Role of the ArcAB two-component system in the resistance of Escherichia coli to reactive oxygen stress

Background The global regulatory system ArcAB controls the anaerobic growth of E. coli, however, its role in aerobic conditions is not well characterized. We have previously reported that ArcA was necessary for Salmonella to resist reactive oxygen species (ROS) in aerobic conditions. Results To investigate the mechanism of ROS resistance mediated by ArcAB, we generated deletion mutants of ArcA and ArcB in E. coli. Our results demonstrated that both ArcA and ArcB were necessary for resistance to hydrogen peroxide (H2O2), a type of ROS, and their function in this resistance was independent from H2O2 scavenge. Mutagenesis analysis of ArcA indicated that ROS resistance was mediated through a distinct signaling pathway from that used in anaerobic conditions. An abundant protein flagellin was elevated at both the protein and mRNA levels in the ΔarcA mutant as compared to the wild type E. coli, and deletion of flagellin restored the resistance of the ΔarcA mutant to H2O2. The resistance of the ΔarcA mutant E. coli to H2O2 can also be restored by amino acid supplementation, suggesting that a deficiency in amino acid and/or protein synthesis in the mutant contributed to its susceptibility to H2O2, which is consistent with the notion that protein synthesis is necessary for ROS resistance. Conclusion Our results suggest that in addition to its role as a global regulator for anaerobic growth of bacteria, ArcAB system is also important for bacterial resistance to ROS in aerobic conditions, possibly through its influence on bacterial metabolism, especially amino acid and/or protein assimilation and synthesis.


Background
Aerobic bacteria use oxygen as a terminal electron acceptor in oxygen-containing environments for their metabolism. Although aerobic growth has its obvious advantages (e. g. high energy efficiency, abundance of oxygen in the atmosphere, etc), bacteria must deal with the undesired consequences from exposure to oxygen and oxidative environments. Oxygen and its derivatives, such as superoxide and hydrogen peroxide, are often highly reactive and pose a threat to many macromolecules, such as enzymes with iron-sulfur centers, nucleic acids, and lipids. Therefore, bacteria undergoing aerobic growth must be able to sense, respond to, and detoxify reactive oxygen species (ROS), and maintain their structural and functional integrities.
Although the function of the ArcAB system in the anaerobic growth of E. coli has been well characterized, its function is unlikely to be limited to those required for the anaerobic growth of bacteria. For example, the ArcAB system has been reported to be involved in chromosomal replication, stress responses and aging of bacteria [35][36][37]. We have previously reported that ArcA of Salmonella enterica is necessary for its resistance to reactive oxygen and nitrogen species (ROS and RNS) [38]. More recently, ArcA is implicated in the ROS stress response of Haemophilus influenzae [39]. In this report, we analyzed the role of ArcAB in reactive oxygen resistance of E. coli and investigated the mechanism of ROS resistance mediated by the ArcAB two-component system.

ArcAB system is necessary for E. coli to resist hydrogen peroxide (H 2 O 2 )
To determine if the ArcAB global regulatory system plays a role in the survival of E. coli under stress by reactive oxygen species (ROS), we generated deletion mutants of ArcA (the global regulator) and ArcB (the cognate sensor-kinase of ArcA) in E. coli (Table 1). Both arcA and arcB mutant E. coli formed smaller colonies than their parental E. coli, but otherwise showed similar colony morphology. The arcA and arcB mutant E. coli were tested for their growth properties in complete (Luria Bertani broth) or minimal (M9) medium with glucose as carbon source. Overnight culture of each bacterial strain was diluted 1:100 in LB or M9 medium, and the growth of bacteria was measured by the optical density of the culture at 550 nm (OD 550 nm ) every 2 hours for 8 hours and then at 24 hours. This incubation period includes both log phase of growth and stationary phase of bacteria. We found that OD 550 nm of both arcA and arcB mutants appeared to be lower than that of the wild type E. coli during the log phase of growth. However, both mutants had similar bacterial concentrations and growth curves to those of the wild type E. coli when their growth was quantified by plating ( Figure 1B and  1D). Therefore, no gross defect was observed in arcA and arcB mutants in spite of lower OD 550 nm of their cultures. The anaerobic growth of the arcA and arcB mutant E. coli was also tested and compared to that of the wild type E. coli. No defect was detected (data not shown). Similar results were obtained with LB broth and M9 minimal medium, results obtained with LB broth are shown (Figure 1).
We assayed the resistance of the arcA mutant E. coli to hydrogen peroxide (H 2 O 2 ). Overnight culture of the arcA mutant E. coli was exposed to H 2 O 2 , and its survival was compared to that of the wild type E. coli. The arcA mutant E. coli was more susceptible than the wild type E. coli (Figure 1A). Plasmid pRB3-arcA, which carries a wild type allele of arcA in plasmid pRB3-273C [38,40], complemented the survival defects in H 2 O 2 . This indicates that the susceptible phenotype of the arcA mutant E. coli was likely due to the deletion of the arcA allele ( Figure 1A). Assays performed with log-phase culture of the arcA mutant E. coli yielded similar results (data not shown). Similar results were obtained with LB broth and M9 minimal medium, results obtained with LB broth are shown ( Figure 1).
The same analysis was carried out for ArcB, the cognate sensor-kinase of the ArcAB system. The arcB mutant E. coli survived less than the wild type parental strain ( Figure  1C). We had attempted to clone a wild type allele of arcB into plasmid pRB3-273C to complement the arcB mutant E. coli. However, the cloning efficiency was unusually low as compared to similar cloning attempts we had conducted with the plasmid vector. Of a total of 7 recombinant plasmids we eventually obtained from several transformations, 5 contained mutations at the start codon of arcB and the remaining 2 had mutations that produced truncations early in the ORF (data not shown). This indicates that an over-expression of arcB from a plasmid is probably toxic to E. coli. As an alternative, we constructed a revertant of the arcB mutant E. coli, in which a wild type arcB allele replaced the deleted arcB allele (see Materials and Methods). The revertant mutant of arcB was shown to have the same resistance to H 2 O 2 as the wild type E. coli ( Figure 1C).  Figure 3). However, unlike the original plasmid pRB3-arcA, plasmid pRB3-arcD2A did not render the complemented arcA mutant E. coli more resistant to H 2 O 2 than the wild type E. coli ( Figure 3).  Figure 4. By peptide mass fingerprinting using MALDI-TOF, the prominent protein in Figure 4 was identified as flagellin encoded by fliC, while the other two less abundant proteins were identified as oligopeptide ABC transporter substrate-binding protein (OppA) and glutamate and aspartate transporter subunit (GltI) (data not shown). The levels of these proteins were quantified in the H 2 O 2 -treated and control untreated samples of the wild type and arcA mutant E. coli (Table  2).

Response of flagellin, OppA and GltI to H 2 O 2 is altered in the
Flagellin is the only one among the 10 most abundant proteins that responded to H 2 O 2 treatment. In the wild type, un-treated E. coli flagellin was detected at a lower level than in the arcA mutant E. coli, and H 2 O 2 treatment further decreased the flagellin level (p < 0.05, Student's ttest, Table 2 and Figure 4). In the arcA mutant E. coli H 2 O 2 treatment also decreased flagellin level, however, the decrease was not statistically significant ( Table 2). Therefore, compared to the wild type the E. coli, arcA mutant displayed higher flagellin levels both constitutively and following H 2 O 2 treatment, and its flagellin level did not respond to H 2 O 2 treatment as that in the wild type E. coli.
The response of OppA and GltI expression was different from that of flagellin. In the untreated bacteria levels of both GltI and OppA appeared to be higher in the arcA mutant than in the wild type E. coli (p < 0.05, Student's ttest for GltI,    Two-dimensional gel electrophoresis analysis of whole cell proteins of the wild type and arcA mutant E. coli

Relative expression
WT ΔarcA H 2 O 2 exposure reduced the fliC mRNA level progressively ( Figure 5). The difference in fliC mRNA levels between the wild type and arcA mutant E. coli decreased with longer exposure periods and no difference could be detected by 120 minutes of exposure ( Figure 5). To determine if ArcA directly regulates fliC expression, we expressed and purified recombinant ArcA from aerobic cultures of E. coli and carried out electrophoretic mobility shift assay of the fliC upstream sequence. No specific binding was detected (data not shown).

Deletion of flagellin increased the survival of the ΔarcA mutant E. coli
Flagellin is one of the most abundant proteins in E. coli, and we have shown that its level was higher in the arcA mutant E. coli both constitutively and upon H 2 O 2 exposure ( Figure 4 and To facilitate a direct comparison between the resistance of the wild type and arcA mutant E. coli to H 2 O 2 with or without amino acid supplementation, we carried out a disc diffusion assay, and bacterial resistance to H 2 O 2 was measured by the diameter of the zone of inhibition (ZOI). Without amino acid supplementation the ZOI of the arcA mutant E. coli was significantly larger than that of the wild type E. coli (Figure 7). With amino acid supplementation, sizes of the ZOI reduced for both the wild type and the arcA mutant E. coli, and the difference in the sizes of the ZOI between wild type and arcA mutant E. coli diminished with amino acid supplementation (Figure 7). We tested single amino acids and combinations of various amino acids, and none of the combinations tested was able to complement the susceptibility of the arcA mutant E. coli as the total amino acids (data not shown).

Antibiotic that inhibits protein synthesis increased susceptibility of E. coli to H 2 O 2
To test if protein synthesis is important for bacterial survival and if protein synthesis inhibition is detrimental to bacteria under reactive oxygen stress, we assayed the resistance of E. coli to H 2 O 2 in the presence of chloramphenicol, an antibiotic that inhibits peptide bond formation and hence protein synthesis. Without H 2 O 2 or antibiotic, wild type E. coli grew approximately 2log 10    Starting bacterial concentration and chloramphenicol was more detrimental to E. coli than either H 2 O 2 or chloramphenicol alone, and the bacterial concentration decreased by nearly 4log 10 (Figure 8, left half, cross-hatched bar). This indicates that chloramphenicol enhanced the bactericidal activity of H 2 O 2 . To determine if this enhanced bactericidal activity is due to the bacteriostatic activity of chloramphenicol, we tested the effect of ampicillin, an antibiotic that inhibits the bacterial cell wall synthesis, in the same assay. When added alone, ampicillin had similar effect on bacterial growth as chloramphenicol did (Figure 8

Discussion
Although the ArcAB system has been extensively investigated for its role as the global control system of E. coli in anaerobic growth, its role, if any, in aerobic growth is much less understood. We have previously reported that ArcA is necessary for the pathogenic bacterium Salmonella enterica to resist reactive oxygen and nitrogen species under aerobic conditions [38]. In this report, we used E. coli as our model to further explore the role of both ArcA and ArcB in ROS resistance, and to investigate the mechanism of ROS resistance mediated by the ArcAB two-component system. Here we demonstrate that deletion mutants of ArcA and ArcB were more susceptible to H 2 O 2 , suggesting that both ArcA and ArcB were necessary for E. coli to resist the stress caused by H 2 O 2 (Figure 1), and that their functions were not limited to anaerobic growth of bacteria. Interestingly, we have not detected any growth defects of arcA or arcB mutant E. coli under anaerobic conditions (data not shown) and to our knowledge no such defect has been reported in the literature. In addition, an arcA mutant of Salmonella enterica grew normally in anaerobic medium [38]. This further indicates that ArcAB has wider roles in the physiology and metabolism of enteric bacteria besides its well-characterized regulation of anaerobic growth of bacteria.
The signaling pathway of the ArcAB system under anaerobic conditions has been extensively characterized [25][26][27][28][30][31][32][33][34]42,44]. The membrane-bound sensor-kinase ArcB is activated by reduced quinones under anaerobic conditions, and subsequently activates its cognate transcriptional regulator ArcA by phosphorylating ArcA at Asp54 [30,42,25]. Matsushika and Mizuno previously reported that ArcB can also phosphorylate ArcA directly through His292 under aerobic conditions [45], however, its physiological relevance to E. coli has not been reported. Our results on the role of ArcAB in ROS resistance suggest that ArcAB can be activated by novel signals other than reduced quinones and anaerobic conditions, and the activation is independent of phosphorylation at Asp54 of ArcA as demonstrated under anaerobic conditions [41,42,46], since phosphorylation-defective ArcA expressed from a plasmid fully complemented an arcA mutant E. coli for its susceptibility to H 2 O 2 ( Figure 3). We would like to point out that our analysis was conducted using a phosphorylation-mutant ArcA (Asp54 → Ala) expressed from a plasmid. It is yet to be determined if a mutant carrying a corresponding mutation of arcA in the chromosome is susceptible to H 2 O 2 . (Our attempts to generate a mutant arcA encoding an Asp54 → Ala mutation in the chromosome were unsuccessful due to technical difficulties. Similar to what we observed for arcB, plasmids carrying arcA were prone to mutations during cloning.) We have also noticed that the wild type ArcA expressed from a plasmid confers a stronger H 2 O 2 resistance phenotype than the phosphorylation-defective ArcA. The arcA mutant E. coli complemented in trans with a wild type arcA allele demonstrated higher H 2 O 2 resistance than the wild type E. coli (Figure 1 and 3), while the same mutant E. coli complemented with a phosphorylationdefective arcA allele has the same H 2 O 2 resistance as the wild type E. coli (Figure 3).
In addition to novel signals and signaling pathways that may mediate the function of the ArcAB system in the ROS resistance, the ArcAB system may also regulate a distinct set of genes under aerobic conditions. Under anaerobic conditions ArcA mostly negatively regulates genes involved in the TCA cycle and electron transport [26][27][28]. Under aerobic conditions, a microarray study by Oshima et al. demonstrated that expression of a large number of genes in the arcA or arcB mutant E. coli was altered [23].
Our results suggest that levels of at least three proteins (flagellin, GltI and OppA) were altered in the arcA mutant E. coli both constitutively and in response to H 2 O 2 treatment ( Figure 4 and Table 2). Our further analysis on the messenger RNA level of fliC indicates that the RNA levels are higher in the arcA mutant E. coli and corresponded to the protein levels, suggesting that the regulation is likely on the transcriptional or post-transcriptional level ( Figure 5). Oshima et al. did not detect a significant alteration in the expression of fliC in their microarray analysis, although flagellar synthesis was identified as a system that was affected in the arcA mutant but not the arcB mutant E. coli [23]. The discrepancy is possibly due to the differences in experimental conditions (shaking bacterial cultures at 120 rpm vs. 225 rpm) and detection methods (microarray vs. Real-Time Reverse Transcriptase PCR and 2-D gel electrophoresis). Since we detected an elevation of both mRNA and protein levels of flagellin in the arcA mutant E. coli (Figures 4 and 5), we believe that our observation is valid. The regulation of ArcA on flagellin is likely to be indirect, as we did not detect specific binding of recombinant ArcA protein to the upstream sequence of fliC (data not shown).
Given that the ArcAB system regulates a large number of genes in E. coli, its role in the ROS resistance is likely to be complex. We have demonstrated that mutation of ArcA or ArcB did not alter the H 2 O 2 scavenging ability of E. coli (Figure 2), however, the precise molecular mechanism on how ArcA regulates ROS resistance in E. coli is yet to be elucidated. ArcA was reported to be necessary for the ROS resistance of Haemophilus influenzae due to its regulation of Dps, a ferritin-like small protein that was previously reported to be involved in ROS resistance of Salmonella [39,47]. The mechanism of the ROS resistance mediated by ArcA is likely to be different in E. coli, since dps is expressed close to the wild type level in the arcA or arcB mutant (84% and 99% respectively), and our preliminary microarray analysis with Salmonella arcA mutant indicated that dps responded normally to H 2 O 2 in the arcA mutant (unpublished results). One possible clue on the mechanism of how ArcAB contributes to the ROS resistance of E. coli came from our proteomic analysis that showed altered expression of flagellin, GltI and OppA between the wild type and arcA mutant E. coli ( Table 2). The constitutive GltI and OppA levels are higher in the arcA mutant than in the wild type E. coli, suggesting that the mutant may have a higher need for amino acid transport. In contrast to the GltI and OppA levels in the wild type E. coli that increased 6-and 24-fold respectively in response to H 2 O 2 exposure (possibly due to a higher need for amino acid transport under ROS stress), the level of neither protein in the arcA mutant increased under the same condition (Table 2). A higher level of flagellin in the arcA mutant likely put further constraint on the protein synthesis, and as a result the arcA mutant E. coli might have become less fit under H 2 O 2 stress. Our genetic study demonstrating that deletion of fliC "rescued" the survival defect of the arcA mutant E. coli under H 2 O 2 stress ( Figure  6) supports the hypothesis.
ROS stress conditions induce growth arrest in E. coli. Chang et al. has reported that in growth arrest induced by either glucose-lactose diauxie, entry into stationary phase, or H 2 O 2 treatment, genes involved in amino acid biosynthesis pathways are down-regulated except those of histidine and arginine biosynthesis [24]. Recently, Jang and Imlay have shown that H 2 O 2 damages enzymes with ironsulfur and impairs bacterial metabolism, especially the biosynthesis of leucine [48]. This down regulation of amino acid synthesis may cause a strain on the protein synthesis of bacteria. Our results indicate that protein synthesis is important for E. coli to survive H 2 O 2 treatment. Chloramphenicol, an antibiotic inhibiting protein synthesis, reduced the survival of both the wild type and arcA mutant E. coli after H 2 O 2 treatment, while ampicillin did not ( Figure 8). Consistently, amino acid supplementation enhanced the survival of E. coli after H 2 O 2 treatment (Figure 7). This is in agreement with the report by Calioz and Touati that amino acid supplementation facilitates the survival of superoxide dismutase-deficient E. coli under aerobic conditions [49].
Although our results and results from other investigators suggest that protein synthesis and amino acid availability are important for E. coli to survive ROS stress and the global regulatory system ArcAB plays a role this aspect of ROS stress resistance, protein synthesis and amino acid availability may be only one aspect of the pleiotropic effect of ArcAB system on E. coli, since chloramphenicol-treated arcA mutant was still more susceptible than the similarly treated wild type E. coli. Further studies are necessary to elucidate more molecular mechanisms that control the ROS resistance mediated by the ArcAB global regulatory system.

Conclusion
The global regulatory system ArcAB of E. coli regulate many important functions of bacteria including anaerobic growth, motility, and cell division. Here we demonstrate that ArcAB regulates ROS resistance under aerobic condition, and the signalling pathway of this regulation is distinct from that under anaerobic conditions. The ArcAB system may regulate protein and amino acid synthesis and transport that influence the fitness of E. coli under ROS stress.

Reagents
Growth media for bacteria were purchased from Becton Dickinson and Company (Franklin Lakes, NJ). Anaerobic peptone-yeast medium was obtained from Anaerobe Sys-tems (Morgan Hills, CA). Chemicals and antibiotics were purchased from Sigma-Aldrich Chemical Co. (St. Louis, MO) unless otherwise indicated. Restriction and modifying enzymes for manipulating DNA were purchased from the New England Biolabs (Beverly, MA). Custom oligonucleotides were purchased from Sigma Genosys (The Woodlands, TX).

Bacterial strains and plasmids
E. coli strain K12 isolate MG1655 (gift from Dr. Sydney Kustu, University of California) was used as the parental strain in all analyses described in this report. Mutagenesis was carried out using the one-step mutagenesis method by Datsenko and Wanner [50]. Mutant bacterial strains and sequences of oligonucleotides used for mutagenesis are listed in Table 1. In the arcA mutant, the wild type arcA allele was replaced by a kanamycin-resistance cassette (Kan r ). In the arcB mutant, the wild type arcB allele was replaced by a chloramphenicol-resistance cassette (Cm r ). Each mutation was transduced into fresh E. coli by general transduction with phage P1 before further analysis. In the fliC mutant, the wild type fliC allele was replaced by Cm r , which was subsequently removed to generate a non-polar mutant [50]. The arcA/ fliC mutant was prepared by transducing arcA::kan from the arcA mutant into the fliC non-polar mutant E. coli. A revertant of arcB mutant E. coli was generated through a two-step process. First, a mutant, arcB(Kan r ), was generated in which Kan r was inserted downstream to the arcB coding sequence without affecting the arcB open reading frame. Subsequently, phage P1 was prepared from arcB(Kan r ) and used to transduce the arcB mutant E. coli. Kanamycinresistant and chloramphenicol-sensitive colonies were selected, in which the deletion mutant arcB allele in the arcB mutant E. coli was replaced by a wild type allele from arcB(Kan r ). The genome structure surrounding the arcB allele was determined to verify that wild type arcB allele was restored. The resultant bacterial strain was referred to as arcB-rev.
Plasmid pRB3-arcA used to complement the arcA mutant E. coli was described previously [38]. Plasmid pRB3-arcD2A was constructed using megaprimer method as described [51]. Briefly, a 260-bp section of the arcA gene that included the Asp54 was amplified using mutagenesis primer 5'-CAACCTGGTGATCATGGCGATCAATCTGCC-3' and an arcA primer 5'-CAACGCTACGACGCTCTTC-3'. Sequence in bold in the mutagenesis primer introduced an aspartate to alanine mutation (Asp → Ala) at amino acid 54 in ArcA. The PCR product was used as a megaprimer to amplify plasmid pRB3-arcA together with a vector primer 5'-GTTTTCCCAGTCACGAC-3'. The PCR product was subsequently digested with KpnI and cloned into KpnI-digested plasmid pRB3-arcA to replace the wild type arcA gene with the corresponding sequence that introduced an Asp54 → Ala mutation. The resulting plasmid pRB3-arcD2A contained the same sequence as the original plasmid pRB3-arcA except that GAT which codes for Asp54 of ArcA was mutated to GCG which codes for Ala.

Survival assays of bacteria after exposure to oxidative and other stresses
Survival of E. coli after H 2 O 2 and other stress conditions was assayed as described previously [38,52]. E. coli was cultured in 2 ml of Luria Bertani (LB) broth at 37°C overnight with shaking at 225 rpm. Antibiotics were added as appropriate. Twenty microliters of overnight cultures were added to 2 ml of LB containing one of the following chemicals: hydrogen peroxide, sodium chloride, or sodium dodecyl sulfate (SDS). Cultures in all assays were grown aerobically by shaking at 225 rpm. After exposure to H 2 O 2 or other stresses, aliquots of cultures were diluted and plated in triplicates. Bacterial colonies were enumerated as colony-forming units (CFU) after overnight incubation to determine the bacterial concentration. Disc diffusion assay was carried out as described previously [52]. Briefly, approximately 1 × 10 6 cfu bacteria were plated onto M9 minimal agar plates and paper discs of 1/ 4" diameter loaded with 10 μl of 30% H 2 O 2 were placed in the center of plates onto the bacterial lawn. Plates were incubated overnight at 37°C, and the diameter of the inhibitory zone on each plate was measured.

Scavenging of H 2 O 2 by E. coli
Wild type, the arcA and the arcB mutant E. coli were cultured overnight in LB broth at 37°C with shaking at 225 rpm. Twenty microliters of overnight bacterial culture was diluted in 1 mL of fresh LB broth containing 2 mM of H 2 O 2 that had been pre-warmed to 37°C. An aliquot of 100 μL was taken as the 0 minute sample, and rest of the cultures were incubated at 37°C with shaking. Subsequently, aliquots were taken at 10' intervals. Aliquots of bacterial cultures were used for plating to determine the bacterial concentration, and the rest of the samples were used to determine the concentration of H 2

Real-Time Reverse Transcriptase Polymerase Chain Reaction (RT-PCR) analysis of gene expression
To analyze the expression of fliC messenger RNA, we cultured the wild type and arcA mutant E. coli in LB broth to log phase and divided each culture into two aliquots. One of the aliquots was exposed to 5 mM H 2 O 2 and samples were taken after different exposure periods. The other aliquot was used as an unexposed control. Total RNA was purified from E. coli using TRIzol reagent (Invitrogen, Carlsbad, CA) followed by digestion with DNase I (Qiagen, Valencia, CA) and purification by RNeasy kit (Qiagen, Valencia, CA). Subsequently, 1.5 μg RNA were reverse-transcribed using M-MLV reverse transcriptase (Promega, Madison, WI), and cDNA samples were used for Real-Time Reverse Transcriptase PCR analysis (RT-PCR). RT-PCR was performed using the iQ SYBR Green PCR supermix (Bio-Rad, Hercules, CA) in an iCycler (Bio-Rad, Hercules, CA). Primers 5'-GGCGGAACTAAC-CCAGCTTCA-3' and 5'-TGCTCCAGTCGCCATTGTCA-3' were used for the RT-PCR analysis of fliC expression. The 16S ribosomal RNA level was determined with primers 5'-GGGACCTTCGGGCCTCTTG-3' and 5'-ACCGTGTCT-CAGTTCCAGTGTGG-3', and was used to normalize expression levels of fliC from different samples. Q-Gene program and Relative Expression Software Tool (REST) were used for data analysis of threshold cycle numbers from the iCycler [54,55]. Mean values of normalized expression and standard error measurements were determined as described [54]. Comparisons of mean normalized expression were used to calculate expression ratios. REST was used to obtain statistical significance (pvalue) as described [55].

Bacterial extracts and two-dimensional (2-D) gel electrophoresis
E. coli was cultured in LB broth overnight at 37°C with shaking. Overnight bacterial culture was diluted 1:100 in fresh LB and cultured for 4 hours at 37°C with shaking, and then split into two aliquots. Hydrogen peroxide was added to 5 mM to one of the aliquots, and both aliquots were further incubated for 2 hours at 37°C with shaking. Bacterial cultures were chilled on ice immediately and spun down. Bacterial pellets were then resuspended in 8 M urea and 4% CHAPS in 10 mM Tris 8.0 and sonicated. The insoluble fraction was removed by centrifugation, and soluble lysate was used for 2-D gel electrophoresis. Two-dimensional gel electrophoresis of E. coli proteins was performed with the Zoom IPG Runner system following the manufacturer's instructions (Invitrogen, Carlsbad, CA). One hundred fifty micrograms of cellular proteins were diluted in rehydration buffer (8 M urea, 4% CHAPS and 0.5% pH 3-10 ampholytes) and loaded onto each pH 3-10 ZOOM strip (Invitrogen, Carlsbad, CA). The first dimension electrophoresis was carried out at 200 V for 20', 450 V for 15', 750 V for 15' and 2000 V for 60'. After isoelectric focusing, ZOOM strips were reduced and alkylated with 125 mM iodoacetamide and electrophoresed on NuPAGE Novex 4-12% Bis-Tris ZOOM gels (Invitrogen, Carlsbad, CA) at 100 V for 90'. Proteins were visualized by staining with ProteomIQ reagents (Proteome Systems, Woburn, MA), and then scanned with a HP Scanjet 5530 scanner (Hewlett-Packard, Palo Alto, CA). Individual proteins were quantified using Image-Quant (Amersham Biosciences, Piscataway, NJ) and normalized against the total protein content of the gel.

Mass spectrometry analysis of protein spots
Protein spots of interest were excised from gels and washed with 50% acetonitrile in 50 mM ammonium bicarbonate twice for 15' each. The gel spots were then dehydrated in acetonitrile for 30' and dried in a speed vac for 10'. Thirty microliters of 50 mM ammonium bicarbonate containing 0.3 μg of trypsin (Sigma-Aldrich, St Louis, MO) were added to each sample, and samples were incubated at 37°C for 16 hours. Digested peptides were extracted from gel spots by two washes of 50% acetonitrile/0.1% trifluoroacetic acid, and purified with Ziptips (Millipore, Billerica, MA). Purified peptides were eluted from Ziptips with 50% acetonitrile/0.05% trifluoroacetic acid with 10 mg/ml alpha-cyano-4-hydroxycinnamic acid, and spotted on a sample plate to obtain mass spectra using an Axima CFR Plus MALDI-ToF mass spectrometer (Shimadzu Biotech, Columbia, MD). Each spectrum was calibrated externally using the ProteoMass peptide MALDI-MS calibration kit (Sigma-Aldrich, St Louis, MO).
Peptide fingerprints obtained for each sample were used to search the databases at NCBI and SWISS-PROT using MASCOT search engine http://www.Matrixscience.com. Search parameters used were variable carbamidomethyl and propionamide modifications of cysteines and oxidation of methionines. A peptide tolerance window of 0.5 daltons was used for all searches. Once an identification was made with a statistically significant score, data were accepted when the peptide coverage of the protein was at least 20%, and the molecular weight and isoelectric point of the protein matched those observed on the 2D gel electrophoresis.