Oxidative stress antagonizes fluoroquinolone drug sensitivity via the SoxR-SUF Fe-S cluster homeostatic axis

The level of antibiotic resistance exhibited by bacteria can vary as a function of environmental conditions. Here, we report that phenazine-methosulfate (PMS), a redox-cycling compound (RCC) enhances resistance to fluoroquinolone (FQ) norfloxacin. Genetic analysis showed that E. coli adapts to PMS stress by making Fe-S clusters with the SUF machinery instead of the ISC one. Based upon phenotypic analysis of soxR, acrA, and micF mutants, we showed that PMS antagonizes fluoroquinolone toxicity by SoxR-mediated up-regulation of the AcrAB drug efflux pump. Subsequently, we showed that despite the fact that SoxR could receive its cluster from either ISC or SUF, only SUF is able to sustain efficient SoxR maturation under exposure to prolonged PMS period or high PMS concentrations. This study furthers the idea that Fe-S cluster homeostasis acts as a sensor of environmental conditions, and because its broad influence on cell metabolism, modifies the antibiotic resistance profile of E. coli.


The SoxR regulon confers PMS-mediated protection against quinolones
To investigate the molecular basis of the oxidative stress/quinolone antagonism, we choose phenazine-methosulfate (PMS) as RCC, and norfloxacin as fluoroquinolone. To test whether SoxR intervenes in the PMS/norfloxacin antagonism, wt and ΔsoxR strains were exposed to norfloxacin, alone or in combination with sub-inhibitory concentrations of PMS (1/8 and 1/16 of the MIC PMS of each strain; MIC PMS were 140 μM for wt; 40 μM for ΔsoxR). Adding PMS at a concentration of 1/8 MIC PMS led to a drastic enhancement of norfloxacin resistance level of the wt strain ( Fig 1A). Indeed, when treated with norfloxacin concentration between 100 and 180 ng/mL, the wt strain exposed to PMS (concentration at 1/8 MIC PMS value) reached OD 600 values between 0.4-0.6, while in the absence of PMS, OD 600 values barely reached 0.1 (Fig 1A). Even at 1/16 MIC PMS concentration, PMS remained a potent antagonist of norfloxacin ( Fig  1A). In sharp contrast, PMS exerted no antagonistic effect on norfloxacin in the ΔsoxR mutant ( Fig 1B). SoxR activates the expression of the soxS gene, which in turn activates target genes involved in antibiotic resistance and superoxide resistance [16][17][18][19][20][21][22][23][24][25][26]. Interestingly, we showed that expression of an IPTG-inducible allele of soxS on a plasmid enhanced fluoroquinolone resistance levels of both wt and ΔsoxR mutant (Table 1), mimicking the PMS antagonistic effect, further strengthening the conclusion that the PMS/norfloxacin antagonism involves SoxS-activated genes.
The effect of the PMS/norfloxacin drug combination was analyzed in mutants of two targets of the SoxRS regulatory system, micF and acrA. The micF gene encodes a small RNA that negatively regulates translation of the OmpF porin, hence its induction could potentially prevent OmpF-mediated entry of PMS [27]. However, PMS-mediated drug antagonism was still observed in the ΔmicF mutant ( Fig 1C). The acrA gene encodes a major multidrug efflux pump (AcrAB-TolC) that includes fluoroquinolones (FQ) as a substrate [28,29]. The PMS antagonistic effect on FQ action was completely lost in the ΔacrA mutant ( Fig 1D). By qRT-PCR, we found that expression of the acrA and acrB genes was up-regulated (2-to 3-fold) in cells treated with either PMS only, or with both PMS and norfloxacin ( Table 2). In contrast, treatment with norfloxacin only exerted, if anything, a down-regulation of 1.4-fold of acrA and acrB gene expression (Table 2). Altogether these results support the view that PMS protects E. coli from norfloxacin, by activating AcrAB-mediated efflux of the fluoroquinolone via the SoxRS regulatory system.

The SUF machinery is required to sustain growth in the presence of PMS
SoxR being an Fe-S cluster-containing protein, we went further in analyzing the importance of ISC and SUF Fe-S clusters biogenesis systems in PMS stress and in the PMS/norfloxacin antagonism. Previous work has reported that the iscRSUA and sufABCDSE operons are induced by PMS [30,31], which we confirmed here by using PiscR::lacZ and PsufA::lacZ gene fusions (2.7 and 3.3-fold increase in the presence of 20 μM PMS, respectively) (S1 Fig). Then, the effect of PMS on the growth kinetics of LB grown cultures was tested. PMS (30 μM) was added during the exponential growth phase and OD 600 values were recorded (Fig 2). Two hours after addition of PMS, we observed that the ΔsufABCDSE and the ΔsoxR mutants stopped growing whereas growth of the wt strain did not (Fig 2A and 2C). In the presence of PMS, growth of the ΔiscUA mutant appeared to be slower than the wt strain but this is a general feature of the ΔiscUA mutant even under non-stressed conditions ( Fig 2B) (in untreated condition, the doubling times of the ΔiscUA mutant and the wt strain were 35 and 28 min, respectively), and more importantly, like the wt, it did not cease growing (Fig 2B). RCC such as PMS are predicted to enhance intracellular level of superoxide [32]. An enhanced level of intracellular superoxide stress can be obtained by using strains lacking superoxide dismutase SodA or/and SodB. As expected, the ΔsodA, ΔsodA ΔsodB mutants and to some extent the single ΔsodB mutant showed hypersensitivity to PMS (S2 Fig). Introduction of the ΔsufABCDSE mutation in ΔsodA, ΔsodB, or ΔsodA ΔsodB mutants increased drastically the sensitivity to PMS (S2 Fig). In contrast, combining ΔiscUA mutation with ΔsodA, ΔsodB or both ΔsodA ΔsodB mutations did not enhance PMS sensitivity (S2 Fig). Altogether, these results established the importance of the SUF system in allowing E. coli to resist PMS stress and by inference, revealed that ISC was of no help in these conditions.

Impact of the SUF and ISC machineries on the PMS-mediated protection from norfloxacin
Contribution of ISC and SUF in PMS/norfloxacin antagonism was then investigated. First, we showed that the PMS-induced resistance to norfloxacin occurred in the wt strain and the ΔiscUA mutant, whereas it was not observed in the ΔsufABCDSE mutant (Fig 3). Sub-inhibitory concentration of PMS (1/16 MIC PMS concentration) permitted both the wt and ΔiscUA strains to grow in the presence of increasing concentrations of norfloxacin (120 ng/mL, 140 ng/mL, 160 ng/mL) (Fig 4 and Table 3). In contrast, sub-inhibitory concentration of PMS did not protect the ΔsufABCDSE mutant above 120 ng/mL norfloxacin (Fig 4). All growth parameters (lag phase, growth rate, final OD 600 ) were more severely affected in the ΔsufABCDSE mutant than in the wt and ΔiscUA strains (Fig 4 and Table 3), illustrating that the protection against norfloxacin afforded by PMS depends upon a functional SUF system.

SUF, but not ISC, is required for SoxR maturation under PMS stress
Results above showed that both SUF and SoxR were required for the PMS/norfloxacin antagonism. A hypothesis was that PMS generates conditions during which SUF is required for synthesizing and carrying Fe-S clusters to SoxR, hence permitting its transcriptional activator function to fire anti-fluoroquinolone defense genes, including the AcrA pump. To test which system, ISC or SUF, was used to mature SoxR, we used a chromosomal PsoxS::lacZ fusion at the lac locus, and β-galactosidase level as a proxy for monitoring SoxR maturation efficiency [33]. In the wt strain, induction of the PsoxS::lacZ fusion reached a plateau at 180 min after adding PMS (30 μM) (Fig 5A). The β-galactosidase level in the treated strain was 22-fold higher than in the untreated cells ( Fig 5A). In the ΔsoxR strain, no induction of PsoxS::lacZ expression was observed ( Fig 5A). In the ΔsufABCDSE mutant, expression of the PsoxS::lacZ plateaued ca 30 min after PMS addition ( Fig 5B). The maximal level of β-galactosidase reached Cultures were incubated 18 hours at 37˚C with shaking. Plates were read for OD 600 in Tecan Infinite. The experiment was repeated at least three times. The means and standard deviations are shown. Asterisks represent the statistical significance calculated using the Bonferroni method ( �� p<0.01; � p<0.05; ns p>0.05). in the ΔsufABCDSE mutant was 2.5-fold lower than in the wt strain. In the ΔiscUA mutant, at all time points, the induction pattern of PsoxS::lacZ was identical to the one observed in the wt strain ( Fig 5C). All these results suggest that during the 30 min period after PMS exposure, the existing pool of [2Fe-2S]-bound SoxR was sufficient to activate soxS expression, in either ΔsufABCDSE or ΔiscUA mutants. In contrast, after 30 min SoxR-mediated activation occurred only in SUF containing strains. It is interesting to note that in the ΔsufABCDSE mutant, SoxR activity ceased shortly before growth stopped. It is tempting to speculate that the former is the cause of the latter (Fig 2A and Fig 5B). We then analyzed the correlation between PMS concentration and induction of the SoxRS regulon in the different genetic backgrounds. Expression of the PsoxS::lacZ fusion was measured 2 hours after incubation with different concentrations of PMS (3.4, 6.8, 10, 20 and 30 μM). At low PMS concentrations, i.e. 3.4 μM and 6.8 μM, PsoxS::lacZ expression was induced to the same extent in all wt, ΔiscUA and ΔsufABCDSE strains ( Fig 5D). In contrast, at PMS concentrations of 10 μM and above, PsoxS::lacZ expression was much stronger in the wt and ΔiscUA strains than in the ΔsufABCDSE mutant ( Fig 5D). In fact, in the ΔsufABCDSE mutant, expression of PsoxS::lacZ plateaued when using PMS concentrations of 10 μM and above ( Fig 5D). Interestingly, defect in PsoxS::lacZ fusion expression within the ΔsufABCDSE strain was suppressed by a high soxR gene dosage, consistent with the notion that SoxR is a poor substrate for ISC system and that increased SoxR level would compensate inefficient maturation by ISC (Fig 6).
Last, we wished to investigate the contribution of ISC and SUF pathways to SoxR maturation in the absence of PMS. To this purpose we assessed the expression of the PsoxS::lacZ fusion in two mutant strains, each lacking a component of the reduction system of SoxR, RsxC and RseC. In these strains, the Fe-S cluster of SoxR is thought to remain mostly oxidized and SoxR active even in the absence of redox-active molecules [34]. In the absence of PMS, activation of PsoxS::lacZ transcription was increased (3-fold) in the ΔrsxC and ΔrseC mutants when compared to the wt strain (Fig 7). In the ΔrsxC ΔiscUA, ΔrsxC ΔsufABCDSE, ΔrseC ΔiscUA and ΔrseC ΔsufABCDSE double mutants, expression of the PsoxS::lacZ fusion was increased to the same extent (3-to 4-fold) as compared to the wt strain (Fig 7). Together, these results indicate that under non-stressful conditions both Fe-S biogenesis systems, ISC and SUF, are able to sustain SoxR maturation.
Altogether, this series of results show that under non-stress inducing conditions, SoxR can acquire its cluster from either ISC or SUF system. However, under stress inducing conditions, SUF is the system that targets Fe-S cluster to SoxR, which subsequently get oxidized and permits SoxR to activate expression of its targets.

The role of SoxR maturation in PMS/fluoroquinolone antagonism
We asked whether different levels of induction of the SoxRS regulon would translate in different level of norfloxacin resistance. Results showed that a concentration of 3.4 μM PMS was not

SufA is required for SoxR maturation during PMS stress
Once formed within the scaffold component of the Fe-S clusters biogenesis pathway, clusters are targeted to apo-protein clients via a series of carriers, which might show some level of redundancy depending upon the growth conditions [35,36]. Therefore, contribution of carriers in SoxR maturation was investigated. We showed above that expression of the PsoxS::lacZ fusion was identical in the ΔiscUA mutant and in the wt strain, ruling out a role for IscA in SoxR maturation under PMS stress ( Fig 5C). In the ΔsufA mutant, expression of the PsoxS:: lacZ fusion was identical to the wt strain after the first hour of incubation with PMS (Fig 8). In contrast, after longer exposure to PMS (2 and 3 hours), expression of the PsoxS::lacZ fusion was drastically reduced when compared to the wt strain. Actually, the profile of expression of the PsoxS::lacZ fusion was identical in the ΔsufA and in the ΔsufABCDSE mutants (Figs 5B and 8). These results indicate that SufA is required for SoxR maturation upon PMS stress.

The stress responding carriers, NfuA and ErpA, are dispensable for SoxR maturation during PMS stress
NfuA and ErpA are additional Fe-S carriers that can cooperate under oxidative stress conditions [37][38][39]. Moreover, they are able to receive Fe-S clusters made within either ISC or SUF systems [37][38][39]. Given that SoxR is a stress responding regulator, it was of interest to evaluate   Table 3. The erpA gene is an essential gene in E. coli [38]. Therefore a conditional allele, in which the endogenous erpA gene is under the control of the ParaBAD promoter (ParaBAD::erpA), was transduced in the PsoxS::lacZ fusion-containing strain. Strikingly, whether cells were grown in glucose or in arabinose, PMS induction of the PsoxS::lacZ fusion was observed in the wt and ParaBAD::erpA strains, indicating that ErpA was dispensable for SoxR maturation (S5 Fig panels A and B). To confirm the dispensability of ErpA for SoxR maturation, we used the CRISPR interference method to control erpA expression [40]. For this purpose, we used a plasmid producing a catalytically inactive version of Cas9 (pdCas9) and a plasmid encoding a single guide RNA (pRBS-erpA) targeting the non-template DNA strand of the UTR of erpA containing the ribosome-binding site. In the presence of the inducer anhydrotetracycline (aTc), cells carrying the pRBS-erpA exhibited a growth defect (S5 Fig panel C). In the presence of both aTc and PMS, PsoxS::lacZ expression was the same in wt cells carrying the pRBS-erpA or the control vector, psgRNA (S5 Fig panel E). To verify that the pRBS-erpA plasmid was indeed preventing erpA expression, we used the strain PM2040 that contained an PerpA::lacZ gene fusion. We showed that in the presence of aTc, expression of PerpA::lacZ was almost null (2 Miller units) in cells carrying the pRBS-erpA, while β-galactosidase activity of 36 Miller units was measured in cells carrying control vector psgRNA (S5 Fig panel D). Altogether, results obtained by two different methods to deplete ErpA led us to conclude that ErpA is not required for SoxR maturation under PMS stress.

Discussion
Understanding the influence of environmental conditions on level of antibiotic resistance is a prerequisite to monitor and control bacterial antibiotic resistance. Previously, we showed that iron limitation enhanced level of resistance of E. coli to aminoglycosides, and that Fe-S cluster biogenesis regulation played a key role in this unexpected link [8,41]. Here we show that Fe-S homeostasis connects ROS producing compound, RCC, and resistance level to fluoroquinolones.

PLOS GENETICS
Fe-S cluster homeostasis and antibiotic resistance E. coli synthesizes ca. 150 Fe-S proteins, the maturation of which depends upon ISC or SUF machineries. E. coli synthesizes five Fe-S bound transcriptional regulators, namely FNR, NsrR, IscR, YeiL, and SoxR [16,42]. Study of maturation of NsrR and IscR, which sense NO and Fe-S cluster demand, respectively, showed that these two related targets (i.e. ca. 40% sequence identity) are matured by ISC under normal conditions and by SUF under stress conditions [35]. In contrast, work by Kiley's lab showed that ISC, but not SUF, was responsible for the maturation of FNR, a Fe-S transcriptional regulator sensing anaerobic/aerobic switch [43]. Here, we found SoxR to be matured mostly by SUF, a situation somehow complementary to that observed with FNR. SoxR being matured by SUF is consistent with the fact that SUF is synthesized and functional in vivo under oxidative stress [30,31,[44][45][46].
The question then arises of what prevents ISC to act on SoxR. Previous in vitro transcription analyses showed that expression of both isc and suf operons is induced by PMS, which we confirmed here using lacZ fusions (S1 Fig) [30]. This rules out the possibility that defect in SoxR maturation was due to differential expression of the isc and suf operons. Increased copy number of the soxR gene suppressed the defect in soxS activation of the ΔsufABCDSE mutant, consistent with the idea that SoxR can be a low affinity substrate for ISC under PMS stress. Note however that SoxR protein levels may be lowered in a ΔsufABCDSE mutant as SoxR is positively autoregulated. Yet, use of the ΔrsxC and ΔrseC mutants allowed us to show that in the absence of PMS, ISC is able to maturate SoxR. Hence, we propose that under PMS stress, SoxR is a poor substrate for ISC because ISC system itself is intrinsically susceptible to oxidative stress, possibly as a result of PMS-mediated damages to some of the Isc proteins. That SoxR can be maturated by both machineries is consistent with the fact that SoxR orthologs can be found in bacterial species that have only SUF such as Streptomyces coelicor, or in bacterial species that have only ISC, such as Pseudomonas aeruginosa [24,47,48].
During the delivery step, Fe-S clusters are transferred from the ISC or SUF systems to the apo-targets. Multiple studies have concluded that ErpA carrier is the ultimate carrier used by most, if not all, cellular proteins, including IspG/H, formate dehydrogenase N, nitrate reductase, succinate dehydrogenase, complex I, or hydrogenases 1 and 2 [37,38,[49][50][51]. Surprisingly, in the present study, maturation of SoxR was found totally independent of ErpA. SoxR maturation was also independent of the ErpA-NfuA delivery pathway we recently identified as being important for combating oxidative stress [37]. This series of observation was totally unexpected and opens new perception of Fe-S biogenesis under stress conditions. An explanation for the ErpA/NfuA independent maturation of SoxR might be that it binds a 2Fe-2S cluster, whereas all other proteins tested so far bind 4Fe-4S clusters (with the noticeable exception of IscR). That delivery factors such as A-type carriers (IscA, SufA, ErpA) be required for 4Fe-4S containing targets was already suggested by our study comparing IscR and NsrR maturation maps, and was also proposed on the basis of the in vivo maturation of the 2Fe-2S containing ferredoxin Fdx and from in vivo studies in yeast [35,52,53]. Altogether, our study is fully consistent with the pioneer work of Schwartz and collaborators [54], which demonstrated that IscS was not needed for SoxR maturation under paraquat stress. However, the results presented here go far beyond this original finding in describing both the basic machinery and the accessory proteins that transfer the cluster from the scaffold to SoxR. We previously reported that E. coli making clusters with SUF showed an enhanced resistance to aminoglycoside and here we report that it also shows an enhanced resistance to fluoroquinolone [8]. Beyond this apparent similarity, opposite molecular causes are found. Because the SUF system is less efficient than the ISC for maturating pmf-producing Fe-S cluster containing Nuo and Sdh respiratory complexes, uptake of aminoglycoside was reduced and bacteria exhibited aminoglycoside phenotypic resistance. On the contrary, under RCC stress, https://doi.org/10.1371/journal.pgen.1009198.g007 because the SUF system is more efficient than the ISC one for maturating SoxR, cells can expel fluoroquinolone and show enhanced fluoroquinolone resistance. An enigma then remains: AcrAB-mediated fluoroquinolone efflux being itself energized by pmf, fluoroquinolone (and possibly PMS) efflux by AcrAB should have been reduced in PMS-treated cells that should have been scored as hypersensitive instead of hyperresistant. Measuring the actual drop in pmf induced by DIP and by PMS, and assessing how much pmf is required for AG uptake and fluoroquinolone efflux will probably help to solve this apparent conundrum.
Drug/drug interaction might mimic situations in natural settings wherein bacteria face multiple antibacterial chemicals. In the case of DIP/aminoglycoside antagonism, one could speculate that iron limitation can be interpreted as a signal to adapt to a hostile environment, by reducing all pmf-dependent exchanges. RCC and fluoroquinolone share structural similarity as they are both heterocyclic compounds and could be present within the same ecological niche. In this regard, Pseudomonas aeruginosa provides a good illustration of evolved adaptation to such compounds: the MexGHI-OpmD pump excretes both fluoroquinolone and 5-methylphenazine-1-carboxylate, an intermediate of pyocyanin biosynthesis, which is structurally similar to PMS and activates the mexGHI-opmD operon that belongs to the SoxR regulon [55][56][57][58]. As a matter of fact, pyocyanin was also found to antagonize activity of many types of antibiotics including fluoroquinolone [5,59]. Thus, a possibility is that in E. coli, the fluoroquinolone exporting AcrAB pump will export PMS as well. The very low MIC PMS value of acrA mutant supports this view. Very recently, the RCC/fluoroquinolone antagonism was shown to be conserved in P. aeruginosa, and importantly it occurred in P. aeruginosa biofilms that are an important cause for persistent and antibiotics-resistant infections [5,60]. Altogether these results illustrate the medical importance of the RCC/fluoroquinolone antagonism. As a last note, a recent system-based analysis predicted ROS as potential adjuvants potentiating antibacterial activity [61]. On the contrary, the present and previous studies identified antagonism between ROS producers RCC and several quinolones (norfloxacin, ciprofloxacin,

PLOS GENETICS
levofloxacin, novobiocin and moxifloxacin). Hence, one should be cautious when using ROS producing chemical as antibiotic adjuvant.

Bacterial strains and growth conditions
The transcriptional PsoxS::lacZ fusion was constructed as described in Ezraty et al. [33]. The PsoxS promoter region fused to lacZ encompassed a region from the 111 nucleotides upstream the transcriptional soxS start site to the 21 first nucleotides of the soxS-coding region. The E. coli K-12 strain MG1655 and its derivatives used in this study are listed in S1 Table. Deletion mutations were introduced by P1 transduction. Transductants were verified by PCR, using primer pairs hybridizing upstream and downstream of the deleted gene. E. coli strains were grown at 37˚C in Luria-Bertani (LB) rich medium. Isopropyl β-D-1-thiogalactopyranoside (IPTG) (0.1 mM), arabinose (0.2%), glucose (0.2%) and anhydrotetracycline (aTc) (2 μM) were added when required. Solid media contained 1.5% agar. Antibiotics were used at the following concentrations: chloramphenicol 25 μg/mL, kanamycin 30 μg/mL, and ampicillin 50 μg/mL.

MIC determination
To determine MIC PMS , PMS was dissolved in LB medium and diluted in LB to reach concentration ranging from 0 to 300 μM in 20 μM increments. One hundred microliters (100 μL) of each concentration of PMS tested were added in a 96-well microplate. Each well was inoculated with 100 μL of a fresh LB bacterial inoculum of 2 × 10 5 c.f.u./mL, obtained from a dilution of a mid-log phase E. coli growing culture. To determine MIC Norfloxacin of strains carrying the pSoxS and pTrc99A plasmids, norfloxacin was dissolved in LB medium and diluted in LB to reach concentration ranging from 800 ng/mL to 1800 ng/mL in 100 ng/mL increments for the strains carrying the pSoxS, and 160 ng/mL to 360 ng/mL in 20 ng/mL increments for the strains carrying the pTrc99A. One hundred microliters (100 μL) of each concentration of norfloxacin tested were added in a 96-well microplate. Each well was inoculated with 100 μL of a fresh LB (supplemented with ampicillin and IPTG) bacterial inoculum of 2 × 10 5 c.f.u./mL, obtained from a dilution of a mid-log phase E. coli growing culture. Ampicillin and IPTG were added to the LB medium used to dilute the E. coli culture. Microplates were incubated at 37˚C for 18 h under aerobic conditions and agitation 170 rpm. The microplates were then read at OD 600nm and MIC was defined as the lowest drug concentration that exhibited complete inhibition of E. coli growth. The experiment was repeated at least three times.

PMS-mediated protection against norfloxacin
Wells of a 96-well microplate containing 100 μL of LB supplemented or not with norfloxacin and PMS, were inoculated with 100 μL of a fresh LB bacterial inoculum of 2 × 10 5 c.f.u./mL. The range of norfloxacin final concentrations used was adapted to each strain depending on their sensitivity, and is given in the corresponding figures. PMS was used at the indicated final concentration. Microplates were incubated at 37˚C for 18 h under aerobic conditions and agitation (170 rpm). The microplates were then read at OD 600nm .

Plasmid construction
Plasmids pSoxR and pSoxS were constructed by PCR amplification of the coding region of soxR and soxS from E. coli MG1655 chromosomal DNA using the following primer pair: NcoI-soxR/BamHI-soxR and NcoI-soxS/BamHI-soxS, respectively (S2 Table). The PCR product was then digested by NcoI and BamHI and cloned into the NcoI/BamHI linearized pTrc99A vector. The plasmids were verified by sequencing. The pRBS-erpA plasmid was cloned as described by Larson et al. [40]. Briefly, we designed primers hybridizing at erpA RBS region (Ec-F and Ec-R), which were used to PCR-amplify the whole plasmid (psgRNA). The plasmid was verified by sequencing using primers Ec-F colony and Ec-R colony. The sgRNA obtained was complementary to the erpA non-template strand.

RNA preparation and reverse transcription
The E. coli wt strain (BE1000) was grown in LB. When the culture reached mid-exponential phase, the culture was divided in four aliquots, three were treated for 30 min with norfloxacin only (30 ng/mL), with PMS only (30 μM), or with both norfloxacin and PMS, while the last sample remained untreated. Three biological independent experiments were performed. RNAs were prepared from E. coli strain cultures (10 mL) grown in appropriate conditions. The cells were harvested and frozen at -80˚C. Total RNAs were isolated from the pellet using the Maxwell 16 LEV miRNA Tissue Kit (Promega) according to the manufacturer's instructions and an extra TURBO DNase (Invitrogen) digestion step to eliminate the contaminating DNA. The RNA quality was assessed by tape station system (Agilent). RNA was quantified spectrophotometrically at 260 nm. For cDNA synthesis, 1 μg total RNA and 0.5 μg random primers were used with the GoScript Reverse transcriptase according to the manufacturer instruction (Promega).

Quantitative real-time-PCR for transcriptional analyses
Quantitative real-time PCR (qPCR) analyses were performed on a CFX96 Real-Time System (Bio-Rad). The reaction volume was 15 μL and the final concentration of each primer was 0.5 μM. The cycling parameters of the qRT-PCR were 98˚C for 2 min, followed by 45 cycles of 98˚C for 5 s, 55˚C for 10 s, 72˚C for 1 s. A final melting curve from 65˚C to 95˚C is added to determine the specificity of the amplification. To determine the amplification kinetics of each product, the fluorescence derived from the incorporation of EvaGreen into the doublestranded PCR products was measured at the end of each cycle using the SsoFast EvaGreen Supermix 2X Kit (Bio-Rad). The results were analyzed using Bio-Rad CFX Maestro software, version 1.1 (Bio-Rad). The RNA16S gene was used as a reference for normalization. For each point a technical duplicate was performed. The amplification efficiencies for each primer pair were comprised between 80 and 100%. The primers used for qRT-PCR are reported in the S2 Table. β-Galactosidase assay β-Galactosidase assays were carried out as described previously by J.H. Miller [62].

Test for E. coli sensitivity to redox-cyclic compounds
The E. coli strains were grown overnight in LB and inoculated (1/100) in fresh LB medium. The cultures were grown to an OD 600 of 0.2, and were each split into two flasks, one with PMS added (time zero), while the other was left untreated. Cultures were further incubated at 37˚C and growth was monitored by following OD 600 . To test the E. coli sensitivity to redox-cyclic compounds on plates, overnight cultures were diluted in sterile PBS and 5 μL were directly spotted onto LB plates containing PMS. The plates were incubated overnight at 37˚C before growth was scored.

ErpA depletion
The LL401 strain carrying the chromosomal copy of erpA gene under the ParaBAD promoter [38] was grown overnight in LB. Fresh LB medium supplemented with glucose (0.2%) was then inoculated at 1/100. The strain carrying the plasmids allowing controlling erpA expression by CRISPRi, pdCas9 and pRBS-erpA, was grown overnight in LB and then inoculated (1/ 100) in fresh LB supplemented with anhydrotetracycline (aTc) (2 μM).
Supporting information S1   A-B) The E. coli strains carrying the chromosomal PsoxS::lacZ fusion, wt (BE1000) (black circles) and ParaBAD::erpA (AG048) (black crosses) (A-B), were grown overnight in LB and then inoculated (1/100) in fresh LB medium with glucose (0.2%) (A) or arabinose (0.2%) (B). The cultures were grown to an OD 600 of 0.2, and were each split into two flasks, PMS (30 μM) was added (time zero) in one (solid line), and the other was left untreated (dotted line). All cultures were further incubated at 37˚C with shaking. β-galactosidase activity was monitored and expressed as Miller units. (C) Growth of the E. coli wt strain (BE1000) containing the chromosomal PsoxS::lacZ fusion and carrying the plasmids pdCas9 together with the plasmid allowing erpA extinction, pRBS-erpA (white triangles), or the empty control vector, psgRNA (black circles). Cells were grown overnight in LB without anhydrotetracycline (aTc) and then inoculated (1/100) in fresh LB medium supplemented with aTc (2 μM). Growth was recorded by measuring OD 600 following time. (D) The E. coli strain possessing the chromosomal PerpA::lacZ fusion (PM2040) and carrying the plasmid pdCAS9 together with the plasmid allowing erpA extinction, pRBS-erpA or the control vector, psgRNA. Cells were grown overnight in LB without anhydrotetracycline (aTc) and then inoculated (1/100) in fresh LB medium supplemented with aTc (2 μM). When cultures reached an OD 600 of 0.2, aliquots were taken to assay β-galactosidase activity that is expressed in Miller units. (E-F) The E. coli wt strain possessing the chromosomal PsoxS::lacZ fusion (BE1000) and carrying the plasmid pdCas9 together with either the plasmid allowing erpA extinction, pRBS-erpA (white triangles), or the empty control vector, psgRNA (black circles) were grown overnight in LB and then inoculated (1/100) in fresh LB medium supplemented (E) or not (F) with aTc (2 μM). The cultures were grown to an OD 600 of 0.2, and were each split into two flasks, PMS (30 μM) was added (time zero) in one (solid line), and the other was left untreated (dotted line). All cultures were further incubated at 37˚C with shaking and β-galactosidase activity was monitored and expressed as Miller units. All the experiments were repeated at least three times. The means and standard deviations are shown (A, B, D, E and F) and a representative experiment is presented in panel C. (DOCX)