Identification and Characterization of the Sodium-binding Site of Activated Protein C*

Activated protein C (APC) requires both Ca2+ and Na+ for its optimal catalytic function. In contrast to the Ca2+-binding sites, the Na+-binding site(s) of APC has not been identified. Based on a recent study with thrombin, the 221–225 loop is predicted to be a potential Na+-binding site in APC. The sequence of this loop is not conserved in trypsin. We engineered a Gla domainless form of protein C (GDPC) in which the 221–225 loop was replaced with the corresponding loop of trypsin. We found that activated GDPC (aGDPC) required Na+ (or other alkali cations) for its amidolytic activity with dissociation constant (K d (app)) = 44.1 ± 8.6 mm. In the presence of Ca2+, however, the requirement for Na+ by aGDPC was eliminated, and Na+ stimulated the cleavage rate 5–6-fold withK d (app) = 2.3 ± 0.3 mm. Both cations were required for efficient factor Va inactivation by aGDPC. In the presence of Ca2+, the catalytic function of the mutant was independent of Na+. Unlike aGDPC, the mutant did not discriminate among monovalent cations. We conclude that the 221–225 loop is a Na+-binding site in APC and that an allosteric link between the Na+ and Ca2+ binding loops modulates the structure and function of this anticoagulant enzyme.

Protein C is a multi domain vitamin K-dependent plasma serine protease zymogen, which upon activation by the thrombin⅐TM 1 complex, down-regulates the blood coagulation cascade by selectively inactivating factors Va and VIIIa (1)(2)(3). Monovalent and divalent cations regulate the protein C anticoagulant pathway at the level of both the protein C zymogen activation and the catalytic function of activated protein C (APC). In the zymogen activation process, binding of Ca 2ϩ to the N-terminal Gla domain and the first epidermal growth factor like domain-1 is required for efficient assembly of the protein into the activation complexes on cell surfaces (4,5). In addition to the Gla domain, Ca 2ϩ binding to a high affinity site in the catalytic domain induces a conformational change in the activation peptide of protein C that is required for rapid activation by the thrombin⅐TM complex (6). Following the activation, the non-catalytic domains of APC remain covalently linked to the serine protease domain to mediate the Ca 2ϩ -and Gla-dependent binding of APC to its specific cofactors, protein S or the endothelial cell protein C receptor (7)(8)(9). Although several previous studies have demonstrated that Ca 2ϩ and other divalent cations stimulate the amidolytic and esterolytic activity of APC (10,11), it is not known if occupancy of the Ca 2ϩ -binding site in the catalytic domain of APC is also required for its physiological function.
Similar to Ca 2ϩ , the monovalent cation Na ϩ regulates the protein C anticoagulant pathway at the level of both zymogen activation and APC function. In this case, it is believed that Na ϩ allosterically regulates the activity of thrombin and that protein C is activated preferentially by the Na ϩ -free form of thrombin, which is referred to as the slow form (12,13). The Na ϩ -bound or fast form of thrombin is believed to function in the procoagulant pathway by specifically activating fibrinogen (12,13). A dissociation constant of 30 -100 mM for Na ϩ binding to thrombin has been reported (12,14). In contrast to protein C activation, however, it has been demonstrated that bovine APC displays a strict requirement for Na ϩ (or other monovalent cations) for expression of its amidolytic activity toward tripeptide chromogenic substrates (15). In previous studies an apparent dissociation constant of 87-129 mM for Na ϩ binding to bovine APC and activated Gla domainless protein C (aG-DPC) has been reported (15)(16)(17). The exact role of Na ϩ in the activity of APC toward its physiological substrates, factors Va and VIIIa, is not known.
In contrast to the divalent cation-binding sites of APC, the monovalent cation-binding site(s) of APC has not been identified. In the case of thrombin, a single metal-binding site for Na ϩ has been localized to a conserved loop consisting of residues 221-225 in the chymotrypsin numbering system (18,19). The predicted consensus Na ϩ -binding sequence in this loop is conserved in APC but not in trypsin (19). To determine whether this loop in APC binds Na ϩ , we prepared a mutant lacking the Gla domain (GDPC) in which the 221-225 loop of GDPC (Gly 221 -Tyr 225 ) was replaced with the corresponding residues of trypsin (Ala 221 -Pro 225 ). Both the wild type and GDPC mutant (GDPC tryp/loop) were expressed in mammalian cells, and sufficient quantities of both derivatives were isolated for activation and characterization. In addition to this mutant, we recently prepared and characterized another GDPC mutant in which Glu 80 in the Ca 2ϩ binding loop of the molecule was replaced with Lys (E80K) (6). In this previous study, we demonstrated that the 70 -80 loop in this mutant was stabilized in the Ca 2ϩ conformer possibly by Lys 80 forming a salt bridge with Glu 70 (6). In the current study, the catalytic activities of the activated wild type, tryp/loop, and E80K GDPC (aGDPC) derivatives were monitored by their ability to hydrolyze several chromogenic substrates in both the absence and presence of Na ϩ and Ca 2ϩ . We found that unlike the wild type protease, the aGDPC tryp/loop mutant lost its ability to bind Na ϩ and no longer discriminated between various monovalent cations. The affinity of the mutant protease for Ca 2ϩ was also impaired ϳ16-fold. Interestingly, further study suggested that the affinity of the wild type aGDPC for Na ϩ in the presence Ca 2ϩ was improved ϳ20-fold, and the binding of aGDPC E80K to Na ϩ was of the high affinity type independent of Ca 2ϩ . These results suggest that the 221-225 loop is a Na ϩ -binding site in APC, which is allosterically linked to the divalent cation-binding loop of the protease. The allosteric coupling of the two metal ionbinding loops in APC has likely played a role in the divergent evolution of this highly specific anticoagulant enzyme.

EXPERIMENTAL PROCEDURES
Mutagenesis, Expression, and Purification-The expression of Gla domainless protein C (GDPC) and GDPC E80K by the RSV-PL4 expression/purification vector system in human 293 cells has been previously described (6,20). The 221-225 loop (Gly 221 -Tyr 225 ) of GDPC was replaced with the corresponding sequence of trypsin (GDPC tryp/loop) by the polymerase chain reaction mutagenesis approach, and the mutant molecule was expressed in the same vector system as described previously (20,21). Accuracy of the mutations was confirmed by sequencing prior to expression. The wild type and mutant zymogens were purified from the cell culture supernatants as described previously (20).
Activation of GDPC and GDPC Tryp/Loop by the Thrombin⅐TM4 -6 Complex-The initial rate of protein C activation by the thrombin⅐TM4 -6 complex (1 nM thrombin in complex with 100 nM TM) was measured as a function of different concentrations of GDPC or GDPC tryp/loop in 0.1 M NaCl, 0.02 M Tris-HCl, pH 7.5 (TBS), containing 0.1% polyethylene glycol 8000 (PEG 8000), and 2.5 mM Ca 2ϩ at room temperature. After inhibition of thrombin activity by antithrombin, the initial rates of activation were measured from the rate of activated protein C generation in an amidolytic activity assay using 1 mM SpPCa in TBS buffer containing 0.1% PEG 8000 and 2.5 mM Ca 2ϩ . The rate of hydrolysis was measured at 405 nm at room temperature in a V max kinetic plate reader (Molecular Devices, Menlo Park, CA). Fixed reaction times of 5 min were employed for each GDPC concentration. Over this time only the initial rate of activation was measured and less than 1% substrate was activated. The concentration of active protein C derivatives in reaction mixtures was determined by reference to a standard curve, which was prepared by total activation of GDPC at the time of each experiment. This was accomplished by activation of 1 M each protein C derivative with 10 nM thrombin in complex with 100 nM TM4 -6 and 2.5 mM Ca 2ϩ for 90 min at 37°C. Under these experimental conditions, all protein C zymogens were completely activated in less than 30 min. The K m and k cat values of protein C activation were calculated from the Michaelis-Menten equation. Protein C activation was also carried out on thrombin⅐TM complex assembled on Affi-Gel 10 (Bio-Rad) as described previously (22). The active site concentrations of the aGDPC derivatives were determined by an active site-specific immunoassay using BioCap-FPR-ck (biotinyl-⑀-aminocaproyl-D-phenylalanine prolylarginine chloromethyl ketone) (Hematologic Technologies Inc., Essex Junction, VT) as described previously (27). The aGDPC derivatives were desalted by passing through PD-10 (Amersham Pharmacia Biotech) gel filtration columns equilibrated with Chelex-treated (Bio-Rad) 5 mM Tris-HCl, pH 8.0, buffer containing 0.1% PEG 8000 and stored at Ϫ80°C.
Steady-state Kinetics-The steady-state kinetic parameters of wild type and mutant enzyme toward S2266 (D-Val-Leu-Arg-p-nitroanilide) was performed at room temperature in 5 mM Tris-HCl, pH 8.0, contain-ing 0.1% PEG 8000, 2.5 mM CaCl 2 , and 0.2 M of different monovalent chloride salts (LiCl, NaCl, KCl, and choline chloride). The rate of hydrolysis was measured at 405 nm at room temperature in a V max kinetic plate reader as described above. The apparent K m and k cat values for substrate hydrolysis were calculated from the Michaelis-Menten equation. The specificity constant for each chloride salt was expressed as the ratio of k cat /K m . The specificity constant for the bulky monovalent cation choline (Ch ϩ ) was used as a reference to determine the monovalent cation specificity as described by Dang and Di Cera (19). The concentration of S2266 ranged from 20 M to 15 mM depending on the K m values, and the concentration of enzymes ranged from 10 to 100 nM depending on the k cat values.
Dissociation Constant (K d(app) ) for Na ϩ -The values for K d(app) of Na ϩ binding to each protease was determined from the effect of varying concentrations of Na ϩ on the activity of the protease toward three synthetic substrates SpPCa, S2266, and S2238 in both the absence (Chelex-treated buffer) and presence of 2.5 mM Ca 2ϩ . In all experiments a constant ionic strength of 0.2 M was maintained by addition of choline chloride. This procedure has been commonly used in the past to study the effect of monovalent cations on the catalytic function of various enzymes (12,19). The values for K d(app) were calculated from the hyperbolic increase in the rate of substrate hydrolysis as a function of increasing Na ϩ concentrations.
Factor Va Inactivation-The time course of human factor Va inactivation by activated GDPC was measured by a two-stage assay. In the first stage, factor Va (50 nM) was incubated with aGDPC (5 nM) at room temperature in 0.02 M Tris-HCl, pH 7.5, containing 0.15 M NaCl or 0.15 M KCl and 0.1% PEG 8000. This stage of the assay was carried out in both the absence and presence of 2.5 mM Ca 2ϩ . For this purpose, aGDPC and factor Va (in 5 mM Ca 2ϩ ) were passed through two separate PD-10 gel filtration columns equilibrated with Chelex-treated 0.02 M Tris-HCl, 0.1% PEG 8000 and used immediately in kinetic experiments. In the second stage, at different time intervals (0 -12.5 min) the remaining activity of factor Va was determined by measuring its ability to accelerate factor Xa activation of recombinant human prethrombin-1 as described previously (28).
Data Analysis-The apparent K m and k cat values for substrate hydrolysis were calculated from the Michaelis-Menten equation, and the affinity of Na ϩ for each aGDPC derivative (K d(app) ) was determined by nonlinear regression fits of data to a rectangular hyperbola using EN-ZFITTER (R. J. Leatherbarrow, Elsevier, Biosoft). All values are the average of at least 3-5 independent measurements ϮS.D.

Expression and Purification of Recombinant
Proteins-Recombinant wild type and mutant GDPC derivatives were expressed in 293 cells and isolated as described under "Experimental Procedures". SDS-PAGE analysis ( Fig. 1) indicated that both derivatives expressed as two subforms with identical apparent molecular weights that correspond to ␣ and ␤ protein C that are glycosylation variants observed previously with this protein (20). Under reducing conditions, a light chain was also observed with both derivatives. With both proteins a fraction of the protein samples remained non-reducible suggesting that the GDPC derivatives were expressed as a mixture of single and two-chain proteins. This property has been also observed for both recombinant and plasma-derived human protein C in the past (29,30). We have previously demonstrated that both single and double-chain APC derivatives have identical catalytic activities (31). These results suggest that the mutation does not alter the post-translational modifications or the processing of the protein. With the mutant GDPC, however, a minor band migrating at ϳ35 kDa was also observed (Fig. 1). The nature of this band was not characterized.
Comparison of the initial rate of protein C activation by the thrombin⅐TM4 -6 complex as a function of different zymogen concentrations suggested that both GDPC derivatives were activated at a similar rate (Fig. 2). SDS-PAGE analysis of the activated products indicated that both zymogens were completely converted to activated forms (data not shown). This was consistent with the observation that the concentrations of enzymes as determined by the active site titration were similar (within 80%) with the values calculated based on the absorbance at 280 nm.
To determine whether the mutant aGDPC can bind Na ϩ , the effect of increasing concentration of Na ϩ on the activity of aGDPC derivatives toward the chromogenic substrates S2266, S2238, and SpPCa was studied. Since the catalytic domain of APC contains a Ca 2ϩ -binding site (6), these studies were carried out both in the absence and presence of Ca 2ϩ . In the absence of Ca 2ϩ , the amidolytic activity of aGDPC was strongly dependent on the presence of Na ϩ in the reaction buffer (Fig.  3A). The amidolytic activity of aGDPC was enhanced with increasing concentrations of Na ϩ and reached saturation with a K d(app) of 44.1 Ϯ 8.6 mM. In a previous study, the interaction of Na ϩ with bovine aGDPC was shown to be cooperative with a Hill coefficient of 1.5 (17). A similar cooperativity for Na ϩ binding to human aGDPC was also observed in this study in the absence of Ca 2ϩ (data not shown). However, when the ionic strength of medium was adjusted to 0.2 M with choline chloride, no significant cooperativity was observed, and the nonlinear regression fits of data to a rectangular hyperbola was found to be suitable for obtaining the K d(app) values (Fig. 3A). In the presence of Ca 2ϩ , no cooperativity for Na ϩ binding to aGDPC was observed irrespective of whether the ionic strength of medium was adjusted to 0.2 M with choline chloride. In this case an ϳ5-6-fold stimulation of the amidolytic activity of aGDPC was observed at saturating Na ϩ concentrations (Fig.  3A). Interestingly, the affinity of Na ϩ for the protease was also improved ϳ20-fold in the presence of Ca 2ϩ (K d(app) ϭ 2.3 Ϯ 0.3 mM), and at higher concentrations of Na ϩ (Ͼ60 mM) the activity of aGDPC was slightly diminished. A similar improvement in the affinity of Na ϩ for the protease in the presence of Ca 2ϩ was observed at physiological temperature, although the K d(app) values were slightly elevated (141.0 Ϯ 20.9 and 5.3 Ϯ 0.3 mM in the absence and presence of 2.5 mM Ca 2ϩ , respectively). To ensure that this effect of Ca 2ϩ on the Na ϩ binding properties of APC was not a phenomenon related to the Gla domainless form of APC or an effect related to a particular substrate, the amidolytic activity of plasma-derived APC was monitored as a function of Na ϩ in both the absence and presence of Ca 2ϩ with three different chromogenic substrates (S2266, S2238, and SpPCa). In all cases, similar results were obtained (data not shown). Only the results with SpPCa hydrolysis for aGDPC at room temperature are presented in these figures.
It is worth noting that in previous studies, the amidolytic activity of bovine APC displayed a strict requirement for Na ϩ (16). In the current study, we noticed some base-line amidolytic activity for human aGDPC in the absence of Na ϩ and Ca 2ϩ (Fig. 3A). However, we believe that our results are consistent with the literature since a similar base-line activity was also observed in previous studies (15,16). It was previously suggested that the base-line amidolytic activity of APC in the absence of Na ϩ and Ca 2ϩ may be due to the presence of other monovalent cations such as Tris ϩ and/or choline (Ch ϩ ) in the reaction buffer (15,16).
The Na ϩ concentration dependence of the amidolytic activity of aGDPC tryp/loop mutant was studied in a similar fashion. In the absence of Ca 2ϩ , no K d(app) could be estimated for the Na ϩ interaction with the mutant since no saturation of Na ϩ binding to the mutant was observed up to 400 mM (data are presented for up to 200 mM NaCl in Fig. 3B). No attempt was made to increase the concentration of Na ϩ above 400 mM since the effect of high ionic strength on the structure of the enzyme is not known. However, in the presence of Ca 2ϩ , the amidolytic activity of the mutant was insensitive to the absence or presence of Na ϩ (Fig. 3B). These results clearly suggest that the mutant has lost its ability to bind Na ϩ and further suggest that Ca 2ϩ binding to aGDPC has a profound effect on the ability of the protease to interact with Na ϩ .
There are two known Ca 2ϩ -binding sites on aGDPC that can influence the Na ϩ binding properties of the protease. The first Ca 2ϩ -binding site resides in the epidermal growth factor like domain-1 of the light chain and the other was localized to the C-terminal catalytic domain of APC (6,32). The Ca 2ϩ -binding site in the catalytic domain is located on a loop between residues Glu 70 and Glu 80 , analogous to the Ca 2ϩ binding loop in trypsin (33). Previously, we prepared and characterized a GDPC derivative in which Glu 80 was replaced with Lys (E80K) (6). In that study, we demonstrated that the catalytic domain of the mutant was stabilized in the Ca 2ϩ conformer possibly as a result of Lys 80 in the mutant forming a salt bridge with Glu 70 (6). To characterize further the Na ϩ -binding site of APC and determine the Ca 2ϩ -binding site responsible for altering the Na ϩ binding properties of the protease, the amidolytic activity of aGDPC E80K toward SpPCa was studied as a function of different concentrations of Na ϩ . Interestingly, the K d(app) values for Na ϩ binding to this mutant was of high affinity and insensitive to the absence or presence of Ca 2ϩ (5.1 Ϯ 0.7 mM in the absence of Ca 2ϩ and 5.0 Ϯ 0.6 mM in the presence of Ca 2ϩ ) (Fig. 4). The K d(app) of Na ϩ for the mutant did not change even if the amidolytic activity of the mutant was monitored in the presence of 0.1 mM EDTA or EGTA to chelate divalent metal ions.
It is known that there is an ϳ20% Ca 2ϩ stimulation of the amidolytic activity of aGDPC toward chromogenic substrates with a K d(app) of ϳ50 M (6). To determine whether mutagenesis of the 221-225 loop influences the affinity of the 70 -80 loop for binding to Ca 2ϩ , the amidolytic activity of aGDPC tryp/loop was monitored as a function of increasing Ca 2ϩ concentrations. Two interesting observations emerged. First, unlike the small effect of Ca 2ϩ on the amidolytic activity of aG-DPC, Ca 2ϩ stimulated the amidolytic activity of the mutant toward the chromogenic substrates 8 -10-fold (data not shown). Second, the K d(app) for Ca 2ϩ binding to the aGDPC tryp/loop mutant increased to ϳ800 M, representing about 16-fold weaker interaction than that observed for Ca 2ϩ binding to aGDPC. Taken together, these results suggest that occupancy of one metal-binding site of aGDPC with its specific metal cation influences the affinity of the other site for its specific metal ligand. We conclude that the 70 -80 and 221-225 loops of aGDPC are allosterically linked.
It is known that other monovalent alkali cations can substitute for Na ϩ in stimulation of the amidolytic activity of bovine APC and that the activity increases in parallel with increasing cation radius (16,17). To determine whether the mutant can discriminate among various alkali cations, the kinetic constants were determined for hydrolysis of S2266 and SpPCa by the wild type and mutant aGDPC. Although similar results were obtained with both chromogenic substrates, the K m value of the mutant for the substrates in the presence of all cations was elevated to 1.5-3 mM. Due to poor solubility of SpPCa at concentrations above ϳ3 mM, kinetic analysis was only carried out for S2266. As shown in Fig. 5, and consistent with previous findings (16), the order of monovalent cation preference for the optimal catalytic function of aGDPC was K ϩ (ionic radius ϭ 1.52 Å) Ͼ Na ϩ (ionic radius ϭ 1.16 Å) Ͼ Li ϩ (ionic radius ϭ 0.90 Å) (34). In these studies, the specificity constant (k cat /K m ) for the bulky monovalent cation choline (Ch ϩ , ionic radius ϭ 3.5 Å) was set as a reference, and data were analyzed as described by Dang and Di Cera (19). As shown in Fig. 5, unlike aGDPC, the mutant has lost its ability to discriminate among Li ϩ , Na ϩ , and Ch ϩ . However, similar to aGDPC, the mutant displayed ϳ2-3-fold improved specificity constant with K ϩ . These results may suggest that the binding site for K ϩ on APC is distinct from that of Na ϩ . However, this notion was not supported by the observation that Na ϩ competitively inhibited the binding of K ϩ to aGDPC (data not shown). Furthermore, similar to the case with Na ϩ , the K d(app) for K ϩ binding to aGDPC was lower affinity (74.3 Ϯ 2.1 mM) in the absence of Ca 2ϩ and higher affinity (5.4 Ϯ 1.1 mM) in the presence of 2.5 mM Ca 2ϩ . These results strongly suggest that both Na ϩ and K ϩ bind to the same site on the protease and that Ca 2ϩ binding to the 70 -80 loop allosterically modulates the structure of the monovalent cation binding loop of APC.
Next, we compared the effect of Na ϩ and K ϩ on the activity of aGDPC toward its natural substrate factor Va in both the absence and presence of Ca 2ϩ . The assay system required to study this question is not straightforward because factor Va is made up of two subunits, held together by Ca 2ϩ (35,36), and both subunits of factor Va are required for its function. However, it has been reported that the affinity of factor Va for Ca 2ϩ is high and that subunit dissociation requires EDTA or denaturation of the cofactor (36,37). Therefore, we desalted human factor Va (in 5 mM Ca 2ϩ ) by passing it through a PD-10 gel filtration column that was equilibrated with Chelex-treated 20 mM Tris-HCl, pH 7.5, buffer containing 0.1% PEG 8000. Desalted factor Va was used immediately in the kinetic experiments before subunit dissociation and concomitant factor Va inactivation could occur. Previously, we employed this strategy to study the Ca 2ϩ dependence of prothrombin activation by factor Xa in the presence of factor Va (38). Again, the aGDPC inactivation rate of factor Va in the presence of K ϩ was ϳ2-fold more efficient than Na ϩ , and Ca 2ϩ was required for the optimal activity of the protease with both monovalent cations (Fig. 6). The catalytic activity of the mutant toward Va inactivation in the presence of Ca 2ϩ ion was impaired more than 30-fold and again the mutant did not discriminate between either Na ϩ or Ch ϩ ions (Fig. 6). DISCUSSION The effect of Na ϩ and other monovalent cations on the catalytic activity of bovine APC has been studied in the past. It has been demonstrated that the binding of Na ϩ or another alkali metal to bovine APC is required for expression of the amidolytic activity of the protease toward tri-peptide chromogenic substrates (15,17). In the current study, we have identified the 221-225 loop as the monovalent cation-binding site in human APC. Our strategy for the mutagenesis was based on 1) the observation that trypsin does not require Na ϩ for its catalytic activity, and 2) the report that the 221-225 loop is a Na ϩ -binding site in thrombin (18,19). Based on studies with thrombin, Dang and Di Cera (19) identified residue 225 as a key residue that may determine whether or not a serine protease can bind Na ϩ . Serine proteases with a Pro at this position, like trypsin, are unable to bind Na ϩ , but plasma serine proteases with a Tyr at this position, like APC and thrombin, are expected to bind Na ϩ . We initially generated a GDPC mutant in which Tyr 225 was replaced with Pro. However, we could not isolate sufficient quantity of this mutant for complete characterization. This was due to lack of expression of this mutant as an intact molecule. SDS-PAGE analysis of the Tyr 225 3 Pro mutant indicated that the majority of the expressed protein is proteolytically cleaved at two or more sites (data not shown). As an alternative, we replaced all 5 variant residues of the loop (Gly 221 -Pro 225 ) with the corresponding residues of trypsin. We isolated sufficient quantity of this mu-tant to homogeneity and demonstrated that, similar to thrombin, this loop is a Na ϩ -binding site in APC. We should also add that our qualitative results suggest that the Try 225 3 Pro mutant also lost its ability to bind Na ϩ , in agreement with the proposal that Pro 225 does not support Na ϩ binding to the 221-225 loop of serine proteases.
In the previous Na ϩ binding studies with APC, a K d(app) of 87-129 mM for Na ϩ binding to bovine APC or aGDPC was reported (15)(16)(17). In these studies, positive cooperativity (Hill coefficient ϭ 1.5) for Na ϩ binding to bovine aGDPC was observed (17). In the current study, a similar result for the recombinant human aGDPC was observed only if choline chloride was not included in the reaction buffer to compensate for the ionic strength during Na ϩ titrations. However, minimal cooperativity for Na ϩ binding to aGDPC was observed when the total ionic strength in all measurements was adjusted to 0.2 M with choline chloride. Under these conditions, the data fit well to a hyperbolic binding equation, and a K d(app) of ϳ44 mM for Na ϩ binding to aGDPC was obtained. In addition to monovalent cations, previous studies by Hill and Castellino (11, 39 -41) has established that divalent cations also stimulate the amidolytic activity of APC and aGDPC and that separate binding sites for these cations exist on each enzyme. Interestingly, when we included Ca 2ϩ in the reaction buffer, the K d(app) for Na ϩ binding to aGDPC was improved ϳ20-fold suggesting that the Ca 2ϩ -binding site of APC allosterically modulates the Na ϩ binding loop of the molecule. Further support for an allosteric link between the two metal-binding sites was provided by the observation that the affinity of the aGDPC tryp/loop mutant for Ca 2ϩ decreased ϳ16-fold. These results indicate that the conformation of the Na ϩ -and Ca 2ϩ -binding sites are interdependent and that the two metal ions allosterically regulate the structure and function of this anticoagulant enzyme.
Previously, Hill and Castellino (11,39,40) reported that APC contains a single divalent cation-binding site outside of the Gla domain that is critical for the optimal expression of its amido- lytic activity. Recently, we demonstrated that this divalent cation-binding site is located in the catalytic domain of APC in a loop between the residues Glu 70 -Glu 80 , which is also known as the Ca 2ϩ binding loop in trypsin (6,33). The binding of Ca 2ϩ to this site is essential for rapid activation of protein C by the thrombin⅐TM complex (6). In thrombin, which does not bind Ca 2ϩ , Glu 70 is not conserved, but it is replaced with Lys. In the crystal structure of thrombin, an internal salt bridge between Lys 70 and Glu 80 stabilizes this loop (42). Previously, we prepared and characterized a GDPC mutant in which Glu 80 was replaced with Lys (E80K) so that the loop was stabilized in the Ca 2ϩ conformer, possibly by Lys 80 in the mutant forming a salt bridge with Glu 70 (6). Interestingly, the affinity of this mutant for Na ϩ was of high affinity (5 mM) and was not sensitive to the absence or presence of Ca 2ϩ in the reaction buffer. Results obtained with this mutant support the hypothesis that the conformation of the monovalent and divalent cation-binding sites are allosterically linked. Furthermore, no cooperativity for Na ϩ binding to this mutant was observed in either absence (10 M EDTA) or presence of 2.5 mM Ca 2ϩ . It is possible that the cooperativity observed for the Na ϩ binding to aGDPC in the absence of Ca 2ϩ is due to binding of Na ϩ to the 70 -80 loop and partially substituting for Ca 2ϩ . In support of this hypothesis, cooperativity is observed only for Na ϩ (ionic radius 1.16 Å) which has similar ionic radius as Ca 2ϩ (1.14 Å), but not with any other alkali cation.
Several previous studies by Di Cera and co-workers (12,18) suggest that Na ϩ binding to the 221-225 loop of thrombin modulates the activity and specificity of this enzyme in an allosteric fashion. These authors believe that the Na ϩ -bound form (fast form) has improved catalytic activity toward chromogenic substrates and specifically clots fibrinogen, whereas the Na ϩ -free form (slow form) specifically activates protein C (12). It is interesting to note that the catalytic activity of the loop mutant of aGDPC was impaired but that the activation of the zymogen form of the mutant by the thrombin⅐TM complex was not impaired and rather improved slightly. These results suggest that Na ϩ binding to either thrombin or protein C is not required for recognition and rapid activation by the thrombin⅐TM complex. It is not known if APC, similar to thrombin, can exist in two distinct slow and fast forms in plasma. The concentration of Na ϩ in blood is very close to the K d(app) of this cation for thrombin, so that the two forms of thrombin may exist in equilibrium under physiological conditions (12,14). In the case of APC, however, the K d(app) for Na ϩ is ϳ25-30-fold lower than the concentration of this cation in blood, which suggests that APC may predominantly exist in the fast form under physiological conditions.
The other interesting observation of this study is that the catalytic activity of aGDPC in the presence of K ϩ was 2-3-fold more efficient than Na ϩ . This was true whether the activity was monitored by hydrolysis of chromogenic substrates or by the inactivation of the natural substrate factor Va. Both thrombin and factor Xa, unlike APC, function more efficiently in the presence of Na ϩ . The molecular basis for the improved catalytic property of APC in the presence of K ϩ was not understood. Since K ϩ is primarily an interacellular metal ion, it is not known if it can play a role in up-regulation of the catalytic activity of APC in plasma. However, it is worth noting that similar to Na ϩ , the affinity of K ϩ for binding to aGDPC was improved dramatically in the presence of Ca 2ϩ . The local concentration of K ϩ at the site of clot formation and/or inflammation due to K ϩ release from aggregated platelets and/or lysed cells may increase to a level that could specifically activate the APC anticoagulant pathway. Further study will be required to determine whether there is validity for this hypothesis.
Finally, the Ca 2ϩ binding loop on the N-terminal ␤-barrel and the Na ϩ binding loop on the C-terminal ␤-barrel of APC are located 15-20 Å away from the catalytic triad and ϳ30 Å apart from each other (Fig. 7). The observation that stabilization of these loops by various metal ions allosterically modulates the catalytic function of APC provides further insight into how coagulation proteases and other more specialized serine proteases may have evolved from a common ancestral fold into highly specific enzymes with very diverse functions. Whereas the key catalytic residues are conserved in all serine proteases, other residues at the vicinity of the primary specificity pocket (S1) or remote from this pocket have diversified. There is overwhelming evidence in the literature that variant residues restrict the specificity of coagulation proteases by enabling them to specifically interact with additional residues surrounding the P1 site of various peptide substrates. In recent years, homology modeling based on the known crystal structures and subsequent mutagenesis studies have identified a number of variant residues critical for determination of the P3-P3Ј specificity of coagulation proteases (27,(43)(44)(45)(46)(47). It is also becoming clear that the binding of specific cofactors or substrates to variant residues and loops remote from the catalytic pocket allosterically regulates the specificity and function of coagulation serine proteases (48 -51). However, except for a few studies with small synthetic substrates and inhibitors, the molecular basis of the coagulation protease specificity with the natural macromolecular substrates have remained largely elusive. The results presented in this study suggest that this may partly be due to evolution of complex allosteric links among different surface loops in coagulation serine proteases that FIG. 7. MOLSCRIPT plot of the catalytic domain of activated protein C derived from the crystal structure of the activated GDPC-PPACK complex (52). The non-catalytic epidermal growth factor-like domains-1 and -2 and the inhibitor PPACK were removed. The 70 -80 Ca 2ϩ loop is located on the N-terminal six-stranded ␤-barrel with a Ca 2ϩ ion positioned at the geometric center of the loop (shown in red). The putative Na ϩ binding loop (from Cys 220 -Gly 226 ) is located on the C-terminal six-stranded ␤-barrel with a Na ϩ ion positioned at the geometric center of the loop (shown in yellow). The catalytic triad residues His 57 , Asp 102 , and Ser 195 are shown in red at the interface of two ␤-barrels.
are not easily identified by homology modeling based on the crystal structures.