Biochemical Characterization of the Structural Zn2+ Site in the Bacillus subtilis Peroxide Sensor PerR*

In Bacillus subtilis most peroxide-inducible oxidative stress genes are regulated by a metal-dependent repressor, PerR. PerR is a dimeric, Zn2+-containing metalloprotein with a regulatory metal-binding site that binds Fe2+ (PerR:Zn,Fe) or Mn2+ (PerR: Zn,Mn). Reaction of PerR:Zn,Fe with low levels of hydrogen peroxide (H2O2) leads to oxidation of two His residues thereby leading to derepression. When bound to Mn2+, the resulting PerR:Zn,Mn is much less sensitive to oxidative inactivation. Here we demonstrate that the structural Zn2+ is coordinated in a highly stable, intrasubunit Cys4:Zn2+ site. Oxidation of this Cys4:Zn2+ site by H2O2 leads to the formation of intrasubunit disulfide bonds. The rate of oxidation is too slow to account for induction of the peroxide stress response by micromolar levels of H2O2 but could contribute to induction under severe oxidative stress conditions. In vivo studies demonstrated that inactivation of PerR:Zn,Mn required 10 mm H2O2, a level at least 1000 times greater than that needed for inactivation of PerR:Zn,Fe. Surprisingly even under these severe oxidation conditions there was little if any detectable oxidation of cysteine residues in vivo: derepression was correlated with oxidation of the regulatory site. Because oxidation at this site required bound Fe2+ in vitro, we suggest that treatment of cells with 10 mm H2O2 released sufficient Fe2+ into the cytosol to effect a transition of PerR from the PerR:Zn,Mn form to the peroxide-sensitive PerR: Zn,Fe form. This model is supported by metal ion affinity measurements demonstrating that PerR bound Fe2+ with higher affinity than Mn2+.

In Bacillus subtilis most peroxide-inducible oxidative stress genes are regulated by a metal-dependent repressor, PerR. PerR is a dimeric, Zn 2؉ -containing metalloprotein with a regulatory metal-binding site that binds Fe 2؉ (PerR:Zn,Fe) or Mn 2؉ (PerR: Zn,Mn). Reaction of PerR:Zn,Fe with low levels of hydrogen peroxide (H 2 O 2 ) leads to oxidation of two His residues thereby leading to derepression. When bound to Mn 2؉ , the resulting PerR:Zn,Mn is much less sensitive to oxidative inactivation. Here we demonstrate that the structural Zn 2؉ is coordinated in a highly stable, intrasubunit Cys 4 :Zn 2؉ site. Oxidation of this Cys 4 :Zn 2؉ site by H 2 O 2 leads to the formation of intrasubunit disulfide bonds. The rate of oxidation is too slow to account for induction of the peroxide stress response by micromolar levels of H 2 O 2 but could contribute to induction under severe oxidative stress conditions. In vivo studies demonstrated that inactivation of PerR:Zn,Mn required 10 mM H 2 O 2 , a level at least 1000 times greater than that needed for inactivation of PerR:Zn,Fe. Surprisingly even under these severe oxidation conditions there was little if any detectable oxidation of cysteine residues in vivo: derepression was correlated with oxidation of the regulatory site. Because oxidation at this site required bound Fe 2؉ in vitro, we suggest that treatment of cells with 10 mM H 2 O 2 released sufficient Fe 2؉ into the cytosol to effect a transition of PerR from the PerR:Zn,Mn form to the peroxide-sensitive PerR: Zn,Fe form. This model is supported by metal ion affinity measurements demonstrating that PerR bound Fe 2؉ with higher affinity than Mn 2؉ .
Bacillus subtilis PerR is a metal-dependent repressor that regulates the peroxide-inducible expression of oxidative stress genes. The PerR regulon includes the major vegetative catalase (KatA), alkyl hydroperoxide reductase (AhpCF), enzymes of heme biosynthesis (HemAXCDBL), a zinc uptake P-type ATPase (ZosA), a Dps-like protein (MrgA), and both the PerR and Fur metalloregulatory proteins. Expression of these genes is elevated in a perR null mutant, and PerR binds directly to operator sites overlapping the promoters for these operons, establishing PerR as a direct repressor of transcription (1)(2)(3).
PerR is a member of the Fur family of metal-dependent regulators, and many Fur family proteins contain one Zn 2ϩ ion per monomer in addition to a regulatory metal ion. The x-ray absorption spectroscopy of Escherichia coli Fur (Fur EC ) revealed Zn 2ϩ bound in a tetrahedral environment composed of two S and two N/O donor ligands (4). Subsequent studies using chemical modification and mass spectroscopy assigned Cys 92 and Cys 95 as Zn 2ϩ ligands (5), a conclusion supported by site-directed mutagenesis studies (6). To date, the only structure available for a protein from the Fur superfamily is Pseudomonas aeruginosa Fur (Fur PA ) that was crystallized in the presence of Zn 2ϩ (7). Fur PA contains two metal-binding sites assigned as a high affinity Zn 2ϩ -binding site and low affinity Fe 2ϩ -binding site, respectively, by accompanying spectroscopy. No cysteine residues were involved in metal coordination in Fur PA , and there is still considerable uncertainty surrounding the functional role of the metal-binding sites. For example, multiple mutants in the corresponding residues in Bradyrhizobium japonicum Fur failed to abrogate metal sensing (8). Moreover putative Zn 2ϩ ligands in Fur PA correspond to residues in PerR required for binding the regulatory metal ion (Fe 2ϩ or Mn 2ϩ ) but not the structurally important Zn 2ϩ (9).
Most PerR-regulated genes are derepressed when cells are exposed to low levels of H 2 O 2 (Ͻ10 M), by limitation for metal ions or by transition into stationary phase during aerobic growth, which leads to an increase in the endogenous production of H 2 O 2 (2,10,11). Physiological studies have revealed that PerR-regulated genes are strongly repressed in medium supplemented with manganese (and lacking added iron) and that, under these growth conditions, the ability of H 2 O 2 to induce expression is greatly reduced. In contrast, in iron-containing growth medium PerR repression is rapidly relieved upon exposure to H 2 O 2 (2).
These results led to a model in which PerR could bind either Mn 2ϩ or Fe 2ϩ as corepressor to generate two forms of PerR that differ in their sensitivity to reactive oxygen species. Biochemical analyses of the reconstituted PerR:Zn,Fe and PerR:Zn,Mn forms of the repressor provide an explanation for this difference: the high sensitivity of PerR:Zn,Fe results from the regiospecific oxidation of two His residues directly coordinated to the activating ferrous ion (9). Thus, in contrast with most peroxide sensors characterized to date (12)(13)(14), PerR does not appear to use thiol disulfide chemistry to sense oxidative stress despite the presence of four conserved cysteine residues in the protein sequence. Previous results from homology modeling of PerR on the Fur PA structure, combined with site-directed mutagenesis, support a model in which these four residues form a high affinity Cys 4 :Zn 2ϩ structural site (9).
Although the Cys 4 :Zn 2ϩ site is not involved in sensing low levels of peroxides, this does not rule out a possible role under conditions of severe oxidative stress. Indeed it has been previously proposed that other PerR homologs sense H 2 O 2 by Cysbased redox reactions as inferred from studies done with 5-10 mM H 2 O 2 (15,16). Analysis of E. coli HSP33 indicates that cells may contain systems that respond specifically to severe oxidative stress conditions. HSP33 is a redox-regulated protein chaperone that contains a Cys 4 :Zn 2ϩ site that is only oxidized in the presence of high levels of peroxides (e.g. 4 mM H 2 O 2 ) and elevated temperatures (e.g. 43°C) (17).
In the present study, we investigated oxidation of the Cys 4 : Zn 2ϩ site and its possible role in peroxide sensing in vivo. By monitoring Zn 2ϩ release, we found a rate constant for the in vitro oxidation of this Cys 4 :Zn 2ϩ site by H 2 O 2 of ϳ0.05 M Ϫ1 s Ϫ1 . Oxidation led to the formation of intramolecular disulfide bonds between cysteine residues in the two CXXC motifs, and the resulting oxidized species could be resolved by SDS-PAGE. The in vitro sensitivity of the Cys 4 :Zn 2ϩ site to oxidation suggests that this site could mediate inactivation of the PerR: Zn,Mn form of the repressor under severe oxidative stress conditions. However, even when treated with high levels of H 2 O 2 there was little if any detectable cysteine oxidation in vivo, and derepression was correlated with oxidation of the regulatory site. Because PerR bound Fe 2ϩ with higher affinity than Mn 2ϩ , we propose a model in which high levels of H 2 O 2 release sufficient Fe 2ϩ into the cytosol to effect a transition of PerR from the PerR:Zn,Mn form to the peroxide-sensitive PerR:Zn,Fe form.
Purification of PerR after Overexpression in E. coli-E. coli cells were harvested from the 10 ml of overnight culture in LB medium and inoculated into 1 liter of fresh LB medium containing 0.4% glucose, 34 mg/liter chloramphenicol, and 100 mg/liter ampicillin. Isopropyl 1-thio-␤-D-galactopyranoside was added to final concentrations of 1 mM at A 600 of ϳ0.6, and the cells were allowed to grow for an additional 2 h. Cells were harvested by centrifugation, and the cell pellets were resuspended in 50 ml of buffer A (20 mM Tris⅐HCl (pH 8.0), 100 mM NaCl, and 5% (v/v) glycerol) containing 10 mM EDTA. After lysis by sonication, cell debris were removed by centrifugation, and the resulting supernatant was loaded onto a 20-ml heparin-Sepharose column. After application of a linear gradient of 0.1-1 M NaCl, PerR-containing fractions were pooled and buffer-exchanged to buffer A containing 10 mM EDTA using a YM-10 ultrafiltration membrane under the pressure of N 2 gas. The resultant PerR was applied to a Mono Q column using an FPLC system (GE Healthcare) and separated using a linear gradient of 0.1-1 M NaCl. The PerR-containing factions were further purified on a Superdex 200 HiLoad 16/60 column (GE Healthcare) equilibrated with Chelex-100-treated buffer A without EDTA. Purification of PerR was also carried out as above in the absence of EDTA or in the presence of DTT. Active PerR (purified in the presence of 10 mM EDTA) contained ϳ1.6 mol of zinc/dimer as judged by inductively coupled plasma MS and could be reconstituted with Mn 2ϩ to give protein with ϳ60% active molecules in a DNA binding assay. PerR purified in the absence of EDTA was also dimeric and contained stoichiometric zinc (ϳ1.9 mol of zinc and ϳ0.1 mol of iron/dimer) but was largely inactive due to oxidation of the regulatory metal site as reported previously (9).
Determination of Protein Concentration-The concentration of purified PerR was determined by measuring A 277 nm using the calculated value of ⑀ 277 nm ϭ 10,400 M Ϫ1 cm Ϫ1 . The concentration of proteins in the cell crude extract was determined by Bio-Rad protein assay and/or Bio-Rad DC protein assay (for samples containing SDS) using bovine serum albumin as standard.
Construction of FLAG-tagged PerR in B. subtilis-A partial perR gene corresponding to about 350 bp from the 3Ј-end of perR excluding the stop codon was amplified by PCR using forward primer 5Ј-GCAGGTACCGCTCTGGAAGGGAAATTTCCTA-ACATGAGC-3Ј and reverse primer 5Ј-GCACGGCCGATGAT-TTTCTTTTTTCGAACACTCTTGGCAGAC-3Ј with B. subtilis CU1065 chromosomal DNA as template. These PCR fragments were cloned into the KpnI and EagI sites of epitope protein-tagging integration vector pMUTIN-FLAG (18) resulting in plasmid pJL062. This plasmid was passed through E. coli JM105 (recA), and the resulting plasmids were used for transformation of B. subtilis CU1065. The generation of perR-FLAG, upon integration of pJL062 into perR, was verified by PCR followed by sequencing, and the resulting strain was named HB9612. The perR-FLAG open reading frame was PCR-amplified using forward primer 5Ј-CCATGTAGCCGAAAAGCTTCAAACCC-3Ј and reverse primer 5Ј-GTTTCCACCGGAATTCGCTTGCATGG-3Ј with HB9612 genomic DNA as template. These PCR fragments were cloned into the HindIII and EcoRI sites of pDG1730 resulting in plasmid pJL070. The ScaI digest of pJL070 was used for transformation of HB2080 (CU1065 perR::kan amyE::cat pheA1) or HB9703 (CU1065 perR::tet) to generate transformants containing perR-FLAG in amyE locus designated HB9620 or HB9735, respectively.
Construction of Reporter Fusion Strains-SP␤ phages from HB1122 (19), HB0618 (20), and HB8010 (21) were used for the transduction of CU1065 to give strains containing reporter fusion for mrgA-cat-lacZ, feuA-cat-lacZ, and yciC-cat-lacZ, respectively. The HB9738 strain was made by transduction of HB9735 with SP␤ phage from HB1122. Fig. 1A, purified PerR protein was incubated with 10 mM DTT, 10 mM H 2 O 2 , 10 mM diamide, or 10 mM EDTA at 80°C for 5 min in SDS sample buffer containing 2% SDS and resolved by nonreducing SDS-PAGE using a Tris-Tricine buffer system (22). For Fig. 1B, purified PerR protein was treated with 1 mM DTT, 10 mM H 2 O 2 , or 10 mM diamide at room temperature for 30 min in buffer A, and residual reagents were removed by precipitation and washing with 10% trichloroacetic acid. Recovered samples were modified by 50 mM AMS in the presence of 10 mM EDTA and 2% SDS and separated by non-reducing SDS-PAGE. For Fig. 4B, under anaerobic conditions 10 M PerR (in 20 l) was treated with 100 M Fe 2ϩ or Mn 2ϩ for 5 min and then exposed to varying amounts of H 2 O 2 for 10 min. PerR was recovered by precipitation with 20 l of 20% trichloroacetic acid. After washing with 10% trichloroacetic acid, the pellets were resuspended in buffer A containing 2% SDS, 20 mM EDTA, and 50 mM AMS and incubated for 30 min in the dark. Then the samples were analyzed by non-reducing SDS-PAGE.

SDS-PAGE Sample Preparation and Analysis-For
On-gel Zn 2ϩ Detection-PerR protein (20 g) treated with 10 mM DTT or 10 mM H 2 O 2 was resolved on non-reducing SDS-PAGE gels. The SDS-PAGE gel was soaked in buffer A containing 500 M PAR for 1 min, and subsequently 50 mM H 2 O 2 was added to release Zn 2ϩ . Because the color development (light orange) was transient, a photograph of the gel was taken within 10 min.

Analysis of SDS-PAGE-fractionated Protein by MALDI-TOF
MS-To alkylate reduced Cys residues, the SDS-PAGE gels were directly submerged and incubated in buffer A containing 50 mM EDTA and 50 mM iodoacetamide for 30 min with gentle rocking in the dark. Then the gels were stained with Coomassie Brilliant Blue R and subsequently destained. The stained bands were cut and washed three times by incubating with 100 l of 50% acetonitrile, 50 mM ammonium bicarbonate for 15 min at 37°C. The washed gel pieces were dried for 20 min in a Speed-Vac and rehydrated with 10 -20 l of trypsin solution (20 ng/l trypsin in 9% acetonitrile solution containing 40 mM ammonium bicarbonate). The samples were digested for ϳ4 h ( Fig. 2; partial digestion) or ϳ18 h ( Fig. 3; full digestion) at 37°C. A volume of 0.5 l of sample was mixed with an equal volume of matrix (saturated solution of ␣-cyano-4-hydroxysuccinamic acid in 50% (v/v) acetonitrile and 0.1% (v/v) trifluoroacetic acid) on the target plate and allowed to air dry. MALDI mass spectra were recorded with an Applied Biosystems 4700 mass spectrometer. For the analysis of oxidation of PerR:Zn,Mn (Fig. 5), PerR:Zn,Mn (10 M purified PerR ϩ 100 M MnCl 2 ) was incubated with 0, 0.1, 1, 10, or 100 mM H 2 O 2 for 10 min under anaerobic conditions, and the proteins were recovered by precipitation with 10% trichloroacetic acid, modified using 50 mM iodoacetamide, trypsinized, and analyzed using a Bruker BIFLEXIII mass spectrometer.  6), or treated with 10 mM diamide (lanes 7 and 8) for 30 min at room temperature in 20 mM Tris (pH 8.0) containing 5% glycerol and 100 mM NaCl. After precipitation and washing with 10% trichloroacetic acid, samples were incubated with 50 mM AMS in the presence of 2% SDS and 10 mM EDTA to alkylate reduced Cys residues (lanes 2, 4, 6, and 8). Lane M is protein molecular mass standards. C, purified PerR protein (20 g) was incubated alone, with 10 mM DTT or with 10 mM H 2 O 2 in SDS sample buffer. After non-reducing SDS-PAGE, gels were stained with Coomassie for protein or PAR for Zn 2ϩ ion.
Electrospray Mass Spectrometry Analysis-Electrospray ionization MS was performed using a Bruker Esquire-LC ion trap mass spectrometer (Bruker Daltonics, Breman, Germany). The ion spray voltage was set at 4 kV, the orifice voltage was set at 80 V, and the interface temperature was set at 80°C. PerR (stored in buffer A) was buffer-exchanged with 10 mM ammonium bicarbonate (pH 8.0) by using a Micro Bio-Spin 6 chromatography column. For standard acidic denaturing conditions, 25% (v/v) methanol and 1% (v/v) acetic acid in 5 mM ammonium bicarbonate was used as vector solution. For the detection of Zn 2ϩ -bound PerR, 5% (v/v) methanol in 5 mM ammonium bicarbonate was used as vector solvent. Sample volumes of 20 l were infused into the spectrometer using a syringe pump at a flow rate of 3 l/min. Mass spectra were analyzed and manipulated using the Bruker Daltonics DataAnalysis program.
Measurement of Zn 2ϩ Release by H 2 O 2 Using PAR-Under our buffer conditions (20 mM Tris (pH 8.0), 100 mM NaCl, and 5% glycerol; 25°C) the absorption maximum of Zn 2ϩ -PAR complex was observed at 494 nm, and the ⑀ 494 nm was measured to be 8.5 Ϯ 0.25 ϫ 10 4 M Ϫ1 cm Ϫ1 in the range of 0 -6 M Zn 2ϩ (compared with ⑀ 500 nm ϭ 6.6 Ϯ 0.2 ϫ 10 4 M Ϫ1 cm Ϫ1 for (PAR) 2 -Zn 2ϩ at pH 7 (23)). 5 M PerR (PerR: Zn) in buffer A (or buffer A containing 7 M urea) was treated with 0, 1, 10, or 100 mM H 2 O 2 in the presence of 100 M PAR at 25°C, and released Zn 2ϩ was measured by monitoring the Zn 2ϩ -PAR complex at 494 nm every 1 s for 30 min. Because the affinity of PAR to Zn 2ϩ was high and the rate of association of the Zn 2ϩ with PAR was much faster than that of Zn 2ϩ release from PerR (reported value of second order k for forming Zn 2ϩ -PAR complex at pH 7 is ϳ2 ϫ 10 7 M Ϫ1 s Ϫ1 ), all the released Zn 2ϩ ions were assumed to be present as Zn 2ϩ -PAR complex. The data were fitted to an effective first order rate equation, where A is absorbance at 494 nm, A o is the amplitude of absorbance, k is the second order rate constant, t is time in seconds, and B is the initial absorbance.
␤-Galactosidase Assay-Overnight cultures of cells in MM containing 10 M FeCl 3 and 10 M MnCl 2 were washed with MM and inoculated by 25ϫ dilution into fresh MM containing 10 M FeCl 3 or 10 M MnCl 2 . At an A 600 of ϳ0.6, aliquots of each culture were treated with H 2 O 2 and further incubated for 30 min. Alternatively cells were treated with H 2 O 2 for 10 min, residual H 2 O 2 was removed by centrifugation, and cells were resuspended in fresh MM containing metal ions and further incubated for 30 min. ␤-Galactosidase assays were performed as reported previously (2).   37 and His 91 residues, respectively, which are susceptible to metal-catalyzed oxidation. T5 and T7 contain Met 35 and Met 54 residues, respectively, which are susceptible to nonspecific oxidation. B, MALDI-TOF MS spectrum of the low mobility band (band 2 in Fig. 1A). T11 (containing 96 CXXC 99 ) corresponds to the peak of m/z ϭ 2083.91, T14 (containing 136 CXXC 139 ) corresponds to the peak of m/z ϭ 1369.66, and T14 ϩ T15 corresponds to the peak of m/z ϭ 1497.75. Note that each of these peaks is characterized by a loss of 2 m/z units from the calculated m/z value, indicative of disulfide bond formation between Cys residues. C, MALDI-TOF MS spectrum of high mobility band (band 1 in Fig. 1A). T11* (m/z ϭ 2199.93), T14* (m/z ϭ 1485.70), and T14 ϩ T15*(m/z ϭ 1613.79) all display a 114-Da mass increase from their calculated m/z values, indicative of two carbamidomethyl-modified Cys residues. Purified PerR was subjected to SDS-PAGE, and reduced Cys residues were alkylated by treating the whole SDS-PAGE gel with 50 mM iodoacetamide. Then band 1 and band 2 (see Fig. 1A, lane 1) were cut and analyzed by in-gel tryptic digestion followed by MALDI-TOF analysis. Peaks were assigned based on calculated and observed m/z values (⌬m/z Ͻ 0.1) except for T11, T14, and T14 ϩ T15. Peaks containing an intrapeptide disulfide bond are shown with filled triangles, and peaks containing two carbamidomethyl-modified Cys residues are shown with open triangles and marked with asterisks. aliquots of 10-ml culture were mixed with 1.1 ml of trichloroacetic acid at a given time point before and after H 2 O 2 treatment. After centrifugation, pellets were resuspended in 10% trichloroacetic acid, split into two fractions, and washed with acetone. Each fraction was sonicated in 150 l of 20 mM iodoacetamide or 20 mM AMS solution containing 200 mM Tris (pH 8.0), 5% glycerol, 1% SDS, and 20 mM EDTA. Alkylations were performed at room temperature in the dark for 1 h, and ϳ10 l (corresponding to 30 g of protein as assayed by DC protein assay (Bio-Rad)) of alkylated samples were resolved by SDS-PAGE. After electrophoresis, proteins were blotted to a polyvinylidene difluoride membrane and probed with polyclonal anti-FLAG antibody from rabbit and anti-rabbit antibody conjugated with alkaline phosphatase.

Analysis of in Vivo Oxidation Status of Cys Residues by AMS
Analysis of in Vivo Oxidation of PerR by MS-HB9738 cells were grown overnight in 25 ml of MM containing 10 M FeCl 3 and 10 M MnCl 2 . Cells were washed two times with MM and inoculated into 0.8 liter of MM supplemented with 10 M MnCl 2 . At an A 600 of ϳ0.5, aliquots of 250-ml cell culture were treated with 0, 0.1, or 10 mM H 2 O 2 for 5 min, harvested by centrifugation at 6000 rpm for 10 min, and washed two times with Tris-buffered saline (50 mM Tris (pH 7.4) and 150 mM NaCl) containing 20 mM EDTA (TBS ϩ EDTA). Cells were sonicated in 1.3 ml of TBS ϩ EDTA containing protease inhibitor mixture, 1 mg/ml lysozyme, and 2% Triton X-100. After removal of cell debris by centrifugation at 16,100 ϫ g for 10 min, the resulting supernatants were incubated with 40 l of anti-FLAG M2 affinity gel for 1 h and then washed with 1 ml of TBS ϩ EDTA three times. Anti-FLAG M2 affinity gel suspensions containing PerR-FLAG were modified with 50 mM iodoacetamide in the presence of 20 mM EDTA and 1% SDS for 30 min in the dark. PerR-FLAG proteins were recovered using SDS-PAGE and analyzed by MALDI-TOF MS after in-gel tryptic digestion (see above).

RESULTS
Purified PerR Contains Stably Bound Zn 2ϩ -As reported previously, PerR is susceptible to oxidative inactivation during purification, and this process is exacerbated by the presence of thiol-reducing agents and the absence of a metal-chelating agent (9). Consequently PerR was overexpressed in E. coli and purified in the presence of 10 mM EDTA and in the absence of thiol-reducing agents. As purified, PerR contained one stably bound Zn 2ϩ per monomer as judged by both inductively cou-pled plasma MS metal analysis and the appearance of a protein peak corresponding to a monomeric, Zn 2ϩ -bound species in electrospray ionization mass spectroscopy under non-acidic conditions (see supplemental Fig. 1S).
When analyzed by SDS-PAGE, purified PerR routinely displayed two bands (Fig. 1A). The lower mobility band (band 2) migrated as expected given the molecular mass of the PerR monomer (16.3 kDa) and was increased in intensity when samples were treated with thiol-oxidizing agents (H 2 O 2 or diamide) or the metal chelator EDTA. The higher mobility band (band 1) was increased in intensity by DTT (lane 2). These results suggest that band 1 may contain Zn 2ϩ that remains associated with PerR even during electrophoresis in the presence of SDS. Removal of Zn 2ϩ from PerR was observed when EDTA was present in the SDS sample buffer (lanes 5 and 6) but not when native protein was treated with EDTA (e.g. during purification).
We used the cysteine-alkylating agent AMS to determine whether all four Cys residues in PerR were reduced. For this experiment, samples were treated with DTT, H 2 O 2 , or diamide for 30 min in the absence of SDS prior to analysis of AMS reactivity (in the presence of SDS and EDTA to remove associated Zn 2ϩ ). Treatment of purified PerR with AMS resulted in a single band corresponding to the addition of four AMS moieties to PerR (Fig. 1B, lane 2). The appearance of a single modified band despite the presence of two bands in the starting material (Fig. 1A, lane 1) is consistent with the hypothesis that the difference in mobility of bands 1 and 2 is not due to proteolysis but instead reflects the presence or absence of bound Zn 2ϩ . Oxidation of Cys residues was apparent after treatment of PerR with 10 mM H 2 O 2 for 30 min (as judged by the reduced susceptibility of the protein to AMS modification; lane 6) but not after exposure to diamide (lane 8). Thus, the ability of diamide to oxidize PerR (Fig. 1A, lane 4) requires the presence of harsher conditions (e.g. SDS and heating) relative to H 2 O 2 . Indeed PerR:Zn,Mn retains DNA binding activity even after treatment with 50 mM diamide for 15 min (data not shown).
These results suggest that band 1 corresponds to monomeric PerR containing bound Zn 2ϩ , whereas band 2 represents PerR that lacks bound Zn 2ϩ . This supposition is supported by direct staining of the SDS-PAGE gels with the Zn 2ϩ -binding dye, PAR. Zn 2ϩ was clearly associated with the protein as purified and after DTT treatment but not with the H 2 O 2 -treated sample (Fig. 1C). Moreover in the presence of DTT, incubation of PerR with Zn 2ϩ , but not other divalent cations, greatly increased the fraction of band 1 (data not shown), whereas all four single Cys 3 Ser mutants migrate in the position of band 2 during SDS-PAGE, suggesting that Zn 2ϩ binding has been eliminated (9).
PerR Apoprotein Forms Intrasubunit Disulfide Bonds-We initially speculated that the mobility difference between bands 1 and 2 might originate from an N-or carboxyl-terminal proteolytic cleavage during SDS-PAGE sample preparation. To explore this possibility, we treated an SDS-polyacrylamide gel with iodoacetamide prior to Coomassie Blue staining and destaining (to prevent oxidation of Cys residues) and analyzed the two bands using in-gel tryptic digestion and MALDI-TOF MS analysis. We detected the amino-terminal tryptic peptide (T1) (in Met-excised form) in protein from both bands 1 and 2 implying that there was no difference at the amino terminus (consistent with electrospray ionization MS data of intact protein; see supplemental Fig. 1S). Although the carboxyl-terminal tryptic peptide (T15 and T16 K and ENH) could not be detected under the m/z range of our experiments, we did detect carboxyl-terminal epitope tag in both bands 1 and 2 by immunoblot-ting analysis of PerR-FLAG (data not shown). These data are consistent with the hypothesis that the primary difference between bands 1 and 2 is the presence of bound Zn 2ϩ in band 1.
In the spectrum of PerR isolated from band 2, the tryptic peptides containing 96 CXXC 99 and 136 CXXC 139 (T11, m/z ϭ 2083.91; T14, m/z ϭ 1369.66; and T14 ϩ T15, m/z ϭ 1497.75) were each detected with a loss of 2 m/z units from the calculated values. This mass change is indicative of an intrafragment disulfide bond formation. No peaks corresponding to interfragment disulfide bond formation were detected in this air-oxidized protein. In contrast, peptides from band 1 contained reduced Cys residues as judged by their quantitative modification by iodoacetamide. The modified peptides (labeled as T11*, m/z ϭ 2199.93; T14*, m/z ϭ 1485.70; and T14 ϩ T15*, m/z ϭ 1613.79) display a 114-Da mass increase (57 Da for each carbamidomethyl-modified Cys residue). Thus, PerR in band 1 is fully reduced as expected for a Cys 4 :Zn 2ϩ site, whereas PerR in band 2 is oxidized. In this case, air oxidation leading to the appearance of band 2 likely occurs after unfolding of the protein by SDS because AMS modification (Fig. 1B) and iodoacetamide modification followed by electrospray ionization MS (data not shown) indicate that all four Cys residues are fully reduced in the protein as purified.

Oxidation of PerR Apoprotein with H 2 O 2 Leads to Both Disulfide Bond and Sulfonic Acid Formation-
To monitor disulfide bond formation during H 2 O 2 -mediated protein oxidation, we treated PerR:Zn with either 10 or 100 mM H 2 O 2 and separated the resulting proteins by SDS-PAGE. Under these severe oxidation conditions, two additional protein bands appear (bands 3 and 4; Fig. 3A). Bands 2, 3, and 4 from the protein treated with 100 mM H 2 O 2 were analyzed by MALDI-TOF MS. As before, band 2 corresponds to protein containing two intrasubunit disulfide bonds (Cys 96 -Cys 99 and Cys 136 -Cys 139 ). Note that even after treatment with 100 mM H 2 O 2 , there was no mass increase cor- responding to the oxidation of His 91 (in peptide T11), consistent with the requirement for Fe 2ϩ for modification of ligands in the high sensitivity peroxide-sensing site (9). Under these conditions, peptides containing Met residues are either partially (T7) or fully (T5) oxidized as judged by a ϩ16-Da mass increase. This presumably corresponds to the formation of methionine sulfoxide as also documented previously for the iron-containing form of the protein (9). Protein in band 3 contains one disulfide bond (Cys 136 -Cys 139 ) and two sulfonic acids (at Cys 96 and Cys 99 ). There was no evidence for sulfonic acid formation in the peptide containing Cys 136 and Cys 139 . Protein in the faster migrating band 4 contains cross-links between the tryptic peptide containing Cys 96 and Cys 99 (T11) and that containing Cys 136 and Cys 139 (T14). This band contains approximately equal levels of two distinct species. The 3452.39-Da peptide corresponds to two disulfide cross-links between T11 and T14 (which may include the parallel or antiparallel doubly cross-linked peptides or both), whereas the 3550.40 peak corresponds in mass to one disulfide bond and two sulfonic acids. There was little if any evidence of the non-cross-linked peptides, T11 and T14.
Together these results indicate that the mobility of PerR during SDS-PAGE is a sensitive indicator of conformational changes elicited by Zn 2ϩ binding and by disulfide bond formation. Oxidation leading to disulfide bond formation between vicinal Cys residues (in the two CXXC motifs) led to a lower mobility species (band 2) that was further retarded under conditions leading to sulfonic acid formation (band 3). In contrast, the more compact conformation of PerR due to the presence of bound Zn 2ϩ (band 1) could be mimicked by one or more disulfide linkages between the two CXXC motifs (band 4). These results are consistent with the previously proposed model of an intrasubunit Cys 4 :Zn 2ϩ site as supported by homology modeling of PerR on the Fur PA structure and by the observation that all four PerR Cys 3 Ser mutant proteins migrated at a position corresponding to band 2 during SDS-PAGE analysis (supplemental material in Ref. 9).
Oxidation of PerR Leads to Release of Zn 2ϩ Independent of Metal Binding at the Regulatory Site-Oxidation of the Cys 4 :Zn 2ϩ site in the stress-regulated E. coli chaperone HSP33 occurs under conditions of severe heat and oxidative stress and activates chaperone activity (24). We reasoned that the Cys 4 :Zn 2ϩ site in PerR might also serve a regulatory role in vivo. Specifically under conditions in which PerR exists predominantly in the manganese-cofactored form (PerR:Zn,Mn), oxidation of the Zn 2ϩ site might still allow derepression under severe oxidative stress.
To test this hypothesis, we first measured the rate of Zn 2ϩ release from PerR:Zn in the presence of H 2 O 2 . Release of Zn 2ϩ , as monitored by formation of the colored PAR complex, was dependent on added H 2 O 2 with a second order rate constant of 0.054 Ϯ 0.009 M Ϫ1 s Ϫ1 (Fig. 4A). This corresponds to a half-time of 21.4 min in the presence of 10 mM H 2 O 2 . Even in the presence of 7 M urea, the rate of H 2 O 2 -mediated Zn 2ϩ release was very slow (0.119 Ϯ 0.029 M Ϫ1 s Ϫ1 ; data not shown). These data are consistent with a structural role for the Cys 4 :Zn 2ϩ site that helps maintain a locally folded domain even in the presence of protein denaturants such as urea and SDS.
To monitor oxidation of the Cys 4 :Zn 2ϩ site in the PerR:Zn,Fe and PerR:Zn,Mn forms of the protein we used AMS to monitor the number of reduced cysteines (Fig. 4B). In each case, no significant oxidation of protein was detected by 1 mM H 2 O 2 , and ϳ30% of the protein was fully oxidized by 10 mM H 2 O 2 . Complete oxidation of all four Cys residues could be attained with 100 mM H 2 O 2 . These results suggest that the rate of oxidation of the Cys 4 :Zn 2ϩ site is similar in the PerR:Zn (Fig. 1B,  lane 6), PerR:Zn,Fe, and PerR:Zn,Mn forms of the repressor (Fig. 4B). The rate of oxidative inactivation measured here also explains the previously reported 50% inactivation of DNA binding activity noted for reconstituted PerR:Zn,Mn by treatment with 10 mM H 2 O 2 for 20 min (9). In contrast, the DNA binding activity of the PerR:Zn,Fe form is highly sensitive to peroxide inactivation (estimated k inact , ϳ10 5 M Ϫ1 s Ϫ1 ) due to histidine oxidation at the Fe 2ϩ -binding site (9).
We also monitored the effects of high levels of H 2 O 2 on reconstituted PerR:Zn,Mn protein by MALDI-TOF MS studies (Fig. 5). Consistent with the AMS modification studies, iodoacetamide modification was reduced in the 10 mM H 2 O 2 -treated samples and eliminated in the samples treated with 100 mM H 2 O 2 . There was a near quantitative ϩ16-Da mass shift in the two peptides containing Met residues (T5 and T7) but no evidence of significant oxidation of the peptide containing His 91 (T11). Thus, PerR:Zn,Mn is insensitive to oxidation in vitro at the Mn 2ϩ -binding site, even under severe conditions, and is instead inactivated by oxidation of the structural Cys 4 :Zn 2ϩ site.
Inactivation of PerR:Zn,Mn Under Severe Oxidative Stress in Vivo Is Not Mediated by Cys Oxidation-In numerous studies, we have observed that the induction of PerR-regulated genes is highly sensitive to the metal composition of the growth medium. In medium containing iron or iron and manganese, induction is efficient with complete derepression elicited by Ͻ100 M H 2 O 2 or by exposure to NO-generating agents. In contrast, in low iron minimal medium supplemented with manganese, the resulting PerR:Zn,Mn form of the repressor is relatively insensitive to peroxide-and NO-mediated inactivation (2,10,25).
To determine whether oxidation of the Cys 4 :Zn 2ϩ site might contribute to PerR regulation in vivo, we first determined the levels of H 2 O 2 needed to inactivate the PerR:Zn,Mn form of the repressor. Whereas induction in iron-containing minimal medium is readily observed with 0.1 mM H 2 O 2 , it took 10 mM H 2 O 2 to induce expression in the manganese-supplemented minimal medium (Fig. 6A). Similar results were observed in a strain expressing PerR-FLAG rather than PerR (data not shown). Although Fur also contains a structural Zn 2ϩ ion, we were unable to detect peroxide-mediated inactivation of Fur (Fig. 6A). Similarly there was no evidence for derepression of a reporter fusion (P yciC -cat-lacZ) regulated by the third Fur paralog in B. subtilis, Zur (data not shown). The high selectivity of peroxide inactivation is consistent with previous transcriptome analyses: treatment of cells with micromolar levels of H 2 O 2 led to derepression of PerR-regulated genes but not genes controlled by Fur or Zur (10).
Based on our in vitro analyses, it seemed likely that the in vivo inactivation of PerR:Zn,Mn by 10 mM H 2 O 2 was due to oxidation of Cys residues associated with the structural Zn 2ϩ site. To test this hypothesis directly, we used AMS modification to trap reduced Cys residues before and after treatment of cells with 10 mM H 2 O 2 . Under all conditions tested (Fig. 6B and data not shown), all four Cys residues were fully modified by AMS. In some experiments, there was a faint band corresponding to unmodified PerR after treatment, consistent with formation of either disulfide or sulfonic acid modifications (data not shown). However, this band was always a minor species, and it seemed unlikely that this small amount of oxidation was sufficient to account for the observed derepression (Fig. 6A). These data led us to reject the hypothesis that inactivation of PerR in vivo, in the presence of 10 mM H 2 O 2 , is principally mediated by Cys oxidation.  Fig. 2. CXXC motif-containing peptides (T11 and T14) are exclusively detected as fully alkylated forms up to 1 mM H 2 O 2 treatment. In samples treated with 100 mM H 2 O 2 (E), CXXC motifcontaining peptides are detected without alkylation at Cys residues. Note that T11 shows no increase in oxidation at His 91 residue despite the full oxidation at Cys 96 and Cys 99 residues by 100 mM H 2 O 2 treatment (T11 contains a small amount of oxidation as judged by the small peak with a ϩ16-Da mass shift. This oxidation is present in the purified PerR protein and is not further increased by peroxide treatment of Mn 2ϩ -supplemented protein. This corresponds to His 91 oxidation and correlates with the observation that the protein, as purified, is ϳ60% active for DNA binding.) T7 containing Met 54 gains ϩ16 m/z units, indicative of methionine sulfoxide formation. By analogy, the gain of ϩ16 m/z units on T5 seems to indicate the oxidation at Met 35 rather than His 37 .

In Vivo Oxidation of PerR:Zn,Mn Is Correlated with Regulatory Site Oxidation-
To determine what modification(s) might be associated with PerR:Zn,Mn inactivation in vivo, we immunoprecipitated FLAG epitope-tagged PerR from B. subtilis cells after treatment with either 0.1 or 10 mM H 2 O 2 (Fig. 7A). MALDI-TOF analysis of in situ trypsin-digested PerR was consistent with the presence of four reduced Cys residues under all conditions: in each case the T11 and T14 tryptic peptides were fully alkylated by iodoacetamide. In the sample treated with 10 mM H 2 O 2 , the T11 peptide appeared as a doublet with ϳ30 -40% oxidation corresponding to a species previously shown to represent His 91 oxidation (9). The other histidine residue previously shown to be oxidized by Fe 2ϩ -mediated hydroxyl radical modification is in peptide T5. This peptide appeared as a doublet (with a 16-Da mass shift) in all three samples, likely due to methionine sulfoxide formation during sample preparation and analysis (Met oxidation is also consistent with the doublet noted for T7, the other Met-containing peptide). With this background oxidation and the poor recovery of highly oxidized T5 peptides noted during matrix desorption, it is difficult to assess the level of His 37 oxidation. Nevertheless these results indicate that the in vivo oxidation of PerR:Zn,Mn leads to oxidation of His 91 in the regulatory metal-binding site but not to significant Cys oxidation.
Treatment of reconstituted PerR:Zn,Mn (in the absence of Fe 2ϩ ) with 10 -100 mM H 2 O 2 oxidized the Cys 4 :Zn 2ϩ site but did not lead to appreciable His oxidation (Fig. 5). In contrast, in vivo oxidation in cells containing PerR:Zn,Mn led to His 91 oxidation with no significant Cys oxidation. Because this pattern of oxidation is characteristic of PerR:Zn,Fe (9), we suggest that treatment of cells with 10 mM H 2 O 2 led to the release of sufficient intracellular Fe 2ϩ to allow a shift in the active species of PerR from PerR:Zn,Mn to PerR:Zn,Fe.
PerR:Zn Binds Fe 2ϩ with Higher Affinity than Mn 2ϩ -We next determined the affinity of PerR:Zn for Fe 2ϩ and Mn 2ϩ  using a fluorescence anisotropy-based DNA binding assay (Fig.  8). The apparent K d for activation by Fe 2ϩ (ϳ0.1 M) is significantly greater than that for Mn 2ϩ (ϳ2.8 M). Although lower absolute affinities were observed in several replicate experiments (particularly for Fe 2ϩ , perhaps due to trace oxygen contamination), the data are generally consistent with a significant preference for Fe 2ϩ relative to Mn 2ϩ . Indeed we previously demonstrated that 10 M Mn 2ϩ is unable to protect PerR:Zn against oxidative inactivation in the presence of 10 M Fe 2ϩ , and only partial protection is afforded by 100 M Mn 2ϩ (9). We also estimated the affinity for Mn 2ϩ using electrophoretic mobility shift assay experiments. By inclusion of Mn 2ϩ in the binding and running buffers as well as in the polyacrylamide gel, we could measure the apparent DNA binding affinity of PerR as a function of Mn 2ϩ concentration (supplemental Fig. S2). Under these assay conditions, DNA binding affinity increased 40-fold as the Mn 2ϩ concentration was increased from 1 to 100 M. Together these studies indicate that PerR:Zn is activated for DNA binding by Mn 2ϩ in the low micromolar range and by Fe 2ϩ with even higher affinity.

DISCUSSION
Cells respond to oxidative stress by the inducible synthesis of a variety of protective enzymes and proteins. Most peroxide sensors characterized to date use cysteine thiolates to detect peroxides, often leading to the formation of either intra-or intermolecular disulfides (12)(13)(14). In contrast, B. subtilis PerR has been shown recently to detect low levels (Ͻ100 M) of H 2 O 2 by iron-catalyzed oxidation of either of two His residues, His 37 and His 91 (9). These two His residues are required for Fe 2ϩ binding and are predicted, by homology modeling, to be direct ligands to the regulatory metal ion. PerR also contains a tightly bound Zn 2ϩ ion (1) proposed to exist as part of a Cys 4 :Zn 2ϩ site (9). In the present study we investigated the redox sensitivity of this Cys 4 :Zn 2ϩ site and assessed the role of this site in sensing high levels of H 2 O 2 both in vitro and in vivo.
PerR retained tightly bound Zn 2ϩ even after purification in the presence of EDTA. In the course of our studies, we noted a correlation between conditions favoring Zn 2ϩ binding and the appearance of a more rapidly migrating form of PerR during SDS-PAGE (band 1). Using MALDI-TOF analysis of in situ digested protein samples, we demonstrated that non-reducing SDS-PAGE provides a sensitive assay for monitoring the status of the Cys 4 :Zn 2ϩ site. When this site was intact, PerR migrated as a 14.5-kDa protein, whereas when Zn 2ϩ was lost the resulting protein migrated at the expected position of 16.4 kDa. Loss of Zn 2ϩ was accompanied by oxidation of PerR to a disulfidebonded form. When treated with high levels of H 2 O 2 (in the presence of protein denaturant and heat) two additional bands appeared including a slowly migrating band in which PerR contains two sulfonic acid residues and one disulfide bond (band 3) and a more rapidly migrating band containing one or two long range disulfide bonds (band 4).
Because many Fur homologs are known or predicted to contain a structural Zn 2ϩ site, our results may provide an explanation for the observation that many Fur homologs migrate as a doublet during SDS-PAGE. Examples include B. subtilis and Bacillus cereus Fur (26,27), Streptomyces reticuli FurS (15), and Streptomyces coelicolor CatR (16). However, analysis of E. coli Fur by amino-terminal sequencing and by electrospray ionization MS revealed that the fast migrating band is a proteolytic product of intact protein lacking the amino-terminal nine amino acids (28,29). Recently it has been noted that Zn 2ϩ binding is required for stable folding of the carboxyl-terminal domain of E. coli Fur and for protein dimerization (30).
Previously we (1) and others (15,16) had suggested that PerR and its homologs might sense H 2 O 2 by Cys-based redox reactions. However, each of these studies used high levels of H 2 O 2 (5-10 mM), and the physiological role, if any, of this Cys-based oxidation has not been established. In the present study we demonstrated that the structural Cys 4 :Zn 2ϩ site in PerR was remarkably resistant to H 2 O 2 with an observed second order rate constant for Zn 2ϩ release of ϳ0.054 M Ϫ1 s Ϫ1 (corresponding to a half-time for inactivation of ϳ20 min by 10 mM H 2 O 2 ). A similar oxidation rate was determined by several assays including Zn 2ϩ release studies (using PerR:Zn; Fig. 4A), monitoring of Cys status by AMS modification of PerR:Zn,Fe and PerR:Zn,Mn (Fig. 4B), and monitoring of Cys status by iodoacetamide modification of PerR:Zn,Mn followed by MALDI-TOF MS (Fig. 5) and by measuring oxidative inactivation of DNA binding activity for PerR:Zn,Mn (9). Significantly the Cys 4 : Zn 2ϩ site in PerR was even less reactive than free cysteine (2-20 M Ϫ1 s Ϫ1 ; Refs. 31-33), the major low molecular weight thiol in B. subtilis (34). Thus, these Cys residues are actually protected against peroxidative attack by Zn 2ϩ coordination. Comparison of the rate reported here for oxidation of the Cys 4 :Zn 2ϩ site (k inact ϳ 0.05 M Ϫ1 s Ϫ1 ) with the rate of peroxide-mediated inactivation of PerR:Zn,Fe (k inact ϳ 10 5 M Ϫ1 s Ϫ1 ) highlights the fact that the structural Zn 2ϩ and regulatory Fe 2ϩ sites have vastly different sensitivities toward oxidation.
The oxidative modification of thiolates coordinated to Zn 2ϩ has been implicated in several other redox-sensitive proteins. The S. coelicolor RsrA antifactor is regulated by disulfide stress. Treatment with diamide, a thiol-specific oxidant, leads to Zn 2ϩ release and inactivation of the anti-. This results in activation of R and expression of thiol reductants such as thioredoxin and thioredoxin reductase (35,36). E. coli HSP33 is a redox-regulated protein chaperone that contains a high affinity Cys 4 :Zn 2ϩ site (17). Oxidation of this Cys 4 :Zn 2ϩ site, accompanied by Zn 2ϩ release, activates this protein chaperone, which can functionally replace the redox-sensitive DnaK protein under severe oxidative stress conditions (24). Activation of HSP33 requires harsh conditions including high levels of peroxides (e.g. 4 mM H 2 O 2 ) and elevated temperatures (e.g. 43°C). B. subtilis also contains an HSP33 ortholog (37), but the redox regulation of this protein has not yet been characterized.
Because the Cys 4 :Zn 2ϩ site in PerR is structurally similar to the regulatory site in HSP33, we speculated that oxidation at this metal center might provide a backup mechanism for induction under severe oxidative stress conditions for iron-starved cells (in which PerR would be present in the PerR:Zn,Mn form). Indeed in vitro oxidation of reconstituted PerR:Zn,Mn by 10 or 100 mM H 2 O 2 was correlated with Cys oxidation and not with oxidation of His residues in the regulatory metal-binding site (Fig. 5). In contrast, in vivo treatment of manganese-supplemented cells with 10 mM H 2 O 2 (a level sufficient to induce a PerR-regulated reporter; Fig. 6) did not lead to PerR Cys oxidation (Fig. 7). The lack of Cys oxidation in vivo may be explained by the lower effective concentration of H 2 O 2 in vivo due to the presence of catalase, peroxidases, and competition from low molecular mass thiols such as cysteine. Instead induction was correlated with oxidation of at least one of the two His residues previously implicated in peroxide sensing (9). Oxidation of these His residues is diagnostic for PerR:Zn,Fe. It is likely that exposure of cells to high levels of peroxide releases sufficient Fe 2ϩ to allow PerR to switch from the relatively insensitive Per-R:Zn,Mn form to the thermodynamically preferred and much more peroxide-sensitive PerR:Zn,Fe form.
To evaluate this model it is necessary to determine (i) the relative affinity of PerR:Zn for Fe 2ϩ and Mn 2ϩ and (ii) the levels of free metal ions in the cytosol under different growth conditions. PerR:Zn binds Fe 2ϩ with higher affinity than Mn 2ϩ as judged by both metal-dependent DNA binding measurement (Fig. 8) and the ability of Mn 2ϩ to protect against Fe 2ϩ -catalyzed protein inactivation (9). This is consistent with the observation that PerR:Zn,Mn only forms in cells grown under ironlimiting conditions. Although it is difficult to measure the levels of free metal ions in the cell, it is possible to make reasonable estimates. For example MntR, the B. subtilis protein responsible for regulating Mn 2ϩ levels in the cell binds two Mn 2ϩ per monomer with apparent dissociation constants (at pH 8.0) of 0.2-2.0 and 5-13 M (38). Somewhat lower affinities were reported based on EPR measurements of Mn 2ϩ binding by the Bacillus anthracis MntR ortholog with average dissociation constants in the range of ϳ50 M (39). Because MntR represses Mn 2ϩ uptake when in its Mn 2ϩ -bound state, it is reasonable to assume that intracellular levels of free Mn 2ϩ are normally maintained at levels sufficient to saturate PerR (to generate the PerR:Zn,Mn form of the repressor). The levels of exchangeable Fe 2ϩ in the cytosol are difficult to measure directly, but in E. coli chelatable iron levels of 10 -30 M have been measured by EPR in the presence of desferrioxamine (40,41). If similar values pertain to B. subtilis, this would explain why, in the presence of both iron and manganese in the growth medium, PerR is predominantly in the peroxide-sensing PerR:Zn,Fe form. Apparently in cells grown in manganese-supplemented minimal medium (with no added iron), the intracellular level of iron is lowered sufficiently to allow the PerR:Zn,Mn form of the repressor to predominate (2,25). Yet because of the relatively higher affinity for Fe 2ϩ , a small increase in intracellular free Fe 2ϩ could shift the repressor from the PerR:Zn,Mn to the Per-R:Zn,Fe form. Treatment of cells with 10 mM H 2 O 2 is likely to damage iron-sulfur clusters (42) and perhaps other non-heme iron proteins, leading to elevation of free Fe 2ϩ in the cell.