Introduction

Addressing bacterial infections caused by emerging and drug-resistant pathogens represents a major global health priority. Bactericidal antibiotics can exert their effects on cells by directly or indirectly causing DNA damage or triggering the production of highly destructive hydroxyl radicals (1–3). This, in turn, initiates a protective mechanism known as the SOS response, which enables bacterial survival against the lethal impacts of antibiotics by activating intrinsic pathways for DNA repair (4–7). The activation of DNA repair processes relies on specific genes, such as recA, which encodes a recombinase involved in DNA repair, and lexA, a repressor of the SOS response that can be inactivated by RecA (8).

Studies have demonstrated that a single exposure to fluoroquinolones, a type of antibiotic that induces DNA breaks and triggers the SOS response, leads to the development of bacterial resistance in Escherichia coli (E. coli) through a RecA and SOS response-dependent mechanism (9). Given the crucial role of RecA in the SOS response, inhibiting RecA activity to deactivate the SOS response presents an appealing strategy for preventing the evolution of bacterial resistance to antibiotics (10). Similarly, exposure to fluoroquinolones induces the SOS response and mutagenesis in Pseudomonas aeruginosa, and the deletion of recA in this pathogen results in a significant reduction in resistance to fluoroquinolones (11).

Unlike fluoroquinolones, β-lactam antibiotics induce a RecA-dependent SOS response in E. coli through impaired cell wall synthesis, mediated by the DpiBA two-component signal system (12). The development of antibiotic resistance, triggered by exposure to β-lactams, has been extensively investigated using the cyclic adaptive laboratory evolution (ALE) method. Mutations that arise during cyclic ALE experiments are attributed to errors occurring during continued growth, necessitating multiple rounds of β-lactam exposure to drive the evolution of resistance in E. coli cells (13,14). However, the precise roles of RecA and SOS responses in the development of resistance under short-term β-lactam antibiotics exposure remain unclear.

Recently, there has been a growing interest in understanding the impact of the stress-induced accumulation of reactive oxygen species (ROS) on bacterial cells (15,16). While exploring methods to harness ROS-mediated killing has the potential to enhance the effectiveness of various antibiotics (17–19), the role of ROS in antimicrobial activity has become a topic of controversy following challenges to the initial observations (20,21). The generation of ROS has been found to contribute to the development of multidrug resistance, as prolonged exposure to antibiotics in cyclic ALE experiments is known to generate ROS, leading to DNA damage and increased mutagenesis (22,23). Nevertheless, there is still limited knowledge regarding the consequences of ROS accumulation in bacteria when the activity of RecA or the SOS response is suppressed. Here, in the present work, we observed a fast evolution of multi-drug resistance in an E. coli strain with the recA mutation, following a single exposure to β-lactam antibiotics. This evolution can be attributed to three mechanisms: compromised DNA repair, transcriptional repression of genes associated with antioxidative processes, and the excessive accumulation of ROS.

Results

Single exposures to β-lactam antibiotics trigger a fast evolution of antibiotic resistance in the recA mutant strain

To investigate the impact of the SOS response on bacterial evolution towards β-lactam resistance, we generated a recA mutant strain (ΔrecA) from the E. coli MG1655 strain (LZ101). Initially, we conducted an ALE experiment using a slightly modified treatment protocol (24) on the wild type and ΔrecA strains. During a period of three weeks, the cells were subjected to cycles of ampicillin exposure for 4.5 hours at a concentration of 50 µg/mL (10 times the MIC) each day (Fig. S1A) (25). As anticipated based on previous studies, the intermittent ampicillin treatments over the course of three weeks resulted in the evolution of antibiotic resistance in the wild type strain (Fig. S1B). However, a significantly accelerated development of resistance was found in the ΔrecA strain, with the average time to resistance reduced to 2 days (Fig. S1B). More importantly, we observed that resistance even emerged in the ΔrecA strain after a single exposure for 8 hours to ampicillin (Fig. 1A-C). Meanwhile, after 8 hours of treatment with 50 μg/mL ampicillin, the survival rates of both wild type and ΔrecA strain were consistent (Fig. S2). To ensure that the emergence of resistance we observed was not illusory due to technical issues during the recA knockout process, we employed another ΔrecA strain (JW2669-1) provided by the Coli Genetic Stock Centre (CGSC) with the same killing procedure. The results from both bacterial strains were consistent (Fig. S3A and B). To further investigate, we treated both the wild type and ΔrecA cells with other β-lactams, including penicillin G and carbenicillin, at concentrations equivalent to 10 times the MIC (1 mg/mL and 200 µg/mL, respectively) for 8 hours (26,27). Consistently, these treatments also led to a fast evolution of antibiotic resistance in the ΔrecA strain (Fig. S4A and B).

Figure 1.

To assess the stability of this accelerated antibiotic resistance acquired by the ΔrecA strain, we conducted a study wherein the ΔrecA resistant isolates, originating from the initial 8-hour treatment with ampicillin, were continuously cultivated in a medium devoid of antibiotics for a period of seven days. Our findings revealed that once resistance was established, resistance remained stable and was able to be passed on to subsequent generations even in the absence of ampicillin (Fig. 1D). Moreover, we performed a complementation experiment by introducing a plasmid containing recA under its native promoter into the ΔrecA strain prior to Step i in Fig. 1A, that is, before exposing the cells to ampicillin. Interestingly, this complemented strain displayed a comparable MIC to the isogenic wild type strain and maintained its sensitivity even after ampicillin treatment for up to 8 hours (Fig. 1E).

Notably, the evolution of antibiotics is influenced by mutation and selection, among other factors. To further investigate the primary drivers behind the rapid development of antibiotic resistance, we employed rifamycin, an unrelated antibiotic. Our findings indicated that following an 8-hour exposure to ampicillin, the MIC of rifampicin reached about 100 µg/mL for the ΔrecA strain, marking a 10-fold increase compared to the wild-type strain (Fig. 1F). Moreover, we measured and calculated the mutation frequency by employing rifampicin as the selective agent in both the wild type and ΔrecA strains after the 8-hour ampicillin exposure. This was done by dividing the number of colony-forming units (CFU) per milliliter on rifampicin plates by the number of CFU per milliliter on LB plates (28). Our results revealed a significantly higher mutation frequency in the ΔrecA strain in contrast to the wild type (Fig. 1G). These findings together demonstrate that a singular exposure to β-lactam antibiotics markedly boosts mutation frequency and thereby triggers a fast evolution of antibiotic resistance that is stable and heritable in the ΔrecA strain.

Rapid appearance of drug resistance associated DNA mutations in recA mutant resistant isolates

In bacteria, resistance to most antibiotics requires the accumulation of drug resistance associated DNA mutations developed over time to provide high levels of resistance (29). To verify whether drug resistance associated DNA mutations have led to the rapid development of antibiotic resistance in recA mutant strain, we randomly selected 15 colonies on non-selected LB agar plates from the wild type surviving isolates, and antibiotic screening plates containing 50 µg/mL ampicillin from the ΔrecA resistant isolates, respectively, and performed whole-genome sequencing. We found that abundant drug resistance associated mutations occurred within all resistant isolates, including the promoter of the β-lactamase ampC (P_ampC) in 8 isolates, the ampicillin-binding target PBP3 (ftsI) in 1 isolate, and the AcrAB-TolC subunit AcrB (acrB) mutations in 6 isolates (Fig. 2A). A mutation in gene nudG was detected in wild type surviving isolates after the single exposure to ampicillin (Fig. 2A), which is involved in pyrimidine (d)NTP hydrolysis to avoid DNA damage (30). Other mutations were listed in the Table S1.

Figure 2.

The presence of P_ampC mutations was accompanied by a significant increase in the production of β-lactamase in bacteria (Fig. S5). This leads to specific resistance to β-lactam antibiotics. The gene acrB codes for a sub-component of the AcrAB-TolC multi-drug efflux pump, which is central in Gram-negative bacteria (28,31). Mutations in AcrAB-TolC enhance the efflux of antibiotics and confer resistance to multiple drugs (28). Consequently, after short-term exposure to ampicillin, the ΔrecA isolates carrying the acrB mutations exhibited resistance to other types of antibiotics, such as chloramphenicol and kanamycin (Fig. 2B and C). Treatment with high concentrations of 1-(1-Naphthylmethyl) piperazine (NMP), an efflux pump inhibitor (EPI) that competitively blocks TolC-composed efflux pumps, successfully restored the sensitivity of ΔrecA resistant isolates to ampicillin, bringing them back to the equivalent concentrations found in the wild type (Fig. 2D). Collectively, these results suggest that a single exposure to β-lactam antibiotics can lead to drug resistance associated DNA mutations that play a crucial role in the accelerated development of antibiotic resistance in the ΔrecA strain.

Hindrance of SOS-independent DNA repair in recA mutant resistant isolates

The impaired ability to repair DNA damage can lead to an accelerated accumulation of DNA mutations. The cellular response to DNA damage, such as the SOS response, often influences the balance between bacterial susceptibility and resistance. We thereafter asked whether the absence of recA and subsequently induced inhibition of the SOS response can contribute to the rapid development of antibiotic resistance.

To investigate this, we first tested the ability of various mutants involved in different pathways of the SOS response to evolving antibiotic resistance following a single treatment with ampicillin for 8 hours at 50 µg/mL. A mutant form of the SOS master regulator LexA (lexA3), which is incapable of being cleaved and thus defective in SOS induction, did not exhibit antibiotic resistance evolution (Fig. 3A). Additionally, the deletion of either DpiB or DpiA (encoded by citB or citA, respectively), inhibiting the DpiBA two-component signal transduction system, did not result in the development of antibiotic resistance after ampicillin exposure (Fig. 3A). Moreover, the deletion of several downstream effectors of the SOS response, including those involved in cell division inhibition (SulA and YmfM encoded by sulA and ymfM) (32) (Fig. 3A), and DNA repair (DNA Pol II, DNA Pol IV, and DNA Pol V encoded by polB, dinB and umuDC) (33) also did not lead to the evolution of antibiotic resistance (Fig. 3A). Interestingly, these findings indicate that the fast evolution of antibiotic resistance is supposed to occur in an SOS-independent manner in the absence of recA.

Figure 3.

In addition to the DNA repair components associated with the SOS response, DNA Pol I plays a role in processing RNA primers during lagging-strand synthesis and filling small gaps during DNA repair reactions (34). Since DNA Pol I (encoded by polA) has been demonstrated as an essential gene required for the growth of E. coli in rich medium, including the LB medium (35,36), we next utilized Single Molecule Localization Microscopy (SMLM) to precisely locate the chromosome and DNA Pol I in a dynamic manner. During an 8 hour exposure to ampicillin, we observed the formation of multinucleated filaments in both the wild type and ΔrecA strain, indicating a pause in cell division and suggesting a time period for bacterial DNA repair to take place (Fig. 3B and C) (37). However, the expression level of DNA Pol I was significantly suppressed in the ΔrecA strain compared to the wild type strain after 4 hours of ampicillin exposure (Fig. 3D). More notably, the super-resolution colocalization analysis revealed a significantly lower ratio of co-localization between the chromosome and DNA Pol I in the ΔrecA strain (Fig. 3E), which together demonstrate that the hindrance of DNA repair caused by the absence of recA is crucial for the fast evolution of antibiotic resistance in an SOS-independent manner.

Repression on antioxidative-related gene transcription drives the fast evolution of antibiotic resistance in the recA mutant strain

To further comprehend the fast evolution of β-lactam resistance that was observed in this study, we investigated the changes in gene transcription induced by ampicillin treatment using a comprehensive transcriptome sequencing approach (total RNA-seq). Our analysis revealed significant alterations in the transcriptomic profiles of both the wild type and ΔrecA strain isolates following a single treatment with ampicillin, compared to untreated controls (Fig. 4A). Specifically, we identified changes in the expression of 161 and 248 coding sequences (with log₂FC > 2 and P value < 0.05) in the wild type and ΔrecA strains, respectively. Principal component analysis (PCA) demonstrated a notable disparity in the effects of ampicillin on the ΔrecA strain compared to the wild type strain (Fig. 4B). Additionally, a Venn diagram analysis confirmed that 138 and 225 genes were uniquely regulated by ampicillin exposure in the wild type and ΔrecA strains, respectively (Fig. 4C).

Figure 4.

To elucidate the differential expression of genes associated with specific biological functions, we conducted Gene Ontology (GO) enrichment analyses (Fig. S6A and B) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses (Fig. S6C and D). Our findings indicate that ampicillin profoundly impacted persistence pathways in the wild type strain, specifically affecting pathways related to quorum sensing, flagellar assembly, biofilm formation, and bacterial chemotaxis (38,39). Conversely, in the ΔrecA strain, a distinct functional category associated with the oxidative stress response exhibited significant and unique down-regulation. This category included activities such as sulfate transporter activity, iron-sulfur cluster assembly, oxidoreductase activity, and carboxylate reductase activity.

To identify specific genes showing significant fold changes (log₂FC > 2 and P value < 0.05), we used volcano plots to visualize the comprehensive changes in gene expression across the genome (Fig. 4D). We examined the transcription levels associated with the SOS response system and found that the transcription of several proteins in the wild type strain can be significantly induced by the single exposure to ampicillin, including citB and dinB (Fig. 4E). However, in the recA mutant strain, antibiotic exposure does not affect the transcription levels of any SOS system-related proteins, suggesting that antibiotic exposure induced the SOS response in the wild type strain but not in the ΔrecA strain. More importantly, we discovered that the induction of the transcription level of the DNA Pol I was significantly suppressed after the single treatment of ampicillin in the ΔrecA strain compared with that in the wild type strain (Fig. 4E). This is consistent with our imaging results and further supports the notion that an SOS-independent evolutionary mechanism dominates the development of antibiotic resistance in recA mutant E. coli.

Further, significant downregulation in the transcription of antioxidative-related genes in the ΔrecA strain was detected, including cysJ, cysI, cysH, soda, and sufD (Fig. 4F). This downregulation suggested an excessive accumulation of reactive oxygen species (ROS) due to compromised cell antioxidative defences. It has been previously reported that the induction of mutagenesis can be stimulated by the overproduction of ROS during antibiotic administration, leading to the evolution of antibiotic resistance both in vivo and in vitro (22,23). Therefore, we hypothesized that elevated ROS levels might be responsible for the observed fast evolution of antibiotic resistance in the ΔrecA strain. To test it, we first examined the level of ROS generation in the wild type and ΔrecA strains treated with ampicillin for 8 hours by using the fluorescent probe DCFDA/H2DCFDA. We found that ROS levels significantly increased in both the wild-type and ΔrecA strain after 8 hours of ampicillin treatment. However, ROS levels in the ΔrecA strain showed a significant further increase compared to the wild-type strain (Fig. 4G). Additionally, with the addition of 50 mM glutathione (GSH), a natural antioxidative compound, no significant change in ROS levels was observed in either the wild-type or ΔrecA strain before and after ampicillin treatment (Fig. 4H). Further, we supplemented the wild type and ΔrecA strains with 50 mM GSH) and treated them with ampicillin at a concentration of 50 μg/ml for 8 hours. Remarkably, the addition of GSH prevented the development of resistance to ampicillin in the ΔrecA strain (Fig. 4I), without impairing the bactericidal effectiveness of ampicillin (Fig. 4J).

Apart from the SOS response, bacterial cells coordinate other DNA repair activities through a network of regulatory pathways, including base excision repair (BER) (40–42). The excessive generation of ROS results in elevated levels of deoxy-8-oxo-guanosine triphosphate (8-oxo-dGTP), an oxidized form of dGTP that becomes both highly toxic and mutagenic upon integration into DNA. The presence of 8-oxo-dG can induce SNP mutations, especially those occurring in guanine, which can be actively rectified by the BER repair pathway (42). Notably, BER glycosylases MutH and MutY can identify and repair these 8-oxo-dG-dependent mutations; however, when MutY and MutH are inactivated, unrepaired 8-oxo-dG can lead to the accumulation of SNP mutations within cells (42). As a result, we conducted a further assessment of the transcription levels of the BER repair pathway in both the wild-type and ΔrecA strain before and after a single exposure to ampicillin. We discovered that after an 8- hour treatment of ampicillin, three DNA repair-associated proteins, including MutH, MutY, and MutM, were notably suppressed in the ΔrecA strain (Fig. 4K).

Finally, we sequenced the surviving ΔrecA isolates and found that the addition of GSH inhibited drug resistance associated mutations in the ΔrecA strain, which was detected in the ΔrecA resistant isolates including genes of the promoter of ampC and acrB (Fig. 4L). Given the DNA repair impairment resulting in the generation of ROS, it would have been expected for genes involved in the oxidative stress response to be induced in RecA-deficient cells. However, the repression of antioxidative-related genes indicated the involvement of transcriptional repressors that might be regulated by RecA. Consequently, we examined the transcription levels of all transcriptional repressors in both the wild type and ΔrecA strain. Remarkably, we observed a significant upregulation of H-NS, a crucial transcriptional repressor (Fig. 4M). This finding suggests that the upregulation of H-NS could potentially account for the downregulation of antioxidant gene transcription levels in the ΔrecA strain.

Discussion

In this study, we present a novel perspective challenging the conventional belief that disabling RecA to inhibit the SOS response can prevent bacteria from developing antibiotic resistance (10). While the SOS response and RecA have been extensively studied for their roles in antibiotic resistance evolution (43), we observed a fast evolution of multi-drug resistance in the ΔrecA strain triggered by single exposures to β-lactam antibiotics, which was driven by three mechanisms, including the impaired DNA repair, transcriptional repression of antioxidative-related genes, and excessive accumulation of ROS (Fig. 5).

Figure 5.

We initially suspected that the decrease in DNA repair levels was caused by the suppression of the SOS response, mediated by the deficiency of RecA during rapid resistance development. We therefore examined the response to antibiotic exposure in various E. coli mutants lacking specific genes from the SOS response. However, we discovered that these strains remained susceptible after short-term exposure to ampicillin, providing evidence that the decrease in DNA repair levels, regulated by RecA in an SOS-independent manner, served as a mechanism for the accelerated development of antibiotic resistance. Furthermore, we conducted super-resolution imaging studies to analyze the expression and localization of DNA Pol I. We observed a significant inhibition of DNA Pol I induction and its involvement in DNA repair. Moreover, a notable suppression of the BER repair system was observed in the recA mutant bacteria upon short-term exposure to antibiotics. While we do not yet have evidence explaining why the absence of recA might affect the transcriptional regulation of the BER repair system, this evidence is sufficient to demonstrate that the absence of RecA can lead to the accumulation of beneficial mutations within bacteria through downregulation of DNA repair capacity. The regulation of DNA repair by RecA is a significantly complex process (44), and although we specifically demonstrated the regulation of DNA Pol I and BER repair system by RecA under antibiotic exposure, there should be other repair mechanisms impaired and contributed to this fast evolution of resistance, which needs to be investigated in future.

Antibiotics exert their bactericidal effects by targeting specific intracellular components, and the production of ROS is considered an alternative mechanism underlying quinolone-induced bacterial killing. However, emerging evidence suggests that lethality may arise from primary antibiotic-induced damage or a secondary response to lethal stress mediated by ROS. ROS generation can also contribute to the development of multidrug resistance, as it can directly damage DNA, leading to genetic mutations. In this study, we demonstrate that excessive ROS accumulation drives the evolution of antibiotic resistance in bacteria with impaired DNA repair capabilities. Significantly, we discovered for the first time that antibiotic exposure can significantly increase the transcription of H-NS in the absence of recA. E. coli H-NS is a DNA-binding protein known for its involvement in transcriptional repression (45). It binds to DNA in a sequence-independent manner, with a preference for A-/T-rich regions in curved DNA, thereby impeding transcription (46). Previous findings report that H-NS can down-regulate the transcriptional expression of cysIJH through the mediation of cysB (47,48). While we have established a potential correlation between the increased transcription of H-NS genes induced by antibiotic exposure in RecA deficient cells and the decreased transcription of genes related to antioxidative processes, further investigations are imperative to elucidate the regulatory mechanisms employed by RecA in H-NS transcription and its inhibitory effects on antioxidative gene transcription.

In clinics, the resistance mechanism to β-lactams is not only associated with β-lactamase acquisition but also with increased expression of efflux pumps (44). Our research for the first time demonstrates the fast evolution of antibiotic resistance through the rapid accumulation of acrB mutations and enhanced functionality of the AcrAB-TolC efflux pump in RecA-deficient bacteria exposed to a single dose of ampicillin. This finding supports the clinical observations that the clinical antibiotic resistance is accelerating (50).

DNA repair is a key target for many anti-cancer drugs that aim to destroy tumor cells by inducing DNA damage (51). Homologous recombination (HR) plays a crucial role in DNA repair, and Rad51, a vital protein in the HR pathway of eukaryotic cells, belongs to the recA/RAD51 gene family (51–53). Clinical trials are currently assessing a range of Rad51 inhibitors. Interestingly, studies have revealed that inhibitors targeting Rad51 can effectively inhibit the activity of RecA (54). Despite RecA and Rad51 sharing only approximately 30% sequence homology, the filaments they form and the conformational changes they induce in DNA are remarkably similar (54). Notably, inhibitors of RecA have demonstrated the ability to enhance the antimicrobial effects of certain antibiotics, including fluoroquinolones (55,56). Our findings have significant clinical implications, especially for patients undergoing cancer treatment. The combination therapy involving β-lactam antibiotics and RecA inhibitors could potentially accelerate the evolution of multi-drug resistance at an alarming rate.

Our findings on the mechanism of the fast evolution of multidrug resistance in E. coli after a single exposure to a specific antibiotic can be applied more broadly. The overproduction of ROS is a common effect of many antibiotics and other strategies used to combat infections, such as antimicrobial photodynamic therapy (aPDT) and cold atmospheric plasma (CAP) (57,58). These approaches rely on ROS-mediated damage as the primary source of stress-induced damage. Thus, maintaining the transcription of genes involved in the oxidative stress response or the combined treatment with drugs and antioxidants shows promising potential to prevent stress-induced mutagenesis that is typically triggered by various classes of antibiotics.

Acknowledgements

This work was supported by the Australian Research Council (ARC grant no: APP1165135), Science and Technology Innovation Commission of Shenzhen (KQTD20170810110913065), Australia China Science and Research Fund Joint Research Centre for Point-of-Care Testing (ACSRF658277, SQ2017YFGH001190).

Additional information

Author contributions

L.Z., G.P., Q.S., A.R., E.H., A.O. and D.J. designed the experiments. L.Z., N.C., A.B., E.H., and D.J. wrote the manuscript with input from all co-authors. L.Z., G.P., Q.S., and Y.Y.C. conducted experiments and analysed data.

Competing interests

The authors declare that they have no competing interests.

Materials and Methods

Bacterial strains, medium and antibiotics

Bacterial strains and plasmids used in this work are described in Table S2 and Table S3. Luria-Bertani (LB) was used as broth or in agar plates. E. coli cells were grown in LB liquid medium or on LB agar (1.5% w/v) plates at 37°C, unless stated otherwise, antibiotics were supplemented, where appropriate. Antibiotic stock solutions were prepared by dissolving antibiotics in MilliQ filter sterilising, including ampicillin (50 mg/mL), penicillin G (100 mg/mL), carbenicillin (20 mg/mL), kanamycin (50 mg/mL) and tetracycline (10 mg/mL). Chloramphenicol stock solution was prepared in 95% EtOH (25 mg/mL). Antibiotic solutions were stored at -20°C (long-term) or 4°C (short-term).

Treatment with antibiotics to induce evolutionary resistance

For the single exposure to antibiotic experiment, an overnight culture (0.6 mL; 1 x 10⁹ CFU/mL cells) was diluted 1:50 into 30 mL LB medium supplemented with antibiotics (50 μg/mL ampicillin, 1 mg/mL penicillin G, or 200 μg/mL carbenicillin) and incubated at 37°C with shaking at 250 rpm for 0, 2, 4, 6 and 8 hours, respectively. After each treatment, the antibiotic-containing medium was removed by washing twice (20 min centrifugation at 1500 g) in fresh LB medium (See Fig. 1A for method overview).

To test resistance, the surviving isolates were first resuspended in 30 mL LB medium and grown overnight at 37°C with shaking at 250 rpm. The regrown culture was then plated onto LB agar and incubated overnight at 37°C. Single colonies were isolated and grown in LB medium for 4-6 hours at 37°C with shaking at 250 rpm, which were then used to test the resistance or stored at -80°C for future use.

For the ALE antibiotic treatment experiments, an overnight culture (0.6 mL; 1 x 10⁹ CFU/mL cells) was diluted 1:50 into 30 mL LB medium supplemented with 50 μg/mL ampicillin and incubated at 37°C with shaking at 250 rpm for 4.5 hours. After treatment, the antibiotic-containing medium was removed by washing twice (20 min centrifugation at 1500 g) in a fresh LB medium. The remaining pellet was resuspended in 30 mL LB medium and grown overnight at 37°C with shaking at 250 rpm. Ampicillin treatment was applied to the regrown culture and repeated until resistance was established, as confirmed by MIC measurement.

Antibiotic susceptibility testing

The susceptibility of E. coli cells to antibiotics was measured using minimum inhibitory concentration (MIC) testing (60). In brief, overnight cultures were diluted and incubated at 37°C for 4-6 hours with shanking at 250 rpm. Cells were then diluted 1:100 and incubated with increasing concentrations of antibiotics in the Synergy HT BioTek plate reader (BioTek Instruments Inc., USA) at 37°C for 16 hours. It was programmed to measure the OD hourly at 595 nm (Gen5 software, BioTek Instruments Inc., USA). The minimum inhibitory concentration was determined as the concentration of antibiotic where no visible growth was observed.

Survival assays

Overnight cultures of E. coli were prepared from single colonies in LB medium and incubated at 37°C with shaking (250 rpm). The overnight cultures were diluted 1:50 in fresh LB medium containing 50 µg/mL ampicillin to initiate antibiotic treatment. Cultures were incubated at 37°C for the indicated times under shaking conditions. Following treatment, cells were collected by centrifugation (1500 g, 20 minutes), and the antibiotic-containing medium was removed by washing twice with fresh LB medium. Serial dilutions of the washed cultures were prepared, and 25 µL of each dilution was plated onto LB agar plates. Plates were incubated overnight at 37°C, and CFUs were counted the following day to evaluate bacterial survival rates.

Mutation frequency

Overnight cultures inoculated from single colonies in LB medium were diluted 1:1,000,000 and incubated at 37°C with shaking until the OD₆₀₀ reached to 1∼1.3. This extreme dilution minimizes the presence of pre-existing stationary phase mutants. The total number of colony-forming units per ml (CFU/ml) was determined by plating on LB agar. To count mutants, cells were centrifuged and plated on LB agar with an appropriate concentration of antibiotic. LB plates were incubated for 24 hours at 37°C and selective plates were incubated for 48-72 hours at 37°C (61). The mutation frequency was then determined as the CFU/ml on LB + selective antibiotic agar plates divided by the CFU/ml on LB agar plates.

Construction of recA deletion mutant

Lambda Red recombination was used to generate the gene recA deletion in the E. coli K-12 strain, followed by previously reported methods with modifications (62,63). Primers (recA-FWD and recA-REV, Table S4) were designed approximately 50 bp upstream and downstream to the gene recA on the chromosome to amplify the tetracycline cassette as well as the flanking DNA sequence needed for homologous recombination. Phusion polymerase (NEB) was used to amplify the DNA sequence (Table S4), and the reaction was cleaned up using a PureLink™ PCR purification kit (ThermoFisher Scientific) as per the manufacturer’s instructions. Electro-competent E. coli MG1655 containing the recombinase plasmid pKD46 was transformed with 50 ng of amplified DNA a 30°C. The transformation was plated onto LB agar plates containing 10 μg/mL tetracycline and incubated overnight at 37°C. PCR was used to confirm the insertion of the tetracycline resistance cassette at the correct site on the chromosome using primers upstream and downstream to the gene recA. The newly constructed mutant strains were cured of plasmid pKD46 by incubating LB streak plates at 42°C overnight. Loss of the plasmid was confirmed by lack of ampicillin sensitivity on LB agar plates. Mutant strains were made electro-competent, and 50 μL of cells were transformed with plasmid pCP20 and incubated on 100 μg/mL ampicillin plates at 30°C overnight. A few colonies were then restreaked onto LB plates and incubated overnight at 42°C. PCR products confirmed the loss of cassette and plasmid.

β-lactamase assay

The amount of β-lactamase was measured using a β-lactamase Activity Assay Kit (Sigma-Aldrich, US). Briefly, cells were collected by centrifugation at 10,000 g for 10 min, and the pellet was resuspended with 5 µL of assay buffer per mg of sample. Then, 48 μL of the sample was mixed with 2 μL of nitrocefin. The β-Lactamase activity was monitored by measuring the absorbance at 490 nm for 30 min at 28°C. The level of β-lactamase was determined by the absorbance at OD₃₉₀.

Whole genome sequencing

Resistant clones were isolated by selection using LB agar plates with the supplementation of ampicillin at 50 μg/mL, and non-resistant clones were isolated from the LB agar plates without the supplementation of antibiotics. Chromosomal DNA was extracted and purified using the PureLink™ Genomic DNA mini kit following the manufacturer’s instructions (ThermoFisher Scientific). Whole genome sequencing (WGS) was conducted following the Nextera Flex library preparation kit process (Illumina). Briefly, genomic DNA was quantitatively assessed using Quant-iT picogreen dsDNA assay kit (Invitrogen, USA). The sample was normalised to the concentration of 1 ng/μL. 10 ng of DNA was used for library preparation. After tagmentation, the tagmented DNA was amplified using the facility’s custom-designed i7 or i5 barcodes, with 12 cycles of PCR. The quality control for the samples was done by sequencing a pool of samples using MiSeq V2 nano kit - 300 cycles. After library amplification, 3 μL of each library was pooled into a library pool. The pool was then cleaned up using SPRI beads following the Nextera Flex clean-up and size selection protocol. The pool was then sequenced using a MiSeq V2 nano kit (Illumina, USA). Based on the sequencing data generated, the read count for each sample was used to identify the failed libraries (i.e., libraries with less than 100 reads).

Moreover, libraries were pooled at different amounts based on the read count to ensure equal representation in the final pool. The final pool was sequenced on Illumina NovaSeq 6000 Xp S4 lane, 2 × 150 bp. WGS read quality was assessed using FASTQC (version 0.11.5) and trimmed using Trimmomatic (version 0.36) with default parameters and trimmed of adaptor sequences (TruSeq3 paired-ended). Reads were aligned to the E. coli MG1655 genome (http://bacteria.ensembl.org/Escherichia_coli_str_k_12_substr_mg1655_gca_000005845/Info/Index/, assembly ASM584v2) and then analysed variants following GATK Best Practices for Variant Discovery (HaplotypeCaller) (64). Further genome variant annotation was conducted using the software SnpEff (65).

Global transcriptome sequencing

After ampicillin treatment for 0 and 8, surviving isolates were immediately washed and harvested for global transcriptome sequencing. Total RNA was extracted from the cell pellets using a PureLink RNA mini kit (Invitrogen) as per the manufacturer’s instructions. The global transcriptome sequencing was processed and analysed by Genewiz, Jiangsu, China. Primers used in this work are listed in Table S4. RNA-Seq read quality was assessed using FASTQC and trimmed using Trimmomatic with default parameters. Reads were aligned to the E. coli MG1655 genome (http://bacteria.ensembl.org/Escherichia_coli_str_k_12_substr_mg1655_gca_000005845/Info/Index/, assembly ASM584v2) and then counted using the RSubread aligner with default parameters (66). After mapping, differential expression analysis was carried out using strand-specific gene-wise quantification using the DESeq2 package (67). Further normalisation was conducted using RUVSeq and the RUV correction method, with k = 1 to correct for batch effects, using replicate samples to estimate the factors of unwanted variation (68). Absolute counts were transformed into standard z-scores for each gene over all treatments, that is, absolute read for a gene minus mean read count for that gene over all samples and then divided by the standard deviation for all counts over all samples. Genes with an adjusted P value (P_adj) of ≤0.05 were considered differentially expressed. PseudoCAP analysis was conducted by calculating the percentage of genes in each classification that were differentially expressed (log₂FC ≥ ±2, P_adj ≤ 0.05).

ROS measurement

Intracellular ROS accumulation was measured using the Cellular ROS Assay Kit (Abcam, ab113851) according to the manufacturer’s protocol. Overnight cultures of both wild type and ΔrecA strains were grown in LB medium. Cells were treated with 50 µg/mL ampicillin for 8 hours at 37°C with shaking (250 rpm). After the treatment, cells were washed twice with 1x buffer to remove residual antibiotics and debris. The collected cells were resuspended in 1x buffer and incubated with the fluorescent ROS probe DCFDA (2’,7’–dichlorofluorescin diacetate)/H2DCFDA at a final concentration of 10 µM for 30 minutes at 37°C in the dark to prevent probe degradation.

After incubation, 500 µL of the stained cells were immediately analysed using a CytoFLEX LX flow cytometer (Beckman Coulter). Fluorescence was detected in the FITC channel (excitation at 488 nm, emission at 525 nm). Data were analysed using FlowJo software (Version X, Tree Star Inc.), with fluorescence intensity serving as an indicator of intracellular ROS levels. Appropriate controls, including unstained cells and cells without ampicillin treatment, were included to ensure accurate ROS measurement.

Single-molecule localisation imaging and data analysis

Single-molecule localisation imaging was performed on a custom-built Stochastic Optical Reconstruction Microscope (STORM) with an Olympus IX81 microscope frame, a 100x magnification NA 1.45 objective (Olympus) and an EMCCD camera (DU-897, Andor) as described previously (69–71). In summary, 35 mm cell culture dishes (0.17 mm No.1 coverglass) were cleaned with 1 M KOH for 30 minutes in an ultrasonic cleaning machine, followed by three washes with MilliQ water. The dishes were air-dried with high-purity nitrogen blowing and sterilised by UV exposure for 30 minutes. E. coli cells were fixed with NaPO4 (30 nM), formaldehyde (2.4%), and glutaraldehyde (0.04%) at room temperature for 15 minutes, followed by 45 minutes on ice. Samples were then centrifuged to collect the pellet cells, and the supernatant was discarded. Cell pellets were washed twice with phosphate-buffered saline (PBS), pH 7.4. Cells were resuspended in 200 μL of GTE buffer and kept on ice until 200 µL was placed onto the coverslip bottom of the cleaned 35 mm culture dish. To label the bacterial chromosome, a Click-iT EdU kit was used prior to fixation following the manufacturer’s instruction (ThermoFisher) and as described before. To label DNA polymerase I, fixed cells were blocked and permeabilised with blocking buffer (5% wt/vol bovine serum albumin (Sigma-Aldrich) and 0.5% vol/vol Triton X-100 in PBS) for 30 min and then incubated with 1 μg/mL primary antibody against DNA polymerase I (ab188424, Abcam) in blocking buffer for 60 min at room temperature. After washing with PBS three times, the cells were incubated with 2 μg/mL fluorescently labelled secondary antibody (Alexa 647, A20006, ThermoFisher) against the primary antibody in the blocking buffer for 40 min at room temperature. After washing with PBS three times, the cells were postfixed with 4% (wt/vol) paraformaldehyde in PBS for 10 min and stored in PBS before imaging. STORM image analysis, drift correction, image rendering, protein cluster identification and images presentation were performed using Insight342, custom-written Matlab (2012a, MathWorks) codes, SR-Tesseler (IINS, Interdisciplinary Institute for Neuroscience) (72), and Image J (National Institutes of Health).

Statistical analysis

Statistical analysis was performed using GraphPad Prism v.9.0.0. All data are presented as individual values and mean or mean ± SEM. A one-tailed unpaired Student’s t-test using a 95% confidence interval was used to evaluate the difference between the two groups. A probability value of P < 0.05 was considered significant. Statistical significance is indicated in each figure. All remaining experiments were repeated independently, at least six with similar results.

Data availability

Sequence data supporting this study’s findings have been deposited in the GEO repository with the GEO accession number GSE179434.