Genome-Wide Screening of Oxidizing Agent Resistance Genes in Escherichia coli

The use of oxidizing agents is one of the most favorable approaches to kill bacteria in daily life. However, bacteria have been evolving to survive in the presence of different oxidizing agents. In this study, we aimed to obtain a comprehensive list of genes whose expression can make Escherichia coli cells resistant to different oxidizing agents. For this purpose, we utilized the ASKA library and performed a genome-wide screening of ~4200 E. coli genes. Hydrogen peroxide (H2O2) and hypochlorite (HOCl) were tested as representative oxidizing agents in this study. To further validate our screening results, we used different E. coli strains as host cells to express or inactivate selected resistance genes individually. More than 100 genes obtained in this screening were not known to associate with oxidative stress responses before. Thus, this study is expected to facilitate both basic studies on oxidative stress and the development of antibacterial agents.


Introduction
In nature, bacteria live under various environmental stresses such as oxidative stress, which is one of the most common challenges for bacteria living in aerobic conditions. Reactive oxygen species (ROSs), reactive nitrogen species (RNSs), and reactive chlorine species (RCSs) are three major sources of oxidative stress [1]. In bacteria, ROSs including superoxide anions (O 2 •− ), hydrogen peroxide (H 2 O 2 ), and hydroxyl radicals ( • OH) can be formed not only endogenously during electron transfer in the respiratory chain [2], but also exogenously under different conditions such as UV exposure [3]. Although ROSs at certain low concentrations have been demonstrated to be essential for several physiological processes and the cellular redox balance [4], a high level of ROSs has been proven to cause damages to nucleic acids, proteins, fatty acids, and other cellular components [2,3,5]. Different from free radicals such as O 2 •− and • OH, the non-radical ROS H 2 O 2 is a strong oxidant and has high activation energy, which makes it reactive with transition metal centers, selenoproteins, and selected thiol proteins [6]. An excess presence of H 2 O 2 has been implicated in the interruption of iron homeostasis in cells [7]. For instance, a high level of H 2 O 2 can inactivate the Escherichia coli Isc iron-sulfur assembly system [8]. Oxidative stress caused by H 2 O 2 can also induce DNA degradation in bacteria and then inhibit bacterial growth [9]. Differently, hypochlorite (HOCl) mainly targets proteins which contain sulfur, aromatic rings, and primary amines, resulting in protein aggregation and degradation [10]. Due to their disinfecting effects on bacteria, H 2 O 2 and HOCl, have been commonly used to kill bacteria in daily life. Moreover, antibiotics such as landomycin E, which can induce a rapid generation of H 2 O 2 , have been applied to treat bacterial infections [11,12].
To protect themselves from H 2 O 2 damages, bacteria have evolved various defense pathways. Besides enzymatic scavengers such as catalase, superoxide dismutase, and peroxidase [13], the stress-induced regulon OxyR is one of the most well-studied approaches 1 mM increments from 1 to 20 mM, or 0.25 mM increments from 0 to 1 mM in LB or M9 minimal medium. The lowest concentration at which bacteria could not grow was recorded as the corresponding MICs.

Bioinformatical Analyses
The subcellular localization of proteins was obtained from the EcoCyc E. coli Database by typing individual gene names into the database [24]. The identified proteins were classified into functional categories according to their annotated functions in the UniProt-GOA Database [25] and analyzed by DAVID Bioinformatics Resources by typing individual gene names into the database [26]. Protein-protein functional interaction networks were analyzed with the STRING database [27], in which active interaction sources from experiments, databases, co-expression, neighborhood, gene fusion, and co-occurrence were selected. The minimum required interaction score of medium confidence was chosen.

Genome-Wide Screening of Resistance Genes against HOCl
Similar screenings were performed with HOCl (in the form of NaOCl) as the oxidizing agent. The MIC was 1 mM for the no-insert control strain. Then, we used a 2-fold MIC (2 mM NaOCl) to screen the whole ASKA library. Compared to H 2 O 2 , there were fewer resistance genes for HOCl. Only 114 strains could grow in 2 mM NaOCl. We divided them into categories based on their biological functions (Table 3).
Bioinformatical analyses were performed on these candidate genes ( Figure 2). Compared to the H 2 O 2 results, a higher number of inner membrane-associated proteins appeared to be coded by HOCl-resistance genes. Similar to H 2 O 2 -resistance genes, the functions of these candidate genes also cover a wide range of biological processes and focus on stress responses, membrane functions, metabolism, and gene expression. A little higher proportion of genes involved in membrane functions were identified for HOCl-resistance.
Threonine deaminase; carries out the first step in the synthesis of isoleucine [41]. rho Transcription termination factor; required for one of the two major types of termination of RNA transcription [42]. ydhL DUF1289 domain-containing protein.
mlaB Intermembrane phospholipid transport system protein; forms a stable complex with MlaF, MlaE, and MlaD and is required for the stability of this complex [43]. citG Triphosphoribosyl-dephospho-CoA synthase [44]. Table 3. List of HOCl-resistance genes from the genome-wide screening.
Bioinformatical analyses were performed on these candidate genes ( Figure 2). Compared to the H2O2 results, a higher number of inner membrane-associated proteins appeared to be coded by HOCl-resistance genes. Similar to H2O2-resistance genes, the functions of these candidate genes also cover a wide range of biological processes and focus on stress responses, membrane functions, metabolism, and gene expression. A little higher proportion of genes involved in membrane functions were identified for HOCl-resistance. The protein-protein interaction network only showed two clusters: (1) the respiration chain including hydrogenase (hyaA, hyaC, hyaE, hyaF) and ubiquinol oxidase (cbdA); (2) flagellar biosynthesis and mobility including fliP, fliS, motB, cheW, and tap, which was found also in H2O2-resistance genes. To identify genes which mediate stronger resistance to HOCl, we used a 3-fold MIC for the wide-type strain (3 mM NaOCl) to screen the identified 114 candidate genes. Only 23 strains could grow, which expressed elaA, exbD, frmB, hyaC, hyaE, ilvA, marC, motB, prpE, rnc, rsuA, sanA, sucC, tap, ybhR, ycbB, yccJ, ycdK, yedQ, yfdY, yhdJ, yoaE, and ytfI. They were selected for further validation. A summary of the current knowledge about these genes is presented in Table 4. Similar to the H2O2 results, membrane transport and gene expression are also two important functions of HOCl-resistance genes. Furthermore, cellular redox balances and DNA damage responses are two unique processes of HOCl-resistance genes. More details are described in the Discussion section. Table 4. List of genes mediating stronger HOCl-resistance.

Gene
Known or Projected Functions prpE Propionate-CoA ligase; catalyzes the synthesis of propionyl-CoA from propionate and CoA [31].
ycbB Periplasmic L,D-transpeptidase; plays a role in the protective remodeling of peptidoglycan during cell envelope stress [48]. yccJ PF13993 family protein.
sanA Multi-copy expression of sanA complements the vancomycin sensitivity of an E. coli K-12 mutant with outer membrane permeability defects [60]. rsuA Pseudo-uridine synthase that is responsible for pseudouridylation of 16S rRNA at position 516 [61].

Effects of Overexpressing Selected Resistance Genes in the MG1655 Strain
The host strain for the ASKA library is AG1, which was engineered for high transformation efficiency. Therefore, the AG1 strain may have potential inferences for the screening. To validate the effects of selected candidate genes which made cells resistant to concentrations that were 3-fold the MICs for wild-type cells in the first screening, we overexpressed each of them (20 genes for H 2 O 2 reported in Table 2 and 23 genes for HOCl reported in Table 4) in the K-12 strain MG1655, which is commonly used for E. coli physiology studies. The LB medium is a rich medium containing amino acids that react with oxidizing agents to potentially affect cellular responses. Therefore, we used both the LB medium and the M9 minimal medium (0.2% glucose) to determine MICs in the validation experiments. MICs of H 2 O 2 and NaOCl for the no-insert control strain (MG1655 with pCA24N empty vector) were 2 mM for H 2 O 2 and 2 mM for NaOCl. Then, we determined the MICs of H 2 O 2 or NaOCl for each resistance gene overexpressed in MG1655 cells (Figure 3). All the genes did not show significant differences in the MICs using the LB medium and the M9 minimal medium. The results were consistent with those of the ASKA library screening. All candidate resistance genes allowed MG1655 cells to grow at concentrations at least 3-fold the MICs for the control strain. For H 2 O 2 , there were 10 genes that allowed cells to grow in the presence of even higher concentrations. Besides katE and katG, which encode two known catalases, overexpression of rho made cells resistant to 12 mM H 2 O 2 , while overexpression of appY, mcbR, or yncG increased the MIC to 5-fold (10 mM H 2 O 2 ) of the MIC for the control strain. On the other hand, overexpression of only one gene (yoaE) required a slightly higher MIC (8 mM NaOCl) for HOCl resistance.

Effects of Inactivating Selected Resistance Genes in the E. coli Genome
To further confirm the resistance induced by the selected genes in MG1655 overexpression tests, we determined the MICs of H 2 O 2 or NaOCl in E. coli cells by inactivating each individual gene. For this purpose, we utilized the Keio collection, which contains strains with each non-essential E. coli gene inactivated. Firstly, we determined the MICs of H 2 O 2 or NaOCl for the wild-type control of the Keio collection, i.e., the BW25113 strain. The MICs of BW25113 cells were 5 mM for H 2 O 2 and 6 mM for NaOCl. Then, we determined the MICs of H 2 O 2 or NaOCl for each resistance gene inactivated in BW25113 cells ( Figure 4). As shown, inactivation of most candidate genes (except for prpE for HOCl resistance) made E. coli cells more sensitive to H 2 O 2 or NaOCl compared to the wild-type strain, confirming that resistance genes identified from the ASKA library could help cells survive in the presence of oxidizing agents. Compared to gene overexpression testing, in which all the candidates showed highly significant effects (Figure 3), gene inactivation testing did not produce highly significant effects (except for ∆katE). This observation indicates that there are other defense mechanisms which can compensate for the inactivated resistance genes. The lowest concentration at which bacteria could not grow was recorded as the corresponding MIC. Each strain was tested in three biological replicates. All the differences between MICs of the candidate genes and those of the control were highly significant (p < 0.001).

Effects of Inactivating Selected Resistance Genes in the E. coli Genome
To further confirm the resistance induced by the selected genes in MG1655 overexpression tests, we determined the MICs of H2O2 or NaOCl in E. coli cells by inactivating each individual gene. For this purpose, we utilized the Keio collection, which contains strains with each non-essential E. coli gene inactivated. Firstly, we determined the MICs of H2O2 or NaOCl for the wild-type control of the Keio collection, i.e., the BW25113 strain. The MICs of BW25113 cells were 5 mM for H2O2 and 6 mM for NaOCl. Then, we determined the MICs of H2O2 or NaOCl for each resistance gene inactivated in BW25113 cells ( Figure 4). As shown, inactivation of most candidate genes (except for prpE for HOCl resistance) made E. coli cells more sensitive to H2O2 or NaOCl compared to the wild-type strain, confirming that resistance genes identified from the ASKA library could help cells survive in the presence of oxidizing agents. Compared to gene overexpression testing, in which all the candidates showed highly significant effects (Figure 3), gene inactivation

MICs of HOCl (mM)
LB medium M9 minimal Figure 3. MICs of selected genes overexpressed in MG1655. MIC determination was performed by varying the concentration of the oxidizing agents, with 1 mM increments from 1 to 20 mM in the LB medium or the M9 minimal medium. The lowest concentration at which bacteria could not grow was recorded as the corresponding MIC. Each strain was tested in three biological replicates. All the differences between MICs of the candidate genes and those of the control were highly significant (p < 0.001).
Antioxidants 2021, 10, x FOR PEER REVIEW 9 of 15 testing did not produce highly significant effects (except for ∆katE). This observation indicates that there are other defense mechanisms which can compensate for the inactivated resistance genes. The lowest concentration at which bacteria could not grow was recorded as the corresponding MIC. Each strain was tested in three biological replicates. Significant differences (p < 0.05) are marked with *, and highly significant differences (p < 0.001) are marked with **.

Summary of the Study
Aiming to identify genes in the whole genome of E. coli cells whose expression can induce resistance to H2O2 or HOCl, this study utilized the ASKA library and further validated candidate genes in common E. coli K-12 strains. In total, ~4200 ORFs from the ASKA library were tested individually. Besides some well-known genes such as katG and katE for oxidative stress responses, this study identified a number of genes (105 genes for H2O2

MICs of HOCl (mM)
LB medium M9 minimal Figure 4. MICs of selected genes inactivated in BW25113 cells from the Keio collection. MIC determination was performed by varying the concentration of the oxidizing agents, with 1 mM increments from 1 to 10 mM for both H 2 O 2 and HOCl and 0.25 mM increments from 0 to 1 mM (if MICs were lower than 1 mM) in LB medium or M9 minimal medium. The lowest concentration at which bacteria could not grow was recorded as the corresponding MIC. Each strain was tested in three biological replicates. Significant differences (p < 0.05) are marked with *, and highly significant differences (p < 0.001) are marked with **.

Summary of the Study
Aiming to identify genes in the whole genome of E. coli cells whose expression can induce resistance to H 2 O 2 or HOCl, this study utilized the ASKA library and further validated candidate genes in common E. coli K-12 strains. In total,~4200 ORFs from the ASKA library were tested individually. Besides some well-known genes such as katG and katE for oxidative stress responses, this study identified a number of genes (105 genes for H 2 O 2 and 63 genes for HOCl) which had not been shown to associate with oxidative stress responses before. On the other hand, some well-known response genes such as oxyR for H 2 O 2 responses and hypT for HOCl responses were not identified in our study. To confirm this result, we determined the MICs of MG1655 cells expressing oxyR or hypT individually using the same protocol as that for candidate genes. The results showed that the engineered cells had the same MICs as those of WT MG1655 cells. One possible reason is that these response proteins are activated upon oxidation by oxidizing agents [20,62] and, when they were overexpressed in our screening, the average oxidation stoichiometry decreased below the level necessary for their activation. Thus, our study nicely complements previous oxidative stress studies, providing new information in this field.
We identified 217 candidate genes for H 2 O 2 resistance and 114 candidate genes for HOCl resistance from our genome-wide screening. Only 27 genes were identified in both sets of candidate genes, including agar, agaS, cheW, exbD, fliS, gloA, hyaA, ilvA, leuA, motB, nrfC, prpE, sbmC, slyB, sucC, ubiD, upp, yagM, ybhC, ydeQ, ydfB, ydfD, ydhL, yhaJ, yodA, yoeE, and yrbB. A summary of the current knowledge of these genes is presented in Table 5. Most of them are involved in stress responses, gene expression, membrane transport, and cell mobility. These overlaps indicate shared mechanisms in oxidative stress responses. However, we also found a large number of genes specific for H 2 O 2 or HOCl, indicating distinct mechanisms in oxidative stress responses. Further studies will be implemented to explore the mechanisms of the genes associated with oxidative stress responses.
yodA Metal-binding protein; may function as a periplasmic zinc chaperone delivering zinc to apo-enzymes in this compartment [65]. ydfB Uncharacterized gene. fliS Flagellar biosynthesis protein; substrate-specific chaperones of the flagellar export system [66]. gloA Glyoxalase I; catalyzes the first of two sequential steps in the conversion of methylglyoxal to D-lactate [67]. ydfD A lysis protein encoded by the Qin prophage [68]. slyB Outer membrane lipoprotein [69]. yoeE TonB-dependent receptor plug domain-containing protein; may be regulated by Fur regulon [70]. sbmC DNA gyrase inhibitor; protects cell from DNA damage cause by DNA-bound gyrase [71].

H 2 O 2 -Resistance Genes
Twenty genes were identified to mediate stronger H 2 O 2 resistance (Table 2). Among them, 13 genes encode proteins which function in the cytosol, while the others encode proteins located in cell membranes. Most of these genes have been functionally studied. Besides two well-known catalase genes katG and katE, the functions of other genes are diverse. For instance, appY and mcbR function as transcriptional regulators; leuA, prpE, ilvA, and citG encode enzymes involved in cell metabolism; kefC, metN, sapC, yqhA, and mlaB encode membrane proteins.
Unsurprisingly, katG and katE in E. coli induce stronger H 2 O 2 resistance than other genes. Interestingly, although the strain with katG overexpression showed a higher MIC (15 mM) than the one overexpressing katE (12 mM), the ∆katE strain (MIC 0.5 mM) apeared more sensitive to H 2 O 2 than the ∆katG strain (MIC 2 mM). One possible explanation for this paradox is that a threshold concentration of H 2 O 2 is required for katG expression [76]. Moreover, the katG gene is regulated by the OxyR regulon [77], while the expression of the katE gene is permanently induced in aerobic environment [78]. Thus, katE can quickly protect cells when katG is inactivated.
The gene rho encodes the transcription termination factor Rho, which is responsible for the termination of over half of the transcripts [79] and is related to several important physiological processes in E. coli [42]. It mediates a strong H 2 O 2 resistance (MIC 12 mM) when overexpressed but is not very sensitive to H 2 O 2 when inactivated. This could be explained by the previous finding that the activity of Rho in bacteria could be altered under stressful conditions [80]. Overexpression of rho could compensate the negative effects brought by a dysfunctional Rho under stress conditions.
The transcriptional regulator OxyR has been known as the major regulon for responses to H 2 O 2 stress [14]. In this study, two more transcriptional regulators, AppY and McbR, were also identified. The overexpression of appY and mcbR allowed E. coli cells to grow in 10 mM H 2 O 2 . AppY was found to function as a transcriptional activator of energy metabolism genes under stressful conditions such as anaerobiosis and phosphate starvation [33]. It was reported that a AppY-defective E. coli strain was more sensitive to H 2 O 2 than the wild-type strain [81], which is consistent with our result. On the other hand, McbR has been demonstrated to be involved in H 2 O 2 responses in avian pathogenic E. coli by downregulating the expression of the stress response genes yciF and yciE [82].
In addition to the genes discussed above, yncG, a gene for a putative glutathione Stransferase, was also shown to induce obvious H 2 O 2 resistance in our tests. Although one previous study demonstrated that YncG does not exhibit GSH activity when expressed in cell-free systems [38], YncG may have a different function in vivo, i.e., a GSH-dependent peroxidase activity similar to that of another putative glutathione S-transferase, GST B1-1 [83].

HOCl-Resistance Genes
Twenty-three genes were identified to mediate stronger HOCl resistance (Table 4). Among these genes, 12 genes encode proteins functioning in the cytosol, while the others encode proteins located in membranes. Surprisingly, all of these 23 genes have not been mentioned as parts of any known HOCl response mechanisms. Some of them have been indicated to be activated under stress conditions. For example, ycbB encodes the L,Dtranspeptidase with a role in protecting outer membranes during cell envelope stress [48]; The yedQ gene encodes a probable inner membrane protein with predicted diguanylate cyclase activity [59]. Expression of sanA is implicated in strengthening membrane permeability in stationary-phase stress responses [60]. It is reasonable that these three genes encode proteins in membrane systems, as HOCl has been found to damage the cell envelope system of bacteria [19]. Besides them, we also identified eight genes (marC, ybhR, exbD, yfdY, motB, yoaE, hyaC, and tap), which encode proteins in membrane systems. Among them, yfdY was indicated as a participant in biofilm formation, which is a defense mechanism against HOCl in E. coli [55].
In addition to genes coding for membrane-associated proteins, we also identified genes involved in other biological processes such as metabolism (sucC in the citric acid cycle; prpE in propionate metabolism) [31], amino acid synthesis (ilvA in isoleucine biosynthesis) [41], DNA and RNA modifications (yhdJ for methylation of genomic DNA; rsuA for pseudouridylation of 16S rRNA) [57,61], and rRNA processing (rnc) [54]. One possible mechanism is that overexpression of these proteins could compensate for their corresponding native proteins which are inactivated by oxidation.
Different from the results for H 2 O 2 , MIC determination tests showed no significant differences among the 23 candidate genes for HOCl. Only yoaE, which encodes a putative transport protein, induced slightly stronger HOCl resistance than other genes. Although there is no previous report indicating the role of YoaE in E. coli stress responses, a recent study demonstrated that the expression of the yoaE gene in Salmonella enterica could be upregulated by CpxR, which plays an important role in repairing bacterial envelope damages [84].

Conclusions
In this study, we performed genome-wide screening of the E. coli ASKA collection and identified 217 candidate genes for H 2 O 2 resistance and 114 candidate genes for HOCl resistance. Among them, 105 genes for H 2 O 2 and 63 genes for HOCl were not shown to associate with oxidative stress responses before. Further studies are necessary to validate the genes here identified, which appear as promising new candidates for oxidative stress studies. Furthermore, because the disinfecting mechanisms of many antibiotics are related to oxidative stress, this study is expected to facilitate antibiotic development.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/antiox10060861/s1, Figure S1: High-resolution image of the protein-protein interaction map for H 2 O 2 resistance genes; Figure S2: High-resolution image of the protein-protein interaction map for HOCl resistance genes.
Funding: This research was funded by the National Institutes of Health (R15GM140433 and P20GM139768) and a Honors College Research Grant from University of Arkansas. The APC was funded by the National Institutes of Health.