The Anticoagulant Thrombin Mutant W215A/E217A Has a Collapsed Primary Specificity Pocket*

The thrombin mutant W215A/E217A features a drastically impaired catalytic activity toward chromogenic and natural substrates but efficiently activates the anticoagulant protein C in the presence of thrombomodulin. As the remarkable anticoagulant properties of this mutant continue to be unraveled in preclinical studies, we solved the x-ray crystal structures of its free form and its complex with the active site inhibitor H- D -Phe- Pro-Arg-CH 2 Cl (PPACK). The PPACK-bound structure of W215A/E217A is identical to the structure of the PPACK-bound slow form of thrombin. On the other hand, the structure of the free form reveals a collapse of the 215–217 strand that crushes the primary specificity pocket. The collapse results from abrogation of the stacking interaction between Phe-227 and Trp-215 and the polar interactions of Glu-217 with Thr-172 and Lys-224. Other notable changes are a rotation of the carboxylate group of Asp-189, breakage of the H-bond between the catalytic residues Ser-195 and His-57, breakage of the ion pair between Asp-222 and Arg-187, and significant disorder in the 186- and 220-loops that define the Na (cid:1) site. These findings explain the impaired catalytic activity of W215A/E217A

Thrombin possesses a paradoxical combination of procoagulant and anticoagulant roles (1). The procoagulant role involves the cleavage of fibrinogen and the platelet receptor PAR1, leading, respectively, to fibrin polymerization and platelet ag-gregation (2). The anticoagulant role unfolds on interaction with thrombomodulin and activation of protein C that eventually inhibits thrombin generation (3). The dual role of thrombin has long raised interest in dissociating its procoagulant and anticoagulant activities (4,5). These efforts have culminated in the design of anticoagulant thrombin mutants that are capable of activating protein C but show significantly reduced activity toward fibrinogen and PAR1 both in vitro and in vivo (6 -8). The mutant W215A/E217A (WE) 1 is by far the most potent anticoagulant thrombin engineered to date (8) and shows a safe and potent anticoagulant/antithrombotic profile in vivo (9). The mutant has drastically reduced (4 -5 orders of magnitude) catalytic activity toward all thrombin substrates, whether chromogenic or natural. However, in the presence of thrombomodulin, the activity of the mutant toward protein C is restored to a level comparable with that of the wild type (8). As a result, WE can circulate in the blood for hours, acting to increase the concentration of the anticoagulant-activated protein C without eliciting any significant fibrinogen clotting or platelet aggregation (9). Because of the importance of WE as a potential treatment for fatal or debilitating thrombo-occlusive events such as myocardial infarction and ischemic stroke, there is a compelling reason to characterize its structural properties. Here we present the structures of WE free and bound to the active site inhibitor H-D-Phe-Pro-Arg-CH 2 Cl (PPACK). The structures offer precious insights into the molecular basis of WE function, which also bear on studies of substrate recognition by thrombin and proteases in general.

MATERIALS AND METHODS
The mutant WE was expressed, purified, and tested for activity as described previously (8). The mutant was concentrated to 10 mg/ml in 50 mM choline chloride, 20 mM MES, pH 6.0. For crystallization with PPACK, the mutant was mixed with the inhibitor at a molar ratio of 1:15 and incubated at room temperature for 1 h. Crystallization was achieved at 25°C by vapor diffusion, with each crystallization well containing 500 l of reservoir solution. Equal volumes of the protein sample and reservoir solution (1 l each) were mixed to prepare the hanging drops. The reservoir solution consisted of 18% polyethylene glycol 8000, 0.2 M zinc acetate, and 0.1 M sodium cacodylate, pH 6.5. The reservoir solution for the free form of WE was composed of 0.1 M CAPS, 0.2 M lithium sulfate, 0.12 M sodium dihydrogen phosphate, and 0.53 M dipotassium hydrogen phosphate, pH 7.9. Diffraction quality crystals of both PPACK-inhibited and free WE were grown within 2 weeks. Crystals were cryoprotected in a solution similar to the reservoir solution but containing 25% glycerol prior to flash freezing. X-ray diffraction data for the WE-PPACK complex were recorded on an R axis image plate detector and processed and scaled with Denzo and Scalepack (10). * This work was supported in part by National Institutes of Health Research Grants HL49413, HL58141, and HL73813. 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 U.S.C. Section 1734 solely to indicate this fact.
The Data for the free form were collected at the Advanced Photon Source (beamline Biocars 14-BMC, Argonne National Laboratory) and processed using the HKL2000 software package (11). The WE-PPACK crystal was orthorhombic of space group P2 1 2 1 2 1 and contained one molecule per asymmetric unit. The free WE crystal was cubic of space group P2 1 3 and contained two molecules per asymmetric unit. Both structures were solved by molecular replacement using the coordinates of the thrombin-PPACK complex (12) as a search model and the program package Crystallography NMR System (13). Crystallographic refinement was carried out by simulated annealing and conjugated gradient minimization using Crystallography NMR System, and model building was performed with the program "O" (14). The autolysis loop could not be resolved in either structure. In the free WE structure, part of the 220-loop in the first monomer and a portion of the 186-loop in the second monomer were not included in the model because of their weak electron density. This observation is notable and underscores disorder in two critical loop regions of the molecule that are always well ordered in the wild type (12,15). Except for this difference, the two monomers in the asymmetric units could be refined without constraints to nearly identical conformations, showing a r.m.s. deviation of 0.42 Å over 270 equivalent C␣ atoms and no regions of significant difference in backbone configuration. Refinement carried out with noncrystallographic symmetry restraints produced essentially the same results, with no improvement in R cryst and R free . Monomer B, with the 220-loop intact, was chosen for structural comparisons. Weak densities were also detected in the N-and C-terminal regions, which were not included in subsequent refinement. Crystal contacts involved the exosite I region of one monomer and exosite II region of the second monomer in which the closest distance was 6.0 Å. The contacts did not involve any of the regions showing significant structural changes compared with the PPACK-bound form. The two monomers of free WE in the asymmetric unit make lattice contact with seven molecules related crystallographically, but the closest contact to the mutated residues 215 and 217 in either monomer is a distance of approximately 25 Å. Water molecules were added in the final stage of the refinement process. They were subjected to visual inspection to check their positioning in electron density and allowed to refine freely. Water molecules with a temperature factor (B factor) Ͼ80 Å 2 were excluded from subsequent refinement. Structural comparisons were computed using LSQMAN (16). The final refinement and model quality statistics are presented in Table I. Coordinates of the structures of the free and PPACK-bound forms of WE have been deposited in the Protein Data Bank (accession codes 1TQ0 for free WE and 1TQ7 for WE-PPACK).

RESULTS
The backbone traces of the structures of WE free or bound to PPACK are shown in Fig. 1. The structure of WE bound to the active site inhibitor PPACK is practically identical to that of the wild type (12,15) and especially to the PPACK-bound slow form (SL) reported recently (15). The autolysis loop is quite disordered, as is usually found in thrombin structures at this resolution (2.4 Å). Remarkably, the backbone region of the 215-217 strand hosting the mutated residues 215 and 217 retains the wild-type architecture of the SL structure (Fig. 2). Valence screening of the solvent molecules using WASP (17) returns no values Ͼ0.5 valence units for possible bound Na ϩ . Close examination of the Na ϩ -binding site of WE-PPACK reveals no evidence of bound Na ϩ , even though the complex was crystallized in the presence of 0.1 M sodium cacodylate. Remarkably, the water molecules in this region overlap 1:1 with those identified in the SL structure, recently solved at a resolution of 1.55 Å (15). These findings are consistent with the lack of Na ϩ binding and allosteric transduction in the WE mutant (8).
The structure of free WE shows notable changes relative to the wild type in its free form (15) and provides a molecular basis for understanding the devastating effects of the Ala replacement of Trp-215 and Glu-217 on substrate recognition. Because of the modest resolution of the crystal structure (2.8 Å), the autolysis loop is not visible nor are water molecules in the primary specificity pocket, the Na ϩ site, or the active site region. Also subject to considerable disorder are the 186-and 220-loops that define the Na ϩ site. The 220-loop is ordered only in one monomer in the asymmetric unit, whereas the 186-loop is ordered only in the other monomer. Disorder in these loops has never been reported in Ͼ150 structures of thrombin deposited in the Protein Data Bank. The disorder is a result of the WE mutation and is likely exacerbated by the low resolution of the structure. As a consequence of the disorder in the 186-and 220-loops, the ion pair between Asp-222 and Arg-187 cannot be defined. Formation of this ion pair is a hallmark of the Na ϩinduced conformational transition of thrombin from the slow to the fast form (15). In contrast to the disorder documented in the 186-and 220-loops, well defined density exists for other important regions of the enzyme. The 215-217 strand hosting the replaced residues Trp-215 and Glu-217 collapses into the primary specificity site and obliterates access of the Arg of substrate in the S1 cavity (Fig. 3). The collapse leaves behind Phe-227, normally in stacking interaction with Trp-215, and causes a shift in the position of its benzene ring. Contributing to the collapse also is the abrogation of important polar interactions of Glu-217 with Thr-172 and Lys-224, which stabilize the fast form of thrombin (15). The WE mutant shows a 35,000fold drop in the k cat /K m for the hydrolysis of the chromogenic substrate H-D-Phe-Pro-Arg-p-nitroanilide (FPR), which is the cleavable analog of PPACK (8). This drastic perturbation of substrate recognition is explained by the collapse of the 215-217 strand that comes in close contact (within 3 Å) with the 189 -192 strand. Coupled to this change, the carboxylate group of Asp-189 in the primary specificity pocket reorients almost 90°and aligns parallel to the backbone of the 189 -192 strand instead of being perpendicular to it as in the wild type. The orientation of Asp-189 precludes ionic interaction with the guanidinium group of the incoming Arg of substrate. The side chain of Glu-192 is oriented as in the PPACK-inhibited form of WE and is in the same orientation as the PPACK-bound fast form (FL) of the wild type (15). The catalytic Ser-195, on the other hand, experiences a main chain movement of 0.5 Å and a rotation in its side chain of approximately 70°relative to the PPACK-inhibited form, causing a breakage of the H-bond with the catalytic His-57. This orientation is similar to that seen in the free slow form(s) of the wild type (15). Some of the structural changes in the free form of WE are reminiscent of those observed in the S structure of the wild type reported recently (15). Specifically, the absence of the Asp-222: Arg-187 ion pair, the breakage of the H-bond interaction between Ser-195 and His-57, and the reorientation of the side chain of Asp-189 are also seen in the S structure. It is possible that these changes are a direct manifestation of the lack of Na ϩ binding in the WE mutant. In addition to these structural signatures that pertain to the slow form of thrombin, the structure of WE contains features that are entirely the result of the double mutation. The replacement of Trp-215 and Glu-217 with Ala causes a detachment of the 215-217 strand that crushes the primary specificity pocket. The movement likely causes disorder in the 186-and 220-loops that define much of the Na ϩ site and the entry point of the cation (18). These additional changes explain why the catalytic properties of WE are com-promised Ͼ1,000-fold relative to those of the slow form of the wild type. DISCUSSION The WE mutant of thrombin is currently being evaluated in preclinical studies as an effective anticoagulant and antithrombotic drug (9). A remarkable property of this mutant is that it shows very little catalytic activity toward chromogenic or natural substrates of thrombin. The k cat /K m values for substrate hydrolysis are compromised 1000-fold for PAR1, 20,000-fold for fibrinogen, and up to 35,000-fold for the chromogenic substrate FPR (8). The k cat /K m value for the hydrolysis of protein C in the absence of thrombomodulin is Ͻ1 M Ϫ1 s Ϫ1 (8). The mutation involves two residues that play a critical role in substrate recognition. Trp-215 is highly conserved in the entire realm of serine proteases, and both Trp-215 and Glu-217 are absolutely conserved in thrombins from hagfish to human (19). The crystal structure of the free form of WE provides a striking explanation for the drastic drop in catalytic activity. Mutation of FIG. 2. Stereo view of the active site, primary specificity pocket, and Na ؉ -binding site of the thrombin mutant WE. The PPACKinhibited WE structure (blue) is superimposed to the SL structure (red) of the wild type (15). Notwithstanding the drastic difference in atomic resolution (2.4 Å for WE-PPACK and 1.55 Å for SL), the two structures are remarkably similar overall (r.m.s. deviation ϭ 0.4 Å). There is no evidence of bound Na ϩ in the WE-PPACK structure, and there is a notable 1:1 correspondence for the water molecules in the Na ϩ site between the two structures. Relevant side chains are labeled. In the WE structure, the side chain of Lys-224 moves away from residue 217 because of the E217A mutation.
Trp-215 abrogates the important stacking interaction with the benzene ring of Phe-227, and mutation of Glu-217 abrogates polar interactions of its side chain with Thr-172 and Lys-224. As a result, the 215-217 strand detaches from its position as a wall of the primary specificity pocket and crashes on the neighboring 189 -192 strand. This prevents the Arg at the P1 position of incoming substrate to access the side chain of Asp-189 at the bottom of the S1 site. The compromised catalytic activity of WE is advantageous in vivo because it ensures that the mutant can circulate in the blood without causing fibrinogen clotting or platelet aggregation. When the mutant WE interacts with thrombomodulin, however, its activity toward protein C increases 60,000-fold, to a level comparable with that of the wild type (8). The mutant thus acts as a molecular switch that turns on its catalytic activity only in the presence of thrombomodulin and protein C. This is the basis of its potent anticoagulant effect in vivo (9).
What is the mechanism that rescues the catalytic activity of WE toward protein C in the presence of thrombomodulin? It is tempting to speculate that thrombomodulin corrects the collapse of the 215-217 strand and restores access to the primary specificity pocket. If so, thrombomodulin should be able to significantly enhance the catalytic activity of WE toward FPR. However, the k cat /K m value for FPR hydrolysis by WE improves only 3-fold in the presence of saturating (100 nM) concentrations of thrombomodulin (Fig. 4). The 3-fold increase is also observed in the presence of saturating (100 M) concentrations of hirugen (Fig. 4). These effects are analogous to those observed with the wild type and argue against thrombomodulin being an allosteric effector of thrombin (20). It is therefore unlikely that thrombomodulin induces significant changes in the primary specificity pocket of WE. An alternative hypothesis is that thrombomodulin acts as a scaffold for presenting protein C to the thrombin active site in a correct orientation for cleavage (20,21). This hypothesis explains the effects of thrombomodulin in the wild type and does not call for conformational changes in the mutant WE, consistent with the data in Fig. 4; however, the hypothesis fails to explain how the collapsed active site of WE is functionally restored to interact with protein C. A plausible scenario is that formation of the ternary complex WE-thrombomodulin-protein C causes a large induced fit transition in the active site of the enzyme that cannot take place when either protein C or thrombomodulin are separately bound. The concerted action of protein C and thrombomodulin turns WE into a molecular switch that only delivers its anticoagulant function.
An important implication of the results reported in this study is that no relevant information on the molecular properties of WE would have emerged from the structure of the mutant bound to PPACK. The thermodynamic definition of a binding constant in the ground state or the transition state involves the chemical potentials of the bound and free species; hence, structural analysis of substrate recognition cannot be based exclusively on information from the enzyme-substrate complex. This becomes even more obvious in the case of mutant enzymes, in which the effect of the structural perturbation may be confined to the free form. A direct demonstration of this possibility is offered by the thrombin mutant Y225I. Although its catalytic activity toward FPR and fibrinogen is compromised Ͼ10,000-fold, its crystal structure bound to PPACK is identical to that of PPACK-inhibited wild type (15,22). Crystallization of thrombin in its free form is beset by many difficulties, including the need to prevent autodigestion. In a recent , is superimposed to the PPACK-inhibited form (blue). The 215-217 strand in the free form collapses into the primary specificity pocket and clashes with the Arg residue at the P1 position of PPACK (green). The r.m.s. deviation between free WE and WE-PPACK in the 215-221 segment is 2.5 Å. The r.m.s. deviation between the two monomers in the asymmetric unit of the free WE structure in the same segment is 0.5 Å. Also notable is the rotation of the side chain of Asp-189 in the free form that aligns almost parallel to the backbone as well as the shift in the side chain of Ser-195 that moves away from its H-bonding partner His-57. study (23), removal of the site of autocatalytic cleavage in exosite I ensured crystallization of the free form of thrombin for the first time. In the case of thrombin mutants with severely compromised catalytic activity, there is in principle no need to remove this site of cleavage. The crystal structure of the anticoagulant mutant E217K has recently been solved in the free form (24). As for the WE mutant presented in this study, the structure of free E217K reveals changes in the active site region, resulting from the drastic charge reversal at residue 217, that explain the perturbed functional properties of the mutant. The structure of E217K bound to PPACK was not presented in the study, but it would have been a critical comparison with those of Y225I-PPACK (22) and WE-PPACK reported here.
The recent structure of thrombin in its free slow form (15, 23) has fostered new interest in the crystallographic investigation of the molecular basis of thrombin function and regulation. We have now demonstrated that it is possible, and indeed highly desirable, to crystallize thrombin mutants with compromised catalytic activity in both their free and PPACK-bound forms. The information on the free form is highly relevant to the biochemical properties of such mutants. Comparison of the free and bound forms affords a deeper understanding of the molecular basis of substrate recognition, which has a bearing on the study of protease specificity in general.