Functional Irreplaceability of Escherichia coli and Shewanella oneidensis OxyRs Is Critically Determined by Intrinsic Differences in Oligomerization

ABSTRACT LysR-type transcriptional regulators (LTTRs), which function in diverse biological processes in prokaryotes, are composed of a conserved structure with an N-terminal DNA-binding domain (DBD) and a C-terminal signal-sensing regulatory domain (RD). LTTRs that sense and respond to the same signal are often functionally exchangeable in bacterial species across wide phyla, but this phenomenon has not been demonstrated for the H2O2-sensing and -responding OxyRs. Here, we systematically examined the biochemical and structural determinants differentiating activator-only OxyRs from dual-activity ones by comparing OxyRs from two Gammaproteobacteria, Escherichia coli and Shewanella oneidensis. Our data show that EcOxyR could function as neither an activator nor a repressor in S. oneidensis. Using SoOxyR-based OxyR chimeras and mutants, we demonstrated that residues 283 to 289, which form the first half of the last C-terminal α-helix (α10), are critical for the proper function of SoOxyR and cannot be replaced with the EcOxyR counterpart. Crystal structural analysis reveals that α10 is important for the oligomerization of SoOxyR, which, unlike EcOxyR, forms several high-order oligomers upon DNA binding. As the mechanisms of OxyR oligomerization vary substantially among bacterial species, our findings underscore the importance of subtle structural features in determining regulatory activities of structurally similar proteins descending from a common ancestor.

R eactive oxygen species (ROS), including superoxide (O 2 2 ), hydrogen peroxide (H 2 O 2 ), and hydroxyl radical (ÁOH), can damage biomolecules such as DNAs, proteins, and lipids (1). In living organisms, oxidative stress caused by ROS is inevitable because they can be generated endogenously as metabolic by-products of cellular oxygen respiration in addition to those coming from environments (1). Basal defenses in bacteria, composed of mostly ROS-scavenging enzymes, are sufficient to cope with ROS formed during routine aerobiosis. However, when intracellular levels of ROS exceed safe limits due to exogenous contribution, oxidative stress sensing and responding systems are activated to coordinately regulate expression of a set of genes to ensure that ROS concentrations are restrained at an acceptable level and damages are promptly repaired (2). The primary members of these genes encode ROS detoxification enzymes (catalases, superoxide dismutase, and various peroxidases), iron-sequestering proteins, and damage control proteins (3).
OxyR, one of the major ROS-sensing and -responding systems identified 35 years ago in enteric bacteria Escherichia coli and Salmonella enterica serovar Typhimurium (S. Typhi), is an LysR-type transcriptional regulator (LTTR) that is characterized by having an N-terminal DNA-binding domain (DBD) and a C-terminal regulatory domain (RD) (2,4,5). As best illustrated in E. coli, OxyR (EcOxyR) becomes activated when a disulfide bond is formed between two conserved cysteine residues located in the RD as a result of oxidization by H 2 O 2 (6)(7). The structural changes in the RD induced by the disulfide bond formation lead to conformational rearrangement of the DBD with an altered DNA-binding affinity (8)(9). Because EcOxyR functions as an activator of its regulon only, we refer to it as a type I OxyR in this study (8).
Although it is a type I OxyR that was initially identified and studied, further investigations into its homologs from other bacteria revealed surprising variations in their functional modes. OxyRs of corynebacteria function as a repressor only (type III) for more than 20 genes, including those for ROS detoxification enzymes and iron-sequestering proteins (10,11). Most OxyRs belong to the dual-control (type II) group, acting as not only an activator of peroxide-scavenging enzymes under oxidative stress conditions but also a repressor of the same target genes under nonstress conditions. Type II OxyRs occur in a large variety of bacteria, such as Shewanella, Pseudomonas, Neisseria, Xanthomonas, and Deinococcus, to name a few (12)(13)(14)(15)(16)(17).
Intriguingly, despite the differences in functional modes, OxyRs characterized to date are similar in overall structure and recognize similar DNA motifs composed of two tandem ATAG-N 7 -CTAT (N represents any nucleotide) repeats with a 7-to 10-bp interval (4,9,15,18,19). Moreover, all OxyRs examined to date, regardless of their regulatory effects, are capable of binding to promoter regions of their target genes in both reduced and oxidized forms (9,15,19,20). It has been proposed that as a repressor (in the reduced state), OxyRs bind to a more extended region in proximity of the core DNA motifs than in the oxidized state, thus occluding RNA polymerase binding (21). However, given that OxyRs in reduced and oxidized forms coexist in the cell, recent reports have provided evidence to suggest that OxyRs in both redox states interact with the same DNA sequence but differ from each other in binding affinity (15,19).
Despite overall similarities in sequences, structures, and activation mechanisms, bacterial OxyRs are generally not interchangeable, with exceptions of the same type from the same or closely related species (5,15). While it has been suggested that the functional irreplaceability is a result of intrinsic structural differences among OxyR orthologues (15,22), up until now, little is known about the underlying mechanisms.
In this study, we attempted to unravel mechanisms for functional irreplaceability by carrying out comparative analyses of type I EcOxyR and type II OxyR of Shewanella oneidensis, a representative of a large group of Gram-negative facultative Gammaproteobacteria renowned for their respiratory versatility and the potential application in biogeochemical circulation of minerals and bioelectricity (23,24). We demonstrate that EcOxyR could function neither as an activator nor a repressor in S. oneidensis. The crystal structure of the SoOxyR RD in its reduced form shows an antiparallel dimer similar to OxyRs from E. coli and other bacterial species. It is also observed that SoOxyR RD dimers further interact through an a-helix at the Sun et al. C terminus to form tetramers and other high-order oligomers. Indeed, SoOxyR/EcOxyR chimeras and SoOxyR mutants demonstrated that the last a-helix of SoOxyR is essential for its proper regulatory activity. Furthermore, DNA gel shift assays indicated that SoOxyR has a much stronger tendency than EcOxyR to form oligomeric assemblies, which is presumably due to cooperative DNA binding mediated by the last a-helix as revealed in the crystal structure of SoOxyR. Overall, our findings provide a mechanistic explanation for the functional nonexchangeability between SoOxyR and EcOxyR, and they highlight the importance of OxyR oligomerization, the mode of which may vary widely among related bacterial species, on the regulatory activity of these transcriptional regulators.

RESULTS
EcOxyR has no physiological activity in S. oneidensis. We have previously shown that both S. oneidensis and E. coli oxyR null mutants exhibit severe plating defects on LB plates (substantially impaired viability) (17,25) (Fig. S1A in the supplemental material). Functional nonexchangeability between SoOxyR and EcOxyR was demonstrated, as they failed to reciprocally complement the phenotypes of the opposite oxyR mutant (17). Unlike type II SoOxyR, which functions as both repressor and activator for some of its regulon members such as katB, type I EcOxyR could not activate expression of these genes (17,19). Given that the repressing activity of OxyRs could not be detected from cell viability when growing on agar plates, we set out to determine whether EcOxyR could repress expression of the katB gene.
DNA fragments for both EcOxyR and SoOxyR, as well as all OxyR variants used in this study, were amplified and cloned into integrative vector pHGM01 (26). Protein constructs for OxyR variants tested in this study, including truncations, chimeric hybrids, and point mutants, are shown in Fig. 1, with additional information given in Table S1. Throughout the study, point mutations were presented in subscript, and all others were presented in superscript. The verified vectors were then introduced into the DSooxyR strain for chromosomal integration, resulting in strains with the oxyR variants under the control of the oxyR promoter (P oxyR ). In this way, all OxyR variants under test are supposed to be produced at levels similar to that of OxyR in the wild type (WT). Indeed, expression of EcOxyR and SoOxyR was found to be comparable by using fusion proteins with green fluorescent protein (GFP) linked to the C terminus, which could be detected in the cytoplasm by confocal microscopy ( Fig. S1B and C). Because WT and the DoxyR strains of S. oneidensis and E. coli expressing a copy of their own oxyR gene integrated into the chromosome, namely, DSooxyR/pSoOxyR and DEcoxyR/ pEcOxyR, respectively, were indistinguishable from each other in all experiments (Fig. S1A), only data for the DSooxyR/pSoOxyR strain (regarded as WT) are presented ( Fig. 2A).
We then compared katB expression levels in DSooxyR cells expressing either SoOxyR or EcOxyR with integrative lacZ reporters used before (19). As expected, DSooxyR/pSoOxyR exhibited a repressing effect on katB expression in normal growing cells up to the mid-exponential phase and substantially elevated expression levels after cells were challenged by H 2 O 2 (Fig. 2B). In the absence of SoOxyR, the katB gene was expressed at levels between the repressed and the activated caused by the regulator. Clearly, DSooxyR/pEcOxyR neither exhibited an H 2 O 2 -responsive effect nor showed any repressing activity in S. oneidensis (Fig. 2B). This was not due to the shortage of the reduced proteins because EcOxyR C199S , a sensory residue point mutant locked in the reduced form, exhibited the same effect (Fig. 2B). These observations were supported by catalase staining analysis of cells prepared similarly, as KatB is the only catalase detectable by the method (17). Neither EcOxyR nor EcOxyR C199S could affect KatB levels in DSooxyR cells (Fig. 2C). Furthermore, the effects of EcOxyR and EcOxyR C199S on H 2 O 2 degradation of the DSooxyR strain were assessed. Cells expressing proteins of interest at the mid-exponential phase were collected and disrupted by sonication in order to avoid interference of H 2 O 2 induction on catalase expression and of peroxidases which require electron transport. The resultant cell extracts were aliquoted, adjusted to contain the same amount of protein, and mixed with 1 mM H 2 O 2 for assaying the remaining H 2 O 2 in the reaction at the indicated time points. In line with the failure of suppressing the plating defect, neither EcOxyR nor EcOxyR C199S had a significant impact on H 2 O 2 degradation (Fig. 2D). All together, these data indicate that EcOxyR does not have detectable repressing and activating activity in S. oneidensis.
DBD of SoOxyR is essential for repressing activity in S. oneidensis. Given that the regulatory activity of OxyR is ultimately realized by interaction between the DBD domain and its target DNAs, we set out to determine whether the DBD domain could function to some extent on its own. For simplicity, SoOxyR and EcOxyR fragments were presented in regular and underlined superscript, respectively; for example, OxyR DBD (truncated mutation) and OxyR DBD represent the DBD domain of SoOxyR and EcOxyR, respectively (Fig. 1). Characterization of OxyR DBD and OxyR DBD demonstrated that the DBD domain of OxyRs alone does not possess any regulatory activity in S. oneidensis ( Fig. S2A and B).
To investigate the mechanism for functional differences between SoOxyR and EcOxyR, we constructed hybrid protein OxyR DBD-RD (SoOxyR DBD and EcOxyR RD) (Fig. 1). Clearly, OxyR DBD-RD did not elicit significant difference in viability (Fig. S2C), indicating that it could not function as an activator for the SoOxyR regulon. Indeed, catalase Residues that are redox active and act as a location marker were shown. There were two parameters for activity, physiological impacts and response to H 2 O 2 . The former is represented by "Both" (repressing and activating), "No" (no effect), "Rep" (repressing), and "Act" (activating), while the latter is represented by positive (1) and negative (2) symbols. Point mutation variants are given in Table S2 in the supplemental material. Sun et al. staining revealed that catalase production was not induced in cells expressing this hybrid OxyR when challenged by H 2 O 2 (Fig. 3A). Moreover, the KatB levels in DSooxyR cells producing OxyR DBD-RD were indistinguishable from that of WT, suggesting that OxyR DBD-RD could function as a repressor (Fig. 3A). This notion was supported by results from the expression assay ( Fig. 3B): OxyR DBD-RD repressed katB expression and was not responsive to H 2 O 2 treatment, and results from H 2 O 2 degradation assay ( Fig. S2D) revealed reduced H 2 O 2 removal rates for the DSooxyR strains producing OxyR DBD-RD . Thus, OxyR DBD-RD can function as a repressor but not an activator for expression of the SoOxyR regulon. Hybrid protein OxyR DBD-RD (E. coli DBD and S. oneidensis RD) was then constructed for cross-examination (Fig. 1). OxyR DBD-RD was not responsive to H 2 O 2 as expected (Fig. 3A) and could not correct the plating defect of either DSooxyR or DEcoxyR strains (Fig. S2C). Moreover, this protein differed from OxyR DBD-RD in that it did not repress KatB production ( Fig. 3A and B; Fig. S2D). These data altogether indicate that the SoOxyR DBD domain is essential for the repressing effect of SoOxyR, implying that the  Cultures of indicated strains prepared to contain approximately 10 9 CFU/mL were regarded as the undiluted (dilution factor, 0), which were subjected to 10-fold serial dilution. Five microliters of each dilution was dropped onto LB plates. Results were recorded after 24 h of incubation. Expression of the oxyR genes was driven by the SoOxyR promoter. pOxyR and pOxyR Ot represents that each strain expresses oxyR of its own or of the other's. (B) Impacts of OxyRs on expression of katB by using integrative lacZ reporters. Cells at the midexponential phase were used for all assays unless otherwise noted. Cells directly taken as untreated (Un) and incubated with 0.2 mM H 2 O 2 for 2 min as treated (T). Asterisks indicate statistically significant difference of the values compared (n = 4, *, P , 0.05; **, P , 0.01; ***, P , 0.001). (C) Catalase detected by staining and activity assay. Cells were either directly used or incubated with 0.2 mM H 2 O 2 for 30 min. Cell lysates containing the same amount of protein were subjected to 10% nondenaturing PAGE. Catalase (KatB) was revealed by catalase staining as described in Materials and Methods. (D) For H 2 O 2 degradation assay, cells were adjusted to the same optical density and disrupted by sonication. The resultant cell extracts were mixed with 1 mM H 2 O 2 and assayed for the remaining H 2 O 2 in the reaction mixture at the indicated time points, which was normalized to give relative amounts to the original (100%). Experiments were performed at least three times or specified as in panel B, with representative results (A and C) or the average 6 error bars representing standard deviation being presented (D). mechanisms underlying the functional difference between type I and type II OxyRs are more profound than expected.
The crystal structure of SoOxyR C203S RD reveals a reduced redox center and potential to oligomerize. To better understand why EcOxyR cannot function as a replacement of SoOxyR, we determined the crystal structure of the RD C203S of SoOxyR, which should be locked in the reduced form with a single point mutation C203S (17), as the mutant could no longer form the disulfide bond C203-C212. The crystal structure of RD C203S of SoOxyR was solved to a resolution of 2.4 Å by molecular replacement ( Fig. 4A; Table S2). The RD C203S of SoOxyR polypeptide chain, which contains residues 91 to 304 of the full-length protein, was traced unambiguously except for the last four residues at the C terminus and a six-residue loop from amino acids (aa) 179 to aa184 that were structurally disordered. Each RD C203S molecule is made of six a-helices and eight b-strands that fold into two domains, RD-I (residues 91 to 164 and 274 to 304) and RD-II (residues 165 to 273). RD-I is made of a central five-stranded b-sheet with two a-helices stacked against one face and one a-helix against the other. RD-II, which hosts the redox center, is comprised of a central three-stranded b-sheet that is surrounded by three loosely organized a-helices. RD-I and RD-II are connected by two loop linkers comprised of residues 165 to 168 and 264 to 273, with an extensive interdomain interface that is primarily mediated by two helices (i.e., a7 and a9) from RD-II and several loops from RD-I ( Fig. 4A and B). Multisequence alignment indicates that RD-I is better conserved within the LysR family in primary sequences compared to RD-II (Fig. 4B). It is also evident that RD-II has a much higher fraction of structured loops (Fig. 4B). According to a DALI search, the structure of the reduced S. oneidensis OxyR RD monomer is closely related to those of the reduced OxyR2 of Vibrio vulnificus (PDB ID 5X0V; Z score, 31.5; root mean square deviation [RMSD], 1.2 Å), the Pseudomonas aeruginosa full-length OxyR C199D (PDB ID 4X6G; Z score, 23.5; RMSD, 2.2 Å), the reduced P. aeruginosa OxyR RD (PDB ID 4Y0M, Z score, 23.5; RMSD, 2.4 Å), the RD C199S of EcOxyR (PDB ID 1I69, Z score, 22.3; RMSD, 2.3 Å), and the RD of reduced Neisseria meningitides OxyR (PDB ID 3JV9; Z score, 21.6; RMSD, 2.3 Å), with no major differences in conformation.
As expected for a reduced form, the disulfide residue pair C203S and C212 in SoOxyR RD are found at the opposite ends of the helix a7, with the 2OH and the 2SH side chain measured to be 13 Å apart from each other. The side chains of both C203S and C212 point away from solvent by tucking into cavities that are partially polar for C203S but completely nonpolar for C212. Based on the structural context, the thiol side chain of residue 203 in a wild-type protein is likely to be more accessible for oxidative modification than C212. In the reduced EcOxyR RD, the side chain of C199S is also tucked in as in SoOxyR RD, but the side chain of C208 is exposed to solvent due to a sharp bend in the middle of the helix connecting the redox pair residues (27) (Fig. S3A to C). It is worth noting that the positioning of the two redox pair residues in the VvOxyR2 RD, also in the reduced form, is almost identical to SoOxyR (27). Like VvOxyR2, SoOxyR has an -EH-dipeptide in front of C203 instead of a -GH-found in EcOxyR. A previous study on VvOxyR2 indicated that the glutamic acid residue is able to enhance H 2 O 2 sensitivity (27). The active site residues surrounding C203S, including T104, T133, H202, and R271, are also conserved in SoOxyR as in VvOxyR2 (11).
There are six SoOxyR RD C203S molecules in each crystallographic asymmetric unit (Fig. S4A). Using the PDBePISA server (https://www.ebi.ac.uk/msd-srv/prot_int/cgi-bin/piserver), which considers both noncrystallographic and crystallographic interactions, a series of intermolecular interfaces are identified (Table S3). Through interface-I, RD C203S monomers assemble into dimers with an average buried surface area of 1,268 Å 2 . This dimer interface, which is primarily mediated by a5, loop aa219-223 and a8 ( Fig. 4C; Fig. S4B), is commonly observed in other bacterial OxyR RD crystal structures (9,18). The second-largest interface (i.e., interface II), which buried a surface area of 561 Å 2 on average, further assembles RD C203S dimers into tetramers or dimers of dimers. Interface II is mediated by a10 located at the very C-terminal end of the protein ( Fig. 5A; Fig. S4C). a10 from two neighboring subunits pack against each other in an antiparallel manner, forming an interface that is largely hydrophobic in nature ( Fig. 5A and D). The SoOxyR RD C203S tetramer arrangement differs considerably from the tetrameric structure of the full-length PaOxyR (PDB ID 4X6G) in which the second dimer interface is mediated by the DBD domain. Nevertheless, the C-terminal a-helix of PaOxyR is implicated in high-order molecular interactions in both full-length C199D and the reduced PaOxyR RD crystal structures ( Fig. 5B and C), suggesting that the C-terminal a-helix may also play a biological role in OxyR oligomerization and gene regulation. The C-terminal a-helix in the two PaOxyR structures makes parallel interaction, whereas this C-terminal a-helix makes antiparallel interactions in the SoOxyR RD C203S structure. It is worth noting that of the six RD C203S molecules in each crystal asymmetric unit (Fig. S4A), molecular pairs A and C, B and D, and E and F can each assemble into an infinitely long helical fiber through alternating interface I and interface II dimer interactions around the 3 2 crystallographic symmetry (Fig. 5E).
In addition to the two interfaces mentioned above, interface III, ;479 Å 2 in size, creates another dimer through interactions mediated primarily by the redox helix (i.e., helix a7) (Fig. S4D). Because interface III is smaller than the first two, and also because this interaction is only observed between two out of the six molecules in a crystal asymmetric unit, we consider it not stable and, therefore, not biologically important. Other interfaces identified by PDBePISA are increasingly weaker and also asymmetric in nature and, therefore, were not considered further in this study.
Fragmentation effect of the RD domains of SoOxyR and EcOxyR for functional exchangeability. Given the overall structural similarity in the RD domain of OxyRs, we attempted to determine the maximal length of the SoOxyR RD domain that could be replaced by its E. coli counterpart without affecting its regulatory activity. Based on the structure comparison, the following 7 hybrid proteins were constructed without disrupting secondary structure elements: OxyR S132E (chimera protein, SoOxyR and EcOxyR sequences before and after residue 131, respectively; between b2 and a6), OxyR S185E (between b5 and a7), OxyR S220E (between h 2 and a8), OxyR S245E (between b6 and a9), OxyR S270E (between b7 and b8), OxyR S280E (between b8 and a10), and OxyR S298E (after a10) ( Fig. 1; Table S1). Among these chimeric OxyRs, only OxyR S298E displayed full activity of SoOxyR ( Fig. 6A and B; Fig. S5A), suggesting that the vast majority of the RD domain contributes to functional nonexchangeability. The ability of OxyR S298E to respond The continued polymerization of SoOxyR dimers through the RD domain could lead to the formation of tetramers, hexamers, octomers, and so on. These RD dimers are related by a crystallographic 3 2 symmetry axis as indicated. a10, which plays a critical role in mediating polymerization, is highlighted in magenta. Residue 91, which is directly connected to the DBD domain, is shown as red spheres. Therefore, dimerization by a10 helps to bring two DBD domains into proximity to facilitate DNA binding. Sun et al. ® to H 2 O 2 was confirmed by catalase staining (Fig. S5B), implying that the fragment between residues 280 and 298 is crucial for the activating effect of SoOxyR. Although all other chimeras were either nonresponsive to H 2 O 2 or unable to function as an activator in S. oneidensis, they could be divided into two groups. The first three, OxyR S132E , OxyR S185E , and OxyR S220E , showed some repressing activity, albeit not as robust as SoOxyR. Clearly, the shorter the E. coli fragments in the chimeras, the less the repressing effect. Despite this, their impacts on viability of DSooxyR cells were not apparent (Fig. S5A). The next three, OxyR S245E , OxyR S270E , and OxyR S280E , exhibited repression even stronger than SoOxyR, similar to OxyR DBD-RD (Fig. 6B). Consistently, OxyR S245E , OxyR S270E , and OxyR S280E further sensitized cells of DSooxyR on the LB agar plates (Fig. S5A), clearly due to lowered catalase production (Fig. S5B). This observation suggests that the fragment of EcOxyR RD after residue 245 introduces an impact on repressing activity of OxyR stronger than its SoOxyR counterpart. Thus, the RD domains of SoOxyR and EcOxyR appear to affect activity in a manner of fragmentation: the RD domain is composed of multiple fragments, and each of them primarily associates with a specific activity, altogether amounting to the terminal effect of OxyR.
Impacts of residues within the last a-helix of SoOxyR. The substantial difference in activity between OxyR S280E and OxyR S298E prompted us to focus on a10, which is implicated in tetramer assembly based on the crystal structure of the SoOxyR RD. A serial of truncations of SoOxyR was generated, from SoOxyR D2832 (truncation lacking all residues after R283) to SoOxyR D2952 , with each increasingly longer by 3 residues (Fig. 1; Table S1). Characterization of these truncations revealed that SoOxyR D2952 and OxyRs of Bacteria ® SoOxyR D2922 functioned indistinguishably from the whole protein in suppressing the plating defect of the DSooxyR strain. In contrast, the remaining three, SoOxyR D2832 , SoOxyR D2862 , and SoOxyR D2892 , did not show any significant improvement (Fig. S6A). Interestingly, there were variations in katB expression in cells with these OxyR truncations. While the activating effect of SoOxyR D2952 was the same as the intact protein, SoOxyR D2922 showed activity of approximately 80% (Fig. 6B). Although SoOxyR D2832 , SoOxyR D2862 , and SoOxyR D2892 were unable to fully activate expression of the katB gene, they exhibited H 2 O 2 -responding ability, which increased with the length of the mutated proteins. In attempts to narrow down the sequence region crucial to activity of SoOxyR, we tested SoOxyR D2902 and found that it was able to correct the plating defect while retaining 65% of activating capacity ( Fig. 6B; Fig. S6A). These results suggest that residues from 283 to 289 are crucial for activating the function of OxyRs, while residues after 291 are dispensable. Consistent with this finding, the crystal structure of SoOxyR RD shows that residues 284 to 290 from a10 are implicated in a dimerdimer interaction that is critical for SoOxyR polymerization (Fig. 5A and E).
To verify this, we first used OxyR chimeras covering residues from 286 to 289 ( Fig. 1; Table S1). Residues from 286 to 289 in SoOxyR have the sequence of 286 TFRT 289 , which is aligned to 281 LYEQ 284 in EcOxyR. While OxyR S286E and OxyR S287E did not show a complementary effect on the plating defect, OxyR S288E and OxyR S289E conferred cells complete suppression (Fig. S6A). The katB promoter activity assays demonstrated that OxyR S286E and OxyR S287E could still function as a repressor in cells grown normally and, at the same time, had some ability to respond to exogenous H 2 O 2 (Fig. 6B). In contrast, OxyR S289E was the same as SoOxyR, but OxyR S288E appeared modestly impaired in activating katB expression. These data again testified the essential role of 286 TFRT 289 , especially the first three residues, in the proper functioning of SoOxyR, presumably by maintaining the correct conformation of a10 that allows tetramer formation.
To further verify the importance of residues in the proximity of R288, we carried out point mutational analyses of SoOxyR. To begin with, we mutated R288 to E as in E. coli and P. aeruginosa (Fig. 4B). The resulting SoOxyR R288E (point mutation) was indistinguishable from SoOxyR in terms of both viability and katB expression ( Fig. 6B;  Fig. S6A). Then, alanine scanning was conducted for residues from 285 to 291. None of these SoOxyR mutants was significantly different from SoOxyR in functionality ( Fig. S6A and B), implying that single mutations in this region are tolerable. However, when the R288E mutation was combined with any of these alanine mutations, activity was affected substantially, depending on residues. Only one double mutant, SoOxyR R288E-T289A , exhibited the characteristics of SoOxyR, whereas all others had impaired activity ( Fig. 6B; Fig. S6A and B). Among these, SoOxyR R288E-F287A and SoOxyR R288E-L290A were reduced to ;30%, while the remaining three retained at least 60% of the wild-type activity. The crystal structure shows that R288 is at the tetramer interface, and the aliphatic portion of the R288 side chain stacks against the side chain of L285, forming part of a large hydrophobic patch that connects the two neighboring dimers together (Fig. 5D). Considering the important role of R288 in tetramer formation, it is possible that although a single mutation, R288E, can be tolerated, the simultaneous mutation of a neighboring hydrophobic residue, either F287A or L290A, would result in disruption of SoOxyR tetramers.
Helix 10 is critical for oligomerization upon DNA binding. It is well-known that OxyR mutants impaired in DNA binding, oligomerization, or disulfide bond formation lack transcriptional activity (28). Because the last helix of OxyRs appears to be important for oligomerization based on structure analysis, we tested DNA binding of representative OxyR mutants studied above. Recombinant OxyR mutants with hexahistidine (His 6 ) tag at the N terminus were expressed in E. coli and purified by Ni 21 affinity chromatography as before (19). Electrophoretic mobility shift assay (EMSA) results showed that both SoOxyR and EcOxyR were able to interact with the katB promoter, contrasting with the gyrB promoter used as the negative control (Fig. 7). Negative results were also obtained from the DBD domain of SoOxyR and the target DNA fragment, indicating that they do not interact with each other specifically. Clearly, SoOxyR differs from EcOxyR in that it generates supershift bands (Fig. 7), which represent DNA oligomer (i.e., tetramer, hexamer, octamer, etc.) complexes (19). Together with the gel filtration results (Fig. S7), this observation indicates that SoOxyR oligomerizes much more effectively than EcOxyR upon binding to the katB promoter. Moreover, we observed a significant difference in EMSA results from OxyR DBD-RD and OxyR DBD-RD , composed of DBD and RD from different bacteria (Fig. 7). The former, which functions as a repressor constitutively, displayed substantially impaired binding capacity, a scenario in line with the previous finding that OxyR in the activating form has a higher affinity for target DNAs than that in the repressing form (18). The latter not only was weak in binding but also impaired in oligomerization, which was supported by the data from gel filtration (Fig. S7). These observations were generally supported by results from three representative chimeric OxyR mutants, OxyR S185E , OxyR S280E , and OxyR S298E , which have no activity, repressing activity only, and full activity, respectively (Fig. 6B).
The three OxyR variants carrying mutations in the last a-helix mutants, OxyR D2862 , OxyR D2902 , and OxyR R288E1F287A , were then examined by EMSA. The former one differed from the latter two in that it could not function as an activator and retained marginal capacity for response to H 2 O 2 (Fig. 6B). Consistently, the latter two exhibited substantially higher DNA affinity (Fig. 7). Notably, OxyR R288E1F287A failed to bind to DNA probes at 100 nM (Fig. 7), an observation consistent with its reduced activating activity (Fig. 6B). More importantly, despite these differences, all of these OxyR mutants could not form octamer (Fig. 7), strongly supporting that the last helix plays an important role in oligomerization.
The crystal structure of SoOxyR RD C203S and the observation of DNA supershifts led us to a cooperative DNA-binding model by SoOxyR (Fig. 5E). Through dimerization mediated by a10, SoOxyR dimers can further oligomerize into tetramers, hexamers, octamers, or even larger linear complexes. In these large linear complexes, SoOxyR dimers are related to each other by 3 2 screw rotation symmetry as observed in the crystal structure of SoOxyR RD C203S . Interestingly, two DBD domains from two adjacent SoOxyR dimers would be brought into close proximity according to this supramolecular assembly model. EcOxyR may utilize a different mechanism for the activating mode of DNA binding, thus explaining the lack of supershifts. Consistent with this model, SoOxyR mutants with reduced or no activating activity did not produce supershifts in the DNA-binding assay.

DISCUSSION
The purpose of this study was to unravel the mechanism underpinning the functional nonexchangeability of EcOxyR and SoOxyR, representatives of type I and II FIG 7 Impacts of residues within the last a-helix of SoOxyR. In vitro interaction of His-tagged OxyR variants and the katB promoter sequence revealed by using EMSA. His 6 -tagged OxyR variants were expressed in E. coli, and proteins in soluble fractions were purified by Ni 21 affinity chromatography. The digoxigenin-labeled DNA probes of 188 bp that cover the OxyR-binding motif were prepared by PCR. The EMSA was performed with 10 nM probes and various amounts of proteins as indicated. The shift bands without arrow, with red arrow, and with blue arrow represent dimer, tetramer, and octamer, respectively. Experiments were performed at least three times with representative results being presented.
OxyRs of Bacteria ® OxyRs, respectively. Given identical activating mechanisms and considerable similarity in amino acid sequences and overall structures, we anticipated that some short fragments, at least some residues, were likely responsible for the functional difference. We were surprised when it turned out not to be the case. The segment-swapping analyses indicated that the differences between EcOxyR and SoOxyR appeared to be comprehensive and profound. Based on our analyses using truncation and point mutations, we conclude that the functional irreplaceability of SoOxyR by EcOxyR cannot be easily resolved by point or short-segment swapping mutations. These results underscore the need to test more OxyRs for their ability to take the role of their counterparts in other bacteria.
EcOxyR lacks repressing activity for its H 2 O 2 -responding target genes (2,6). Given that OxyR proteins in the reduced and oxidized forms are present at the same time in the cell and repression is carried out by reduced OxyR proteins, one may imagine that EcOxyR proteins in the reduced form were not mounted to levels sufficiently high to block transcription in S. oneidensis. However, the failure of EcOxyR C199S to repress katB expression eliminates this possibility. Similar scenarios about type II OxyR have been reported before in bacteria such as P. aeruginosa and Neisseria meningitidis, highlighting intrinsic differences between OxyR analogues possessing activator-only and dualcontrol activities (15,29). It has been suggested that OxyR proteins may exist in vivo as a number of potential reaction intermediates to disulfide formation, including S-OH, S-NO, and S-SG on C199 of EcOxyR, offering additional options beyond an on/off redox regulation between oxidized (S-S) and reduced (S-H) forms (30). These intermediates may have different activities, resulting in a hierarchical response and regulation on the regulon members. As a result, repression of catalase may have something to do with a graded response to the stress. However, until now, the model of intramolecular disulfide bond formation prevails and has been widely accepted based on enormous amounts of evidence (31).
We showed here that the SoOxyR DBD domain is essential for repression, an observation somewhat surprising because of relative higher sequence identity of the DBD domain (45% versus 35% full length) and highly similar DNA motifs for EcOxyR and SoOxyR. EcOxyR and chimeric OxyR DBD-RD (E. coli DBD) could not affect regulation, although they are able to interact with target DNAs. Despite this, the RD domain of EcOxyR converts the DBD domain of SoOxyR into a repressor. It should be noted that only a small share of the type II OxyR regulon is subjected to dual-activity regulation, indicating that type II OxyRs are able to differentiate the promoters for these regulon members from the rest of the regulon despite the high similarity of all DNA sequences to which they bind. For example, there are only 2, 1, and 1 genes under dual regulation in S. oneidensis (katB and dps encoding iron sequestering protein), P. aeruginosa (katA encoding major catalase), and N. meningitidis (kat encoding major catalase), respectively (15,17,29). This convinces us that activation is likely the core functioning mechanism of OxyRs, but evolution has elegantly honed these regulators for their roles in differentially mediating expression of genes containing similar DNA motifs in order to adapt to environments where they thrive.
By solving the RD domain structure of SoOxyR, we identified significant differences between SoOxyR and two other OxyRs, EcOxyR and PaOxyR, whose structures are well defined (9,11,18). EMSA results revealed that all OxyR mutants except those composed of the DBD domain only are able to interact with the target DNA fragment, suggesting that these proteins can properly dimerize. Clearly, this is in perfect agreement with the structural data. However, during formation of dimers of dimers, SoOxyR displays some unique features. The overall tetramer arrangement of SoOxyR appears to be different considerably from that of EcOxyR and PaOxyR. We speculate that these differences may result in distinct oligomerization status of SoOxyR revealed in EMSA. In addition to dimer and tetramer forms, a significant share of SoOxyR appears to be assembled into octamer (Fig. 7). As full activating activity is only observed from mutants capable of forming octamer (Fig. 7), it is reasonable to assume that this type of oligomerization is critical for SoOxyR to act as an activator. It is worth mentioning that OxyR of P. aeruginosa (a gammaproteobacterium per se) is an abnormal type II OxyR because it is phylogenetically clustered with betaproteobacterial OxyRs rather than with those from Gammaproteobacteria (22,29). Moreover, the last residue of the C-terminal a-helix within PaOxyR is cysteine, and more importantly, this Cys residue has been suggested to be involved in peroxide sensing in P. aeruginosa (22). Thus, PaOxyR is not a genuine type II OxyR defined in this study, and its deviation from SoOxyR may not be sufficient to support the link between physiological function and structure.
Perhaps the most important finding of the structural analysis is revelation of involvement of the C-terminal a-helix in oligomerization. The loss of the entire a10 (OxyR D2832 ) almost completely disabled SoOxyR to respond to H 2 O 2 . However, SoOxyR without half of this a-helix (OxyR D2892 ) retains H 2 O 2 response and repressing, but not activating, activity. In addition, while mutations in a10 impair both repressing and activating activities, they do not abolish the ability to respond to H 2 O 2 . Thus, a10 may not be directly associated with conformational changes induced by disulfide bond formation. EMSA results show that regardless of regulatory effects, SoOxyR variants carrying mutations in this region lose the signature oligomerization pattern observed from the wild-type SoOxyR, confirming the importance of a10 in assembly of tetramer and octamer. Notably, SoOxyR variants that display activating activity are largely present in tetramer only. While this further validates the role of a10 in oligomerization, it indicates that octamer complexation is not absolutely required for activating activity.
It is also important to note that our oligomerization model (Fig. 5E) and the tetramer model proposed for PaOxyR (18) and the C. glutamicum OxyR (CgOxyR) (11) are not mutually exclusive. In the crystal structure of full-length PaOxyR and CgOxyR, two RD-interfaced homodimers further dimerize through their DBD interface to form an asymmetric tetramer. It is possible that each OxyR dimer from our linear oligomer model (Fig. 5E) is actually part of a tetramer like those observed in the PaOxyR and CgOxyR crystal structures. Therefore, through the C-terminal a-helix, OxyR tetramers can form bigger assemblies, such as octamers, dodecamers, etc. Indeed, octamers were observed for full-length CgOxyR in solution when the protein was in the oxidized state (11). Interestingly, both full-length PaOxyR and CgOxyR tetramers were found to dimerize through their C-terminal a-helices in crystal.
Gene repression by type II OxyRs in bacteria has been investigated before, but the physiological relevance as to this phenomenon remains elusive (15,29,32). In addition to catalase, iron-sequestering protein Dps is also under dual control of OxyR in S. oneidensis (17). Given that S. oneidensis is renowned for unusually high respiratory versatility because of particular richness and abundance in iron-containing proteins, for instance, more than 40 c-type cytochromes, the speculation is that OxyR downregulates production of iron-containing catalase and Dps when cells are not challenged by H 2 O 2 (19,33,34). In this way, the biosynthesis of iron proteins involved in metabolism and electron transport responsible for respiratory versatility gains priority so as to maintain fitness in environments (35,36). Given that the ultimate output of transcriptional regulation is realized by interaction between regulators and their target DNAs, this complex dual control presents an extremely elaborated maneuver for tuning gene expression in response to environmental cues in different prokaryotic cells. To this end, our study represents an important step toward a better understanding of the mechanisms underlying the different functional modes of the OxyR proteins.

MATERIALS AND METHODS
Bacterial strains, plasmids, and culture conditions. All bacterial strains and plasmids used in this study can be found in Table S1 in the supplemental material. Information about all of the primers is available upon request. All chemicals were obtained from Sigma-Aldrich unless otherwise noted. E. coli and S. oneidensis were grown in lysogeny broth (LB, Difco, Detroit, MI) under aerobic conditions at 37 and 30°C for genetic manipulation. When necessary, the growth medium was supplemented with chemicals at the following concentrations: 0.3 mM 2,6-diaminopimelic acid (DAP), 50 mg/mL ampicillin, 50 mg/mL kanamycin, 15 mg/mL gentamicin, 100 mg/mL streptomycin, and 2,000 U/mL catalase on plates.
OxyRs of Bacteria ® Growth in liquid medium was monitored by recording values of optical density at 600 nm (OD 600 ), as all strains used in this study were morphologically similar. Both LB and defined medium MS (25) were used for phenotypic assays in this study, and comparable results were obtained with respect to growth.
Knock-in and expression of oxyR variants. The mutagenesis procedure for constructing in-frame deletion (26) was used to knock in DNA sequences encoding OxyR variants after the oxyR promoter (P oxyR ) in S. oneidensis. In brief, the target gene sequences were amplified by PCR with primers containing attB and the gene-specific sequence. The fragments were introduced into plasmid pHGM01 by using Gateway BP Clonase II enzyme mix (Invitrogen) according to the manufacturer's instructions, resulting in mutagenesis vectors, which were maintained in E. coli DAP auxotroph WM3064. The vectors were subsequently transferred into the S. oneidensis DoxyR strains via conjugation. Integration of the mutagenesis constructs into the chromosome was selected by resistance to gentamicin and confirmed by PCR. Verified transconjugants were grown in LB broth in the absence of NaCl and plated on LB containing 10% sucrose. Gentamycin-sensitive and sucrose-resistant colonies were screened by PCR for deletion of the target gene. Strains carrying OxyR-variant knock-in were verified by sequencing the mutated regions.
To assess protein levels of OxyR variants, constructs for generating GFP fusion proteins were prepared. In brief, DNA fragments encoding the OxyR variants under test and gfp genes were PCR amplified with specifically designed primers, allowing the first-round products to be joined together by a second round of PCR as described previously (37). The final PCR products were cloned into the vectors under the control of P oxyR for knock-in as described above. Expression and localization of GFP fusions were visualized as described previously (38). For quantification of fluorescence, mid-log-phase cultures of each test strain carrying GFP fusions were collected, washed with phosphate-buffered saline containing 0.05% Tween 20, and disrupted by French pressure cell treatment. Throughout this study, the protein concentration of the resulting cell lysates was determined using a Bradford assay with bovine serum albumin (BSA) as a standard (Bio-Rad) when necessary. A volume of 100 mL cell lysates was transferred into black 96-well plates at various time intervals, and fluorescence was measured using a fluorescence microplate reader (M200 Pro Tecan) with excitation at 485 nm and detection of emission at 515 nm.
Analysis of gene expression. Activity of the promoter (P katB ) for the major catalase katB gene was assessed using a single-copy integrative lacZ reporter system as described and used previously (39). Cells grown to the mid-exponential phase under normal or H 2 O 2 -challenging conditions (specified in the text and/or figure legends) were collected, and b-galactosidase activity was determined by monitoring color development at 420 nm using a Synergy 2 Pro200 multidetection microplate reader (Tecan), and the data were presented as Miller units.
Droplet assays. Droplet assays were employed to evaluate viability and growth inhibition on plates. Cells grown in LB to the mid-log phase were collected by centrifugation and adjusted to 10 9 CFU/mL, which was set as the undiluted (dilution factor, 0). Tenfold serial dilutions were prepared with fresh LB medium. Five microliters of each dilution were dropped onto LB plates containing agents such as catalase when necessary. The plates were incubated for 24 h or longer in dark before being read. All experiments were repeated at least three times.
Analysis of catalase. To assess catalase levels, S. oneidensis cells grown in LB the mid-exponential phase were incubated with 0.2 mM H 2 O 2 for 30 min and then collected by centrifugation and disrupted by French pressure cell treatment. Throughout this study, the protein concentration of the resulting cell lysates was determined using a Bradford assay with BSA as a standard (Bio-Rad). Aliquots of cell lysates containing the same amount of protein were subjected to 10% nondenaturing polyacrylamide gel electrophoresis (PAGE). Catalases were detected by using the corresponding activity-staining methods (40).
Activity of catalase was also assayed in a more quantitative approach as described previously (41). Briefly, mid-exponential-phase cells in liquid medium were collected, washed twice in 50 mM KH 2 PO 4 buffer (pH 7.0), resuspended in the same buffer, and then disrupted by sonication. Ten microliters of cell extracts containing 40 ng/mL protein was added to 90 mL KH 2 PO 4 and 100 mL 20 mM H 2 O 2 in a 200-mL volume. Decomposition of H 2 O 2 was measured at 240 nm with absorbance readings taken at 15-s time intervals for a total time of 3.5 min in a Tecan M200 Pro microplate reader. The unit of activity of each sample is expressed as mmol H 2 O 2 decomposed per min and per mg of protein (mmol Á min 21 Á mg 21 ). Each sample was tested in quadruplicate for each strain assayed.
Expression and purification of OxyR variants. All OxyR variants under test were purified as Histagged soluble proteins as described before (19). In brief, E. coli BL21(DE3) strains transformed with pET28a carrying target genes were grown in LB to the mid-log phase and then induced with 0.2 mM IPTG (isopropylb-D-thiogalactopyranoside) at 25°C for 6 h to produce high levels of His 6 -OxyR variants. Cell pellets were treated by French press, and His 6 -OxyR variants were purified from crude cell lysates using a nickel-ion affinity column (GE Healthcare). After removal of contaminant proteins with washing buffer containing 20 mM imidazole, the His-tagged OxyR variants were collected in elution buffer containing 100 mM imidazole. The eluted fractions were concentrated and dialyzed using a buffer containing 20 mM Tris-HCl, pH 8.0, 250 mM NaCl, 350 mL 2-mercaptoethanol (2-ME), and 1 mM NaN 3 and further purified by gel filtration using a Superdex 200 column (Pharmacia) run on an Äkta fast protein liquid chromatography (FPLC) system (Pharmacia). The different oligomeric states of SoOxyR were resolved with a short analytical gel filtration column (GFC 300), which allows quick injections and monitoring with great resolution. The peak fractions were collected and analyzed by SDS-PAGE. Fractions containing purified proteins were concentrated to 8 mg/mL and stored at 4°C. The identity of purified proteins was confirmed with tandem mass spectrometry (MS/MS) analysis.
Site-directed mutagenesis. Site-directed mutagenesis was employed to generate OxyR proteins carrying point mutations. The oxyR gene within the vectors used for knock-in or for expression and purification was subjected to the modification by using a QuikChange II XL site-directed mutagenesis kit (Stratagene) as described previously (42).
Crystallization and structure determination. Purified SoOxyR C203S (point mutation) RD was crystallized at 20°C by hanging-drop vapor diffusion. Drops were made by combining 1 mL of SoOxyR C203S RD with 1 mL of mother liquor containing 4.0 M ammonium acetate and 0.1 M Bis-Tris propane, pH 7.0. Rodshaped crystals appeared within 3 to 4 days. Crystals were harvested and flash frozen in cold nitrogen stream at 2173°C using mother liquor supplemented with 20% (vol/vol) glycerol as a cryoprotectant. The diffraction data were collected from single crystals at the Life Sciences Collaborative Access Team (LS-CAT) at the Advanced Photon Source (APS). Images were collected using a 1°oscillation angle at a wavelength of 0.98 Å. The data were processed using HKL-2000 (43) ( Table S2). The structure was determined by molecular replacement using the coordinates of the reduced OxyR of E. coli (PDB ID 1I69) as a search model. The structure model was built using Autobuild (PHENIX) (44) and Coot (45) and refined with phenix.refine. Merohedral twining was observed with a twining fraction of 43.5%, and therefore, the twin law (h, -h-k, -l) was applied during refinement in PHENIX. The final structure was refined to a 2.4-Å resolution (R work , 17.34%, and R free , 20.36%). There are six molecules in each asymmetric unit. All the structure figures were prepared using the program PyMOL unless otherwise specified (PyMOL Molecular Graphics System, version 2.0; Schrödinger, LLC). All structural alignment calculations were done using DALI (46). The final model of the OxyR regulatory domain S. oneidensis was deposited in the PDB with PDB ID 7L4S.
DNA-binding analyses. To test the interaction between OxyR and promoter regions of its target genes, electrophoretic mobility shift assays (EMSAs) were conducted as previously described (19). DNA probes covering the predicted OxyR-binding sites were obtained by PCR, during which the doublestranded product was labeled with digoxigenin-ddUTP (Roche Diagnostics). The digoxigenin-labeled DNA probes were mixed with serial dilutions of purified OxyR of various concentrations in binding buffer (4 mM Tris-HCl [pH 8.0], 40 mM NaCl, 4 mM MgCl 2 , and 4% glycerol) containing 0.75 mg of poly(dI-dC) at room temperature for 15 min. The DNA/protein mixtures were loaded on 7% native polyacrylamide gels for electrophoretic separation, and the resulting gel was visualized with the UVP image system.
Other analyses. Student's t test was performed for pairwise comparisons with statistical significance set at the 0.05 confidence level. Values were presented as means 6 standard deviation.
Data availability. The crystallographic coordinates and associated structure factors for S. oneidensis OxyR are available at the Protein Data Bank (PDB) under accession code 7L4S.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.