Thermodynamic Linkage between the S1 Site, the Na+ Site, and the Ca2+ Site in the Protease Domain of Human Activated Protein C (APC)

The serine protease domain of activated protein C (APC) contains a Na+ and a Ca2+ site. However, the number and identity of the APC residues that coordinate to Na+ is not precisely known. Further, the functional link between the Na+ and the Ca2+ site is insufficiently defined, and their linkage to the substrate S1 site has not been studied. Here, we systematically investigate the functional significance of these two cation sites and their thermodynamic links to the S1 site. Kinetic data reveal that Na+ binds to the substrate-occupied APC withK d values of ∼24 mm in the absence and ∼6 mm in the presence of Ca2+. Sodium-occupied APC has ∼100-fold increased catalytic efficiency (∼4-fold decrease in K m and ∼25-fold increase in k cat) in hydrolyzing S-2288 (H-d-Ile-Pro-Arg-p-nitroanilide) and Ca2+ further increases thisk cat slightly (∼1.2-fold). Ca2+binds to the protease domain of APC with K d values of ∼438 μm in the absence and ∼105 μmin the presence of Na+. Ca2+ binding to the protease domain of APC does not affect K m but increases the k cat ∼10-fold, and Na+ further increases this k cat∼3-fold and decreases the K m value ∼3.7-fold. In agreement with the K m data, sodium-occupied APC has ∼4-fold increased affinity in binding top-aminobenzamidine (S1 probe). Crystallographically, the Ca2+ site in APC is similar to that in trypsin, and the Na+ site is similar to that in factor Xa but not thrombin. Collectively, the Na+ site is thermodynamically linked to the S1 site as well as to the protease domain Ca2+ site, whereas the Ca2+ site is only linked to the Na+site. The significance of these findings is that under physiologic conditions, most of the APC will exist in Na2+-APC-Ca2+ form, which has 110-fold increased proteolytic activity.

Activated protein C (APC) 1 circulates in blood as an inactive zymogen, protein C (PC), which is a disulfide-linked two-chain vitamin K-dependent glycoprotein with a M r value of ϳ62,000 (1,2). The N-terminal light chain of PC contains a ␥-carboxyglutamic acid (Gla) domain and two epidermal growth factor domains (EGF1 and EGF2), whereas the heavy chain contains the latent serine protease domain (3)(4)(5). Although ␣-thrombin alone can activate PC to APC, the rate is inappropriately slow for a physiologically important reaction (1). The physiologic activator for conversion of PC to APC is the thrombin-thrombomodulin complex at the endothelial surface that may also involve the endothelial PC receptor (6 -8). During conversion of PC to APC, a short peptide consisting of 12 residues in human (2) and 14 residues in bovine (9) is removed from the N terminus of the heavy chain. The resultant enzyme is a serine protease that inactivates two key cofactors of coagulation, namely, factors Va and VIIIa (8). This represents a major anticoagulant mechanism involved in the regulation of hemostasis (8).
The Gla domain of the light chain of APC contains several Gla residues, which are required for calcium binding to this domain (10). In addition, the EGF1 domain contains a high affinity Ca 2ϩ -binding site in the native PC (or APC) molecule (8). Further, Gla domainless PC also contains a single high affinity Ca 2ϩ -binding site (11), and by mutagenesis experiments this site is implicated to be located in the protease domain (12). Thus, it appears that removal of the Gla domain may impair binding of calcium to the EGF1 domain. The protease domain of APC also contains a Na ϩ -binding site, and residue c225 2 has been implicated in determining the Na ϩinduced allosteric regulation of catalytic activity of this as well as of other serine proteases (13).
The proteolytic activity of APC, including the hydrolysis of amide and ester substrates, is regulated by Na ϩ and Ca 2ϩ 1 The abbreviations used are: APC, activated protein C; PC, protein C; S-2288, H-D-Ile-Pro-Arg-p-nitroanilide; PEG, polyethylene glycol 8000; Gla, ␥-carboxyglutamic acid; Ch ϩ , choline; FPRck, D-Phe-Pro-Arg-chloromethylketone; PABA, p-aminobenzamidine; pNA, p-nitroaniline; EGF, epidermal growth factor; APC(S), APC saturated with S-2288. 2 For comparison, the chymotrypsin amino acid numbering system is used. Thus, the numbers with the prefix "c" (e.g. c225) refer to the chymotrypsin equivalents for the protease domain of APC. Where insertions occur, the chymotrypsin numbering is followed by a capital letter, such as A. (13)(14)(15)(16)(17)(18)(19). Initially, Castellino and co-workers (14 -18) published a series of elegant kinetic and biophysical studies describing the potentiation of amidolytic and esterolytic activities of bovine APC. In these studies, they observed a sigmoidal dependence of the reaction velocity at a subsaturating concentration of substrate as a function of Na ϩ . In the absence of structural information available at that time, these authors proposed that APC contained two sites or classes of Na ϩ sites that may be allosterically linked. In this report, we present data that support a heterotropic allosteric linkage of the single Na ϩ -binding site in the protease domain to the S1 site of the substrate. Further, in a recent study, He and Rezaie (19) attempted to establish linkage between the protease domain Ca 2ϩ site and the Na ϩ site. Here, we present data which strongly indicate that occupancy of the S1 site in APC increases the affinity of Na ϩ ϳ4-fold. He and Rezaie (19) did not address this point and only measured the affinity of Na ϩ at a single subsaturating concentration of substrate. Such data led to obtaining a high K d app value for Na ϩ binding and therefore precluded determination of an accurate linkage between the Na ϩ site and the Ca 2ϩ site (19). Moreover, although it is recognized that the c221-c225 loop in APC is most likely involved in binding to Na ϩ (13,19), the site is not structurally defined. Here, we present crystallographic evidence that the Na ϩ site in APC involves not only the c221-c225 loop but also the c183-c189 loop. Based upon our findings, we predict that the Na ϩ site in factors VIIa and IXa will be similar to that in APC but not thrombin.

EXPERIMENTAL PROCEDURES
Reagents-H-D-Ile-Pro-Arg-p-nitroanilide (S-2288) was obtained from Diapharma Inc. Polyethylene glycol 8000 (PEG) and p-aminobenzamidine (PABA) were purchased from Sigma. D-Phe-Pro-Arg-chloromethylketone (FPRck) and diisopropyl fluorophosphate were obtained from Calbiochem. Human PC was purified as a by-product of the purification of factor VII (20,21). A total of ϳ25 mg of PC obtained from six isolations, each starting with 8 liters of plasma, was used for these studies. It showed a single band (ϳ62,000 Da) on nonreduced and two bands (ϳ41,000 and ϳ21,000 Da) on reduced SDS-PAGE (22). For Ca 2ϩ binding studies, FPRck-APC and Gla domainless FPRck-APC were prepared and purified using Mono Q fast flow anion exchange resin as outlined by Esmon et al. (23,24) and Mather et al. (25). Decarboxylated FPRck-APC was obtained by thermal decarboxylation of Gla residues as described earlier for prothrombin (26) and factor X (27). For kinetic studies, human APC was obtained either from Enzyme Research Laboratories Inc. or from Hematologic Technologies, Inc. Both proteins gave comparable results. Gla domainless APC was prepared by chymotrypsin treatment of APC as outlined by Hill and Castellino (18). It was also purified according to their method with the exception that we used a Mono Q fast protein liquid chromatography column instead of the Q Sepharose column (18). Each protein sample was freed of Na ϩ by dialysis and/or by a desalting column exactly as described by Wells and Di Cera (28). Although this step resulted in some loss of APC activity, it was necessary for preparing APC free of Na ϩ . The final concentration of Na ϩ after these steps was Ͻ1 mM as measured by a conductivity meter as outlined by Wells and Di Cera (28). The proteins were frozen at Ϫ80°C in 20-l aliquots, thawed, and used immediately. This freezing and thawing step did not result in a measurable loss of activity. All of the final kinetic data presented in this paper were obtained using APC obtained from Enzyme Research Laboratories or Hematologic Technologies, Inc.
Amino Acid Sequencing and Gla Analysis-Automated Edman degradation of each protein component was performed using an Applied Biosystems model 477A gas phase sequencer. Approximately 0.5 nmol of protein was loaded on the filter cartridge. The proteins from SDS gels were transferred to polyvinylidene difluoride membranes as described by Rosenberg (29). The N-terminal sequence analysis of Gla domainless APC or Gla domainless FPRck-APC revealed two sequences of approximately equimolar amounts, one corresponding to the heavy chain (Leu-Ile-Asp-Gly-Lys) and the other corresponding to the modified light chain (Ser-Lys-His-Val-Asp). This indicates that Gla domainless APC does not contain the Gla domain (N-terminal 41 residues of the light chain) of APC. Decarboxylated APC contained Ͻ0.5 Gla/mol as com-pared with APC, which contained 7.9 Gla/mol as measured by the specific 3 H incorporation (30). APC concentrations were determined from the absorbance at 280 nm using an extinction coefficient of 14.5 for 1% solution (2).
SDS-PAGE Analysis-SDS gel electrophoresis was performed using the Laemmli buffer system (22). The acrylamide concentration was 15%, and the gels were stained with Coomassie Brilliant Blue dye. All of the proteins used in the present study were ϳ98% pure.
Measurement of S-2288 Amidolytic Activity of APC Proteins-The concentration of APC or Gla domainless APC used was between 4 and 40 nM. The S-2288 concentration ranged from 10 M to 3 mM. The buffer used was 50 mM Tris-HCl, pH 7.4, containing 0.1% PEG and various salt combinations given in the legends to the appropriate figures. Chloride salt of choline (Ch ϩ ), a bulky monovalent cation, was used to keep the ionic strength constant to 0.2 M. p-Nitroaniline (pNA) release was measured continuously (⌬A 405 nm /min) for up to 30 min using a Beckman DU65 spectrophotometer equipped with a Soft-Pac kinetics module. An extinction coefficient of 9.9 mM Ϫ1 cm Ϫ1 at 405 nM was used in calculating the amount of pNA released (31). All of the reactions were performed in triplicate. The K m and V max values were obtained using the Enzyme Kinetics program from Erithacus Software. The K d app of binding of Na ϩ or Ca 2ϩ to the substrate-bound APC was calculated from the kinetic data using the following equation.
where V is the apparent maximum velocity under a specified cation concentration represented by x, V is the V max at saturating concentration of the cation, b is the maximum velocity at 0 concentration of the cation, and K d app is the K d of the specified cation for the APC(S) (substrate bound APC). The values for the correlation coefficient (r), a measure of the strength of the goodness of fit, were obtained using the program SPSS, version 10.0. Ca 2ϩ Binding to Decarboxlated FPRck-APC and Gla Domainless FPRck-APC-Calcium ion activity was determined by using a Ca 2ϩspecific electrode and a model 601A digital Ionlyzer (Orion Research). Titrations of the protein at 29 M in 4 ml of Tris/NaCl, pH 7.4 (50 mM Tris-HCl, 200 mM NaCl), were performed by adding 1-2-l increments of 40 mM CaCl 2 at room temperature. In these titrations, bound Ca 2ϩ was taken as the difference between the measured free Ca 2ϩ concentration and the total added. The data were fitted to the following equation using the program GraFit from Erithacus Software.
where K d app is the dissociation constant for binding of Ca 2ϩ to the protein, Ca B is the maximum Ca 2ϩ binding capacity, and Ca f and Ca b represent the free and bound concentrations of Ca 2ϩ , respectively. Determination of K d PABA of Binding of PABA to APC-The K d app of binding of PABA to APC was determined by its ability to competitively inhibit S-2288 hydrolysis in the absence and presence of Na ϩ and/or Ca 2ϩ . The details are provided in the legend to Fig. 6. The IC 50 (concentration of PABA required for 50% inhibition) was determined by fitting the data to the following IC 50 four-parameter logistic equation from Halfman (32) given below.
where y is the rate of pNA release in the presence of a given concentration of PABA represented by x, a is the maximum rate of pNA release in the absence of PABA, and s is the slope factor. Each point was weighted equally, and the data were fitted to Equation 3 using the nonlinear regression analysis program GraFit from Erithcus Software.
To obtain K d PABA values for the interaction of PABA with APC, we used the following equation as described by Cheng and Prusoff (33) and further elaborated by Craig (34).
where [S] is the S-2288 concentration. The K m values obtained under different conditions (listed in Table II) were used to obtain the K d PABA value.
Further Refinement of the Gla Domainless FPRck-APC Structure-The x-ray intensity data and the atomic coordinates (Protein Data Bank code 1aut) used for refinement were provided earlier by Mather et al. (25). The crystallization drop contained 500 M total calcium concentration, which could provide sufficient free Ca 2ϩ for it to bind to the c70 -c80 loop in the protease domain. Examination of the (2F obs Ϫ F cal ) and (F obs Ϫ F cal ) electron density maps revealed that Ca 2ϩ and appropriate solvent water molecules could be added to the structure. Similarly, based upon the Na ϩ site in thrombin (35) and factor Xa (36), we added Na ϩ and appropriate solvent water molecules in that region of the (2F obs Ϫ F cal ) electron density map. The structure was refined using XPLOR (37) and iterative use of computer graphics using the program O (38). In all, three rounds of refinement were performed. The resultant structure has an R factor of 17.8% compared with the starting R factor of 18.4%. The final structure has 149 water molecules, one calcium ion, and one sodium ion. All of the water molecules had significant electron density when the (2F obs Ϫ F cal ) electron density maps were contoured at the 1 level. The coordinates are being deposited with the Research Collaboratory for Structural Bioinformatics.

RESULTS
Na ϩ Potentiation of S-2288 Hydrolysis by APC in the Absence of Ca 2ϩ -In earlier studies, Na ϩ has been shown to be the physiologically relevant and an effective monovalent ion in potentiating the amidolytic activity of APC (13)(14)(15)18). To further investigate the mechanism of Na ϩ -mediated potentiation of S-2288 hydrolysis, we determined K m app and VЈ max at several concentrations of NaCl. We used 200 mM NaCl as the highest salt concentration in these experiments. When the concentration of Na ϩ was less than 200 mM, Ch ϩ was used as a compensating ion to keep the ionic strength constant to 0.2 M. One mM EDTA was present in each buffer to eliminate the effect of divalent cations. These data are presented in Fig. 1A. The values of K m app and VЈ max were calculated for each salt concentration using the Enzyme Kinetics program from Erithacus Software. The results indicate that Na ϩ affects both the K m app and VЈ max of this reaction (Fig. 1A). The K m app and VЈ max values obtained at each Na ϩ concentration are given in Table I. When VЈ max are plotted as a function of Na ϩ concentration ( Fig. 1B), the midpoint of the curve should yield K d app of interaction of Na ϩ with APC saturated with S-2288 (APC(S)) in the absence of Ca 2ϩ ; this value was calculated to be 23.9 Ϯ 2 mM.
Ca 2ϩ Potentiation of S-2288 Hydrolysis by APC in the Absence of Na ϩ -These data are presented in Fig. 2. The kinetic data indicate that Ca 2ϩ does not change the K m app but increases the k cat ϳ10-fold (Table II). From the data of Fig. 2B, we calculate the K d app of Ca 2ϩ in its interactions with APC(S) to be 438 Ϯ 31 M.
Na ϩ Potentiation of S-2288 Hydrolysis by APC in the Presence of Ca 2ϩ -These data are presented in Fig. 3. The kinetic data indicate that Na ϩ decreases the K m app ϳ3.7-fold and increases the k cat ϳ2.8-fold. Importantly, Ca 2ϩ decreases the K d app of Na ϩ interaction with APC(S) to 6.2 Ϯ 1 mM as compared with ϳ24 mM (Fig. 1) when Ca 2ϩ is absent. Thus, Ca 2ϩ site is thermodynamically linked to the Na ϩ site.
Ca 2ϩ Potentiation of S-2288 Hydrolysis by APC in the Presence of Na ϩ -These data are presented in Fig. 4. As noted previously (Fig. 2), Ca 2ϩ does not change the K m app but increases the k cat slightly (1.2 Ϯ 0.1-fold) in the presence of Na ϩ . Because the change in VЈ max is very small, these data could not be used to obtain the K d app of Ca 2ϩ interaction with APC(S). However, we later show that in the presence of Na ϩ this K d app is ϳ105 M as compared with ϳ438 M (Fig. 2) when Na ϩ is absent. Thus, as noted above, the Na ϩ site is linked to the Ca 2ϩ site.
The above experiments were repeated using Gla domainless APC. Potentiation of S-2288 hydrolytic activity at several subsaturating concentrations of Na ϩ (Ϯ 5 mM Ca 2ϩ ) or Ca 2ϩ (Ϯ 200 mM Na ϩ ) was indistinguishable from that obtained with APC. These data are consistent with extensive studies conducted by Hill and Castellino (17) using bovine APC and Gla domainless APC. Thus, the effects of Na ϩ and/or Ca 2ϩ in potentiating the activity of both human and bovine APC and Gla domainless APC are comparable. Importantly, our results for the first time reveal that the substrate binding is allosterically linked to the binding of Na ϩ .
Binding of Ca 2ϩ to Human Decarboxylated FPRck-APC and Gla Domainless FPRck-APC-Bovine Gla domainless APC has  Table I. The concentration of APC used was from 4 to 40 nM; for consistency, the data were normalized to a 40 nM enzyme concentration. The pNA release was measured as outlined under "Experimental Procedures." For these experiments, two stock buffers were made. One buffer was 50 mM Trizma base, 0.1% PEG, 1 mM EDTA (acid form), and 200 mM ChCl adjusted to pH 7.4 with HCl. In the second buffer, 200 mM ChCl was replaced by 200 mM NaCl, and the pH was again adjusted to 7.4 with HCl. These two buffers were mixed in appropriate proportions to yield the desired concentrations of Na ϩ and Ch ϩ . Thus, the monovalent ion concentration was always 200 mM, and the buffer containing 200 mM Ch ϩ had no added source of Na ϩ . B, VЈ max as a function of Na ϩ . The data were fitted to Equation 1. The K d of Na ϩ for APC(S) was calculated to be 24 Ϯ 2 mM. The values of b and V (converted to k cat ) are given in Table II. been reported to contain one high affinity Ca 2ϩ site in the presence of Na ϩ (11). By mutagenesis experiments, a Ca 2ϩbinding site in the protease domain of human APC has also been implicated (12). Here, we have performed direct Ca 2ϩ binding studies with human decarboxylated FPRck-APC and Gla domainless FPRck-APC. These data are presented in Fig. 5 and reveal that the Gla domainless FPRck-APC contains a single Ca 2ϩ -binding site with a K d app value of 105 Ϯ 11 M, whereas the decarboxylated FPRck-APC contains two Ca 2ϩbinding sites with K d app values of 120 Ϯ 19 M. Because Gla domainless FPRck-APC has only one high affinity Ca 2ϩ -binding site compared with the two in the decarboxylated FPRck-APC, it is likely that the EGF1 domain residues 42-49 containing Asp-46 and Gln-49 that are implicated in binding to Ca 2ϩ (39) are flexibly disordered in the Gla domainless FPRck-APC and cannot participate in forming the high affinity site in this domain. Thus, the Ca 2ϩ effects observed on the S-2288 amidolytic activity of APC are due to the Ca 2ϩ -binding site in the protease domain and not in the EGF1 domain.
The data presented thus far establish the interdependence of the binding of Na ϩ and Ca 2ϩ in the protease domain of APC. APC(S) can be converted to the state with both ions bound (sodium-APC(S)-calcium) by acquiring either Na ϩ (via APC(S)calcium) or Ca 2ϩ (via sodium-APC(S)). The ratio of the apparent K d Na in the presence and absence of Ca 2ϩ is 3.9 and that of apparent K d Ca in the presence and absence of Na ϩ is 4.2. Therefore, within experimental error, the net sum of binding energy over the cycle appears to be 0. This establishes an accurate thermodynamic linkage between the Na ϩ site and the protease domain Ca 2ϩ site of APC as opposed to the 16 -20-fold ratio proposed by He and Rezaie (19). Furthermore, because Na ϩ affects the K m app of S-2288 hydrolysis, there appears to be a link between the substrate-binding and the Na ϩ -binding site.  Fig. 1, the data were normalized to a 40 nM APC concentration. Ca 2ϩ did not affect the K m app , and its value ranged from 1723 to 1705 M (Table II). B, VЈ max as a function of Ca 2ϩ . As in Fig. 1, the data were fitted to Equation 1. The K d app of Ca 2ϩ for APC(S) was calculated to be 438 Ϯ 47 M. The correlation coefficient for each curve was Ͼ0.95.  . Each reaction mixture contained 5 mM Ca 2ϩ and appropriate concentrations of Ch ϩ to maintain constant ionic strength. Thus, at 0 mM Na ϩ , the Ch ϩ concentration was 185 mM. As in Fig. 1, the data were normalized to a 40 nM APC concentration. The K m app values ranged from 466 M (in the presence of Na ϩ ) to 1705 M (in the absence of Na ϩ ) and are given in Table II. B, VЈ max as a function of Na ϩ . The K d app of Na ϩ in binding to APC(S) in the presence of Ca 2ϩ was calculated to be 6.2 Ϯ 1 mM. The correlation coefficient for each curve was Ͼ0.95.
Both in the absence and presence of Ca 2ϩ , Na ϩ decreases the K m app of S-2288 by ϳ3.8-fold. Using K m app as an approximation of substrate affinity, one can complete this part of the linkage using the thermodynamic principles (40). Therefore, apparent K d Na and apparent K d Ca to the free form of APC in the absence of substrate or the other cation can be calculated; these values are ϳ95.1 mM and ϳ443 M, respectively.
Effects of Na ϩ and Ca 2ϩ on the Interaction of PABA with APC-We next investigated whether Na ϩ and/or Ca 2ϩ affect the S1 part of the active site in APC. We used the S1 site probe PABA for these studies. The interaction of PABA with APC was determined under each of the four salt conditions. These data are presented in Fig. 6 and summarized in Table III. Under a saturating concentration of Na ϩ , the affinity of PABA for APC was increased ϳ4.2-4.4-fold as compared with that in its absence. Further, Ca 2ϩ had essentially no effect on the apparent K d PABA in the absence or presence of Na ϩ . These K d PABA data agree well with the K m data presented in the previous sections where Na ϩ decreased the K m ϳ4-fold and Ca 2ϩ had no effect. Our PABA data are similar to the PABA binding to bovine APC, where the K d PABA is 13 M in the presence and 95 M in the absence of Na ϩ (17). However, the absolute apparent K d PABA value for human APC is ϳ79 M in the presence and ϳ333 M in the absence of Na ϩ (Table III). Thus, it would appear that compared with human APC, bovine APC binds PABA with ϳ6and ϳ3.5-fold higher affinity in the presence and absence of Na ϩ , respectively. Importantly, our data establish that Na ϩ binding is linked to the substrate binding through its S1 site.
Location of the Protease Domain Ca 2ϩ Site and the Na ϩ Site in the Gla Domainless FPRck-APC Crystal Structure-To provide a structural basis for the linkage between cation-binding sites and the S1 site, we refined the APC structure after including Ca 2ϩ and Na ϩ in the protease domain. A schematic representation of the protease domain of human APC depicting the Ca 2ϩ site and the Na ϩ site are presented in Fig. 7A. Details of the Ca 2ϩ -and the Na ϩ -binding sites are presented in Fig. 7  (B and C, respectively). The Ca 2ϩ -binding site is similar to that in trypsin (41) and involves the carboxylates of Glu-c70 and Glu-c80, carbonyl oxygens of c72 and c75, and two water molecules. The Na ϩ -binding site is similar to that in factor Xa (36) and involves carbonyl oxygens of c184A, c185, c221A, and c224 3 and two water molecules. DISCUSSION The purpose of the present study was to structurally define the protease domain Ca 2ϩ site as well as the Na ϩ site and their roles in regulating the catalytic activity of human APC. The spatial relationship of the Ca 2ϩ site, the Na ϩ site, the autolysis loop, the Asp-c189 S1 site, and the catalytic triad is provided in Fig. 7A. Dang and Di Cera (13) recently reported that several serine proteases, including those in coagulation, possess a func- 3 In APC (25), the residues in the c183 loop are numbered as c184, c184A, c185, c186, c186A, and c187. In factor Xa (36), the equivalent residues in this loop are numbered as c184, c185, c185A, c185B, c186, and c187. The residues in the c221 loop of APC are numbered as c221, c221A, c222, c223, c224, and c225. The equivalent residues in this loop of factor Xa are numbered as c221, c222, c222A, c223, c224, and c225. The residues involved in binding to Na ϩ in APC and factor Xa are given in italics in this footnote.  Fig. 6 using Equation 4. b A similar apparent K d PABA value was obtained when the Na ϩ concentration was increased to 485 mM (Fig. 6). c A similar apparent K d PABA value was obtained when the Na ϩ concentration was 485 mM and the Ca 2ϩ concentration was 5 mM (Fig. 6).  Table III. tional Na ϩ site. X-ray crystal structures of thrombin (35,42) and factor Xa (35, 36) are reported where the Na ϩ site in these molecules is defined. The Na ϩ site in thrombin uses a single loop involving the carbonyl oxygen atoms of residues c221A and c224 as well as four water molecules, whereas the Na ϩ site in factor Xa uses two loops involving the carbonyl oxygen atoms of FIG. 7. Location of the Na ؉ site as well as the Ca 2؉ site in the protease domain of APC. A, overall fold of the protease domain. ␤-Sheets are in red, ␣-helices are in blue, the Ca 2ϩ -binding loop is in white, the two Na ϩ -binding loops are in yellow and white, and the autolysis loop is in magenta. The active site residues (Asp-c102, His-c57, and Ser-c195) are labeled D, H, and S, respectively. The N and C termini are marked N and C, respectively. The positions of residues that are important in binding to Na ϩ as well as to Ca 2ϩ are labeled. The residue numbers are based upon chymotrypsin numbering as outlined by Mather et al. (25). B, details of the Ca 2ϩ -binding site in the protease domain. The main chain carbonyl groups of c72 and c75 and the carboxyl side chains of residues c70 and c80 along with two water molecules that coordinate to Ca 2ϩ are depicted. The Ca 2ϩ -binding loop is colored by atom type. Ca 2ϩ is shown as a white sphere, and water molecules are shown as red spheres. The dotted lines represent coordination of Ca 2ϩ with its ligands. w, oxygen of water. C, details of the Na ϩ -binding site and its relationship to the S1 site. The two Na ϩ -binding loops, c183-c189 (colored red) and c221-c225 (colored magenta), are shown. The Na ϩ is coordinated by the carbonyl oxygen atoms of c184A, c185, c221A, c224, and two water molecules. Tyr-c225 residue, an important determinant for Na ϩ binding (13), and the S1 specificity binding pocket residue, Asp-c189, that is linked to the Na ϩ site as well as to the Tyr-c228, are also depicted. Na ϩ is shown as a blue sphere, and water molecules are shown as red spheres. The dotted lines represent hydrogen bonds as well as coordination of Na ϩ with its ligands. w, oxygen of water. residues c185, c185A, c222, and c224 as well as two water molecules. The nature of the residue c225 plays an important role in orienting the carbonyl oxygen atom of c224 toward the Na ϩ coordination shell (35,36,42). As shown in Fig. 7C, the Na ϩ site in APC resembles that of factor Xa but not thrombin. This may be due to the insertion of three residues in the c183 loop of thrombin. As a result of this insertion, the carbonyl oxygens of this loop in thrombin are spatially distant and unable to coordinate with Na ϩ . Instead, the cavity in thrombin is filled with water molecules, two of which are optimally situated to coordinate with Na ϩ .
In the presence or absence of Ca 2ϩ , Na ϩ increased the affinity of S-2288 to APC (Table II) by ϳ4-fold and binding of PABA (Table III) by a similar fold. Because Ca 2ϩ does not influence the K m app value of S-2288 and has essentially no effect on the PABA binding, it would appear that these effects are primarily mediated through the S1 site in APC. These results can be readily rationalized, because the Na ϩ site in APC is directly linked to the S1 specificity pocket residue Asp-c189 (Fig. 7C). Occupancy of the Na ϩ site could rigidify the c189 side chain as well as the c183-c189 loop for optimal interaction with the basic P1 residue (Arg) of the substrate or the basic amino groups of the S1 probe, PABA.
In the absence of Na ϩ , Ca 2ϩ increased the k cat of S-2288 hydrolysis ϳ10-fold (Table II). In the presence of Na ϩ , however, Ca 2ϩ had a minimal effect (ϳ1.2-fold) on the hydrolysis of S-2288. Of interest, Ca 2ϩ in the presence or absence of Na ϩ had no effect on the K m app of S-2288 hydrolysis as well as on the PABA binding. Thus, Ca 2ϩ does not affect the ground state binding of substrates/inhibitors to APC. The ϳ10-fold increase in k cat for S-2288 hydrolysis thus appears to be due to a decrease in the transition state energy. All of the catalytic effects of Ca 2ϩ on APC are remarkably similar to those observed earlier for factor Xa (43). This is not surprising because both proteases are structurally similar.
Of interest is the fact that the Na ϩ site and the protease domain Ca 2ϩ site are thermodynamically linked. Thus, Ca 2ϩ increases the affinity of Na ϩ by ϳ4.0-fold, and Na ϩ increases the affinity of protease domain Ca 2ϩ binding by a similar fold. The thermodynamic linkage between the Na ϩ site and the protease domain Ca 2ϩ site of APC is depicted in Fig. 8. This part of the linkage is supported by the data presented in Figs. 1, 2, 3, and 5. Thus, although the effects of Na ϩ and Ca 2ϩ at individual steps of the thermodynamic cycle depicted in Fig. 8 are different, the overall change in apparent affinity for Na ϩ or Ca 2ϩ is the same regardless of the pathway followed in going from APC(S) to the sodium-APC(S)-calcium state or the pathway followed in going from APC to the sodium-APC-calcium state.
A major observation of significance in this paper is that the S1 site in APC is linked to the Na ϩ site. Hence, occupancy of the S1 site increases the affinity of Na ϩ ϳ4-fold, and Na ϩ increases the affinity (K m ) of the substrate by a similar fold. This part of the thermodynamic linkage between the S1 site and the Na ϩ site in APC is also depicted in Fig. 8 and is supported by the data presented in Figs. 1, 3, and 6. Consequently, although the effects of Na ϩ and substrate at individual steps of this part of the thermodynamic cycle are different, the overall change in apparent affinity for Na ϩ or substrate S-2288 is the same regardless of the pathway followed in going from APC to the sodium-APC(S) state or the pathway followed in going from APC-calcium to the sodium-APC(S)-calcium state. Note that Ca 2ϩ is not thermodynamically linked to the S1 site. Thus, in going from APC to the APC(S)-calcium state, the affinity of Ca 2ϩ or the substrate for APC is not altered regardless of the pathway followed in Fig. 8. Similarly, in going from sodium-APC to the sodium-APC(S)-calcium state, the affinity of Ca 2ϩ or the substrate for sodium-APC is not altered regardless of the pathway followed in Fig. 8.
The plasma concentration of Ca 2ϩ is ϳ1 mM and that of Na ϩ is ϳ150 mM. Under these conditions, ϳ85% of the formed APC will exist in the sodium-APC-calcium state. Binding of the substrate further drives this equilibrium in the direction that favors APC binding to Na ϩ and Ca 2ϩ . Thus, the overall effect of the linkage between the S1 site, the Na ϩ site, and the Ca 2ϩ site is that during physiologic hemostasis, the APC formed will mostly exist in the sodium-APC-calcium state that has maximum anticoagulant activity. Notably, the linkage between Ca 2ϩ and Na ϩ and the S1 site are essentially the same as proposed for factor Xa earlier (43). This similarity between the two proteases (APC and factor Xa) is consistent with the structural resemblance in the Ca 2ϩ site, the Na ϩ site, and the S1 site.
The Ca 2ϩ site in the APC protease domain (Fig. 7B) is typical of that found in trypsin-like proteases (41). It is of note that the c70 -c80 calcium loop in APC contains three basic and two hydrophobic Trp residues (25). The serine proteases that bind Ca 2ϩ contain primarily acidic and polar residues in this loop, which facilitate attraction of a positively charged divalent cation. Thus, trypsin contains four acidic and three polar residues (41), and elastase contains two acidic and seven polar residues (44). Similarly, human factor IXa contains five acidic and five polar residues and is thought to bind Ca 2ϩ in this loop (45). Human factor VIIa contains six acidic and three polar residues and contains a Ca 2ϩ -binding site in this loop (46). Furthermore, the Gla domainless PC or APC was reported to contain only a single Ca 2ϩ -binding site (11), which was thought to be located in the EGF1 domain (39). Based upon the above considerations, Bajaj et al. (45) had initially proposed that PC or APC might not contain a Ca 2ϩ -binding site in the protease domain. However, recent mutagenesis experiments (12), direct binding studies (Fig. 5), and crystal structure data (Fig. 7B) clearly reveal the presence of a Ca 2ϩ -binding site in the c70 -c80 loop of APC. Thus, the Ca 2ϩ site in the Gla domainless PC or APC previously attributed to the EGF1 domain is most likely the Ca 2ϩ site in the protease domain.
Structural features of the c70 -c80 Ca 2ϩ -binding loop of APC are shown in Fig. 9. A unique feature of this loop appears to be that all three basic residues and the two hydrophobic Trp residues are exposed to the solvent. The other hydrophobic residue Leu-c73 is buried in a hydrophobic pocket involving spatially close neighbors. The Trp-c76 is stabilized by two hydrogen bonds and van der Waals' interactions involving the side chain carbon atoms of Lys-c78. Similarly, Trp-c79 is held in place by a hydrogen bond involving the carbonyl oxygen of Location of Leu-c73 in a hydrophobic pocket comprised of Val-c32, Leu-c40, and Trp-c141 is shown in white. Two hydrophobic residues and three basic residues, namely, Trp-c76, Trp-c79, Arg-c74, Arg-c75, and Lys-c78 are exposed to the solvent. These residues are stabilized by hydrogen bonds and van der Waals' interactions and are discussed in the text. The dotted lines represent hydrogen bonds as well as coordination of Ca 2ϩ with its ligands. Magenta sphere (Ca), calcium; red sphere (w), oxygen of water.
Gly-c69. The hydroxyl group of Tyr-c71 in this loop is hydrogenbonded to a water molecule as well as to the side chain of Thr-c22. Further, the phenyl ring of Tyr-c71 is stabilized through van der Waals' interactions with the side chain carbon atoms of Arg-c24. Thus, it is not unreasonable to conclude that the binding of Ca 2ϩ to this loop provides favorable energy for the two Trp residues to stay exposed to the solvent. Interestingly, these two Trp resides have been implicated to be important for the activation of PC by the thrombin-thrombomodulin complex (8). CONCLUDING

REMARKS
In this report, we provide evidence that the Na ϩ site in APC is thermodynamically linked to the S1 site as well as to the protease domain Ca 2ϩ site. Such information was critical in defining the accurate thermodynamic cycle for the linkage of the Na ϩ site to the Ca 2ϩ site as well as to the S1 site (Fig. 8).
Our data also indicate that most of the APC under physiologic conditions will exist as sodium-APC-calcium, which has maximal biologic activity. These conclusions with respect to function are supported by the structural data. The Ca 2ϩ site in APC is similar to other trypsin-like proteases (41), and the Na ϩ site is similar to factor Xa (36) but not thrombin (35). Based upon the sequence comparison (47) and structural homology between factor Xa (36), factor IXa (48), and factor VIIa (46), we postulate that factors IXa and VIIa will have Na ϩ sites similar to that of factor Xa and APC. If so, then Na ϩ appears to play a structural role in maintaining the fully active conformer of the protease domain in these vitamin K-dependent proteins.