Activation-induced Exposure of the Thrombin Anion-binding Exosite INTERACTIONS OF RECOMBINANT MUTANT PROTHROMBINS WITH THROMBOMODULIN AND A THROMBIN EXOSITE-SPECIFIC ANTIBODY*

The activation of serine protease zymogens involves conformational changes that increase the affinity of substrate binding and the activity of the catalytic cen- ter. The activation of prothrombin is particularly complex and requires several cleavages in the proenzyme region in addition to the conserved activation cleavage after Areao. To understand how these cleavages lead to the exposure of the thrombin anion-binding exosite, a major macromolecular recognition site, interactions of recombinant human prothrombin derivatives with thrombomodulin, and an exosite-specific antibody were studied by competition binding and immunoprecipitation. By either method, the anion-binding exosite is not functional on prethrombin 2, which cleaved lacks fragment 1.2, nor on meizothrombin, which is cleaved only after Areno. In contrast, the exosite is fully exposed on meizothrombin des-F1, which is cleaved after both and (Fl). tion, 5 rn) were co-precipitated with a rabbit antiserum against a carboxyl-terminal peptide of human thrombomodulin (27). Immunopre- cipitation experiments were performed with protein A-Sepharose beads at

Conformational changes during zymogen activation are essential for most serine proteases to form high affinity complexes with substrate and for the enzyme-substrate complex to reach the transition state (1). In the case of trypsinogen, surface loops become reoriented after activation to trypsin, permitting the binding of substrate residues on the amino-terminal side of the scissile bond (2). For thrombin, however, the requirements for substrate recognition are more stringent. Thrombin function toward macromolecules requires both the serine active site and the anion-binding exosite that is some distance from the catalytic center (3). This exosite is essential Health Grant HLBI14147 (Specialized Center of &search in Thrombo-*This research was supported in part by National Institutes of sis). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  for interactions with substrate residues on the carboxyl-terminal side of the scissile bond but is not available in prothrombin. Analogous interactions do not seem to be important for the activity of trypsin. During blood coagulation, prothrombin is converted to thrombin by factor Xa that cleaves both the Ar$20-Ile321 bond and the Arg271-ThP72 bond ( Fig. 1) (4)(5)(6)(7)(8)(9)(10)(11). (Numbering is from the amino terminus of human prothrombin) (12). Depending on the reaction conditions, either bond may be cleaved first. Cleavage after Ar$20 yields meizothrombin. This cleavage site is homologous to the activation cleavage sites of trypsinogen and other serine protease zymogens. Alternatively, cleavage after releases an amino-terminal polypeptide termed fragment 1.2 (F1.2)l and yields the catalytically inactive prethrombin 2. Cleavage of both bonds by factor Xa yields active thrombin, which consists of two chains linked by a disulfide bond. In the presence of factor Va, calcium ions, and phospholipids, the pathway favored by factor Xa appears to be through meizothrombin to thrombin. The conversion of meizothrombin to thrombin needs not be direct, however, because meizothrombin and thrombin could cleave the Arg165-Ser156 bond to release fragment 1 (Fl) and yield meizothrombin des-F1.
The substrate specificity of the prothrombin intermediates illustrates the complexity of the thrombin activation mechanism. Meizothrombin and thrombin have similar activity toward small amide substrates, but human meizothrombin does not interact with macromolecules such as fibrinogen (13) and thrombomodulin (14). Cleavage of peptidyl amides requires the binding of substrate residues on the amino-terminal side of the scissile bond to appropriate subsites on the enzyme surface. These sites appear to be unavailable in prothrombin but are fully functional in meizothrombin. Removal from meizothrombin of F1.2 enables the exosite to bind substrate residues on the carboxyl-terminal side of the scissile bond. This is important for thrombin to efficiently cleave macromolecules in addition to small amides and esters.
To understand the mechanism by which proteolytic activation leads to the exposure of the thrombin exosite, we studied interactions of recombinant mutant prothrombins with thrombomodulin and a thrombin exosite-specific antibody. The results indicated that the exosite was not accessible in either meizothrombin or prethrombin 2 and that cleavage after either in meizothrombin or Ar$'O in prethrombin 2 was essential for interactions with several macromolecules. Thus, the conformational changes upon thrombin activation were much more extensive than those accompanying trypsin activation, The abbreviations used are: F1.2, prothrombin fragment 1.2; DIP, diisopropyl phosphoryl; F1, prothrombin fragment 1.

3726
Conformation of the Thrombin Anion-binding Exosite creating binding sites for substrate residues on both sides of the scissile bond. These results also suggest that the prothrombin F1 fragment regulates the accessibility of the thrombin anionbinding exosite.
EXPERIMENTAL PROCEDURES Materials-Human plasma thrombin (3200 NIH unitslmg) was from Haematologic Technologies Inc., Essex Junction, VT. Human plasma prothrombin was provided by Drs. G. J. Broze and J. P. Miletich (Washington University). Human fibrinogen was from Kabivitrum, Stockholm, Sweden. Echis carinatus venom was from Sigma.
Ecarin, the prothrombin activator from E . carinatus venom, was purified as described previously (14,15). A rabbit anti-human thrombin antibody was provided by Dr. P. W. Majerus (Washington University). The polyclonal antibody recognized thrombin, prothrombin, and prothrombin activation intermediates and inhibited thrombin-clotting activities (16). Antithrombin autoantibody D from a patient with recurrent arterial thrombosis was purified by ammonium sulfate precipitation, Q-Sepharose ion-exchange chromatography, and protein GSepharose affinity chromatography, as described previously (17,18). The purified IgG fractions inhibited thrombin-clotting activity with an ICso of 1.8 JIM but had no effect on thrombin-antithrombin I11 interaction (17). Recombinant Prothrombins-Expression and characterization of recombinant human wild type prothrombin and mutant prothrombins K52E,2 R68E, R70E, and S205Awere described previously (19,20). The active-site mutant thrombin S205A had no detectable catalytic activities but bound to thrombomodulin normally. Plasmids pPTH66E and pPTK154E were constructed to express mutant prothrombins H66E and K154E, in which residues His 66 and Lys 154, respectively, of the human thrombin B chain were replaced by Glu. Plasmids pPTAE25 and pPTAPQE23-25 were constructed to express mutant prothrombins AE25 and APQE23-25, in which residue Glu 25 and residues Pro 23, Gln 24, and Glu 25, respectively, of the human thrombin B chain were deleted. Mutant prothrombin H66E was stably expressed in CV-1 cells. Mutant prothrombins K154E, AE25, and APQE23-25 were stably expressed in human embryonic kidney 293 cells.
Expression and Purification of Prethrombin 2-Plasmid pAPT was constructed to express human prethrombin 2 by deleting sequences for the Gla and kringle domains of prothrombin. The codon for Val" (GTG) in the parent plasmid pCMWT (19) was changed to encode Ala17 (GCA), thereby creating an NsiI restriction site. A second NsiI site was introduced by changing the codons for G 1~~~~-G l y~~~-A r g~~~ (GAA GGG CGT) to encode Asp-Ala-Phe (GAT GCATTC). The NsiI fragment was excised from the resultant plasmid to yield plasmid pAPT, which encoded a polypeptide containing the putative signal peptide of prothrombin (Met-43-Phe"6) followed by the thrombin A chain and B chain (WT2-GluST9). Signal peptidase was predicted to cleave the Ala-17-Phe-16 bond, using a computer program described by Folz and Gordon (21). Prethrombin 2 was stably expressed in CV-1 cells and purified from conditioned medium by Amberlite CG-50 ion-exchange chromatography (22), heparin-Sepharose CL-GB affinity chromatography (23), and high pressure liquid chromatography ion-exchange chromatography (TSK SP-5PW, TOSOHAAS, Montgomeryville, PA). Purity of recombinant prethrombin 2 was confirmed by SDS-polyacrylamide gel electrophoresis analysis and silver staining. Protein was quantitated by optical absorption at 280 nm using an extinction coefficient (1 mg/ml) of 1.95 (11). The amino-terminal sequence of the purified protein was determined by automated Edman degradation using an Applied Biosystems model 470A protein Sequencer (24).
Thrombomodulin Binding-Competition equilibrium binding was performed with CV-l(TMnc) cells expressing recombinant human thrombomodulin or control CV-1 cells (25). 12sI-Labeled diisopropyl phosphoryl (DIPFthrombin at 0.3 rn was incubated with cells at 4 "C for 2 h with increasing concentrations of unlabeled competitor. Binding medium contained 1.8 m~ calcium chloride. Cells were washed with cold phosphate-buffered saline (10 m~ phosphate, pH 7.4,150 m~ NaCl) and solubilized with 1 ml of 1 M NaOH. Bound 1261-labeled DIP-thrombin was quantitated by y spectroscopy. The competitive binding equation for the two ligands uersus one receptor model was used to fit the data (26).
thrombin (12). In the trivial names for mutant recombinant prothrom-Superscripted numbers of amino acid residues are for human probins and the thrombins derived from them, the altered amino acid the human thrombin B-chain. The first residue of the thrombin B-chain residues are numbered sequentially beginning with the first residue of is Ile321 of prothrombin. Metabolic Labeling and Immunoprecipitation-Confluent CV-1 cells and 293 cells expressing recombinant prothrombins or control CV-1 cells were metabolically labeled with [36Slcysteine (100 pCi/ml/106 cells) for 8 h at 37 "C. Labeled conditioned media were pretreated either with crude E. carinatus venom or with purified ecarin as described previously (14). Immunoprecipitation was performed with a rabbit anti-human thrombin polyclonal antibody (16) and an anti-thrombin autoantibody (17).
In experiments to detect thrombomodulin complex formation, [36Slcysteine-labeled prothrombin mutant S205A in conditioned medium containing 1.8 m~ calcium chloride was pretreated with crude E. carinatus venom to produce various prothrombin intermediates. Control CV-1 cells or CV-l(TMnc) cells expressing recombinant thrombomodulin were solubilized in lysis buffer (100 n m Tris-HC1, pH 8.1; 0.6% (v/v) Triton X-100). After removing cell debris by centrifugation, cell lysate (200 pl) was added to the mixture of prothrombin derivatives (600 pl). Proteins bound to thrombomodulin (final concentration, 5 rn) were co-precipitated with a rabbit antiserum against a carboxyl-terminal peptide of human thrombomodulin (27). Immunoprecipitation experiments were performed with protein A-Sepharose beads at 4 "C as described previously (26). Fig. 1. Previous studies have shown that human meizothrombin has normal activity toward small amides but very low activities for fibrinogen, platelets, and thrombomodulin (13,14). Binding sites for these macromolecules are located in the thrombin anion-binding exosite (28), suggesting that a single cleavage after of prothrombin is not sufficient for the exposure of the thrombin exosite and that the presence of prothrombin fragment F1.2 may prevent access to the thrombin exosite. To test whether removal of F1.2 from prothrombin is sufficient to expose the exosite, we expressed recombinant human prethrombin 2 by site-directed mutagenesis. The amino-terminal sequence of purified recombinant prethrombin 2 was determined to be Phe-Thr-Ala-Thr-Ser-Glu-Qr-Gln-Thr-Phe, corresponding to that of human thrombin A chain with an extra Phe residue at the amino terminus (12). As expected, recombinant prethrombin 2 was catalytically inactive. After activation with E. carinatus vernom, the resultant thrombin had activity toward fibrinogen, protein C, and platelets similar to that of thrombin prepared from recombinant wild type prothrombin (data not shown).

Expression of Human Prethrombin 2-Selected proteolytic derivatives of human prothrombin are illustrated schematically in
Binding of Prethrombin 2 to Thrombomodulin-The affinity affinity is similar to that observed for plasma-derived thrombin or recombinant thrombin binding to recombinant human thrombomodulin (14,19,26). These results indicate that simply removing F1.2 from prothrombin is not sufficient to expose the thrombomodulin binding site and that additional cleavage of the peptide bond at Ar$20-Ile321 is also required. Binding of Meizothrombin des-Fl to Thrombomodulin-The binding of meizothrombin des-F1 to thrombomodulin was assessed by complex formation and immunoprecipitation (Fig. 3).
[35SlCysteine-labeled conditioned medium was prepared from control CV-1 cells or cells expressing active-site mutant prothrombin S205A (19). This active-site mutant allowed us to generate various stable prothrombin intermediates without further autoproteolytic cleavages (14). Meizothrombin S205A, meizothrombin des-F1 S205A, and thrombin S205A were immunoprecipitated by an anti-human thrombin polyclonal antibody (Fig. 3). Under nonreducing conditions, meizothrombin S205A exhibited a single band of 72 kDa (Fig. 3 A , lane 2 1; under reducing conditions, it exhibited two bands of 50 and 32 kDa, corresponding to the prothrombin F1.2 plus the thrombin A chain and the thrombin B chain, respectively (Fig. 3B, lane 2). Meizothrombin des-F1 S205A migrated under nonreducing conditions as a species of -54 kDa (Fig. 3 A , lane 2), and it was reduced to the thrombin B chain and the small polypeptide consisting of F2 plus the thrombin A chain migrating near the dye front (Fig. 3B, lane 2).
When cell lysate containing thrombomodulin was added to the mixture of prothrombin intermediates, both meizothrombin des-F1 S205A and thrombin S205A, but not meizothrombin S205A, were co-precipitated by an anti-thrombomodulin antibody (Fig. 3, A and B, lanes 3). A nonspecific band was present in the reactions with or without thrombomodulin (Fig. 3, A and  B, lanes 3 and 4 ) . These results indicate that removal of the F1 fragment from meizothrombin enables meizothrombin des-F1 to bind to thrombomodulin, suggesting that the presence of the F1 fragment prevents access of the exosite in meizothrombin by macromolecules. This conclusion was further supported by the following studies with an anti-thrombin autoantibody. Localization of the Epitope for the Autoantibody D in the Thrombin Anion-binding Exosite-In a previous report (171, an autoantibody (D) from a patient with recurrent arterial thrombosis was shown to inhibit the fibrinogen clotting and thrombomodulin binding activities of thrombin, suggesting that the antibody recognized an epitope in the thrombin exosite. To further localize the epitope for this autoantibody, the interaction of purified antibody D with various recombinant mutant thrombins was examined by immunoprecipitation (Fig. 4). Antibody D recognized wild type thrombin, mutant thrombins K52E and K154E, and active-site mutant thrombin S205A. In contrast, the antibody did not recognize mutant thrombins H66E, R68E, and R70E, nor deletion mutant thrombins AE25 and APQE23-25. Residues His 66, Arg 68, Arg 70 and residues Pro 23, Gln 24, Glu 25 are located in two adjacent surface loops (loop segments Lys 65-Glu 76 and Phe 19-Leu 27, respectively) in the anion-binding exosite, whereas residues Lys 52 and Lys 154 are located in other surface loops near the exosite (29). As controls, the rabbit anti-human thrombin antibody ( H ) precipitated wild type thrombin and all mutant thrombins, whereas an antibody fraction isolated from normal subjects (N) did not immunoprecipitate recombinant thrombins (Fig. 4). Differences in the intensity of autoradiographic bands among the recombinant thrombins or thrombin derivatives correlated with the different levels of prothrombin expressed in the stable cell lines. These results indicate that antibody D binds to a region in the thrombin exosite that overlaps the previously identified thrombomodulin binding site (19). The conclusion was further supported by separate fibrinogen clotting assays in which mutant thrombin R70E was completely resistant to the inhibition of antibody D at concentrations up to 150 &ml, whereas plasma-derived thrombin was inhibited in a dose-dependent manner (data not shown). These results are consistent with the previous finding that antibody D inhibits thrombin interaction with fibrinogen and thrombomodulin but not thrombin amidolytic activity (17).
Binding of Antibody D to Prothrombin and Prethrombin 2-Antibody D was employed to detect conformational changes of the thrombin exosite during prothrombin activation. Neither  autoantibody D (D), or a negative control I g G fraction from normal individuals (N). Immunoprecipitated proteins were analyzed on 10% polyacrylamide gels under reducing conditions. recombinant wild type prothrombin (Fig. 5, lane 5) nor prethrombin 2 (Fig. 5, lane XI ) was immunoprecipitated by antibody D. After activation with E . carinatus venom, however, the resultant thrombin forms were precipitated (Fig. 5, lanes 8 and  14). These results indicate that the thrombin anion-binding exosite is not available to antibody D in prethrombin 2. This conclusion is consistent with the inability of recombinant prethrombin 2 to bind thrombomodulin (Fig. 2). either with crude E . carinatus venom or purified ecarin. Ecarin cleaved the single peptide bond at Ar$20-Ile321 of prothrombin (15) and yielded meizothrombin S205A (14). Crude E. carinatus venom variably cleaved in addition the Arg155-Ser156 and Ar871-Th?72 bonds to generate meizothrombin des-F1 S205A and thrombin S205A. All of these forms were immunoprecipitated efficiently by rabbit polyclonal antibody H (Fig. 6, A and   B, lanes X and 3 ) .

Binding ofArttibody D to Meizothrombin and
Thrombin S205A was recognized by antibody D (Fig. 4; Fig.   6 A , lane 2 ) , indicating that mutation of the catalytic serine did not affect the binding of the antibody to the thrombin exosite. In contrast, antibody D bound poorly to meizothrombin S205A (Fig. 6, A and B, lanes 4 ) . As measured by densitometry, no more than 20% of meizothrombin S205A was precipitated by antibody D compared with that precipitated by polyclonal antibody H (Fig. 6 A , lanes 3 and 4 ) . The difference was more dramatic in the reducing gel (Fig. 6B, lanes 3 and 4 ) .
This suggests that in meizothrombin the epitope for antibody D in the exosite is not fully exposed. This is consistent with our previous observations that human meizothrombin has very low affinity for thrombomodulin (14). Interestingly, meizothrombin des-F1 S205A was precipitated with similar efficiency by antibody D and by antibody H (Fig. 6, A and B, lanes X and 21, suggesting that the thrombin exosite became further exposed upon the removal of the F1 fragment from meizothrombin. These results are in agreement with the finding that meizothrombin des-F1 can bind to thrombomodulin (Fig. 3). DISCUSSION Thrombin is a remarkable serine protease that plays a n important role in a variety of biological events. Thrombin-specific function is determined by its unique structural features. These include several enlarged insertion surface loops, which restrict the access of substrates to the serine active center, and clusters of positively charged residues in the anion-binding exosite, which constitutes a major substrate recognition site. A number of studies have demonstrated that the exosite contributes significantly to thrombin interactions with fibrinogen, thrombomodulin, heparin cofactor 11, the thrombin platelet receptor, and hirudin (28). In this report, we used recombinant prothrombins to study conformational changes of the thrombin exosite induced during prothrombin activation and to understand how activation cleavages convert prothrombin to an active enzyme.
Reorientation of surface loops with substrate binding function is an important step in zymogen activation. Our results indicate that the conformational changes upon thrombin activation are much more extensive than those accompanying trypsin activation. The cleavage of an activation hexapeptide from trypsinogen exposes a new Ile-Val amino terminus and enables trypsin to interact with substrates. Structures determined by x-ray crystallography have shown major conformational changes in the trypsin surface loops that interact with substrate residues on the amino-terminal side of the scissile bond (2). Similar conformational changes in corresponding thrombin surface loops appear to be created by the analogous cleavage after A q f Z 0 in prothrombin, because meizothrombin is able to bind small amide substrates (13) and a thrombin active-site inhibitor, dansylarginine N-(3-ethyl-1,5-pentanediyl)amide (30). Interactions with these peptidyl amides depend on subsites of thrombin that bind substrate residues on the aminoterminal side of the scissile bond. In addition, the affinity of dansylarginine N-(3-ethyl-1,5-pentanediyl)amide was reported to be 30 times higher for bovine thrombin than for bovine prethrombin 2 (31), further illustrating the conformational changes induced by cleavage of the peptide bond after Ar$20. In this study, however, we have demonstrated functionally significant conformational changes in the thrombin exosite, which is distant from the active site and interacts with substrate residues on the carboxyl-terminal side of the scissile bond. The corresponding surface loops in trypsinogen do not appear to change in conformation upon activation (2). Our results are consistent with a recent study in which structural differences between thrombin and prethrombin 2 around the active site and near the exosite were observed by monitoring the fluorescence of tryptophan residues (32).
Binding studies with both thrombomodulin and autoantibody D have shown that the thrombin exosite is not accessible in human meizothrombin and that an additional cleavage at removing the F1 fragment is required for thrombin to interact with these macromolecules (Figs. 3 and 6). Although the kringle domains of prothrombin are generally considered to be important for interactions within the prothrombinase complex (331, the precise function of each kringle domain remains unknown. Studies with recombinant mutant prothrombins lacking kringle domains have indicated that these kringle domains are not necessary for prothrombin binding to phospholipid membranes (34). We have shown that the presence of the F1 fragment prevents access to the thrombin exosite by macromolecules, thereby inhibiting the function of human meizothrombin. This may represent a regulatory function of kringle 1 in the prothrombin activation fragment.
In this study, human recombinant prethrombin 2 was expressed and purified. As determined by competition binding studies, the affinity of prethrombin 2 for human recombinant thrombomodulin was at least 20-fold lower than the affinity of thrombin (Fig. 2). The results are consistent with a previous observation that the affinity of bovine prethrombin 2 for rabbit thrombomodulin was approximately 10-fold lower than that of bovine thrombin (35). Studies with autoantibody D demonstrated that the exosite was not accessible in human recombinant prethrombin 2 (Fig. 5). It has been reported, however, that hirugen binds to bovine thrombin and prethrombin 2 with similar affinity (36). Hirugen (37) is a synthetic dodecapeptide corresponding to the carboxyl-terminal portion of hirudin, which interacts with the thrombin exosite (38, 39). It is possible that relatively small size and greater flexibility enable hirugen to bind to regions on prethrombin 2 that are not accessible to macromolecules such as thrombomodulin.
The binding of meizothrombin des-F1 to thrombomodulin (Fig. 3) suggests that this complex could have functional significance. The cleavage site at Arg155-Ser156 between the F1 and F2 fragments is conserved among prothrombins of all species thus far studied (12,40-42), suggesting evolutionary pressure to maintain this prothrombin intermediate. In addition, the 3-dimensional structure of a noncovalent F2-thrombin complex shows that F2 interacts with the putative heparin binding site near the carboxyl-terminal a-helix of thrombin (43). Meizothrombin des-F1 may, therefore, be resistant to heparin inhibition. In a previous study, bovine meizothrombin des-F1 was reported to catalyze protein C activation in a thrombomodulindependent manner (44). Further studies are required to test the stability of the human meizothrombin des-F1-thrombomodulin complex and to determine if the complex has an unexpected substrate specificity.