Defining the binding determinants of Shewanella oneidensis OxyR: Implications for the link between the contracted OxyR regulon and adaptation

It is well-established that OxyR functions as a transcriptional activator of the peroxide stress response in bacteria, primarily based on studies on Escherichia coli. Recent investigations have revealed that OxyRs of some other bacteria can regulate gene expression through both repression and activation or repression only; however, the underlying mechanisms remain largely unknown. Here, we demonstrated in γ-proteobacteriumShewanella oneidensis regulation of OxyR on expression of major catalase gene katB in a dual-control manner through interaction with a single site in the promoter region. Under non-stress conditions, katB expression was repressed by reduced OxyR (OxyRred), whereas when oxidized, OxyR (OxyRoxi) outcompeted OxyRred for the site because of substantially enhanced affinity, resulting in a graded response to oxidative stress, from repression to derepression to activation. The OxyR-binding motif is characterized as a combination of the E. coli motif (tetranucleotides spaced by heptanucleotide) and palindromic structure. We provided evidence to suggest that the S. oneidensis OxyR regulon is significantly contracted compared with those reported, probably containing only five members that are exclusively involved in oxygen reactive species scavenging and iron sequestering. These characteristics probably reflect the adapting strategy of the bacteria that S. oneidensis represents to thrive in redox-stratified microaerobic and anaerobic environments.

The ability of bacteria to alter gene expression patterns in response to environmental stresses is critical for their survival. This is particularly true about oxidative stress because its causing agents, reactive oxygen species (ROS), 2 are omnipresent; in addition to those coming from environments, they are generated endogenously as metabolic by-products of cellular oxygen respiration (1). ROS, including superoxide (O 2 . ), hydrogen peroxide (H 2 O 2 ), and hydroxyl radical ( ⅐ OH), damage biomolecules such as lipids, proteins, and DNA (1). When ROS levels exceed safe limits, sensing and responding systems are triggered to coordinately regulate expression of a set of genes to ensure that the ROS concentrations are restrained at an acceptable level and damages are promptly repaired (2). The primary members of these genes encode ROS detoxification enzymes (alkylhydroperoxide reductase (Ahp, renamed NADH peroxidase), catalases, and various peroxidases), iron-sequestering proteins (Dps in particular), and oxidative damage-repairing macromolecules (3). One of the major ROS sensing and responding systems is OxyR, a LysR-type transcriptional regulator (LTTR) consisting of an N-terminal DNA-binding domain (DBD) and C-terminal regulatory domain (RD) (2,4). In model bacterium Escherichia coli, xyR in a tetrameric arrangement is able to activate the transcription of its target genes through the DBD extremely rapidly, within 1-2 min after E. coli cells are exposed to H 2 O 2 (5). The activation of OxyR by H 2 O 2 oxidation is through the formation of an intracellular disulfide bond between the two conserved cysteine residues located in the RD. This induces structural changes of the RD, leading to conformational rearrangement of the DBD to alter the DNA-binding affinity (6,7). The OxyR-binding motif in E. coli is characterized by four OxyR-binding tetranucleotide sequences spaced by heptanucleotides (consensus, ATAGntnnnanCTAT-N 7 -ATAGntnnnanCTAT) (8). It has been proposed that each subunit of the OxyR homotetramer specifically binds to one of four tetranucleotide sequences, and because of the spacing, the four subunits interact with four adjacent major grooves on one side of the DNA helix (6,8).
In recent years, studies into OxyRs of other bacteria have revealed some variations to the E. coli OxyR model. In Deinococcus radiodurans, OxyR is activated by oxidization at the conserved single cysteine residue to a sulfenic acid under peroxide stress (9). In several bacterial species, such as Neisseria, OxyRs function in a dual-control manner (both a repressor and an activator) for major H 2 O 2 scavenging proteins and an activator only for other members of their regulons; as a consequence, OxyR-deficient strains are more resistant to H 2 O 2 (10 -13). More surprisingly, Corynebacteria OxyRs function primarily as a repressor for more than 20 genes, including those for ROS detoxification enzymes and iron-sequestering proteins (14 -16). In general, the majority of OxyRs are homologous in sequence (ϳ30 -35% in identity) and similar in structure and thus functionally exchangeable (10 -16).
Shewanella, a group of Gram-negative facultative ␥-proteobacteria, thrive in redox-stratified environments and are renowned for their respiratory versatility (17). Because of the potential application in bioremediation, biogeochemical circulation of minerals, and bioelectricity, these bacteria have been intensively studied, especially the model species Shewanella oneidensis (18). The S. oneidensis OxyR (SoOxyR) resembles those of Neisseria in that it represses expression of catalase gene katB under non-stress conditions but functions as an activator for all of its target genes when cells are challenged by H 2 O 2 (19). Nevertheless, three unexpected phenotypes resulting from the SooxyR disruption were observed (11). First, although the SooxyR mutant under non-stress conditions degrades H 2 O 2 more rapidly, it still carries defects in both viability (plating defect) due to increased sensitivity to H 2 O 2 and growth (18 -20), phenotypes reported in bacteria with OxyR being a positive regulator, such as E. coli (21,22). Second, SoOxyR is also able to respond to organic peroxides, functionally intertwined with OhrR, a regulator specific for organic peroxides (23). Third, SoOxyR is unable to complement the E. coli oxyR mutation and vice versa, although they share a comparable sequence similarity, as mentioned above (11,13,19). Hence, SoOxyR apparently mediates cellular response to oxidative stress with unprecedented mechanisms.
In our continued efforts to determine factors accountable for the characteristics of the SoOxyR mutant, we took on in this study to assess the biochemical properties of SoOxyR. We demonstrate that SoOxyR proteins, present in reduced (SoOxyR oxi ) and oxidized (SoOxyR red ) forms, recognize a single site featured by a combination of the OxyR-binding motif of E. coli and palindromic structure. When present, SoOxyR oxi dictates regulation because of higher affinity for the site than SoOxyR red . Consequently, SoOxyR oxi in excess exerts great inhibitory effects on growth but, interestingly, not on viability. Furthermore, evidence is provided to suggest that the SoOxyR regulon appears contracted significantly, having only five members.

Recombinant His-tagged OxyR proteins exist in reduced and oxidized forms
S. oneidensis OxyR, as a H 2 O 2 -responsive transcriptional regulator, empowers cells to confront oxidative stress by regulating gene expression through reversible formation of an inter-molecular disulfide bond (19). To characterize the biochemical properties of this regulator in more detail, we expressed recombinant SoOxyR with a hexahistidine (His 6 ) tag at the N terminus in E. coli. A large fraction of the protein was soluble, from which the protein was purified by Ni 2ϩ -affinity chromatography, as revealed as two bands of appropriate molecular mass on SDS-PAGE (Fig. 1A). The scenario suggests that the protein may exist in both reduced and oxidized forms, SoOxyR red and SoOxyR oxi , respectively. To confirm that both bands are SoOxyR proteins, they were excised from an SDS-polyacrylamide gel and digested with trypsin, and the resulting peptides were analyzed with LC-MS/MS. The result verified their identities in that the peptides matched SoOxyR with 78% coverage of the expected peptides (Fig. S1). To investigate whether oxidation underlies the two bands, the protein samples were treated with the reducing agent DTT of various concentrations and then applied to SDS-PAGE (Fig. 1B). Clearly, SoOxyR oxi can be converted to SoOxyR red by DTT, and complete conversion was achieved with 3.0 mM. We then performed chemical cross-linking assays to evaluate whether SoOxyR forms oligomers at high concentrations (1). Indeed, when treated with the cross-linker ethylene glycol bis(succinimidyl succinate) (EGS), oligomerization of SoOxyR, dimerization in particular, was evident (Fig. 1C). These results indicate that SoOxyR, probably functioning in an oligomeric structure, exists in both reduced and oxidized forms.

SoOxyR oxi always functions as an activator, whereas SoOxyR red is conditional as a repressor
E. coli OxyR senses H 2 O 2 via the reversible formation of disulfide bond between the two conserved cysteines (Cys-199 and Cys-208). In SoOxyR, the counterparts of these two cysteines are Cys-203 and Cys-212, respectively; mutation of either residue to serine locks SoOxyR into the reduced form (19,24). In addition, the SoOxyR L197P mutant is found to be locked into the oxidized state (25). To determine physiological activity of SoOxyR of different states, we assessed the ability of SoOxyR variants, SoOxyR WT (wildtype), SoOxyR C203S , and SoOxyR L197P , to complement the SooxyR null mutant (⌬SooxyR). We placed the SooxyR alleles under the control of isopropyl-␤-D-thiogalactopyranoside (IPTG)-inducible promoter P tac within pHGE-Ptac, and the resulting vectors were introduced into ⌬SooxyR. Spotting assays of series dilution cultures were performed to OxyR regulation in S. oneidensis examine the plating defect resulting from the SoOxyR loss on LB plates, as it is the most dramatic phenotype (18,19). Because the promoter is slightly leaky (26,27), both SoOxyR WT and SoOxyR L197P produced in the absence of IPTG were sufficient to fully correct the plating defect of ⌬SooxyR, but SoOxyR C203S could not. Additionally, induced SoOxyR C203S reduced the viability of ⌬SooxyR cells, as evidenced by decreased survival of undiluted cultures on LB plates ( Fig. 2A). We also noticed that SoOxyR L197P in excess impeded growth when IPTG was at 0.02 mM, with much more severe effects being observed from 0.2 mM. Overproduced SoOxyR WT appeared to inhibit growth too, but the effect was much less significant.
In parallel, impacts of the SoOxyR variants on H 2 O 2 sensitivity were assessed by a diffusion disc assay, which was carried out in the presence of 0.02 mM IPTG. The results revealed that the ⌬SooxyR strain had an inhibition zone smaller (11 mm) than the wildtype (16 mm) ( Fig. 2B and Fig. S2), a phenomenon reported before (19). Whereas production of SoOxyR C203S substantially augmented the zone of ⌬SooxyR, ϳ34 mm in diameter, SoOxyR L197P virtually eliminated the zone. Given that in S. oneidensis, KatB is the catalase predominantly accounting for decomposing H 2 O 2 and therefore resistance (19) (Fig. 2B), we subsequently evaluated the influence of SoOxyR variants on KatB production with catalase staining (Fig. 2C) Fig. 2B). These data are in excellent agreement with previous findings (18,19). In ⌬SooxyR/SoOxyR L197P , the amount of KatB produced was drastically increased, migrating as the supercomplex on the gel (Fig.  2C). In contrast, KatB in ⌬SooxyR/SoOxyR C203S was undetectable either in the treated or untreated cells, indicating that KatB expression is repressed by SoOxyR C203S consistently. For further confirmation, catalase activities of these samples were assayed. The cell extracts (the same as for catalase staining) were mixed with 10 mM H 2 O 2 . Decomposition of H 2 O 2 was measured at 15-s time intervals for a total time of 3.5 min to calculate catalase activity. As shown in Fig. 2C, the results correlated well with those of catalase staining.
We further examined impacts of these SoOxyR variants on expression of ahpC that is subject to SoOxyR activation only by qRT-PCR (Fig. 2D). Like katB, ahpC was responsive to H 2 O 2 in the wildtype, with substantially enhanced transcription upon H 2 O 2 treatment. In ⌬SooxyR, although the responsiveness was abolished, ahpC differed from katB in that its transcription was down-regulated, indicating that SoOxyR does not function as a repressor for ahpC. This was supported by the fact that SoOxyR C203S failed to elicit detectable impacts on ahpC transcription. As expected, we observed that transcription of ahpC in So⌬oxyR/SoOxyR L197P was substantially up-regulated. Altogether, these results indicate that SoOxyR proteins are present in the reduced and oxidized states, both of which mediate reg- A, droplet assays for viability and growth assessment. Cultures of the indicated strains prepared to contain ϳ10 9 colony-forming units/ml were regarded as undiluted (dilution factor, 0) and were subjected to 10-fold series dilution. Five microliters of each dilution was dropped on LB plates containing IPTG at the indicated concentrations. Results were recorded after a 24-h incubation. Expression of OxyR variants was driven by IPTG-inducible P tac . B, H 2 O 2 susceptibility by disc diffusion assay. One hundred microliters of cultures was spread onto an agar plate containing 0.02 mM IPTG. Paper discs (diameter, 8 mm) containing 10 l of 10 M H 2 O 2 were placed on top of the agar. The plates were incubated at 30°C for 16 h prior to analysis. The diameters of the zones of clearing (halo, in millimeters) generated by the peroxides were measured. C, catalase detected by staining and activity assay. Cells of the mid-log phase either directly used (non-treated) or incubated with 0.2 mM H 2 O 2 for 30 min (treated) were used for the assay. Cell lysates containing the same amount of protein were subjected to 10% nondenaturing PAGE (top) and an activity assay (bottom). Staining was performed as described under "Experimental procedures." For the catalase activity assay, 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. The unit of activity of each sample is expressed as catalase unit (mol of H 2 O 2 decomposed per min and per mg of protein). D, analysis of ahpC and katB transcripts by qRT-PCR. Cells of the mid-log phase incubated with 0.2 mM H 2 O 2 for 2 min were used for total RNA extraction. The cycle threshold (C T ) values were averaged and normalized against the C T value of the 16S rRNA gene. RTA, relative transcript abundance. In all panels, experiments were performed at least three times, with the average Ϯ S.E. (error bars) of representative results presented.

OxyR regulation in S. oneidensis
ulation; SoOxyR oxi appears to activate the regulon in general, but SoOxyR red only works for some as a repressor.

Analysis of the promoter region of the katB gene
Data presented thus far have demonstrated that katB expression is constitutively activated by SoOxyR L197P and repressed by SoOxyR C203S . We wondered whether SoOxyR interacts with different DNA sites for these two opposing effects on katB expression. To this end, we first mapped the transcription start site (TSS) for the katB gene with 5Ј-rapid amplification of cDNA ends (5Ј-RACE) (Fig. 3A). The result revealed one major elongated primer product initiating at the position corresponding to A44, the adenine 44 bp upstream of the translation initiation codon ATG (Fig. 3B).
We then used an integrative lacZ reporter system to evaluate the activity of a series of katB promoters. DNA fragments of various lengths upstream of the katB coding sequence (from Ϫ362, Ϫ232, Ϫ188, Ϫ152, Ϫ108, or Ϫ70 to Ϫ1) were amplified and cloned into the reporter vector pHGEI01 (Fig. 3C). The resulting vectors, verified by sequencing, were introduced into relevant S. oneidensis strains for integration and subsequent removal of the antibiotic marker (28). ␤-Galactosidase assay revealed that activities of the fragments no less than 188 bp were comparable under any given condition, indicating that they are the same in functionality; to simplify description, only data from 188 bp (P katB-188 ) were shown (Fig. 3C). P katB-188 was responsive to H 2 O 2 in the wildtype but not in ⌬SooxyR. Moreover, P katB-188 activities in the untreated and treated wildtype cells were substantially lower and higher than those in ⌬SooxyR, respectively. These data are consistent with the catalase-staining results (Fig. 2C), indicating that this 188-bp fragment contains all elements required for both repression and activation of SoOxyR. In contrast, P katB-152 and P katB-108 exhibited considerable activities but were unable to respond to H 2 O 2 , indicating that these two promoter fragments are functioning and lack the elements for SoOxyR regulation. The activity of P katB-70 was barely detectable, indicating that the promoter in the fragment is destroyed. These results suggest that the SoOxyR binding sites lie between Ϫ108 and Ϫ188 bp upstream of the katB gene.

Mapping the binding sites of OxyR oxi and OxyR red
An electrophoretic mobility shift assay (EMSA) was performed to detect the binding ability of purified His 6 -tagged SoOxyR WT , SoOxyR C203S , and SoOxyR L197P to a katB promoter fragment covering Ϫ300 to Ϫ1 in which the predicted OxyRbinding sites are centered. All SoOxyR variants bound well to this fragment in comparison with negative control P 16S (promoter sequence of the 16S rRNA gene) (Fig. 4A), indicating that the sites with which SoOxyR variants interact are included. It is noteworthy that SoOxyR variant OxyR L197P had the strongest affinity for the katB promoter, giving an apparent band shift even at 2.5 nM, in contrast to SoOxyR C203S , with which band shift was observed at 50 nM and above.
To pinpoint the precise binding sequences of SoOxyR WT , SoOxyR C203S , and SoOxyR L197P , a DNA-footprinting experiment with the DNA fragment was performed. The fragments were labeled with 6-carboxyfluorescein (6-FAM), incubated with increasing concentrations of SoOxyR variants, and subjected to DNase I digestion. Upon the addition of 4.5 g of SoOxyR WT , a region of protection corresponding to the area from Ϫ126 to Ϫ173 with respect to the ATG became clearly evident (Fig. 4B). Interestingly, SoOxyR L197P and SoOxyR C203S protected almost the same region as SoOxyR WT , but the region expanding forward 7-10 bp with respect to the direction of the ATG (Fig. 4B). It was evident that the protected region contains a perfect palindromic sequence, TTAAAGTAGATT(GTAA-G)AATCTACTTTAA. In contrast, whether the region also covers a sequence pattern mimicking the OxyR-binding motif of E. coli (ATAGntnnnanCTAT-N 7 -ATAGntnnnanCTAT) was unknown, as this motif cannot be easily identified by sequence analysis.

OxyR regulation in S. oneidensis OxyR oxi outcompetes OxyR red in DNA binding
Because EMSAs do not measure binding under equilibrium conditions, fluorescence anisotropy was chosen to examine DNA binding in a quantitative manner at equilibrium (29). Fluorescein-labeled, 60-bp oligonucleotides that contain the entire binding motif were synthesized. As shown in Fig. 5A, the fluorescence anisotropy of the target DNA increased with increasing amounts of SoOxyR L197P and SoOxyR C203S , indicating that both proteins bound to this DNA probe. The ability of these proteins to bind the labeled DNA was clearly different based on the K d values calculated from the anisotropy data. SoOxyR L197P bound to the probe with a K d of 0.261 Ϯ 0.03 M, whereas SoOxyR C203S bound with a K d of 1.041 Ϯ 0.12 M. The significant (p Ͻ 0.001; t test) difference in K d values indicates that SoOxyR oxi binds to its target DNA sequence with an affinity about 4 times higher than that of SoOxyR red .
Given these differences, we reasoned that high-affinity SoOxyR oxi would overshadow SoOxyR red in physiological impacts when they are coexisting in the cell. To test this, we calibrated promoter activity of SooxyR with the integrative lacZ reporter in comparison with integrative vector carrying P tac -lacZ (Fig. S3). When P tac is on the chromosome, its activity in the presence of 0.02 mM IPTG was comparable with that of P oxyR under normal conditions. Constructs of SoOxyR L197P and SoOxyR C203S whose expression was driven by P tac were then introduced into the wildtype and integrated onto the chromosome. In the presence of IPTG at varying concentrations, impacts of SoOxyR L197P and SoOxyR C203S on growth of the wildtype were assessed. As shown in Fig. 5B, SoOxyR C203S expressed from the IPTG-inducible promoter had undetectable effects on growth, indicating that it could not interfere with SoOxyR WT in regulation. When produced at low levels (IPTG OxyR regulation in S. oneidensis at 0.01 mM), effects of SoOxyR L197P were not evident. However, in the presence of IPTG at 0.02 (equivalent to the oxyR native promoter) and 0.2 mM, SoOxyR L197P inhibited growth, and inhibitory effects increased with levels of the inducer. The impact is on growth rather than viability because the growth defect resulting from SoOxyR L197P production remained in the presence of catalase. These data suggest that SoOxyR oxi outcompetes SoOxyR red in regulation in vivo.

S. oneidensis OxyR-binding motif is complex
Although LTTRs bind to 15-17-bp palindromic regions as dimeric proteins proposed initially, they are active in tetramer form and protect large regions of DNA (up to 60 bp) (4). In line with this, OxyR binding motifs determined to date are up to 50 bp and low in conservation (10,13,14,30,31). To determine the pattern of the binding motif of SoOxyR, we tested promoter regions of H 2 O 2 -responsive genes, including ahpC, ccpA, dps, and katG-1, for in vitro binding of purified SoOxyR L197P protein by EMSA. All of the target DNA probes produced distinct retarded bands when SoOxyR proteins were added (Fig. 6A), validating that SoOxyR directly binds to these promoters. From the promoter regions of katB, katG-1, ahpC, dps, and ccpA, we deduced the SoOxyR consensus sequence with AlignACE and MDScan. Three putative binding motifs were obtained, and they are all palindrome-like sequences and nested (Fig. 6B). Importantly, these motifs reside in the region defined by DNase I footprinting presented above (Fig. 4B). Motif SoOxyR-M13 (13 bp, GATTGTAAGAATC in the katB promoter region) is in the central region, which extends 8 and 13 bp on both ends to form SoOxyR-M29 and SoOxyR-M39, respectively. Although the bioinformatics tools failed to identify a motif resembling the EcOxyR-binding motif, putative tetranucleotide sequences were manually marked out from these sequences (Fig. 6B).
To test the pattern of the SoOxyR-binding motif, we constructed a series of promoter mutants, P katB -M1-M6, to drive expression of E. coli lacZ (Fig. 6C). M1 and M2 were designed to test the importance of the 13-bp central palindromic sequence (SoOxyR-M13), lacking (by deletion) and losing (by point mutations) the feature, respectively. ␤-Galactosidase assays revealed that M1 lost response to H 2 O 2 both negatively and positively, resembling that of P katB in ⌬SooxyR, whereas M2 retained such ability, albeit significantly compromised (Fig. 6D). M3 aimed to estimate the essentiality of the 7-bp linker between the first two tetranucleotide sequences, whereas M4 altered the structure of the 29-bp palindromic sequence (SoOxyR-M29). The change in the length of the linker (M3) abolished the responding ability of the promoter to H 2 O 2 , but M4 largely behaved as the wildtype (Fig. 6D). M5 and M6 were used to test the significance of the first and last tetranucleotide sequences, which have the highest and lowest conservation, respectively. The promoter activities showed that destruction of the first tetranucleotide sequence (M5) largely rendered SoOxyR unable to activate katB expression upon H 2 O 2 treatment, whereas M6 only slightly affected the characteristics of the promoter (Fig. 6D). These data, collectively, suggest that the SoOxyR-binding motif is probably composed of four tetranucleotide sequences separated by a 7-bp linker, the last of which appears less conserved with respective to regulation. Additionally, the palindromic sequence of the central 13-bp but not the extended 29-bp region is critical for regulation.
We then constructed the SoOxyR binding weight matrix based on the experimentally verified SoOxyR-binding sequences (Fig. 6B). Screening for the other SoOxyR binding sites in the S. oneidensis genome was carried out using RSAT (32). In total, 11 putative SoOxyR-binding sites were identified with weight values greater than 7, a cut-off for reliable prediction (Table 1). Except for genes like ahpC, ccpA, dps, katG-1, and katB that have been verified by EMSA, all other genes on this list except for def (polypeptide deformylase) were not reported to be associated with oxidative stress response (1). More importantly, weight values of putative motifs for these genes are relatively low (highest, 14.4); there is a sharp gap between them and those of verified sites, such as katB (lowest, 26.6). Although further

OxyR regulation in S. oneidensis
investigation is needed, these data suggest that S. oneidensis OxyR possesses an extremely contracted regulon compared with that of E. coli, which consists of more than 30 genes (30).

Discussion
In many Gram-negative bacteria, oxidative stress response is critically mediated by the H 2 O 2 -responsive transcriptional regulator OxyR, a member of the LTTR family (2). Upon induction by H 2 O 2 , OxyR undergoes a conformational change, allowing the oxidized regulator to function as an activator for genes encoding proteins involved in protection against ROS, a scenario best illustrated in E. coli, on which most of our understanding about this regulator is built (1,33). However, this turns out to be one side of the regulatory effects of OxyR proteins. In recent years, OxyRs that work as a repressor only or through a dual-control mechanism have been found in Corynebacteria and Neisseria as exemplary bacteria, respectively (9 -16).
Clearly, the S. oneidensis OxyR resembles those of Neisseria, working as both a repressor and an activator. Nevertheless, it carries novel features that have never been observed from OxyRs studied to date, including compromised resistance to exogenous H 2 O 2 , intertwined regulation with OhrR, and nonexchangeable functionality with E. coli OxyR (18 -20, 23).
With SoOxyR variants that are locked in reduced and oxidized forms, we verified that SoOxyR oxi exclusively functions as an activator, whereas SoOxyR red as a repressor is conditional, repressing katB expression under non-stress conditions only. In the wildtype cultivated without H 2 O 2 , katB is expressed at the basal level, which is so low that KatB catalase activity is barely detectable (Fig. 2C). Although deficient in activation, SoOxyR C203S (SoOxyR red ) was fully functional for H 2 O 2 -independent repression, a feature consistent with the proposal that SoOxyR acts as a repressor when not in its activated form (19). Depletion of SoOxyR prompts katB expression to a constitutive Experiments were performed in the same manner as Fig. 4A. B, predicted OxyR-binding motifs in S. oneidensis based on the verified promoter sequences. The motifs of the 13-and 29-bp palindromic sequences predicted by AlignACE and MEME are marked with black and blue dashed lines, respectively. Tetranucleotide sequences are underlined based on the E. coli OxyR consensus. C, mutational analysis of the katB regulatory region. Transcriptional fusion constructs are diagramed as in Fig. 4C. In P katB , the motif of 13 bp is in blue and extended to form the 29-bp motif, the additional nucleotides are in red, and tetranucleotide sequences are underlined. In P katB mutants, a dot represents the corresponding nucleotide deleted, and nucleotides underlined and in boldface type were mutated. Promoter activity was measured by ␤-galactosidase assay and is presented as Miller units. Experiments were performed at least three times, with the average Ϯ S.E. (error bars) presented.

Table 1
Top OxyR-binding sites predicted to be in S. oneidensis  OxyR regulation in S. oneidensis intermediate level, which can be substantially elevated further when oxidized SoOxyR is present. Clearly, this occurs through direct binding of SoOxyR to the promoter region of katB based on the mutational analysis of the katB promoter variants (Fig. 6).
Although SoOxyR-binding motifs determined to date are low in sequence conservation, a common feature is that they are relatively long (up to 60 bp) (4,10,13,14,30,31). For repressoronly Corynebacterium OxyRs, OxyR-binding regions identified by EMSA are ϳ50 bp long with multiple T-N 7 -A motifs of very low sequence similarity, a characteristic for sequences recognized by LTTRs (15). In the case of activator-only and dualcontrol OxyRs, OxyR-binding motifs are primarily characterized by four tetranucleotide sequences (ATAG) spaced by heptanucleotides (8,13). It is worth noting that OxyR-binding motifs of some bacteria, such as Pseudomonas aeruginosa, have been proposed to be 15 bp (31). However, these motifs, at least in the case of P. aeruginosa, reside in the middle of 37-bp DNA fragments resembling the E. coli OxyR-binding consensus sequence (8,31).
Data presented here indicate that S. oneidensis OxyR-binding motifs corroborate the model of four tetranucleotide sequences (ATAG) spaced by heptanucleotides (Fig. 6). Disruption of tetranucleotide sequences except for the last one and the distance in between abolishes both repression and activation. The last tetranucleotide sequence in the katB promoter region is less critical, which may be due to low conservation among target promoters. Intriguingly, there are novel features in S. oneidensis OxyR-binding motifs. Two nested palindromic sequences, 13 and 29 bp, are located in the middle of the sites protected by SoOxyR from DNase I digestion. Whereas the palindrome feature of the 13-bp sequence is important for regulation, the 29-bp one only exhibits a slight impact. Seemingly, these palindromic features are dispensable for repression of SoOxyR red . Nevertheless, the 13-bp palindromic sequence may underlie, at least in part, the functional discrepancy between OxyRs of E. coli and S. oneidensis. However, this merits further investigation.
All OxyR proteins 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 (8,13,15). Given that H 2 O 2 can be generated as a metabolic by-product of cellular oxygen respiration and abiotically (18), cells are living in an environment where a balance between the production and removal of ROS is maintained. Therefore, it is conceivable that OxyR proteins in reduced and oxidized forms co-exist in the cell (Fig. 1). An obvious strategy for OxyR regulation is that OxyR red and OxyR oxi interact with the same site but differ from each other in binding affinity. In E. coli, OxyR oxi has greater affinity than OxyR red to the consensus sequence (8,33). In this study, we showed that the strategy is also employed by dual-control OxyR. An in vitro binding assay revealed that Oxy-R oxi exceeds OxyR red in binding affinity more than 4 times (Fig.  5A). In parallel, in vivo analysis demonstrated that OxyR oxi overwhelms OxyR red in its regulatory effects when both proteins are produced at similar levels (Fig. 5B). It is worth mentioning that OxyR in excess negatively influences growth but not viability. We have shown previously that the growth defect of the oxyR mutant is largely attributable to reduced efficacy of oxygen respiration, an indirect effect of the oxyR mutation (20). We speculate that the same mechanism accounts for the growth inhibition of excessive OxyR; efforts to test this notion are under way.
The size of the OxyR regulon varies greatly, ranging from dozens (most bacteria in which the subject has been studied) to over 100 (P. aeruginosa) (31). Among the OxyR-binding sites predicted by RSAT (Regulatory Sequence Analysis Tools) (32), five verified members of the regulon are on the top, far above the remaining several based on RSAT weight values. This implies that the regulon of S. oneidensis OxyR is probably small. On the contrary, a microarray analysis of S. oneidnesis cells in response to H 2 O 2 -induced oxidative stress has revealed that a vast number of genes are differentially transcribed (19). This indicates that oxidative stress imposes a global impact on gene expression mostly in an indirect manner. Most significantly, those associated with iron/heme biology except for Dps and thioredoxin/glutaredoxin systems involved in OxyR reduction, two groups of well-known OxyR-dependent genes in E. coli and many other bacteria, are not affected by the OxyR loss in S. oneidnesis (19, 30 -31).
The physiological relevance of why the OxyR regulon contracts so much is intriguing. It is tempting to suggest that this may be associated with redox-stratified niches where Shewanella thrive. As Shewanella respire an array of electron acceptors, mostly under microaerobic or anaerobic conditions, endogenous and exogenous ROS may not often amount to lifethreatening levels. As such, these bacteria evolve a concise and sufficient protection system against oxidative stress in their living environments. In line with this, although small, the regulon contains CcpA, cytochrome c 551 peroxidase, which substantiates H 2 O 2 removal under anaerobic conditions but is dispensable under aerobic conditions (34).

Bacterial strains, plasmids, and culture conditions
All bacterial strains and plasmids used in this study can be found in Table 2. 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: 2,6-diaminopimelic acid, 0.3 mM; ampicillin, 50 g/ml; kanamycin, 50 g/ml; gentamycin, 15 g/ml; streptomycin, 100 g/ml; and catalase on plates, 2000 units/ml. Growth in liquid media was monitored by recording values of optical density at 600 nM (A 600 ), as all strains used in this study were morphologically similar. Both LB and defined medium MS (18) were used for phenotypic assays in this study, and comparable results were obtained with respect to growth.

DNA manipulations
For inducible gene expression, the gene of interest was generated by PCR and introduced into the inducible plasmid pHGE-Ptac under the control of promoter P tac (35). For OxyR, OxyR regulation in S. oneidensis for example, the resulting vector, pHGE-Ptac-oxyR was verified by sequencing and transferred into the relevant strains via E. coli WM3064-mediated conjugation for chromosome integration as described before (28). To calibrate activities of P tac integrated on the chromosome, P tac was placed before the lacZ gene within integrative vector pHGEI01. The resulting vector, after verification by sequencing, was introduced into the S. oneidensis wildtype for chromosome integration. The same strategy was used to construct integrative lacZ reporter vectors carrying P katB variants.

Analysis of gene expression
The activity of target promoters was assessed using a singlecopy integrative lacZ reporter system as described previously (28). Briefly, fragments of varying length (indicated in the relevant figures) containing the sequence upstream of the target operon were amplified, cloned into the reporter vector pHGEI01, and verified by sequencing. The resultant vector in E. coli WM3064 was then transferred by conjugation into relevant S. oneidensis strains, in which it integrated into the chromosome, and the antibiotic marker was removed subsequently (36). Cells grown to the mid-log phase under conditions specified in the text and/or figure legends were collected, and ␤-galactosidase activity was determined by monitoring color development at 420 nm using a Synergy 2 Pro200 Multi-Detection Microplate Reader (Tecan) presented as Miller units.
For qRT-PCR, S. oneidensis cells were grown in LB with the required additives to the mid-log phase and collected by centrifugation, and RNA extraction was performed using the RNeasy minikit (Qiagen) as described before (37). RNA was quantified by using a NanoVue spectrophotometer (GE Healthcare). The analysis was carried out with an ABI7300 96-well qRT-PCR system (Applied Biosystems) as described previously (37). The expression of each gene was determined from three replicates in a single real-time qRT-PCR experiment. The cycle threshold (C T ) values for each gene of interest were averaged and normalized against the C T value of the 16S rRNA gene, whose abundance is relatively constant during the log phase. The relative abundance of each gene is presented.

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 colony-forming units/ml, which was set as undiluted (dilution factor, 0). 10-Fold series dilutions were prepared with fresh LB medium. Five microliters of each dilution was dropped onto LB plates containing agents such as catalase and/or IPTG when necessary. The plates were incubated for 24 h or longer in the dark before being read. All experiments were repeated at least three times.

Disc diffusion assays
Disc diffusion assays were done similarly to those done previously (19). Briefly, overnight cultures were diluted into LB and grown to the mid-log phase (ϳ0.6 of A 600 ). One hundred microliters of cultures was spread onto an agar plate containing required chemicals. Paper discs (diameter, 8 mm) containing 10 l of 10 M H 2 O 2 were placed on top of the agar. The plates were incubated at 30°C for 16 h before analysis. The diameters of the zones of clearing (zones of inhibition, in millimeters) generated by the peroxides were measured. All assays were done in triplicate using independent cultures, and the resulting zones of inhibition were averaged.

Analysis of catalase
To assess catalase levels, S. oneidensis cells grown in LB the mid-log 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 PAGE. Catalases were detected by using the corresponding activity-staining methods (38).
Activity of catalase was also assayed in a more quantitative approach, as described previously (19). Briefly, mid-log-phase cells in liquid medium were collected, washed twice in 50 mM KH 2 PO 4 buffer (pH 7.0), and resuspended in the same buffer and then disrupted by sonication. Ten microliters of cell extracts containing 40 ng/l protein was added to 90 l of KH 2 PO 4 and 100 l of 20 mM H 2 O 2 in a 200-l 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 mol of H 2 O 2 decomposed per min and per mg of protein (mol⅐min Ϫ1 ⅐mg Ϫ1 ). Each sample was tested in quadruplicate for each strain assayed.

Site-directed mutagenesis
The coding sequence of S. oneidensis oxyR was amplified by PCR and cloned in pET28 for expressing recombinant proteins with an N-terminal His 6 tag. The resulting pET28-oxyR, after sequence verification, was used as the template for site-directed mutagenesis to express OxyR C203S and OxyR L197P with a QuikChange II XL site-directed mutagenesis kit (Stratagene), as described previously (39). The resulting pET28-oxyR T607A and pET28-oxyR T613C was verified by sequencing.

Expression and purification of OxyR variants
Full-length S. oneidensis OxyR WT , OxyR C203S , and OxyR L197P were purified as His-tagged soluble proteins. E. coli BL21(DE3) strains carrying pET28-oxyR, pET28-oxyR T607A , or pET28-oxyR T613C grown in LB to the mid-log phase were induced with 0.2 mM IPTG at 25°C for 6 h to produce high levels of His 6 -OxyR variants. His 6 -OxyR variants were purified from crude cell lysates by French pressure cell treatment over 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 eluted in elution buffer containing 100 mM imidazole. The proteins were concentrated by ultrafiltration (10-kDa cutoff), exchanged into 20 mM Tris-HCl (pH 8.0) containing 150 mM NaCl, and further purified by gel filtration using a Superdex 200 column (Pharmacia) run on an Äkta fast protein liquid chromatography (FPLC) system (Pharmacia). The peak fractions were analyzed by 12% SDS-PAGE, followed by staining with Coomassie Brilliant Blue R250. Identification of purified proteins was confirmed with MS/MS analysis as described before (40).

Chemical cross-linking of OxyR
Chemical cross-linking with ethylene glycol bis-succinimidylsuccinate (Pierce) was used to examine the solution struc-ture of OxyR. Cross-linking was initiated by the addition of EGS in dimethyl sulfoxide to a final concentration of 10 mM, with 2.5 g of purified His 6 -OxyR in a total volume of 20 l. The reaction mixtures were incubated at room temperature for 30 min, quenched with 50 mM glycine, and analyzed by 12% SDS-PAGE.

DNA binding analyses
To test interaction between OxyR and promoter regions of its target genes, EMSAs were conducted as described previously (41). DNA-binding assays and conditions were similar to those reported previously, although a fluorescent label was used in place of radioactivity to detect the promoter fragment. DNA probes covering the predicted OxyR binding sites were obtained by PCR, during which the double-stranded product was labeled with digoxigenin-ddUTP (Roche Diagnostics). The digoxigenin-labeled DNA probes were mixed with serial dilutions of purified OxyR of varying concentrations in binding buffer (4 mM Tris-HCl (pH 8.0), 40 mM NaCl, 4 mM MgCl 2 , 4% glycerol) containing 0.75 g of poly(dI⅐dC) 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.
Fluorescence anisotropy was chosen as a method of DNA-binding analysis in equilibrium (42). DNA oligonucleotides 5Ј-end-labeled with 6-FAM were ordered from Sangon (Shanghai) and resuspended in annealing buffer (10 mM Tris, pH 7.5, 50 mM NaCl, 0.05 mM EDTA). OxyR of increasing protein titrations was incubated with a solution containing 10 nM 6-FAM-labeled DNA, 100 nM competitive nonspecific DNA, 100 mg/liter BSA, 100 mM BisTris (pH 6.5), 150 mM NaCl, and 5% glycerol and added to a constant volume of 25 l in a 384well black assay plate (Costar). Fluorescence anisotropy of the fluorescein-labeled DNA was observed via excitation at 485 nm and emission at 520 nm, using a plate reader (M200Pro, Tecan). Measurements were made in triplicate, and reported values are the averages for three separate triplicate runs. Data were plotted as average fluorescence anisotropy values as a function of protein concentration by using GraphPad Prism version 7. The K d (dissociation constant) was calculated based on the equation generated by the best-fit curve using a single-ligand binding equation, r ϭ ([P] ϫ ⌬r)/(K d ϩ [P]), where r is the anisotropy value, ⌬r is the change in anisotropy, [P] is the concentration of added protein, and K d is the dissociation constant of the protein with the DNA probe. Each OxyR variant was analyzed at each experiment by three individual replicates.

5-RACE analysis of transcripts
The transcriptional start site of the katB gene of S. oneidensis was determined using a 5Ј-RACE kit (Invitrogen) as described recently (43). RNAs from S. oneidensis cells of the mid-log phase were extracted and quantified as described above. Reverse transcription was conducted on preprocessed RNAs without 5Ј-phosphates followed by two rounds of nested PCRs. The PCR products were subjected to agarose gel separation, with purification of the 5Ј-RACE products, and inserted into the pMD19-T vector (TaKaRa) for direct DNA sequencing. The first DNA base adjacent to the 5Ј-RACE adaptor was regarded as the transcription start site.

OxyR regulation in S. oneidensis DNase I footprinting
DNase I footprinting analysis was carried out as described elsewhere (44). DNA fragments of 152 bp with the predicted OxyR-binding site were generated by PCR using 6-FAM primers labeled at the 5Ј-end. Binding reactions for DNase I footprinting were conducted as for the EMSAs. The digested DNA was subjected to DNA sequencing and analyzed with Peak Scanner software (Applied Biosystems).

Other analyses
AlignACE (45) and MDScan (46) were used to identify OxyR-binding motifs in the putative promoter regions of genes. Sequence logos were generated using WebLogo (47). Student's t test was performed for pairwise comparisons with statistical significance set at the 0.05 confidence level. Values are presented as means Ϯ S.E.