DNA repair in Mycobacterium tuberculosis revisited


  • Editor: Josep Casadesus

Correspondence: Brigitte Gicquel, Unité de Génétique mycobactérienne, Institut Pasteur, 25-28 rue du docteur Roux, Paris 75015, France. Tel.: +33 145 688 828; fax: +33 145 688 843; e-mail: bgicquel@pasteur.fr


Our understanding of Mycobacterium tuberculosis DNA repair mechanisms is still poor compared with that of other bacterial organisms. However, the publication of the first complete M. tuberculosis genome sequence 10 years ago boosted the study of DNA repair systems in this organism. A first step in the elucidation of M. tuberculosis DNA repair mechanisms was taken by Mizrahi and Andersen, who identified homologs of genes involved in the reversal or repair of DNA damage in Escherichia coli and related organisms. Genes required for nucleotide excision repair, base excision repair, recombination, and SOS repair and mutagenesis were identified. Notably, no homologs of genes involved in mismatch repair were identified. Novel characteristics of the M. tuberculosis DNA repair machinery have been found over the last decade, such as nonhomologous end joining, the presence of Mpg, ERCC3 and Hlr – proteins previously presumed to be produced exclusively in mammalian cells – and the recently discovered bifunctional dCTP deaminase:dUTPase. The study of these systems is important to develop therapeutic agents that can counteract M. tuberculosis evolutionary changes and to prevent adaptive events resulting in antibiotic resistance. This review summarizes our current understanding of the M. tuberculosis DNA repair system.


Pathogenic bacteria are constantly exposed to a multitude of hostile conditions, with host defense systems and antibiotic treatments continuously changing their environments. Intracellular pathogens, such as Mycobacterium tuberculosis, are confronted by a variety of potentially DNA-damaging assaults in vivo, primarily from host-generated antimicrobial reactive oxygen intermediates and reactive nitrogen intermediates (RNI) (Adams et al., 1997; MacMicking et al., 1997; Rich et al., 1997; Akaki et al., 2000; Nathan & Shiloh, 2000; Warner & Mizrahi, 2006). It is therefore extremely important for bacteria to have DNA damage repair and reversal mechanisms that can efficiently counteract the detrimental effects of these challenges. However, due to the technical difficulties in working with a slow-growing pathogen such as M. tuberculosis, the study of DNA repair systems in this organism has advanced more slowly than for other bacteria. As a result, most assumptions about M. tuberculosis DNA repair are still based on homology rather than on functional studies (Mizrahi & Andersen, 1998). Genes encoding proteins required for nucleotide excision repair (NER), base excision repair (BER), recombination, and SOS repair and mutagenesis have been identified. In particular, a full complement of genes known to be directly involved in the repair of oxidative and alkylation damage are present in M. tuberculosis. In contrast, homologs of genes involved in mismatch repair (MMR) have not been detected (Fig. 1). This potentially has significant implications in genome stability. However, in contrast to Helicobacter pylori, where the absence of a functional MMR system was correlated to a markedly high level of genetic diversity (Kang & Blaser, 2006), M. tuberculosis genomes are very stable. The nucleotide sequences of the M. tuberculosis genes studied generally display 99.9% similarity (Liu et al., 2006; Dos Vultos et al., 2008). Nevertheless, the capacity of M. tuberculosis to adapt to host conditions has been demonstrated by the appearance of strains that are multidrug resistant (defined as resistant to both isoniazid and rifampicin, with or without resistance to any other antituberculosis drugs) and, more recently, extensively drug-resistant (defined as resistant to isoniazid and rifampin as well as at least two of the six primary classes of second-line drugs, one being a fluoroquinolone and the other an injectable drug). Our interest in DNA repair genes arose from the observation that M. tuberculosis W-Beijing strains, a family linked with an increased risk of drug resistance, could be divided into several branches as a function of unique missense alterations accumulated in three putative antimutator genes (encoding proteins that decrease the overall mutation rate), including two mutT-type genes, mutT2 and mutT4, involved in oxidative damage (Rad et al., 2003). This suggested a possible link between the success in the survival of this family and changes in DNA repair systems. Allelic variation in bacteria arises from random mutations, which may or may not be subject to selective pressure, horizontal gene transfer or recombination events. No studies have shown evidence of recent horizontal transfer in M. tuberculosis to date; however, M. tuberculosis seems to have undergone exchange of genetic material in the distant past (Liu et al., 2006; Rosas-Magallanes et al., 2006). Therefore, other mechanisms, such as defects in DNA repair, recombination and replication (3R), seem to have fuelled the more recent evolution of M. tuberculosis. An earlier study by Werngren & Hoffner (2003) investigated whether the W-Beijing family of M. tuberculosis strains developed resistance to rifampicin at a higher rate than other strains, but they did not detect any significant differences. We now know that members of this family show a particularly high level of variation in the nucleotide sequences encoding their DNA repair proteins (Dos Vultos et al., 2008). The limited knowledge of the DNA repair profile of the strains included in the study therefore prevented conclusions from being drawn. As such, the characterization of M. tuberculosis DNA repair systems may allow us not only to address the reason for the clonal nature of M. tuberculosis strains in comparison with other pathogenic bacteria (Sreevatsan et al., 1997) but also to detect potential selective advantage markers for certain families of strains. In this review, we summarize DNA repair systems, homologs of which are found in other bacteria, and discuss findings obtained from studies carried out over the last decade on M. tuberculosis DNA repair components (Table 1).

Figure 1.

 Major DNA repair pathways in bacteria. Bacteria have developed different mechanisms to counteract DNA damage caused by various sources. Such mechanisms include BER, NER, HR repair and NHEJ. MMR homologs were not found in Mycobacterium tuberculosis, and so this pathway is not shown in the figure. Homologs of DNA repair proteins that are found in M. tuberculosis are listed beside each pathway.

Table 1.   Major findings of the last decade and questions raised
 Major findings of the last decadeImportant questions for future study
 Oxidative stressMutants deficient in components of the GO repair pathway show increased mutation frequencies but different patterns of mutation (Dos Vultos et al., 2006; Jain et al., 2007).Is oxidative damage the driving force behind mutations in the M. tuberculosis genome? What is the nature of interactions between GO repair system components and DNA polymerases?
 Alkylation stressThe presence of Mpg, a 3-methylpurine-DNA glycosylase, initially thought to be present only in mammalian cells. This protein shows limited variability between M. tuberculosis strains (Cole et al., 1998; Mizrahi & Andersen, 1998; Dos Vultos et al., 2008).Given the existence of an extensive network of genes/molecules involved in the alkylation damage response, does Mpg have a more important role in dealing with purine deamination effects than with alkylation damage?
 UracilThe characterization of a bifunctional dCTP deaminase:dUTPase, the first to be demonstrated outside the archae kingdom (Helt et al., 2008).Is the bifunctional dCTP deaminase:dUTPase a mechanism to control dUTP concentration in the nucleotide pool, or is it also an important mechanism for dTTP biosynthesis?
Mismatch repairLack of mismatch repair facilitates genome evolution in mycobacteria (Springer et al., 2004).Does MMR deficiency promote genomic diversity, stability or tolerance in mycobacteria, or all of the above? Is the absence of MMR complemented by different levels of processivity of DNA polymerases?
DNA polymerasesDnaE2, and not a member of the Y family of error-prone DNA polymerases, is the primary mediator of survival through inducible mutagenesis (Boshoff et al., 2003).Can DnaE2 serve as a target for therapeutic intervention?
NERUvrB deficiency results in decreased survival in bone marrow-derived macrophages (Darwin & Nathan, 2005).Besides counteracting DNA damage, is NER also involved in compensating mechanisms for MMR deficiency?
Recombinational repairNonhomologous end-joining repair in mycobacteria (Della et al., 2004).Does NHEJ promote mycobacterial survival during latency or reactivation from latency?
RecA and the DNA damage responseThe existence of a recA-independent mechanism of gene regulation, which is important for DNA repair in M. tuberculosis (Rand et al., 2003).What is the advantage of possessing several different mechanisms of DNA damage induction?

BER genes

BER is one of the most important DNA repair systems in nature. In this pathway, the recognition and excision of base damage is initiated by the action of DNA glycosylases. This event is followed by strand cleavage of the sugar-phosphate backbone, either by apurinic/apyrimidinic-lyase (AP-lyase) activity inherent to many DNA glycosylases or by an AP endonuclease. The repair process is completed by the successive actions of phosphonucleotide kinase or a 3′- or 5′-deoxyribosephosphodiesterase; a DNA polymerase then fills in the new base and a DNA ligase seals the gap (Seeberg et al., 1995; Slupphaug et al., 2003).

BER and oxidative stress

The production of reactive oxygen species (ROS), such as hydrogen peroxide (H2O2), superoxide and hydroxyl radicals, from exogenous and endogenous sources can induce many different types of DNA damage, including single- and double-strand breaks, AP sites and base damage (Demple & Harrison, 1994). DNA bases are susceptible to oxidation mediated by ROS. The low redox potential of guanine makes this base particularly vulnerable and leads to the generation of various oxidized guanine products (Neeley & Essigmann, 2006; David et al., 2007). The high GC content of the M. tuberculosis genome may render it more susceptible to guanine-specific stress (O'Sullivan et al., 2005). One of the most frequently occurring stable and abundant oxidized base lesions in DNA is 8-oxo-7,8-dihydroguanine (8-oxoG). This lesion has strong promutagenic properties (Demple & Harrison, 1994) and, during replication, 8-oxoG frequently mispairs with the nucleotide A (Shibutani et al., 1991), leading to G→T/C→A transversions (Cheng et al., 1992). Formamidopyrimidine (faPy) is another frequently oxidized purine lesion, causing mostly cytotoxic effects (Boiteux & Laval, 1983).

In Escherichia coli, the repair of 8-oxoG and faPy lesions is initiated by the activity of a DNA glycosylase of the formamidopyrimidine–DNA glycosylase (Fpg)/endonuclease VIII (Nei) family. Escherichia coli Fpg, also known as MutM, catalyzes the excision of 8-oxoG and other purines with oxidative damage. Biochemical analysis of Mycobacterium smegmatis Fpg showed that its substrate specificities were similar to its counterparts in E. coli (Jain et al., 2007). The principal substrates of E. coli Nei, however, are oxidized pyrimidines. Putative orthologs of Nei have only been identified in M. tuberculosis and E. coli (Melamede et al., 1994; Bhagwat & Gerlt, 1996; Cole et al., 1998; Eisen & Hanawalt, 1999; Boiteux et al., 2002; Zharkov et al., 2002). The endonuclease III Nth, together with Nei, probably forms part of the repair system targeting 8-oxoG. 8-OxoG residues incorporated opposite G are the preferred substrates of the Nth protein, whereas Nei favors the excision of 8-oxoG mispaired with an A or G (Hazra et al., 2000; Matsumoto et al., 2001). Nth has been implicated in the response to the inhibition of translation in M. tuberculosis (Boshoff et al., 2004). In addition to Fpg (MutM)/Nei, the DNA glycosylase MutY and the 8-oxo-dGTPase MutT are involved in handling the potential mutagenic effects of guanine oxidation. MutY removes adenine mispaired with 8-oxoG, and MutT scavenges the nucleotide pool for oxidized dGTPs. This triplet of DNA repair enzymes constitutes the Go system (Michaels & Miller, 1992). Inactivation of each of these genes individually in E. coli confers an increased mutator phenotype, with E. coli mutM described as a weak mutator and mutY and mutT mutants described as moderate and strong mutators, respectively (Michaels et al., 1992; Fowler et al., 2003). The study of null mutants for the putative nudix hydrolases MutT1, MutT2, MutT3 and Rv3908 (MutT4) in M. tuberculosis revealed that only the absence of the mutT1 product leads to an increased mutator phenotype (Dos Vultos et al., 2006). However, the increased mutation rate in ΔmutT1 mutants occurred at a much lower frequency in M. tuberculosis than in other bacteria. This could be due to the presence of other enzymes capable of counteracting this kind of oxidative stress. The same study suggested that MutT2 can act as an 8-oxo-GTPase, which could explain the low mutation frequency observed in MutT2 mutants. Mycobacterium tuberculosis harbors no less than four copies of the mutM/nei genes and four MutT homologs have been annotated in mycobacterial genomes, whereas only a single mutY homolog has been found (Cole et al., 1998; Mizrahi & Andersen, 1998). Thus, mycobacteria appear to be very well equipped to handle potentially high levels of 8-oxoG in the intracellular pool of dNTPs. Consistent with this, MutY-deficient M. tuberculosis strains have only a mildly increased mutation frequency in the presence of low concentrations of H2O2, whereas fpg disruption has no apparent effect on the frequency of spontaneous mutations or on growth rate (personal observations, unpublished data). One of our most striking observations was that neither MutY nor Fpg deficiency increased the frequency of G→T/C→A transversions; rather, these mutants displayed an increased frequency of A→G/T→C transversions in untreated cells. Exposure to H2O2 resulted in a marked increase of C→G/G→C changes associated with a decrease in the frequency of C→T/G→A changes in the mutY mutants. Similarly, the frequency of G→T/C→A transversions is not increased in M. smegmatis (Morlock et al., 2000; Jain et al., 2007). Jain et al. (2007) suggested that, given that Fpg (MutM) plays a major role in the excision of 8-oxoG, deficiency in this protein would allow the potential misincorporation of this altered base to persist. The propensity of mycobacterial DNA polymerases to insert G instead of C could then lead to 8-oxoG:G mispairing, which is not efficiently excised by MutY. The subsequent excision of 8-oxoG from the mispair would result in a consequent fixation of C→G/G→C mutations. An alternative hypothesis would be that C→G/G→C changes arise as a consequence of DNA replication before MutY- or Fpg-mediated repair. The patterns observed for M. tuberculosis MutY-deficient mutants suggest that the first of these potential mechanisms is the most likely to occur.

Although M. tuberculosis BER pathways induced by oxidative stress still need to be fully elucidated, it is already clear that these bacteria are well adapted to environments where oxidative damage is prevalent. This is illustrated by the shear number of putative genes generating products that are potentially involved in repairing such damage and by the restricted diversity of the nucleotide sequences of these genes (Dos Vultos et al., 2008). The analysis of rpoB mutations in both clinical (O'Sullivan et al., 2005) and laboratory (Morlock et al., 2000) strains has led to the conclusion that oxidative damage is not the main driving force underlying variation in the M. tuberculosis genome. In addition to being interesting in itself, this ubiquitous system seems to display characteristic differences in mycobacteria. A major challenge for the future will be to determine the conditions under which deficiencies in BER can lead to an adaptive advantage of M. tuberculosis.

BER and alkylation stress

The production of nitric oxide (NO) and other RNI by macrophages helps to control infection by M. tuberculosis. The reaction of NO· with oxygen produces nitrous anhydride (N2O3), which nitrosates amines and amides to produce compounds that can form potent DNA-alkylating agents (Friedberg et al., 1995; Durbach et al., 2003). In E. coli, alkylated bases are repaired by two DNA glycosylases, products of the tagA and alkA genes. The tagA gene encodes 3-methyladenine DNA glycosylase I, which is constitutively produced and is highly specific (Bjelland et al., 1993). AlkA encodes 3-methyladenine DNA glycosylase II (O'Brien & Ellenberger, 2004), a component of the adaptive response controlled by ada. The ada gene encodes O6-alkylguanine-DNA alkyltransferase I. A second type of O6-alkylguanine-DNA alkyltransferase is encoded by the ogt gene in E. coli (Friedberg et al., 1995).

In M. tuberculosis, ada and alkA are predicted to encode fused proteins, which seems to occur in many Gram-positive bacterial species (Eisen & Hanawalt, 1999), and are part of an operon including ogt (Cole et al., 1998). Mycobacterium tuberculosis mutants lacking this alkylation damage repair and reversal operon are hypersensitive to the genotoxic effects of N-methyl-N′-nitro-N-nitrosoguanidine (MNNG). Exposure to UV radiation, mitomycin C and H2O2 results in the upregulation of this operon (Boshoff et al., 2003). However, this attenuation was not observed in vivo in survival studies in mice, possibly indicating that nitrosative stress does not induce cytotoxic DNA damage in the murine model (Durbach et al., 2003).

Mpg encodes a 3-methylpurine-DNA glycosylase, which was initially thought to exist only in mammalian cells. However, genome sequencing has revealed homologs of genes encoding Mpg in Bacillus subtilis, Borrelia burgdorferi and M. tuberculosis. Mpg initiates DNA BER and has a wide range of substrates, including damaged or altered bases in duplex DNA (Cole et al., 1998; Biswas et al., 2002). In B. subtilis, this enzyme plays a minor role in the repair of DNA alkylation damage and a major role in preventing the mutagenic effect of deaminated purines and cyclic etheno-adducts (Aamodt et al., 2004). The production of NO and other RNIs by macrophages are involved in controlling M. tuberculosis infection. NO·, along with its many essential roles in vivo, can also undergo reactions that may result in cytotoxic or mutagenic events by a number of mechanisms. NO can react with oxygen, for example, to form the nitrosating agent N2O3, which in turn can react with amines, thiols and other available nucleophiles. N2O3, which is also formed from acidic nitrite, can nitrosate primary amine functional groups on DNA bases, leading to direct DNA damage via deamination. Mycobacterium tuberculosis Mpg may therefore also have a more important role in the repair of damage caused by purine deamination than in alkylation damage repair, particularly given the already extensive network of molecules involved in the alkylation damage response.

BER and uracil

Uracil residues can be introduced into genomic DNA by polymerases using dUTP rather than dTTP or by deamination of the existing dCMP residues (Tye et al., 1978; Duncan & Weiss, 1982; Mosbaugh, 1988). The high percentage of GC in M. tuberculosis genome may make it particularly susceptible to guanine oxidation and cytosine deamination. The deamination of cytosine can be enhanced by UV radiation, certain intercalating agents or by the introduction of a mismatched base or alkylated base paired to cytosine. The lack of MMR in M. tuberculosis may render it more sensitive to cytosine deamination, although this has yet to be confirmed. RNIs, which, as mentioned above, help to control M. tuberculosis infection, may also promote deamination. In E. coli, the amount of dUTP in the nucleotide pool is controlled by a dUTP pyrophosphatase encoded by the dut gene (Friedberg et al., 1995). Mutations in the dut gene are either lethal or increase the amount of uracil residues incorporated into genomic DNA (Wang & Weiss, 1992; Guillet et al., 2006). Given the importance of its biological role, inhibiting M. tuberculosis dUTPase might be an effective means to treat tuberculosis in humans. However, comparison of M. tuberculosis and human dUTPase structures reveal few differences in the active site that could be exploited for species specificity (Chan et al., 2004). Additionally, a bifunctional dCTP deaminase : dUTPase has been identified in M. tuberculosis, the first to be detected outside the archae kingdom (Helt et al., 2008). It is possible that this protein resulted from gene duplication events combined with gain or loss of bifunctional activity. For instance, the existence of a dCTP deaminase in a bacterium with such a high GC content may need to be counterbalanced directly with a mechanism controlling the nucleotide pool dUTP concentration. Nevertheless, it demonstrates the importance of both dTTP biosynthesis and the control of dUTP levels in mycobacteria. The uracil–DNA glycosylase Ung (or UDG) removes this base from the DNA (Lindahl, 1974). Ung is important for the prevention of mutations, for increasing resistance to RNIs generated by acidified nitrite, and for the multiplication of GC-rich bacteria within macrophages (in the M. smegmatis and Pseudomonas aeruginosa models) (Venkatesh et al., 2003). Like its E. coli counterpart, mycobacterial Ung efficiently excises uracil (Purnapatre & Varshney, 1998). Bacillus subtilis phage PBS-1/2, which naturally contains uracil in its genome, encodes the early gene product, UDG inhibitor (Ugi). This protein protects the B. subtilis genome from host UDG by forming an highly specific and stable complex. Ugi has been tested as a potential inhibitor to control the growth of mycobacteria. However, unlike E. coli Ung, which forms a physiologically irreversible complex with Ugi, mycobacterial Ung forms a dissociable complex (Purnapatre & Varshney, 1998; Acharya et al., 2003). Ung – a highly conserved protein – was the only UDG characterized in mycobacteria until the recent characterization of a thermo-tolerant UDG (MtuUdgB). In addition to uracil, MtuUdgB excises ethenocytosine and hypoxanthine from double-stranded DNA (dsDNA). Moreover, although this enzyme is described as having a broad substrate specificity, it rescues the C→T mutator phenotype of E. coli ung. In Pyrobaculum aerophilum, UdgB may act as a backup UDG. Indeed, MtuUdgB is inhibited by its products, and thus likely to act preferentially when uracil levels are unbalanced (Sartori et al., 2002; Srinath et al., 2007). Future studies will be needed to determine whether this enzyme plays a more primary role.

Lack of recognized MMR components

The absence of recognized MMR homologs in the M. tuberculosis DNA repair system is particularly striking. It thus seems unlikely that these bacteria undergo this kind of repair mechanism. There is some evidence that MutS1 and MutL were present in the last common ancestor of all species. The absence of these genes from specific bacteria is thus thought to result from gene loss. The observation of several parallel losses of these genes suggests that these genes are particularly unstable, and consequently easily lost, or that there is some selective advantage to the absence of this kind of repair (Eisen & Hanawalt, 1999). This phenomenon, together with the consequent increase in mutation rate, may serve as a starting point for the evolution of antibiotic resistance, fitness for survival and pathogenicity, possibly conferring a selective advantage under certain stressful situations. However, M. tuberculosis consists of strictly clonal populations, in contrast to the high genetic diversity observed for H. pylori strains that are also MMR deficient (Kang & Blaser, 2006). The potential advantage of loss of MMR in M. tuberculosis, given that it does not appear to result in an increased frequency of mutation, is thus unclear. The resulting reduced fidelity in genome maintenance may contribute to the adaptive ability of this bacterium, which otherwise seems to exist in genetic isolation. However, MMR activity could exist without sequence homology to recognized MMR components; indeed, the hunt for MMR activity in M. tuberculosis is ongoing.

The absence of MMR may be an advantage in the presence of alkylating agents, such as MNNG, which generates O6-methyl-guanine (O6-MeG). In the absence of a functional MMR system, O6-MeG is tolerated by the cell and becomes a mutagenic lesion rather than a lethal one (Harfe & Jinks-Robertson, 2000). The absence of MMR might then, for example, allow adaptation of the mycobacteria to the harsh alkylating conditions within macrophages. Consistent with this, MNNG does not exert a cytotoxic effect on Ada : AlkA-deficient M. tuberculosis in vitro (Durbach et al., 2003). The control of the mutagenic effects of MMR deficiency may in turn be limited in M. tuberculosis by the rate of DNA synthesis, which is slower than that observed in E. coli (Hiriyanna & Ramakrishnan, 1986). This slower processivity may promote increased fidelity of the DNA polymerases, therefore reducing genetic diversity in an MMR-deficient background (Radman, 1998).

However, intrachromosomal homologous recombination (HR) in mycobacteria occurs between divergent loci whenever suitable stretches of high sequence identity are available for initiation or termination of this process (Springer et al., 2004). Given that the MMR system is known to limit recombination between homologous sequences (Matic et al., 1996), the absence of its components might result in increased diversity of the several repetitive sequence elements and repeat regions in the mycobacterial genome. Consistently, M. tuberculosis harbors several repetitive units that show a considerable organizational diversity (Groenen et al., 1993; Supply et al., 1997; Frothingham & Meeker-O'Connell, 1998). Thus, MMR deficiency may play a role in the evolution of M. tuberculosis strains.

AP endonucleases

Most of the DNA glycosylases considered above recognize and remove damaged or inappropriate bases from genomic DNA by hydrolysis of the N-glycosidic bond, giving rise to AP sites. These sites can also arise spontaneously in DNA. They are noncoding during semi-conservative DNA synthesis, and strongly inhibit most replicative polymerases in the template DNA (Friedberg et al., 1995; Boiteux & Guillet, 2004). These lesions are repaired by ubiquitous AP endonucleases, including XthA, which is found in M. tuberculosis (Cole et al., 1998). The xthA gene encodes a protein that is involved in the BER pathway in bacteria (White et al., 1976). Escherichia coli mutants defective in this enzyme are hypersensitive to UV radiation and H2O2 (Demple et al., 1983; Sammartano & Tuveson, 1983; Souza et al., 2006). Its biological importance in M. tuberculosis is highlighted by the fact that no variations in xthA have been observed in clinical strains (Dos Vultos et al., 2008). Furthermore, a second gene, encoding the endonuclease IV End (Nfo), is present in M. tuberculosis and may fulfill the same function (Cole et al., 1998). The presence of two enzymes with this function indicates that dealing with abasic sites is a major aspect of DNA repair. This is also reflected in the high number of DNA glycosylases found in M. tuberculosis and the consequent frequent formation of AP.

DNA polymerases

Gaps resulting from AP endonuclease activity are filled principally by polymerase I, PolA (Dianov & Lindahl, 1994). Mycobacterium tuberculosis possesses a DNA-dependent PolA (Cole et al., 1998). In E. coli, the overexpression of the gene encoding DnaE (an α subunit of DNA polymerase III) restores viability in a conditional lethal PolA-deficient strain (Witkin & Roegner-Maniscalco, 1992). Mycobacterium tuberculosis contains two apparently functionally redundant DNA polymerases involved in replication, DnaE1 and DnaE2, but does not possess PolII (Cole et al., 1998). This redundancy could be explained by the main function of either DnaE1 or DnaE2 being error-prone DNA repair. Deletion of dnaE1 is lethal; thus, this gene is presumed to be the replicative polymerase. In M. tuberculosis, the DNA polymerase DnaE2 is upregulated by UV-induced DNA damage. DnaE2 deletion results in hypersensitivity to damage and eliminates damage-induced mutagenesis in vitro. It is associated with late-stage attenuation and a reduced rate of drug resistance mutation in a murine infection model. DnaE2 thus seems to be a major mediator of induced mutagenesis in mice, playing a role in the emergence of drug resistance (Boshoff et al., 2003). These data suggest that DnaE2, and not a member of the Y family of error-prone DNA polymerases (Ohmori et al., 2001), is the primary mediator of survival through inducible mutagenesis. More recent findings of specific mutations in the rpoB gene associated with increased dnaE2 expression in fitness-impaired rifampicin-resistant M. tuberculosis strains suggests a role for DnaE2 in the adaptation of these strains. It was suggested that the probability of adaptation is increased by increasing the expression of this error-prone polymerase (Bergval et al., 2007). Other genes, dnaZX, dnaQ and dnaN, encoding the γτ, ɛ and β subunits of polymerase III, respectively, are also found in the M. tuberculosis genome (Cole et al., 1998). Additionally, M. tuberculosis harbors dinP and dinX genes encoding the predicted DNA polymerases polV and polIV. These products are error-prone DNA polymerases involved in mutagenesis in E. coli (Tang et al., 1999; Wagner et al., 1999; Friedberg et al., 2005), through translesion DNA synthesis (TLS). TLS is an alternative DNA damage tolerance mechanism carried out by specialized polymerases that can replicate past template lesions. Polymerases involved in this process can be mutagenic (error-prone) (Friedberg et al., 2002). Neither dinP nor dinX are induced following DNA damage (Brooks et al., 2001; Rand et al., 2003). Thus, M. tuberculosis may rely on more specific repair mechanisms for survival following DNA damage. The lack of a ‘high-fidelity’ DNA polymerase, such as PolII, is not consistent with this notion. It is therefore likely that M. tuberculosis has additional polymerases capable of performing TLS. The study of DNA polymerase interactions with other DNA repair components is proving particularly relevant to M. tuberculosis research. The possibility that higher fidelity DNA polymerases compensate for the lack of MMR and the existence of potential interactions between polymerases and components of the GO repair system are just two aspects of interest that could be addressed. DNA polymerases isolated from clinical strains are encoded by one of the groups of DNA repair-related genes showing highest diversity in nucleotide sequence. This may reflect selective pressures acting upon these genes (Dos Vultos et al., 2008). These genes are thus evolving, and probably play a role in the adaptation of M. tuberculosis strains to host conditions.

DNA ligases

The nicks resulting from excision repair or from the action of another damaging agent may be sealed by the formation of a phosphodiester bond by DNA ligases. This reaction requires energy, provided by either ATP or NAD+. Most eubacteria, like E. coli, have DNA ligases that use NAD+ as their source of energy. Escherichia coli carries a single DNA ligase, which uses NAD+ as a nucleotide cofactor and is encoded by the ligA gene. Mycobacterium tuberculosis has one NAD+-dependent DNA ligase gene, ligA, and three different ATP-dependent DNA ligase gene homologs, ligB, ligC and Rv0938 (ligD). Mycobacterium tuberculosis LigA may be essential for repair and recombination, and therefore for viability, making it an attractive target for antibiotic therapy (Cole et al., 1998; Wilkinson et al., 2003; Gong et al., 2004; Srivastava et al., 2005b). However, although LigA seems to be essential for mycobacterial viability, only low levels of the protein are required for growth. As such, inhibition of enzymatic activity would have to be very efficient for this ligase to be useful as an antibiotic target (Korycka-Machala et al., 2007). Nonetheless, compounds such as glycosylamines, tested in assays with M. tuberculosis LigA, T4 ligase and human ligase I, have been shown to specifically inhibit M. tuberculosis LigA activity (Srivastava et al., 2005a). Similar findings were obtained for N-substituted tetracyclic indole (Srivastava et al., 2007). LigD and LigC, together with a Ku homolog, appear to have major roles in the nonhomologous end-joining (NHEJ) repair pathway, as discussed below (Della et al., 2004).

Single-stranded DNA (ssDNA)-binding protein (SSB)

The ssb gene encodes a putative ssDNA-binding protein, which protects transiently formed ssDNA from nuclease and chemical attack, and prevents it from forming aberrant secondary structures. SSB plays an essential role in several DNA transactions, including replication, repair and recombination (Trovcevic et al., 1991; Kelman et al., 1998; Sarov-Blat & Livneh, 1998; Yuzhakov et al., 1999; Genschel et al., 2000). Indeed, ssb is part of a set of 262 genes thought to be the minimal requirement for cellular life (Mushegian & Koonin, 1996; Saikrishnan et al., 2003). This gene seems to be important for protection against DNA damage, as demonstrated by its upregulation upon exposure to capreomycin, mitomycin C or H2O2, and for inhibition of translation (Boshoff et al., 2003, 2004; Fu & Shinnick, 2007). SSB proteins from mycobacteria display high affinity for cognate RecA, whereas E. coli SSB do not (Reddy et al., 2001). Furthermore, in M. tuberculosis, SSB participates in the recruitment of Ung to damaged DNA (Purnapatre et al., 1999; Handa et al., 2001; Acharya & Varshney, 2002).

NER genes

NER is an alternative pathway to BER, in which genomic damage is repaired by means of incisions in regions flanking the damaged DNA, leading to the excision of an oligonucleotide rather than a single base. This system can recognize a wider range of damaged bases than the BER system, using endonucleases with lower specificity. This pathway involves the three components of the excinuclease ABC (UvrA, UvrB and UvrC), the DNA helicase II UvrD and the transcription repair-coupling factor (TRCF), encoded by the mfd gene. Gene expression analysis has shown that M. tuberculosis seems to rely on uvrB for protection against NO and UV radiation (Darwin et al., 2003; Darwin & Nathan, 2005). However, not all uvr genes, such as uvrA, C, and D– involved in the repair of UV-induced damage – were identified in the same screen. It remains to be determined whether they are individually dispensable for resistance to these damaging agents and whether other gene products may be able to substitute for their functions. The product of uvrB appears to be essential for M. tuberculosis to cause premature death in immunocompetent mice. Additionally, uvrB deficiency results in a decreased M. tuberculosis survival rate in bone marrow-derived macrophages (Darwin & Nathan, 2005). Uvr gene expression is increased in M. tuberculosis isolated from human macrophages, demonstrating the importance of the uvr system for M. tuberculosis survival upon infection (Graham & Clark-Curtiss, 1999). Additionally, uvr genes are involved in the response to H2O2 damage (Cabusora et al., 2005) and are upregulated upon exposure to UV irradiation (Boshoff et al., 2003), and TRCF is required for NER activity upon transcriptional arrest due to base damage in the template DNA strands (Friedberg et al., 1995). Mycobacteria possess all the genes encoding the proteins involved in these mechanisms. However, although most bacteria carry just one gene encoding DNA helicase II, M. tuberculosis has two putative genes encoding this protein: uvrD1 and uvrD2 (Cole et al., 1998). UvrD1 is a weak DNA-unwinding enzyme on its own. It also acts as a highly active DNA-dependent ATPase and is able to form a stable binary complex with the tailed duplex DNA helicase substrate (Curti et al., 2007; Sinha et al., 2007). Ablation of UvrD1 sensitizes mycobacteria to cell death caused by UV and ionizing radiation and to single chromosomal breaks generated by I-SceI endonuclease (Sinha et al., 2007). UvrD1 needs Ku to activate its latent helicase activity, and seems to play an essential role in the repair of several types of DNA damage, including chromosomal site-specific double-strand breaks (DSBs). Interestingly, uvrD1 seems to be essential for mycobacterial repair of UV damage and is upregulated by H2O2-induced stress (Boshoff et al., 2003), demonstrating its role in NER. This function of UvrD1 is independent of Ku. The role of UvrD1 in mycobacterial DNA repair is of particular interest in light of recent findings, showing that the absence of UvrD1 attenuates M. tuberculosis in a mouse model of infection (Curti et al., 2007). UvrD2 is a Ku-independent helicase (Sinha et al., 2007), which is upregulated upon exposure to UV radiation (Boshoff et al., 2003). A recent genetic study has shown that UvrD2 is essential for M. smegmatis survival (Sinha et al., 2008). The same study suggests that while mycobacterial UvrD1 is genetically reminiscent of E. coli UvrD, which is nonessential, UvrD2 appears to be functionally similar to PcrA which is essential for the viability of B. subtilis and Staphylococcus aureus. The authors suggest that there is a functional overlap between bacterial UvrD and PcrA-like helicases, with PcrA alone able to perform all the essential functions. They speculate that UvrD1 and UvrD2 might have overlapping functions, UvrD2 being pluripotent with respect to essential tasks. However, these two proteins show the same level of conservation in clinical strains. Thus, UvrD1 may be equally as important as UvrD2 (Dos Vultos et al., 2008) in other roles, such as in the repair of different types of DNA damage, including chromosomal site-specific DSBs.

As well as directly counteracting DNA damage caused by host conditions, NER may also be involved in compensating for MMR deficiency (Guthlein et al., 2008). This recently proposed notion offers exciting new perspectives for future studies on NER.

Recombinational repair genes

DNA DSBs can be generated during replication or caused by exposure to DNA-damaging agents. The repair of DSB is essential: even a single DSB is lethal to dividing cells, if it is not repaired in a timely fashion. There are two major DSB repair pathways: HR and NHEJ. In HR, a second intact copy of the broken chromosomal segment serves as a template for DNA synthesis across the break (Shuman & Glickman, 2007). NHEJ does not rely on a homologous DNA template, which means it can operate in situations with only one chromosomal copy available. NHEJ requires the broken ends to be brought into contact. This process is aided by the DNA-end-binding protein Ku, followed by the sealing of at least one of the broken strands by a specialized DNA ligase [LigD in M. tuberculosis (Della et al., 2004)]. LigD is a large, multifunctional enzyme comprising an ATP-dependent ligase (LIG) domain, a polymerase (POL) domain and a phosphoesterase (PE) domain (Della et al., 2004; Gong et al., 2005; Pitcher et al., 2005, 2007b; Akey et al., 2006; Malyarchuk et al., 2007). Whereas HR is generally error free, NHEJ can be either faithful, if the ends are sealed directly, or mutagenic, if the ends are remodeled by nucleases or polymerases before being sealed by DNA ligases (Shuman & Glickman, 2007). The efficiency of plasmid-based NHEJ of blunt and 5′-overhang DSBs in mycobacteria is several hundred times lower after deletion of Ku and c. 100 times lower after the deletion of LigD (Gong et al., 2005; Shuman & Glickman, 2007). A physical interaction between mycobacterial Ku and LigD has been demonstrated using several different methods. In particular, the DNA-Ku-LigD ternary complex has been detected by native gel electrophoresis (Weller et al., 2002; Shuman & Glickman, 2007) and Ku identified in a genome-wide yeast two-hybrid screen for LigD partners (Gong et al., 2005; Shuman & Glickman, 2007). Moreover, Ku stimulates the ability of LigD to seal DSB substrates in vivo (Shuman & Glickman, 2007). The absence of known NHEJ components sensitizes mycobacteria to desiccation and ionizing radiation during the stationary phase (Pitcher et al., 2007c). Mycobacteria possess a backup mechanism for LigD-independent error-prone repair of blunt-end DSBs. Indeed, blunt-end NHEJ efficiency in LigD-deficient cells is reduced 20-fold when the ligC locus is deleted (Gong et al., 2005). As mentioned above, UvrD1 is also Ku-dependent. However, the exact mechanisms underlying the role of in DSB repair remain to be determined.

The NHEJ pathway was thought to be restricted to eukaryotes but was recently identified in prokaryotes. It seems to be broadly relevant to bacterial physiology, as demonstrated in the protection of the B. subtilis spore DNA under stressful conditions (Moeller et al., 2007). However, the NHEJ repair pathway is not conserved in all bacterial species. Several possible hypotheses have thus been proposed to explain its presence in certain species. In mycobacteria, NHEJ may be important during the latent phase of human tuberculosis, assuming that no daughter chromatid is present for HR. NHEJ would then be the only pathway available to repair DSB, promoting mycobacterial survival during latency or reactivation from latency. Even though expression profiles of M. tuberculosis cells suggest that LigD and Ku transcription is not regulated during infection (Rengarajan et al., 2005), NHEJ may nevertheless be important for survival within macrophages, where they are continually exposed to the genotoxic defense mechanisms of the host cell. As mentioned, oxidative damage in particular induces DSB under such conditions. This may explain why M. tuberculosis has retained a functional NHEJ apparatus that has been lost by other species. This mechanism would also promote mutagenesis, serving as a potential selective advantage under certain conditions, for example, for adaptation to genetic stresses induced by the macrophage environment or for the acquisition of antibiotic resistance (Shuman & Glickman, 2007; Pitcher et al., 2007a).

Two major RecA-dependent pathways are involved in initiating HR: the RecBCD and RecFOR pathways. RecBCD is a bipolar helicase that splits the duplex into its component strands and digests them until it reaches a recombinational hotspot (chi site) (Singleton et al., 2004). RecBCD may have been acquired by M. tuberculosis by gene transfer. The high nucleotide diversity of these genes observed in clinical strains may reflect the adaptive evolution of these genes in M. tuberculosis (Dos Vultos et al., 2008). Corynebacterium species belonging to the same order that classifies Mycobacterium species, Actinomycetales, have extremely stable genomes. This may be due to the fact that such species lack recBCD genes (Nakamura et al., 2003). Indeed, differences in genome variability between mycobacterial species might be attributed to the presence or absence of such genes. RecFOR is important in postreplication daughter-strand gap repair (Morimatsu & Kowalczykowski, 2003). Both pathways provide an ssDNA molecule coated with RecA, allowing the formation of a homologous molecule. RecA then promotes strand exchange and the resulting joint molecules can then be processed by the RuvABC complex or the RecG helicase. RuvA and ruvC are upregulated upon UV damage (Boshoff et al., 2003). RuvC encodes a dimeric endonuclease that resolves the Holliday junction to form duplex products by introducing symmetric nicks in two of the four DNA strands (Iwasaki et al., 1991; Sharples et al., 1999). This gene, unlike its E. coli counterpart, is induced by DNA damage in M. tuberculosis. Previous findings have demonstrated its importance for H. pylori survival in macrophages, suggesting that it may also play an important role in other pathogens, including M. tuberculosis (Brooks et al., 2001; Loughlin et al., 2003). RecR interacts with both RecO and RecF, but no three-component complex has been detected (Webb et al., 1997; Morimatsu & Kowalczykowski, 2003). RecR may therefore play a central role in this pathway. This is strengthened by the fact that recR is upregulated in M. tuberculosis upon capreomycin treatment (Fu & Shinnick, 2007). Furthermore, both recR and recF are upregulated upon translational inhibition (Boshoff et al., 2004), suggesting that, despite the absence of recJ, this pathway is active in M. tuberculosis.

RecN also appears to be important for recombinational repair (Meddows et al., 2005) in response to DSB. A homolog of the recN gene was found in M. tuberculosis, but so far, no studies on this protein have been performed in M. tuberculosis. However, this gene does not seem to be induced by DNA damage (Brooks et al., 2001; Rand et al., 2003).

RecA, LexA and the SOS response

RecA is a central component of the bacterial response to DNA damage. It has a direct role in recombinational DNA repair, either by the RecBCD pathway or the RecFOR pathway. In both cases, RecA forms a nucleoprotein filament on regions of ssDNA. This protein searches for homologous sequences through its ability to simultaneously bind dsDNA. Strand exchange then initiates the actual process of recombination (Kowalczykowski et al., 1994).

RecA also has a key regulatory role in response to DNA damage. In conjunction with the repressor protein LexA, RecA controls the expression of other genes that promote survival following DNA damage – the so-called SOS response. The nucleoprotein filament formed by RecA in ssDNA regions upon DNA damage can stimulate autocatalytic cleavage of LexA. Therefore, repression by LexA is alleviated and SOS gene expression is increased (Little & Mount, 1982). Mycobacterium tuberculosis has, with the exception of polB and umuD, all of the genes required for a functional SOS system, such as that found in E. coli (Cole et al., 1998; Mizrahi & Andersen, 1998). These genes include those encoding the RuvABC complex, and also the genes encoding the DNA polymerases DinP and DinX, the uvrA, uvrB and uvrD genes all discussed above. The remaining dinG gene and the key regulators RecA and LexA are discussed below.

RecA is a highly conserved and ubiquitous protein (Karlin et al., 1995). It may thus be expected to perform similar functions in all bacteria. Davis and colleagues first identified a recA-like gene in M. tuberculosis in 1991 through DNA hybridization using recA gene sequence from E. coli. The gene was cloned and expressed, but only showed partial complementation of E. coli recA null mutants (Davis et al., 1991). In vitro studies have shown that the M. tuberculosis RecA protein exhibits several differences from other bacterial RecA homologs. Unlike E. coli RecA (EcRecA), homologous pairing and strand exchange promoted by M. tuberculosis RecA (MtRecA) is greatly dependent on the pH of the medium (Vaze & Muniyappa, 1999). The MtRecA protein also shows a reduced affinity for ATP and reduced efficiency of ATP hydrolysis (Kumar et al., 1996). Indeed, the overall structure of MtRecA is very similar to that of EcRecA – exhibiting high sequence similarity and similar folds and topology – whereas the crystal structures of MtRecA complexed with an ATP analog are consistent with reduced levels of interaction (Datta et al., 2000). The role of the M. tuberculosis RecA protein in DNA repair and integration of exogenous nucleic acids by HR was demonstrated for the first time (Frischkorn et al., 1998) by its ability to fully restore the phenotype of M. smegmatis recA mutants.

The structure of the M. tuberculosis recA gene is unusual in that the ORF is approximately twice the size of its conserved homologs (Davis et al., 1991). Further studies revealed that the genetic locus encoding the functional protein comprises a single ORF, which is interrupted by an intein-encoding sequence (Davis et al., 1992, 1994). The presence of such inteins in many obligate pathogens suggests that the RecA intein may play a role in mycobacterial functions related to pathogenesis or virulence. However, another study of 39 mycobacterial strains belonging to 32 different species demonstrated that the presence of inteins does not correlate with specific species characteristics such as pathogenicity or growth rate (Saves et al., 2000). Alternatively, these inteins may be responsible for the low efficiency of HR in M. tuberculosis and other slow-growing mycobacteria. Thus, this unusual organization of the ORF, from which the active RecA protein is generated, may affect its function in M. tuberculosis. However, several studies demonstrated that the presence of an intein does not seem to interfere with RecA function. Mycobacterium tuberculosis recA (with or without its intein) fully restored both DNA repair and integration of exogenous nucleic acids by HR in M. smegmatis recA null mutants, despite a small amount of precursor protein being detected in addition to spliced RecA (Frischkorn et al., 1998; Papavinasasundaram et al., 1998). Therefore, the relatively low frequency of HR in M. tuberculosis is unlikely to be due to the presence of the intein-coding sequence in its recA gene.

As for most bacterial species, M. tuberculosis RecA (MtbRecA) is induced by DNA-damaging agents. Analysis of the DNA sequences required for MtbRecA expression suggested that two promoters may be present in recA (Movahedzadeh et al., 1997b). Later studies confirmed that recA is expressed from two promoters, both of which are inducible by DNA damage, although by different mechanisms. One of these promoters is regulated in the classical way by the repressor LexA. The other promoter is DNA damage-inducible, independently of both LexA and RecA (Davis et al., 2002b). This dual DNA damage-inducible regulation of recA expression in M. tuberculosis differs from the mechanisms present in E. coli and B. subtilis, where recA is expressed from a single promoter (Weisemann & Weinstock, 1991; Cheo et al., 1993). A novel mechanism of inducing DNA damage repair in M. tuberculosis, independent of LexA and RecA, has also been described.

RecA is cotranscribed with another gene, recX, in many eubacteria including M. tuberculosis, for which a regulatory role related to recA function was proposed (Papavinasasundaram et al., 1997). Further studies in M. tuberculosis indicated that RecX is a negative regulator of recA. RecX interacts directly with RecA, resulting in supression of ATPase and strand exchange, processes that are central to HR (Venkatesh et al., 2002).

LexA is another important regulatory component of the SOS system. In E. coli, LexA is a repressor protein, which in the induced state binds to a specific sequence, termed the SOS box. The SOS box is located upstream of, and thereby restricts the expression of, the genes it regulates. Following DNA damage and LexA autocatalytic cleavage promoted by RecA, LexA no longer binds to the SOS boxes. This increases transcription of the genes of the SOS regulon. The basic principles of this regulatory mechanism are found in many other bacteria, although the DNA sequence of the LexA-binding site varies. A motif similar to the SOS box originally identified in B. subtilis has been found to bind LexA in M. tuberculosis (Movahedzadeh et al., 1997a). More recently, a novel consensus sequence, defining the mycobacterial SOS box, has been described (Davis et al., 2002a). This led to the identification of a number of LexA-regulated genes in M. tuberculosis, only four of which are known to be involved in DNA repair: recA, lexA, ruvC and dnaE2. Although this is quite a small number of genes, previous studies have also shown that certain SOS genes, including recN, dinP and dinG, are not inducible by DNA-damaging agents. Other inducible genes did not bind to LexA (uvrA and ssb). This again supports the idea that an alternative mechanism of DNA damage induction, not controlled by LexA, exists in M. tuberculosis (Brooks et al., 2001).

Sub-MICs of certain antibiotics – in particular, compounds with a primary mode of action being DNA damage – enhance mutation rates in bacteria (Gillespie et al., 2005). This is mostly the result of transcriptional changes in the genes responsible for DNA repair and preservation of the genome integrity, such as the SOS response genes. A recent study investigated induction of the SOS system in M. tuberculosis following treatment with fluoroquinolones (O'Sullivan et al., 2008). Fluoroquinolones represent the only ‘new’ class of active drugs currently available to treat drug-resistant tuberculosis (Migliori et al., 2008). Fluoroquinolones target DNA gyrase, generating lethal dsDNA breaks (Drlica & Zhao, 1997) that induce the SOS response (Phillips et al., 1987). The low-fidelity polymerases induced during this response enhance M. tuberculosis survival through the emergence of quinolone resistance mutations. Therefore, the use of fluoroquinolones could compromise tuberculosis treatment by increasing the selection pressure to resistance. Subinhibitory quinolone treatment of M. tuberculosis induced lexA and recA expression, although the overall gene expression response did not show a consistent pattern of SOS upregulation. This response following treatment with quinolone differs from the response observed following treatment with other DNA-damaging agents and from the response observed in other organisms (Mesak et al., 2008). Further studies are required to further elucidate the role of this important response in the acquisition of quinolone (and other antibiotic) resistance mutations in M. tuberculosis. Until new drugs for fighting tuberculosis become available, the prevention of further development of drug resistance will be crucial.

Another SOS response gene found in M. tuberculosis is dinG. The dinG gene encodes a helicase (Lewis & Mount, 1992; Koonin, 1993). Helicases are motor proteins that move directionally along the nucleic acid phosphodiester backbone, using energy from NTP hydrolysis to separate two annealed nucleic acid strands. These proteins thus play essential roles in cell survival. Another helicase, RecQ, is SOS inducible in E. coli. It appears to have several functions in the initiation of recombination, resolution of recombination intermediates and suppression of illegitimate recombination. RecQ is also required for proper induction of the ‘SOS’ response to stalled replication forks (Hishida et al., 2004; Chow & Courcelle, 2007). Notably, RecQ is absent in M. tuberculosis; however, some helicase domains characteristic of RecQ are found in UvrD2 (Sinha et al., 2008).

Recent studies have investigated the mechanisms underlying the response to DNA damage in M. tuberculosis in global analyses of gene expression following DNA damage in both wt and recA M. tuberculosis. Surprisingly, the majority of genes induced by DNA damage did not have known roles in DNA repair. However, most of the inducible genes with a known or predicted function in DNA repair were not dependent on recA for induction; indeed, these genes were induced to the same extent in the recA mutant as in the wild type. These findings thus suggest that this is a novel mechanism of gene regulation, and that recA-independent SOS induction is important for DNA repair in M. tuberculosis (Rand et al., 2003). One group of genes is thus strictly dependent on recA for induction, mediated exclusively by SOS boxes, and therefore, presumably regulated solely by LexA. These genes include lexA itself, as well as ruvA, ruvB and dnaE2. Expression of most of the other inducible DNA repair genes was found to be fully independent of recA. This is rare in bacteria and previously described only for Acinetobacter calcoaceticus (Rauch et al., 1996). This group of genes includes the majority of the inducible genes known or predicted to be involved in DNA repair, such as: uvrA, uvrB, nei, ogt, xthA, ung, uvrD, tagA, dnaE1 and ligB, and Rv1907, RV2191, Rv3201c, Rv3202c, Rv3644c (encoding putative DNA polymerases or helicases). Some genes appear to be subject to regulation by both recA-dependent and -independent mechanisms, including recA itself, radA, ruvC and ssb. Recently, a motif (RecA-NDp) has been identified that seems to define the RecA/LexA-independent promoter in M. tuberculosis (Gamulin et al., 2004).

It remains unclear why M. tuberculosis has several different mechanisms for the induction of DNA damage repair. It is possible that each mechanism responds to different stimuli. For example, the RecA-NDp-mechanism may be important for survival inside macrophages, because the majority of genes under the control of this promoter are induced to a significant level of expression in the macrophage environment (Gamulin et al., 2004).

Variation and conservation in M. tuberculosis DNA repair genes

We previously carried out an analysis of single-nucleotide polymorphisms (SNPs) in 3R genes in a comprehensive selection of M. tuberculosis complex strains from different countries worldwide. The analysis yielded significantly high levels of polymorphisms in these genes than in house-keeping genes, with a large number of these polymorphisms predicted to be potentially deleterious (Dos Vultos et al., 2008). It was also possible to clearly distinguish between 80% of the clinical isolates assessed, a task that is not feasible using current strain-typing techniques. The comparison of site frequency spectra of synonymous and nonsynonymous variants, together with analysis of Ka/Ks ratios, suggested that these sets of genes are subject to general negative/purifying selection, which might enable suboptimal function of the 3R system. In turn, this relaxed fidelity of 3R genes may lead to the generation of adaptive variants, some of which might survive. However, four DNA repair genes (mpg, xthA, ssb and ruvC) did not show any variation in their nucleotide sequence in a study of clinical strains (Dos Vultos et al., 2008), highlighting the essential functions of these genes.

In the same study, the ada/alkA gene was found to be highly polymorphic in a large proportion of the clinical strains assessed. The ada/alkA variants have a 79 AMBER codon in the ada part of the operon, suggesting that induction of both alkA and ogt differ in these variants (Friedberg et al., 1995; Nouvel et al., 2006, 2007). Polymorphisms in clinical strains associated with epidemics, such as the Haarlem (Mardassi et al., 2005) and W-Beijing (Rad et al., 2003) genotypes, have been described for the ogt gene. Mycobacterium tuberculosis needs a solid repair system to counteract the alkylating conditions to which it is subjected. Several aspects of M. tuberculosis alkylation repair remain unresolved. In particular, it remains unclear whether the activity of the fused proteins in M. tuberculosis differs from the activity of these proteins encoded and expressed as separate enzymes in other bacteria. Similarly, whether this expression pattern affects the genome stability of M. tuberculosis remains to be determined. Nevertheless, M. tuberculosis alkylation repair is clearly important. The SNP analysis study revealed that some of these DNA repair genes seemed to be subject to strong selective pressures, showing greater sequence variation than housekeeping genes; other genes, however, were perfectly conserved (Dos Vultos et al., 2008). Future studies should provide insight into the relevance of these observations in adaptive evolution.

Concluding remarks

Although still in its infancy, the study of M. tuberculosis DNA repair has already demonstrated the uniqueness of this system. The absence of known homologs for genes involved in MMR is consistent with this and seems indicative of the potential for M. tuberculosis to act as a natural mutator. However, the absence of MMR may have been an advantage for mycobacteria, for example by increasing the duplication rate of DNA repair genes. The presence of a robust error-prone repair system in M. tuberculosis, consisting of DnaE2 and proteins involved in NHEJ, may be expected to predispose bacteria to increased genome variability under stressful situations. The upregulation of several DNA repair genes upon exposure to various DNA-damaging agents demonstrates the importance of DNA repair for M. tuberculosis to survive in its host. Other DNA repair genes have been shown to be essential for the in vivo growth of mycobacteria. Additionally, several mutator phenotypes arising from deficiencies in DNA repair components have recently been demonstrated in M. tuberculosis. SNP analysis of DNA repair genes in clinical strains suggests that some of those strains may be natural mutators, and therefore predisposed to acquire drug resistance.

The improved understanding of M. tuberculosis DNA repair systems may allow us to identify the components that are essential for the growth of mycobacteria and use them as targets for antibiotherapy. It may also allow the identification of strains that are more capable of evolving, and/or resisting treatment.