Genome-wide screen reveals cellular functions that counteract rifampicin lethality in Escherichia coli

ABSTRACT The antibacterial activity of rifamycins specifically relies on the inhibition of transcription by directly binding to the β-subunit of bacterial DNA-dependent RNA polymerase (RNAP). However, its killing efficacy is substantially diminished in most gram-negative bacteria. To systematically reveal the cellular functions that counteract rifamycin-mediated killing in the gram-negative model organism Escherichia coli, we performed a genome-wide Tn5 transposon-mediated screen to identify mutants with altered susceptibility to rifampicin. Combined with targeted gene knockouts, our results showed that the β-barrel assembly machinery plays a crucial role in restricting rifampicin from entering the cell, whereas mutants deficient in other cellular permeability barriers, such as lipopolysaccharide and enterobacterial common antigen, had no such effect. At bactericidal concentrations, the killing efficacy of rifampicin was strongly influenced by cellular functions, including iron acquisition, DNA repair, aerobic respiration, and carbon metabolism. Although iron acquisition de facto has a strong impact or dependence on cellular redox, our results suggest that their effects on rifampicin efficacy do not rely on hydroxyl radical formation. We provide evidence that maintenance of DNA replication and transcription-coupled nucleotide excision repair protects E. coli cells against rifampicin killing. Moreover, our results showed that sustained aerobic respiration and carbon catabolism diminish rifampicin’s killing efficacy, and this effect relies on the inhibition of transcription but not on translation. These findings suggest that the killing efficacy of rifamycins is largely determined by cellular responses upon the inhibition of RNAP and may expand our knowledge of the action mechanisms of rifamycins. IMPORTANCE Rifamycins are a group of antibiotics with a wide antibacterial spectrum. Although the binding target of rifamycin has been well characterized, the mechanisms underlying the discrepant killing efficacy between gram-negative and gram-positive bacteria remain poorly understood. Using a high-throughput screen combined with targeted gene knockouts in the gram-negative model organism Escherichia coli, we established that rifampicin efficacy is strongly dependent on several cellular pathways, including iron acquisition, DNA repair, aerobic respiration, and carbon metabolism. In addition, we provide evidence that these pathways modulate rifampicin efficacy in a manner distinct from redox-related killing. Our findings provide insights into the mechanism of rifamycin efficacy and may aid in the development of new antimicrobial adjuvants.

of rifamycins) is a cornerstone of short-course antituberculosis therapy because of its sterilizing activity against both replicating and nonreplicating antibiotic-tolerant Mycobacterium tuberculosis.This sterilizing activity is barely observed in most bacteria that are de facto sensitive to rifampicin (4).
Early investigations attributed this difference in antimicrobial activity to the efficacy of drug uptake, as most gram-negative bacteria show lower susceptibility to rifamycins owing to the presence of the outer membrane (OM) (2).However, the penetration of the antibiotic through the cell membrane may account for rifamycin susceptibility in terms of minimum inhibitory concentration (MIC) but not for killing efficacy, especially when considering that rifamycins do not exhibit a typical concentration-dependent killing effect on Escherichia coli (2).In fact, even at concentrations that could completely inhibit RNAP, the killing efficacy of rifampicin remains mild in E. coli (5)(6)(7)(8)(9).Moreover, biochemical studies have demonstrated that the difference in rifampicin's killing efficacy is unlikely to stem from variations in their binding site on RNAP, as RNAPs from both gram-negative and gram-positive bacteria have comparable sensitivities to rifampicin (9,10).These results indicate that rifamycin lethality may depend on cellular responses to RNAP inhibition (11,12).
Mounting evidence indicates that bacterial metabolic states and responses have a strong impact on antibiotic efficacy (12,13).The immediate downstream events of RNAP inhibition are the prevention of transcription and translation, both of which are major energy-consuming processes of cellular activities (7,14,15).Previous studies have shown that blocking RNA synthesis by rifampicin leads to a reduced rate of aerobic respiration and metabolism (5).Although these cellular responses could protect bacteria from bactericidal antibiotics, their role in rifampicin's efficacy remains unclear.
In bacteria, the cellular response to DNA damage has been associated with antibiotic efficacy, as well as other lethal stresses (16)(17)(18)(19).In E. coli, the deletion of recA results in hypersensitivity to rifampicin killing (6).In mycobacteria, independent lines of evidence suggested a role of reactive oxygen species (ROS) in rifampicin-induced lethal DNA damage (17,(20)(21)(22)(23).However, because rifampicin inhibits aerobic respiration and does not stimulate ROS production in E. coli (5,24), the molecular mechanisms underlying DNA damage and damage repair upon rifamycin action remain elusive.
To systematically reveal the cellular functions that influence rifamycin lethality, we performed genome-wide screening of the E. coli K12 strain MG1655.Through the construction of a high-density Tn5 transposon mutation library (71.4/kb) and deploy ing high-throughput transposon insertion sequencing (Tn-Seq) (25), we comprehen sively identified the mutants with altered fitness upon exposure to different rifampicin concentrations (0.25× MIC to 20× MIC).Combined with targeted gene knockouts, our results demonstrated that the efficacy of rifampicin strongly relies on cellular functions, including iron acquisition, DNA repair, aerobic respiration, and carbon metabolism.Further, we provided evidence that these pathways modulate rifampicin's efficacy in a manner that is distinct from redox-related killing.These findings provide insights into the mechanism of action of rifamycins and may aid in the development of new antimicrobial strategies to improve rifamycin efficacy.

Selection of transposon mutants altering rifampicin efficacy
A genome-wide screen was deployed to comprehensively identify the cellular functions involved in modulating the efficacy of rifampicin in the model organism, E. coli K12 strain MG1655.The Tn5 transposon insertion mutant library was constructed as previously described (26).Genomic DNA from two biological replicates was subjected to Tn-Seq.The normalized read count of the two biological repeats exhibited a Pearson correlation coefficient of 0.91, demonstrating the high reproducibility of this methodology (Fig. 1A; Table S1).The results showed that this library contained ~310,000 randomly distributed Tn5 insertions, corresponding to an insertion density of 71.4/kb (Fig. 1B).
The E. coli K12 strain MG1655 had a rifampicin MIC of 8 mg/L (Fig. S1A).Most bacterici dal antibiotics exhibit concentration-and time-dependent killing effects.However, an intriguing characteristic of rifampicin in E. coli is the absence of concentration-depend ent or time-dependent effects (Fig. 1C).This property indicates that a mutant with a reduced MIC to rifampicin would only be selected by Tn-Seq at subinhibitory concentra tions of rifampicin (sub-MIC) but not at concentrations above the MIC (11).Thereafter, to identify the mutants with reduced MIC, rifampicin at 0.25× MIC was selected as a screen condition.In addition, to identify the cellular functions involved in modulating rifampicin efficacy, 4× MIC and 20× MIC were also selected.The pooled library was grown aerobi cally in the middle exponential phase and exposed to 2 (0.25× MIC), 32 (4× MIC), and 160 (20× MIC) mg/L rifampicin.At 1 and 3 h post-treatment, a sample containing about 3 × 10 6 viable cells (~10 times the library diversity) was recovered by plating, and genomic DNA was extracted from the pooled outgrowths for Tn-Seq (Fig. 1C).The results of the two biological replicates under each condition were highly reproducible (Pearson correlation coefficient of 0.9 to 0.99) (Fig. 1D).
Next, we used the nonparametric resampling method of the TRANSIT software to identify mutants that were differentially represented between input and post-treatment samples (Table S2) (27,28).For each gene, the normalized read counts at all the insertion sites and all replicates under each condition were summed.The difference of the summed read counts between the input and post-treatment samples was calculated for each mutant and expressed as a log 2 fold change (log 2 FC).The significance of this difference was calculated using a permutation test.According to the criteria of log 2 FC > 1 or log 2 FC < −1 and adjusted P-value <0.05, a total of 365 genes were selected.Of the identified genes, 216 showed reduced fitness (log 2 FC < −1) upon exposure to rifampicin, suggesting a role in counteracting the action of rifampicin in E. coli (Table S3).

The impact of cellular permeability barriers on rifampicin susceptibility
In accordance with the speculation that the absence of a concentration-dependent killing effect may lead to the selection of mutants with distinct fitness changes between sub-MIC and above-MIC concentrations of rifampicin, we identified 58 genes showing altered fitness exclusively under 0.25× MIC rifampicin (Table S3).Functional categories related to outer membrane assembly (OMA, false discovery rate [FDR] = 0.08) and biosynthesis of enterobacterial common antigen (ECA, FDR = 4.5 × 10 −8 ) were signifi cantly enriched among these genes (Fig. 2A; Table S4).OMA is a critical process that maintains the integrity of OM (29), a defining feature of gram-negative bacteria that poses a selective permeability barrier that has evolved to restrict toxic compounds from entering the cell (30).All the selected OMA genes, including bamB, bamE, skp, and lptC, showed reduced fitness upon Tn5 insertion (log 2 FC −2.3 to −5.2).BamB and BamE belong to the β-barrel assembly machinery (BAM complex, BamA-E), responsible for the assembly and insertion of outer membrane proteins (OMPs).Skp is a periplasmic chaperone that functions to remove β-barrel OMPs that have stalled during assembly on the BAM complex (30).To validate these results, we constructed ΔbamB and Δskp mutants in E. coli and found that the inactivation of these genes resulted in eightand twofold reductions in MIC, respectively (Fig. 2B and C).The increased rifampicin susceptibility of these mutants was completely restored by the expression of bamB or skp (Fig. 2B and C).Therefore, these results indicate that the BAM complex plays a crucial role in intrinsic resistance to rifampicin.
ECA is a surface carbohydrate antigen composed of repeating units of trisaccharide, which limits the accessibility of hydrophobic compounds to OM (31).However, our results showed that all the selected mutants with Tn5 insertion in the ECA synthesis genes resulted in increased fitness (log 2 FC 1.3 to 2.9), suggesting that ECA may not function as a barrier for rifampicin uptake (Fig. 2A).Similarly, a large-scale chemical genetic screen also showed that deletion of rffG, rffH, or wecA in E. coli K12 led to reduced sensitivity to sub-MIC rifampicin (32).
While most of the enriched genes showed consistent changes in susceptibility to different drug concentrations, three genes (waaPG and galU) belonging to LPS core region biosynthesis showed opposing fitness changes between sub-and above-MIC concentrations of rifampicin (Table S4).Moreover, although there were seven selected genes (waaBCDEFOQ) belonging to the LPS core region biosynthetic pathway that showed consistent changes in susceptibility to different drug concentrations (33), the effects were opposite between waaBOQ (increased fitness change) and waaCDEF (decreased fitness change) (Table S4).These results indicate that the role of LPS in rifampicin susceptibility may not be akin to its function as a permeability barrier.

Iron acquisition modulates rifampicin lethality in a hydroxyl radical-inde pendent manner
Our results showed that the genes involved in iron metabolism exhibited the most marked changes in fitness and enrichment (Fig. 4A; Table S4).E. coli encodes multiple systems, including iron chelating, export, import, and sequestration, to maintain iron homeostasis (Fig. 4B) (34).Among these systems, the FetAB iron exporter showed the most profound fitness reduction upon Tn5 insertion (log 2 FC ~−7), suggesting that shrinking the iron pool may contribute to intrinsic tolerance to rifampicin (Fig. 4A and  B) (35).Similarly, mutations in either the ferric citrate importer FecABCDE (log 2 FC ~5) or the TonB complex (encoded by tonB, exbB, and exbD), which energizes iron importers, led to increased fitness (Fig. 4A and B) (36).However, one exception was the FepBCDG ferric enterobactin importer, which exhibited reduced fitness upon Tn5 insertion (Fig. 4A).Although an early study showed that strains with Tn5 insertion in fepD, -C, and -G were unable to grow under iron starvation conditions, its impact on intracellular iron levels under normal growth conditions remains unclear (37).To validate these results, fetA, fepD, fecA, and exbB were individually deleted in E. coli.We found that the deletion of these genes did not affect rifampicin MIC (Fig. 4C), thus excluding its role in rifampicin uptake (38,39).The survival abilities of these mutants upon rifampicin treatment were highly consistent with the fitness change observed on the screen.The survival pheno types of these mutants could be completely restored by the expression of the gene or operon in the mutant strain (Fig. 4D through G).
Iron ions can trigger the generation of hydroxyl radicals (OH .) via the Fenton reaction, which can contribute to cell death induced by bactericidal antibiotics (40,41).However, we found that the addition of the hydroxyl radical scavenger thiourea had no significant effect on the killing efficacy of rifampicin in the wild-type and ΔfetA strains (Fig. 4H).Hydrogen peroxide (H 2 O 2 ) functions as a substrate to induce hydroxyl radical formation via the Fenton reaction (42).Our results showed that the deletion of katG, which encodes the primary scavenger of H 2 O 2 (43), has no impact on the survival of wild-type E. coli upon rifampicin treatment (Fig. 4I).Collectively, these results suggest that iron metabo lism contributes to rifampicin lethality in a hydroxyl radical-independent manner in E. coli.

Maintenance of DNA replication and transcription-coupled DNA repair protect E. coli cells against killing by rifampicin
Increasing evidence shows that DNA damage, especially DSBs, is a causative factor for cell death induced by bactericidal antibiotics (11,17).Although rifampicin is not a typical bactericidal agent in E. coli, the reduced fitness of the mutants deficient in recombination repair of DSB (recABC, recG, and ruvA) indicates the engagement of DSB in rifampicin lethality (Fig. 5A) (44).To validate these results, recA and recG were individually deleted in E. coli, and the results showed that the mutants became more sensitive to rifampicin killing without affecting the MIC (Fig. 5B through D).Early studies established that DSBs induced by bactericidal antibiotics could be generated by incomplete base excision repair (BER) (17,18,(45)(46)(47).However, our results showed that genes involved in the BER pathway, including the previously validated mutM, mutY, mutT, mazG, and nth, were not selected in our screen (Table S4).These results, taken together with the irrelevance of hydroxyl radicals in rifampicin-induced killing (Fig. 4H and I) (5,24), indicate that rifampicin induces lethal DNA damage in an oxidation-independent manner.
The transcription complex is the primary source of replication fork pausing, which can cause genome instability (48,49).E. coli rep encodes an adenosine triphosphate (ATP)-dependent DNA helicase involved in preventing replication fork stalling by facilitating replication through the transcribed regions (48,49).The rifampicin-hypersen sitive phenotype of rep suggests that inhibition of RNAP may cause replication fork pausing, which requires fork bypassing for survival (Fig. 5A).In addition, recent progress has established that the TCR pathway accounts for a vast majority of DNA repair events (50,51).In this model, UvrA and UvrD associate with RNAP, forming a surveillance complex that can timely backtrack RNAP and initiate nucleotide excision repair (NER) upon encountering DNA lesions during transcription.In accordance with this scenario, our results showed that Tn5 insertion in uvrA or uvrB resulted in a marked fitness reduction (log 2 FC down to −5.2) among the selected DNA repair genes.Further, we found that the deletion of uvrA in E. coli not only led to hypersensitivity to rifampicin but also resulted in a twofold reduction in MIC (Fig. 5E and F).These phenotypes were completely restored by the expression of uvrA in the mutant strain.Importantly, the deletion of E. coli uvrA had no impact on ampicillin killing (Fig. 5G), indicating a unique role of NER in the action of rifampicin.Moreover, the perturbation of TCR upon rifampicin action could also be signified by the increased survival of greAdeficient mutant (Fig. 5H).According to recent studies, the inactivation of greA in E. coli could stimulate RecABCDmediated DNA repair by promoting the RNAP backtrack (52,53).

Sustained cellular respiration and catabolism diminish rifampicin efficacy
A previous study established that the lethality of most bactericidal antibiotics is associated with accelerated respiration.However, the action of rifampicin is linked to the inhibition of respiration in various organisms, regardless of its effect on killing (5).The effect of cellular respiration on the lethality of rifampicin remains unclear.Our results showed that genes related to the respiration chain NADH dehydrogenase I (NDH-I, encoded by nuoA-N), succinate dehydrogenase (encoded by sdhA-D), and bo-type oxidase (encoded by cyoA-D) were associated with reduced fitness upon Tn5 insertion (Fig. 3B; Tables S3 and S4).Given that deficiency in NDH-I or bo-type oxidase leads to significantly reduced oxygen consumption in E. coli (54), these results indicate that sustained aerobic respiration may counteract rifamycin killing.To test this hypothe sis, we constructed an unmarked ΔnuoA-N strain in E. coli, and the results showed that the mutant became more sensitive to rifampicin (Fig. 6A).The deletion of nuoA-N in E. coli resulted in a growthdeficient phenotype in enriched medium (Fig. S1B), consistent with previous findings that NDH-1 predominates when E. coli is grown at high oxygen tension (54,55).The deletion of nuoA-N in E. coli did not affect the MIC of rifampicin (Fig. S1A), suggesting a role of NDH-1 in rifampicin's killing, rather than drug entry.Moreover, our results indicated that the NDH-1-added survival effect specifically relies on the inhibition of transcription by rifampicin, and the deletion of nuoA-N in E. coli resulted in decreased killing by ampicillin or gentamycin and did not affect the sensitivity to spectinomycin, chloramphenicol, and tetracycline (Fig. 6B; Fig. S1C) (5).Collectively, these results suggest a role for sustained aerobic respiration in counteracting rifampicin-induced killing.
Aerobic respiration plays a pivotal role in cell metabolism by recycling NAD + and generating ATP.To dissect the mechanisms underlying NDH-I-induced survival, we measured ATP and NADH/NAD + in the ΔnuoA-N strain.Our results showed that NDH-I deficiency resulted in a greater than twofold increase in the NADH/NAD + ratio, whereas its effect on cellular ATP was very mild (Fig. 6C and D).Given that the regeneration of reducing equivalents is critical for the maintenance of cellular catabolism, these results indicate that the effect of NDH-I depletion on rifampicin killing may be attributable to reduced catabolism (56).This speculation could be supported by the selection of genes enriched in carbon metabolism (eco01200, FDR = 3.6 × 10 −3 ) (Fig. 3B; Table S4).These metabolic genes mainly belonged to the tricarboxylic acid (TCA) cycle (including gltA, sdhB, mdh, and fumC), pentose phosphate pathway (tpiA, talB, tktA, rpiA, rpe, and gnd), and glycolysis (pgi, pfkA, and pykF), almost all of which exhibited reduced fitness upon Tn5 insertion (except for gnd encoding for 6-phosphogluconate dehydrogenase).To validate our hypothesis further, we constructed a mutant deficient in gltA, which encodes the first committed enzyme of the TCA cycle.The results showed that, upon exposure to rifampicin, the survival rate of the ΔgltA mutant was reduced by three orders of magnitude compared to the wild type (Fig. 6E).This phenotype was completely restored by gltA expression in the mutant strain.These results provide evidence that sustained NAD + recycling and carbon catabolism counteract the killing effect of rifampicin.

DISCUSSION
Rifampicin plays a pivotal role in shortening the course of antituberculosis therapy owing to its sterilizing activity against M. tuberculosis (4).However, owing to its low bactericidal activity and rapid development of resistance, its clinical application in infectious diseases caused by most gram-negative bacteria is limited.In this study, we comprehensively identified and evaluated the cellular functions that modulate the action of rifampicin in the model organism E. coli and provided unique insights into its lethal effects.
Our results showed that E. coli mutants deficient in DSB-repair genes recA-C, recG, and ruvA became more sensitive to rifampicin killing, suggesting a role of lethal DNA damage in rifampicin killing (6).Although the precise mechanisms underlying rifampicininduced DNA damage remain elusive, our study revealed the unique characteristics of its action.First, in accordance with the inability of rifampicin to induce ROS in E. coli (5,24), we found that the well-characterized BER genes that contribute to ROS-induced DSB were not selected upon rifampicin action (17,18,(45)(46)(47).These findings, taken together with the results showing the irrelevance of katG and thiourea in rifampicin's killing, signify distinct oxidation-independent DNA damage events.Further, we provided genetic evidence that rifampicin-induced DNA damage may stem from perturbations in DNA replication and TCR.The implication of Rep in rifampicin lethality suggests that the inhibition of RNAP may result in a stalled transcriptional complex, leading to replication fork collision and pausing (48,49).The participation of TCR in rifampicin's action could be exemplified by the hypersensitivity of the mutants deficient in UvrA/UvrB, as well as the protective effect posed by GreA depletion.In E. coli, UvrA, UvrD, and RNAP form a surveillance complex that can timely backtrack RNAP and initiate NER upon encoun ter with a DNA lesion during transcription (50,51).By contrast, the presence of GreA impedes RNAP backtracking (52,53).Our results demonstrated that the deletion of uvrA in E. coli resulted in a twofold reduction in rifampicin MIC but did not affect ampicillin sensitivity, suggesting a unique role of NER in rifampicin action.
Early studies have established the role of the ferrichrome transport system (FhuA-TonB) encoded by some gram-negative bacteria in the uptake of rifamycin derivatives CGP 4832 and rifabutin, but not of rifampicin (38,39,57).While our results are consistent with these findings, showing the unchanged rifampicin MIC of the mutant deficient in the TonB complex and other iron transporters, it also suggests a novel role of iron metabolism in rifamycin killing.Moreover, the evidence presented in this study suggests that the impact of iron on rifampicin's killing does not rely on the generation of hydroxyl radicals, which is strikingly different from the well-established role of iron in ROS-medi ated cell death (40,41).Together, these observations suggest an alternative route for iron-related cell death.Notably, antibiotic lethality under anaerobic conditions can be induced by reactive metabolic byproducts in E. coli (58).Considering the inhibitory effect of rifampicin on cellular respiration (5), further studies should address the physiological role of iron under oxygen-limited conditions.
For most bactericidal antibiotics, the inhibition of bacterial respiration counteracts killing (5).However, our results indicate that reduced aerobic respiration contributes to rifampicin-induced killing of E. coli.We found that the depletion of NDH-I in E. coli resulted in increased killing by rifampicin, which could be attributed to increased NADH/ NAD + and reduced carbon catabolism.Moreover, our results demonstrated that this effect relies on the inhibition of transcription but not translation.It is possible that the reduced but sustained respiration and catabolism add to survival by providing building blocks for the repair of cellular damage induced by rifampicin.
Penetration through the OM is the rate-limiting step in the action of rifampicin (1).In Enterobacterales, LPS and ECA are the key surface components that limit the accessibility of the hydrophobic compounds to OM (30,31,33).Integral OMPs also play a role in OM permeability owing to the selective transport of small hydrophilic molecules (29,30).Our results showed that the BAM complex contributes substantially to the impermeability of rifampicin.However, we found that mutants deficient in ECA biosynthesis universally exhibited increased fitness under sub-MIC rifampicin, suggesting that ECA may not function as a barrier to rifampicin entry (32).Intriguingly, our results indicated that the impact of LPS on rifampicin efficacy is mild and may have opposing effects between sub-MIC and above-MIC concentrations of rifampicin.These results suggest that the contribution of these permeability barriers to rifampicin entry differs substantially.single wig file.According to the file, Pearson correlation and distribution of reads can be calculated and visualized with Python and R. Fitness change was determined by carrying out a pairwise comparison of each rifampicin-treated sample with the input data.Reads in the 5% N-terminal and 5% C-terminal of the gene sequence were discarded (61).The "Resampling" method of TRANSIT software (v3.2.0) was used to identify mutants that were differentially represented between input and post-treatment samples.Read counts, P-value (adjusted by using the method of FDR), and log 2 FC between the input and post-treatment were calculated using default parameters.The genes with an adj P-value of less than 0.05 and log 2 FC of more than 1 or less than −1 were selected.

Gene enrichment analysis
Pathway enrichment analysis was performed on the website (https://david.ncifcrf.gov).Enrichment in pathway terms was calculated for each gene cluster.GO terms with an FDR value of less than 0.1 were selected.

Gene deletion and complementation
Single gene deletion in E. coli was constructed by allelic transduction from Keio collection strains using classical P1 phage transduction followed by Kan R cassette excision (62).For the deletion of nuoA-N, DNA fragments containing flanking regions of the nuoA-N cluster and Kan R cassette were PCRamplified and transformed into E. coli using electroporation.Recovered cells were selected for kanamycin-resistant homologous recombinants.To avoid potential offtarget effects, the selected mutant containing the ΔnuoA-N::Kan R allele was used as a donor for P1 transduction to generate the ΔnuoA-N mutant.The plasmid was cured, and the Kan R cassette was removed to generate the unmarked nuoA-N deletion mutant.The mutant genotype was verified by PCR and sequencing.For complementation of the gene deletion mutants, the coding region of the target gene was PCRamplified and cloned into a low-copy plasmid pZS*24 (63).The resulting plasmid was transformed into the mutant strain by electroporation.

Antibiotic susceptibility
To measure MIC, strains were grown to the exponential phase and diluted to ~10 5 CFU per mL in 200-µL LB media containing antibiotics at indicated concentrations.Growth was monitored by OD 600 and expressed as percent of growth relative to the untreated control.Ninety percent inhibitory concentration (MIC 90 ) was determined by plotting the percent of growth versus concentration of antibiotic.The growth inhibition dynamics were approximated by nonlinear regression using the inhibitor versus response model of GraphPad Prism 9.0.Three biological replicates were performed.

Antibiotic killing assay
Bacterial cultures were grown overnight in LB and then diluted 1:200 in 10-mL LB in 50-mL flasks.For the complementary strains, cultures were grown in LB containing 20 µg/mL kanamycin; gene expression was induced by 1 mM isopropyl-1-thio-β-D-galac topyranoside upon inoculation of the seeding culture.At an OD 600 of 0.3 to 0.4, aliquots of 1 mL culture (in a 14-mL tube) were treated with antibiotics at 37°C.Aliquots were taken, serially diluted, and plated onto LB agar plates to determine CFU.Survival was expressed as the ratio compared with pre-treatment.

ATP qualification
ATP concentration was measured by using the BacTiter-Glo Microbial Cell Viability Assay kit (Promega) and FlexStation 3 (Molecular Devices) following the manufacturer's protocols.Samples (100 µL) at an OD 600 of 0.4 to 0.6 were mixed with an equal volume of BacTiter-Glo reagent in 96-well opaque plates and incubated at room temperature for 5 minutes.The relative light units were recorded by FlexStation 3 and normalized by OD 600 .

NADH/NAD + assay
The qualification of NADH and NAD + was performed by using the NAD/NADH-Glo Assay kit (Promega) and FlexStation 3 following the manufacturer's protocols.The samples (100 µL) at an OD 600 of 0.4 to 0.6 were mixed with an equal volume of NAD/NADH-Glo detection reagent in 96-well opaque plates and incubated at room temperature with gentle shaking for 60 minutes.RLU values were recorded by FlexStation 3 and normal ized by OD 600 .

Statistical analysis
Statistical analysis was performed on log 10 -transformed data (for antibiotic killing experiments) or untransformed data (for NADH and ATP assays) using unpaired twotailed Student's t-test using GraphPad Prism 9.0 software.A statistical difference between wild type and a gene deletion mutant is marked (*P < 0.05; **P < 0.01; ***P < 0.001).

FIG 1
FIG 1 Quality control of the Tn-Seq and screening strategy.(A) Correlation coefficients of the number of reads per insertion site for two biological replicates of the input transposon library.Shown are log 10 (reads number +1).(B) Distribution of transposon sequence reads across the E. coli MG1655 genome.Tn-Seq read data from the input transposon library are shown.(C) Survival of the pooled library exposed to rifampicin at the concentrations of 0.25×, 1×, 4×, and 20× MIC.Survival was determined by monitoring colony-forming units (CFUs) and expressed as the ratio compared with pre-treatment.Data are shown as the mean ± SE of three independent experiments.Samples collected before treatment and at 1 and 3 h post-treatment were subjected to Tn-Seq (arrows).(D) The cluster heatmap shows Pearson correlation coefficient matrices that measure the relevance of the corresponding samples.

FIG 2
FIG 2 Genes and pathways identified under sub-MIC.(A) The heatmap of the log 2 FC values of enriched genes selected under 0.25× MIC.Enriched pathways are labeled on the left: OMA, outer membrane assembly; ECA, biosynthesis of enterobacterial common antigens.(B and C) Rifampicin susceptibility (MIC 90 ) of E. coli MG1655, the mutant strains, and complementary strains harboring the indicated plasmids.The line represents the nonlinear regression of growth that is dependent on the concentration of rifampicin.Data are shown as the mean ± SE of at least three independent experiments.

FIG 3
FIG 3 Genes and pathways identified under bactericidal concentrations of rifampicin.(A) Heat map of log 2 FC values of the identified genes.Genes not identified in the condition are shown in white.(B) Pathways enriched under bactericidal concentrations of rifampicin.eco03018 (RNA degradation), GO:0009244 (LPS core region biosynthetic process), eco01200 (carbon metabolism), GO:0043213 (bacteriocin transport), GO:0006811 (ion transport), eco00190 (oxidative phosphorylation), and GO:0009314 (response to radiation).

FIG 4
FIG 4 Iron metabolism modulates rifampicin lethality in a hydroxyl radical-independent manner.(A) Heatmap of the selected genes involved in iron metabo lism.GO:0006811, ion transport; GO:0043213, bacteriocin transport.(B) Illustration of the E. coli Fe acquisition system selected in our screen.The TonB-ExbB-ExbD complex provides energy to Fep and Fec.(C) Rifampicin susceptibility (MIC 90 ) of E. coli strains.(D through I) Survival of the indicated E. coli strains exposed to 32 mg/L rifamycin with (H) or without the addition of 100 mM thiourea.Data are shown as the mean ± SE of at least three independent experiments.*P < 0.05, **P < 0.01, and ***P < 0.001 using an unpaired t-test compared to the counterpart of wild type.

FIG 5
FIG 5 Survival of mutants deficient in DSB-repair function or transcription-coupled DNA repair (TCR) upon exposure to rifampicin.(A) Heatmap of the selected genes related to DNA repair and replication.(B and E) Rifampicin susceptibility (MIC 90 ) of indicated E. coli strains.(C-D and F-H) Survival of the indicated E. coli strains treated with 32 mg/L rifampicin (C, D, F, and H) or 100 mg/L ampicillin (G).Data are shown as the mean ± SE of at least three independent experiments.*P < 0.05, **P < 0.01, and ***P < 0.001 using an unpaired t-test compared to the counterpart of wild type.

FIG 6
FIG 6 Sustained cellular respiration counteracts rifampicin-induced killing.(A) Survival of E. coli strains treated with 32 mg/L rifampicin.(B) Survival of E. coli strains treated with 100 mg/L ampicillin or 16 mg/L gentamycin.(C and D) The NADH/NAD + and ATP levels in E. coli MG1655 and the ΔnuoA-N mutant.RLU, relative light units.(E) Survival of E. coli strains treated with 32 mg/L rifampicin.Data are shown as the mean ± SE of at least three independent experiments.*P < 0.05, **P < 0.01, and ***P < 0.001 using an unpaired t-test compared to the counterpart of wild type.