Compromised base excision repair pathway in Mycobacterium tuberculosis imparts superior adaptability in the host

Tuberculosis caused by Mycobacterium tuberculosis (Mtb) is a significant public health concern, exacerbated by the emergence of drug-resistant TB. To combat the host’s dynamic environment, Mtb encodes multiple DNA repair enzymes that play a critical role in maintaining genomic integrity. Mtb possesses a GC-rich genome, rendering it highly susceptible to cytosine deaminations, resulting in the occurrence of uracils in the DNA. UDGs encoded by ung and udgB initiate the repair; hence we investigated the biological impact of deleting UDGs in the adaptation of pathogen. We generated gene replacement mutants of uracil DNA glycosylases, individually (RvΔung, RvΔudgB) or together (RvΔdKO). The double KO mutant, RvΔdKO exhibited remarkably higher spontaneous mutation rate, in the presence of antibiotics. Interestingly, RvΔdKO showed higher survival rates in guinea pigs and accumulated large number of SNPs as revealed by whole-genome sequence analysis. Competition assays revealed the superior fitness of RvΔdKO over Rv, both in ex vivo and in vivo conditions. We propose that compromised DNA repair results in the accumulation of mutations, and a subset of these drives adaptation in the host. Importantly, this property allowed us to utilize RvΔdKO for the facile identification of drug targets.


Introduction
The bacterium responsible for causing tuberculosis (TB) disease -Mycobacterium tuberculosis -is among the most notorious human pathogens prevalent across the world. Although multiple drugs are available to treat TB, the emergence of multidrug-resistant (MDR) and eXtensively drug-resistant (XDR) tuberculosis is a major cause of concern [1]. Almost half-a-million cases of MDR/Rifampicin resistant-TB were reported in 2019 [2]. While the success rate of treatment for drug-susceptible TB is~85%, and only~57% for MDR-TB cases. These numbers decline further for patients co-infected with HIV [2, 3].
The acquisition of drug resistance in Mtb is not a simple mechanism. It is a conglomeration of genetic events that occur sequentially, described as a probable pre-resistance state that predisposes pathogen to eventual antibiotic resistance [4]. Drug resistance in Mtb can arise due to mutations in the direct targets of antibiotics or the drug-activating/modifying enzymes. Mutations in the regulatory regions of genes that cause overexpression of drug resistance-conferring genes, including but not restricted to efflux pumps, also help in decreasing the intracellular concentration of drugs. However, in 10-40% of clinical cases, drug resistance cannot be explained by mutations in the direct targets, suggesting hitherto unknown mechanisms employed by the bacilli [5]. Beijing family strains have a higher propensity to develop drug resistance [6]. It is speculated that the higher mutation rate of the Beijing strains could be due to polymorphisms in the genes involved in DNA repair, DNA replication, and recombination [7,8]. In fact, mutations in DNA repair/replication genes such as dnaQ, alkA, nth and recF have been identified in drug-resistant strains [9][10][11].
When Mtb enters the host cell, it encounters stress conditions in the form of reactive nitrogen and oxygen species, capable of damaging the nucleotides resulting in the spectrum of mutations in the genome and challenging the genomic integrity in bacteria [12][13][14][15]. Mtb genome is GC rich, which renders this pathogen more susceptible to cytosine deaminations resulting in C to U change. Inability or failure to correct such mutations prior to replication would result in the accumulation of CG to TA mutations in the genome [16]. Uracils may also arise in the genome because of their direct incorporation by DNA polymerases during replication [17]. Uracil DNA glycosylase (UDG), initiates the uracil excision repair pathway by catalyzing uracil removal from DNA. Analysis of the genome sequence of Mtb unveiled genes that encode proteins involved in uracil excision repair, namely Ung and UdgB, are reported to be non-essential for in vitro growth of Mtb [18,19] (https://mycobrowser.epfl.ch/). Ung belongs to a family I of UDGs which act on the single and double-stranded DNA and are highly specific in recognizing and removing uracils whereas UdgB is a family V UDG that acts exclusively on double-stranded DNA. UdgB is a thermotolerant enzyme that excises uracil, hypoxanthine, and ethenocytosine from the double-stranded DNA. UdgB is an Fe-S cluster containing protein whose activity can be regulated by the availability of iron [20]. Previously we have shown that simultaneous deletion of ung and udgB in Mycobacterium smegmatis (Msm), showed a synergistic effect on the accumulation of mutations [21].
We hypothesized that deletion of DNA base excision repair genes may help bacteria in better adaptability under stress conditions. We adopted the route of generating gene replacement mutants for both the uracil DNA glycosylases (ung and udgB) individually (RvΔung and RvΔudgB) or together (RvΔdKO) to fast forward the in vitro and in vivo evolution of the pathogen. Based on the results presented here, we propose that deletion of DNA repair pathway genes translates into the higher accumulation of random mutations in the bacteria under stress. A selected subset of these mutations aids in the adaptation of the organism to varied selection pressure such as the host immune response.

Ethics statement
Animal experiments protocol was approved by the Animal Ethics Committee of the National Institute of Immunology, New Delhi, India. The approval (IAEC#409/16) is as per the guidelines issued by Committee for Control and Supervision of Experiments on Animals (CPCSEA), Government of India.

Generation of gene replacement mutants and growth pattern analysis
Uracil DNA glycosylase gene replacement mutants were generated using a specialized transduction method. Briefly, upstream and downstream regions of ung were amplified using specific primers and the amplicons were cloned individually in pGEM-T vector. Upstream and downstream flanks were digested with appropriate restriction enzymes and subcloned in pYUB854 to generate pYUB854-ung [22]. pYUB854-ung was digested with PacI and ligated into phAE159 at the compatible PacI site to generate recombinant phAE159-ung. pYUB1471 was digested with PflMI to obtain 1.6 kb oriE+ λ cos sites DNA fragment. PCR amplicons were digested with appropriate restriction enzymes and ligated with oriE+λ cos and chloramphenicol resistance cassette to generate AES. AES was linearized and ligated with phAE159 in the compatible PacI site to generate recombinant phAE159-udgB [23]. Transduction was performed as described previously and colonies obtained after transduction were screened for recombination at the native locus by performing multiple PCR reactions. For in vitro growth measurements, Rv, RvΔung, RvΔudgB, and RvΔdKO were grown either in 7H9-ADC or Sauton's minimal medium upto A 600~0 .6 and inoculated into a fresh medium at final of A 600~0 .1 in technical quadruplet. CFUs were enumerated at day 0, 3, 6, and 9 on 7H11-OADC plates. Growth kinetics was performed using two independent biological experiments and each biological experiment was performed in quadruplet. Statistical analysis (one way ANOVA) was performed using n = 4 for each biological experiment. Graphpad Prism software was employed for performing statistical analysis.

Preparation of whole-cell extract and western blotting
Cultures were inoculated in 7H9-ADC medium at A 600~0 .1 and grown till A 600~0 .8. Cells were centrifuged at 4000 rpm for 10 min at room temperature (RT) and resuspended in the lysis buffer containing protease inhibitors. Cells were transferred into bead beating tubes containing zirconium beads and bead-beating was performed until cell lysis. Lysates were centrifuged twice at 13000 rpm for 45 min at 4˚C. The supernatant was transferred into a fresh Eppendorf tube and protein estimation was performed using Bradford assay. Cell lysates (10,50, and 100 μg) of Rv, RvΔung, RvΔudgB, and RvΔdKO were loaded on 16% Tris-Tricine gel containing 6M urea and transferred to nitrocellulose membranes. Membranes were blocked using 5% BSA prepared in 1XPBST 20 for 2 h, and incubated overnight at 4˚C with α-PknB (1:10000), α-Ung (1:2000), and α-UdgB (1:2000) antibodies, respectively. Membranes were washed thrice using 1XPBST 20 and incubated with anti-rabbit secondary antibody, DARPO (1:10000) for 1h at RT, and washed thrice with 1XPBST 20 . Nitrocellulose membranes were incubated with chemiluminescence reagents and exposed to X-ray film and developed.

UDG activity assay
DNA oligomers, SSU9 and GU9 (10 pmols) were 5' end-labeled using 10 μCi of γ-[ 32 P]ATP and T4 polynucleotide kinase. After labeling, oligomers were passed through Sephadex G50 mini-columns to remove free γ-[ 32 P]ATP, and the oligomers were eluted in 25 μl of 1X Ung assay buffer (20 mM Tris-HCl (pH 7.5), 100 mM NaCl, 10% glycerol and 2 mM β-mercaptoethanol). To determine Ung activity, 10 μg of Rv, RvΔung, RvΔudgB, and RvΔdKO cell lysates were incubated with radiolabeled SSU9 and GU9, in the presence of 1X Ung buffer and the reactions were incubated at 37˚C for 1 h. Reactions were stopped by adding 10 μl of 0.1 N NaOH, and the abasic sites were cleaved by heating at 90˚C for 30 min. Samples were vacuum dried and resuspended in 8 μl formamide dye. Samples were boiled for 10 min at 95˚C, centrifuged, and loaded on 15% polyacrylamide-8M urea gels. For assessing the activity of UdgB, 25 ng of Ugi was added to the lysates of Rv, RvΔung, RvΔudgB, and RvΔdKO and incubated for 30 min prior to performing the UDG assays. Gels were exposed overnight on X-ray film and an autoradiogram was developed after 24 h. Similar procedure was employed for performing Ung or UdgB activity assay using Rv, RvΔung, RvΔung::ung, RvΔudgB, RvΔudgB::udgB, RvΔdKO and RvΔdKO::ung-udgB.

Generation of complementation constructs
Ung and udgB independently and together were cloned in pST-Ki vector [24]. Ung and udgB were PCR amplified using specific primers from the Rv genomic DNA. Amplicons were digested with NdeI and HindIII restriction enzymes and cloned into the corresponding sites in pST-Ki to generate pST-Ki-ung and pST-Ki-udgB. To generate a double complementation construct harboring both the ung and udgB, pST-Ki-udgB was digested with ScaI-HpaI, and pST-Ki-ung was digested with SnaBI and treated with Antarctic phosphatase. pSTKi-ung was ligated with udgB fragment to generate pST-Ki-ung-udgB.

Analysis of mutation rates
Antibiotic sensitivity of Rv, RvΔung, RvΔung::ung, RvΔudgB, RvΔudgB::udgB, RvΔdKO, and RvΔdKO::ung-udgB was determined by spotting them on rifampicin, ciprofloxacin, isoniazid, and no antibiotic-containing plates. Colonies that were antibiotic sensitive were selected for spontaneous mutation rate analysis experiments. Three antibiotic-sensitive cultures of Rv, RvΔung, RvΔudgB,RvΔdKO, RvΔung::ung, RvΔudgB::udgB and RvΔdKO::ung-udgB were grown in 7H9-ADC medium to A 600~0 .6 and cells were inoculated in 10 ml fresh medium (at 50,000 cells/ml)containing 15% sterile culture filtrate of H37Rv. Cultures were grown for 15 days by incubation at 37˚C at 200 rpm. On the 15 th day, appropriate dilutions were plated on 7H11-OADC plates in the absence of antibiotic for enumerating the total number of bacteria, while 1 ml was plated on either rifampicin-containing (2 μg/ml) or ciprofloxacin-containing (1.5 μg/ml) or isoniazid (10 μg/ml) 7H11-OADC for enumerating antibiotic-resistant colonies. Spontaneous mutation rates were calculated using David's fluctuation test [25,26]. For determination of mutation frequency under oxidative stress conditions, Rv, RvΔung, RvΔung::ung, RvΔudgB, RvΔudgB::udgB, RvΔdKO, and RvΔdKO::ung-udgB strains were inoculated at A 600 0.1 and grown to A 600~0 .6 and 50 μM CHP added for 24 h. Subsequently, (24 h later) cells were collected by centrifugation and plated on no antibiotic or rifampicin-containing (10 μg/ ml) medium for calculating the mutation frequency. For analysis under nitrosative stress, cells were incubated for 48 h in an acidic medium containing 6 mM sodium nitrite, at A 600~0 .6. Mutation frequency was calculated by plating after 48 h post sodium nitrite addition on noantibiotic or rifampicin-containing medium. Spontaneous mutation rate or mutation frequency was performed using two independent biological experiments and each biological experiment was performed in triplicates. Statistical analysis (one-way ANOVA) was performed using n = 3 or n = 6. Graphpad Prism software was employed for performing statistical analysis. Data represents one of the two biological experiments.

Guinea pig infection
Cultures of Rv, RvΔung, RvΔudgB, RvΔdKO, and RvΔdKO::ung-udgB were grown to A 600~0 .8. Cells were centrifuged at 4000 rpm for 10 min at room temperature, resuspended in saline, and passed through a 26½ gauge needle to ensure single-cell suspension. Cells (1 x 10 8 ) were suspended in 15 ml saline for infection. Female outbred Hartley guinea pigs were challenged through an aerosol route using a Madison chamber calibrated to deliver~100 bacilli/lung. No antibiotic treatment was given to guinea pigs during or after infection. CFUs were enumerated at day 1 post infection (p.i) for assessing the implantation of bacilli and 56-days p.i for assessing the survival in the lungs and spleen. Statistical analysis was performed at day 1 using n = 3 (per strain) and n = 7 (per strain) at 56-days p.i. Statistical analysis (One-way ANOVA) was performed using Graphpad Prism software. Data represents mean and standard error mean.

Genomic DNA extraction for WGS library preparation
Independent colonies of Rv, RvΔung, RvΔudgB, and RvΔdKO were selected randomly for whole genome sequencing, and no bias was given in terms of size or morphology while selecting colonies for sequencing. We picked independent colonies of Rv (n = 3), RvΔung (n = 4), RvΔudgB (n = 4) and RvΔdKO (n = 3) that were grown in vitro. Besides, we performed the whole genome sequencing of the colonies that were isolated from guinea pig lungs 56-days p.i. Rv (n = 11), RvΔung (n = 8), RvΔudgB (n = 8) and RvΔdKO (n = 9). We performed the whole genome sequencing of ciprofloxacin-resistant RvΔdKO (n = 13). Rv, RvΔung, RvΔudgB, and RvΔdKO in vitro cultures, or colonies obtained from guinea pig lung homogenate, or ciprofloxacin-resistant colonies, were grown till A 600~0 .8 in 7H9-ADC medium (30 ml) and genomic DNA extracted as per manufacturer's instructions (Qiagen). The integrity of the isolated genomic DNA was analyzed by agarose gel electrophoresis, and the DNA was quantified using a Qubit fluorometer. Libraries were prepared using QIAseq FX DNA library Kit. Sequencing was performed using the Illumina Hiseq 4000 platform, generating paired-end reads of~101 bp length. A total of 63 samples covering Rv, RvΔung, RvΔudgB, and RvΔdKO strains were sequenced. Fastq paired-end files were aligned on the H37Rv reference genome using Bowtie2 [27]. Generated SAM files were converted into BAM files using SAM Tools [28]. Qualimap was used to evaluate the alignment of the data as well as other quality points [29]. SNPs were extracted from BAM files using VarScan software [30]. Annotation of SNPs was performed using SnpEff toolbox [31]. All 63 VCF files were combined to create a matrix that includes chromosome position, a nucleotide position in the reference genome, the identified SNP in the newly sequenced genomes, SNP biotype, amino acid change, and respective genes/intergenic region (S1, S3, S5, S7 and S9 Tables). For the identification of unique SNPs under different conditions, the laboratory strain of Rv was used as a reference.

Ex vivo and in vivo competition experiments
Peritoneal macrophages were isolated from BALB/c mice 72 h post intraperitoneal thioglycollate injection. 5 x 10 5 cells/well were seeded in the wells of a 24 well plate in RPMI medium. Cells were infected with Rv, RvΔdKO, and RvΔdKO::ung-udgB independently, or together at 1:1 ratio, at an M.O.I of 1:5. CFUs were enumerated 4 h p.i to determine the uptake. Cells were lysed at 36 h p.i in 0.05% SDS. Intracellular bacteria obtained after lysis were washed thrice using 1XPBS to remove SDS, and half the bacteria were used for CFU enumeration on 7H11-OADC plain plates while the remaining half was used for the next round of infection. The same procedure was followed for the third round of infection. RvΔdKO strain is hygromycin-resistant whereas Rv is sensitive to antibiotics. Colonies obtained on 7H11-OADC plates from mixed infection were patched on hygromycin-containing plates to score for RvΔdKO or RvΔdKO::ung-udgB. Ex-vivo infection was performed using two independent biological experiments and each biological experiment was performed in triplicates. Statistical analysis (Unpaired t-test) was performed using n = 3 for each biological experiment. Graphpad Prism software was employed for performing statistical analysis. Data represents one of the two biological experiments. Inbred BALB/c mice were challenged with Rv which has pST-Ki integrative plasmid that provides kanamycin resistance and RvΔdKO independently, or together at a 1:1 ratio. Briefly, 2 x 10 8 cells of Rv or RvΔdKO or a 1:1 mix of both were suspended in 15 ml saline for performing infections through the aerosol route in mice. Similarly, 1 x 10 8 cells of Rv and RvΔdKO mix (1:1) or Rv and RvΔdKO::ung-udgB mix (1:1) were suspended in 15 ml saline for performing infections through the aerosol route in guinea pigs. CFUs were enumerated 1-and 56-days p.i., on 7H11-OADC with or without hygromycin/kanamycin. For in-vivo infection, statistical analysis was performed at day 1 using n = 3 (per strain) mice or guinea pigs and n = 8 mice or n = 7 guinea pigs (per strain) at 56-days p.i. Statistical analysis (Unpaired t-test) was performed using Graphpad Prism software. Data represents mean and standard deviation.

MIC determination
Rv,RvΔung, RvΔudgB, and RvΔdKO strains were grown in 7H9-ADC medium to A 600~0 .6 and diluted in fresh 7H9-ADC medium to obtain A 600~0 .0006. Medium (100 μl of 7H9-ADC) was added to 96-well plate and ciprofloxacin was serially diluted. Cells (100 μl) were added to these wells and the plate was incubated for 5 days at 37˚C, followed by the addition of 20 μl of resazurin (0.2% w/v in AMQ) for 24 h.

List of DNA oligomers
A list of oligonucleotides used in the study is provided in S11 Table. Source data file for all the figures is provided as S12 Table. Results

Generation and characterization of uracil DNA glycosylases mutants in Mtb
To examine the impact of deletion of UDGs in the pathogen's survival and adaptability, we set out to delete ung, udgB, and both together in the drug-sensitive laboratory strain H37Rv (Rv).
The ung and udgB genes at the native loci were replaced with hygromycin resistant or chloramphenicol resistant cassette with the help of a specialized transduction method to generate RvΔung and RvΔudgB, respectively (Fig 1A and 1C). Recombination at the native loci was confirmed using multiple sets of PCR reactions (Fig 1B and 1D). Subsequently, to generate the combinatorial mutant, the native udgB gene in the RvΔung was replaced with chloramphenicol resistant cassette to generate RvΔdKO ( Fig 1E). Western blot analysis using α-Ung and α-UdgB antibodies confirmed the deletion of ung, udgB, and both ung-udgB in the mutants ( Fig 1F).
To ensure that the deletion of ung and udgB resulted in a loss of UDG activity, we performed biochemical assays. While Ung excises uracil residues from both the single-stranded and double-stranded contexts, UdgB removes uracil residues exclusively from the doublestranded substrates. We used both the single-stranded (SSU9) and double-stranded (GU9) DNA oligomers having uracil at the 9 th position as substrates for monitoring the UDG activities ( Fig 1G). Incubation of the 5'-32 P-radiolabeled substrate(s) with the cell-free extract would result in the formation of abasic site, which can be subsequently cleaved to generate 32 P labeled smaller size product that migrates faster (Fig 1G and 1H). As expected, the radiolabeled band corresponding to the product could not be detected in RvΔung and RvΔdKO lysates with either SSU9 or GU9, suggesting the loss of UDG activity. Since the UDG activity of Ung is overwhelming compared with UdgB, to assess the activity of UdgB, it is necessary to inhibit the activity of Ung, which the addition of Ugi can achieve. Ugi is a small protein encoded by the Bacillus subtilis phage that specifically inhibits the activity of Ung at a 1:1 ratio [32]. Preincubation of lysates with Ugi abrogated the activity of Ung, which was apparent from the absence of the UDG activity on SSU9 ( Fig 1H). Incubation of lysates from Rv and RvΔung with GU9 in the presence of Ugi resulted in the formation of the desired product indicating the activity of UdgB. However, no product formation was detected in RvΔudgB and RvΔdKO lysates suggesting successful deletion of udgB in these strains.

Characterization of complementation strains of uracil DNA glycosylases
We next sought to investigate the growth kinetics of RvΔung, RvΔudgB, and RvΔdKO in nutrient-rich or limiting mediums. There were no discernible growth defects either in a rich or limiting medium under in vitro conditions suggesting that ung and udgB independently and together are dispensable for the in vitro growth (Fig 2A and 2B). Similar growth phenotypes were observed when the strains were subjected to various in vitro conditions of oxidative, nitrosative, and hypoxic stresses (S1A-S1D Fig). For the generation of complementation strains, ung and udgB genes were independently cloned into the integrative vector pST-Ki [24,33] under a constitutive promoter (P smyc ) to generate pST-Ki-ung and pST-Ki-udgB constructs, respectively. Next, we created a construct that expresses both ung and udgB by excising udgB and the promoter from pST-Ki-udgB and sub-cloning it into the unique SnaBI site pST-Ki-ung to generate pST-Ki-ung-udgB ( Fig 2C). pST-Ki-ung, pST-Ki-udgB and pST-Kiung-udgB constructs were electroporated into RvΔung, RvΔudgB, and RvΔdKO strains to generate RvΔung::ung, RvΔudgB::udgB, and RvΔdKO::ung-udgB, respectively. UDG activity assays were performed to characterize the complementation strains. Complementation of RvΔung and RvΔdKO with pST-Ki expressing ung resulted in the restoration of UDG activity, which was apparent by the presence of radiolabelled product upon incubation of either SSU9 or GU9 with the lysates (Fig 2D). Similarly, we observed the complementation of UdgB activity in the presence of Ugi when lysates from RvΔudgB::udgB, and RvΔdKO::ung-udgB were used ( Fig  2E). Collectively, these data confirm the generation of the desired gene deletion mutants and the corresponding complementation strains.

Uracil DNA glycosylases mutants exhibit hypermutability
Since ung and udgB encode the key base excision repair enzymes, we sought to determine the accumulation of mutations in the genomes of RvΔung, RvΔudgB, and RvΔdKO under the normal laboratory growth conditions (Fig 3A) by performing whole-genome sequencing (WGS). Analysis of the data revealed that the laboratory Rv strain showed 92 SNPs, which include synonymous, non-synonymous, non-coding, and intergenic, compared with the reference H37Rv genome (S1 and S2 Tables PE and PPE family of proteins in Mtb represent~10% of its coding regions. Due to the presence of repetitive sequences and high GC rich content, these regions cannot be sequenced and aligned with high confidence [34], which may contribute to a higher number of SNPs observed in the laboratory Rv strain compared with the reference genome. RvΔung, RvΔudgB, and RvΔdKO were generated in laboratory Rv background (Rv in vitro); therefore, in subsequent analysis, we have used Rv in vitro as the reference genome. Next, we analyzed the unique SNPs found in RvΔung, RvΔudgB, and RvΔdKO compared with Rv in vitro (Fig 3B and S2C and S3 and S4 Tables). Interestingly, we observed very few SNPs in RvΔung, RvΔudgB, and RvΔdKO strains; most of them were also found in the PPE and PGRS genes. The observed SNPs did not show specific mutation spectrum such C➔T or G➔A mutations, suggesting that the deletion of UDGs did not result in accumulation of significant mutations when cultured in a nonselective complete medium (S2C Fig).
Acquisition of spontaneous mutations under stress conditions can either lead to cell death if the mutations are deleterious or improved survival if the mutations are advantageous [35]. To examine the acquisition of spontaneous mutations in the absence of ung, udgB, or both, we performed mutation rate analysis using a fluctuation test in the presence of rifampicin and isoniazid (Figs 3C and 3D and S2D) [25]. RvΔung, RvΔudgB, and RvΔdKO strains' spontaneous mutation rate was significantly higher than Rv (Fig 3D and 3E). Deletion of ung, udgB, or both resulted in a 3.57, 4.3, and 22.28-fold increase in the spontaneous mutation rate in the presence of rifampicin, respectively. While the phenotype was completely restored in RvΔudgB::ungB and RvΔdKO::ung-udgB; ung complementation resulted only in partial restoration. Similarly, we observed a 5.32, 22.11, and 37.17-fold increase in the spontaneous mutation rate of RvΔung, RvΔudgB, and RvΔdKO strains compared with Rv in the presence of isoniazid (Fig 3D and 3E). The phenotype was restored in the complementation strains. Collectively the data suggest that deletion strains exhibit a higher spontaneous mutation rate (Figs 3D and 3E and S2E). The Rifampicin Resistance Determining Region (RRDR) sequence of the spontaneous rifampicin resistant colonies revealed multiple mutations in the loci (S3B Fig). In Rv, most of the mutations were in RRDR, in RvΔung, RvΔudgB, we also found mutations outside RRDR. Interestingly, in RvΔdKO, we found mutations in Ile-488 outside the RRDR region, a residue suggested to be involved in interactions with rifampicin [36].

PLOS PATHOGENS
(CHP) (source of ROS). In the presence of acidified sodium nitrite RvΔung and RvΔdKO showed higher mutation frequency than Rv and RvΔudgB; most likely due to increased deamination of cytosines, which would accumulate in the absence of ung (Fig 3F). On the other hand, in the presence of CHP, mutation frequency observed with RvΔudgB and RvΔdKO were higher compared with Rv and RvΔung, which is likely because, besides uracils, UdgB is known to excise hypoxanthine and ethenocysteine from DNA (Fig 3F). The complementation strains restored the phenotype in both the stress conditions. These results suggest the differential abilities of Ung and UdgB in repairing DNA damage in the host.

Uracil DNA glycosylase mutants exhibit hypervirulent phenotype
In vitro data suggested that the UDG mutants have a higher spontaneous mutation rate (Fig 3D  and 3E), thus we sought to evaluate the stress induced mutagenesis of all the strains under hostile in vivo conditions. Towards this we investigated the impact of deleting UDGs on the survival of the pathogen in the host using a guinea pig infection model. We first assessed the survival of  Table). Since RvΔdKO exhibit a better survival in the guinea pig lungs and showed a higher accumulation of SNPs, therefore, we sorted the observed SNPs in RvΔdKO based on the nature of mutation ( Fig  4F). The analysis showed that 34% of mutations are synonymous, 58% mutations are non-synonymous, 6.4% are intergenic mutations, and 1.6% have non-sense mutations.
Further analysis showed that genes such as vapC47, gadB, eccD4, PE-PGRS9, and fadD4 possess non-synonymous mutations. Deletion of any of these genes provide a growth advantage to the pathogen [38]. Interestingly, a non-sense mutation in the Rv3437 at Glu-69 position was identified in RvΔdKO. Disruption of this gene also provides a growth advantage to the bacteria (https://mycobrowser.epfl.ch/) (S4D Fig). However, we did not identify mutations in the drug resistance-conferring loci such as inhA, rpoB, or katG. This is most likely because guinea pigs were not subjected to antibiotics treatment before or after the infection. Spectrum analysis of mutations showed a clear predisposition towards G➔A and C➔T mutations, which is most likely due to these strains' inability to repair modified cytosines and a few other bases (Fig 4E  and S8 Table). The mutation spectrum analysis was performed by counting the number of independent mutations per gene; therefore, C➔T mutation is different from G➔A mutation. These results suggest that the UDG mutants show a superior ability to evolve and survive under in vivo stresses enforced by the host.
There exists a possibility that the hypervirulent phenotype observed in RvΔdKO (S4B Fig) could be due to the spontaneous reversion of attenuating mutations. To negate this possibility, we reperformed the guinea pig infection experiment with Rv, RvΔdKO, and the complementation strain RvΔdKO::ung-udgB. Implantation of Rv, RvΔdKO and RvΔdKO::ung-udgB at day 1 p.i in guinea pigs' lungs was comparable (Fig 4G and 4H). While the CFUs obtained 56-days p. i were similar for Rv, and RvΔdKO::ung-udgB, we noticed an~0.5 log fold increase in the CFUs of RvΔdKO in lungs and spleen (Fig 4H). Hematoxylin and eosin staining performed to examine the gross histopathology showed the presence of well-formed granulomas in all three strains (Fig 4I). The results are consistent with the earlier guinea pig infection experiment (S4B Fig). The ability of complementation strain to restore the phenotype suggested that RvΔdKO's hypervirulence is not due to the spontaneous reversion of attenuating mutations.

The uracil DNA glycosylase mutants accelerate the acquisition of antibiotic resistance
The data show that the RvΔdKO exhibits a survival advantage over the wild-type parent under drug selection conditions or the host-imposed stress. Mutations in UDGs could help in the faster evolution of drug resistance. If this were indeed the case, RvΔdKO could also be an excellent resource to ascertain mutations at the genome-scale that could result in resistance against a candidate drug. To check the prediction, as a test, we set out to identify mutations that would make the strain resistant to second-line antibiotic ciprofloxacin. Ciprofloxacin MIC values obtained for Rv, RvΔung, RvΔudgB, and RvΔdKO were comparable (MIC-0.31 μg/ml), suggesting that at the outset both the strains were sensitive to the antibiotic (S4E Fig). The fluctuation test revealed that the mutation rate was 8.31, 8.82 and 96.88-fold higher in RvΔung, RvΔudgB, and RvΔdKO, respectively compared with the Rv strain (Fig 5A and 5B), and the complementation of ung and udgB independently or together rescued the phenotype (Fig 5A  and 5B). WGS of 13 independent ciprofloxacin-resistant colonies showed that RvΔdKO acquired many mutations compared with the antibiotic-sensitive naïve RvΔdKO parent ( Fig  5C and S9 Table). As anticipated, most of the mutations were G➔A or C➔T (Fig 5D and S10  Table). Circos plot constructed using Rv as the reference genome showed synonymous, nonsynonymous, and intergenic mutations (Fig 5E).
Additionally, a mutation was also present in the intergenic region of Rv600-rpoB (Fig 5F). However, the biological impact of the identified mutation is unclear. Importantly, the analysis

PLOS PATHOGENS
showed mutations in gyrA, nanT, and ureC, in multiple ciprofloxacin-resistant strains (�40%). In addition to mutations in gyrA, a direct target of fluoroquinolones, we identified novel mutations; an example is mutation in ureC that encodes for urea amidohydrolase. Also we found mutations in nanT, which encodes an integral membrane protein involved in sialic acid transport. A➔ G or G➔A mutation in gyrA that results in Asp➔Asn or Gly at the 94 th position occur in all fluoroquinolone-resistant isolates [39]. Together, the results suggest that the absence of UDGs results in faster evolution of antibiotic resistance by the acquisition of mutations and therefore can be an excellent tool for identifying targets responsible for drug resistance.

RvΔdKO strain displays superior fitness ex vivo and in vivo
The fitness of a bacterial strain derives from its ability to survive better than the other bacteria in a given environment. Independent survival of both Rv and RvΔdKO was comparable in the peritoneal macrophages (pF), RAW, and THP1 cells (S5A-S5C Fig). Subsequently, to evaluate the strains' relative fitness, we infected pF with Rv, RvΔdKO and, RvΔdKO::ung-udgB independently or in combination wherein we competed Rv with RvΔdKO or Rv with RvΔdKO::ung-udgB. We reasoned that the differences between Rv and RvΔdKO or Rv with RvΔdKO::ung-udgB strains would only be apparent if we continuously compete strains against each other by performing consecutive rounds of infections with the bacterial population recovered from the preceding infections (Fig 6A). Thus, bacilli obtained 36 h after the first round of infection were used to infect fresh pF cells and the whole process was repeated twice over (Fig 6A). When the experiment was performed with Rv, RvΔdKO, and RvΔdKO::ung udgB strains independently, their proficiency in all the sets was comparable (Fig 6B). The total bacillary survival in Rv + RvΔdKO or Rv + RvΔdKO::ung-udgB competition experiment was also comparable with independent infection (Fig 6B). However, when the strains were rivaled against each other, RvΔdKO outcompeted Rv with each subsequent round of infection (Fig 6C). While the survival abilities were comparable after the first round of infection (Fig 6C; 2 nd set), after completing second and third rounds of infection, RvΔdKO showed its clear dominance over Rv (Fig 6C; 3 rd and final set). On the other hand, competition between Rv and RvΔdKO::ung-udgB did not show discernable differences in the survival, suggesting; a. expression of ung and udgB from the L5 att site rescued the phenotype; b. the survival advantage is indeed due to the absence of UDGs.
In vivo, necrosis of infected cells and subsequent spread of infection to the neighboring cells is responsible for the increased bacillary load. If a strain has superior fitness over the other, one would expect that it would show relatively higher CFUs with time. Even though, independently Rv and RvΔdKO strains showed comparable CFUs 56 days p.i., when competed against each other RvΔdKO exhibited a decisive advantage over Rv (Fig 6D). We next performed a competition experiment between Rv and RvΔdKO or Rv and RvΔdKO::ung-udgB using the guinea pigs model of infection. CFU plating at 56 days p.i suggests~0.5 to 1 log fold increase in the survival of RvΔdKO in comparison with Rv in lungs and spleen, respectively (Fig 6E). In case of Rv and RvΔdKO::ung-udgB, no apparent difference was observed (Fig 6E). Percent CFUs analysis showed an equal deposition of strains at day 1 in mice and in guinea pigs lungs. Survival difference between Rv and RvΔdKO became evident at day 56 p.i in the lungs and spleen (Figs 6E and 7A). Together, these results suggest that RvΔdKO displays a decisive edge over the wild-type parent (Fig 7B). We speculate that this is due to the strain's ability to develop mutations that provide a survival advantage. Collectively, deletion of BER genes ung and udgB drives the accumulation of mutations in the genome under stress conditions and assists in enhanced adaptation.

Discussion
Mtb encodes for multiple DNA repair pathways such as base excision repair (BER), nucleotide excision repair (NER), homologous recombination (HR), Non-Homologous End Joining (NHEJ), and Single Strand Annealing (SSA) pathways. These pathways are crucial for maintaining the genomic integrity [40]. Mutations in DNA repair genes and their roles in the

PLOS PATHOGENS
emergence of antibiotic resistance were previously reported in various pathogens such as Pseudomonas aeruginosa, Helicobacter pylori, Neisseria meningitides, and Salmonella typhimurium. Among the P. aeruginosa strains isolated from cystic fibrosis (CF) patients, 36% had deletion or point mutations in DNA repair genes, mutS, mutL, and mutY [41]. Within H. pylori isolates from dyspeptic patients, 25% of strains showed high mutation frequency due to altered DNA repair and/or proofreading [42]. Similarly, 11% of epidemic strains of N. meningitidis possess defects in mismatch repair pathway genes mutL or mutS [43]. Recently, Mfd, a transcriptioncoupled NER factor, was implicated in the evolution of antibiotic resistance in Bacillus subtilis, S. typhimurium, and P. aeruginosa [44]. These findings establish a role for DNA repair genes in imparting drug resistance. Thus, we set out to investigate how the deletion of the DNA repair pathway in Mtb affects the survival in the host. We also sought to determine the contribution of the host environment in the acquisition of mutations.
Uracil residues in DNA arise due to the deamination of cytosine residues, which in the absence of repair results in CG➔TA mutations in the genome. Deletion of ung and udgB in M. smegmatis compromised bacterial survival under in vitro stresses [21]. However, in Mtb, the ung and udgB deletion strains did not show any survival defects in either in vitro or in vivo (S1 and S5 Figs). These results are in line with previously published work where Mtb genes involved in NHEJ, HR, and BER pathways were found to be dispensable for in vitro growth as well as in vivo survival. Deletions of genes of the NHEJ and HR double-stranded DNA repair pathways independently or together did not compromise Mtb survival in different animal infection models [45]. Deletion of AP-endonucleases, endA, and xthA, independently or together had no impact on mycobacterial survival in a guinea pig model of infection [46]. In the current study, while the survival of ung (RvΔung) and udgB (RvΔudgB) mutants individually was comparable to the wild type parent, double mutants of ung and udgB (RvΔdKO) showed improved survival in the guinea pig model of infection (Figs 4H and S4B). WGS analysis of the wild type and mutants grown under laboratory conditions did not show significant SNPs accumulation (Fig 3B). Suggesting that in the absence of external stress, the presence or absence of a particular repair gene may be of little consequence as the alternate DNA repair pathways may substitute for the deficiency. However, strains subjected to nitrosative or oxidative, or antibiotic stress, repair mutants displayed hypermutability (Figs 3D-3F). In line with this, WGS analysis of the wild type and mutant bacteria from the guinea pig's lungs showed a significant accumulation of mutations in the mutants compared with the wild type ( Fig 4D). Non-synonymous or non-sense mutations in the genes that confer growth advantage to the bacteria upon deletion were identified in the RvΔdKO suggesting that the better survival of the is attributed to the accumulation and selection of advantageous mutations (S4D Fig). The difference was apparent when we compared RvΔdKO with Rv. DNA repair genes such as UDGs or other BER pathway genes, upon deletion, exhibit approximately 20% higher mutation rate, which would help in bacillary survival [47,48]. How is such a property advantageous to the pathogen under physiological conditions? Many DNA repair enzymes, including ung and udgB are downregulated during hypoxia [49]. This may explain the broad mutation spectrum observed in the strains isolated from the infected guinea pig lung granulomas, an oxygen depleted environment (Fig 6F). An advantage of such a regulation could be that when Mtb reactivates from dormancy, it carries a repertoire of population that offers better fitness to the bacterium to sustain itself for active growth in the changed physiological environment. The bacterium that reactivate from dormancy may also offer potential advantages in evolving/ acquiring drug resistance. In contrast, deletion of genes such as dnaE1 or endoMS/nucS results in a very high mutation rate that eventually compromises the strain's fitness [50,51].
A classical way to isolate drug-resistant mutants is to select spontaneous resistant mutants in vitro by growing the laboratory strains in drug-containing media. Subsequent WGS analysis of bacteria of the resistant colonies yields information on the loci responsible for resistance. Such methods have been used to identify multiple drugs' mechanism, including bedaquiline and pyrazinamide [34]. However, selecting for spontaneous resistant mutants can be timeconsuming and at times may not be effective. Mutants of DNA repair gene intrinsically show higher mutation rates. Thus, we speculated that the double mutant of DNA repair genes (RvΔdKO) would be an excellent strain to expedite drug resistance selection. In line with this, we observed an~97-fold higher mutation rate when selecting for resistance against the second-line drug ciprofloxacin (Fig 5A-5F). Most importantly, 11 out of the 13 resistant mutants sequenced showed mutations in gyrA, a well-known target of fluoroquinolones. Interestingly, in addition to mutations in the drug's direct target, we have identified mutations in nanT, ureC, rv2414c and intergenic region of rv600-rpoB. Further investigations are necessary to determine the role of these novel mutations in conferring fluoroquinolone resistance. Our results demonstrate that mutants of DNA repair genes can be successfully employed for faster acquisition of drug resistance, which can be employed in studies targeting the identification of mechanisms of drug resistance.
Fitness in evolutionary theory is measured using the competition assays wherein two strains compete for the same niche, which eventually results in the selection of the strain that has higher capabilities for adaptation [52]. Results show that RvΔdKO when competed against Rv, displayed enhanced proficiency in its ability to survive both in vitro and in vivo, suggesting that the mutant strain's fitness is superior to that of the wild-type strain (Fig 6). Our data suggest that bacteria harboring SNPs that are not favorable for survival are eliminated. Only those that acquire SNPs that confer a survival advantage are retained. We conclude that under duress, either due to host-induced stresses or antibiotic treatment, the RvΔdKO mutant can accumulate mutations that would accelerate the process of its natural selection. While DNA repair genes such as UDG are critical for maintaining genome integrity, their compromised function may help in the accumulation of mutations that provide survival advantage under unfavorable conditions. Collectively, data suggests that the deletion of UDGs results in the accumulation of mutations in the genome under various stress conditions that eventually aids in superior adaptability of the pathogen in the host.