Staphylococcus aureus PerR Is a Hypersensitive Hydrogen Peroxide Sensor using Iron-mediated Histidine Oxidation*

Background: PerR is a metal-dependent H2O2 sensor in many Gram-positive bacteria. Results: Staphylococcus aureus PerRSA, previously known as a Mn2+-specific repressor, uses Fe2+ to sense very low levels of H2O2. Conclusion: The apparent lack of Fe2+-dependent repressor activity of PerRSA is due to the hypersensitivity of PerRSA under aerobic conditions. Significance: Cells expressing hypersensitive PerRSA are less virulent than those expressing PerRBS. In many Gram-positive bacteria PerR is a major peroxide sensor whose repressor activity is dependent on a bound metal cofactor. The prototype for PerR sensors, the Bacillus subtilis PerRBS protein, represses target genes when bound to either Mn2+ or Fe2+ as corepressor, but only the Fe2+-bound form responds to H2O2. The orthologous protein in the human pathogen Staphylococcus aureus, PerRSA, plays important roles in H2O2 resistance and virulence. However, PerRSA is reported to only respond to Mn2+ as corepressor, which suggests that it might rely on a distinct, iron-independent mechanism for H2O2 sensing. Here we demonstrate that PerRSA uses either Fe2+ or Mn2+ as corepressor, and that, like PerRBS, the Fe2+-bound form of PerRSA senses physiological levels of H2O2 by iron-mediated histidine oxidation. Moreover, we show that PerRSA is poised to sense very low levels of endogenous H2O2, which normally cannot be sensed by B. subtilis PerRBS. This hypersensitivity of PerRSA accounts for the apparent lack of Fe2+-dependent repressor activity and consequent Mn2+-specific repressor activity under aerobic conditions. We also provide evidence that the activity of PerRSA is directly correlated with virulence, whereas it is inversely correlated with H2O2 resistance, suggesting that PerRSA may be an attractive target for the control of S. aureus pathogenesis.

Reactive oxygen species, which are produced endogenously as a by-product of aerobic metabolism or exogenously by microbial competitors and eukaryotic hosts, can cause oxidative stress to bacteria by damaging cellular constituents (1,2).
To cope with reactive oxygen species, bacteria have evolved sophisticated oxidative stress response systems including transcription factors that efficiently sense specific reactive oxygen species and induce appropriate defense systems (2)(3)(4). For example, OxyR in the Gram-negative model bacterium Escherichia coli senses H 2 O 2 using cysteine oxidation, and activates transcription of ϳ20 genes, including genes involved in H 2 O 2 detoxification (1). Whereas many Gram-negative bacteria use OxyR as the major H 2 O 2 sensor, many Gram-positive bacteria use PerR as a functional equivalent of OxyR (5,6).
Bacillus subtilis PerR (PerR BS ) 4 is a member of Fur family of metal-dependent regulators and is the prototype for a group of metal-dependent peroxide sensing repressors (5). PerR BS contains a structural Zn 2ϩ coordinated by four cysteine residues (Cys 4 :Zn site, Site 1) and a second regulatory metal binding site (Site 2) composed of three N-donor ligands (His-37, His-91, and His-93) and two O-donor ligands (Asp-85 and Asp-104). Although the binding of either Fe 2ϩ (PerR BS :Zn,Fe) or Mn 2ϩ (PerR BS :Zn,Mn) at Site 2 activates PerR BS to bind DNA, only PerR BS :Zn,Fe can sense low levels of H 2 O 2 . Unlike cysteine thiol-based peroxide sensors such as OxyR and OhrR, PerR BS senses H 2 O 2 by metal-catalyzed histidine oxidation. Reaction of Fe 2ϩ , bound to Site 2, with H 2 O 2 leads to the rapid oxidation of either His-37 or, to a lesser degree, His-91 (two of the Site 2 ligands) with concomitant loss of iron binding (7). Structurally, this results in an opening of the DNA-binding competent caliper-like conformation, leading to a loss of DNA binding and thus allowing the induction of genes that are normally repressed by active PerR BS :Zn,Fe (8).
Staphylococcus aureus is a major human pathogen commonly causing nosocomial and community-acquired infectious diseases worldwide. S. aureus, which can be found as part of the normal skin flora and in anterior nares of the nasal passages, can cause a spectrum of illnesses from minor skin and soft tissue infections to more invasive and serious diseases such as pneumonia, meningitis, osteomyelitis, endocarditis, toxic shock syndrome, bacteremia, and sepsis (9). As a facultative anaerobic Gram-positive bacterium, S. aureus also uses PerR for the control of oxidative stress response (10,11). The S. aureus PerR (PerR SA ) regulon is similar to that described for PerR BS and includes genes encoding KatA, AhpCF, MrgA, Fur, and PerR, as well as others encoding thioredoxin reductase (TrxB), bacterioferritin comigratory protein (Bcp), and an iron storage protein ferritin (Ftn). Despite the similarity of the H 2 O 2 -dependent regulation of the PerR SA regulon, Fe 2ϩ was reported to be completely ineffective as a corepressor for the PerR SA -regulated genes, which were only repressed by Mn 2ϩ . Indeed, expression of the PerR SA -regulated genes is induced rather than repressed by added Fe 2ϩ (11)(12)(13)(14). These observations led to the conclusion that PerR SA is a Mn 2ϩ -specific repressor and further suggest that PerR SA may use a fundamentally different and iron-independent mechanism to sense H 2 O 2 . Despite the importance of PerR SA in the regulation of virulence factors of S. aureus, the mechanism by which PerR SA senses H 2 O 2 has not been elucidated.
Here we have analyzed the metal-dependent H 2 O 2 sensing mechanisms of PerR SA in vitro and in vivo in comparison with those of PerR BS . PerR SA , like many other Fur family proteins, contains a structural Zn 2ϩ site coordinated by four cysteine residues, which is resistant to oxidation by physiologically relevant H 2 O 2 concentration. Contrary to the suggestion that PerR SA is a Mn 2ϩ -specific repressor, the regulatory metal binding site (composed of His-43, Asp-91, His-97, His-99, and Asp-110) can bind Fe 2ϩ even with higher affinity than Mn 2ϩ when measured under anaerobic conditions. Moreover, the Fe 2ϩbound PerR SA , but not the Mn 2ϩ -bound form, can sense H 2 O 2 by Fe 2ϩ -dependent oxidation of His-43 and His-97. In cells grown under aerobic conditions most of PerR SA is detected in the fully oxidized state, whereas cells grown under oxygen-limited conditions exhibit Fe 2ϩ -dependent repression of the PerR SA regulon. The exquisite sensitivity of PerR SA to inactivation likely explains the previous observation of an apparent lack of Fe 2ϩ -dependent repressor activity. Finally, we provide evidence that the high H 2 O 2 sensitivity of PerR SA (in comparison to PerR BS ) is important for H 2 O 2 resistance under aerobic conditions and that the low sensitivity of PerR BS (in comparison to PerR SA ) increases virulence of S. aureus in host.

Experimental Procedures
Bacterial Strains, Media, and Growth Conditions-The bacterial strains used in this study are listed in Table 1. E. coli, B. subtilis, and S. aureus were grown in Luria-Bertani (LB) media at 37°C with appropriate antibiotics unless otherwise indicated. As metal-limited minimal media (MLMM), MOPSbuffered minimal medium was used for B. subtilis (15), and phosphate-buffered minimal medium was used for S. aureus (10). Oxygen-limited cultures were grown in 15-ml rubber screw-top tubes with the addition of 0.2% potassium nitrate. To facilitate oxygen-limited growth of S. aureus in MLMM, 1% Chelex-treated tryptone was added.

Construction of a perR Deletion Mutant Strain of S. aureus
Newman-The perR SA ::cat cassette was constructed by joining PCR using two 1-kb DNA fragments (upstream and downstream of perR SA ORF) and a fragment with a chloramphenicol resistance marker (cat) from pDG1661. This cassette was cloned into the BamHI and EcoRI sites of pMAD (16) having a  (17). For the expression of perR SA -FLAG or perR BS -FLAG fusions in S. aureus, the DNA fragment of pJL070 containing perR BS -FLAG and that of pJL361 containing perR SA -FLAG, were each cloned into BamHI and EcoRI sites of pLL29 (18) producing pJL1434 and pJL1430, respectively. Mutant alleles of perR SA -FLAG were generated by the QuikChange method (Stratagene) using pJL1430 as templates. Each of these plasmids was integrated into the chromosome of S. aureus RN4220 with the help of int gene encoded in pLL2787, and then transferred into the perR null mutant S. aureus Newman strain (LS0085) by phage transduction using ⌽11 (19). For the construction of reporter fusion plasmids, the lacZ gene from pDG1661 was PCR-amplified with the introduction of an NcoI site just after the KpnI site, and cloned into the KpnI and EcoRI sites of pCN33 resulting in pJL901. Then, DNA fragment containing the mrgA or katA promoter regions was introduced into the BamHI and NcoI sites of pJL901. The resulting reporter fusion plasmids were introduced into S. aureus Newman by electroporation after passage through S. aureus RN4220. To increase the copy number of perR SA -FLAG, perR SA -FLAG was cloned into the BamHI and EcoRI sites of the high copy number plasmid pCN48 (20) resulting in pJL643. This plasmid was introduced into the perR null mutant S. aureus Newman (LS0085) by electroporation after passage through S. aureus RN4220, generating a strain named LS0166. PerR BS -FLAG and PerR SA -FLAG proteins are fully functional as judged by reporter fusion assays, and were used for complementation of the perR null mutant strains and pulldown assays (15,21).
Overexpression and Purification of Proteins-The PCR-amplified DNA fragments containing the perR SA ORF were digested with BspHI and BamHI, and cloned into the NcoI and BamHI sites of pET-16b (Novagen) producing a plasmid named pJL203. Mutant alleles of perR SA were generated by the QuikChange method (Stratagene) using pJL203 as template. E. coli oxyR was cloned into the NdeI and BamHI sites of pET-11a (Novagen) producing a plasmid named pJL1282. The encoded proteins were overexpressed using E. coli BL21(DE3) pLysS cells. Wild type (WT) PerR SA proteins were purified after overexpression in E. coli BL21(DE3) pLysS cells harboring pJL203 as previously described for PerR BS proteins (15). Briefly, the cell lysates were clarified by centrifugation and then PerR SA was purified by heparin-Sepharose and Mono-Q chromatography using buffer A (20 mM Tris-HCl, pH 8.0, 0.1 M NaCl, and 5% glycerol (v/v)) containing 10 mM EDTA with the application of a linear gradient of 0.1-1 M NaCl. Further purification was performed using a Superdex-200 (HiLoad 16/60) column equilibrated with Chelex-treated buffer A. The concentration of PerR SA was determined using a molar extinction coefficient of 10,430 M Ϫ1 cm Ϫ1 at 280 nm.
Enzyme Assays-On-gel catalase activity was assayed using 1:1 mixture of 5% ferric chloride and 5% potassium ferricyanide after gel-soaking in 2 mM H 2 O 2 . ␤-Galactosidase assays were performed with or without 100 M H 2 O 2 treatment for 30 min as described previously (15), except that lysostaphin (10 g/ml) was used for the lysis of S. aureus cells. Measurement of Zn 2ϩ release by H 2 O 2 was performed as described previously (15) using 2.5 M dimeric PerR SA and 100 M 4-(2-pyridylazo)resorcinol (PAR). The Zn 2ϩ content of purified PerR SA by PAR assay was measured using a molar extinction coefficient of 85,000 M Ϫ1 cm Ϫ1 at 494 nm for Zn 2ϩ -PAR complex (15).
Western Blot Analysis-At A 600 ϭ 0.6, 10-ml cultures were harvested by centrifugation after the addition of 1.1 ml of trichloroacetic acid. Then, cells were resuspended in 500 l of 10% trichloroacetic acid and sonicated. After recovering sonicated samples by centrifugation, the pellets were resuspended with 60 l of alkylating buffer (100 mM iodoacetamide, 0.5 M Tris, pH 8.0, 5% glycerol, 100 mM NaCl, 2% SDS) and incubated for 1 h in the dark to alkylate-free thiols. Alkylated samples of 10 l (75 g of protein) were separated on 13.3% non-reducing SDS-PAGE gel using a Tris-Tricine buffer system and blotted to a polyvinylidene difluoride membrane. FLAG-tagged proteins were probed with mouse monoclonal anti-FLAG antibody and anti-mouse antibody conjugated with alkaline phosphatase (Sigma).
Fluorescence Anisotropy (FA) Experiments-FA experiments were performed using an LS55 luminescence spectrometer (PerkinElmer Life Sciences) installed in an anaerobic chamber (Coy). A 6-carboxyfluorescein (6-FAM)-labeled katA-PerR box DNA fragment was generated by annealing 5Ј-6-FAM-TTAAATTATAATTATTATAAATTGT-3Ј (Integrated DNA Technology) and its unlabeled complement. FA measurements ( ex ϭ 492 nm, slit width ϭ 15 nm; em ϭ 520 nm, slit width ϭ 20 nm, integration time ϭ 1 s) were performed in 3 ml of Chelex-treated anaerobic buffer A. The percentage activity and K d for DNA of purified PerR SA were determined to be ϳ20% and ϳ1 nM, respectively, by titration of PerR SA into 3 ml of buffer A containing 10 nM DNA and 1 mM manganese as reported previously (7). For the metal binding and H 2 O 2 sensitivity assays, buffer A containing 100 nM DNA and 100 nM active dimeric PerR SA were used, and FA was measured after each addition.
MALDI-TOF and LC-ESI MS/MS Mass Analyses-To analyze in vivo oxidation of PerR SA (Fig. 4), LS0166 cells expressing PerR SA -FLAG were grown in MLMM containing 50 M FeSO 4 or 50 M MnCl 2 . At an A 600 of ϳ1, cells were treated with or without 100 M H 2 O 2 for 2 min and lysed with 50 g of lysostaphin for 1 h in 0.5 ml of buffer A. PerR SA -FLAG protein recovered using anti-FLAG M2-agarose beads (Sigma) was incubated with 100 mM iodoacetamide in the presence of 50 mM EDTA and 1% SDS for 1 h in the dark, and separated on SDS-PAGE gel. The protein bands corresponding to PerR SA -FLAG were analyzed by MALDI-TOF MS using a 4700 Proteomics Analyzer instrument (Applied Biosystems) after in-gel tryptic digestion as described previously (15,22). The MALDI-TOF MS analysis of in vitro oxidation of purified PerR SA (Fig. 3D) was performed as described previously (7, 15) using a Voyager-DE STR instrument (Applied Biosystems). The analysis of protein oxidation after overexpression in E. coli (Fig. 5) was performed as previously described (22), except that sample preparations for Fig. 5B were carried out in an anaerobic chamber. The sites of oxidation were identified by LC-MS/MS analyses using an Agilent nanoflow-1200 series HPLC system connected to a linear ion trap mass spectrometer (Thermo Scientific).
Caenorhabditis elegans Killing Assay-NGM agar plates (5.5 cm diameter) spread with 30 l of A 600 ϭ 1 culture of E. coli OP50 (laboratory nematode food), S. aureus expressing no PerR (LS0093), S. aureus expressing PerR SA (LS0088), or S. aureus expressing PerR BS (LS0134) were used. For each assay 90 L4 stage C. elegans were used in triplicate of 30 worms/plate. The plates were incubated at 25°C, and scored for live and dead worms at least every 24 h as described previously (23). For each assay, the survival of worms was calculated by the Kaplan-Meier method, and survival differences were tested by using OASIS (24).

Results
Structural Zn 2ϩ and Regulatory Metal Binding Sites of PerR SA -PerR SA is highly similar (67% sequence identity) to PerR BS , but previous results have highlighted some striking differences in their response to added metal ions (11)(12)(13)(14). To provide a structural context for our investigation of PerR SA reactivity we generated a homology model of PerR SA based on the known structure of PerR BS (25). As shown in Fig. 1, A and B, PerR SA retains four highly conserved cysteine residues (Cys-102, Cys-105, Cys-142, and Cys-145) corresponding to those involved in high affinity structural Zn 2ϩ binding (Site 1) in most Fur family proteins as well as in PerR BS (5,15,26). PerR SA also has five other residues (His-43, Asp-91, His-97, His-99, and Asp-110), which correspond to the N/O donor ligands for the regulatory metal binding (Site 2) in PerR BS (7,8). To investigate the role of these predicted metal-binding residues in protein function, we generated PerR SA mutants and examined the in vivo repressor activities of WT and mutants using a PerR-regulated katA promoter-lacZ fusion (P katA -lacZ) (Fig. 1C). As expected, the P katA -lacZ reporter fusion was repressed in cells expressing WT PerR SA -FLAG but derepressed in the perR null mutant cells. Note that the repression levels of P katA -lacZ reporter fusion by WT PerR SA -FLAG were similar to those observed with the WT S. aureus strain, indicating that the perR null mutant strain complemented with WT PerR SA -FLAG behaves like WT strain. All nine mutants exhibited no repressor activity for the P katA -lacZ reporter fusion, and furthermore, these mutant proteins were present at levels greater than WT protein (Fig. 1D) indicative of a loss of repression of the autoregulated perR promoter. These results indicate that these amino acid residues proposed to be metal ligands are essential for in vivo repressor function.
Previously we have demonstrated that the structural Zn 2ϩ binding status of PerR BS can be monitored by mobility difference on non-reducing SDS-PAGE: monomeric PerR BS containing bound Zn 2ϩ migrates faster than PerR BS lacking bound Zn 2ϩ (15). To investigate the Zn 2ϩ binding status of PerR SA , WT and mutant proteins were separated on SDS-PAGE after overexpression in E. coli. WT and Site 2 mutants migrated with the mobility characteristic of the Zn 2ϩ -bound form, whereas all four Site 1 mutants migrated with the mobility characteristic of the Zn 2ϩ -lacking form (Fig. 1E). This result indicates that PerR SA contains a tightly bound Zn 2ϩ coordinated by four cysteine residues, and that mutations at the proposed regulatory metal binding site do not affect the Zn 2ϩ binding.
PerR SA was shown previously to be a Mn 2ϩ -specific repressor, which suggested that this protein might use an Fe 2ϩ -independent H 2 O 2 sensing mechanism (11,14). We therefore wondered whether the cysteine residues coordinating Zn 2ϩ might serve a role in peroxide sensing. To test this, we measured the rate of Zn 2ϩ release from purified PerR SA upon H 2 O 2 treatment (Fig. 1F) by monitoring the formation of a Zn 2ϩ -PAR complex as reported previously (15). The second-order rate constant of Zn 2ϩ release and the Zn 2ϩ content of PerR SA were determined to be ϳ0.05 M Ϫ1 s Ϫ1 and ϳ0.8 Zn 2ϩ /monomer, respectively, which are comparable with those of PerR BS (15). The slow rate of H 2 O 2 -mediated Zn 2ϩ release, along with the retention of Zn 2ϩ despite the use of 10 mM EDTA during the purification procedures, further supports the idea that the Zn 2ϩ site of PerR SA plays a structural rather than a peroxide sensing role. All these data together indicate that PerR SA has a structural Zn 2ϩ site coordinated by four cysteine residues and a regulatory metal binding site composed of His-43, Asp-91, His-97, His-99, and Asp-110.
In Vivo Repressor Activity of PerR SA in Comparison with PerR BS -To investigate the difference in metal-and H 2 O 2sensing ability of PerR BS and PerR SA , the repressor activities of PerR proteins were examined in MLMM using an mrgA promoter lacZ-fusion (P mrgA -lacZ). As reported previously, PerR BS repressed the P mrgA -lacZ reporter fusion in the presence of either Fe 2ϩ or Mn 2ϩ , and Fe 2ϩ -dependent repression was relieved upon H 2 O 2 treatment ( Fig. 2A) (7). PerR SA repressed the P mrgA -lacZ reporter fusion in the presence of Mn 2ϩ but not in the presence of Fe 2ϩ , consistent with the previous finding that PerR SA is a Mn 2ϩ -dependent repressor (Fig. 2B) (11). Interestingly, however, ␤-galactosidase expression was increased by about 2-fold in the presence of Fe 2ϩ as reported previously (12) and further increased upon H 2 O 2 treatment. The previous observation that this Fe 2ϩ -dependent induction of the mrgA gene is not observed with the perR null mutant S. aureus (12)  To test whether differences in cellular milieu between B. subtilis and S. aureus affect the repressor activity of PerR proteins, PerR BS and PerR SA were expressed both in B. subtilis (Fig. 2, C and E) and S. aureus (Fig. 2, D and F). PerR BS expressed in S. aureus repressed the S. aureus P mrgA -lacZ reporter fusion with even higher repressor activity than PerR SA , and responded normally to H 2 O 2 as in B. subtilis (Fig. 2D). PerR SA expressed in B. subtilis or in S. aureus repressed the B. subtilis P mrgA -lacZ fusion or S. aureus P mrgA -lacZ fusion, respectively, although not quite as efficiently as PerR BS (Fig. 2, C and D). Interestingly, the repression levels of both the B. subtilis and S. aureus P mrgA -lacZ reporter fusions by PerR SA were similar to those by PerR BS treated with H 2 O 2 for 30 min. Furthermore, PerR SA responded poorly to H 2 O 2 (ϳ1.2-fold induction) both in B. subtilis and S. aureus when compared with the responsiveness of PerR BS to H 2 O 2 (more than 3-fold induction). Thus it is likely that the Release of Zn 2ϩ from PerR SA :Zn (5 M purified PerR SA ) was measured by monitoring Zn 2ϩ -PAR complex at 494 nm after treatment of 0, 1, 10, and 100 mM H 2 O 2 as described previously (15). The Zn 2ϩ content was calculated using a molar extinction coefficient of 85,000 M Ϫ1 cm Ϫ1 at 494 nm for Zn 2ϩ -PAR complex (15). difference in responsiveness to metal and H 2 O 2 between PerR SA and PerR BS is due to differences between the PerR proteins rather than the cellular environments. In summary the in vivo data indicate that Fe 2ϩ addition appears to lead to apparent activation or derepression, rather than repression, of PerR SAregulated genes under our experimental conditions.
In Vitro PerR SA Senses H 2 O 2 by Iron-mediated Histidine Oxidation-To test whether PerR SA is activated to bind DNA by both Mn 2ϩ and Fe 2ϩ , we measured the apparent affinity of PerR SA for Fe 2ϩ and Mn 2ϩ using a fluorescence anisotropybased DNA-binding assay (Fig. 3A). Because PerR is immediately oxidized in the presence of Fe 2ϩ under aerobic conditions as reported previously (7), all the FA experiments were performed under anaerobic conditions. Consistent with the observed Mn 2ϩ -dependent repressor activity of PerR SA in vivo, the DNA-binding affinity of PerR SA was increased by the addition of Mn 2ϩ . The apparent K d for the Mn 2ϩ -dependent activation of PerR SA was determined to be 9 M, which is slightly weaker than that of PerR BS (ϳ3 M). Interestingly, despite the apparent lack of Fe 2ϩ -dependent repressor activity of PerR SA in vivo, Fe 2ϩ could also increase the DNA binding of PerR SA in a concentration-dependent manner. The apparent K d for the Fe 2ϩ -dependent activation of PerR SA (0.1 M) appeared to be the same as that of PerR BS . These results therefore suggest that the apparent lack of Fe 2ϩ -dependent repressor activity and poor responsiveness to H 2 O 2 of PerR SA in vivo is not due to a decreased Fe 2ϩ binding affinity of PerR SA per se.
Because both Fe 2ϩ and Mn 2ϩ increase the DNA binding affinity of PerR SA , we next investigated the effect of H 2 O 2 on the DNA-binding activity of different metal-bound forms of PerR SA (PerR SA :Zn,Fe and PerR SA :Zn,Mn) under anaerobic conditions (Fig. 3, B and C) To test the hypothesis that H 2 O 2 leads to a metal-dependent modification of PerR SA , we analyzed the effect of H 2 O 2 on different metallated forms of PerR SA using MALDI-TOF MS (Fig.  3D). PerR SA :Zn,Mn displayed no detectable changes in tryptic peptide peaks after 100 M H 2 O 2 treatment. However, H 2 O 2treated PerR SA :Zn,Fe exhibited a significant decrease in the intensity of the T5 peptide (Tyr-36 to Arg-70) and T9* peptide (Phe-90 to Lys-107) with a concomitant increase in the intensity of two tryptic peptides corresponding to T5 ϩ 16 and T9* ϩ 16. The sites of oxidation responsible for this 16-Da mass increase were mapped to His-43 (corresponding to His-37 in PerR BS ) in the T5 peptide and His-97 (corresponding to His-91 in PerR BS ) in the T9* peptide using LC-ESI MS analysis (Fig. 1A and data not shown). Note that the T9* peptide also contains Cys-102 and Cys-105, which were detected in their fully alkylated form, indicative of no oxidation at cysteine residues after 100 M H 2 O 2 treatment, consistent with the structural role for these zinc-coordinating cysteine residues. These data indicate that PerR SA , like PerR BS , senses low levels of H 2 O 2 by Fe 2ϩmediated oxidation of either of two histidine residues, His-43 and/or His-97, which are used as regulatory metal binding ligands.
Apparent Lack of Fe 2ϩ -dependent Repressor Activity of PerR SA in Vivo Is Due to the Hypersensitivity of PerR SA to Ironmediated Oxidation under Aerobic Conditions-Since we observed the Fe 2ϩ -dependent DNA-binding activity and Fe 2ϩmediated histidine oxidation of PerR SA in vitro, we wondered whether H 2 O 2 sensing by histidine oxidation would also occur in vivo. To monitor the oxidation of PerR SA in vivo, we analyzed the oxidation status of PerR SA recovered by immunoprecipitation from S. aureus cells grown in MLMM supplemented with Fe 2ϩ or Mn 2ϩ using MALDI-TOF MS (Fig. 4). Interestingly, even without H 2 O 2 treatment, almost all of the T5 peptide was detected as oxidized form (T5 ϩ 16) and a significant amount of the T9* peptide was also detected as oxidized form (T9* ϩ 16) for PerR SA from cells grown in MLMM supplemented with Fe 2ϩ or both Fe 2ϩ and Mn 2ϩ . However, less oxidation was observed at both T5 and T9* peptides for PerR SA from cells grown in MLMM supplemented with Mn 2ϩ or no metal ion, and no significant further oxidation of these peptides was detected upon H 2 O 2 treatment. These observations indicate that under aerobic growth conditions the majority of PerR SA : Zn,Fe exists in an oxidized form even without external addition of H 2 O 2 .
To compare the sensitivity of PerR SA with those of well known H 2 O 2 sensors, E. coli OxyR and PerR BS , we analyzed protein oxidation in E. coli using MALDI-TOF MS as described previously (22). As noted for PerR SA recovered from S. aureus cells (Fig. 4), PerR SA from aerobically grown E. coli cells exhibited a significant oxidation at the T5 peptide even in the absence of H 2 O 2 treatment (Fig. 5A). However, PerR SA from E. coli cells grown under oxygen-limited conditions exhibited no detectable oxidation at both T5 and T9* peptides, and oxidation of these peptides was observed upon external H 2 O 2 treatment (Fig. 5B). In contrast, under aerobic conditions, PerR BS exhibited less oxidation at both T5 and T11* peptides when compared with PerR SA (Fig. 5C), and E. coli OxyR exhibited no detectable oxidation of both T19 and T20 peptides, which contain the peroxidatic cysteine (Cys-199) and resolving cysteine (Cys-208), respectively (Fig. 5D). These results indicate that PerR SA is more sensitive than PerR BS or E. coli OxyR to oxidation by low levels of H 2 O 2 , which are normally encountered during the aerobic growth of E. coli.
The above observations suggest that the poor H 2 O 2 responsiveness and the apparent lack of Fe 2ϩ -dependent repressor activity of PerR SA can be overcome under oxygen-limited growth conditions where limited amounts of H 2 O 2 are generated. Consistent with this hypothesis, PerR SA exhibited an increased repressor activity under oxygen-limited growth conditions (Fig. 6A) when compared with that under aerobic growth conditions (Fig. 2D). Furthermore, under these conditions PerR SA responded to H 2 O 2 (ϳ3-fold induction) much like PerR BS , as judged by an increased ␤-galactosidase activity upon H 2 O 2 treatment. We also investigated the metal-dependent   Fig. 6B, addition of either Fe 2ϩ or Mn 2ϩ enabled PerR SA to repress the P mrgA -lacZ reporter fusion, and ␤-galactosidase activity was increased by more than 2-fold upon H 2 O 2 treatment in the presence of Fe 2ϩ or in the presence of both Fe 2ϩ and Mn 2ϩ , but only slightly (ϳ1.2-fold) in the presence of only Mn 2ϩ . These data demonstrate that PerR SA functions as a Fe 2ϩ -dependent repressor and senses H 2 O 2 in an Fe 2ϩ -dependent manner in vivo under oxygen-limited growth conditions (Fig. 6), as observed in vitro under anaerobic conditions (Fig. 3). We also note that the efficient sensing of H 2 O 2 in the presence of both Fe 2ϩ and Mn 2ϩ is consistent with the higher affinity of PerR SA for Fe 2ϩ than Mn 2ϩ as observed in vitro (Fig. 3A). In general, these results indicate that PerR SA behaves in oxygen-limited cells much like PerR BS does in B. subtilis, and that the apparent lack of Fe 2ϩ -dependent repressor activity (and thus poor H 2 O 2 responsiveness) of PerR SA under aerobic conditions is due to an efficient oxidation of PerR SA by low levels of endogenous H 2 O 2 . Oxidation status of proteins was analyzed by MALDI-TOF MS after SDS-PAGE fractionation and in-gel tryptic digestion as reported previously (22). Asterisks represent peptides containing carboxyamidomethylated cysteine residue(s).   suggested that resistance to oxidative stress is an important factor for the survival and persistence of S. aureus (14). To investigate whether the difference in sensitivity of PerR proteins to oxidation affects the H 2 O 2 resistance of S. aureus, we measured the growth of cells in the presence and absence of H 2 O 2 (Fig. 7A). As expected the perR null mutant S. aureus cells exhibited an increased H 2 O 2 resistance compared with those expressing PerR SA . Also, compared with cells expressing PerR SA , S. aureus cells expressing PerR BS (which is less sensitive to Fe-mediated oxidation than PerR SA ) exhibited an increased H 2 O 2 sensitivity. These data indicate that the ability of PerR SA to respond to very low levels of H 2 O 2 encountered during aerobic growth is important for the H 2 O 2 resistance of S. aureus. Consistent with this, high levels of KatA activity were detected in the perR null mutant S. aureus cells, but low levels of KatA activity were detected from cells expressing PerR BS , when compared with those expressing PerR SA (Fig. 7B).
Although the H 2 O 2 defense enzymes, such as KatA and AhpC, which are under the control of PerR SA play important roles in the nasal colonization and infection by S. aureus, it has been reported that they are not important for virulence (11,14). However, interestingly, PerR is known to be required for virulence in other models of infection including murine skin abscess (11,14), fruit fly (27), and zebrafish (28). We used a C. elegans model, which has been widely used as an invertebrate animal model for S. aureus pathogenesis (23), to investigate whether the difference in sensitivity of PerR proteins affects the virulence of S. aureus (Fig. 7C). As noted for other models of infection, the perR null mutant S. aureus was attenuated in the C. elegans model. Unexpectedly, S. aureus cells expressing PerR BS killed C. elegans more rapidly than did those expressing PerR SA . These results suggest that the virulence of S. aureus is somewhat directly correlated with the activity of PerR, given that heterologous PerR BS , which is less sensitive to oxidation than PerR SA , increases the virulence of S. aureus.

Discussion
PerR and PerR-like regulators have been described in a wide variety of organisms since the first characterization in B. subtilis (4,5,17,29,30). However, to date, the H 2 O 2 -sensing mechanism of PerR proteins has only been extensively studied for PerR BS (7,15,21). Here we demonstrate that PerR SA , previously regarded as a Mn 2ϩ -specific repressor, senses H 2 O 2 using the same Fe 2ϩ -dependent histidine oxidation mechanism previously described for PerR BS . Moreover we show that the apparent lack of Fe 2ϩ -dependent repressor activity, and the consequent Mn 2ϩ -specific repressor activity of PerR SA in vivo, is due to the hypersensitivity of PerR SA to H 2 O 2 under aerobic conditions, rather than due to a decreased Fe 2ϩ binding affinity of PerR SA per se.
Several lines of evidence indicate that PerR SA is a more sensitive H 2 O 2 sensor than either PerR BS or E. coli OxyR. The majority of PerR SA in aerobically grown S. aureus is detected in an oxidized form (Fig. 4), whereas only partial oxidation is observed with PerR BS from aerobically grown B. subtilis (7). However, the interpretation of this result can be complicated by potential differences in the levels of endogenous H 2 O 2 between these two species. Therefore, we directly compared the levels of oxidized PerR proteins when both were expressed in either S. aureus or B. subtilis. Indeed the direct measurement of KatA activity (Fig. 7B) and reporter fusion assays (Fig. 2, C and D (1). Normally, under these routine aerobic growth conditions, OxyR is inactive: OxyR is activated when the intracellular H 2 O 2 concentration reaches ϳ200 nM (1,32,33). PerR BS has a second-order rate constant of ϳ10 5 M Ϫ1 s Ϫ1 for inactivation by H 2 O 2 , which is comparable with that of E. coli OxyR (7). Consistent with this, PerR BS and E. coli OxyR exhibit no significant oxidation in aerobically grown E. coli (Fig.  5, C and D). In contrast, a significant oxidation of PerR SA is observed in aerobically grown E. coli without external H 2 O 2 treatment, indicating that endogenously produced H 2 O 2 is sufficient to oxidize PerR SA (Fig. 5A). Collectively these indicate that PerR SA senses very low levels of H 2 O 2 (as little as ϳ50 nM) as generated during normal aerobiosis in E. coli, levels that do not significantly oxidize PerR BS or E. coli OxyR. Corroborating with this, it has recently been reported that OxyR2 from Vibrio vulnificus is activated under normal aerobic growth conditions, whereas OxyR1, an E. coli OxyR homologue, is only activated by exogenous H 2 O 2 (34).
The efficient sensing of H 2 O 2 and induction of defense enzymes have been considered crucial for pathogens that have to fight against H 2 O 2 assault by macrophages or neutrophils (35,36). However, the perR null mutant S. aureus strain, which is more resistant to H 2 O 2 than the wild type by constitutive expression of H 2 O 2 defense enzymes, exhibits attenuated virulence in our C. elegans model of infection (Fig. 7C) as observed with other models of infection (11,27,28). Moreover, the H 2 O 2 -sensitive S. aureus strain by the expression of PerR BS (Fig. 7, A and B) is not attenuated in C. elegans (Fig. 7C), consistent with the previous finding that neither KatA nor AhpC are required for resistance to neutrophil-dependent killing or virulence of S. aureus (14). Instead, the expression of PerR BS , which is less sensitive to H 2 O 2 compared with PerR SA , increases the virulence suggesting that the activity of PerR positively correlates with the virulence of S. aureus. This may imply that the inactivation of PerR by H 2 O 2 , rather than the direct poisoning of bacteria by H 2 O 2 , can be exploited by phagocytic cells that wish to reduce the virulence of S. aureus. It is not clear why inactivation of PerR activity reduces the virulence of S. aureus. One possible explanation would be poor growth of S. aureus in the iron-limited host environment. Derepression of PerR-regulon is likely to lead to Fe 2ϩ deficiency due to the elevated expression of KatA, which consumes Fe 2ϩ , and Fur, which represses Fe 2ϩ uptake, as observed with B. subtilis (37). Alternatively, or in addition, active PerR may be involved in the induction of the virulence factor, either directly or indirectly. Indeed it has been shown that Streptococcus pyogenes PerR regulates an extracellular virulence factor, MF3, directly (38). All together, our findings that the activity of PerR is directly linked to the virulence of S. aureus suggests that PerR SA can be an attractive target for a novel approach to design new drugs for S. aureus treatment (39).
In contrast to virulence, H 2 O 2 resistance is inversely correlated with the activity of PerR because the H 2 O 2 defense systems are derepressed by H 2 O 2 -mediated PerR inactivation. Our results clearly show this relationship (Fig. 7, A and B). As reported previously the perR null mutant S. aureus exhibits increased resistance to H 2 O 2 , whereas S. aureus expressing PerR BS exhibits an increased sensitivity to H 2 O 2 than that expressing PerR SA , presumably due to the hyper-repression of PerR SA -regulated genes by PerR BS . Although KatA and AhpC are not required for virulence, they are known to play important roles for survival under aerobic conditions and especially for colonization at the anterior nares, which are the primary ecological niche for S. aureus (14). Considering that most of the PerR SA is fully oxidized and no significant further derepression of PerR regulon is triggered by H 2 O 2 treatment under aerobic conditions, it is likely that S. aureus, as a facultative anaerobic bacterium, has evolved PerR SA to sense low levels of endogenous H 2 O 2 normally encountered under aerobic environment, rather than to sense higher levels of external H 2 O 2 produced by microbial competitor or by host immune system.