Transposon-Directed Insertion-Site Sequencing Reveals Glycolysis Gene gpmA as Part of the H2O2 Defense Mechanisms in Escherichia coli

Hydrogen peroxide (H2O2) is a common effector of defense mechanisms against pathogenic infections. However, bacterial factors involved in H2O2 tolerance remain unclear. Here we used transposon-directed insertion-site sequencing (TraDIS), a technique allowing the screening of the whole genome, to identify genes implicated in H2O2 tolerance in Escherichia coli. Our TraDIS analysis identified 10 mutants with fitness defect upon H2O2 exposure, among which previously H2O2-associated genes (oxyR, dps, dksA, rpoS, hfq and polA) and other genes with no known association with H2O2 tolerance in E. coli (corA, rbsR, nhaA and gpmA). This is the first description of the impact of gpmA, a gene involved in glycolysis, on the susceptibility of E. coli to H2O2. Indeed, confirmatory experiments showed that the deletion of gpmA led to a specific hypersensitivity to H2O2 comparable to the deletion of the major H2O2 scavenger gene katG. This hypersensitivity was not due to an alteration of catalase function and was independent of the carbon source or the presence of oxygen. Transcription of gpmA was upregulated under H2O2 exposure, highlighting its role under oxidative stress. In summary, our TraDIS approach identified gpmA as a member of the oxidative stress defense mechanism in E. coli.


Introduction
Escherichia coli is a Gram-negative facultative anaerobic bacterium. It is a frequent member of the normal human microbiota but can also be a pathogen causing food poisoning, urinary tract infection and even septic shock [1]. The burden of diarrheal infections by pathogenic strains of E. coli is immense; in 79 low-income countries alone, more than 200 million episodes of childhood diarrhea due to E. coli and Shigella occur each year [2]. In high-income countries, E. coli is the primary cause of blood stream infections, accounting for 27% of the documented bacteremia episodes [3]. The emergence of antibiotic resistance in Gram-negative bacteria is also concerning and a recent study of 203 countries identified E. coli as the leading pathogen for deaths associated with antimicrobial resistance in 2019 [4].
Reactive oxygen species (ROS), and more specifically hydrogen peroxide (H 2 O 2 ), have a strong impact on bacterial pathogenesis. Millimolar H 2 O 2 can be produced by certain strains of Lactobacilli of the human normal microbiota [5]. H 2 O 2 production prevents the colonization by pathogens of the urinary tract [6]. Similarly, H 2 O 2 is produced by phagocytes during the oxidative burst, a necessary step for the killing of pathogens [7]. The effect of H 2 O 2 on bacteria has been partially studied, but a complete picture of how H 2 O 2 affects bacteria and the bacterial response has not been elucidated for any bacterial species. Previous studies on H 2 O 2 tolerance, using DNA microarrays and RNA-seq, identified genes regulated under H 2 O 2 exposure [8][9][10] in E. coli. These studies permitted a better understanding of the regulation of numerous genes and pathways affected by H 2 O 2 exposure. In particular, OxyR, a specific H 2 O 2 -responsive transcription factor, and SoxR which senses oxidative stress and nitric oxide, have been identified as playing an important role in resistance to H 2 O 2 [9,11]. OxyR senses hydrogen peroxide through the oxidation of its cysteine residues, which orchestrate a conformational change allowing it to regulate the expression of 38 genes [9,12]. The iron-sulfur cluster of SoxR is oxidized by redox cycling compounds or superoxide, leading to the activation of the transcription factor which regulates the expression of 11 genes, which includes SoxS, another transcription factor that further regulates 34 genes [9,13]. However, transcriptional analyses do not identify genes required for survival in oxidative conditions. Diverse mutagenesis techniques were used to identify the genes involved in E. coli survival under H 2 O 2 exposure, but only a limited number of genes were identified each time [14][15][16].
The combination of transposon mutagenesis and high-throughput sequencing is a powerful technique that allows interrogation of the whole genome and represents a new standard for global functional genomic studies [17]. Here, we performed transposondirected insertion-site sequencing (TraDIS) [18] to identify genes implicated in tolerance to exogenous H 2 O 2 exposure. A similar approach was used on Salmonella Typhimurium to identify genes implicated in H 2 O 2 tolerance, deepening the understanding of how the bacteria survive oxidative burst [19,20]. The results of our study highlighted the role of gpmA, which encodes a phosphoglycerate mutase, an enzyme of the glycolysis, under H 2 O 2 exposure. This is the first study identifying gpmA as a factor of H 2 O 2 tolerance in E. coli.

Bacterial Strains, Media and Growth Conditions
All bacterial strains and plasmid used in this study are documented in Table 1. E. coli strains were cultured at 37 • C in Luria-Bertani (LB) (Becton & Dickinson, Basel, Switzerland) broth or on Luria-Bertani Agar (Becton & Dickinson). H 2 O 2 35% w/w (Acros Organics, VWR Life Science, Nyon, Switzerland) was added at the indicated final concentration. LB was supplemented with 0.4% glucose, 0.4% glycerol, 0.4% sodium acetate, 0.4% sodium citrate, 50 mM sodium nitrate (Sigma-Aldrich, St. Louis, MI, USA, Merck and Cie, Schaffhausen, Switzerland) where indicated. Minimal medium M9 plates, constituted by M9 salts (VWR Life Science), 0.1 mM CaCl 2 (Sigma-Aldrich), 0.2 mM MgSO 4 (Sigma-Aldrich), 1.5% (w/v) agar (Carl Roth, Arlesheim, Switzerland), were used when indicated. Antibiotics were used when indicated at the following concentrations: ampicillin 100 µg/mL (10044, Sigma-Aldrich), kanamycin 50 µg/mL (PanReac AppliChem, VWR, Switzerland). For anaerobic assays, bacteria were grown in deoxygenated LB with the corresponding antibiotic and every step was performed under anaerobic condition (Coy Laboratory Products, Labgene scientific SA, Châtel-Saint-Denis, Switzerland). Overnight cultures and agar plates were grown overnight at 40 • C in anaerobic chamber. The TraDIS screen was performed using a library of transposon mutants previously generated in E. coli strain BW25113 [18]. The E. coli strain MG1655 is referred in this paper as the wild-type (WT). The strain BEFB02 with oxyR deletion was a kind gift from B. Ezraty. Other gene deletions were obtained from the Keio collection [21] and transduced in a MG1655 background by P1 transduction.

TraDIS
The TraDIS library was thawed and diluted in 50 mL of LB broth to reach an OD 595 of 0.02 (approximatively 8 × 10 8 CFU). H 2 O 2 was added to H 2 O 2 -treated samples to reach a concentration of 2.5 mM whereas pure medium was added in untreated controls. The experiment was performed in duplicates. Bacteria were grown at 37 • C in aerating conditions (250 mL flask, shaking 250 rpm) until an OD 595 = 1.
Bacteria were collected and the DNA was extracted using a DNeasy Blood & Tissue Kit (Qiagen) according to the manufacturer's instructions. Samples were prepared for sequencing as described previously [18]. Briefly, genomic DNA was fragmented by ultrasonication, fragments were end-repaired using the NEBNext Ultra I kit (New England Biolabs, Notting Hill, Australia) and transposon fragments enriched by PCR using primers specific for the transposon and adapter. Samples were quantified by qPCR using the NEBNext Library Quant Kit for Illumina kit (New England Biolabs) according to the manufacturer's instructions and sequenced using an Illumina MiSeq with 150-cycle v3 cartridges.
The TraDIS data were analyzed using Bio::TraDIS pipeline [24] with the following parameters: 50 reads per gene as minimal threshold and 5% trim at each side of gene to avoid the consideration of meaningless transposon insertions that can occur within gene extremities. Sequencing

P1 Transduction
Strains from the Keio library were grown with 50 mg/mL kanamycin. The deletions of genes of interest from the corresponding Keio library mutant were transduced in E. coli MG1655 using phage P1 as previously described [25]. P1 phage was a kind donation from G. Panis (University of Geneva). The deleted mutants were verified using PCR with appropriate gene-specific primers (Supplementary Table S1).

H 2 O 2 Susceptibiliy Assessed by Disk Diffusion Assay
To assess the susceptibility to H 2 O 2 and other oxidants, we used disk diffusion assay as previously described [10]. Briefly, an overnight culture of bacteria was diluted in LB to McFarland 0.5 using a Densimat (bioMérieux, Marcy-l'Étoile, France) and LB agar plates were inoculated using a sterile cotton swab. Sterile cellulose disks (5 mm diameter) were placed on the plate and 10 µL of 1 M H 2 O 2 diluted in sterile water was added to the center of the disk. Other oxidants were used at the following concentrations: methylhydroquinone (Sigma-Aldrich) MHQ 0.5 M in water; methyl viologen dichloride hydrate, also called paraquat, (Sigma-Aldrich) PQ 1 M in water; diamide (Sigma-Aldrich) DI 0.2 M in water; menadione (Sigma-Aldrich) K3 360 mM in DMSO; cumene hydroperoxide (Sigma-Aldrich) CHP 0.25 M in DMSO; sodium hypochlorite (Sigma-Aldrich) NaOCl 5%; ciprofloxacin (Sigma-Aldrich) CIP 0.5 µg/µL in water; ampicillin AMP 1µg/µL in water.
Plates were incubated at 37 • C for 18 h and the diameter of inhibition was measured in mm. The area of inhibition was calculated as: [diameter of inhibition/2] 2 × 3.14. To compare the effect of different oxidants, data were normalized as following: [area of inhibition of the interested mutant] × 100/[area of inhibition of the WT].

H 2 O 2 Survival Assay
For survival assay, the susceptibility of E. coli to H 2 O 2 was tested in liquid medium. Briefly, overnight cultures were diluted to 2 × 10 7 CFU/mL in 10 mL LB. 1 mL of H 2 O 2 diluted in LB was added to the bacterial suspension to reach a final concentration of 2.5 mM. The corresponding control received 1 mL LB without H 2 O 2 . Bacteria were grown at 37 • C, 180 rpm. At indicated time points, 20 µL of each sample were serially diluted in LB by 10-fold dilutions. 10 µL of each dilution were spotted on LB agar supplemented with 100 U/mL of bovine liver catalase (Sigma-Aldrich). Plates were incubated overnight at 37 • C. Percent survival was calculated as [CFU from H 2 O 2 -treated sample/CFU from untreated sample] × 100.

Expression Levels Assessed by qRT-PCR
Overnight cultures were diluted in 10 mL of LB to OD 595 0.02. These fresh cultures were grown at 37 • C, 180 rpm for 2 h to reach exponential phase. Bacterial suspension was divided in 2 mL samples, and 200 µL of H 2 O 2 diluted in LB was added to reach the final concentrations indicated in the figures. The same volume of LB was added in the corresponding control conditions. Samples were incubated at 37 • C for 10 min. Subsequently, 1 mL was stabilized with 2 mL RNAprotect Bacteria Reagent (Qiagen, Hombrechtikon, Switzerland). RNA was purified using RNeasy Plus Mini Kit (Qiagen) according to the manufacturer instructions with on-column DNA digestion by RNase-Free DNase Set (Qiagen).
Quantitative PCR (qRT-PCR) was performed on RNA samples as previously described [10]. Briefly, the cDNA was produced by reverse-transcribing 500 ng of total RNA using a mix of random hexamers and oligo d(T) primers and Primescript reverse transcriptase enzyme (Takara Bio, Saint Germain-en-Laye, France). The efficiency of each pair of primers was tested with four serial dilutions of cDNA. Oligonucleotides are indicated in Table 2. PCR reactions (10 µL volume) contained 1:20 diluted cDNA, 2 × Power SYBR Green Master Mix (Thermo Fisher, Fisher Scientific AG, Reinach. Switzerland), and 300 nM of forward and reverse primers. PCRs were performed on a SDS 7900 HT instrument (Thermo Fisher) with the following parameters: 50 • C for 2 min, 95 • C for 10 min, and 45 cycles of 95 • C for 15 s, 60 • C for 1 min. Each reaction was performed in three replicates on 384-well plate. Raw Ct values obtained with SDS 2.2 (Thermo Fisher) were imported into Excel and normalization factors were calculated using the GeNorm method as described by Vandesompele et al. [26]. The absence of residual genomic DNA in RNA samples was verified by performing PCR reactions without RTase with the primer pair gyrB_N. Significance was assessed by one-way ANOVA with ad hoc Tukey's multiple comparisons test.

H 2 O 2 Degradation Mesurements by Amplex Red
Overnight cultures were diluted in LB to McFarland 1.0 using a Densimat (bioMérieux) and further diluted 10 fold in fresh LB. 10 mL were grown in a Falcon 50 at 37 • C for 2 h to reach exponential phase of growth. Pellets were washed with DPBS (Gibco Thermo Fisher) and resuspended to reach OD 595 = 0.1 in DPBS. 1 mL of H 2 O 2 diluted in sterile water was added to 10 mL of bacterial suspension for a final concentration of 1 mM of H 2 O 2 . At indicated time points, 10 µL were taken from each sample and diluted 1:200 in DPBS; 100 µL of each sample were transferred into a 96-well black plate with clear bottom (Corning). Amplex Red (Thermo Fisher) was used to detect H 2 O 2 according to manufacturer's instructions. Briefly, 100 µL of Amplex Red mix was added to each well for a final concentration of 27.5 µM Amplex Red and 0.1 UI/mL horseradish peroxidase (Sigma-Aldrich). The plate was incubated for 10 min at 37 • C and the fluorescence (excitation 535 nm, detection 595 nm) was read in a Spectramax Paradigm (Molecular Devices, Wokingham, UK). A H 2 O 2 calibration curve was generated by 1:2 serial dilutions of H 2 O 2 in DPBS (from 0.11 mM to 1.07 × 10 −4 mM) and used to calculate the H 2 O 2 concentration of the samples by linear regression.

Complementation of gpmA
The E. coli MG1655 gpmA gene with its native promoter was amplified from genomic DNA using KOD DNA polymerase (Toyobo) and the primers in Table 2 (gene ID Ecocyc database: EG11699). The single amino acid replacement of the 11 th histidine by an alanine (gpmA His11Ala) was obtained by overlap PCR using primers described in Table 2. The pWSK29 plasmid [23] was a kind gift from M. Roch (Geneva University). The plasmid and the PCR products were digested with the restriction enzymes EcoRI and KpnI (Thermo Fisher) and were gel-purified using QIAquick gel cleanup kit (Qiagen). T4 ligase (New England Biolabs) was used for the ligations and the ligation products (pWSK29 with either gpmA or gpmA His11Ala) were transformed in TOP10 Chemically Competent E. coli (C404010, Thermo Fisher). The coding region of the two cloned plasmids was verified by Sanger sequencing. Plasmids were electroporated in either WT or ∆gpmA strains.
The complemented ∆gpmA strains (pWSK29 with either gpmA or gpmA His11Ala) were compared to the WT and the ∆gpmA strains harboring the pWSK29 (empty) plasmid on LB agar plates containing ampicillin 100 µg/mL.

TraDIS Was Performed under Sublethal H 2 O 2 Exposure
To determine the optimal dose of H 2 O 2 to apply for the TraDIS experiment, we tested different concentrations of H 2 O 2 on the E.coli strain BW25113, the strain used to generate the TraDIS library [18] ( Figure 1A). The application of 2.5 mM H 2 O 2 increased the lag phase by 70 min, whereas 5 mM or more resulted in complete absence of bacterial growth. The growth rate of bacteria during the exponential phase (between OD 0.2 and 1.6) was identical when treated with 2.5 mM compared to no treatment.
We performed the TraDIS experiment in similar conditions with 2.5 mM H 2 O 2 . The H 2 O 2 -treated condition reached OD = 1 approximatively 140 min after untreated controls. We used the genome browser Artemis to observe the insertion site of the transposons ( Figure 1B). To analyze the comparative fitness of each gene under both conditions, we performed fitness analysis with the Bio::TraDIS pipeline.
In TraDIS and other Tn-seq techniques, the fitness of each gene deletion is assessed by sequencing. Mutants that are less fit in a given condition will be outcompeted and therefore less abundant, which is approximately measured by insertion frequency. To ensure the relevance of our data, we scrutinized the TraDIS data for an impact on oxyR. The transcription factor OxyR is a well-described H 2 O 2 sensor that regulates E. coli antioxidant response and deletion of the gene has been shown to increase sensitivity against oxidative stress [31]. As expected, the frequency of insertions was significantly reduced after exposure to H 2 O 2 indicating oxyR mutants are less fit than the wild type in the presence of H 2 O 2 ( Figure 1C). Using the values derived for oxyR as a threshold, we identified nine genes that displayed higher fold-change values suggesting of a role for each of these genes in H 2 O 2 tolerance (Table 3). Several genes were already described in the oxidative stress response. . Representative examples of essential gene (murI in red box), non-essential genes (btuB, in green box) and genes with a reduced fitness in the H2O2 condition (dps in orange box), (N = 2) (C) Fitness analysis of the TraDIS data, H2O2-treated condition compared to control. Each dot represents a gene, X-axis represents the difference in number of insertions in the H2O2 condition compared to control, Y-axis represents the statistical significance. Nine genes (in pink) displayed a significant and more extreme change than the H2O2-sensor gene oxyR.  No transposon insertion was significantly overrepresented in the H 2 O 2 condition, suggesting that no gene deletion is protective against H 2 O 2 in these conditions. This analysis considered only transposon insertions inside the coding regions of genes. Transposons disrupting promoters or altering the expression of genes such as polar effect were not considered by this analysis.

H 2 O 2 Susceptibility of Single-Gene Deletion Identified by TraDIS
Single-deletion mutants of genes identified by the TraDIS experiments were tested against H 2 O 2 to evaluate the sensitivity of each mutant. To ensure the absence of undesired mutations, cognate E. coli strain MG1655 mutants were created by P1 phage transduction of the relevant mutations from the Keio collection [21]. The susceptibility to H 2 O 2 of the single-gene deletion mutants was assessed by disk diffusion assay. The mutant deleted for the catalase katG, known as the principal H 2 O 2 scavenger at high concentration [32], was used as positive control. As expected, the deletion of oxyR led to a dramatic increase of the inhibition diameter generated by H 2 O 2 ( Figure 2). The deletion of gpmA increased the sensitivity to H 2 O 2 to the same extent as the katG deletion. Similarly, loss of hfq also increased significantly the sensitivity to H 2 O 2 . Other genetic deletions did not significantly alter the H 2 O 2 susceptibility in the disk diffusion assay.
Antioxidants 2021, 10, x FOR PEER REVIEW 8 of 19 No transposon insertion was significantly overrepresented in the H2O2 condition, suggesting that no gene deletion is protective against H2O2 in these conditions. This analysis considered only transposon insertions inside the coding regions of genes. Transposons disrupting promoters or altering the expression of genes such as polar effect were not considered by this analysis.

H2O2 Susceptibility of Single-Gene Deletion Identified by TraDIS
Single-deletion mutants of genes identified by the TraDIS experiments were tested against H2O2 to evaluate the sensitivity of each mutant. To ensure the absence of undesired mutations, cognate E. coli strain MG1655 mutants were created by P1 phage transduction of the relevant mutations from the Keio collection [21]. The susceptibility to H2O2 of the single-gene deletion mutants was assessed by disk diffusion assay. The mutant deleted for the catalase katG, known as the principal H2O2 scavenger at high concentration [32], was used as positive control. As expected, the deletion of oxyR led to a dramatic increase of the inhibition diameter generated by H2O2 (Figure 2). The deletion of gpmA increased the sensitivity to H2O2 to the same extent as the katG deletion. Similarly, loss of hfq also increased significantly the sensitivity to H2O2. Other genetic deletions did not significantly alter the H2O2 susceptibility in the disk diffusion assay.

∆gpmA Mutant Was More Sensitive to H 2 O 2 but Not to Other Oxidants
Single-deletion mutants were tested against other oxidants by disk diffusion assay ( Figure 3). The deletion of oxyR and hfq led to an increase of the inhibition area of a wide range of oxidants ( Figure 3C,E). The deletion of gpmA led to a hypersensitivity to H 2 O 2 but not to other oxidants or antibiotics ( Figure 3B,E). This pattern was highly similar to the sensitivity of the ∆katG mutant used as positive control ( Figure 2E).

ΔgpmA Mutant Was More Sensitive to H2O2 but Not to Other Oxidants
Single-deletion mutants were tested against other oxidants by disk diffusion assay ( Figure 3). The deletion of oxyR and hfq led to an increase of the inhibition area of a wide range of oxidants ( Figure 3C,E). The deletion of gpmA led to a hypersensitivity to H2O2 but not to other oxidants or antibiotics ( Figure 3B,E). This pattern was highly similar to the sensitivity of the ΔkatG mutant used as positive control ( Figure 2E).  Figure S1). The significance of the difference with the WT is represented on the normalized data by stars (mean +/− SD, N = 3). *, **, ***, **** correspond to p < 0.05, 0.01, 0.001, and 0.0001, respectively.  Figure S1). The significance of the difference with the WT is represented on the normalized data by stars (mean +/− SD, N = 3). *, **, ***, **** correspond to p < 0.05, 0.01, 0.001, and 0.0001, respectively.
The deletion mutants of the other genes identified by TraDIS were also tested against these oxidants (Supplementary Materials Figure S1). The ∆dksA mutant was more sensitive to methylhydroquinone, cumene peroxide, diamide and ciprofloxacin, and the ∆nhaA mutant was slightly more sensitive to diamide and ciprofloxacin. This suggests that these mutants, despite no increased sensitivity to H 2 O 2 in these conditions, were more sensitive to other oxidative stresses. Other mutants did not display significant differences compared to WT.

gpmA Is Upregulated by H 2 O 2 Exposure
In a previous study, we performed a RNA-seq analysis of E. coli BW25113 after a 10 min exposure to a sublethal concentration (2.5 mM) of H 2 O 2 [10]. Among the ten genes identified by TraDIS, only two genes, dps and gpmA, were significantly dysregulated by H 2 O 2 ( Figure 4A). In these settings, gpmA was upregulated over fourfold. As gpmA is part of the glycolysis reaction in E. coli, we extracted the transcriptomic data for the glycolysis and the TCA cycle (Supplementary Materials Figure S2). Other enzymes from the glycolysis (pgi, pfkAB, fbaAB, pgk) were also upregulated, suggesting an increased activity of glycolysis following exposure to H 2 O 2 .
The deletion mutants of the other genes identified by TraDIS were also tested against these oxidants (Supplementary Materials Figure S1). The ΔdksA mutant was more sensitive to methylhydroquinone, cumene peroxide, diamide and ciprofloxacin, and the ΔnhaA mutant was slightly more sensitive to diamide and ciprofloxacin. This suggests that these mutants, despite no increased sensitivity to H2O2 in these conditions, were more sensitive to other oxidative stresses. Other mutants did not display significant differences compared to WT.

gpmA Is Upregulated by H2O2 Exposure
In a previous study, we performed a RNA-seq analysis of E. coli BW25113 after a 10 min exposure to a sublethal concentration (2.5 mM) of H2O2 [10]. Among the ten genes identified by TraDIS, only two genes, dps and gpmA, were significantly dysregulated by H2O2 ( Figure 4A). In these settings, gpmA was upregulated over fourfold. As gpmA is part of the glycolysis reaction in E. coli, we extracted the transcriptomic data for the glycolysis and the TCA cycle (Supplementary Materials Figure S2). Other enzymes from the glycolysis (pgi, pfkAB, fbaAB, pgk) were also upregulated, suggesting an increased activity of glycolysis following exposure to H2O2.
We confirmed the impact of H2O2 on gpmA expression in the MG1655 strain used in this study by qRT-PCR. The gpmA RNA was upregulated following sublethal exposure of H2O2 in a dose-dependent manner ( Figure 4C). Induction of gpmA expression was less impressive than the catalase katG, a known H2O2-responsive gene ( Figure 4B).  We confirmed the impact of H 2 O 2 on gpmA expression in the MG1655 strain used in this study by qRT-PCR. The gpmA RNA was upregulated following sublethal exposure of H 2 O 2 in a dose-dependent manner ( Figure 4C). Induction of gpmA expression was less impressive than the catalase katG, a known H 2 O 2 -responsive gene ( Figure 4B).

Catalase Activity Is Not Involved in the Increased Sensitivity of ∆gpmA to H 2 O 2
As the ∆gpmA mutant displayed a similar sensitivity to oxidants compared to the ∆katG mutant ( Figure 3E), we measured catalase expression and activity in the presence of H 2 O 2 . The ∆gpmA mutant did not exhibit a growth defect compared to the WT in liquid LB (Supplementary Figure S3), so we first assessed the sensitivity of the gpmA mutant to H 2 O 2 in liquid LB medium by counting surviving bacteria after H 2 O 2 exposure ( Figure 5A). Two hours after the addition of 2.5 mM H 2 O 2 , a 100-fold difference in the number of surviving bacteria in the gpmA and the oxyR mutant compared to the WT was observed ( Figure 5B).

Catalase Activity Is Not Involved in the Increased Sensitivity of ΔgpmA to H2O2
As the ΔgpmA mutant displayed a similar sensitivity to oxidants compared to the ΔkatG mutant ( Figure 3E), we measured catalase expression and activity in the presence of H2O2. The ΔgpmA mutant did not exhibit a growth defect compared to the WT in liquid LB (Supplementary Figure S3), so we first assessed the sensitivity of the gpmA mutant to H2O2 in liquid LB medium by counting surviving bacteria after H2O2 exposure ( Figure  5A). Two hours after the addition of 2.5 mM H2O2, a 100-fold difference in the number of surviving bacteria in the gpmA and the oxyR mutant compared to the WT was observed ( Figure 5B).  We measured the expression levels of the three enzymes of E. coli that are known to degrade H 2 O 2 , the alkyl hydroperoxide reductase encoded by ahpC, the catalase/hydroperoxidase HPI encoded by katG and the catalase HPII encoded by katE. We compared the WT and the ∆gpmA strain, in presence or in absence of H 2 O 2 ( Figure 5C). There was no significant difference in the expression of the three genes between the WT and the ∆gpmA strain. Upregulation of ahpC and katG was observed after the addition of 2.5 mM of H 2 O 2 in both strains. There was no significant difference in katE expression after H 2 O 2 exposure, which was expected as it is not regulated by OxyR but by RpoS and upregulated during the stationary phase of bacterial growth [33].
To test if the catalase activity was affected by the deletion of gpmA, we measured the degradation of 1 mM of H 2 O 2 of the WT and the ∆gpmA using the H 2 O 2 -sensitive probe Amplex Red. There was no difference in H 2 O 2 degradation between the WT and the ∆gpmA strains. The ∆katG strain, which is defective for the main H 2 O 2 scavenger at high concentration, was unable to degrade H 2 O 2 . Altogether, this suggests that the higher sensitivity of the ∆gpmA to H 2 O 2 is independent of catalase activity.

Other Carbon Sources Cannot Compensate the H 2 O 2 Hypersensitivity of ∆gpmA Mutant
In LB medium, amino acids are the main source of carbon and there is virtually no glucose [34]. We wondered if the supplementation with metabolites entering the central metabolism at different levels could affect the H 2 O 2 sensitivity of the ∆gpmA mutant. The addition of alternative carbon source did not significantly modify the H 2 O 2 susceptibility of the WT or the ∆gpmA mutant ( Figure 6A). We also tested H 2 O 2 sensitivity in M9 minimal media with these metabolites as the only carbon source. There was no difference in the sensitivity of the WT in M9 + glucose compared to LB + glucose. The WT strain was slightly more sensitive in M9 + acetate compared to M9 + glucose. The ∆gpmA strain displayed a higher sensitivity in the M9 conditions compared to LB conditions. M9 plates with citrate as the sole source of carbon led to limited growth even after 48 h and were therefore not measurable. Addition of 0.5% pyruvate led to a complete disappearance of the zone of inhibition (data not shown) probably because pyruvate reacts with H 2 O 2 to produce CO 2 , acetate and water [35].
In Salmonella Typhimurium, the ∆gpmA mutant was more susceptible to H 2 O 2 than the WT in aerobic conditions, but not in anaerobic conditions, and the addition of the electron acceptor nitrate restored the hypersusceptibility of ∆gpmA [20]. We tested the H 2 O 2 susceptibility of E. coli WT and ∆gpmA in anaerobic conditions. Interestingly, it appeared that the WT was slightly more sensitive to H 2 O 2 in anaerobic conditions than in aerobic conditions suggesting that the exposure to oxygen protect in part against H 2 O 2 damage. However, the ∆gpmA mutant did not display any difference in H 2 O 2 sensitivity between anaerobic and aerobic conditions and the difference between the ∆gpmA mutant and the WT was maintained in anaerobic conditions. As E. coli is also able to use other electron acceptors than oxygen for respiration, we tested the addition of sodium nitrate in anaerobic conditions, but this did not change the area of inhibition induced by H 2 O 2 compared to the anaerobic condition without nitrate (data not shown).
Altogether, we explored a potential impact of factors affecting glycolysis following H 2 O 2 exposure, but we did not observe significant changes in conditions of low oxygen or using different carbon sources.

The Function of gpmA Is Necessary for H 2 O 2 Tolerance
The ∆gpmA was complemented by native gpmA gene including its natural promoter using the low copy plasmid pWSK29. The complemented strain displayed similar H 2 O 2 susceptibility than the WT strain ( Figure 7A). A mutation previously described as to be necessary for the function of gpmA, namely the substitution of the histidine 11 residue by an alanine [36], resulted in restauration of the hypersensitivity to H 2 O 2 . These data suggest that the function of gpmA is necessary to reach the WT levels of tolerance against H 2 O 2 .

The Function of gpmA Is Necessary for H2O2 Tolerance
The ΔgpmA was complemented by native gpmA gene including its natural promoter using the low copy plasmid pWSK29. The complemented strain displayed similar H2O2 susceptibility than the WT strain ( Figure 7A). A mutation previously described as to be necessary for the function of gpmA, namely the substitution of the histidine 11 residue by an alanine [36], resulted in restauration of the hypersensitivity to H2O2. These data suggest that the function of gpmA is necessary to reach the WT levels of tolerance against H2O2.
E. coli possess a secondphosphoglycerate mutase encoded by gpmM, which presents no sequence similarity with gpmA [37]. Contrary to gpmA, the expression level of gpmM was slightly downregulated after the addition of H2O2 in a previous RNA-seq dataset (Figure 7B). This suggests that following exposure to H2O2, gpmA represents the principal form of phophoglygerate mutase. We tested the ΔgpmM mutant for H2O2 sensitivity. Con- E. coli possess a secondphosphoglycerate mutase encoded by gpmM, which presents no sequence similarity with gpmA [37]. Contrary to gpmA, the expression level of gpmM was slightly downregulated after the addition of H 2 O 2 in a previous RNA-seq dataset ( Figure 7B). This suggests that following exposure to H 2 O 2 , gpmA represents the principal form of phophoglygerate mutase. We tested the ∆gpmM mutant for H 2 O 2 sensitivity. Contrary to gpmA, the deletion of gpmM did not increase the sensitivity to H 2 O 2 ( Figure 7C

Discussion
The production of H2O2 by phagocytes from the human immune system a tobacilli species of the normal microbiota are essential for the prevention of co from various opportunistic pathogens. Although H2O2 effects on bacteria have ied for years, the mechanisms by which H2O2 exerts its antimicrobial activity completely understood [14,16].
Our TraDIS analysis identified 10 mutants with fitness defect upon H2O2 implicating a role for these genes under H2O2-induced oxidative stress. Only t ten genes, oxyR, gpmA and hfq, showed a significantly higher susceptibility to H knocked-out. This could be due to the differences in the settings between the periment and the disk diffusion assay. For example, the DNA-binding protei by dps protects DNA from H2O2 damage through iron sequestration and this more important in stationary phase of growth [38]. However stationary phase

Discussion
The production of H 2 O 2 by phagocytes from the human immune system and by Lactobacilli species of the normal microbiota are essential for the prevention of colonization from various opportunistic pathogens. Although H 2 O 2 effects on bacteria have been studied for years, the mechanisms by which H 2 O 2 exerts its antimicrobial activity is still incompletely understood [14,16].
Our TraDIS analysis identified 10 mutants with fitness defect upon H 2 O 2 exposure, implicating a role for these genes under H 2 O 2 -induced oxidative stress. Only three of the ten genes, oxyR, gpmA and hfq, showed a significantly higher susceptibility to H 2 O 2 when knocked-out. This could be due to the differences in the settings between the TraDIS experiment and the disk diffusion assay. For example, the DNA-binding protein encoded by dps protects DNA from H 2 O 2 damage through iron sequestration and this defense is more important in stationary phase of growth [38]. However stationary phase cultures of each knockout was treated with H 2 O 2 in liquid medium, their respective growth was not different compared to the WT, except for the ∆oxyR strain (data not shown). The majority of genes we identified by TraDIS (oxyR, dps, rpoS, dksA, hfq, polA) have already been reported to respond to oxidative stress in E. coli. The transcription factor OxyR is a well described sensor of H 2 O 2 , which regulates an extensive and coordinated antioxidant transcriptional response [9,39]. The RNA polymerase subunit RpoS regulates the general stress response and was previously described to be activated by oxidative stress [40], and the deletion of this gene increases sensitivity to H 2 O 2 [41]. The RNA polymerase accessory protein DksA senses oxidative stress through its cysteine residues and participates to the transcriptional response against oxidative stress [42]. Hfq, a RNA-binding protein that affects many cellular processes influences both the small RNA OxyS and the translation of rpoS described above in E. coli [43,44]. The DNA polymerase I encoded by polA is implicated in DNA repair and non-functional PolA increases H 2 O 2 -sensitivity [45,46]. The polA, rbsR, dps, oxyR, corA, rpoS genes were also identified in a similar experiment performed previously on Salmonella enterica serovar Typhimurium under sublethal H 2 O 2 exposure [19]. The dksA and nhaA mutants, despite showing no increase in sensitivity to H 2 O 2 , were slightly more sensitive to other oxidants than WT using disk diffusion assay. Other validation experiments, such as competition assay with WT under H 2 O 2 stress, might better reflect the TraDIS experimental conditions.
On the other hand, several genes previously identified in the literature as necessary for H 2 O 2 tolerance were not identified by this TraDIS experiment. For example, xthA, whose deletion mutant is more sensitive to H 2 O 2 [47], displayed a decreased fitness in H 2 O 2 condition but did not reach the threshold of significance. An explanation could lie in the fact that unlike antibiotics, H 2 O 2 is rapidly degraded by bacteria. The duration of the exposure to H 2 O 2 performed for the TraDIS may have been insufficient to identify all genes implicated in H 2 O 2 tolerance. Secondly, the stress was applied against pooled mutants in liquid where other mutants could provide cross-protection for susceptible mutants. For example, the catalase KatG, which is known to protect against H 2 O 2 , was not identified by TraDIS, probably because of this phenomenon. This was also the case in a previous study that used Tn-seq with H 2 O 2 in Salmonella Typhimurium, where none of the catalase genes were identified [19]. Thus, our TraDIS data only identified those mutants that showed fitness defects despite cross-protection and inherent H 2 O 2 degradation.
The TraDIS experiment also identified genes that, to our knowledge, were not previously associated with E. coli H 2 O 2 sensitivity. The magnesium ion transporter encoded by corA had been shown to be more sensitive to lactoperoxidase-thiocyanate stress but not to H 2 O 2 [48]. rbsR controls the transcription of the operon involved in ribose catabolism and transport and the salvage pathway of purine nucleotide synthesis [49]. corA and rbsR were also identified in a similar Tn-seq experiment using H 2 O 2 on Salmonella enterica serovar Typhimurium [19]. The Na + :H+ antiporter nhaA is implicated in other stress responses against sodium ion, pH homeostasis and in maintaining antibiotic tolerance under starvation [50]. The glycolysis enzyme gpmA has been previously identified by a Tn-seq experiment following H 2 O 2 exposure in the Gram-negative bacteria Salmonella enterica serovar Typhimurium [20].
When tested with diverse oxidants that damage bacteria through different modes of action, ∆gpmA was specifically more sensitive to H 2 O 2 , like the ∆katG strain. However, it was not through a differential expression of H 2 O 2 -scavenging genes or a decreased catalase activity of the strain, suggesting a different mode of action. Moreover, the upregulation of gpmA by sublethal exposure of H 2 O 2 suggests the importance of gpmA in H 2 O 2 tolerance. Under oxidative stress, some enzymes of the central metabolism have been shown to be upregulated. The glucose-6-phosphate isomerase encoded by pgi have been shown to be regulated by the oxidative stress sensitive regulators SoxRS [51]. Similarly, in the TCA cycle, the aconitase acnA and the fumarase fumC are regulated by SoxRS and are upregulated under H 2 O 2 exposure [52,53]. The hypersensitivity to H 2 O 2 of ∆gpmA mutant could be complemented with the low-copy plasmid pWSK29 expressing the WT gpmA gene under its native promoter but not if the histidine 11 was mutated to an alanine. This strongly suggests that the function of gpmA affects E. coli tolerance to H 2 O 2 .
Surprisingly, addition of other metabolites or the absence of oxygen did not abolish the difference in H 2 O 2 sensitivity between the WT and the ∆gpmA mutant. These data contrast with previous work on Salmonella enterica serovar Typhimurium, where other metabolites entering metabolism downstream of gpmA reaction (for a scheme of glycolysis, see Supplementary Figure S2) could complement the increased sensitivity of a ∆gpmA mutant and where anoxic environment abolished the difference of H 2 O 2 susceptibility between WT and ∆gpmA mutant [20]. In the same study, metabolomics approach showed that H 2 O 2 exposure led to an increase of glycolysis and fermentation that was important in Salmonella H 2 O 2 tolerance. This contrasts with previous metabolomics analysis on E. coli after H 2 O 2 treatment which reported a decrease of metabolites related to glycolysis and TCA cycle, changes that were common to other stress conditions such as heat shock and cold stress [54]. Altogether, this suggests a different metabolic adaptation to H 2 O 2 stress between E. coli and Salmonella and a difference of gpmA function. More research is needed to better understand the mechanisms of gpmA effects in H 2 O 2 tolerance in E. coli and in other organisms.
Contrary to vertebrates that only possess one phosphoglycerate mutase, some eubacteria, among which relevant pathogens including E. coli, encode two enzymes that display no sequence similarity [55]. The 2,3-bisphosphoglycerate-dependent phosphoglycerate mutase (or dPGM), encoded by gpmA is common to bacteria and vertebrates, whereas 2,3-bisphosphoglycerate-independent phosphoglycerate mutase (or iPGM) encoded by gpmM is shared by bacteria and higher plants. As the double deletion of gpmA and gpmM have been suspected non-viable in E. coli, the glycolysis function is assumed by gpmM in the gpmA-deleted strain and vice versa [56]. The deletion of E. coli gpmM did not affect H 2 O 2 sensitivity, suggesting that only gpmA function has a role under H 2 O 2 exposure. This led to the hypothesis that gpmM could be damaged by H 2 O 2 and its function is replaced by gpmA under H 2 O 2 exposure. This happens for other enzymes of the TCA cycle, the aconitase and the fumarase, where oxidative-resistant isoforms (acnA, fumC) replace oxidative-sensitive isoforms (acnB, fumA, fumB), after H 2 O 2 exposure [52,53]. Cysteine residues can be more prone to oxidation by H 2 O 2 than other amino acids [57]. GpmM possesses two cysteine residues, which can result in H 2 O 2 -induced damage from oxidation of these residues. As GpmA does not possess cysteine residues, it could be more resistant to H 2 O 2 than GpmM. The cysteine residues of GpmM are not implicated in active sites described in current models (Ecocyc, Uniprot). While Cys397 seems buried and is not conserved in Gram-positive bacteria, Cys424 seems to be more accessible on the protein models and is present in both Gram-negative (P. aeruginosa, Salmonella enterica, K. pneumoniae) and Gram-positive bacteria (S. aureus, B. subtilis). Additional studies are needed to evaluate their potential implication in oxidative stress susceptibility.

Conclusions
This work was aimed at expanding the knowledge of which genes are implicated in H 2 O 2 tolerance. The main finding of this study was that a functional gpmA gene is required for tolerance to H 2 O 2 . This is the first time that gpmA was highlighted as an important contributor to the E. coli tolerance to H 2 O 2 , and it links defense against oxidative stress to central metabolism.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/antiox11102053/s1, Table S1: Primers used to validate the gene replacement by the kanamycin cassette from the Keio collection, Figure S1: Sensitivity of the deletion mutants of the TraDIS exposed to various oxidants, Figure S2: Schematic diagram of H 2 O 2 -induced transcriptional changes of glycolysis and TCA cycle, Figure S3: The deletion of gpmA did not affect bacterial growth.