effect. Fig. 4A also shows that the mutation K193A had a large effect on the Arg 506 cleavage rate constant, whereas the mutations K191A and K192A had only moderate effects. In contrast, the double mutation D214A/E215A increased the rate constant for Arg 306 cleavage but not for Arg 506 cleavage. With that exception, the effects of most of these mutations were largest for cleavage at Arg 506 (k 506 ) and apparently much less for cleavage at Arg 306 (k 306 , Fig. 4). This selective effect of these mutations in APC on the Arg 506 cleavage is illustrated in Fig. 4C, where the ratio of the Arg 506 cleavage rate constant (k 506 ) to the Arg 306 cleavage rate constant (k 306 ) is shown for wild type and mutant APCs. Wild type APC cleaves Arg 506 (k 506 ) 31ϫ faster than it cleaves Arg 306 following Arg 506 cleavage (k 306 ). APC mutants that had reduced activity mainly lost activity for Arg 506 cleavage. This is reflected in the reduced ratio (k 506 /k 306 , Fig. 4C). This reduced ratio is clear for the mutations K193A, R229A, and R230A. The triple mutation, K191A/K192A/K193A, had an even greater reduction in ratio in an apparent additive effect of the three individual mutations. The double mutation R229A/R230A also manifested an additive effect of the two individual mutations. This graph indicates that the double mutation D214A/E215A had a significant effect on the ratio of k 506 to k 306 . But, as noted above, this is apparently because of an increase in the Arg 306 cleavage rate rather than a decrease in the Arg 506 cleavage rate.

DISCUSSION
Mutational analysis using conservative alanine mutations to probe protein-protein interactions is a well established method for characterizing the molecular details of the interaction (9,24). However, it cannot be conclusively ruled out that mutations alter folding properties of the protein or exert their effects at a distance from the interaction site rather than alter a specific protein-protein interaction. To minimize this possibility, we analyzed the kinetic parameters of mutant APCs for cleavage of a small chromogenic substrate (Table I), and for selected APC mutants we also measured the rate of inactivation by plasma serpins. In both cases, kinetic parameters equivalent to those of wild type APC suggest that no gross structural changes were caused by the mutations. As exceptions, significant changes in these parameters for two mutants (R222A and D227A) caused us to remove them from consideration in our structure-function analysis. In support of this, the x-ray crystallographic structure of APC (7) illustrates that all of these mutated residues are fully surface exposed, except for Arg 222 and Asp 227 , which are only partially exposed, so mutation of them might be expected to have structural consequences. Therefore, for the remaining mutants, it is reasonable to assume that functional changes caused by mutations are the result of alteration of specific protein-protein interactions.
Accordingly, our results provide insights into APC exosite interactions with FVa. Residues Arg 229 , Arg 230 , and Lys 193 play the largest role in FVa interactions, whereas residues Lys 191 , Lys 192 , Asp 214 , and Glu 215 appear to play minor roles. The rate of cleavage at Arg 506 in FVa is significantly affected by mutation of each of these residues, whereas cleavage at Arg 306 is only mildly affected. Conversely, residues Lys 217 , Lys 218 , FIG. 4. Effect of APC mutations on rate constants for cleavages at Arg 506 and Arg 306 in FVa. A, cleavage rate constants for Arg 506 . B, cleavage rate constants for Arg 306 . Cleavage rates are apparent second-order rate constants in units of M Ϫ1 sec Ϫ1 , which were determined as described under "Materials and Methods." C, ratio of the cleavage rate constant for Arg 506 to that for Arg 306 (k 506 /k 306 ). The ratio for wild type APC is presented at the top of the bar graph and the dashed vertical line indicates the ratio for wild type APC (ratio ϭ 31).
Trp 231 , Glu 232 , Lys 233 , and Trp 234 do not appear to play any significant role in FVa interactions with APC.
Friedrich et al. (11) mutated Arg 229 (CHT 74) to Gln, a naturally occurring type II protein C mutation, and concluded that this residue played, at most, a minor role in the interaction of APC with the Arg 506 cleavage site of FVa. This is in contrast to the significant reduction in APC activity we observed here for the R229A mutant. But the mutation was not the same, so direct comparison is not possible. In fact, both naturally occurring mutations of Arg 229 (R229Q and R229W) are still potential hydrogen bond donors. In contrast, the mutation R229A removes all hydrogen bonding possibility at residue 229. Therefore, it is possible that these mutations (R229Q and R229W) maintain part of the functional contact provided by Arg 229 , which could also involve hydrogen bonding. Thus, mutation to these naturally occurring mutants, while reflecting what happens in individuals with these mutations, would not necessarily elucidate the normal functional role of the wild type Arg 229 in FVa interactions. We conclude that Arg 229 does indeed play a significant role in interactions of APC with FVa. Furthermore, Arg 230 clearly plays a significant role in interactions of APC with FVa. Thus, the calcium-binding loop is central in the interaction of APC with FVa.
Friedrich et al. (11) also concluded that residues Lys 191 , Lys 192 , and Lys 193 (CHT 37-39) were involved in cleavage of Arg 506 in FVa by APC, whereas residues Lys 217 and Lys 218 (CHT 62 and 63) were not involved in this interaction. This agrees well with our results, but we mutated residues 191, 192, and 193 individually as well as together, whereas Friedrich et al. (11) only mutated them all at once. From these individual mutations we were able to determine that the three Lys residues did not contribute equally to this interaction. Rather, Lys 193 played the largest role in the interaction of APC with FVa.
In a previous study (9), we identified several residues in the autolysis loop of APC (residues 301-316, equivalent to chymotrypsin 142-153) that also play a very significant role in FVa interactions and further distinguish between cleavages at Arg 506 and Arg 306 . This surface loop is also part of the extended positive exosite on APC, and these residues are nearly contiguous in tertiary space with the residues identified in the current study as FVa binding residues. Another residue, Arg 352 , identified as a site for FVa interaction, is also contiguous with the positive exosite (10). Together these studies map out an extended FVa binding site on APC that arches around the active site triad of the serine protease but does not indiscriminately involve the entire so-called positive exosite, as various residues that define the positive exosite, such as Lys 217 , Lys 218 , and Lys 233 , as well as other non-positive residues that fall within this area, such as Ser 190 , Trp 231 , Glu 232 , Trp 234 , and Glu 309 , do not appear to play significant direct roles in FVa interactions.
The correspondence of anticoagulant plasma-based assay data (dilute prothrombin time assay) with FVa inactivation rate constant data was generally excellent, with one significant exception. The mutation D214A/E215A had a k 506 value that was similar to the wild type value and a k 306 value that was almost double the wild type value, whereas its anticoagulant activity was about 50% of wild type. We ruled out the possibility that this mutation plays a role in other APC interactions that are present only in the plasma-based assay, such as with protein S, factor VIII, or serpin inhibitors. The dilute prothrombin time anticoagulant assay was done in FVIII-deficient plasma and protein S-deficient plasma (data not shown), and the same relative loss of activity was seen in both assays, confirming that the defect of D214A/E215A in plasma was not due to interactions with FVIII or protein S. Also, as mentioned, the rates of inactivation of wild type APC and D214A/E215A APC that were measured in plasma were virtually identical. It is possible that the proximity of these residues to the catalytic triad, both in sequence and in tertiary structure, plays a role in the difference between clotting activity and FVa inactivation activity. We suggest that residues Asp 214 and Glu 215 may have a minor role in FVa interactions.
It is intriguing that most of the mutants presented in this work and others (9,11) that were defective in FVa inactivation primarily affected Arg 506 cleavage rather than Arg 306 cleavage in FVa. Pellequer et al. (25) modeled APC bound to the Arg 506 cleavage site, and the model showed extensive contacts between the positive exosite of APC and FVa with Arg 506 bound in the S1 site of APC. This APC/FVa model allowed for contacts with certain basic residues of the APC autolysis loop that were suggested by mutation to be involved in FVa binding (9) and it also allowed that Arg 229 and Arg 230 could interact with FVa. In addition loop 37 (Lys 191 , Lys 192 , and Lys 193 ) was modeled to be in contact with FVa. The data in this paper are consistent with Pellequer's APC/FVa complex model because we show that Lys 193 , Arg 229 , and Arg 230 are all critical for interactions with FVa during cleavage of Arg 506 .
In summary, including our previous work (9), we have now identified a total of seven residues in the positive exosite of APC that appear to play a major role in FVa binding (Lys 193 , Arg 229 , Arg 230 , Arg 306 , Lys 311 , Arg 312 , and Arg 314 ) and at least four that appear to play a minor but significant role (Lys 217 , Lys 218 , Lys 307 , and Glu 308 ), with two more at the periphery of the positive exosite that may play a minor role (Asp 214 and Glu 215 ). These results serve to outline an extensive FVa binding exosite on the surface of the protease domain of APC. In future work it will be interesting to see how this FVa binding exosite compares to exosites for other APC/PC molecular partners such as thrombomodulin and FVIIIa.