Important Role of the Cys-191–Cys-220 Disulfide Bond in Thrombin Function and Allostery*

Little is known on the role of disulfide bonds in the catalytic domain of serine proteases. The Cys-191–Cys-220 disulfide bond is located between the 190 strand leading to the oxyanion hole and the 220-loop that contributes to the architecture of the primary specificity pocket and the Na+ binding site in allosteric proteases. Removal of this bond in thrombin produces an ∼100-fold loss of activity toward several chromogenic and natural substrates carrying Arg or Lys at P1. Na+ activation is compromised, and no fluorescence change can be detected in response to Na+ binding. A 1.54-Å resolution structure of the C191A/C220A mutant in the free form reveals a conformation similar to the Na+-free slow form of wild type. The lack of disulfide bond exposes the side chain of Asp-189 to solvent, flips the backbone O atom of Gly-219, and generates disorder in portions of the 186 and 220 loops defining the Na+ site. This conformation, featuring perturbation of the Na+ site but with the active site accessible to substrate, offers a possible representation of the recently identified E* form of thrombin. Disorder in the 186 and 220 loops and the flip of Gly-219 are corrected by the active site inhibitor H-D-Phe-Pro-Arg-CH2Cl, as revealed by the 1.8-Å resolution structure of the complex. We conclude that the Cys-191–Cys-220 disulfide bond confers stability to the primary specificity pocket by shielding Asp-189 from the solvent and orients the backbone O atom of Gly-219 for optimal substrate binding. In addition, the disulfide bond stabilizes the 186 and 220 loops that are critical for Na+ binding and activation.

Thrombin is a Na ϩ -activated, allosteric serine protease involved in blood clotting (1,2) and is composed of two polypeptide chains of 36 (A chain) and 259 (B chain) residues covalently linked through the Cys-1-Cys-122 disulfide bond (3,4). The shorter A chain runs in the back of the B chain that has the typical fold of serine proteases, with two six-stranded ␤-barrels of similar structure that pack together asymmetrically to accommodate at their interface the residues of the catalytic triad His-57, Asp-102, and Ser-195 (5). The B chain is stabilized by three disulfide bonds, Cys-42-Cys-58, Cys-168 -Cys-182, and Cys-191-Cys-220, that connect contiguous strands shaping the Na ϩ binding site, the primary specificity pocket, and the active site. These bonds are highly conserved among serine proteases, but the Cys-191-Cys-220 disulfide bond is absent in cathepsins and granzymes. Previous studies using reduction with dithiothreitol have addressed the role of disulfide bonds in the mechanism of thrombin unfolding (6). The role of disulfide bonds in thrombin function and allostery, however, remains unexplored. Indeed, remarkably little is known on the role of conserved disulfide bonds for serine proteases in general (7), and no structural information on relevant mutants lacking one or more such bonds is currently available.
Previous mutagenesis studies on trypsin have addressed the role of the Cys-191-Cys-220 bond and offered conflicting reports. In one case elimination of the disulfide bond with the C191A/C220A mutation was found to be inconsequential on the function, structure ,and stability of the enzyme and affected k cat but not K m for the hydrolysis of chromogenic amide substrates (8). In another case removal of the disulfide bond with the same C191A/C220A mutation was associated with a loss of catalytic function toward chromogenic amide substrates due mostly to increased K m , with a concomitant slight shift in substrate specificity from trypsin-like to chymotrypsin-like (9). Given the similarity of reagents and experimental conditions used in these previous studies, it is difficult to rationalize the origin of the discrepancy or to offer conclusive answers about the role of the Cys-191-Cys-220 disulfide bond. Studies on tissue-type plasminogen activator have shown that the Cys-136 -Cys-201 bond, which is not present in Na ϩ -activated allosteric proteases, is inconsequential on function (10). The Cys-1-Cys-122 bond in thrombin was also found to be dispensable for function (11,12), although several naturally occurring mutations of prothrombin involve residues of the A chain (13)(14)(15)(16) and are associated with severe bleeding. Because thrombin is a Na ϩ -activated enzyme, unlike trypsin (17,18), the possibility exists that the conserved disulfide bonds play a role in the allosteric properties of the enzyme that control its numerous and opposing functions in vivo (19 -21).
The Cys-42-Cys-58 disulfide bond anchors the catalytic His-57 and may influence the response to Na ϩ at the level of the active site. The Cys-168 -Cys-182 disulfide bond anchors the 170-strand that hosts Thr-172, one of the most important residues linked to Na ϩ binding (22), and residues that stabilize the environment of Tyr-225 and part of the Na ϩ binding pocket. The Cys-191-Cys-220 disulfide bond deserves special atten-tion because it anchors the 220-loop to the 190-strand and ensures communication between the Na ϩ site and the oxyanion hole (23). In addition, the bond provides integrity to the primary specificity pocket by shielding Asp-189 from the solvent and burying the RGD sequence of which Asp-189 is a member (24). Here we demonstrate that this disulfide bond plays an important role in thrombin function and allostery, and We provide the first crystal structure that directly documents the molecular defects associated with its removal.

MATERIALS AND METHODS
The thrombin mutant C191A/C220A was expressed, purified, and tested for activity as described previously (22,25). The mutant was concentrated to 5 mg/ml in 50 mM choline chloride, 20 mM Tris, pH 7.5, at 25°C. The k cat /K m for the hydrolysis of the chromogenic substrates H-D-Phe-Pro-Arg-p-nitroanilide (FPR), 2 H-D-Phe-Pro-Lys-p-nitroanilide (FPK), and H-D-Phe-Pro-Phe-p-nitroanilide (FPF) was determined as reported (26 -28) under experimental conditions of 5 mM Tris, 0.1% PEG8000 at 25°C in the presence of 200 mM NaCl (fast form) or choline chloride (slow form). Interactions with the physiologic substrates fibrinogen, PAR1, PAR4, and protein C and the inhibitor antithrombin were determined as reported elsewhere (22, 29 -32) under experimental conditions of 5 mM Tris, 0.1% PEG8000, 145 mM NaCl, pH 7.4, at 37°C. Activation of protein C was studied in the presence of 5 mM CaCl 2 with or without 100 nM thrombomodulin. Antithrombin inhibition was studied with or without 0.25 units/ml of heparin, which maximized the k on rate.
For crystallization with H-D-Phe-Pro-Arg-CH 2 Cl (PPACK), the mutant was mixed with the inhibitor at a molar ratio of 1:10 and incubated at room temperature for 1 h. Crystals were obtained using the hanging drop vapor-diffusion method at 22°C, with each crystallization well containing 1 ml of reservoir solution. Equal volumes of the protein sample and reservoir solution (1 l) were mixed to prepare the hanging drops. The reservoir solution contained the following: for the free mutant: 25% PEG4000, 0.2 M lithium sulfate, 0.1 M Tris, pH 8.5; for the PPACK-bound mutant: 20% PEG3350, 0.2 M zinc acetate. Diffraction quality crystals (ϳ0.75 ϫ 0.1 ϫ 0.1 mm) were grown in both cases within 2 weeks. Crystals were cryoprotected in a solution similar to its reservoir buffer but containing additional 25% glycerol before flash-freezing. X-ray data were collected on an ADSC Quantum-315 CCD detector of the BIOCARS Beamline 14BMC at the Advanced Photon Source, Argonne National Laboratories (Argonne, IL). Data processing including indexing, integration, and scaling was performed using the HKL2000 package (33). Both the free and PPACK-bound crystals were orthorhombic and contained one molecule per asymmetric unit. The space group was P2 1 2 1 2 for the free form and P2 1 2 1 2 1 for the PPACK-bound form. Both structures were solved by molecular replacement using the coordinates of the PPACKbound slow form of human thrombin 1SHH (22) as the starting model. Refinement and electron density map generation were carried out using CNS (35), and 5% of the reflections were selected randomly and set aside as a test set for cross-validation (35). Model building and analysis of the structure were carried out with the program O (36). Additional refinement with the program CCP4 was performed for the PPACK-bound structure. Results of data collection and refinement are summarized in Table 1. Coordinates of the structures of the C191A/C220A thrombin mutant in the free and PPACK-bound forms have been deposited in the Protein Data Bank (accession codes: 2PGB for the free form and 2PGQ for the PPACK-bound form).

RESULTS
Removal of the Cys-191-Cys-220 disulfide bond has profound effects on the interaction of thrombin with chromogenic ( Table 2) and natural (Table 3) substrates. Activity toward the highly specific substrate FPR drops ϳ50-fold in the slow and fast forms due mainly to an increase in K m . Interestingly, the k cat increases relative to wild type especially in the slow form. FPR binding is perturbed in the C191A/C220A mutant, but once docking into the active site is secured, hydrolysis proceeds at a rate that is even faster than in the wild type. The less specific substrate FPK replaces Arg at P1 with Lys (37). Activity toward this substrate is also compromised due to increased K m and reduced binding but without the increase in k cat observed for FPR. Finally, the chymotrypsin-specific substrate FPF, for which thrombin shows considerable activity unlike trypsin (27), experiences only a modest drop in k cat /K m in the slow form and a more pronounced drop in the fast form. The limited solubility of this substrate prevents independent assessment of the indi- vidual Michaelis-Menten parameters. In all cases, the Na ϩ effect on substrate hydrolysis is significantly reduced compared with wild type and almost vanishes for FPF. The results suggest that removal of the Cys-191-Cys-220 disulfide bond perturbs the environment of the primary specificity pocket probed by the positively charged P1 residue of substrate. The library of chromogenic substrates FPR, FPK, and FPF provides valuable insights on the effect of P1 substitutions on the perturbed functional properties of the double mutation C191A/C220A.
Reduced Na ϩ activation can originate from compromised Na ϩ binding or reduced transduction of binding into enhanced catalytic activity. Direct measurements of Na ϩ binding help distinguish between the two alternatives (5,22). Attempts to monitor Na ϩ binding directly from stopped-slow fluorescence measurements (38) failed to produce any significant spectral change (data not shown) for the C191A/C220A mutant. Because all nine Trp residues of thrombin contribute to the signal linked to Na ϩ binding and the residues are distributed over the entire surface of the enzyme (38), it is unlikely that the C191A/C220A mutation affects the structure of the enzyme as a whole in a way that silences all fluorophores. Most likely, Na ϩ binding to the mutant is too weak to detect in the range of NaCl concentration explored up to 400 mM under conditions where the wild type binds Na ϩ with an affinity K A ϭ 160 M Ϫ1 (38).
Perturbation induced by the C191A/C220A mutation extends to the interaction with physiologic substrates ( Table 3). Release of fibrinopeptides A (FpA) and B (FpB) from fibrinogen is reduced 150-and 290-fold, respectively. PAR1 cleavage is compromised ϳ100-fold, and specificity toward PAR4 drops ϳ600-fold. The effect on protein C activation is less pronounced compared with the procoagulant (fibrinogen) and prothrombotic/signaling (PAR1, PAR4) substrates, as typically observed for mutations that affect Na ϩ binding and activation and stabilize thrombin in the anticoagulant slow form (5). However, the drop in k cat /K m for protein C activation in the presence of thrombomodulin is too large (Ͼ15-fold) to make the C191A/ C220A mutant a compelling candidate as an anticoagulant thrombin mutant for studies in vivo (39 -41). Perturbation of fibrinogen, PAR1, and PAR4 cleavage is comparable with that seen for FPR (Table 2), vouching for a molecular origin of the effect limited to the primary specificity site and possibly to the Na ϩ site. The conclusion is supported by the observation that inhibition by antithrombin is reduced ϳ400-fold with or without heparin. The inhibitor is known to make extensive contacts with the 186 and 220 loops of the Na ϩ site in addition to the primary specificity pocket (42).
Functional studies point to significant perturbation of the primary specificity pocket and the Na ϩ site upon removal of the Cys-191-Cys-220 disulfide bond. Direct assessment of the perturbation comes from the x-ray crystal structures of the mutant C191A/C220A in the free (CCF) and PPACK-bound (CCB) forms. The resolution of CCF and CCB is very high, 1.54 and 1.8 Å, respectively, and both structures contain a single molecule in the asymmetric unit. Paradoxically, CCF crystallized at higher resolution than the PPACK-inhibited form CCB and yielded by far the highest resolution structure ever reported for free thrombin and one of the highest resolution structures ever reported for thrombin in general.
CCF and CCB are very similar (root mean square 0.42 Å) (Fig.  1). The structure of CCB is practically identical to that of wildtype thrombin bound to PPACK, with r.m.s.d. of 0.33 and 0.32 Å versus the PPACK-bound slow (SL) and fast (FL) forms (22), respectively. The whole backbone of the autolysis loop is well defined in the electron density map and assumes a conformation similar to that seen in the SL structure. All side chains are clearly resolved with the exception of Lys-145, Trp-148, and Lys-149e. The backbones of the 186 and 220 loops around the sites of mutation are also well defined in the density map and do not differ significantly from the conformations observed in SL (Fig. 2). The C␤ atoms of Ala-191 and Ala-220 in CCB superimpose well with the C␤ atoms of Cys-191 and Cys-220 in the wild-type structures FL and SL. In summary, the presence of PPACK produces a conformation that shows no obvious departure from those of the slow and fast forms of wild-type bound to the same active site inhibitor. The structure of CCB fails to explain the severe impairment of substrate recognition documented by functional studies (Tables 2 and 3).

JOURNAL OF BIOLOGICAL CHEMISTRY 27167
Our recent work on the W215A/E217A mutant (43) demonstrated that PPACK has a stabilizing structural effect on thrombin, even in the presence of mutations that produce large perturbation of function. We, therefore, endeavored to obtain a structure of the C191A/C220A mutant in the free form, devoid of potential bias caused by the presence of an active site inhibitor. As seen for W215A/E217A (43), the structure of CCF shows some significant changes relative to CCB that help explain the perturbed functional properties of the mutant. The overall fold of CCF is very similar to wild type, with r.m.s.d. of 0.34 and 0.38 Å relative to the free slow (S) and fast (F) forms (22), respectively. CCF folds in an active conformation, with access to the active site and primary specificity pocket unperturbed. However, portions of the 186 and 220 loops feature significant disorder (Fig. 2) that is not seen in the CCB structure. Specifically, the side chains of residues Glu-186b, Lys-186d, and Arg-187 are completely disordered, and part of the backbone of Gly-186c is not visible in the density map. In the 220 loop the side chains of Glu-217, Asp-221, Arg-221a, Asp-222, and Lys-224 are highly disordered, and only the backbone O atom of Gly-219 is visible in the density map. This atom makes an important H-bond with the guanidinium group of Arg at P1 of substrate (3,22) but is flipped in the CCF structure. The flip may contribute to the reduced electrostatic coupling with Arg or Lys residues at P1. Disorder in these loops makes it impossible to locate the ion pair between Arg-187 and Asp-222, which is a hallmark of the fast form of thrombin (22,44). The lack of density in this region is in contrast with the extremely high (1.54 Å) resolution of the structure and is testimony to an intrinsic conformational property of the C191A/C220A mutant. Also notable is the disorder in the autolysis loop that is corrected by the presence of PPACK (Fig. 1). On the other hand, well defined density exists at the site of mutation Ala-220. The side chain of Ala-191 superimposes well with the C␤ of Cys-191 in the free slow (S) and fast (F) form structures (22), but the C␤ of Ala-220 is 1.1 Å away from the corresponding atom of Cys-220 in those structures. The deviation exposes to solvent the side chain of Asp-189 in the primary specificity pocket, likely increasing its pK a with a resulting weakening of the electrostatic coupling with the positively charged moieties of Arg or Lys at P1 of substrate. The increased solvent exposure of Asp-189 and the flip of the backbone O atom of Gly-219 explain the increase in K m for FPR and FPK hydrolysis (Table 2), the reduced k cat /K m for physiologic substrates, and the reduced k on for antithrombin binding (Table 3). Because the Phe group at P1 of FPF does not engage Asp-189 or Gly-219 in electrostatic coupling, no perturbation of k cat /K m is observed relative to the slow form of wild type. The change relative to the k cat /K m value in the fast form is due to the stabilizing effect of Na ϩ on the side chain of Asp-189 (22), which is not produced in the C191A/ C220A mutant because of the perturbed environment of the Na ϩ site. Na ϩ binding has no effect on the orientation of the backbone O atom of Gly-219 (22). Other signatures of the CCF structure reveal a conformation similar to the slow form S of wild type (22). The side chain of the catalytic Ser-195 is oriented as in the S structure and is not H-bonded to the catalytic His-57 (Fig. 2). The distance between N⑀2 of His-57 and O␥ of Ser-195 is 3.58 Å in CCF, similar to 3.70 Å in the S structure and significantly longer than 3.09 Å in the fast form F structure (22). Another noteworthy feature of the CCF structure is the paucity of water molecules within the active site and primary specificity pocket regions. Again, this is all the more significant in view of the high resolution of the structure and the fact that a well organized network of water molecules does exist in these regions in the F structure but not in the S structure of wild type (22).

DISCUSSION
The thrombin mutant C191A/C220A reveals several features of relevance to thrombin and serine proteases in general. Removal of the Cys-191-Cys-220 disulfide bond compromises ϳ100-fold the catalytic activity of the enzyme and reduces the Na ϩ effect on substrate hydrolysis. The effect is seen to a similar extent on all substrates used, chromogenic or physiologic, except for FPF that does not engage Asp-189 and Gly-219 in electrostatic coupling. The origin of the structural perturbation is to be found in the primary specificity pocket and the adjacent Na ϩ site.
Previous functional studies on the trypsin mutant C191A/ C220A (8,9) have failed to provide conclusive answers on the role of the Cys-191-Cys-220 disulfide bond. In one report the disulfide bond was found not to be essential for the function, structure, and stability of the enzyme, and its removal affected k cat but not K m for the hydrolysis of chromogenic amide substrates (8). In a subsequent report, removal of the disulfide bond increased K m for chromogenic amide substrates and slightly shifted substrate specificity from trypsin-like to chymotrypsinlike (9). The results reported here on the thrombin mutant C191A/C220A show that the chromogenic substrate FPR is cleaved with higher k cat compared with wild type but with significantly increased K m . Hydrolysis of FPK shows reduced k cat and K m . In the case of FPF, no perturbation of hydrolysis is observed in the absence of Na ϩ . In fact, when Na ϩ is removed from the buffer, the k cat /K m values for FPR and FPF hydrolysis differ 27-fold for the mutant, as opposed to ϳ500-fold in the wild type. Significantly, the mutant is more specific for FPF than FPK, whereas the wild type prefers FPK ϳ30-fold over FPF. Hence, removal of the Cys-191-Cys-220 disulfide bond with the C191A/C220A mutation in a Na ϩ -activated protease like thrombin recapitulates many of the contrasting features observed with trypsin (8,9), including the intriguing shift in specificity toward chymotrypsin-like substrates. Dominant involvement of the primary specificity pocket is confirmed by perturbation of fibrinogen and PAR cleavage that is affected to the same extent as FPR and FPK. Protein C activation is compromised to lesser extent, presumably because this substrate engages thrombin with minimal interactions in the presence of thrombomodulin (31). Also noteworthy is the large effect on antithrombin binding, which suggests perturbation of the 186 and 220 loops in addition to the primary specificity pocket.
Evidence that the Cys-191-Cys-220 disulfide bond stabilizes the Na ϩ binding environment in addition to the primary spec-ificity pocket comes from the observation of highly perturbed Na ϩ binding and activation in the C191A/C220A mutant. No fluorescence change could be detected up to 400 mM NaCl that would support direct Na ϩ interaction with the enzyme, and the extent of Na ϩ activation was reduced (FPR, FPK) or abolished (FPF) compared with wild type (Table 2). Most likely, these properties are the result of a binding interaction too weak to measure.
The molecular origin of perturbation due to removal of the Cys-191-Cys-220 disulfide bond is revealed by the x-ray crystal structures of the free (CCF) and PPACK-inhibited (CCB) forms of the C191A/C220A mutant. The structure of CCF shows disorder in the autolysis loop and portions of the 186 and 220 loops defining the Na ϩ site. The disorder is not seen in the structure of CCB. The presence of the active site inhibitor corrects the structural perturbation induced by the mutation and produces an architecture similar to that of wild type, as recently documented for the mutant W215A/E217A (43).
The structure of CCF is among the highest resolution structures ever reported for thrombin. The absence of steric constraints for substrate binding to the active site makes this structure particularly relevant to thrombin allostery. There are three dominant species of thrombin, two in the Na ϩ -free (E and E*) and one in the Na ϩ -bound (E:Na ϩ ) conformations (5,38). E and E:Na ϩ correspond to the slow (E) and fast (E:Na ϩ ) forms originally defined by Wells and Di Cera (1), and their structures have been solved recently (22). The structure of CCF contains features observed in the slow form of thrombin E, whereas others are unique to the mutant. The H-bond between His-57 and Ser-195 is broken, as in the slow form. However, removal of the disulfide bond causes disorder in the channel connecting the Na ϩ site to the active site where a number of water molecules are clearly detected in the wild type (22,45,46). The channel in CCF is shaped as in the wild-type but is conspicuously devoid of ordered water molecules. The lack of covalent bridging of the 220 and 190 strands affects accessibility of the channel and presumably increases the rate of exchange of water molecules between the channel and bulk solvent. That also explains the increased exposure of Asp-189 to solvent and its perturbed pK a , with a resulting weakening of the electrostatic coupling with positively charged groups (Arg and Lys) at the P1 position of substrate. The disorder in the 220 loop and in the side chains of residues in the 186 loop coupled with the inability to detect any  (22). Disorder in the side chains of residues in the 186-loop and around Glu-217 and Gly-219 in the CCF structure (A) is corrected by the presence of PPACK (stick model in green) in the CCB structure (B). Disorder in the Na ϩ binding site (186 and 220 loops) suggests that the conformation of CCF is unable to bind Na ϩ , in agreement with functional data on the mutant. spectral signature of Na ϩ binding suggest that C191A/C220A is stabilized in a conformation unable to bind Na ϩ but capable of binding substrate to the active site. E* differs from E insofar as it cannot bind Na ϩ and is negligibly populated under conditions of physiological interest (38). A previous spectroscopic study has speculated that E* may also be inactive toward substrates (47), and a recent crystal structure of the D102N mutant of thrombin has revealed a self-inhibited conformation with the Na ϩ site and active site occluded (48). However, the existence of inactive forms of thrombin is not supported by the mechanism of Na ϩ activation (1,5) or by the kinetic profile of substrate hydrolysis in the absence of Na ϩ . The structure of CCF offers a representation of E* as a conformation of thrombin that is active toward substrate much like the slow form E but is unable to bind Na ϩ due to disorder in the 186 and 220 loops. Ongoing kinetic studies will shed light on which of the currently available structures of E*, if any, reflects the properties of the enzyme in solution (5).