Deficiency of Double-Strand DNA Break Repair Does Not Impair Mycobacterium tuberculosis Virulence in Multiple Animal Models of Infection

Abstract

Mycobacterium tuberculosis persistence within its human host requires mechanisms to resist the effector molecules of host immunity, which exert their bactericidal effects through damaging pathogen proteins, membranes, and DNA. Substantial evidence indicates that bacterial pathogens, including M. tuberculosis, require DNA repair systems to repair the DNA damage inflicted by the host during infection, but the role of double-strand DNA break (DSB) repair systems is unclear. Double-strand DNA breaks are the most cytotoxic form of DNA damage and must be repaired for chromosome replication to proceed. M. tuberculosis elaborates three genetically distinct DSB repair systems: homologous recombination (HR), nonhomologous end joining (NHEJ), and single-strand annealing (SSA). NHEJ, which repairs DSBs in quiescent cells, may be particularly relevant to M. tuberculosis latency. However, very little information is available about the phenotype of DSB repair-deficient M. tuberculosis in animal models of infection. Here we tested M. tuberculosis strains lacking NHEJ (a Δku ΔligD strain), HR (a ΔrecA strain), or both (a ΔrecA Δku strain) in C57BL/6J mice, C3HeB/FeJ mice, guinea pigs, and a mouse hollow-fiber model of infection. We found no difference in bacterial load, histopathology, or host mortality between wild-type and DSB repair mutant strains in any model of infection. These results suggest that the animal models tested do not inflict DSBs on the mycobacterial chromosome, that other repair pathways can compensate for the loss of NHEJ and HR, or that DSB repair is not required for M. tuberculosis pathogenesis.

INTRODUCTION

Mycobacterium tuberculosis is one of the world's most successful pathogens, which reflects the ability of this pathogen to adapt to the human host. M. tuberculosis is capable of persisting within the human host for decades in a clinically latent state, during which M. tuberculosis is presumed to be quiescent and nonreplicating. This capacity for latency indicates that M. tuberculosis is able to neutralize, tolerate, or repair damage caused by the products of the mammalian immune system. These immune effectors include microbicidal products of macrophages, such as activated nitrogen intermediates and reactive oxygen species (ROS). These agents are capable of damaging a variety of bacterial cellular constituents, including bacterial DNA.

There is substantial evidence that a variety of bacterial pathogens, including M. tuberculosis, use DNA repair pathways to repair DNA damage inflicted by the host (1). M. tuberculosis strains lacking nucleotide excision repair (NER) through deletion of the uvrB gene are attenuated in the mouse model of infection, and this attenuation is reversed in mice deficient for ROS and reactive nitrogen intermediate generation (2). M. tuberculosis strains deficient for DnaE2 are also attenuated in the mouse model and fail to evolve rifampin resistance in vivo (3). Finally, M. tuberculosis strains lacking uvrD1 and uvrA are severely attenuated in the mouse model (4). Importantly, as the deletion of uvrA alone should cripple the NER pathway, the lack of epistasis between UvrD1 and UvrA for pathogenesis indicates that the attenuation is due to non-NER functions of UvrD1 (4), possibly indicating that the loss of multiple DNA repair pathways is required to fully sensitize M. tuberculosis to host clastogenic pressure. A single study examined a Mycobacterium bovis BCG ΔrecA strain in mouse lungs and spleens and found no attenuation phenotype (5).

It has become clear that mycobacteria, including M. tuberculosis and Mycobacterium smegmatis, differ substantially from Escherichia coli in their complement of double-strand DNA break (DSB) repair pathways (6). DSB repair in bacteria, based predominantly on studies in E. coli, was long believed to occur solely through RecA-dependent homologous recombination (HR). However, additional pathways of DSB repair, including nonhomologous end joining (NHEJ) and single-strand annealing (SSA), are known to operate in Saccharomyces cerevisiae and mammalian cells but were thought to be absent from bacteria. The first clue that some bacteria, including mycobacteria, may express an NHEJ pathway of DSB repair came from the bioinformatic observation that some bacteria encode orthologs of Ku, the end binding protein that is a core NHEJ factor in eukaryotes (7). Subsequent biochemical and genetic studies demonstrated the existence of a bona fide NHEJ pathway in bacteria (8–10). The core components of mycobacterial NHEJ are the Ku protein and the LigD DNA ligase, encoded by Rv0937c and Rv0938, respectively, in M. tuberculosis. Deletion of ku or ligD cripples the NHEJ pathway, as measured by either recircularization of linear plasmids (9–13), I-SceI-induced chromosomal breaks (14, 15), desiccation (16), or ionizing radiation-induced DNA damage (14, 15).

The NHEJ pathway can repair DSBs without a homologous chromosomal template, which is required to direct homologous recombination (17). As such, NHEJ is important for repair of DSBs that arise when DNA replication is absent. This property is evident in mycobacteria, in that Ku- or LigD-deficient mycobacteria are indistinguishable from wild-type cells in clastogen sensitivity during logarithmic growth but become sensitized in late stationary phase (14, 16). Similarly, ku-deficient Bacillus subtilis spores are sensitized to dry heat, a stress that is known to induce DNA breaks (18, 19). A major hypothesis that has been advanced for the role of mycobacterial NHEJ in pathogenesis is that this repair pathway may repair DSBs that arise in the M. tuberculosis chromosome during latent tuberculosis (TB) infection; however, this hypothesis has remained untested due to the lack of an animal model that faithfully recapitulates human latent TB infection. However, animal models do exist in which bacterial replication is slowed or halted, as measured by the response to antimicrobials that target replicating bacilli (20–24), and in which the animals develop the hypoxic, necrotic granulomas seen in human latent TB infection (25, 26).

To expand our understanding of the role of DSB repair in M. tuberculosis pathogenesis, we tested the role of NHEJ and HR, alone and in combination, during M. tuberculosis infection in C57BL/6J mice, C3HeB/FeJ mice (25), and guinea pigs. Importantly, unlike standard mice, C3HeB/FeJ mice and guinea pigs develop necrotic granulomas, which are the pathological hallmark of human latent TB infection. In addition, these DSB repair pathways were tested in a murine hollow fiber model of TB infection (27) in which the mice contain granulomas characterized by tissue hypoxia (28), which is believed to be an important microenvironmental condition encountered by M. tuberculosis during latent TB infection (29).

MATERIALS AND METHODS

Strain construction.

All strains used in this study and their relevant features and genotypes are listed in Table 1. The M. tuberculosis ΔligD strain has been previously described (9). All gene deletions were constructed using a temperature-sensitive mycobacteriophage. To delete M. tuberculosis ku (Rv0937c) from the M. tuberculosis Erdman chromosome, we amplified the 5′ and 3′ flanking regions of the gene by PCR using the following primer sets: for the 5′ flanking region, ku-5-1 (ATCTAGAAGCTTGCCCGAAGGCGATCGAACC) and ku-5-2L (AAGGCCTACTAGTGCTGCTCACCGGCTCATCC), and for the 3′ flanking region, ku-3-2L (ATCTAGAAGCTTCGCGCTCGAAGGCCAACTC) and ku-3-1L (AAGGCCTACTAGTCGTCAGCCGCGCTGTACC) (restriction sites are in bold). PCR amplification products were sequenced to exclude mutations arising during amplification and inserted on either side of a hygromycin resistance cassette. The ku deletion produced by this strategy deletes all but the first and last 12 amino acids of the encoded protein.

TABLE 1.

Strains used in this study

Strain	Genotype	Source
WT	M. tuberculosis Erdman	Lab stock
MGM131	M. tuberculosis Erdman ΔligD::hyg	8
MGM152	MGM131 attb::pMSG319 (LigD)	This work
MGM167	M. tuberculosis Erdman Δku::hyg	This work
MGM173	MGM167/pMSG362 (Δku complemented strain)	This work
MGM1979	M. tuberculosis Erdman ΔrecA::zeo	This work
MGM1989	MGM167 ΔrecA::zeo (recA ku double mutant)	This work
MGM1994	MGM1989/pMSG362 (Ku) (recA ku double mutant complemented with ku)	This work

To delete M. tuberculosis recA (Rv2737c), we amplified the 5′ and 3′ flanking regions of the gene by PCR using the following primer sets: for the 5′ flanking region, oDB91 (ACTAGTATTGTGCGCCGACTTCGA) and oDB92 (AAGCTTTCGCCGAGGCGCATC), and for the 3′ flanking region, oDB94 (GGTACCGACCATTCTCTAGCCATCGAATCC) and oDB95 (TCTAGAGTTCACCTACGAGGGCGAG) (restriction sites are in bold).

The recA deletion allele deletes the entire recA coding sequence except for the first 32 amino acids and the last 59 amino acids of the encoded protein. Note that the RecX start codon overlaps the end of the RecA gene and the RecX start codon is preserved in the recA deletion allele. These PCR products were sequenced to exclude mutations arising during amplification and inserted on either side of a zeocin resistance cassette in plasmid pMSG360Zeo.

To delete recA in the Δku background, we transduced MGM167 Δku::hyg with the recA::zeo phage described above.

Confirmation of gene deletions by Southern hybridization.

To confirm the chromosomal deletion of ku (Rv0937c), chromosomal DNA from candidate hygromycin-resistant transductants was fragmented with SmaI. Southern hybridization was executed using a radiolabeled DNA fragment from the 3′ end of the ku gene. Hybridization to wild-type chromosomal DNA is predicted to yield a 1,901-bp hybridization product, whereas hybridization to Δku::hyg is predicted to yield a 3,524-bp product.

To confirm the chromosomal deletion of recA (Rv2737c), chromosomal DNA from candidate zeocin-resistant transductants was fragmented with either NcoI or EcoRI. Southern hybridization was executed using a radiolabeled DNA fragment from the 3′ end of the recA gene. Hybridization to wild-type chromosomal DNA is predicted to yield a 5,712-bp (NcoI) or 13,247-bp hybridization product (EcoRI), whereas hybridization to ΔrecA::zeo is predicted to yield a 3,603-bp (EcoRI) or 3,185-bp (NcoI) product. Figure 1 shows the confirmation of the strain identity by Southern hybridization for the Δku, ΔrecA, and ΔrecA Δku strains.

FIG 1.

Strain construction by deletion of ku, recA, or both recA and ku. (A) Deletion of ku. An autoradiogram of SmaI-digested chromosomal DNA hybridized to a radiolabeled DNA probe derived from the 3′ end of the ku gene is shown. M. tuberculosis Erdman wild type (wt) or eight candidate hygromycin-resistant transductants (lanes 1 to 8) from transduction of wild-type M. tuberculosis Erdman with phMSG345 is shown. The predicted sizes of the hybridization products are as follows: wild type, 1,901 bp; Δku strain, 3,524 bp. The strain in lane 8 was designated MGM167 and used in subsequent experiments. (B) Deletion of recA. An autoradiogram of NcoI- or EcoRI-digested chromosomal DNA hybridized to a radiolabeled DNA probe corresponding to the 3′ end of the recA gene is shown. Results are for the M. tuberculosis Erdman wild type or two candidate zeocin-resistant transductants (lanes 1 and 2) obtained by transduction of wild-type Erdman with phDB11. The predicted sizes of the hybridization products obtained with NcoI are 5,712 bp (wild type) and 3,185 bp (ΔrecA mutant), and the predicted sizes of the hybridization products obtained with EcoRI are 13,247 bp (wild type) and 3,603 bp (ΔrecA mutant). (C) Deletion of recA in the Δku strain. The results for EcoRI-digested chromosomal DNA from five candidate zeocin-resistant transductants (lanes 1 to 5) with the same hybridization strategy described for panel B are shown. (D) DNA damage sensitivity of M. tuberculosis strains. The wild-type, Δku, ΔrecA, and ΔrecA Δku strains were exposed or not exposed to 0.2% MMS, and regrowth was monitored by determination of the OD₆₀₀.

DNA damage sensitivity.

M. tuberculosis at an optical density at 600 nm (OD₆₀₀) of ∼0.5 was treated with either 0 or 0.2% methyl methanesulfonate (MMS) for 1 h. Cells were collected by centrifugation, washed three times, and resuspended in the original volume of medium. Washed cells were diluted to an OD₆₀₀ of ∼0.03 in Middlebrook 7H9 medium plus oleic acid-albumin-dextrose-catalase. The growth of the bacteria in liquid cultures was monitored by measuring the OD₆₀₀ every 24 h.

Aerosol infections of mice.

All animal experiments were carried out in strict accordance with the recommendations in the Public Health Service Policy on Humane Care and Use of Laboratory Animals (30). The protocol was approved by the Institutional Animal Care and Use Committee (IACUC) of the Memorial Sloan Kettering Cancer Center (for all mouse experiments) or the Institutional Animal Care and Use Committee at the Johns Hopkins University School of Medicine (for all experiments with guinea pigs and mouse hollow fiber infections). Six- to 8-week-old female C57BL/6J and C3HeB/FeJ mice were purchased from The Jackson Laboratory. M. tuberculosis was grown to log phase (OD₆₀₀ = 0.5), washed twice with phosphate-buffered saline containing 0.05% Tween 80, and briefly sonicated to disperse clumps. C57BL/6J mice were exposed to 8 × 10⁷ CFU of the appropriate strain in a Middlebrook aerosol exposure system (Glas-Col), which delivers approximately 100 CFU per animal. Bacterial burdens were determined by plating serial dilutions of lung and spleen homogenates on 7H10 agar medium. The value for each time point represents the average bacterial count from 3 to 5 mice. Plates were incubated at 37°C in 5% CO₂ for 3 to 5 weeks before colonies were enumerated.

Mouse hollow fiber infection.

Six- to 8-week-old hairless, immunocompetent female SKH1 mice (Charles River Laboratories) were used for hollow fiber infections, as previously described (27). An equal volume of each of the axenically grown log-phase cultures (OD₆₀₀ = 0.5) of the wild type, the four strains listed above, and three additional control strains was mixed, and this pool of eight equally represented strains was diluted in order to deliver ∼10³ total bacilli per fiber. At days 1, 21, and 42 after fiber implantation, the fibers were retrieved and their contents were cultured on Middlebrook 7H10 plates (Fisher) and incubated at 37°C in 5% CO₂ for 4 weeks. The genomic DNA was extracted as described previously (28), after scraping all colonies from plates inoculated with undiluted samples. The relative abundance of each strain in the pool was quantitated over time by quantitative PCR (qPCR; iCycler) using primers specific for each strain in the pool. The primers used to amplify each strains were as follows: 5′-CGTGTTCCTGCTGCTATTGA-3′ and 5′-GACGCCCCAGCTTATAACTTC-3′ for the Δku (Rv0937c) strain, 5′-TTCGAGGTGTTCGAGGAGAC-3′ and 5′-CACCACCTCACTGGGTTTCT-3′ for the ΔligD (Rv0938) strain, 5′-CTGCGTTATCCCCTGATTCT-3′ and 5′-TCAATAGCAGCAGGAACACG-3′ for the ku-complemented strain, and 5′-AGAAACCCAGTGAGGTGGTG-3′ and 5′-CCCCAAGGGGTTATGCTAGT-3′ for the ligD-complemented strain. The abundance of each strain was normalized to the abundance of the entire pool, as reflected in the amplification product obtained with primers that amplify the sigA gene, as previously described (27).

Guinea pig infection.

Female outbred Hartley guinea pigs (weight, 250 to 300 g) were purchased from Charles River Laboratories. Guinea pig infections were performed using the mixed strain pool described above and an inhalation exposure system (Glas-Col), as previously described (31, 32), calibrated to deliver ∼1,000 total bacilli to the lungs of individual animals. Lung homogenates were cultured on Middlebrook 7H11 plates. Genomic DNA was extracted after scraping all colonies from plates inoculated with undiluted samples, as described previously (32). Relative strain abundance was quantitated using quantitative PCR, as described above for mouse hollow fiber infections.

RESULTS

Infection of C57BL/6J mice with NHEJ-deficient M. tuberculosis.

After confirmation of the genotypes of the Δku, ΔrecA, and Δku ΔrecA M. tuberculosis strains by Southern hybridization of chromosomal DNA (Fig. 1A to C), we sought to confirm the expected DNA damage sensitivity conferred by recA deletion. We exposed the Δku, ΔrecA, and Δku ΔrecA strains to MMS and measured regrowth in liquid medium. As expected, the ΔrecA strain was highly sensitized to MMS (Fig. 1D). Deletion of ku in either the M. tuberculosis wild-type or the M. tuberculosis ΔrecA background did not additionally sensitize the strains to MMS, which is consistent with our prior observations that (i) NHEJ repairs double-strand breaks, which are not the major lesion produced by MMS, and (ii) the NHEJ system does not contribute to clastogen resistance in log-phase cells (14).

We proceeded to test the effect of NHEJ deficiency on M. tuberculosis growth and persistence in animal tissues. We infected C57BL/6J mice by aerosol infection with M. tuberculosis lacking ku or ligD, the two factors essential for the mycobacterial NHEJ pathway (6, 9, 10, 12). Initial titers of wild-type M. tuberculosis, the ΔligD strain, and the ΔligD strain complemented with ligD 24 h after aerosol infection were 168, 191, and 126 CFU/mouse, respectively (Fig. 2A). Consistent with prior reports, the titers of wild-type M. tuberculosis increased by 4.4 log₁₀ units over the first 3 weeks of infection and then persisted at a constant titer for the duration of the 9-month experiment in lungs (Fig. 2A), spleen (Fig. 2B), or liver (Fig. 2C). The titers of both the ΔligD strain and the ΔligD strain complemented with ligD increased by 4.5 log₁₀ units over the first 3 weeks of infection and persisted at a constant titer in all three organs, similar to the findings for the wild type (Fig. 2A to C). We also measured the weights of infected mice to detect any potential difference in disease severity that may occur independently of the bacterial load. We did not detect any difference in weight between mice infected with wild-type M. tuberculosis and mice infected with M. tuberculosis ΔligD (Fig. 2D).

FIG 2.

Phenotype of M. tuberculosis ΔligD in C57BL/6J mice. (A to C) Titers of the wild-type M. tuberculosis strain, the ΔligD strain, or the ΔligD strain complemented (comp) with ligD in the lungs (A), spleens (B), and livers (C) of C57BL/6J mice after aerosol infection. The x axis shows the number of days after infection, and the y axis shows the bacterial loads in the organs plotted on a logarithmic axis. (D) Body weight of mice infected with the wild-type strain, the ΔligD strain, or the ΔligD strain complemented with ligD over the course of the infection shown in panels A to C.

We next tested the phenotype of M. tuberculosis lacking the gene encoding the Ku protein. Our characterization of mycobacterial NHEJ has revealed a backup NHEJ pathway that is Ku dependent but LigD ligase independent (12). Therefore, it is possible that the lack of a phenotype in the ligD null strain could be due to the LigD-independent NHEJ pathway. To test this idea, we infected mice with M. tuberculosis Δku by aerosol infection. The titers of the wild-type M. tuberculosis and Δku strains in the lung (Fig. 3A), spleen (Fig. 3B), or liver (Fig. 3C) were indistinguishable over the course of the 266-day infection. We also measured host morbidity from the infection, as virulence can be dissociated from bacterial load (33). The median survival of wild type-infected mice was 269 days, and that of Δku strain-infected mice was 290 days, which is not a statistically significant difference (Fig. 3D; P = 0.2820 by the log rank test).

FIG 3.

Phenotype of M. tuberculosis Δku in C57BL/6J mice. (A to C) Titers of the wild-type strain, the Δku strain, or the Δku strain complemented (comp) with ku in the lungs (A), spleens (B), and livers (C) of C57BL/6J mice after aerosol infection. (D) Survival curves for mice infected with the wild-type (dashed line) and Δku (solid line) strains.

Role of RecA in M. tuberculosis infection of C57BL/6J mice.

A single prior study has examined the requirement for RecA during mycobacterial infection of mice. This study examined a BCG ΔrecA strain and found no effect on virulence in either BALB/c or nude mice with intravenous infection (5). To test the role of RecA in M. tuberculosis pathogenesis, we deleted recA (Rv2737c) from the M. tuberculosis chromosome and examined the effect of this null mutation on virulence in C57BL/6J mice.

We infected C57BL/6J mice with wild-type M. tuberculosis or the ΔrecA strain by aerosol. The lung titers at 24 h after aerosolization for the wild-type and ΔrecA strains were 143 and 221 CFU/mouse, respectively. The two strains reached similar titers at 3 weeks of infection (Fig. 4A). Over the course of the 203-day infection, the ΔrecA strain did decline in titer such that at the 203-day time point the titers of the ΔrecA strain were 3.73-fold lower than the titers of the wild-type strain, a difference which was statistically significant (P = 0.008). However, this difference was not reproducible (see below).

FIG 4.

Phenotype of ΔrecA M. tuberculosis in C57BL/6J mice. (A) Titers of the wild-type or ΔrecA strain in the lungs of C57BL/6J mice after aerosol infection. (B) Titers of the wild-type strain, the ΔrecA Δku strain, or the ΔrecA Δku strain complemented with ku in the lungs of C57BL/6J mice after aerosol infection. The difference between the two mutant strains and the wild type at the 147-day time point was not statistically significant.

The lack of a substantial phenotype in the Δku, ΔligD, and ΔrecA strains could be because the two pathways of DSB repair are redundant, masking a phenotype for either single mutant. In M. smegmatis, repair of ionizing radiation-induced DSBs in late stationary phase is partially NHEJ dependent, whereas NHEJ does not contribute to ionizing radiation resistance in log-phase cells (14). As such, NHEJ would be predicted to be important for bacterial survival in the host during states of nonreplication. Although the bacterial titer of M. tuberculosis in mouse lung is stable over the chronic phase of infection, recent evidence indicates that M. tuberculosis is still replicating during this phase, although at a reduced rate (34). Thus, both DSB repair pathways might be required in the mouse if there are mixed populations of replicating and nonreplicating cells, with the former executing DSB repair through RecA-dependent HR and the latter executing DSB repair through Ku-dependent NHEJ. To test this idea, we created a Δku ΔrecA M. tuberculosis strain. As a control, we complemented the Δku ΔrecA strain with a plasmid encoding ku. The 24-hour lung burden of the wild-type strain, the Δku ΔrecA strain, and the Δku ΔrecA strain complemented with ku were 69, 221, and 214 CFU/mouse, respectively. The mean lung bacillary burdens of the three strains are presented in Fig. 4B. We observed similar loads at 21 days postinfection, indicating no deficiency in acute bacterial replication. The titers of both the Δku ΔrecA strain and the Δku ΔrecA strain complemented with ku declined slightly at day 147, similar to the results for the ΔrecA strain, although this difference was not statistically significant (Fig. 4A). However, there was no difference in titer between the Δku ΔrecA strains and Δku ΔrecA strain complemented with ku. In addition, at day 203 there was no difference in titer between any of the strains, indicating no deficiency in chronic infection (Fig. 4B).

Phenotype of DSB repair-deficient strains in C3HeB/FeJ mice.

It has long been recognized that mouse strains differ in their susceptibility to mycobacterial infection. The C3HeB/FeJ mouse model differs from both the C57BL/6J and BALB/c mouse models of tuberculosis, in that in the former model mice develop lung lesions with central necrosis marked by tissue hypoxia (25), a key histopathological feature of human tuberculous lesions. We reasoned that this mouse model might exert more vigorous clastogenic pressure on the M. tuberculosis chromosome during infection and thereby might reveal a pathogenesis phenotype for DSB repair-deficient strains that is lacking in the mouse model presented above. To test this idea, we infected C3HeB/FeJ mice by aerosol with wild-type M. tuberculosis Erdman, M. tuberculosis Δku, or M. tuberculosis ΔrecA Δku at 28, 27, or 4 CFU/mouse, respectively. The ku-deficient strain attained bacillary loads similar to those of the wild type in the lungs at weeks 1 and 3 after infection (Fig. 5A) and in the spleen at week 3 (Fig. 5B). At 15 weeks of infection, the mean burdens in both the lung and spleen of the Δku strain were actually higher than those in the lung and spleen of the wild type (Fig. 5), but this difference was not statistically significant due to variability between the numbers of CFU in the lungs of individual animals (Fig. 5B and D). Due to the low seeding of the lungs by the recA ku double mutant, the initial growth of this strain was very low in both the lung and spleen. However, despite reaching a lower lung burden at 3 weeks after infection, the recA and ku double mutant persisted at a constant level in lungs (Fig. 5A and B) and spleen (Fig. 5C and D), indicating that this strain is competent for persistent infection.

FIG 5.

Infection of C3HeB/FeJ mice. (A) Titers of the wild-type, Δku, and ΔrecA Δku strains in the lungs of C3HeB/FeJ mice after aerosol infection. The x axis shows the time after infection (in weeks). (B) Titers of the same strains from the experiment whose results are shown in panel A, but the titers in the individual mice from each time point are plotted. The mean value for each time point is indicated as a horizontal line. (C) Titers of wild-type, Δku, and ΔrecA Δku strains in the spleens of C3HeB/FeJ mice after aerosol infection. The x axis shows the time after infection (in weeks). (D) Titers of the same strains from the experiment whose results are shown in panel C, but the titers in the individual mice from each time point are plotted. The mean value for each time point is indicated as a horizontal line.

The mouse model that we used is remarkable for a granulomatous inflammatory pathology that is more reminiscent of that of human granulomas. To confirm that our infections replicated this previously reported pathology (35), we examined histopathological sections of C3HeB/FeJ mice infected with the wild-type or the Δku strain (Fig. 6A to H). We observed two major lesion types in mice infected with these strains of M. tuberculosis: large encapsulated lesions with centers of necrosis surrounded by foamy macrophages and a fibrous capsule and more widespread unencapsulated lesions. Additionally, we observed cholesterol clefts and foamy macrophages in regions of necrotizing alveolitis. Wild-type- and Δku strain-infected mice showed no obvious differences in lung pathology.

FIG 6.

Histopathology of infected lungs from C3HeB/FeJ mice. Lung sections of C3HeB/FeJ mice infected with either the wild-type strain (A to D) or the Δku strain (E to H) were stained with hematoxylin-eosin and analyzed. (A and E) Large lesions with centers of liquefactive necrosis surrounded by a layer of foamy macrophages and fibrous capsule and also unencapsulated lesions of necrotic alveolitis (white arrow); (B and F) the wall of foamy macrophages; (C and G) magnified images of the boxes in B and F, respectively, showing a center of necrosis surrounded by foamy macrophages, a fibrous capsule, and compressed lung parenchyma; (D and H) foamy macrophages with cholesterol clefts in regions of severe necrotic alveolitis.

Mouse hollow fiber infections.

We next tested DSB repair-deficient strains in the mouse hollow fiber model of M. tuberculosis infection (27). In this model, M. tuberculosis is encapsulated into semidiffusible hollow fibers, which are implanted subcutaneously into mice. Strain abundance over time was monitored by quantitative PCR using primers specific for each strain in the pool. The pool included the Δku strain, the ΔligD strain, their respective complemented strains, and three control strains unrelated to DSB repair. Figure 7A displays the change in cycle threshold (C_T) between days 1 and 21 and between days 21 and 42. In this experimental design, if a mutant is attenuated, its abundance in the pool drops and therefore the cycle threshold associated with detection of that strain rises because more cycles of amplification are required to yield a signal by qPCR. Thus, a positive value in the cycle threshold difference (e.g., between days 42 and 21) indicates attenuation, whereas a negative cycle threshold difference indicates that the amount of the strain measured in the pool is greater than that at the prior time point. At day 21, the relative abundance of neither the ku nor the ligD mutant strain was significantly different from the overall pool at the prior time point, as indicated by C_T differences near 0 (Fig. 7A). Similarly, at day 42, these strains were not attenuated and, in fact, displayed negative C_T differences (Fig. 7A).

FIG 7.

Phenotype of NHEJ-deficient M. tuberculosis in mouse hollow fiber infection and guinea pig. (A) Infection of M. tuberculosis Δku ΔligD and the respective complemented strains in the mouse hollow fiber model. The differences in strain abundance between day 21 (D21) and day 1 (D1) and between day 42 (D42) and day 21 (D21) are shown. A positive normalized cycle threshold difference indicates that the strain is less abundant than the other strains in the pool because a higher C_T value indicates a lower abundance of the target of amplification. (B) Bacterial loads in guinea pig lung or spleen after aerosol infection over the course of the 42-day infection. (C) Quantitation of M. tuberculosis Δku ΔligD and the respective complemented strains in the guinea pig model. Black bars, difference in strain abundance at day 21 compared with that at day 1; gray bars, difference in strain abundance at day 42 compared with that at day 21.

Guinea pig infection.

We next tested the ku- and ligD-deficient strains in guinea pigs by aerosol infection. The experimental design was identical to that in prior studies (31, 32) and similar to that in the hollow fiber experiments. The total bacterial loads of the pooled strains in guinea pig lungs and spleen are shown in Fig. 7B, revealing that the lung bacillary burden increased 10,000-fold by week 3 after infection. The differences in the C_T values for the Δku and ΔligD strains at day 21 or day 42 are shown in Fig. 7C and again indicate no attenuation of either strain at either time point.

DISCUSSION

We have examined the phenotype of M. tuberculosis lacking two double-strand DNA break repair pathways (HR and NHEJ) in multiple animal models of infection. These experiments were motivated by the attenuation phenotypes of other DNA repair-deficient M. tuberculosis strains in mice (3, 4, 36) and in other bacterial pathogens (37–39). In addition, the documented role of the mycobacterial NHEJ system in the repair of DSBs that arise in nonreplicating cells has logical potential relevance to latency, during which M. tuberculosis is thought to be nonreplicating. However, despite our testing of four different pathogenesis models, we were unable to detect a phenotype for the Δku, ΔligD, ΔrecA, or ΔrecA Δku strains in any of these models, as measured by the bacterial burden and, in some cases, host morbidity. While the manuscript was in review, Brzostek et al. reported that an M. tuberculosis H37Rv ΔrecA Δku ΔligD strain is attenuated during infection of THP-1 cells (40). This macrophage phenotype is apparently insufficiently severe to be manifest as a change in bacterial load in the whole animal during infection.

There are several potential explanations for this lack of a phenotype in our experiments. With regard to NHEJ, the first potential explanation for the lack of attenuation is that none of the models includes a state of true latent TB infection. Although the mouse model of M. tuberculosis infection includes a prolonged phase of stable bacterial titers, recent work indicates that there is active bacterial replication during this chronic phase (34). This bacterial replication would provide a chromosomal template to direct homologous recombination and thereby obviate a need for the NHEJ pathway to repair DSBs. However, the lack of attenuation of even the recA ku double mutant suggests that ongoing bacterial replication is not the major explanation. We have recently shown that mycobacteria possess a third DSB repair pathway, in addition to NHEJ and HR, the single-strand-annealing (SSA) pathway (15, 41). The SSA pathway requires the RecBCD helicase-nuclease, as well as the RecO protein, and therefore would be active in the recA and ku double mutant strain. It is therefore possible that DSBs that appear during infection can be repaired via SSA. SSA requires repetitive DNA flanking the DSB, which could be supplied by the repetitive PE-PGRS gene family.

An additional factor that may mask a phenotype of DSB repair-deficient M. tuberculosis is the use of the bacillary load, a relatively gross measure of bacterial fitness (36), as the major study endpoint. In this regard, it is remarkable that the NHEJ pathway is highly mutagenic when it repairs DSBs with both insertions and deletions. The insertions are partially dependent on the LigD polymerase domain, but the agent(s) of deletional NHEJ has yet to be identified (12, 15, 42). The sequence diversity generated by NHEJ-mediated repair may not lead to the loss of bacterial viability and therefore would be detected only by more sensitive assays of genome evolution, such as whole-genome sequencing of an in vivo-passaged strain, as has been performed on wild-type M. tuberculosis in the macaque model (43).

It is also possible that the immune pressure exerted by the animal models examined here is not capable of inflicting DSBs on the bacterial chromosome. None of the models examined included a phase of spontaneous latency, and therefore, the immune pressure that accompanies the establishment of latency in human infection is likely absent from these models. Further examination of these hypotheses would require testing of DSB repair strains in additional models of infection (such as primate models), by whole-genome sequencing, and/or by creation of a completely DSB repair-deficient strain (a Δku ΔrecA ΔrecBCD strain). However, the experiments that we have performed to date do not substantiate a role for DSB repair in mycobacterial pathogenesis.