A multilayered repair system protects the mycobacterial chromosome from endogenous and antibiotic-induced oxidative damage

Significance

Oxidative stress is a common insult to all living organisms, in part because of its potent DNA-damaging effects leading to cell death and mutagenesis. In bacteria, antibiotics are one source of oxidative stress. Some DNA repair systems that counteract oxidative-induced mutagenesis have been described in Escherichia coli as participating in antibiotic killing action. However, in mycobacteria, which include agents of tuberculosis and leprosy, roles of these oxidative DNA repair systems are poorly understood because multiple and redundant enzymes are encoded. This study definitively dissects this redundancy using genetic approaches. We uncover an intricate set of DNA repair systems that defend the mycobacterial chromosome against endogenous oxidative mutagenesis. While some of them protect against antibiotic-induced oxidative killing, others participate in it.

Abstract

Oxidative damage to DNA is a threat to the genomic integrity and coding accuracy of the chromosomes of all living organisms. Guanine is particularly susceptible to oxidation, and 8-oxo-dG (OG), when produced in situ or incorporated by DNA polymerases, is highly mutagenic due to mispairing with adenine. In many bacteria, defense against OG depends on MutT enzymes, which sanitize OG in the nucleotide pool, and the MutM/Y system, which counteracts OG in chromosomal DNA. In Escherichia coli, antibiotic lethality has been linked to oxidative stress and the downstream consequences of OG processing. However, in mycobacteria, the role of these systems in genomic integrity and antibiotic lethality is not understood, in part because mycobacteria encode four MutT enzymes and two MutMs, suggesting substantial redundancy. Here, we definitively probe the role of OG handling systems in mycobacteria. We find that, although MutT4 is the only MutT enzyme required for resistance to oxidative stress, this effect is not due to OG processing. We find that the dominant system that defends against OG-mediated mutagenesis is MutY/MutM1, and this system is dedicated to in situ chromosomal oxidation rather than correcting OG incorporated by accessory polymerases (DinB1/DinB2/DinB3/DnaE2). In addition, we uncover that mycobacteria resist antibiotic lethality through nucleotide sanitization by MutTs, and in the absence of this system, accessory DNA polymerases and MutY/M contribute to antibiotic-induced lethality. These results reveal a complex, multitiered system of OG handling in mycobacteria with roles in oxidative stress resistance, mutagenesis, and antibiotic lethality.

In living cells, DNA is subjected to a constant barrage of oxidative damage from environmental factors or reactive oxygen species (ROS) produced by cellular metabolism (1, 2). Guanine bases are particularly vulnerable to oxidative stress, readily forming 7,8-dihydro-8-oxoguanine (8-oxo-G or OG) in the free nucleotide pool and in chromosomal DNA (3). OG is the most frequent oxidative lesion in DNA experienced by living cells and is generated either by direct oxidation of guanine in the chromosome or incorporation of oxidized nucleotides by DNA polymerases (SI Appendix, Fig. S1) (4). Although genomic OG does not block DNA replication, it is highly mutagenic because of its ambiguous pairing with cytosine and adenine.

To counteract the harmful consequences of OG, most bacteria deploy the GO system (SI Appendix, Fig. S1) composed of MutM, MutY, and MutT enzymes (4, 5). The MutM glycosylase removes genomic OG paired with cytosine (6–8), whereas MutY excises adenine mispaired with OG (9, 10). Individual loss-of-function mutM and mutY mutants in Escherichia coli display high levels of G to T or C to A transversion mutations (11, 12) and a synergistic augmentation of mutagenesis when both MutM and MutY are lost (13, 14). MutT hydrolyzes 8-oxo-deoxyguanosine triphosphate (dGTP) to 8-oxo-deoxyguanosine monophosphate in the deoxynucleoside triphosphate (dNTP) pool, thereby preventing its incorporation by DNA polymerases (15, 16). In E. coli, mutT deletion increases the mutation frequency between 100- and 10,000-fold, depending on the marker tested, by inducing A to C or T to G mutations (16).

In addition to their antimutagenic role, OG genomic incorporation is a key factor of cell death from antibiotics (17, 18). In 2007, Kohanski et al. (19) showed that bactericidal antibiotics share a common mechanism of cellular death by inducing the production of ROS. Foti et al. (20) proposed in 2012 that cytotoxicity of beta-lactams and quinolones in E. coli is caused, in part, by DNA double-strand break formation due to incomplete genomic excision of OG. After antibiotic treatment, OG is incorporated in DNA by translesional polymerases, and double-strand breaks are induced by MutY and MutM activity. Less is known about the role of OG in antibiotic action in mycobacteria. However, a recent paper showed that antibiotic-induced oxidation of deoxycytidine triphosphate (dCTP) causes DNA damage and cell death in stationary-phase mycobacteria (21), clearly establishing that antibiotic-induced oxidative damage occurs in mycobacterial cells. However, the mechanisms of antibiotic-induced oxidative DNA damage, and in particular, the role of OG-mediated killing in mycobacteria, remain unknown.

Mycobacterium tuberculosis (TB) is the causative agent of tuberculosis, which kills more than 1.5 million people per year. During infection, the bacterium is exposed to ROS released by host macrophages (22). Due to the high G + C content and the absence of canonical mismatch repair in TB, GO system could be a key mediator of genome maintenance of the pathogen during infection (23, 24). Interestingly, TB and the nonpathogenic model organism Mycobacterium smegmatis elaborate a seemingly complex system of defense against OG. M. tuberculosis and M. smegmatis encode four MutTs (MutT1 to -4), two MutMs (MutM1 and MutM2), and one MutY (23, 24). Biochemical analysis of MtuMutT1 demonstrated a clear 8-oxo-GTPase activity in vitro, and its expression in E. coli rescued the oxidative stress sensitivity and mutagenesis phenotypes of the mutT mutant (25, 26). mutT1 deletion was reported to increase the mutation frequency more than 10-fold in both M. tuberculosis and M. smegmatis, enhancing expected A to C or T to G mutations (25). Biochemical analysis of MtuMutT2 and MtuMutT4 indicates that MutT2 can hydrolyze dGTP and dCTP (25, 27, 28), whereas MutT4 prefers deoxyadenosine triphosphate (25) and that mutT4 deletion increases A to G or T to C mutations (50-fold) in M. smegmatis but not in TB (25). In vitro, MutY possesses the same substrate specificities as its homolog in E. coli (29, 30), whereas MutM1 excises OG with a higher efficiency when paired with G, C, or T rather than with A (29, 31). Individual mutM1, mutM2, or mutY deletions have minimal effects on M. smegmatis mutation frequency (29, 30, 32). However, the triple mutant shows a strong enhancement of the expected G to T or C to A mutations (32).

In order to uncover the interplay between the different components of mycobacterial GO system, we performed a comprehensive genetic characterization of MutTs, MutMs, MutY, and accessory polymerases in M. smegmatis. We constructed single and multiple mutT mutants to analyze the relative contributions of these genes to mutagenesis and oxidative stress tolerance. We reveal the dominant role of MutT4 in H₂O₂ resistance but did not detect any increase of OG-dependent mutations even when all MutTs are absent. In contrast, we observed a dramatic enhancement of spontaneous G to T or C to A mutations in the ΔmutYΔmutM1 double mutant, whereas the mutM2 deletion, alone or in combination, did not impact the mutation frequency. Loss of mutTs or accessory polymerases did not impact the mutation frequency in the ΔmutYΔmutM12 mutant, demonstrating that the cellular load of OG in mycobacterial cells arises from in situ oxidation of the chromosome rather than incorporation of oxidized guanine from the dNTP pool. We also establish that antibiotic-induced killing is not due to OG genomic incorporation/excision in wild-type M. smegmatis. However, we uncover a role for OG in antibiotic lethality when MutTs are disabled, suggesting that in certain circumstances of high antibiotic-induced nucleotide pool oxidation, OG incorporation mediates antibiotic action in mycobacteria.

Materials and Methods

Bacterial Strains and Growth Conditions.

Strains used in this study are listed in SI Appendix, Table S1. E. coli strains were grown at 37 °C in Luria–Bertani medium. M. smegmatis strains were grown at 37 °C in Middlebrook 7H9 medium supplemented with 0.5% glycerol, 0.5% dextrose, and 0.1% Tween 80. Antibiotics were used at the following concentrations: 2 μg mL⁻¹ streptomycin (sm) and 20 μg mL⁻¹ kanamycin. To perform growth curves, overnight cultures were diluted to an optical density at 600 nm (OD₆₀₀) of 0.05 and grown for 8 h. Cells were again diluted in fresh medium (OD₆₀₀ = 0.001), and growth was measured by monitoring OD₆₀₀ every 3 h from OD₆₀₀ = 0.05. Doubling times have been calculated using trendline equation drawn for the exponential phase of these growth curves.

Strain and Plasmid Constructions.

All plasmid constructions are listed in SI Appendix, Table S2 and constructed in E. coli DH5α. Oligonucleotides used in this work are detailed in SI Appendix, Table S3. The absence of mutations in constructs was verified by DNA sequencing. Plasmids have been introduced into M. smegmatis by electrotransformation. Markerless and in-frame gene deletions were performed as described in Barkan et al. (33) using pAJF067 derivatives containing ∼500-bp regions flanking the gene to be deleted: mutT1 (MSMEG_2390), mutT2 (MSMEG_5148), mutT3 (MSMEG_0790), mutT4 (MSMEG_6927), mutM1 (MSMEG_2419), mutM2 (MSMEG_5545), mutY (MSMEG_6083), dnaE2 (MSMEG_1633), dinB1 (MSMEG_3172), dinB2 (MSMEG_2294/MSMEG_1014), dinB3 (MSMEG_6443), and recA (MSMEG_2723). Open reading frame-flanking DNA fragments were amplified by PCR using M. smegmatis mc²155 genomic DNA as template and the oligonucleotides listed in SI Appendix, Table S3 as primers and were cloned into pAJF067 digested with NdeI using In-Fusion recombination cloning (Takara). Because a second copy of dinB2 is encoded among a large genomic duplicated region (34), a first dinB2 copy together with its ∼1,000-bp flanked regions has been primarily deleted, whereas the only coding seqeunce of the dinB2 second copy has been deleted in a second step. Deletions of genes have been checked by PCR as shown in SI Appendix, Fig. S2. For complementation and overexpression plasmids, ORFs (together with their 5′ flanked region [∼500 bp] for complementation) were amplified by PCR using M. smegmatis mc²155 or M. tuberculosis Erdman genomic DNA as template and the oligonucleotides listed in SI Appendix, Table S3 as primers and were cloned into pDB60 digested with EcoR1 (complementation) or pAJF266 digested with HindIII (overexpression) using recombination-based cloning (In-Fusion).

H₂O₂ and Antibiotic Sensitivity Assays.

For disk diffusion assays, bacteria were grown to exponential phase, diluted at an OD₆₀₀ of 0.01 in 3 mL prewarmed top agar (7H9; 6 mg mL⁻¹ agar), and plated on 7H10. A filter disk was placed and spotted with 2.5 µL of 10 M H₂O₂, 50 mg mL⁻¹ rifampicin (rif), 200 mg mL⁻¹ sm, 10 mg mL⁻¹ ciprofloxacin (cip), or 10 mg mL⁻¹ isoniazid (INH). After incubation at 37 °C for 48 h, the diameter of the growth inhibition zone was measured.

For agar-based assays, bacteria were grown to exponential phase and diluted at an OD₆₀₀ of 0.1. Serial dilutions were performed from 10⁰ to 10⁻⁵ in 7H9, and 5 µL each dilution was plated on 7H10 or 7H10 supplemented with 0.5 mM H₂O₂, 5 µg mL⁻¹ rif, or 0.10 µg mL⁻¹ cip. Experiments were imaged after a 5-d incubation at 37 °C.

For killing assays, exponential-phase cultures were diluted at an OD₆₀₀ of 0.1. Antibiotics were added to cultures at the following concentrations: 100 µg mL⁻¹ rif, 1 µg mL⁻¹ cip, or 25 µg mL⁻¹ INH. After incubation at 37 °C with shaking for 6 h, cells were collected by centrifugation and washed in 7H9. Serial dilutions from 10⁰ to 10⁻⁵ were performed, and 5 µL each dilution was cultured on 7H10 agar. Colonies were counted after incubation at 37 °C for 3 d. For each strain, survival percentage was calculated by using the ratio of (colony forming units [CFU] after the antibiotic treatment/CFU number obtained before the antibiotic treatment) × 100.

Mutagenesis Assay.

Mutagenesis assays were performed by measuring the number of rifampicin-resistant (rif^R) cells in a population. Bacteria were grown to exponential phase and diluted at a calculated OD₆₀₀ of 0.0005. After 16 h of incubation at 37 °C, 5 ml of culture (OD600 around ∼0.3) was concentrated 20 times by centrifugation and pellet resuspention in 250 µl of Middlebrook 7H9 broth medium and 200 µL cultured on 7H10 100 µg mL⁻¹ rif; 100 µL of 10⁻⁶ serial dilution was cultured on 7H10. For H₂O₂-induced mutagenesis, cells were treated with H₂O₂ (2.5 mM) for 2 h, washed, and incubated for 4 h at 37 °C in fresh medium. The mutation frequency has been expressed by the mean number of rif^R cells per 10⁸ CFU from independent cultures. For each strain and condition, the number of independent cultures used to measure the mutation frequency is indicated by the number of gray dots in each bar of graphs (Figs. 2D and 3 A and B). For determination of mutation spectrum, the rpoB gene in rif^R colonies was amplified and sequenced using primers listed in SI Appendix, Table S3.

Fig. 2.

MutT4 is the dominant MutT enzyme for defense against oxidative stress but is not involved in the prevention of OG-dependent mutations. (A and B) M. smegmatis H₂O₂ sensitivities of the indicated strains determined by disk diffusion assay. Results show the diameter of the inhibition zone, expressed relative to the WT set as 100%. (C) Serial dilutions of M. smegmatis cultures on 7H10 (−) or 7H10 containing H₂O₂ (+). (D) Frequency of spontaneous and H₂O₂-induced rif^R colonies. (E) Mutation spectrum expressed in percentage of rif^R mutations in the rpoB gene. Sequencing was performed from at least six independent experiments, and the number of clones sequenced for each condition is listed in the center of the ring. Results shown in bar graphs are means (± SEM) of data obtained from independent experiments symbolized by gray dots. *P < 0.05; **P < 0.01; ***P < 0.001.

Fig. 3.

MutY/MutM1 is the dominant system for the prevention of OG-dependent mutations in M. smegmatis. Frequency of (A) spontaneous and (B) H₂O₂-induced rif resistance (rif^R). Results shown are means (± SEM) of data obtained from independent experiments symbolized by gray dots. (C) Mutation spectrum expressed in percentage of rif^R mutations in the rpoB gene. WT data are the same as in Fig. 2. Sequencing was performed from at least six independent experiments, and the number of clones sequenced for each condition is given within the ring. (D) WT and ΔmutYΔmutM12 H₂O₂ sensitivities by disk diffusion. Pictures are representative of nine independent experiments, and numbers are means (± SEM) of inhibition zone diameters in millimeters. **P < 0.01; ***P < 0.001.

Western Blot.

Lysates were prepared from 2 mL culture at OD₆₀₀ of 0.4. For protein detection, anti-RpoB (Biolegend; 663905; research resources identifier: AB_2566583) and anti-RecA (Pocono Rabbit Farm & Laboratory) (35) were incubated for 1 h at 1:10,000 dilutions.

Statistical Tests.

For all experiments except antibiotic killing assays, statistical analyses were performed using a one-way ANOVA and a Bonferroni posttest. For antibiotic killing assays, Student t tests have been performed between pairs of interest. A statistical difference between the wild type (WT) strain and another is marked by a star above the column, whereas a statistical difference between two other strains is specifically shown by lines connecting both columns (*P < 0.05; **P < 0.01; ***P < 0.001).

Data Availability.

All data and protocols are available in the text and SI Appendix. All bacterial strains are available on request.

Results

MutT4 Is the Only MutT Required for Optimal Growth in Mycobacteria.

M. smegmatis encodes four putative MutTs orthologs. To understand the relative contributions of these proteins to handling of nucleotide oxidation in vivo, we constructed M. smegmatis strains lacking each mutT gene (ΔmutT1, ΔmutT2, ΔmutT3, or ΔmutT4) and all mutT genes in combination (ΔmutT1-4). The doubling time of both ΔmutT4 and ΔmutT1-4 mutants increased by more than 10 min in liquid medium (8% increase), a finding that was also evinced by smaller colony size on agar media (Fig. 1 A and B). The ΔmutT4 growth defect was complemented by an ectopic copy of M. smegmatis or M. tuberculosis mutT4 (Fig. 1C), suggesting a conservation of MutT4 activities between fast- and slow-growing mycobacteria. To test the role of MutT4 phosphohydrolase activity in these phenotypes, we complemented the ΔmutT4 strain with an allele of mutT4 encoding a mutation in the conserved nudix-hydrolase motif (E162A) that is required for E. coli MutT activity (E53A) (36). Expression of mutT4^E162A in the mutT4 mutant failed to restore WT growth rate (Fig. 1C). Overexpression of mutT1, mutT2, or mutT3 in the ΔmutT4 strain also did not restore the WT growth rate, whereas mutT4 overexpression did (SI Appendix, Fig. S5A), suggesting a specificity of the MutT4 activity. RecA is the primary mediator of both homologous recombination and SOS response and therefore, mediates a broad range of DNA damage responses. Accordingly, the slow growth phenotype of the mutT4 deletion was exacerbated by the loss of recA (15% increase in doubling time), indicating the presence of DNA damage in the ΔmutT4 mutant that is compensated by RecA (Fig. 1A). Moreover, we detected a strong induction of RecA in the ΔmutT4 strain (Fig. 1D and SI Appendix, Fig. S3), indicating an activation of the DNA damage response (37) due to mutT4 loss.

Fig. 1.

MutT4 is the only MutT required for optimal growth and genome integrity in mycobacteria. M. smegmatis doubling times calculated from growth curves in (A) 7H9 or (C) 7H9 supplemented with sm. Results shown are means (± SEM) of data obtained from independent experiments symbolized by gray dots. For complementation experiments, the empty integrative pDB60 plasmid or carrying different versions of the mutT4 gene was introduced into WT and ΔmutT4 strains, and cells were grown in presence of sm: WT (PDS62), ΔmutT4 (PDS497), ΔmutT4+mutT4^smeg (PDS498), ΔmutT4+mutT4^E162Asmeg (PDS799), and ΔmutT4+mutT4^TB (PDS499). (B) M. smegmatis colonies after 3 d on agar plates (7H10) incubated at 37 °C. (D) Western blot anti-RecA/RpoB from M. smegmatis lysates. The full blot is presented in SI Appendix, Fig. S3. Pictures and blots are representative of at least three independent experiments. ***P < 0.001.

MutT4 Confers Oxidative Stress Tolerance.

We then tested the role of MutTs in mycobacterial resistance to oxidative stress. Using a disk diffusion assay, allowing detection of both bacteriostatic and bactericidal effects of hydrogen peroxide, the deletion of mutT1, mutT2, or mutT3 did not significantly increase the sensitivity of M. smegmatis to H₂O₂ (Fig. 2A). In contrast, whereas the ΔmutT1-3 mutant was slightly more sensitive than the WT strain, loss of mutT4 conferred a clear sensitivity to H₂O₂. The ΔmutT1-4 strain was no more sensitive than ΔmutT4, indicating that MutT4 is the dominant MutT enzyme in M. smegmatis for resistance to oxidative stress. The ΔmutT4 H₂O₂ sensitivity was complemented by the introduction of an ectopic copy of M. smegmatis or M. tuberculosis mutT4 but not by the catalytic mutant (Fig. 2B). We obtained comparable results by culturing serial dilutions of these strains on agar medium containing H₂O₂ (Fig. 2C and SI Appendix, Fig. S4A). In this agar-based assay, the number of colonies formed by the WT strain was reduced by H₂O₂, and colonies were smaller. Although the mutT4 mutant formed a similar number of colonies as the WT strain on plates containing H₂O₂, the mutT4 deletion enhanced the impact of H₂O₂ on the colony size. This suggests that the mutT4 deletion exacerbates the bacteriostatic effect of oxidative stress. The introduction of an ectopic copy of mutT4 in the ΔmutT4 mutant restored the H₂O₂ sensitivity of the strain to near-WT level. In addition, the triple mutT1-3 deletion, alone or in combination with mutT4 deletion, did not increase H₂O₂ sensitivity. These data demonstrate, under the conditions tested, a dominant role for mutT4 among MutTs in oxidative stress tolerance in M. smegmatis.

MutTs Do Not Prevent OG-Dependent Mutations.

The data above indicate that MutT4 is the dominant MutT enzyme that defends against oxidative stress in M. smegmatis, a role that parallels E. coli MutT. Accordingly, we next investigated the role of mutTs in OG-dependent mutation avoidance. We measured the number of rif^R CFU in WT and mutT-deficient strains. We found approximatively two spontaneous rif^R per 10⁸ CFU in the WT strain but did not detect any significant impact of loss of any or all mutTs on mutation frequency (Fig. 2D). When exposed to oxidative stress, we observed a 10-fold increase of the mutation frequency after H₂O₂ treatment, but again, mutT deletions did not impact the number of rif^R in the bacterial population (Fig. 2D). To determine if the appearance of OG mutagenic signatures (A > C or T > G and G > T or C > A) was affected by loss of MutTs despite the lack of effect on overall mutation rate, we sequenced the rifampin resistance-determining region of the rpoB gene. We did not observe an increase of OG-associated mutations in any mutT mutants either at baseline or with oxidative stress (Fig. 2E). H₂O₂ treatment increased the proportion of G > A or C > T and G > C or C > G mutations by twofold in WT cells but did not induce mutations characteristic of OG. During oxidative stress, mutT4 deletion doubled the proportion of A > G or T > C observed in comparison with the WT, with an additive effect of the mutT1-3 triple deletion (3.5-fold increase in the ΔmutT1-4). These data show that, in conditions of oxidative stress, 1) M. smegmatis does not accumulate OG-associated chromosomal mutations, implying a robust system to avoid OG mutagenesis, and 2) the MutT system does not play a preventive role against OG-dependent mutations, despite having an important role in oxidative stress resistance.

MutM1 and MutY but Not MutM2 Mediate OG-Dependent Mutation Avoidance.

The absence of enhanced mutagenesis in mutT-deficient strains could be explained by a high efficiency of GO system excision pathways (MutY and MutMs) in mycobacteria that would counteract the mutagenic effect of OG incorporation or in situ oxidation (SI Appendix, Fig. S1). To test this possibility, we deleted mutY, mutM1, and mutM2 and tested the effects on mutation frequency. Individually, none of mutY, mutM1, or mutM2 deletions significantly increased the spontaneous or H₂O₂-induced mutation frequency (Fig. 3 A and B). Despite the lack of effect on overall mutation frequency, we observed two- and fourfold augmentation of the proportion of expected OG mutations (G > T or C > T) in the ΔmutY and the ΔmutM1 mutants, respectively (Fig. 3C). Surprisingly, we observed a distinct mutation spectrum in the ΔmutM2 mutant: twofold increase of A > G or T > C mutations.

Although the double ΔmutM1ΔmutM2 and ΔmutYΔmutM2 mutants did not show any significant increase of spontaneous mutation frequency, we found a 10-fold augmentation of the mutation frequency in the ΔmutYΔmutM1 and ΔmutYΔmutM12 mutants. The mutation spectrum in the ΔmutYΔmutM12 strain was dominated by G > T or C > A mutations, which are characteristic of defective excision of chromosomal OG or mispaired adenine by MutM or MutY, respectively (SI Appendix, Fig. S1). However, the deletion of mutY, mutM1, and mutM2 did not impact the M. smegmatis sensitivity to oxidative stress (Fig. 3D), and H₂O₂ treatment did not significantly modify the mutation frequency or the rpoB mutation signature in the triple mutant. These data reveal that 1) MutY and MutM1, but not MutM2, are the primary mediators of the OG-dependent mutation avoidance in M. smegmatis and 2) although MutY/MutM1 prevents OG-dependent mutagenesis, the source of this oxidative stress is from endogenous sources, likely cellular metabolism.

Direct Oxidation in Chromosomal DNA Is the Primary Source of OG.

The presence of OG in DNA can be the consequence of genomic guanine oxidation or the incorporation of oxidized OG in dNTP pool by DNA polymerases (SI Appendix, Fig. S1). In E. coli, the translesion polymerases UmuDC and DinB can utilize OG (20, 38), and in mycobacteria, at least DinB2 has a similar propriety (39). To investigate the provenance of OG in mycobacteria, we measured the impact of the loss of the putative translesion polymerases (DnaE2 and DinBs) and dNTP sanitizers (MutTs) on the elevated mutation frequency in the ΔmutYM12 genetic background. The spontaneous mutation frequency and mutation spectrum were comparable in ΔmutYM12, ΔmutYM12ΔmutT1-4, and ΔmutYM12ΔdnaE2dinB123 strains (Fig. 3 A and C), indicating that the OG load handled by the MutY/M system does not originate in oxidation of the nucleotide pool or the utilization of these nucleotides by accessory polymerases. We observed a modest decrease of the mutation frequency in the ΔmutYM12ΔdnaE2dinB123 mutant compared with ΔmutYM12 after H₂O₂ treatment (90 vs. 48 rif^R per 10⁸ CFU) (Fig. 3B), possibly indicating polymerase utilization of oxidized nucleotides under oxidative stress.

These data indicate that direct oxidation of chromosomal guanines is the main source of the cellular OG load, whereas DNA incorporation of OG from oxidized dNTP pool is a minor pathway in M. smegmatis.

Excision of Genomic OG Is Not the Primary Cause of Antibiotic Killing Effect in Mycobacteria.

Although genomic OG is highly mutagenic, its presence in DNA is not necessarily lethal for cells, as replicative DNA polymerases can bypass this lesion (4). However, accumulation of closely spaced OG in DNA can cause double-strand breaks due to incomplete excision repair by the MutY/MutM system, a process that has been linked to the lethality of antibiotics in E. coli (18, 20). To test if a comparable killing mechanism is present in mycobacteria, we measured the antibiotic sensitivity of both mutY/mutMs and accessory polymerase mutants. Using a disk diffusion assay, the ΔdnaE2dinB123 and ΔmutYM12 mutants showed the same rif (Fig. 4A), cip (Fig. 4B), INH (Fig. 4C), and sm (Fig. 4D) sensitivity as WT. Because disk diffusion assays do not distinguish between growth inhibition and killing, this assay may obscure the role of specific pathways in these two distinct processes. We specifically studied the impact of GO system loss on the bactericidal effect of antibiotics by measuring the survival of M. smegmatis after rif, cip, and INH treatments. Although these three antibiotics induced time-dependent killing in the WT strain, deletion of mutY/mutMs and accessory polymerases did not impact the antibiotic killing effect (Fig. 4 E–G). These results strongly suggest that bactericidal effects of antibiotics in mycobacteria are not due to a higher frequency of OG incorporation/excision.

Fig. 4.

Bactericidal effect of antibiotics is not due to OG genomic excision by the MutY/MutM system in M. smegmatis. (A) Rif, (B) cip, (C) INH, and (D) sm sensitivities of M. smegmatis determined by disk diffusion. Results are means (± SEM) obtained from independent experiments symbolized by gray dots showing the diameter of the inhibition zone, expressed relative to WT (set as 100%). (E) Rif, (F) cip, and (G) INH killing effect in M. smegmatis WT (circles), ΔdnaE2ΔdinB123 (squares), and ΔmutYΔmutM12 (triangles) strains determined by CFU enumeration after 6, 24, or 48 h of treatment in liquid culture. Results shown are means (± SEM) of data obtained from at least six independent experiments. *P < 0.05.

MutTs Protect Mycobacteria against Bacteriostatic and Bactericidal Effects of Antibiotics.

The absence of OG-dependent killing during antibiotic treatment could be explained by the efficiency of the pool sanitizer system (MutTs) in mycobacteria. To test this idea, we measured the sensitivity of ΔmutT mutants to various antibiotics. By disk diffusion, mutT1-3 deletion did not increase sensitivity to rif (Fig. 5A), cip (Fig. 5E), INH (Fig. 5I), or sm (Fig. 5K). Similarly, the bacteriostatic effect of rif and cip on agar media for the ΔmutT1-3 strain was comparable with the WT strain (SI Appendix, Fig. S4 B and C). However, the ΔmutT4 and ΔmutT1-4 mutants showed 15 and 3% increase of the growth inhibition zone size for rif and cip in comparison with the WT strain, respectively (Fig. 5 A and E). We observed a similar increase of the rif and cip sensitivity in the absence of mutT4 in the agar-based assay (SI Appendix, Fig. S4 B and C). Genetic complementation by M. smegmatis mutT4, but not the catalytic mutant, restored WT rif and cip resistance (Fig. 5 B and F and SI Appendix, Fig. S4 B and C). Partial reversion of the rif ΔmutT4 sensitivity was observed by expressing the M. tuberculosis mutT4. Furthermore, overexpression of mutT4, but not mutT1, mutT2, or mutT3, restored rif sensitivity to near-WT levels, demonstrating a specificity of the MutT4 activity for this phenotype (SI Appendix, Fig. S5B). Finally, none of the ΔmutT mutants were more sensitive to sm or INH (Fig. 5 I and K). These results indicate that MutT4 has a role in defending against antibiotic lethality and/or bacteriostasis in mycobacteria.

Fig. 5.

MutT4 protects M. smegmatis against bacteriostatic and bactericidal effects of antibiotics. (A and B) Rif, (E and F) cip, (I) INH, and (K) sm sensitivities of M. smegmatis determined by disk diffusion. Results are means (± SEM) of data obtained from independent experiments symbolized by gray dots showing the diameter of the inhibition zone, expressed relative to the WT (set as 100%). Bactericidality of rif (C and D), cip (G and H), and INH (J) in M. smegmatis as determined by CFU enumeration after 6 h of treatment in liquid culture. Results shown are means (± SEM) of data obtained from independent experiments symbolized by gray dots. For complementation experiments, the empty integrative pDB60 plasmid carrying different versions of the mutT4 gene was introduced into WT, ΔmutT4, and ΔmutT1-4 strains: WT (PDS62), ΔmutT4 (PDS497), ΔmutT4+mutT4^smeg (PDS498), ΔmutT4+mutT4^{E162A smeg} (PDS799), ΔmutT4+mutT4^TB (PDS499), ΔmutT1-4 (PDS829), ΔmutT1-4+mutT4^smeg (PDS830), ΔmutT1-4+mutT4^TB (PDS831), and ΔmutT4+mutT1-4^{E162A smeg} (PDS833). *P < 0.05; **P < 0.01; ***P < 0.001.

To determine if the antibiotic sensitivity of ΔmutT4 reflects bacterial killing, we measured the survival of WT and mutT mutants after 6 h of rif, cip, or INH treatments in liquid medium. Although rif did not kill WT or ΔmutT1-3 strains, ΔmutT4 and ΔmutT1-4 mutants showed a threefold loss of viability (Fig. 5C). Ectopic expression of the M. smegmatis mutT4, but not the catalytic mutant, in the ΔmutT4 strain restored survival to near-WT level, whereas partial complementation was obtained with the TB gene (Fig. 5D). In addition, overexpression of mutT4, but not mutT1-3, increased the resistance for rif lethality of the ΔmutT1-4 strain to a higher level than is observed in WT strain (SI Appendix, Fig. S5C). Cip and INH induced 50- and 3-fold mortality in the WT strain, respectively, and we did not detect significant differences for either ΔmutT1-3 or ΔmutT4 strain (Fig. 5 G and J). However, mutT1-4 quadruple deletion increased the bactericidal effect of cip and INH, suggesting redundancy between mutT4 and at least one of other mutTs. Complementation of the ΔmutT1-4 strain with M. smegmatis and TB mutT4 partially restored resistance to cip lethality, whereas the catalytic mutant did not (Fig. 5H). In addition, overexpression of mutT4, but not mutT1, mutT2, and mutT3, complemented the higher mortality to cip found in the ΔmutT1-4 mutant (SI Appendix, Fig. S5D). Whereas disc diffusion assays showed that MutT4 is the only MutT involved in cip resistance, we observed redundancy between the different mutTs using an assay measuring killing. These results suggest that MutT4 has a dominant role for preventing the bacteriostatic effect of ciprofloxacin and a redundant activity with at least another MutT for protecting against quinolone-induced killing.

Taken together, these results show that 1) MutT4 has a dominant role for resisting rif lethality and cip bacteriostasis in M. smegmatis and 2) MutT4, with at least one other MutT, protects mycobacteria against bactericidal effect of cip and INH.

An OG-Dependent Pathway of Antibiotic Lethality in Mycobacteria.

To test whether the involvement of mutTs in antibiotic action is due to a higher frequency of genomic OG, we queried the role of four putative translesion polymerases or the mutY/mutM system in the ΔmutT genetic background. The ΔdnaE2dinB123 and ΔmutYM12 mutants were not affected for growth (Fig. 6A) or H₂O₂ tolerance (Fig. 6B). Surprisingly, the growth defect and H₂O₂ sensitivity of the ΔmutT1-4 mutant were not reduced by the loss of accessory polymerases or the GO system excision pathway. These results strongly suggest that the growth defect and the H₂O₂ sensitivity found in the mutT4 mutant are not due to OG genomic incorporation/excision.

Fig. 6.

A pathway of OG-dependent antibiotic lethality in mycobacteria. (A) Doubling times of the indicated strains of M. smegmatis calculated from growth curves in 7H9. (B–D) Sensitivities of M. smegmatis to H₂O₂ (B), rif (C), and cip (D) determined by disk diffusion. Killing of the indicated strains of M. smegmatis by rif (E), cip (F), and INH (G) determined by CFU counting after 6 h of treatment in liquid culture. Results shown are means (± SEM) of data obtained from independent experiments symbolized by gray dots. Values obtained with the WT and ΔmutT1-4 strains for antibiotics killing assays are the same as in Fig. 5 and are reported as a reference. *P < 0.05; **P < 0.01; ***P < 0.001. ns, not significant.

We then measured the impact of dnaE2dinB123 and mutYM12 deletions on antibiotic sensitivity in the ΔmutT background. By disk diffusion, the antibiotic sensitivity of the ΔmutT1-4 strain was not reduced by the deletion of the accessory polymerases or the GO system excision pathway (Fig. 6 C and D). Similarly, mortality after rif treatment was comparable between ΔmutT1-4, ΔmutT1-4ΔmutYM12, and ΔmutT1-4ΔdnaE2dinB123 strains (Fig. 6E). However, loss of accessory polymerases did restore the survival of ΔmutT1-4 to near-WT level after cip (Fig. 6F) and INH treatments (Fig. 6G), suggesting that, in absence of MutTs to sanitize the dNTP pool, accessory polymerases do incorporate oxidized nucleotides into DNA. Interestingly, the mutY and mutMs deletion reversed the lethality of cip in the absence of MutTs to the same degree as loss of polymerases (Fig. 6F), indicating that OG excision by the MutY/M system is toxic after polymerase incorporation of oxidized nucleotides. This effect was antibiotic specific as we did not observe the same result with INH (Fig. 6G). Taken together, these results indicate that the absence of mutTs exacerbates both bacteriostatic and bactericidal effects of antibiotics and that the bactericidal effect of cip is attributable to incorporation of chromosomal OG.

Discussion

We have taken a comprehensive genetic approach to interrogate the roles of nucleotide pool sanitization by MutTs; the role of accessory polymerases; and the role of genomic OG excision in mycobacterial oxidative stress resistance, mutagenesis, and antibiotic sensitivity. Our findings reveal a complex network of protein and repair activities that form a multilayered defense against oxidation of nucleotides and the downstream mutagenic consequences of OG (Fig. 7). Our findings indicate that the mycobacterial GO system does not rely predominantly on nucleotide pool sanitization to prevent OG-mediated mutagenesis but rather, uses removal of endogenously oxidized chromosomal lesions through the MutY/M system. In cells lacking the MutT system, agent-specific antibiotic-induced killing, but not antibiotic-induced growth arrest, does proceed through DinB/DnaE2 and MutY/M systems, indicating a role for oxidative nucleotide utilization in antibiotic lethality in mycobacteria.

Fig. 7.

Model of GO system roles in mycobacteria. (A) Roles of the mycobacterial GO system in basal conditions. MutM1 and MutY prevent mutations induced by in situ generation of chromosomal OG by excising OG paired with cytosine (MutM1) or adenine mispaired with OG after a round of DNA replication (rep) (MutY). The phosphohydrolase activity of MutT4 is necessary for optimal growth and prevention of DNA damage, but this effect is independent on OG. The substrate hydrolyzed by MutT4 is unknown, and the role of ROS in its creation has not been established. The causality between the increase of DNA damage and growth defect is not proven and indicated by the dashed line. (B) Roles of the GO system during oxidative stress. Cip generates ROS (red circles) leading to OG generation in the dNTP pool. OG incorporation is prevented by the redundant activity of MutT4 and at least one other MutT, which hydrolyze OG in dNTP pool. When the MutT system is either disabled or overwhelmed, accessory polymerases incorporate OG into DNA, which is processed by MutY and MutMs to produce DNA DSB from incomplete excision, ultimately leading to lethality. INH treatment generates an unidentified nucleotide that causes cell death when incorporated into DNA by accessory polymerases. INH lethality is prevented by the redundant activity of MutT4 and at least one other MutT. Redundant activity of MutT4 and at least one other MutT also prevents A to G and T to C mutations induced by H₂O₂, independently of OG. Finally, MutT4 prevents killing and/or bacteriostasis induced by H₂O₂, rif, or cip independently on OG.

In Situ Generation of Chromosomal OG from Endogenous Metabolism Is the Dominant Oxidative DNA Lesion.

Because OG can pair with adenine and cytosine, three theoretical pathways are predicted to induce mutations (SI Appendix, Fig. S1). Genomic guanines may be oxidized and mispaired with adenine after replication, leading to G > T or C > A mutations (pathway 1). Guanines can also be oxidized in the dNTP pool and incorporated into the genome by DNA polymerases leading to T > G or A > C mutations if OG is paired with template adenine (pathway 2), whereas G > T or C > A arises after replication if OG is paired with cytosine (pathway 3).

We observed a strong enhancement of G > T or C > A mutations in the ΔmutYΔmutM1 mutant, indicating that pathways 1 and/or 3 are dominant in mycobacteria. However, neither loss of mutTs nor accessory polymerases impact the OG signature mutations in the ΔmutYM12 strain, indicating that pathway 1 is the dominant pathway for OG mutagenesis in mycobacteria, rather than the incorporation of OG oxidized from dNTP pool (pathway 3). This result also indicates that oxidative stress from endogenous metabolism produces a substantial cellular load of chromosomal OG, whereas H₂O₂ treatment does not, suggesting that mycobacteria have evolved distinct systems to defend against endogenous and exogenous oxidative stress. In M. tuberculosis, the major antioxidant enzymes are located in the periplasm (SodC) or secreted (SodA, KatG) (40, 41), whereas the central regulator of bacterial oxidative stress response OxyR or SoxR is absent (23, 42).

In E. coli, mutT deletion strongly increases the frequency of spontaneous T > G or A > C mutations (16), indicating that pathway 2 is operative. However, despite the presence of four mutTs in the M. smegmatis and M. tuberculosis genomes, we did not detect any increase of the mutation frequency in the quadruple mutant of M. smegmatis. An interesting hypothesis would be that unexpected mismatch or base excision repair can deal with OG incorporated in front of adenine and prevent any mutagenic effect of loss of the MutT system. M. tuberculosis and M. smegmatis encode several DNA glycosylases (43), including two endonuclease VIII (Nei1 and Nei2) and one endonuclease III (Nth), which can excise OG from DNA in E. coli (44, 45). Mycobacteria do not encode the MutS/MutL mismatch repair system, but a recent study revealed the existence of a noncanonical mismatch repair (MMR) mediated by the archaeal endonuclease NucS homolog (46). However, the capacity of the mycobacterial MMR to deal with OG is unknown. Examination of the genetic interactions between mutTs, neis, nth, or nucS could reveal a role of the MutT system for mutation avoidance in mycobacteria. Surprisingly, Dos Vultos et al. (25) reports an increase of T > G or A > C in the mutT1 mutant in both M. smegmatis and M. tuberculosis. The methods of that manuscript suggest that saturated cultures were plated to measure mutation frequencies, which could indicate the existence of a growth phase-dependent mutagenesis caused by OG in the mutT1 mutant, as we did not observe this effect in our experiments.

An OG-Dependent Antibiotic Killing Pathway in Mycobacteria.

Foti et al. (20) proposed that incomplete genomic excision of closely spaced OG contributes to the bactericidal effect of some antibiotics in E. coli. To test this hypothesis in mycobacteria, we measured the impact of GO system and accessory polymerases deletion on the killing effect of different antibiotic classes: rif, which inhibits RNA polymerase (rifamycins); cip, targeting DNA gyrase (fluoroquinolone); and INH, which blocks cell wall formation (mycolic acid synthesis).

We found that deletion of accessory polymerases or the GO system excision pathway did not impair killing by rif, cip, or INH in M. smegmatis. In TB, susceptibility for rif, INH, and sm cannot be chemically rescued by the addition of an antioxidant, showing that oxidative stress is not the primary mediator of mortality during antibiotic treatment (47). However, an elevated concentration of ROS has been detected in M. smegmatis after treatment with bactericidal antibiotics (21). The same study showed that deletion of double strand break (DSB) repair pathways (RecA and LigD) or an oxidized dCTP sanitizer (MazG) increases cell mortality after antibiotic treatment. Consistent with these results, we found that loss of mutT1-4 increased the bactericidal effect of the three tested antibiotics. The enhanced bactericidal effect of cip was reversed by loss of dnaE2dinB123 or mutYmutM12, clearly indicating that OG genomic incorporation and excision participate in antibiotic lethality in mycobacteria. Taken together, these data strongly suggest that treatment with bactericidal antibiotics induces oxidative stress in mycobacterial cells, but in wild-type cells, a multilayered set of DNA repair and dNTP pool sanitizer systems efficiently deals with genomic consequences of nucleotide oxidation.

In contrast to our findings with cip, the bactericidal effect of INH in the ΔmutT1-4 strain was reverted by the deletion of accessory polymerases but not of the GO system excision pathway. We propose that, after INH treatment, accessory polymerases incorporate “toxic dNTPs,” which cannot be handled by the MutY/MutM system. Furthermore, loss of neither accessory polymerases nor the mutY/M system reduced rifampin-induced mortality in the ΔmutT1-4 strain. This result indicates that genomic incorporation of toxic dNTPs is not the cause of the higher ΔmutT1-4 sensitivity for rif. In addition to its capacity to clean dNTP pool, MutT can hydrolyze 8-oxo-rGTPs to prevent its incorporation by RNA polymerase (48), possibly suggesting that accumulation of mistranslated proteins could lead to cell death in the ΔmutT1-4 mutant after rif treatment.

What Are the Biological Functions of MutT4 and MutM2 in Mycobacteria?

As reported by Hassim et al. (32), we found that a defect of the GO system excision pathway (MutY and MutMs) strongly enhances the spontaneous frequency of G > T or C > A mutations in M. smegmatis. Surprisingly, we observed a redundancy between MutY and MutM1 but not with MutM2, suggesting that MutM2 assumes a different physiologic function than MutM1.

In vitro, mycobacterial and E. coli MutYs share the same capacity to excise adenine opposite to OG (29, 30). In addition, mycobacterial MutM1 can excise OG paired with G, C, or T, but not A, for which the excision by its E. coli homolog is specific (29, 31). However, the redundancy that we found between MutY and MutM1 to prevent G > T or C > A mutations suggests that, in vivo, MutM1 has the capacity to excise OG paired with adenine, as it is the case in E. coli.

In M. tuberculosis, MutM2 has lost its essential catalytic residues, and the protein does not show DNA binding or DNA glycosylase/lyase activities in vitro (31). M. smegmatis MutM2 activity has never been tested in vitro; however, the protein has conserved its essential residues, and the gene is well expressed according to RNA sequencing data (49). Interestingly, we found an increase of the frequency of A > G or T > C mutations in the mutM2 mutant. In addition to OG, E. coli MutM efficiently excised 4,6-diamino-5-formamidopyrimidine (FapyAde) and 8-oxo-A (7, 50), which are both described to lead to A > G or T > C transition mutations (51–53). Subsequent biochemical studies investigating MutM2 capacity to excise FapyAde and 8-oxo-A might help reveal the physiologic function of this enzyme in nonpathogenic mycobacteria.

The role of MutT4 in preventing the bactericidal effect of cip implicates the enzyme in OG hydrolysis. However, the importance of MutT4 for growth and its ability to protect the bacterium against the bacteriostatic effect of rif and cip are not reverted by the deletion of accessory polymerases or of the GO system excision pathway, excluding OG involvement. Interestingly, MutT4 shows a higher efficiency for adenine hydrolysis in vitro, and we detected an increase of the proportion of A > G or T > C mutations in the mutT4 mutant during oxidative stress, a finding previously reported by Dos Vultos et al. (25). This observation suggests that MutT4, in conjunction with MutM2, acts in the avoidance of FapyAde and 8-oxo-A–dependent mutations. Future experiments will test a potential synergetic effect of ΔmutT4ΔmutM2 double deletion on the induction of A > G or T > C mutations in vivo and the ability of MutT4 for FapyAde and 8-oxo-A hydrolysis in vitro.