Molecular Basis for Specificity of the Extracytoplasmic Thioredoxin ResA*

ResA, an extracytoplasmic thioredoxin from Bacillus subtilis, acts in cytochrome c maturation by reducing the disulfide bond present in apocytochromes prior to covalent attachment of heme. This reaction is (and has to be) specific, as broad substrate specificity would result in unproductive shortcircuiting with the general oxidizing thioredoxin(s) present in the same compartment. Using mutational analysis and subsequent biochemical and structural characterization of active site variants, we show that reduced ResA displays unusually low reactivity at neutral pH, consistent with the observed high pKa values >8 for both active site cysteines. Residue Glu80 is shown to play a key role in controlling the acid-base properties of the active site. A model in which substrate binding dramatically enhances the reactivity of the active site cysteines is proposed to account for the specificity of the protein. Such a substratemediated activation mechanism is likely to have wide relevance for extracytoplasmic thioredoxins.

bonds in protein substrates as part of post-translational modification processes, e.g. E. coli CcmG and B. subtilis StoA (6,7). Some TDORs have an isomerase activity that enables them to rearrange disulfide bonds in protein/peptide substrates, e.g. E. coli DsbC (8,9).
Many TDORs consist of a five-stranded mixed ␤-sheet surrounded by four ␣-helices, an arrangement known as the thioredoxin fold, after the original member of the family (1,10). The two cysteine thiols of the thioredoxin TDOR active site motif, CXXC (in which X denotes an unspecified amino acid residue) are positioned at the terminus of an ␣-helix and are correctly orientated to reversibly cycle between reduced (thiol) and oxidized (disulfide) forms. The redox potential of the disulfide/dithiol couple correlates with the function of the protein; low and high potential TDORs act to reduce and oxidize their substrates, respectively (e.g. Refs. 11 and 12).
ResA is an extracytoplasmic thioredoxin-like TDOR from B. subtilis that is involved in cytochrome c maturation (13,14). Bacteria exhibit one of two distinct maturation pathways (Systems I and II) (15,16) and Gram-positive organisms such as B. subtilis have System II, which in this organism, consists of only three dedicated proteins, ResA, ResB, and ResC (13,17). Strains lacking ResA are deficient in c-type cytochromes. This is reversed in strains also lacking BdbD, a TDOR that catalyzes formation of disulfide bonds in the extracytoplasmic compartment of B. subtilis, indicating that ResA is involved in reduction of the apocytochrome prior to covalent heme attachment (13). Soluble ResA (lacking its single transmembrane segment anchor) was found to have a low redox potential and low activity in the insulin assay, indicating that it functions specifically as a reductase in cytochrome c maturation (13).
High resolution structures of soluble ResA in both oxidation states revealed a typical thioredoxin fold, with an additional N-terminal ␤-hairpin and a strand/helix insertion between strand ␤ 2 and helix ␣ 3 (14). Upon reduction, several stabilizing interactions are broken in the vicinity of the active site, whereas only one hydrogen bond, between the Cys 74 thiol and the backbone amide hydrogen of Gly 71 , is formed (note that the numbering of ResA residues in the original structure PDB file (1SU9) and article (14) were in error by Ϫ1. The active site thiol residues are Cys 74 and Cys 77 rather than Cys 73 and Cys 76 ). Overall this results in a destabilization of the reduced structure, relative to the oxidized, consistent with its low redox potential. In reduced ResA, the thiol of Cys 74 is significantly more solvent-exposed than that of Cys 77 , as found in other TDOR structures, but the separation between the S␥ atoms is unusually long. Present only in the reduced structure is a cavity lined by mainly hydrophobic residues (Phe 66 , Asn 68 , Pro 76 , Cys 77 , Phe 81 , Met 84 , Pro 141 , Thr 143 , Gly 158 , Thr 159 , and Met 160 ). At its base is Glu 80 , which is stabilized through hydrogen bonding interactions with Thr 159 and a water molecule that is also hydrogenbonded to the side chain of Cys 77 . The location of the cavity close to the active site led us to propose that it plays an important role in substrate recognition (14). Solution NMR experiments have recently confirmed that redox-related conformational changes occur in this region (18).
Here we present biochemical and structural studies of wildtype and variants of soluble ResA. By labeling with the environment-sensitive fluorescent probe badan and iodoacetate, we demonstrate that the thiol groups of Cys 74 and Cys 77 both have pK a values between 8 and 9. The unusually high pK a of Cys 74 is consistent with the protein's low reactivity at low to neutral pH values. In contrast to thioredoxin, at pH values above 8 both active site cysteines can be readily alkylated. Biochemical and structural studies of E80Q ResA reveal that Glu 80 plays a key role in controlling the acid/base properties of the active site. The data support a model for reduced ResA in which Glu 80 locks the active site thiols in an unreactive conformation, and substrate binding specifically unlocks the active site reactivity. Similar substrate-mediated activation mechanisms are likely to be relevant for many other extracytoplasmic TDORs involved in cytochrome c maturation and other post-translational modification pathways.
Wild-type and variant soluble ResA proteins were purified as previously described (13), except that 20 mM Tris, pH 8.0 was used instead of sodium phosphate, and a Superdex S75 gel filtration column (GE Healthcare) was used in place of a Sephacryl S100 column. The single cysteine variants showed a tendency to form disulfide-linked dimers and so 1 mM DTT was included in the buffer during the gel filtration step.
Thiol Reactivity Experiments-All proteins were prereduced by incubating with 1 mM DTT at 4°C overnight. DTT was removed by gel filtration. Under aerobic conditions, all variants remained reduced for at least 24 h following removal of DTT. For fluorescence labeling experiments, ResA and variants (final conc. 1 M) were added to a 13 M badan solution (Molecular Probes) (20) and fluorescence spectra recorded between 400 and 600 nm (ex. wavelength 390 nm) at 25°C using a PerkinElmer Life Sciences LS55 fluorescence spectrophotometer. Badan was in either a mixed buffer system containing ammonia, potassium acetate, MES, MOPS, and Tris (10 mM each) and 200 mM KCl (21), or single buffer solutions of 150 mM NaCl containing 50 mM MES (pH 5.5-7.0), Tris (pH 7.5-9.0) or CHES (pH 9.5-10.0). Equivalent results were obtained in each.
For carboxymethylation experiments, reduced protein (ϳ40 M) was incubated with 10 mM iodoacetate at room temperature. For kinetic experiments, 30-l samples were removed at different time points, and the reaction quenched by the addition of 10 l of native PAGE loading buffer containing 200 mM DTT. Modified and unmodified proteins were separated using native PAGE run in 5 mM Tris-glycine buffer, pH 8.0. ESI-MS was conducted on badan-modified ResA as previously described for ResA (13).
Determination of Cysteine Thiol pK a Values-The reaction of cysteine side chains with alkylating reagents is well established and occurs only with the ionized thiolate anion (22). Measurement of the rate of alkylation as a function of pH can be used to determine the pK a values of protein cysteine thiol groups (21,23,24), where the observed rate constant is proportional to the extent of thiol deprotonation at a given pH value (21). Reaction of wild-type ResA and variants with badan were carried out under pseudo-first order conditions (see above). Fluorescence data were fitted to a single exponential function to obtain an observed, pseudo-first order rate constant (k 0 ). Where necessary, a double exponential fit was used and the rate constant for the initial reaction was taken as k 0 . pK a values were determined by plotting k 0 values as a function of pH and fitting to an appropriate equation. For single pK a processes, Equation 1 was used, where k SH and k S Ϫ are the rate constants for the protonated and deprotonated forms, respectively (21). For two proton dissociation events, where the pK a values are well separated, Equation 2 was used, where k SHxH , k S Ϫ xH , and k S Ϫ x Ϫ are the rate constants for the fully protonated, and singly and doubly deprotonated forms, respectively (25,26). In cases where two proton dissociation events were cooperatively coupled, Equation 3 was used (see Supplemental Data), where pK av is the apparent average of the two pK a values (see Supplemental Data). For fitting titration data to three proton dissociation events where the latter two are cooperatively coupled, Equation 4 was used (see Supplemental Data).
For the determination of the pK a of Cys 77 in C74A ResA using iodoacetate, samples at pH values between 7.0 and 9.5 were taken after 24 h of incubation in iodoacetate and run on native PAGE. Stained modified and unmodified proteins were quantitated using imaging software. For the pH titration, fraction modified, f, was converted to an observed relative rate constant, k 0 , using Equation 5, where c is a constant that includes a time component. Rate constants were plotted against pH and fitted using the appropriate equation above. Attempts to determine pK a values using uv absorbance spectroscopy to monitor directly the concentration of cysteine thiolates (21) did not give reliable data. This is likely to be a consequence of the overlap of cysteine thiol and tyrosine hydroxyl deprotonation events (ResA has 6 tyrosine residues). Crystallization and Data Collection-ResA proteins were concentrated to 12-17 mg/ml in 20 mM MOPS pH 7.0. Conditions for crystallization of ResA variants were screened by the sitting drop vapor diffusion method at 25°C, employing a 24 condition two-dimensional grid screen (see Supplemental Data), where equal volumes of protein were mixed with well solution. Crystals were soaked in a cryoprotectant solution composed of 30% PEG 4,000, 0.1 M ammonium acetate, 0.1 M sodium citrate, 20% ethylene glycol for up to 5 min before freezing. For the high pH structure of ResA, crystals were grown at pH 5.6 and subsequently equilibrated in an equivalent cryoprotectant for which the MES component was replaced by 0.1 M CHES pH 9.25. X-ray diffraction datasets were collected on beam lines ID-29 and ID14 -2 of the Euro-pean Synchrotron Radiation Facility (Grenoble, France). Diffraction data were indexed and integrated with MOSFLM (27), and subsequently scaled using SCALA (28). Data collection statistics for each dataset are presented in Table 1.
Structure Determination and Refinement-Structure determination was aided using programs of the CCP4 suite (29). Starting phases were obtained by molecular replacement using the program MOLREP (30). Efforts to minimize phase bias are described in Supplemental Data. In each case, monomer A from the reduced structure of wild-type ResA (1SU9) was used as the initial searching model. Refinement of successful molecular replacement solutions was initiated using REFMAC (31) and refined further through alternate rounds of manual building in COOT (32) and restrained refinement in REFMAC. Completed structures were validated with the aid of PROCHECK (33) and SFCHECK (34) and the coordinates submitted to the RCSB Protein Data Bank along with appropriate structure factor files. Assigned PDB ID codes are: 2H1A (C74A); 2H19 (C77A); 2H1G (C74A/C77A); 2H1B (E80Q); 2H1D (high pH ResA). Refinement statistics for each structure are presented in Table  1. See Supplemental Data for further details of crystallization, data collection, and structure refinement.

RESULTS AND DISCUSSION
The Active Site Cysteines of Wild-type ResA Have Unusual Reactivities-The alkylating reagent 6-bromoacetyl-2-dimethylaminonaphthalene (badan) forms a thioether bond with the side chain of cysteine residues, leading to a significant increase in its fluorescence intensity. The spectrum of unbound badan features a band at ϳ540 nm. On addition of ResA (at pH 8.5) the probe fluorescence showed a marked increase over time, with two bands developing at ϳ460 and 520 nm, see Fig. 1A. The probe fluorescence is highly sensitive to its environment, with emission maxima for the ␤-mercaptoethanol adduct ranging from ϳ550 nm in a polar solvent (e.g. water) to 450 nm in hydrophobic solvents (e.g. toluene) (20). From the crystal structure of reduced ResA, it is known that the Cys 74 sulfur atom is reasonably exposed to the bulk solvent and so badan attached to this residue is expected to give a fluorescence signal near to that of the unbound probe. Cys 77 , which is substantially more buried in a hydrophobic pocket, is likely to give rise to the peak at 460 nm. Addition of badan to a cysteine-free C74A/C77A ResA double variant (at pH Ͻ 10) did not cause an increase in intensity, confirming the specificity of the probe for thiol groups. Fluorescence at 450 and 540 nm was plotted against time (off-maxima wavelength data were used to minimize effects from peak overlap), at a range of pH values, see Fig. 1, B and C. The initial increase and subsequent decrease of intensity observed at 450 nm at higher pH is most likely because of local unfolding caused by the addition of two labels at the active site (see Supplemental Data). At pH values above pH 9.5, significantly increased rates of modification were observed, suggesting that the protein undergoes unfolding in this pH range; this was confirmed by pH stability experiments (see Supplemental Data).
Pseudo-first order rate constants were obtained from fits of the data at 540 and 450 nm and plotted against pH (for values below 9.5), Fig. 1D. Data at 450 nm, reporting on Cys 77 , fitted well to a single proton dissociation event, with a pK a of 8.2 Ϯ 0.13. The data at 540 nm, reporting on Cys 74 , could not be satisfactorily fitted to equations describing simple Henderson-Hasselbach titration behavior (Equations 1 and 2). The plot is sigmoidal, but the transition region of the curve is much steeper than expected. This implies that the titration behavior of Cys 74 is strongly cooperatively coupled to another ionizable residue, which binds and releases protons simultaneously with Cys 74 . Attempts to fit the data to an equation describing interacting functional groups (35) indicated that the concentration of the two monoprotonated species must be negligible (see Supplemental Data). Thus, Equation 3 describing the cooperative binding of two protons was employed, giving a good fit with an average apparent pK a of 8.8 Ϯ 0.2. We note that, because of the strong cooperativity observed, it may be more appropriate to consider an apparent overall equilibrium constant (K app ) for the two proton process, where, here, pK app ϭ 17.6. A similarly high pK a value (Ͼ9) was recently reported for the solvent exposed active site cysteine of the C-terminal domain of DsbD (36).
ESI-MS spectra of ResA samples reacted with badan at pH 7 and pH 9 contained peaks at 15,925 and 16,350 Da, with a higher relative intensity of the higher molecular weight peak at higher pH (not shown). These correspond to the unmodified protein (with the N-terminal methionine excised (13)) and doubly labeled ResA, respectively, confirming the stoichiometry of the labeling reaction. The lack of singly modified protein could imply cooperativity in the alkylation reaction or simply that the singly modified protein is not detected. A native gel of ResA following reaction with iodoacetate, Fig. 2A, shows clearly that both cysteines react to give the doubly modified protein, and that this does not occur cooperatively.
In thioredoxin, only the first cysteine residue of the CXXC motif can be modified in the folded protein by iodoacetate (37). The pK a value for this cysteine is in the range 6.7-7.5, significantly lower than the typical value of ϳ8.5-9.0 observed for cysteine residues, while that of the second cysteine is estimated to be Ͼ9 (25,35,37,38). This large separation of pK a values is consistent with the close proximity of the two thiol groups, indicating that the ionization of one significantly influences that of the other. The cysteine thiols of thioredoxin may even share a proton after the deprotonation of the N-terminal cysteine thiol (25). The wide separation of active site thiol pK a values appears to be a general feature of TDORs that act with low specificity (21,39).
One of the most striking features of the structure of reduced ResA is the atypically large separation of ϳ4.5 Å between the cysteine thiols (compared with the 3.7-3.9 Å separation in reduced thioredoxin (40, 41) and 3.5 Å in reduced DsbA (42)). Thus, the two cysteines should have significantly less effect on each other in terms of their acid/base properties and their resulting reactivities. Data indicate that this is the case and ResA has very different properties to thioredoxin and related TDORs: both cysteine thiols are reactive to alkylating reagents; the pK a values for the two cysteine residues of the active site CXXC motif are both above 8 and within ϳ0.5 units of each other; and, the second cysteine residue at the active site has a lower pK a than the first.
Structural Analyses of Active Site Cysteine Variants of ResA-To investigate the properties of the active site thiols individually, C74A and C77A single variants of ResA were generated and isolated, and their crystal structures solved, along with that of the C74A/C77A double variant (see Table 1). Cysteine vari- B shows similar non-denaturing PAGE of C74A ResA with iodoacetate in a mixed buffer system after 24 h; pH as indicated. Band intensity was used to estimate a rate constant, k 0 , for the reaction (see "Experimental Procedures"). C shows a plot of k 0 (left hand abscissa, filled squares) as a function of pH. There are insufficient data points to obtain a very reliable fit, but because the data has the same form as the data from badan labeling of C74A ResA (right hand abscissa, open circles), the iodoacetate data were fitted to Equation 4 (solid line).
ants crystallized in space group P2 1 2 1 2 1 with two molecules per asymmetric unit and were isomorphous with previously reported crystals of reduced wild-type ResA. Fig. 3D-F depict the active sites of C77A, C74A, and C74A/C77A ResA, respectively. Beyond the point of mutation itself, the variants do not show any significant conformational deviations from the wildtype structure (Fig. 3A). The cysteine variant structures show that any biochemical differences observed are not because of inconsequential structural rearrangements and thus provide a valuable control for the interpretation of biochemical data derived from these proteins.
The Reactivities of ResA Active Site Cysteines Are Not Interdependent-Reaction of C77A with badan caused an increase in fluorescence at 500 nm (Fig. 4A), indicating that the environment of the probe at Cys 74 is slightly more hydrophobic in this variant compared with wild-type. C74A exhibited a band at 443 nm, similar to that observed for Cys 77 in the wild-type protein (Fig. 4D). The difference in solvent exposure of the two cysteines results in badan-labeled Cys 77 exhibiting an intensity 8-fold greater than that for Cys 74 . These intensities are also, respectively, 16-and 2-fold greater than the equivalent bands in the wild-type protein spectrum (note that this is not apparent from Fig. 4, A and D because a smaller emission slit width was employed for the variant proteins). Thus, significant selfquenching occurs within the doubly labeled wild-type protein.
Plots of 500 and 450 nm intensity against time (Fig. 4, B and E) show that Cys 74 and Cys 77 exhibit behavior similar to the wildtype protein. As pH increases, local and then global unfolding occurs, as observed for the wild-type protein (see Supplemental Data).
Pseudo-first order rate constants for C77A and C74A were obtained from the data and plotted against pH, see Fig. 4, C and F, respectively. As for the wild-type C74 data, attempts to fit each data set to a one proton dissociation process did not give good fits (see Fig. 4F), but Equation 3, describing two cooperatively coupled proton dissociation processes, gave a significantly improved fit in the latter part of the plot. However, the low pH data are still not well simulated. The best fits (particularly for the less noisy C74A data) were obtained by fitting each data set to Equation 4, which describes, in addition to two cooperatively coupled proton dissociation processes, a lower pH dissociation event. This gave a pK a value of 6.5 Ϯ 0.6 for the initial proton dissociation event (for both plots), which we ascribe to a non-cysteine residue. This dissociation event also very likely occurs in the wild-type protein; the inclusion of this process in the fitting of the wild-type ResA data gave equivalent/improved fits (not shown). In addition to the initial pK a of ϳ6.5, fits of the single cysteine variant data gave average apparent pK a values of 8.48 Ϯ 0.08 and 8.36 Ϯ 0.05, respectively, for Cys 74 and Cys 77 (corresponding to pK app values of 17 and 16.7, respectively). Data from pH stability studies also indicate the presence of a dissociable proton between pH 6.5-7.0 (see Supplemental Data).
To verify the reactivities of the active site cysteines, wild-type ResA, and both single cysteine variants were reacted with Ellman's reagent. This demonstrated similar rates of reaction for each protein (see Supplemental Data). To verify pK a values determined with badan-labeling, the pK a of Cys 77 in C74A ResA was determined using iodoacetate at pH values between 7.0 and 9.5 (note that the reactivity of the individual residues in the wild-type protein cannot be followed separately using this method and so individual pK a values in the wild-type protein could not be measured). Reaction products at different pH values were run on a native gel, Fig. 2B, and observed rate constants (see "Experimental Procedures") were plotted as a function of pH, Fig. 2C. At pH 9.5 and above, a significant increase in rate was observed (not shown) consistent with protein unfolding at high pH. The shape of the plot is very similar to the equivalent plot for the reaction with badan (see inset), and the data were fitted to the same three proton dissociation process, giving an average apparent pK a value for Cys 77 of 8.3 Ϯ 0.1 (pK app ϭ 16.6). Thus, in both single cysteine variants, the deprotonation of the remaining cysteine is cooperatively coupled to another, non-cysteine residue. In the wild-type protein, only the deprotonation of Cys 74 is coupled in this way. Thus, in the C74A variant, the coupling interaction that is normally between Cys 74 and the non-cysteine residue is replaced by a similar interaction between Cys 77 and the non-cysteine residue. The identity of the coupled, non-cysteine residue is unknown but we presume that it is the same residue in wild-type ResA (coupled to Cys 74 ) and in each variant (coupled to Cys 74 and Cys 77 , respectively). The residue must be spatially close to the active site to have a significant effect on the rate of modification, but cannot be so close that its ionization inhibits that of the cysteine residue(s). There are very few ionizable residues close to the active site that are not entirely solvent-exposed (and therefore unlikely to have an unusual pK a or be cooperatively coupled to Cys 74 ). One such residue is Glu 105 ; this partially buried residue is ϳ7.4 Å away from Cys 74 (and ϳ8.7 Å away from Cys 77 ) and is also conserved or conservatively substituted in ResA/CcmG homologues (by sequence and structural alignment). A pK a value Ͼ 8 would be very unusual for a glutamate residue side chain; one possibility is that its deprotonation at lower pH is inhibited and that this is somehow relieved by the deprotonation of Cys 74 . This would be consistent with the very tight cooperativity observed, but further investigation is required to test this possible coupling mechanism. In summary, the data from single cysteine variants demonstrate that the cysteine pK a values are not interdependent; removal of one or the other of the cysteines has only a small effect on the pK a of the remaining cysteine thiol. This is consistent with the large separation between the cysteine thiols of reduced ResA.
Structural Analysis of ResA at High pH-We have previously reported crystal structures of ResA in which the protein was crystallized at pH 5.6 (14). The pK a data presented above clearly indicate that the reduced structure represents the protonated, dithiol, form of the protein. To investigate potential structural rearrangements linked to deprotonation of the active site cysteines, we have determined a 2.4 Å crystal structure of DTTreduced ResA at pH 9.25 (Table 1 and "Experimental Procedures").
Overall, the structures determined at pH 9.25 and pH 5.6 are extremely similar (compare Fig. 3, A and B); individual ResA monomers of the high pH and low pH forms can be superposed with an average root mean square deviation (RMSD) of just 0.24 Å for C ␣ positions and 0.52 Å for all atoms. The arrangement of residues in the active site is also extremely similar. Thus, crystallography does not reveal any significant structural rearrangements upon moving from a low to high pH medium. While the lack of significant structural rearrangements that might stabilize the thiolate form is consistent with the low potential of the protein, we cannot rule out that some structural changes might occur in solution but are not observed in the crystalline form of the protein, due, for example, to crystal packing effects. Also, given the resolution limits, and although we believe it likely, we cannot be absolutely certain that the high pH structure represents the dithiolate form of ResA.

Glu 80 Plays a Key Role in Controlling the Reactivities of the Active Site
Cysteines-Adjacent to the active site of reduced ResA (in thiol and thiolate forms) is a hydrophobic cavity, which is proposed to be important for the interaction of ResA with its substrates (14,18). To test the possibility that the presumed negative charge associated with Glu 80 , which is located at the base of the cavity, is an important determinant for the unusually high pK a values of the active site thiols, a E80Q variant was constructed and isolated. The variant was found to be significantly less stable than wild-type protein at high pH (see Supplemental Data). Reaction with badan revealed that the reactivities of the two cysteines are an order of magnitude greater than for the wild-type protein, and also much better resolved: Cys 74 (monitored though the fluorescence intensity at 550 nm) reacted much more rapidly than Cys 77 (at 440 nm), see Fig. 5A. Although the crystal structure (see below) shows that the protein adopts the wild-type fold, the decreased stability of the variant could indicate greater flexibility around the cysteines. Consistent with this is the shift in the Cys 77 peak maximum to ϳ470 nm. To follow Cys 74 , reactions were carried out at lower protein concentration and temperature (10°C). Kinetic runs (Fig. 5, B and C) were fitted to obtain pseudo-first order rate constants (k 0 ) and the data for Cys 77 were plotted as a function of pH (Fig. 5D). The data fitted well to a single proton dissociation event, indicating that the removal of Glu 80 results in the loss of the low pK a (ϭ ϳ6.5) proton dissociation event observed for the wild-type protein and the single cysteine variants. This is supported by the unfolding data for E80Q (Supplemental Data) which also indicates the loss of the titratable group with a pK a of ϳ6.5. We conclude that this group is the carboxylate side chain of Glu 80 . This pK a is, in general, rather high for a glutamate side chain, but perhaps not for one that is buried in a hydrophobic environment (55). The pK a value for Cys 77 obtained from the fit is 7.4 Ϯ 0.1, i.e. approximately one log unit lower than in the wild-type protein. Because of the lower stability of the protein to high pH, we were unable to obtain sufficient data points in the high pH region to determine the pK a of Cys 74 . Nevertheless, the reactivity data indicate strongly that the pK a of Cys 74 is significantly lower than in the wild-type protein (and we estimate it to be Ͻ8).
We conclude that Glu 80 plays a key role in modulating the reactivity of the protein. It has been noted previously that the equivalent of Glu 80 in CcmG (Glu 86 ) is located close to the position of a conserved aspartate in thioredoxin (Asp 26 in the Escherichia coli protein). This is believed to deprotonate the second cysteine thiol of the active site, thus facilitating resolution of mixed disulfide intermediates (43,44). This raises the possibility that Glu 80 serves a similar function in ResA/CcmG homologues (45). D26A thioredoxin, however, was found to have significantly lower activity and an increased pK a value associated with its N-terminal active site thiol (46). Our data indicate that the negative charge associ-ated with Glu 80 inhibits deprotonation of the cysteine residues, thus elevating their pK a values. Hence, it is unlikely that E80 has the same function as Asp 26 in thioredoxin. Consistent with this is the observation that substitution of Glu 86 in CcmG did not affect cytochrome c maturation (45).
Structural Insights into the Interaction of ResA with Its Substrate(s)-The crystal structure of E80Q in the reduced state was determined at 1.95 Å (see Table 1 and Fig. 3C). The structure not only confirms that the variant is correctly folded in a manner analogous to the wild-type reduced protein, but also reveals several interesting intermolecular contacts, not previously seen in other ResA crystal structures. Some of these contacts occur close to the active site and appear to induce conformational changes to the active site cysteines. We have therefore explored the possibility that these contacts, while not physiological, may nonetheless give structural insight into the interaction of ResA with its redox partners. The structure of E80Q was determined for a monoclinic crystal with 4 monomers per asymmetric unit (chains A, B, C, D; see Table 1). The active sites of monomers B and D are essentially identical to the wild-type structure with no unusual conformations imparted by the replacement of Glu 80 by glutamine. The two remaining monomers show differences.
In monomer A, the side chain of the more solvent-exposed cysteine (Cys 74 ) adopts two alternative conformations and a significant kink is observed in the active site helix (Fig. 6A). This does not appear to result directly from the E80Q substitution (because monomers B and D have single cysteine conformations); rather, we attribute this behavior to a crystal contact with monomer D.
The most striking feature of this contact is the insertion of a side chain (Gln 130 ) from monomer D into the hydrophobic cavity of monomer A. We and others (14,18,47) have previously shown by x-ray crystallography and NMR spectroscopy that the cavity may provide a means for ResA to specifically recognize the conserved histidine of the CXXCH motif. While the intermolecular interaction between these two ResA monomers does not confirm the cavity as being important for physiological interactions with apocytochrome substrates, it does prove that a single residue side chain can be accommodated within the cavity and that such interactions may lead to subtle conformational changes at the active site. The importance of the cavity for substrate binding was also recently proposed by Colbert et al. (18) on the basis of NMR chemical shift data resulting from the addition of a mimetic apocytochrome c peptide to ResA.
Monomer C of the E80Q structure is also subject to a crystal contact in the vicinity of the active site (Fig. 6B). The sulfur-to-sulfur distance in this monomer is just 3.3 Å, the shortest distance observed for any ResA structure. Monomer C contacts a symmetry-related monomer A via Leu 140 (immediately preceding cis-Pro 141 ) and Trp 73 . The backbone of Leu 140 is able to form a pair of hydrogen bonds with the side chain of Asn 112 (from monomer A) by virtue of the cis conformation of Pro 141 , which orientates the Leu 140 carbonyl oxygen away from the interior of the protein and Trp 73 forms a hydrophobic interaction with Ile 108 (from monomer A). The mode by which these contacts induce conformational changes in the cysteine positions is unclear, but contact with the cis-proline and Trp 73 appears to be important; Pro 141 is invariant among thioredoxin-like proteins (in the cis conformation) and Trp 73 is only occasionally substituted by other large aromatics.
It is interesting to note that the pair of hydrogen bonds formed between the backbone atoms of Leu 140 (in monomer C) and the side chain of Asn 112 (in monomer A) closely mimic hydrogen bonds that are universally conserved among structures of TDORs in complex with their substrates. For example, the crystal structure of E. coli CcmG in a mixed disulfide with the N-terminal domain of DsbD (48) revealed hydrogen bonds between the backbone carbonyl of CcmG Ala 143 (equivalent to Leu 140 in ResA) and the backbone amide nitrogen of DsbD Cys 109 , and another between the backbone amide nitrogen of Ala 143 and the backbone carbonyl of Cys 109 . Similar pairs of hydrogen bonds are also apparent in other structures of mixed disulfide complexes (49 -51) even though very few direct intermolecular hydrogen bonds are observed in such complexes. The importance of the cis-proline for TDOR:substrate interactions is further demonstrated by studies of E. coli DsbA proline mutants which are observed to accumulate mixed disulfide complexes with substrates (52) and modeling studies of the interaction of glutathione with glutaredoxin (53). We conclude that the cis-proline appears to play a critical role in the reactivity of TDORs, and that interactions similar to those observed here are likely be important for ResA:apocytochrome c interactions.
The Mechanism of ResA Specificity-The data reported here lead to the following proposal for how ResA achieves specificity. In the absence of an apocytochrome c substrate molecule, the high pK a values of the ResA active site thiols, resulting at least in part from their interaction with Glu 80 , ensure that ResA is unreactive toward nonspecific substrates. ResA must become activated in some way, and we propose that this occurs through binding of a substrate molecule. Substrate binding would result in hydrogen bonding interactions, most likely including a reciprocal pair between the backbone carbonyl and amide nitrogen of Leu 141 and the N-terminal cysteine in the CXXCH motif of the apocytochrome. We suggest that this orientates the substrate molecule and facilitates the docking of the histidine residue side chain (of the apocytochrome CXXCH motif) into the hydrophobic surface cavity, resulting in a likely hydrogen bond interaction between it and Glu 80 at the base of the cavity. This would mask, at least partially, the effect of Glu 80 on the acid/base properties of the active site, thus bringing the pK a values of the active site thiols into the physiological range. . Intermolecular contacts between E80Q monomers that may mimic the interaction between ResA and its redox partners. A, intermolecular contact between chain A (light blue) and chain D (gray) in the 1.95-Å structure of E80Q ResA. The side chain of Gln 130 has become inserted into the hydrophobic cavity found only in the reduced form of ResA. B, intermolecular contact between the active site of chain C (green) and a symmetry related monomer of chain A (aubergine). The interaction gives rise to a significant conformational change in the position of the active site cysteines reducing the distance between the S␥ atoms by 25%.
Structural changes, resulting from substrate binding in the cavity, and other protein-protein interactions involving the Leu 140 -cis-Pro 141 loop and Trp 73 are likely to considerably reduce the separation between the S ␥ atoms of the active site cysteines, causing the pK a values to separate significantly (25). In this form, ResA, with one cysteine now in its thiolate form, is primed for nucleophilic attack on the disulfide bond of the substrate. Experiments designed to further test this model are in progress.
Significance of Substrate-mediated Activation among TDORs-The requirement for specificity is a common one among TDORs on the outside of the cytoplasmic membrane, raising the question of whether related or quite distinct mechanisms have evolved to prevent unproductive short circuiting of thiol-disulfide oxidative and reductive branches.
Here we have identified Glu 80 in ResA as a key residue for preventing reaction of the protein with non-cognate substrates. Amino acid residue sequence alignments show that Glu 80 is not conserved in cytoplasmic TDORs, but is highly conserved among reductive branch extracytoplasmic TDORs, including many that have no role in cytochrome c maturation (see Supplemental Data). In contrast, residues that form the near active site hydrophobic cavity of ResA are generally well conserved only in Gram-positive TDORs involved in System II cytochrome c maturation. Thus, while it is likely that the latter residues are important for binding apocytochrome c substrates, the primary role of Glu 80 appears to be to control active site reactivity. This glutamate residue is likely to be similarly important in other reductive branch TDORs, where it may function as part of substrate-mediated activation mechanisms similar to that proposed here for ResA.