The Fourth Epidermal Growth Factor-like Domain of Thrombomodulin Interacts with the Basic Exosite of Protein C*

Thrombomodulin (TM) functions as a cofactor to enhance the rate of protein C activation by thrombin (cid:1) 1000-fold. The molecular mechanism by which TM improves the catalytic efficiency of thrombin toward protein C is not known. Molecular modeling of the protein C activation based on the crystal structure of thrombin in complex with the epidermal growth factor-like domains 4, 5, and 6 of TM (TM456) predicts that the binding of TM56 to exosite 1 of thrombin positions TM4 so that a negatively charged region on this domain juxtaposes a positively charged region of protein C. It has been hypothesized that electrostatic interactions between these oppositely charged residues of TM4 and protein C facilitate a proper docking of the substrate into the catalytic pocket of thrombin. To test this hypothesis, we have constructed several mutants of TM456 and protein C in which charges of the putative interacting residues on both TM4 (Asp/Glu) and protein C (Lys/Arg) have been reversed. Results of TM-dependent protein C activation studies by such a compensatory mutagenesis approach support the molecular model that TM4 interacts with the basic exosite of protein C.

Thrombomodulin (TM) functions as a cofactor to enhance the rate of protein C activation by thrombin ϳ1000-fold. The molecular mechanism by which TM improves the catalytic efficiency of thrombin toward protein C is not known. Molecular modeling of the protein C activation based on the crystal structure of thrombin in complex with the epidermal growth factor-like domains 4, 5, and 6 of TM (TM456) predicts that the binding of TM56 to exosite 1 of thrombin positions TM4 so that a negatively charged region on this domain juxtaposes a positively charged region of protein C. It has been hypothesized that electrostatic interactions between these oppositely charged residues of TM4 and protein C facilitate a proper docking of the substrate into the catalytic pocket of thrombin. To test this hypothesis, we have constructed several mutants of TM456 and protein C in which charges of the putative interacting residues on both TM4 (Asp/Glu) and protein C (Lys/Arg) have been reversed. Results of TM-dependent protein C activation studies by such a compensatory mutagenesis approach support the molecular model that TM4 interacts with the basic exosite of protein C.
Thrombin is the final serine protease of the coagulation cascade that activates fibrinogen to form blood clots (1)(2)(3). Thrombin can up-regulate its own production by activating platelets (4), factor XI (5), and factors V and VIII (1). It can also down-regulate its own production by forming a complex with thrombomodulin (TM) 1 on the surface of endothelial cells (6). Upon interaction with TM, thrombin can no longer clot fibrinogen but, instead, rapidly activates protein C to activated protein C (6,7). Activated protein C is a vitamin K-dependent plasma serine protease that can shut down the coagulation cascade by degrading factors Va and VIIIa by limited proteolysis (8,9). Structural and functional data have indicated that TM binds to exosite 1 of thrombin to change the specificity of thrombin from a coagulant to an anticoagulant enzyme (10 -15). The occupancy of exosite 1 by fibrinogen, the thrombin receptor on platelets and cofactors V and VIII is required for the recognition and efficient activation of these molecules by thrombin in the clotting cascade (3,16,17). Thus, the binding of TM to exosite 1 competitively blocks the interaction of thrombin with these procoagulant proteins. It has been demonstrated that the binding of epidermal growth factor (EGF)-like domains 5 and 6 of TM to exosite 1 is sufficient to block all coagulant properties of thrombin; however, for catalyzing the activation of protein C, the fourth EGF domain of TM (TM456) is also required for its cofactor function (6,10).
Calcium plays a key role in protein C activation by the thrombin-TM complex. Using Gla-domainless forms of protein C, it has been demonstrated that the Ca 2ϩ -binding site critical for activation of protein C by the thrombin-TM complex is located in the 70 -80-loop of the substrate (18,19) (chymotrypsinogen numbering system (20)). This is the same site in trypsin that is also known to bind Ca 2ϩ (21). Occupancy of this site by Ca 2ϩ is a prerequisite for the ability of TM to catalyze the rapid activation protein C by thrombin (18,19). Interestingly, the binding of Ca 2ϩ to this site of protein C is inhibitory for activation by thrombin in the absence of TM (18). It has been hypothesized that the binding of Ca 2ϩ to this site of protein C is associated with a conformational change in the activation peptide of the substrate and that the Ca 2ϩ -stabilized conformer of the activation peptide is complementary for the active site pocket of thrombin in the presence of TM but noncomplementary in the absence of the cofactor (6).
The exact molecular mechanism by which TM456 improves the catalytic efficiency of thrombin toward protein C in the presence of Ca 2ϩ is not known. Previously, it has been hypothesized that TM may function by inducing allosteric changes in the structure of residues in the extended binding pocket of thrombin (6). Thus, it is believed that the TM-altered conformer of thrombin optimally recognizes the Ca 2ϩ -altered conformer of protein C (6). Recently, the x-ray crystal structure of thrombin in complex with TM456 was resolved (15). Surprisingly, structural data did not indicate any allosteric changes in the active site pocket of thrombin upon binding to the TM fragment (15). Molecular modeling of protein C activation, based on the structure of the complex, suggested that binding of the TM5 domain of the cofactor to exosite 1 of thrombin positions the TM4 domain so that a negatively charged region on this domain juxtaposes a positively charged exosite of protein C (15). Several electrostatic interactions between these oppositely charged regions are postulated to facilitate a proper docking of protein C into the catalytic pocket of thrombin. Based on this model, it has been hypothesized that "substrate presentation" is the primary mechanism by which TM may function (15). A previous study proposed a similar function for TM in protein C activation by thrombin (22).
To examine the validity of this molecular model by a mutagenesis approach, we developed a bacterial periplasmic ex-pression system for TM456 and constructed several mutants of TM456 in which the negative charges of five residues in TM4 (Glu 346 , Asp 349 , Glu 357 , Glu 382 , and Glu 374 ) and two residues (Asp 398 and Asp 400 ) in the interface between TM4 and TM5 were reversed by substituting each residue with a Lys in separate constructs. In addition, a double mutant of TM456 containing Glu 346 3 Arg and Asp 349 3 Lys substitutions was constructed and expressed in the same system. Then the positive charges of basic residues of the protein C exosite including Lys 62 , Lys 63 , Arg 74 , Arg 75 , and Lys 78 were reversed by substituting these residues individually or in a combination of two with Glu in a mammalian expression/purification vector system. The protein C mutants were prepared in Gla-domainless forms and expressed in HEK293 cells as described (23). Kinetic characterization of these mutant molecules by this compensatory mutagenesis approach supports the molecular model that the TM4 domain of TM456 in the complex interacts with the basic exosite of protein C. Moreover, analysis of results suggest that the 70 -80-loop of protein C must be stabilized by the Ca 2ϩ ion before it can productively interact with TM4. This accounts for the unique Ca 2ϩ dependence of protein C activation by thrombin in the presence of TM.
Protein C Activation-The initial rates of GD-PC activation by thrombin were measured in 0.1 M NaCl, 0.02 M Tris-HCl, pH 7.5 (TBS), containing 1 mg/ml bovine serum albumin (BSA), 0.1% polyethylene glycol (PEG) 8000, and 2.5 mM Ca 2ϩ in both the absence and the presence of TM456. In the absence of TM456, the time course of activation of each GD-PC derivative was studied by incubating 1 M substrate with 50 nM thrombin at room temperature. At each time point (5-80 min), samples of reactions were transferred to TBS buffer containing 500 g/ml human antithrombin to inhibit thrombin activity. At this concentration of antithrombin, the activity of thrombin was rapidly inhibited, whereas the amidolytic activity of active GD-PC remained stable for more than 120 min. The amidolytic activities of active GD-PC derivatives in the activation reactions were monitored by hydrolysis of 400 M Spectrozyme PCa (American Diagnostica Inc., Greenwich, CT) in the TBS buffer containing 1 mg/ml BSA and 0.1% PEG 8000. The rate of hydrolysis was measured at 405 nm at room temperature in a V max kinetic plate reader (Molecular Devices, Menlo Park, CA). The concentration of active protein C derivatives in reaction mixtures was determined by reference to standard curves that were prepared by total activation of GD-PC derivatives with excess thrombin at the time of experiments. This was accomplished by total activation of 1 M protein C with 10 nM thrombin in complex with 100 nM TM456 and 2.5 mM Ca 2ϩ for 90 min at 37°C. Under these experimental conditions, the wild type zymogen was completely activated in less than 30 min. The GD-PC mutants, which were resistant to activation by the thrombin-TM456 complex in the presence of Ca 2ϩ , were totally activated with thrombin alone in 1 mM EDTA as described (23). In this case, 1 M of each protein C derivative in TBS containing 1 mM EDTA was activated with 50 nM thrombin for 4 h at 37°C. Time course analysis of the activation reactions indicated that all GD-PC derivatives were totally activated under these conditions.
Characterization of the TM456 Derivatives-The ability of TM456 derivatives to bind thrombin (K d(app) values) as well as to function as a cofactor to enhance the catalytic efficiency of thrombin in activation of the GD-PC derivatives (k cat and K m values) were evaluated. The K d (app) values were determined from the cofactor dependence of the initial rate of the activation of a subsaturating concentration of GD-PC (1 M) in the presence of a fixed concentration of TM456 (ϳ5 nM) and increasing concentrations of thrombin (1-60 nM) as described (25). The K d (app) values were estimated from the saturable dependence of the GD-PC activation rates on thrombin concentration using a hyperbolic equation as described (25). The cofactor dependence of the initial rate of activation also allowed an accurate determination of the concentration of TM456 derivatives, since TM456 devoid from the chondroitin sulfate moiety binds to thrombin with a 1:1 stoichiometry (26). The time course of the initial rate of protein C activation in the presence of TM456 derivatives was determined by incubating each GD-PC derivative (1 M) with thrombin (1-50 nM) in complex with saturating concentrations of each TM456 derivative (Ͼ10 ϫ K d(app) values). Under these experimental conditions, the concentrations of the thrombin-TM456 complexes were in proximity to the concentration of free thrombin. After inhibition of thrombin activity by antithrombin, the initial rate of activation was measured from the rate of activated protein C generation as described above. When a detailed kinetic analysis was done to determine values for K m and k cat of protein C activation, the initial rates of activation were measured as a function of the GD-PC concentration with 1 nM thrombin in complex with a saturating concentration of TM456 as described above.

Expression and Purification of Recombinant Proteins-Wild
type and mutant GD-PC derivatives were expressed in HEK293 cells and purified to homogeneity on the HPC4 immunoaffinity column as described (23). The SDS-PAGE analysis of zymogens indicated that protein C derivatives were expressed as two closely migrating subforms that are glycosylation variants observed previously with this protein (23, 27, 28) (data not shown). TM456 derivatives were expressed in the periplasmic space of the XL1-B strain of E. coli using the pIN-III-pelB expression/purification vector system as described (24). The relative positions of the mutant residues in the fourth epidermal growth factor domain of TM are schematically shown in Fig. 1A. TM456 expressed in this system contains a 12-residue epitope for the Ca 2ϩ -dependent monoclonal antibody HPC4 at its N-terminal domain for an easy purification. The periplasmic extract was prepared by a hypotonic shock and passed through the HPC4 column in the presence of Ca 2ϩ as described (24). The column was extensively washed with TBS containing 1 M NaCl and 2.5 mM Ca 2ϩ followed by elution with TBS containing 5 mM EDTA as described (24). The HPC4 eluates were further purified by anion exchange chromatography on a Mono Q column as described (23). SDS-PAGE analysis under nonreducing conditions suggested that all TM456 derivatives had been purified to homogeneity, migrating as single bands with expected apparent molecular masses of ϳ15 kDa (Fig. 1B). Wild type forms of both mammalian and periplasmic TM456 exhibited comparable cofactor activities (within 80%) as determined by protein C activation assays described below.
Initial Rate of GD-PC Activation-The initial rates of thrombin-catalyzed wild type and mutant GD-PC activation were studied in both the absence and presence of TM456. GD-PC is known to contain a single Ca 2ϩ -binding site with an affinity of ϳ50 M, which is located in the protease domain of protein C in a loop consisting of residues Glu 70 -Glu 80 (19,23). It is known that the binding of Ca 2ϩ to this loop of protein C is inhibitory for activation by thrombin in the absence of TM (6,19). However, the metal ion is an obligatory cofactor for activation by thrombin in the presence of the cofactor (6,19). Since several residues of the 70 -80-loop were targeted for the mutagenesis study, it was essential to evaluate the extent to which the interaction of these mutants with Ca 2ϩ has been altered. Thus, the initial rates of GD-PC activation by thrombin were studied as a function of varying Ca 2ϩ concentrations. To simplify the comparisons of the Ca 2ϩ dependence of activations, the data were normalized to maximal inhibition values as described (23). As shown in Fig. 2, the Ca 2ϩ ion inhibited the activation of all GD-PC derivatives with the exceptions of R75E and R74E,R75E with a similar half-maximal value of 30 -40 M (see legend of Fig. 2). However, these values were elevated ϳ2-3-fold for the R75E and R74E,R75E mutants. Since the half-maximal values for the activation of the R74E mutant were not affected, the altered effect of Ca 2ϩ in the thrombin activation of the double mutant R74E,R75E is also probably due to the R75E mutation.
Next, the time course of the initial rates of GD-PC activation by thrombin alone or by the thrombin-TM456 complex was studied at saturating concentration of Ca 2ϩ (2.5 mM). As shown in Fig. 3A, thrombin activated all GD-PC derivatives with comparable or improved initial rates. The greatest improvement (ϳ6 -7-fold) was observed for the thrombin activation of the R74E,R75E double mutant. A detailed kinetic analysis was not possible, since protein C exhibits a K m of greater than 60 M for thrombin in the presence of Ca 2ϩ (29). In contrast to activation by thrombin alone, the TM-dependent activations of the GD-PC derivatives were all impaired at varying degrees. Under conditions described in the legend to Fig. 3, no TM456dependent activation by thrombin was detected for the R74E and R75E mutants of GD-PC, and the rates were impaired ϳ2-fold for K63E and K78E and ϳ5-fold for the K62E and K62E,K63E mutants (Fig. 3B). These results suggested that the basic residues of both 60-and 70 -80-loops are critical for the thrombin activation of protein C in the presence of TM. It should be noted that in addition to these residues, it is known that basic residues of the 39-loop (Lys 37 -Lys 39 ) are also required for the TM-dependent protein C activation by thrombin (30). Consistent with these results, we found that the activation of a triple GD-PC K37E,K38E,K39E mutant by thrombin alone was normal in both EDTA and Ca 2ϩ , but it was severely impaired in the presence of TM456 (data not shown).
Initial Rates of Protein C Activation by Thrombin in Complex with TM456 Derivatives-The initial rates of wild type and mutant GD-PC activation by thrombin in complex with different TM456 derivatives were measured as described above. As shown in Fig. 4, relative to wild type TM456, the cofactor activities of all TM456 mutants in catalyzing the thrombin activation of wild type GD-PC were impaired. The impairments ranged from a maximum of 90 -95% for the E357K, D398K, and D400K mutants to a minimum of 50 -60% for the E374K and R385E mutants of TM456. These results appear to indicate that charged residues under study in TM4 play a role in the cofactor function of TM in enhancing the thrombin activation of protein C. This is consistent with the literature (25,31).
The rationale for the preparation of the R385E mutant of TM456 was that in the previous molecular model of protein C activation it was proposed that the interaction of Arg 385 of TM4 with a basic residue of the 70 -80-loop of protein C (particularly Glu 80 ) might be responsible for the unique Ca 2ϩ dependence of activation (15). Thus, the initial rate of wild type GD-PC acti- vation by thrombin in complex with saturating concentrations of TM456 derivatives were studied as a function of varying concentrations of Ca 2ϩ . The metal ion enhanced the rate of GD-PC activation by thrombin in complex with wild type TM456 by an order of magnitude with K d(app) of ϳ20 M at room temperature (data not shown). Interestingly, however, no Ca 2ϩ dependence of GD-PC activation was observed by thrombin in complex with D349K, E357K, D398K, and D400K mutants of TM456. Although a similar K d(app) of ϳ20 M for GD-PC activation by thrombin in complex with other mutants of TM456 was observed, the extent of the cofactor effect of Ca 2ϩ was, nonetheless, reduced to ϳ2-fold with these mutants. Similarly, no Ca 2ϩ -dependence of activation was observed for the thrombin activation of the R74E and R75E mutants of GD-PC in the presence of wild type TM456. Since the interaction of thrombin with TM456 is not known to be Ca 2ϩ -dependent (6,19), it follows that the cofactor defect of TM456 derivatives must primarily be caused by their inability to interact with the Ca 2ϩ -stabilized conformer of GD-PC in the complex.
Next, the cofactor functions of the TM456 mutants in enhancing the thrombin activation of the compensatory GD-PC mutants were evaluated. Similar to wild type TM456, none of the mutants exhibited any cofactor activity toward thrombin activation of the 70 -80-loop mutants of GD-PC (data not shown). However, the cofactor functions of both the D349K and E346R,D349K mutants in enhancing the activation of the compensatory 60-loop mutants of GD-PC were improved (Fig. 5, shown for the double mutant only). Relative to activation of wild type GD-PC, the activation of K62E, K63E, and K62E,K63E by the thrombin-TM456 complex were impaired 2-5-fold (Fig. 5). Interestingly, however, the defect in the activation of these mutants by thrombin was restored if the E346R,D349K mutant of TM456 was used as the cofactor in the activation reaction. As shown in Fig. 5, the initial rate of activation of K62E,K63E by thrombin in complex with the double mutant of TM456 was essentially a mirror image of the GD-PC activation rate by thrombin in complex with the wild type TM456. Thus, 100% of the cofactor activity of the Glu 346 and Asp 349 charge reversal mutants of TM could be restored if compensatory mutations were made in residues of the 60-loop of protein C. These results clearly suggest that TM4 provides a site for interaction with protein C in the activation complex.
Evaluation of K d(app) Values-The cofactor dependence of the initial rate of protein C activation was also used to evaluate the affinities of the TM456 derivatives for interaction with thrombin in the ternary complex. With the wild type GD-PC as a The initial rate of GD-PC (1 M) activation by thrombin (1-10 nM) in complex with a saturating concentration of each TM456 derivative (500 nM) was measured at room temperature in TBS containing 1 mg/ml BSA, 0.1% PEG 8000, and 2.5 mM Ca 2ϩ . Following inhibition of thrombin activity by antithrombin, the rate of activated protein C generation was measured by an amidolytic activity assay as described under "Experimental Procedures." substrate, a K d(app) value of 5.7 nM for the interaction of mammalian TM456 with thrombin was observed, which is consistent with previously reported values in the literature (31). The bacterial TM456 exhibited a similar K d(app) of 4.2 nM, suggesting that the periplasmic protein has been properly folded (Table I). The K d(app) values for the TM456 mutants ranged from 5 to 15 nM (Table I). The TM456 E346R,D349K mutant exhibited ϳ2.5-fold impairment in binding to the activation complex. Interestingly, the K d(app) values for the E346R,D349K mutant resembled the value for the wild type TM456 if the K62E,K63E mutant of GD-PC was used as a substrate in the activation reaction. Thus, in contrast to a K d(app) value of 10.1 nM for the thrombin-TM456 E346R,D349K interaction as shown in Table  I, this value was decreased to 4.5 nM if the K62E,K63E mutant of GD-PC was used as the substrate in the reaction. These results further support the conclusion that the TM4 domain of the cofactor in the activation complex provides a binding site for the 60-loop of protein C in the ternary complex. The results further suggest that the interaction of protein C with TM4 in the ternary complex contributes to K d(app) of the thrombin-TM456 interaction. The K d(app) values with the two mutants D398K and D400K, which are located outside TM4 in the ␤2-␤3 loop of TM5 (15) were elevated more than all other mutants. To determine whether the loss of cofactor function with these mutants is due to their impaired binding to thrombin or a specific effect related to the protein C activation complex, the cofactor dependence of the amidolytic activity of thrombin toward Spectrozyme TH was studied at saturating concentrations of TM456 (500 nM) for all mutants. It is known that the occupancy of exosite 1 by hirugen, thrombin receptor on platelets, and TM456 is associated with ϳ60% enhancement in the amidolytic activity of thrombin toward this substrate (32). As shown in Table I, the cofactor functions of all mutants were normal in this assay, since they enhanced the catalytic efficiency of thrombin toward the chromogenic substrate to a similar extent (70 -90%). This is in contrast to protein C activation, in which the cofactor functions of these mutants were dramatically impaired under similar assay conditions (500 nM TM456). Thus, the loss of cofactor function of these mutants is not likely to be due to their impaired interaction with thrombin, but rather their impaired interaction with protein C in the activation complex. These results are also consistent with the essential role of these residues in the cofactor function of TM as previously shown by others (25,31).
K m and k cat Analysis-To study the effect of cofactor mutagenesis on the kinetics of GD-PC activation, the initial rate of protein C activation by thrombin in complex with a saturating concentration of each TM456 derivative (500 nM) was determined as a function of increasing concentrations of wild type GD-PC. Analysis of data by the Michaelis-Menten equation suggested that all TM456 derivatives enhanced the rate of GD-PC activation with similar k cat values (Table I). However, the K m(app) values with all mutants were impaired. These results appear to suggest that the primary defect in the protein C activation reactions with these mutants is due to impaired K m(app) values. Interestingly, the impairment in the K m(app) of GD-PC activation by thrombin in complex with the E346R,D349K mutant of TM456 was totally restored if the compensatory K62E,K63E mutant of GD-PC was utilized as the substrate in the activation reaction (Fig. 6). Thus, in contrast to K m(app) and k cat values of 43.1 Ϯ 3.0 M and 15.1 Ϯ 0.8 mol/min/mol for the GD-PC activation by the thrombin-TM456 E346R,D349D complex, the corresponding values were 8.0 Ϯ 0.3 M and 15.7 Ϯ 0.3 mol/min/mol if the K62E,K63E mutant of GD-PC was used as the substrate in the activation reaction (Fig. 6). These results suggest that the interaction of basic residues of protein C with acidic residues of TM4 primarily decreases the K m of the substrate activation by thrombin in the presence of Ca 2ϩ . Noting that the K m constant in an enzymatic reaction partly represents the affinity of the enzyme-cofactor complex for binding the substrate, these results further suggest that the two acidic residues of TM4 are part of a sequence of TM4 that interact with the basic residues of the 60-loop in protein C. It should be noted that the highest available concentration of GD-PC in these reactions was only 20 M; thus, no complete saturation could be observed for all reactions. Hence, kinetic constants for such reactions are estimates that were derived from the best fit of data to the Michaelis-Menten equation.

DISCUSSION
Previous molecular modeling of protein C activation, based on the structure of the thrombin-TM456 complex, predicted that the binding of TM56 domains of the cofactor to exosite 1 of thrombin may orient TM4 so that a negatively charged region in this domain could contact a basic exosite of protein C in the ternary activation complex (15). Based on this model, it was postulated that electrostatic interactions between the oppositely charged residues of these two regions could facilitate a proper docking of protein C into the catalytic pocket of thrombin (15). In this study, we tested the validity of this molecular model by a compensatory "gain of function" mutagenesis approach. Thus, we reversed the charges of the acidic residues of TM4 in TM456 constructs and expressed the mutant cofactor TABLE I Apparent K d and kinetic constants for thrombin activation of wild type GD-PC in the presence of TM456 derivatives The K d(app) values were determined from the cofactor-dependence of the initial rate of activation of 1 M GD-PC in the presence of a fixed concentration of TM456 (Ϸ5 nM) and increasing concentrations of thrombin (1-60 nM) as described under "Experimental Procedures." K m(app) and V max values were determined from the concentration dependence of the initial rate of wild type (WT) GD-PC activation by thrombin (1 nM) in complex with a saturating concentration of each TM456 derivative (500 nM) according to the Michaelis-Menten equation. The percentage increase in SpTH hydrolysis by thrombin (2 nM) in complex with each TM456 derivative (500 nM) was determined in TBS containing 0.1 mg/ml BSA, 0.1% PEG 8000, and 2.5 mM Ca 2ϩ . ND, not determined. fragments in the periplasmic space of bacteria. We then prepared compensatory mutants of protein C in Gla-domainless forms by reversing the charges of certain basic residues of the putative exosite that is believed to interact with TM4. Based on results of previous TM456 (25,31) and protein C (33,34) mutants, we reasoned that such mutants would exhibit impaired cofactor or substrate activities if they were individually evaluated in appropriate functional assays. However, if the molecular model is correct, it was expected that the defective cofactor effect of certain TM456 mutants in catalyzing the thrombin activation of certain GD-PC mutants would fully, or at least partially, be restored. Indeed, the results suggested that the basic residues of the 60-loop of protein C interact with Glu 346 and Asp 349 of TM4 in the ternary activation complex. This was evidenced by the restoration of full cofactor activity of the double mutant of TM456 in catalyzing the thrombin activation of the K62E,K63E mutant of GD-PC. In addition to basic residues of the 60-loop, the molecular model of the protein C activation predicts that the basic residues of the Ca 2ϩ -binding 70 -80-loop (particularly Arg 74 and Arg 75 ) also interact with the acidic TM4 domain of the cofactor. The results of this and other previous mutagenesis studies (33,34) clearly suggest that these basic residues are essential for the TM-dependent activation of protein C by thrombin. However, by the compensatory mutagenesis approach of this study, we could not provide any information as to possible sites of the interaction of the 70 -80-loop residues of protein C with TM4. The reason for the lack of success in mapping these interaction sites by this approach is not known. One possibility is that the guanidinium groups of Arg 74 and Arg 75 are involved in bifur-cated salt bridges with two acidic residues in TM4, and thus the choice of Lys in the compensatory TM456 mutants may have flawed the strategy. The second possibility is that, with the exception of Glu 346 and Asp 349 , other residues of TM4 mutated in this study are located on ␤-sheet structures and that the mutagenesis of these residues may cause subtle conformational changes in TM4, thereby hindering the proper alignment of these residues with basic residues of the Ca 2ϩloop. Such adverse conformational effects in Glu 346 and Asp 349 residues of TM4 would be minimal, since these residues are more flexible and lie in the C-terminal loop and are also solvent-accessible (14). Finally, the other possibility is that Ca 2ϩ may have a complex electrostatic effect on the conformations of charged residues of the 70 -80-loop in the protein C mutants. For instance, in addition to three basic residues, the 70 -80loop of protein C also has four acidic residues, two of which (Glu 70 and Glu 80 ) are involved in ligating the Ca 2ϩ ion. It is likely that, in the absence of Ca 2ϩ , the acidic and basic residues of this loop are in random and disordered conformations and that the binding of the metal ion to the 70 -80-loop leads to the internalization of acidic residues to ligate Ca 2ϩ and exposure of basic residues to facilitate their interaction with TM4 in the thrombin-TM complex. Thus, reversing the charges of Arg 74 and Arg 75 could interfere with such electrostatic steering of this loop by the Ca 2ϩ ion. In agreement with this hypothesis, previous spectroscopic studies have indicated that binding of Ca 2ϩ to the 70 -80-loop of GD-PC is associated with a conformational change in the loop that is detected by quenching of the intrinsic fluorescence of two Trp 76 and Trp 79 residues in this loop (residues 231 and 234 in the protein C numbering) (35).
Ca 2ϩ plays an intriguing role in protein C activation by thrombin. Whereas the metal ion is required for protein C activation by thrombin in the presence of TM, it inhibits the activation by thrombin alone (6). Based on the molecular model of protein C activation by the thrombin-TM456 complex, it has been proposed that the unique Ca 2ϩ dependence of protein C activation by thrombin may be the result of direct contacts between TM4 side chains and protein C residues that coordinate Ca 2ϩ (particularly between Arg 385 of TM4 and Glu 80 of protein C) (15). The observation that the Ca 2ϩ dependence of protein C activation by thrombin in complex with several mutants of TM456 (D349K, E357K, D398K, and D400K) was completely eliminated supports this hypothesis. However, the results do not agree with the proposal that interactions between Arg 385 of TM and Glu 80 of protein C account for this phenomenon, since protein C activation with the R385E mutant of TM456 remained Ca 2ϩ -dependent. Moreover, we previously demonstrated that the substitution of Glu 80 of protein C with Lys results in a protein C mutant whose activation by the thrombin-TM456 complex is independent of Ca 2ϩ (19). Thus, the unique Ca 2ϩ dependence of protein C activation is mediated by direct contacts between acidic residues of TM4 and basic residues of protein C (particularly Arg 74 and Arg 75 ). This is supported by the observation that the activation of both R74E and R75E mutants of GD-PC was severely impaired and that Ca 2ϩ did not stimulate the reaction.
It is known that TM promotes the catalytic efficiency of thrombin toward activation of protein C in the presence of Ca 2ϩ by 3 orders of magnitude (6). Such an improvement in the rate of protein C activation by thrombin in the presence of TM is accounted for both by a 1-order of magnitude decrease in K m and a 2-order of magnitude increase in V max . The mechanism by which TM improves both kinetic constants of protein C activation by thrombin in the presence of Ca 2ϩ is not known. Results of this mutagenesis study suggest that the phospholipid-independent improvement in the K m of activation is pri- marily mediated by an interaction between the basic exosite of protein C and acidic residues of TM4. This is derived from the observation that the K m of activation was impaired with all mutants of TM456. Based on results presented, we believe that the Ca 2ϩ stabilization of the 70 -80-loop of protein C culminates in the exposure of basic residues required for interaction with TM4 and that such an interaction leads to a 1-order of magnitude improvement in K m . However, the result of this study does not provide insight into how TM improves the k cat of protein C activation by thrombin. The observation that the k cat values for the GD-PC activation by thrombin in complex with all mutants of TM456 were comparable, with only K m values being impaired suggests that the interaction of basic residues of protein C exosite with the acidic residues of TM4 may not contribute to the k cat of the activation reaction. Thus, the open question is how does TM improve the k cat of protein C activation by thrombin? Previously, it was hypothesized that TM456 may induce conformational changes in the active site pocket of thrombin that leads to preferential improvement in the catalytic efficiency of thrombin toward protein C activation (6). Although numerous kinetics (36), mutagenesis (37), and spectroscopic (38) studies have supported this hypothesis, no significant structural rearrangement in the active site pocket of thrombin has been observed in the x-ray crystal structure of the thrombin-TM456 complex (15). Molecular modeling based on structural data has indicated that the interaction of protein C with TM4, and thus "substrate presentation" by TM456 may be the primary mechanism by which the cofactor functions (15). However, the results presented in this study suggest that the energy of such an interaction could only be utilized to improve the K m but not the k cat of the reaction. It should also be noted that our approach is suitable for only probing the ionic electrostatic interaction between protein C and TM4 but not other possible noncovalent interactions between the two proteins. In the case of thrombin, hydrophobic interactions between exosite 1 and TM5 make a significant contribution to the energy of binding (15). Thus, it is possible that similar hydrophobic interactions exist between protein C and TM4 that is associated with an improvement in the k cat of protein C activation by the thrombin-TM complex. It is also possible that the lack of a structural change in the active site pocket of thrombin stems from using an active site inhibited thrombin for preparing crystals for the complex. In any case, further studies are needed to understand exactly how TM improves the k cat of protein C activation by thrombin in the presence of Ca 2ϩ .
In summary, our results clearly show that TM4 in the thrombin-TM456 complex provides a binding site for interaction with the basic exosite of protein C. The interaction of the basic exosite of protein C with acidic residues of TM4 improves the K m of activation in the presence of Ca 2ϩ , and this interaction also accounts for the unique Ca 2ϩ dependence of protein C activation by the thrombin-TM complex.