Introduction

Emerging antibiotic resistance in bacteria is considered a global threat to public health since it can rapidly evolve in response to clinical doses. It has been suggested previously that antibiotic resistance results from SOS response-mediated mutagenesis and horizontal gene transfer in response to antibiotic treatment [1]. When double-stranded DNA is damaged by ultraviolet (UV) irradiation or DNA-damaging agents, it breaks up into single-stranded DNA (ssDNA) [2], the accumulation of which triggers the SOS response through the activation of RecA coprotease, which promotes the autocatalytic cleavage of the LexA repressor and induction of the SOS response genes [3]. In Escherichia coli, more than 40 genes are directly regulated by LexA, including the recombination and repair genes recA, recN, and ruvAB, the nucleotide excision repair genes uvrAB and uvrD, the error-prone DNA polymerase genes dinB encoding Pol IV and umuDC encoding Pol V, and polB encoding DNA polymerase II [4]. The error-prone DNA polymerases induce a hypermutable state that generates genetic diversity to promote the acquisition of antibiotic resistance [5, 6].

RecA is the key regulatory protein involved in the SOS response and DNA repair; therefore, its inactivation leads to increased sensitivity to some antibiotics. Treatment with the fluoroquinolone antibiotic ciprofloxacin has been shown to lead to an increase in recA expression levels as well as activation of the SOS response, as indicated by the induction of error-prone DNA polymerase Pol V [7]. By contrast, deletion of the recA gene from a variety of bacteria causes a two- to eight-fold decrease in the minimum inhibitory concentration (MIC) of levofloxacin, another type of fluoroquinolone [8]. Similarly, the inactivation of RecA reduces the antibiotic-mediated mutagenic effect and increases sensitivity to some antibiotics in E. coli when exposed to sublethal concentrations of antibiotics [9], and strains that lack RecA show significantly increased rates of cell death when treated with quinolones, β-lactams, or aminoglycosides [6]. These bactericidal antibiotics generally induce reactive oxygen species (ROS) formation, resulting in DNA damage and induction of the SOS response [6]. However, Ezraty et al. found that antibiotic exposure did not induce ROS production, with lethality probably resulting from the direct inhibition of cell-wall assembly, protein synthesis, and DNA replication [10], and Ezraty et al. also showed that ROS defense mechanisms are dispensable during treatment with bactericidal antibiotics [10]. Therefore, given the involvement of RecA in DNA repair, the SOS response, horizontal gene transfer [11], the control of swarming ability [12], and biofilm formation [13], it is considered a promising target for the development of a new strategy for preventing antibiotic resistance [14]. However, the impact of RecA inactivation on various kinds of antibiotic sensitivities is not fully understood yet. Therefore, in this study, we attempted to investigate the impact of recA deletion on sensitivities to antibiotics and chemicals with varying drug actions in the wild-type MDS42 strain and the ΔrecA mutant strain of E. coli. To do this, we determined the MIC of each chemical using a previously developed automated culture system for laboratory evolution that allows us to maintain more than 100-independent culture series under various culture conditions [15].

We found that the ΔrecA mutation increased the sensitivities to not only SOS response-inducing agents but also S-(2-aminoethyl)-l-cysteine, l-histidine, ruthenium red, d-penicillamine, carbonyl cyanide 3-chlorophenylhydrazone (CCCP), cerulenin, and l-cysteine. Furthermore, microarray analysis showed that the ΔrecA mutation resulted in decreased expression of nitrogen assimilation regulators as well as amtB, which encodes an ammonium transporter. Together, these results suggest that the inactivation of RecA affects not only the SOS response but also some metabolic stresses, thereby potentiating several antibiotic activities.

Materials and methods

Bacterial strains and growth media

The insertion sequence (IS)-free E. coli strain MDS42 [16] was purchased from Scarab Genomics (Scarab Genomics, Madison, Wisconsin, USA) and used as the wild-type strain. The MDS42 ΔrecA mutant strain was constructed using the Quick and Easy E. coli Gene Deletion Kit (Gene Bridges, Heidelberg, Germany). Deletion of the open reading frame region of recA was performed in accordance with the manufacturer’s instructions. Briefly, the FRT-cmr-FRT cassette (A006; Gene Bridges) was amplified by polymerase chain reaction (PCR) using the primer pairs Primer1/Primer2 and Primer3/Primer4 (see Supplementary Table S1) to add 50-bp-long homology arms that corresponded to the sequences flanking the insertion site on the chromosome. The constructed functional cassette was electroporated into the parental strain and inserted into the target locus by λ-Red recombinase using the expression plasmid pRedET. Then, the selection marker was removed from the chromosome by FLPe recombinase using the 709-FLPe expression plasmid (A106; Gene Bridges). The genome modifications and removal of the selection marker cassette were verified by colony PCR using the primer pair Primer5/Primer6 (Table S1) and were then confirmed by direct Sanger sequencing of the PCR products.

The E. coli strains were cultured in modified M9 minimal medium containing 17.1 g l−1 Na2HPO4·12H2O, 3.0 g l−1 KH2PO4, 5.0 g l−1 NaCl, 2.0 g l−1 NH4Cl, 5.0 g l−1 glucose, 14.7 mg l−1 CaCl2·2H2O, 123.0 mg l−1 MgSO4·7H2O, 2.8 mg l−1 FeSO4·7H2O, and 10.0 mg l−1 thiamine hydrochloride (pH 7.0) [17].

Determination of minimum inhibitory concentrations (MICs)

To comprehensively investigate the effect of the ΔrecA mutation on the susceptibility to different chemicals, we determined the MICs of 217 chemicals with differing mechanisms of action in the wild-type MDS42 strain and the ΔrecA mutant strain. The selected chemicals included a wide range of antibiotics, such as β-lactams, penicillins, cephalosporins, aminoglycoside, tetracyclines, macrolides, amphenicols, rifamycins, and quinolones. In addition, a number of toxic chemicals other than antibiotics were tested, such as sulfonamides, organic acids, amino acid analogs, metals, chelators, cationic surfactants, and receptor inhibitors. The chemicals we tested have a wide range of biological targets, such as peptidoglycan synthesis, the cell membrane, polysaccharides, 30S ribosomes, 50S ribosomes, protein translation, RNA polymerase, DNA, DNA gyrase, folic acid biosynthesis, metabolic enzymes, oxidative phosphorylation, fatty acid biosynthesis, oxidative stress, metal chelators, acid stress, and osmotic balance. Figure 1 shows the number of chemicals for each biological target.

Fig. 1
figure 1

Biological targets of the chemical compounds used in this study. The proportion (and number) of the 217 chemical compounds tested that fell into each biological target category is shown

Supplementary Table S2 lists all of the chemicals that were used in this study and the solvents in which they were dissolved to prepare stock solutions. Chemicals that were not dissolved in the modified M9 medium were added to it at a >20-fold dilution. Cell cultivation, optical density (OD) measurements, and serial dilutions were performed for each chemical using an automated culture system [15] consisting of a Biomek® NX span-8 laboratory automation workstation (Beckman Coulter, Brea, California, USA) in a clean booth connected to a microplate reader (FilterMax F5; Molecular Devices, San Jose, California, USA), a shaker incubator (STX44; Liconic, Mauren, Liechtenstein), and a microplate hotel (LPX220, Liconic). Serial dilutions of each chemical were prepared in 384-well microplates using the modified M9 medium with doubling dilution steps to determine MICs.

MDS42 and the ΔrecA mutant cells were inoculated from the frozen glycerol stock into the modified M9 medium and cultivated overnight at 34 °C and 150 rpm. The OD620 values of the precultures were measured using the automated culture system, and those precultured cells that were calculated to have initial OD620 values of 0.0003 were inoculated into each well (5 μl of diluted overnight culture into 45 μl of medium per well) of 384-well microplates containing serially diluted chemicals to a final volume of 50 µl. After 24 h incubation at 34 °C with agitation at 300 rotations/min, the OD620 of the precultures was measured again. The MIC was defined as the lowest concentration of a chemical that reduced the growth to OD620 < 0.09.

Total RNA purification

Total cellular RNA was isolated from the MDS42 and ΔrecA mutant cells as follows. Cells were inoculated from the frozen glycerol stock into 5 ml of modified M9 medium in test tubes and cultivated overnight at 34 °C and 150 rpm. Then, the overnight cultures were diluted to an OD600 of 0.1 in fresh modified M9 medium in test tubes and cultured. Two volumes of RNAprotect Bacteria Reagent (Qiagen, Hilden, Germany) were added directly to one volume of exponentially growing cultures (OD600 of ~1; mid-exponential growth phase) to stabilize the cellular RNA. Then, the cells were harvested by centrifugation at 5000 × g for 10 min at 25 °C and total cellular RNA was isolated using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. Total RNA was treated with DNase I at room temperature for 15 min and the purified RNA samples were stored at −80 °C until the microarray experiments.

Transcriptome analysis using microarrays

Microarray experiments were performed as described previously [18] using a custom-designed Agilent 8 × 60 K array for E. coli W3110 that included 12 probes for each gene. Briefly, 100 ng of each purified total RNA sample was labeled using the Low Input Quick Amp WT Labeling Kit (Agilent Technologies, Santa Clara, California, USA) with Cyanine3 (Cy3) according to the manufacturer’s instructions. Cy3-labeled cRNAs were fragmented and hybridized to the microarray for 17 h at 65 °C in a hybridization oven (Agilent Technologies), following which the microarray was washed and scanned according to the manufacturer’s instructions. Microarray image analysis was performed using Feature Extraction version 10.7.3.1 (Agilent Technologies) and expression levels were normalized using the quantile normalization method [19]. To exclude quantitatively unreliable data, genes with low expression levels [<1.5 AU (log base 10)] were excluded from the subsequent analysis (representing ~35% of genes). Each experiment was performed in triplicate, starting from independent cultures. The microarray data have been submitted to the National Center for Biotechnology Information’s Gene Expression Omnibus functional genomics data repository under accession number GSE121218.

Results

Effect of the ΔrecA mutation on chemical susceptibility

To reveal the impact of RecA inactivation on antibiotic sensitivities, we attempted to identify chemicals that strongly inhibit the growth of bacteria lacking the recA gene. First, the effect of ΔrecA mutation on cell growth was examined (Fig. S1). The ΔrecA mutant strain only showed a slight decrease in growth compared to the wild-type strain. The growth rates at 34 °C in the modified M9 medium were 0.52 ± 0.01 h−1 in the wild-type strain and 0.50 ± 0.02 h−1 in the ΔrecA mutant strain. Table S2 shows the calculated MIC values for each of the 217 chemicals tested. The ΔrecA mutant strain showed at least twofold decreases in the MICs for 28 of the 217 chemicals compared with the wild-type strain (Table S2). Table 1 shows the MICs of these 28 chemicals for both strains.

Table 1 Minimum inhibitory concentration (MIC; µg ml−1) of each chemical compound tested for which the ΔrecA mutant strain was more sensitive than the MDS42 wild-type strain

We found that the MICs of all seven of the quinolones tested were at least twofold lower for the ΔrecA mutant than for the wild-type strain (Table 1). The ΔrecA mutant also showed decreased MICs for many chemicals that cause DNA damage, i.e., acriflavine, mitomycin C, hydroxyurea, phleomycin, 4-nitroquinoline 1-oxide, ornidazole, and tinidazole (Table 1). However, it should be noted that the ΔrecA mutant did not show decreased MICs for semicarbazide, netropsin, N-methyl-N-nitrosourea, bleomycin, or 9-aminoacridine, all of which also cause DNA damage (Table S2).

The ΔrecA mutant also showed decreased MICs for chemicals that inhibit nucleotide biosynthesis, i.e., 5-azacytidine, 5′-fluoro-5′-deoxyuridine, and azathioprine, although the MICs for cytosine β-d -arabinofuranoside, 6-mercaptopurine, and 3,5-diamino-1,2,4-triazole, which inhibit nucleotide biosynthesis, and for the sulfonamides, which inhibit folic acid biosynthesis (which is a precursor of nucleotide biosynthesis), were comparable between the two strains (Table S2). In addition, the MICs for nitrofurantoin and furaltadone, which target many macromolecules, including DNA and ribosomes, were 4- and 32-fold lower, respectively, in the ΔrecA mutant than in the wild-type strain, and the ΔrecA mutant showed increased sensitives to sodium dichromate, which causes oxidative DNA damage (Table 1).

We also found that the ΔrecA mutation caused increased sensitivities to ruthenium red (an ion channel and Ca2+ binding protein inhibitor), d-penicillamine (a metal chelator), l-amino acids (l-histidine and l-cysteine), and some metabolic inhibitors, i.e., S-(2-aminoethyl)-l-cysteine (a lysine analog), cerulenin (a fatty acid synthesis inhibitor), and carbonyl cyanide m-chlorophenyl hydrazine (CCCP; an oxidative phosphorylation inhibitor) (Table 1). Among the 217 chemicals tested, S-(2-aminoethyl)-l-cysteine was most effective in inhibiting the growth of the ΔrecA mutant, with a 64-fold lower MIC for this strain compared with the wild-type strain (Table 1).

Effect of the ΔrecA mutation on global gene expression

Since our MIC measurements revealed that the ΔrecA mutation resulted in increased sensitivities to chemicals in addition to DNA-damaging agents and DNA synthesis inhibitors, we compared the transcriptomes of the ΔrecA mutant and wild-type strains using DNA microarray analysis. Overall, 19 genes showed ≥2.0-fold increases and 14 genes showed ≥2.0-fold decreases in mRNA levels in the ΔrecA mutant compared with the wild type. These genes and their annotated functions and mRNA ratios are listed in Table 2. The ΔrecA mutant showed significantly decreased expression of nitrogen regulator genes, including a 7.8-fold decrease in glnK, which encodes a nitrogen assimilation regulator; a 4.3-fold decrease in nac, which encodes a nitrogen assimilation control gene; a 3.6-fold decrease in amtB, which encodes an ammonium transporter; a 2.7-fold decrease in glnLG, which encodes a two-component regulatory system for nitrogen regulation; and a 2.1-fold decrease in glnA, which encodes glutamine synthetase. The ΔrecA mutant also showed decreased expressions of recN, which encodes a recombination and repair protein, and sulA, which encodes an SOS cell division inhibitor. By contrast, the ΔrecA mutant had increased expressions of several genes that encode chaperone proteins, including a 3.9-fold increase in ibpB, a 2.5-fold increase in ibpA, a 2.4-fold increase in clpB, and a 2.0-fold increase in htpG.

Table 2 Effect of the ΔrecA mutation on the expression of Escherichia coli genes

Discussion

In this study, we determined MICs for 217 chemicals with a wide range of biological targets in the wild-type and ΔrecA mutant strains of E. coli. We found that the ΔrecA mutant showed increased sensitivities to various chemicals, including not only DNA-damaging agents but also some metabolic inhibitors, such as the lysine analog S-(2-aminoethyl)-l-cysteine and l-histidine. To understand the effect of recA deletion on global gene expression better, we also performed transcriptome analysis, which showed that overall deletion of the recA gene did not affect the global transcriptome profile. However, the ΔrecA mutant strain did show decreased expressions of nitrogen assimilation regulators and their regulons, amtB, which encodes an ammonium transporter, and glnA, which encodes glutamine synthetase.

Previous studies have shown that deletion of the recA gene in a number of bacteria results in increased sensitivities to quinolones, which inhibit DNA gyrase [8, 9]. Similarly, in this study, we found that the recA mutant showed two- to eight-fold lower MICs for all of the tested quinolones compared with the wild type (Table 1). In addition, we found that deletion of the recA gene resulted in decreased MICs for many DNA-damaging agents and sodium dichromate. Since both double-strand breaks in the DNA caused by DNA-damaging agents, such as mitomycin C and nitroimidazole derivatives, and chromate shock induce the SOS response [20], the observation of decreased MICs for these chemicals in the ΔrecA mutant is reasonable. A previous study showed that treatment of an E. coli ΔrecA mutant with UV irradiation and 4-nitroquinoline-1-oxide treatment resulted in rapid chromosome degradation and cell death [21], indicating that the SOS response and/or RecA-dependent recombination are required for the prevention of DNA degradation and cell death. It has also been shown that alterations to the deoxyribonucleoside triphosphate (dNTP) pool caused by the deletion of either ndk, which encodes nucleoside diphosphate kinase, or dcd, which encodes deoxycytidine triphosphate (dCTP) deaminase, result in hypermutability and error catastrophe [22,23,24]. In the present study, we found that the ΔrecA mutant showed increased sensitivities to the nucleic acid synthesis inhibitors 5-azacytidine (a cytidine analog) and 5′-fluoro-5′-deoxyuridine (a pyrimidine analog) (Table 1). Therefore, since dNTP pool imbalances lead to an increased production of mispairing errors and reduced exonucleolytic proofreading of any mispairings [24], treatment with these inhibitors may be more deleterious to the ΔrecA mutant.

Bactericidal drugs such as quinolones, β-lactams, and aminoglycosides have been shown to induce hydroxyl radical formation in a ΔrecA strain, causing DNA damage and cell death, and thereby potentiating the killing efficiency [6]. Furthermore, Thi et al. [9] found that deletion of the recA gene results in increased sensitivities to some bactericidal antibiotics, such as ceftazidime (cephalosporin), fosfomycin (phosphonic antibiotic inhibiting peptidoglycan synthesis), trimethoprim/sulfamethoxazole, and colistin, although there was no evidence of increased sensitivities to other bactericidal drugs, such as ampicillin (β-lactam) and gentamicin (aminoglycoside). However, none of these bactericidal drugs produced ROS, indicating that the ROS defense mechanisms are dispensable during treatment with these drugs [10, 25]. Therefore, the effect of recA inactivation on susceptibilities to bactericidal drugs remains unclear. In the present study, we did not observe any potentiating effects of β-lactams or aminoglycosides on growth inhibition of the ΔrecA strain (Table S2). These differences may have resulted from differences in the media used for the determination of MICs. Previous studies showed that the growth arrest of E. coli cells caused by nutrient limitation triggers the production of oxidative stress proteins e.g. catalases which result in the resistance to hydrogen peroxide [26, 27]. We used the modified M9 minimal medium while Kohanski et al. [6] and Thi et al. [9] used nutrient-rich medium (LB medium). Therefore, the effect of ΔrecA mutation on susceptibilities to β-lactams and aminoglycosides may be suppressed in this study.

We found that the ΔrecA mutant also showed increased sensitivities to chemicals other than DNA-damaging agents, such as a lysine analog [S-(2-aminoethyl)-l-cysteine], l-amino acids (l-histidine and l-cysteine), an ion channel and Ca2+ binding protein inhibitor (ruthenium red), a metal chelator (d-penicillamine), an uncoupling agent that inhibits oxidative phosphorylation (CCCP), and a fatty acid biosynthesis inhibitor (cerulenin). To understand how the absence of recA increases the sensitivities to these chemicals better, we performed microarray analysis. This showed that the ΔrecA mutant had significantly lower expressions of the glnALG operon, which encodes glutamine synthetase (GS), the nitrogen regulatory sensor kinase and response regulator nac, which encodes nitrogen assimilation control, and the glnK-amtB operon, which encodes a nitrogen regulator and ammonium transporter (Table 2). In E. coli, the master regulator of the nitrogen-limited stress response is the NtrBC two-component system, which is encoded by the glnGL genes [28]. Under nitrogen-limited conditions, the nitrogen regulator GlnK and the nitrogen assimilation control protein Nac also activate genes that are involved in nitrogen assimilation [29,30,31,32]. However, the ΔrecA mutant showed decreased expressions of genes that are important for nitrogen assimilation (i.e., glnALG, nac, glnK, and amtB), meaning that the biosynthesis of glutamine and other amino acids may be limited. Interestingly, the ΔrecA mutant also showed increased sensitivity to l-histidine, which is known to inhibit GS encoded by the glnA gene [33]. However, the ΔrecA mutant did not show decreased MICs for l-methionine sulfoximine, bialaphos, or glufosinate, all of which are known as GS inhibitors [34] (Table S2). It was shown that NtrC couples the stringent response and NtrC regulates not only nitrogen assimilation related genes but also argT encoding lysine/arginine/ornithine ABC transporter and hisJQMP encoding histidine ABC transporter in E. coli [35]. Although these gene expressions were not significantly changed in the ΔrecA mutant, decreased expression of glnG in the ΔrecA mutant may also affect lysine and histidine metabolism. Since, we used the minimal medium for MIC measurements, such a deficiency in amino acid biosynthesis may result in an increased sensitivity to the lysine analog, l-histidine, and l-cysteine. These results suggest that deletion of the recA gene results in a limitation of amino acid biosynthesis.

Since bacteria easily evolve antibiotic resistance in response to clinical doses, new strategies for preventing antibiotic resistance are required for public health. One such strategy is drug combination therapy [36]. In some cases, combinations of drugs act synergistically, causing them to kill pathogens more efficiently and, thus, suppressing antibiotic resistance [37, 38]. RecA inhibitors are considered promising for the prevention of antibiotic resistance [14]. Suramin (polysulphonated naphthylurea) is considered a potent and selective inhibitor of RecA and the SOS response in Mycobacterium tuberculosis [39]. In addition, phthalocyanine tetrasulfonic acid (PcTs)-based RecA inhibitors are known to block ATPase, DNA binding, DNA strand exchange, and LexA proteolysis activities of RecA, potentiating the activity of quinolone, β-lactam, and aminoglycoside family antibiotics and reducing the ability of bacteria to acquire antibiotic resistance [14]. Therefore, an additional dosage of a drug that shows synergistic activity to the RecA inhibitor could potentiate the inhibition of antibiotic resistance. In this study, we identified a number of chemical compounds that effectively inhibited the growth of the recA deletion mutant, which may contribute to the development of synergistic combination therapy using RecA inhibitors.