Structural Basis of Redox-coupled Protein Substrate Selection by the Cytochrome c Biosynthesis Protein ResA*

Post-translational maturation of cytochromes c involves the covalent attachment of heme to the Cys- Xxx-Xxx -Cys-His motif of the apo-cytochrome. For this process, the two cysteines of the motif must be in the reduced state. In bacteria, this is achieved by dedicated, mem-brane-bound thiol-disulfide oxidoreductases with a high reducing power, which are essential components of cytochrome c maturation systems and are also linked to cellular disulfide-bond formation machineries. Here we report high-resolution structures of oxidized and reduced states of a soluble, functional domain of one such oxidoreductase, ResA, from Bacillus subtilis . The structures elucidate the structural basis of the protein’s high reducing power and reveal the largest redox-coupled conformational changes observed to date in any thiore-doxin-like protein. These redox-coupled changes alter the protein surface and illustrate how the redox state of ResA predetermines to which substrate it binds. Fur-thermore, a polar cavity, present only in the reduced state, may confer specificity to recognize apo-cyto-chrome c . The described features of ResA are likely to be general for bacterial cytochrome c maturation systems. c -Type cytochromes are crucial for respiration in many biological species They have considerable structural and functional flexibility. Perhaps the best known are the mono-heme proteins that shuttle electrons between respiratory membrane-bound, proton-translocating

c-Type cytochromes are crucial for respiration in many biological species (1). They have considerable structural and functional flexibility. Perhaps the best known are the mono-heme proteins that shuttle electrons between respiratory membranebound, proton-translocating protein complexes, i.e. from the cytochrome bc 1 complex to cytochrome c oxidase. Many bacteria contain multi-heme c-type cytochromes in which the hemes have electron transfer and catalytic activities, e.g. cytochrome c nitrite reductase (2,3). Many bacteria also contain enzymes with heme c and at least one other redox cofactor, e.g. hemocytochromes c (4), flavocytochromes c (5,6), and quinocytochromes c (7). Recent findings that mono-heme c-type cytochromes function as apoptosis-triggering factors in eukaryotes (8) and as sensors of the toxic signaling molecule nitric oxide (9) have sparked new interest in these proteins. Recent years have seen great advances in our understanding of the structure and function of a diverse range of c-type cytochromes (10 -12) and their (interaction with) redox partners (13)(14)(15)(16)(17).
In contrast, relatively little is known about how c-type cytochromes are synthesized. This process, commonly referred to as cytochrome c maturation (CCM), 1 involves the post-translational, covalent attachment of heme to the cysteinyls at the heme-binding site, which usually contains the conserved Cys-Xxx-Xxx-Cys-His sequence motif. Largely on the basis of genetic information, three markedly different CCM systems have been identified. Systems I and II occur predominantly in bacteria, whereas system III is found exclusively in eukaryotes (1, 18 -20).
The type I system of Escherichia coli (Fig. 1A) comprises the membrane-bound proteins CcmABCDEFGH. CcmA, CcmB, and CcmC form an ABC-type transporter, the function of which is unknown (20,21), although it is known that CcmC, together with CcmE, is involved in heme delivery to the cytochrome c heme lyase complex consisting of CcmFH (22)(23)(24)(25)(26). Reduction of the cysteinyls of the Cys-Xxx-Xxx-Cys-His motif prior to heme insertion is accomplished by CcmG and CcmH, but it is not clear whether the electrons flow from CcmG by means of CcmH, or from CcmH via CcmG, to the apo-cytochrome (27,28). In addition to these CCM-specific proteins, DsbA, which catalyzes the formation of disulfide bonds in the periplasm, and DsbB, which releases the electrons obtained from DsbA to the respiratory system by means of the quinol pool, have been reported to be important for formation of a disulfide bond in the apo-cytochrome prior to CCM (29 -33). Finally, DsbD (DipZ), an integral membrane protein that utilizes three distinct domains to transfer electrons from cytoplasmic thioredoxins to the periplasm, supplies CcmGH with the necessary electrons for apo-cytochrome reduction (34 -37).
In common with the type I CCM system, the type II system contains heme delivery/ligation and apo-cytochrome disulfide reductase pathways (18). Remarkably, however, system II seems to comprise just three, rather than eight or nine, CCMspecific proteins (19,38,39). The type II system of Bacillus subtilis (Fig. 1B) comprises the ResA, ResB, and ResC proteins (19,39). The function of ResB and ResC has not been studied on a molecular level but may lie in heme delivery and/or ligation to the apo-cytochrome. ResC contains a tryptophan-rich sequence motif, also found in system I CcmE and CcmF, thought to be involved in heme binding (18,19,40). ResA is a thiol-disulfide oxidoreductase with a very low midpoint redox potential (Ϫ340 mV at pH 7) that is tethered to the membrane via a single transmembrane helix (39). It is thought to reduce the disulfide bond of the Cys-Xxx-Xxx-Cys-His motif of the apo-cytochrome.
The necessary electrons to do this are probably delivered by the integral membrane protein CcdA, a relative of DsbD (18,39,(41)(42)(43). The role of CcdA in apo-cytochrome disulfide reduction is supported by the fact that inactivating mutations in bdbC and bdbD, which encode orthologs of the E. coli DsbB and DsbA proteins, can complement the cytochrome c-deficient phenotype of a CcdA-inactivated mutant (43).
The apparent simplicity of the Bacillus (type II) system in relation to that of E. coli (type I) makes it an attractive system for study. It also raises questions as to whether system II is indeed simpler than system I, whether each system II protein may combine functions of several system I proteins in one, and whether different pathways or underlying chemical principles may occur in the two CCM systems. To begin to address these issues, the three-dimensional structure of the enzymatic domain of ResA has been elucidated in its two biologically relevant redox states. These structures explain the molecular basis of the low redox potential that is fundamental to its function in reducing apo-cytochrome c. Moreover, the observed conformational differences in the two structures suggest a mechanism whereby the redox state of ResA might predetermine which substrate, apo-cytochrome c or CcdA, it binds. This selective substrate recognition may be a general feature of CCM-specific disulfide bond reduction systems.

Seleno-L-methionine Labeling of Soluble and His-tagged Soluble
ResA-Histidine-tagged and wild-type soluble domain constructs of ResA were expressed in E. coli and purified as described previously (39). For over-expression and production of seleno-methionine-containing ResA, the E. coli selenium auxotroph strain B834(DE3) was transformed with plasmids pRAN8 (encoding His-tagged soluble ResA (htsResA)) or pRAN11 (encoding soluble ResA (sResA)) according to standard procedures (44). The B834 strain cells were initially grown on M9 minimal media, supplemented with 50 g/ml methionine, at 37°C. When the cells reached an A 600 ϭ 1.0, the culture was centrifuged at 6000 ϫ g for 10 min at 4°C. Cells were resuspended in 1 liter of M9 minimal media without methionine, incubated at 37°C, and centrifuged at 200 rpm. After 6 h, seleno-L-methionine was added to a final concentration of 50 g/ml and the culture was incubated at 37°C and centrifuged at 200 rpm for 30 min. Over-expression of ResA was then induced by adding isopropyl-1-thio-␤-D-galactopyranoside to a final con-centration of 1 mM and the culture was incubated at 37°C and centrifuged at 200 rpm for a further 10 h. Cells were harvested by centrifugation at 7000 ϫ g for 15 min at 4°C. The incorporation of selenomethionine was verified by electro-spray mass spectrometry. Labeled htsResA was purified as described previously for the unlabelled proteins (39).
Crystallization, Data Collection, and Processing-As ResA was isolated as a mixture of 70% oxidized and 30% reduced states (39), complete oxidation of samples prior to crystallization was carried out by incubation in 10 mM diamide for 3 h at 277 K in the dark. Excess diamide was removed by using a 5-ml HiTrap desalting column that was previously equilibrated with 20 mM potassium phosphate, pH 7.0. Samples were subsequently concentrated to 12 mg/ml in 20 mM Na ϩ /K ϩ phosphate, pH 7.0, and centrifuged for 10 min at 13,000 rpm at 277 K. Crystals were grown by the hanging-drop vapor diffusion method. Crystals of oxidized htsResA and sResA grew in approximately 2 weeks at 277 and 289 K from 24 -27% (w/v) PEG 4000, 0.2 M ammonium acetate, and 0.1 M sodium citrate, pH 5.6 -5.9. The optimal crystallization conditions for both protein forms were identical, and the resulting crystals were indistinguishable, although crystals of the His-tagged protein were not as reproducible. Two distinct crystal forms were obtained for both His-tagged and non-tagged proteins. The first type of crystals grew as extruded hexagonal needles that belonged to space group P6 5 with cell dimensions a ϭ b ϭ 36.6 Å, c ϭ 176.9 Å with one molecule per asymmetric unit. The second crystal form also exhibited morphology strongly indicative of the underlying hexagonal symmetry. These bipyramidal hexagonal crystals belonged to space group P6 5 with cell dimensions a ϭ b ϭ 61.0 Å, c ϭ 165.4 Å, with two monomers per asymmetric unit. Crystals of reduced ResA were obtained from similar conditions as the oxidized form with the addition of 10 -40 mM dithiothreitol. These crystals grew in 2-3 days and belonged to space group P2 1 2 1 2 1 with cell dimensions a ϭ 47.5, b ϭ 59.7, c ϭ 110.1 Å. All ResA crystals could be frozen successfully in a solution containing 20% (v/v) ethylene glycol, 30% (w/v) PEG 4000, 0.2 M ammonium acetate, 0.1 M tri-sodium citrate, pH 5.6. The cryoprotectant solution of reduced crystals was supplemented with 40 mM dithiothreitol.
Native data sets of the first, needle-like crystals of both htsResA and sResA were collected at 1.8 Å resolution at beam line XRD1 at Sincrotrone Trieste, Trieste, Italy. Data sets of the second type of oxidized ResA and of reduced ResA crystals were collected at 1.50 and 1.95 Å resolution, respectively, at BM14 of the European Synchrotron Radiation Facility, Grenoble, France. A multiple-wavelength anomalous dispersion (MAD) data set was collected at 2.37 Å resolution of selenomethionine ResA crystals at the Protein Structure Factory, Berliner Elektronenspeicherring-Gesellschaft fü r Synchrotronstrahlung mbH, Berlin, Germany. All data sets were processed and reduced with the HKL package (45). Data collection statistics are summarized in Table I.
Structure Determination and Refinement-The structure of untagged, oxidized ResA was determined by MAD using the anomalous signal of five incorporated selenium atoms per ResA molecule. Identification of the 10 Se atom substructure and subsequent phasing was carried out with SOLVE (46). The obtained phases were of excellent quality (FOM ϭ 0.82), which allowed for 87% of all amino acids in the asymmetric unit to be built automatically by RESOLVE (47,48). The model was completed using alternating rounds of manual model building using O (49) and automated refinement using programs of the CCP4 suite (50,51). As refinement of the protein model approached completion, refinement was continued using a 1.5 Å resolution native data set, and phases were calculated from the model. Water molecules were incorporated using ARP (52) and, in the latter stages of refinement, the model was further improved by applying individual atomic anisotropic B-factor refinement using REFMAC (53).
Throughout refinement, progress was monitored with the aid of an R free value calculated with 5% of the data (54). Upon completion of the model (R free ϭ 12.3%, R work ϭ 15.0%), a final round of refinement using all of the data was used to give a single crystallographic R factor (R cryst ϭ 12.25%). The final model comprises 2 monomers of ResA (residues 37-173 and 39 -173), 434 molecules of water, and 4 molecules of ethylene glycol. Several residues also exhibit clearly defined alternate conformations.
The structure of reduced ResA was determined by molecular replacement using the CCP4 version of AMORE (55) using as a search model the structure of the oxidized state in which the active-site cysteines were first replaced by alanines. A round of simulated annealing in CNS (56) preceded iterative cycles of manual rebuilding using O (49) and maximum likelihood refinement in CNS (57). The addition of 300 waters and a final round of correlation-based refinement using all of the data completed the model (R free ϭ 21.0%, R work ϭ 18.3%, R cryst ϭ 18.1%). The reduced structure incorporates residues 39 -175 in the first monomer and 39 -174 in the second. Both the oxidized and reduced structures exhibit excellent stereochemistry and geometry as judged with PROCHECK (58). Structure determination and refinement statistics are summarized in Table I.

High-resolution Structures of Oxidized and Reduced ResA-
The crystal structure of oxidized ResA was determined at 2.37 Å by the MAD method using the anomalous signal of five incorporated selenium atoms per monomer and subsequently refined at 1.5 Å (R cryst ϭ 12.25%, R work ϭ 12.3%, R free ϭ 15.0%). The crystals contain two independent molecules in the asymmetric unit of a hexagonal P6 5 cell. These molecules can be superimposed with a root-mean-square deviation of 0.63 Å.
The overall structure of monomeric ResA contains a classical thioredoxin fold, comprising a mixed four-stranded ␤-sheet surrounded by three helices. ResA contains two additions to this motif, one N-terminal ␤-hairpin (residues 36 -63), and one insertion (residues 104 -127), which gives rise to an additional strand and helix between strand ␤2 and helix ␣3 ( Fig. 2A). Similar additions to the thioredoxin fold have been described recently in the structures of Bradyrhizobium japonicum CcmG (59) and TlpA, which is essential for biosynthesis of the cytochrome aa 3 oxidase from B. japonicum (60).
Crystals of dithiothreitol-reduced ResA belong to space group P2 1 2 1 2 1 and contained two molecules per asymmetric unit. The structure of reduced ResA was determined by molecular replacement and refined at 1.95 Å resolution (R cryst ϭ 18.1%, R work ϭ 18.3%, R free ϭ 21.0%). Protein-protein interactions between ResA monomers in reduced crystals differ from those in oxidized crystals, indicating that the presence of two molecules per asymmetric unit is an artifact of crystallization and not a functional feature. Gel filtration (39) and dynamic light-scattering experiments confirm that soluble ResA is monomeric in solution (data not shown). Overall, structural differences between the two redox states are small, as reflected in root-mean-square differences of 0.73 Å for superposition of the two reduced monomers onto one of the oxidized monomers. Nevertheless, significant conformational changes are observed in the vicinity of the active site between the oxidized and reduced proteins (Fig. 2B), and also between the two monomers in the oxidized state (Fig. 2C).
Redox-coupled Conformational Changes in the Active Site-The most obvious difference between the oxidized and reduced forms of ResA is the configuration of the active-site cysteines,  ϭ 61, b ϭ 61, c ϭ 165  a ϭ b ϭ 61, c ϭ 165 a ϭ 48, b ϭ 60 R free is calculated with a 5% subset of the data that was not used for refinement; R work was calculated with the remaining 95% of the data. R cryst , the crystallographic R factor, refers to the final model of ResA for which a final round of refinement was performed using all diffraction data.
e Root-mean-square deviation from ideal stereo chemistry.
Cys-73 and Cys-76. In the oxidized structure, the S␥ atom of each cysteine is separated by ϳ2.16 Å. Continuous electron density was observed between the sulfur atoms, indicating the presence of a covalent bond. In the reduced enzyme, the sulfur of Cys-73 is repositioned by a 228°rotation about the C␣-C␤ bond (1), whereas the C␣ position of this residue remains essentially unchanged. The rotation distances the sulfur atom of Cys-73 from that of Cys-76, which, in conjunction with movement of Cys-76, results in an increase of the sulfur-sulfur distance from ϳ2.16 to ϳ4.5 Å. This is the largest distance between the two active-site sulfur atoms observed so far in any thioredoxin-like protein. This large distance precludes formation of a hydrogen bond between the two cysteines, which is a stabilizing interaction reported for the reduced state of several other thioredoxin-like proteins, i.e. DsbA and human and E. coli thioredoxin (61-64) (see below). In addition to an increase of the sulfur-sulfur distance, the rotation of the Cys-73 side chain forces the S␥ atom toward the protein surface. The partially solvent-exposed thiolate is surrounded by hydrophobic residues and is only involved in a single hydrogen-bond interaction. Thus, the residue seems poised for nucleophilic attack on the disulfide bond of its likely electron acceptor apo-cytochrome c.
These observations are consistent with the accepted catalytic mechanism of thioredoxin-like proteins: they bind their redox partner by means of a hydrophobic surface and subsequently perform a nucleophilic attack on the target disulfide bond via the thiolate of the N-terminal of the two active-site cysteines (Cys-73 in ResA). This process leads to the formation of a covalent mixed disulfide bond between the two proteins, which is resolved by a nucleophilic attack of the second cysteine (Cys-76) on the intermolecular disulfide ( Fig. 3; Ref. 65). This type of mechanism requires the N-terminal of the active-site cysteines to be in the thiolate form, which is stabilized by interaction with the dipole of helix ␣1. In several thiol-disulfide oxidoreductases, this interaction reduces the pK a of the relevant cysteines by at least two pH units. As the helix dipolecysteine interaction is conserved in ResA, it may be anticipated that ResA contains an active-site cysteine with reduced pK a in common with other thioredoxin-like proteins, but definitive conclusions await experimental determination of the pK a of Cys-73.
The increased reactivity expected of the Cys-73 thiolate may explain why this residue remains partly shrouded by hydrophobic residues in the reduced structure rather than being fully exposed to solvent: to do so would not only expose this highly reactive residue to the presumed redox-partner of reduced ResA, apo-cytochrome c, but also to other potentially reactive groups within the extra-cytoplasmic milieu. Unwanted side-reactions could lead to inactivation of ResA by covalent modification of the active-site cysteine or to wasteful electron transfer to oxidizing thioredoxin-like proteins such as the Bacillus DsbA ortholog, BdbD. Hence, the partial shielding of Cys-73 might confer protection against protein inactivation and nonspecific electron transfer.
Structural Basis for the Reducing Power of ResA-One attribute that is crucial for the presumed function of ResA of reducing the disulfide bond of oxidized apo-cytochromes is its low midpoint redox potential of Ϫ340 mV at pH 7. The relative stability of the oxidized enzyme, with respect to the reduced, is largely due to the presence of the active-site disulfide bond. This covalent bond is broken in the reduced structure, as are several other favorable interactions. These include a hydrogen bond between the main-chain amide hydrogen of Lys-77 and the main chain carbonyl oxygen of Glu-74, and a hydrophobic interaction between several residues on the protein surface and in the immediate vicinity of the active site (see below). The reduced state of ResA is further destabilized by development of a negative charge on Cys-73 in an essentially hydrophobic environment. The energetic cost of these rearrangements are only partly compensated for by the generation of a hydrogen bond between the thiol(ate) of Cys-73 and the amide hydrogen of Gly-70 (Fig. 4A), as well as a favorable interaction between the Cys-73 thiol(ate) and the dipole of helix ␣1.
The poor stabilization of the reduced state observed in ResA is in marked contrast to that of DsbA, which has a relatively high redox potential (Ϫ125 mV, pH 7) (66) and catalyzes disulfide bond formation in the periplasm of E. coli (29). In addition to the well described interaction between the helix dipole and cysteine thiolate groups in thioredoxins (67), the reduced form of DsbA is stabilized by the formation of four hydrogen bonds with the active-site thiolate (Fig. 4B). These hydrogen bond donors are the N␦ atom of His-32, the backbone nitrogen of Cys-33, the S␥ of Cys-33, and the carbonyl oxygen of Val-150, respectively. No destabilizing effects have been described for reduced DsbA (64). The active-site interactions of DsbA thus serve to stabilize its reduced state, resulting in a relatively high midpoint redox potential. Based on a structure of its reduced state, the same seems to hold for DsbE (68). The opposite is true for ResA: the reduced structure is destabilized overall, because more favorable interactions are disrupted than formed upon reduction. This is consistent with a significant lowering of the redox potential, which is important for ResA to function as a disulfide bond reductase. Furthermore, it seems to be a general feature of thiol-disulfide oxidoreductases that the degree of stabilization of the active-site thiolate group in the reduced state determines the value of the midpoint redox potential.
Redox-coupled Conformational Changes around the Active Site Suggest a Mechanism for Substrate Selection Based on Surface Complementarity-Reduction of the active-site disulfide bond between Cys-73 and Cys-76 causes the S␥ atoms to move apart by more than 2 Å. This local movement triggers other conformational changes in the vicinity of the active site. Part of the long ␣1 helix, at the N terminus of which Cys-73 is positioned, is subject to a rearrangement (Fig. 2B). The movement of Cys-76 is accompanied by a repositioning of neighboring Lys-77 and Lys-78 by ϳ1-1.8 Å and, as a result, the hydrogen bond that is present in the oxidized structure between the main chain amide proton of Lys-77 and the main chain carbonyl oxygen of Glu-74 is broken. The movement of Cys-76 also forces a buried water molecule to move, and this in turn imposes a significant rotation on the carboxyl group of Glu-79 (Fig. 5, A and B). This glutamate is buried in the oxidized structure where it participates in an intricate hydrogen-bonding network that includes another buried polar residue, Asn-67 (Fig. 5A). Upon reduction, the hydrogen bonding network around Glu-79 is altered, and its side chain moves toward the surface (Fig. 5B). One of its carboxyl oxygen atoms becomes accessible to solvent, whereas the other remains buried. To accommodate Glu-79 in its new position, Gly-157, Thr-158, Met-159, Pro-138, and Pro-75 have all moved apart. Thus, upon reduction of the active site, the parting of a hydrophobic cluster of residues generates a solvent-accessible cavity, at the bottom of which is Glu-79. The cavity is composed of many hydrophobic and some polar/charged residues, i.e. Phe-65, Asn-67, Pro-75, Cys-76, Glu-79, Phe-80, Met-83, Pro-140, Thr-142, Gly-157, Thr-158, Met-159, and Pro-140, and is in part constituted by residues around the active-site cysteines. The fact that the cavity only appears in the reduced state of ResA implies that the protein surface is different in the two redox states, and this is confirmed by surface representations of ResA colored by electrostatic potential (Fig. 5, C and D). In the oxidized form, a patch of hydrophobic residues (Pro-75, Pro-140, Gly-157, and the methyl group of Thr-158) replaces the cavity and shields Glu-79 from solution (Fig. 5A). Consequently, the surface of the oxidized form of ResA is relatively hydrophobic (Fig. 5C). On the other hand, the surface of the reduced state near the active site is relatively basic because of the opening of the cleft, which makes Glu-79 accessible to solvent (Fig. 5, B and D). It should also be noted that the cavity observed in the reduced state is in the immediate vicinity of the active site, being only a few Ångströms away from the partially solvent-exposed Cys-73 thiolate. Access to the Cys-73 S␥ atom is possible from the side of the cleft, but the other side is blocked by Trp-72, the bulky side chain of which can be observed on the right-hand side of the Cys-73 S␥ atom in Fig. 5D. These observations, in combination with the requirement that any redox partner of ResA must be closely juxtaposed to its active site during electron transfer to facilitate formation of a transient mixed disulfide bond (Fig. 3), indicate that the protein surface of ResA may have a major influence on the mode of binding of its redox partners. In particular, the redox-coupled opening and closure of the cleft suggest a rather elegant mechanism by which the redox state of ResA predetermines which substrate, apo-cytochrome c or CcdA, it binds. The fact that the basic cavity is only present in reduced ResA suggests that it may be involved in recognition of the physiological electron acceptor, apo-cytochrome c. The cleft seems to be too small to bind a polypeptide chain, but it is of sufficient size to accommodate an individual amino acid side chain. Such an amino acid would be expected to (i) have both hydrophobic and polar parts to match the character of the cavity, (ii) be spatially close to the Cys-Xxx-Xxx-Cys-His motif, and (iii) be conserved among the apo-cytochrome c sequences of B. subtilis. Taken together, these requirements all point to the histidine of the Cys-Xxx-Xxx-Cys-His motif. Manual positioning of a histidine residue into the pocket with its N⑀ atom in place of the water molecule that is hydrogen-bonded to the carboxyl oxygen of Glu-79 and a subsequent round of unrestrained positional refinement, indicate that a histidine residue would fit securely within the pocket (not shown). The model suggests that the histidine could be reasonably well stabilized by a hydrogen bond between the exposed carboxyl oxygen of Glu-79 and the N⑀ atom of the histidine imidazole group and a van der Waals interaction between the imidazole ring and the hydrophobic residues lining the opening to the cleft. Such a "histidine clamp" mechanism would confer specificity to the recognition of apo-cytochrome c without compromising the capacity to recognize several different sequences (B. subtilis contains four different c-type cytochromes). The histidine clamp would also provide a means of preventing ResA from binding and donating electrons to other thiol-disulfide oxidoreductases, as these proteins do not possess a histidine adjacent to the C-terminal of the active-site cysteines.
The residues involved in redox-coupled conformational changes in ResA are conserved among many CCM-related apocytochrome reductases, and superposition of the oxidized structures of three of such proteins, ResA, CcmG, and TlpA, shows that these parts of the structures are all very similar. By comparison, the polypeptide chains of DsbA and DsbC, both located in the periplasm of E. coli and involved in disulfide bond formation and isomerization, are markedly different in this area (69,70). It thus seems likely that many other CCMspecific proteins will be subject to redox-coupled conformational changes similar to those of ResA, and that the redoxcoupled substrate selection mechanism proposed here may be a general feature of thiol reductases involved in CCM.
From the structures presented here, it is less clear how oxidized ResA might interact with its presumed natural electron donor, CcdA. The surface of CcdA may be more compatible with the smooth hydrophobic surface of oxidized ResA than with the pitted and slightly basic surface of reduced ResA. Moreover, the interaction of ResA and CcdA might somehow be modulated by the central and N-terminal hairpin additions to the classical thioredoxin fold found in ResA and related CCM proteins. These regions adopt different conformations in the two independent molecules of oxidized ResA in the asymmetric unit of the crystals (Fig. 2C), whereas the corresponding parts of the two reduced molecules superpose very well (not shown). The observed flexibility in this region of the oxidized structures may point toward a function in protein-protein interaction whereby this region might become locked in one conformation upon binding of a suitable electron donor. It is interesting to note that the corresponding structural elements of CcmG have been reported to be crucial for CCM in E. coli and were proposed to be involved in the binding of the presumed redox partners of this protein (59).
In conclusion, the crystal structures of the oxidized and reduced states of ResA presented here provide a structural rationale for the strong reducing power of ResA and reveal the largest redox-coupled conformational changes observed so far in a thioredoxin-like protein. These changes alter the protein surface substantially and lead to the appearance of a polar cavity in the reduced state. On the basis of these results it seems likely that the redox state of ResA predetermines to which substrate it binds. It is speculated that the hydrophobic cavity may provide specificity in apo-cytochrome recognition, and that this may be a general mechanism in bacterial cytochrome c maturation systems.