Mycobacterial MazG Safeguards Genetic Stability via Housecleaning of 5-OH-dCTP

Generation of reactive oxygen species and reactive nitrogen species in phagocytes is an important innate immune response mechanism to eliminate microbial pathogens. It is known that deoxynucleotides (dNTPs), the precursor nucleotides to DNA synthesis, are one group of the significant targets for these oxidants and incorporation of oxidized dNTPs into genomic DNA may cause mutations and even cell death. Here we show that the mycobacterial dNTP pyrophosphohydrolase MazG safeguards the bacilli genome by degrading 5-OH-dCTP, thereby, preventing it from incorporation into DNA. Deletion of the (d)NTP pyrophosphohydrolase-encoding mazG in mycobacteria leads to a mutator phenotype both under oxidative stress and in the stationary phase of growth, resulting in increased CG to TA mutations. Biochemical analyses demonstrate that mycobacterial MazG can efficiently hydrolyze 5-OH-dCTP, an oxidized nucleotide that induces CG to TA mutation upon incorporation by polymerase. Moreover, chemical genetic analyses show that direct incorporation of 5-OH-dCTP into mazG-null mutant strain of Mycobacterium smegmatis (Msm) leads to a dose-dependent mutagenesis phenotype, indicating that 5-OH-dCTP is a natural substrate of mycobacterial MazG. Furthermore, deletion of mazG in Mycobacterium tuberculosis (Mtb) leads to reduced survival in activated macrophages and in the spleen of infected mice. This study not only characterizes the mycobacterial MazG as a novel pyrimidine-specific housecleaning enzyme that prevents CG to TA mutation by degrading 5-OH-dCTP but also reveals a genome-safeguarding mechanism for survival of Mtb in vivo.


Introduction
Oxidative damage to DNA and the DNA precursors, deoxynucleotides (dNTPs) is an inevitable mutagenic challenge occurring in normal aerobic metabolism, generating a large amount of reactive oxygen species (ROS) as by-products during respiration or oxidation-reduction reaction [1][2][3]. Oxidative DNA damage is also an important innate immune response mechanism implemented by phagocytes, which produce large amount of ROS and reactive nitrogen species (RNS) as a bactericidal strategy to eliminate microbial pathogens [4,5]. Increasing evidence shows that the nucleotide pool is a significant target for oxidative modification via ROS and substantial portion of the oxidative damage to genomic DNA is caused by incorporation of oxidized dNTPs from the nucleotide pool [3,6,7]. Due to their ambiguous conformation (anti/syn) compared to that of the canonical dNTPs, incorporation of oxidized dNTPs into DNA is known to cause mispairing and mutation, and may be related to carcinogenesis, aging and neurodegeneration [6,[8][9][10]. Recent studies also established that incorporation of oxidized dNTPs into DNA is a major causative mechanism for bacterial cell death induced by bactericidal antibiotics [11,12]. Therefore, like the DNA repair enzymes, elimination of the oxidatively damaged dNTPs from the nucleotide pool is an important defense line for cells to maintain genetic stability.
Cells have evolved a group of non-canonical nucleotide-specific bio-degradation enzymes, named housecleaning enzyme, to eliminate the oxidized non-canonical dNTPs from the nucleotide pool and thus prevent their incorporation into DNA [13,14]. These proteins belong to four structural superfamilies: 1) dUTPase, 2) ITPase, 3) Nudix (nucleoside diphosphate linked to an X moiety, or MutT-like) hydrolase, and 4) all-a NTP pyrophosphohydrolase (MazG NTP-PPase) [14]. The dUTPase and ITPase are NTP phosphatases that target dUTP, an intermediate during dTTP synthesis, and ITP/XTP, the deamination products of purine nucleotides, respectively. MutT is the best-studied Nudix hydrolase specific for oxidatively damaged nucleotides [14,15]. Escherichia coli MutT is the first characterized Nudix enzyme with 8-oxo-dGTP and 8-oxo-GTP as its natural substrates. Deletion of mutT in E. coli results in increased AT to CG mutation in both DNA and mRNA [8,16]. MTH1, the MutT-like protein in humans, is active against 8-oxo-dGTP, 8-oxo-dATP and 2-OH-dATP [17]. Depletion of MTH1 in mice leads to a higher incidence of spontaneous tumorigenesis [18], while in human cells, MTH1 is involved in maintenance of genome stability and suppression of degenerative disorders such as neurodegeneration and carcinogenesis [6,7,19]. However, all the natural substrates for the MutT-like proteins that have been characterized in various organisms so far have been the oxidized purine nucleotides [15].
Oxidized pyrimidine nucleotides likely have a mutagenic effect similar to that of oxidized purine nucleotides. First, dCTP and dTTP can be oxidatively modified by ROS to form 5-OH-dCTP and 5-CHO-dUTP, respectively [20,21]. Second, direct incorporation of 5-OH-dCTP or 5-CHO-dUTP into E. coli cells may cause an increase in mutation frequency, and both of these oxidized nucleotides may be mispaired with adenine rather than guanine leading to CG to TA mutation [10,22]. Furthermore, 5-OH-dCTP is known to be incorporated into DNA more efficiently than 8-oxo-dGTP catalyzed by the exonuclease-free Klenow fragment [10]. Finally, it was found that the amount of 5-OH-dC in normal or oxidized cellular DNA is comparable to that of 8oxo-dG [1,23]. In addition to their role in mutagenesis, oxidized pyrimidine nucleotides also show a highly lethal effect on E. coli, indicating that these non-canonical nucleotides may disturb normal DNA replication and nucleotide metabolism [22]. Therefore, it is reasonable to conclude that cells have evolved housecleaning enzymes to eliminate oxidized pyrimidine nucleotides [10,22]. However, although various enzymes responsible for the removal of oxidized pyrimidine in DNA molecules have been identified [24][25][26], the long-awaited housecleaning enzyme specific for elimination of oxidized pyrimidine nucleotides has yet to be characterized.
MazG-like proteins are widespread in all three domains of life and have been biochemically characterized as NTP-PPase while structurally categorized into the all-a NTP pyrophosphohydrolases superfamily unrelated to the MutT-like housecleaning enzymes [14,27,28]. It was found that E. coli MazG can regulate cellular (p)ppGpp levels and thus, may control programmed cell death under starvation conditions [29]. However, the mechanism whereby MazG regulates the cellular (p)ppGpp levels is still unclear. Structure-based modeling study of MazG from sulfolobus solfataricus suggested that 2-OH-dATP might be its most likely substrate and thus proposed, for the first time, a probable role of housecleaning for this enzyme [27]. Recently, it was reported that RS21-C6, a MazG-like enzyme in mice, showed a preference for degrading dCTP and its derivatives, with 5-I-dCTP as the most preferred substrate in vitro [30]. This attempt to identify a pyrimidine-specific housecleaning enzyme was echoed by structure-based analysis, which found that RS21-C6 binds to 5-methyl dCTP [31]. However, the natural substrate of the MazG family proteins remained unclear because none of the suggested functions involving regulation of pyrimidine synthesis [30], prevention of inappropriate DNA methylation [31], or antimutagenesis by degrading abnormal dCTP [27,30] have ever been verified in vivo.
Although mycobacterial MazG has been characterized as a potent NTP pyrophosphohydrolase capable of hydrolyzing all canonical (d)NTPs in vitro, MazG can also hydrolyze dUTP and 8oxo-dGTP, with their affinity to these substrates being similar to their affinity to the canonical dNTPs (K m ,1 mM). Therefore, it is unlikely that these non-canonical nucleotides are the natural substrate of the mycobacterial MazG in vivo [28]. In this study, we demonstrate that 5-OH-dCTP is a natural substrate of mycobacterial MazG by means of enzymatic and chemical genetic analyses. In addition to confirming the antimutator function of MazG, we show that deletion of mazG in the virulent Mtb strain H37Rv results in reduced survival in activated macrophage and mice. Our results reveal that mycobacterial MazG is a novel housecleaning enzyme involved in a pathway preventing the CG to TA mutation and ensuring the survival of Mtb in vivo.

Mycobacterial mazG is an antimutator
Previously, we demonstrated that lack of the MazG NTP-PPase activity in Msm strain mc 2 155 rendered the bacilli more susceptible to killing by hydrogen peroxide (H 2 O 2 ) [28]. In order to test whether the oxidative stress resistant effect of the mycobacterial MazG is truly attributable to its potential housecleaning function in degrading certain oxidatively damaged dNTP(s), the spontaneous rifampicin-resistance mutation frequencies in wild-type and mazG-null (DmazG::hyg) Msm (bacterial strains used in this study are list in Table S1) were measured under different physiological conditions. We showed that the rifampicinresistance mutation frequency in the mazG-null Msm increased 8.7fold when treated with H 2 O 2 (known to generate hydroxyl radicals which damage the dNTP pool [11,32]), in contrast to merely 2.5 times increase in the wild-type Msm ( Figure 1A). It was also found that, under the oxidative stress conditions, the expression level of recA and dnaE2, which is known to be elevated by damaged DNA (SOS response) [33], was 2-fold and 3-fold higher in the mazG-null Msm than that in the wild-type Msm, respectively ( Figure S1). This suggests that, under oxidative stress, the mazG-null Msm suffers more genetic assaults than does the wild type. On the other hand, during the exponential phase of growth, the

Author Summary
The cellular nucleotide pool is a significant target for oxidation by reactive oxygen species and reactive nitrogen species. Misincorporation of these oxidized non-canonical nucleotides into DNA is known to cause mutations, and may be related to carcinogenesis, aging and neurodegeneration. Cells have evolved a group of bio-degradation housecleaning enzymes that may specifically eliminate certain non-canonical nucleotide from the nucleotide pool and thus prevent their incorporation into DNA. The most well-characterized housecleaning enzymes are the MutTlike proteins which specifically hydrolyze the oxidized purine nucleotides, such as 8-oxo-dGTP and 2-OH-dATP. Lack of MutT activity in cells leads to significant increase of AT-CG mutation and genetic instability. However, housecleaning enzymes specific for oxidized pyrimidine nucleotides are yet to be identified. Here we show that the dNTP pyrophosphohydrolase MazG from mycobacteria is a 5-OH-dCTP-specific housecleaning enzyme. Deletion of mazG in mycobacteria results in increased CG to TA mutation under oxidative stress and in the stationary phase of growth. Both biochemical and chemical genetic analyses demonstrate that 5-OH-dCTP is a natural substrate of mycobacterial MazG. Furthermore, deletion of mazG in Mtb leads to reduced survival in activated macrophages and in the spleen of infected mice. These results reveal a novel housecleaning pathway for mycobacteria to maintain genetic stability and survival in vivo.
rifampicin-resistance mutation frequency of mazG-null Msm is comparable to that of the wild-type Msm ( Figure 1A).
We also measured the rifampicin-resistance mutation frequency in the stationary phase of growth, a stage known to accumulate metabolic byproducts and mutations [34]. It was found that the rifampicin-resistance mutation frequency in 5-day-old mazG-null Msm cultures was 2.5 times greater than that in wild type ( Figure 1A), suggesting a mutator phenotype of mazG-null Msm during the stationary phase of growth. A similar result was observed with 8-day-old cultures (data not shown), indicating prolonged incubation during the stationary phase does not further increase the mutation frequency in mazG-null Msm. To test whether mazG plays the same function in Mtb, we constructed a DmazG::hyg null mutation in the virulent Mtb strain H37Rv by means of allelic exchange and the deletion of this gene was confirmed by Southern blot ( Figure S2). The mazG-null Mtb exhibited the same mutator phenotype as that of the mazG-null Msm ( Figure 1B), showing a 2.5-fold increase in rifampicinresistance mutation frequency compared to that of the wild type under oxidative stress or the stationary phase of growth.
In order to test the cytotoxic effects of H 2 O 2 upon the bacteria studied, we measured the survival rates of the mazG-null strains versus the wild-type strains of Msm and Mtb under the same H 2 O 2 treatment conditions as that for mutation analysis. The survival rate of the mazG-null Msm decreased only slightly after 5 hours of H 2 O 2 treatment compared to that of the wild type, while in the mazG-null Mtb, no significant effect was observed ( Figure 1C). These data suggest that other than the change of mutation frequencies, H 2 O 2 treatment in this study did not induce other major physiological change affecting the survival of the bacilli. Because the wild-type mazG gene complements all of the defective phenotypes of the mazG-null mutants (Figure 1), the antimutator role of mycobacterial MazG is genetically established.

Mycobacterial MazG prevents CG to TA mutation
It has been shown that incorporation of different oxidized dNTPs into DNA preferentially induces a specific spectrum of mutation, e.g., 8-oxo-dGTP leads to AT to CG mutation [35,36] while 5-OH-dCTP induces GC to AT mutation [9,10,22]. Therefore, we compared the mutation spectra between the mazG-null Msm and its parental strains to infer the probable substrate of mycobacterial MazG.
We sequenced the cluster I region of the rpoB gene [37] from randomly isolated rifampicin-resistant colonies. All of the sequences contained single nonsynonymous nucleotide variations. Of these, .99% were located within the cluster I region (the remaining mutations occurred outside of the cluster I region) and caused mutations in the well characterized rifampicinresistance mutation hot spots (Table S2 and S3). Of the mutations detected, the frequency of CG to TA mutation exhibited a significant difference between the wild type and the mazG-null mutant ( Table 1). Among the rifampicin-resistant mutants derived from the exponential-phase cells, the CG to TA mutation frequency in wild-type Msm increased from 2.2610 28 in the untreated samples to 6.8610 28 in the H 2 O 2 treated cultures (,3-fold increase), while in the mazG-null Msm, the frequency of this type of mutations increased significantly from 0.8610 28 to 14.3610 28 (,18-fold increase). Meanwhile, the rifampicin-resistant mutants of the wild-type Msm exhibited a CG to TA mutation frequency in the stationary-phase cells similar to that in the exponential-phase cells (1.9610 28 and 2.2610 28 , respectively), suggesting that the CG to TA mutation rate is likely constant during replication in the wild-type Msm. However, the CG to TA mutation in the mazG-null Msm increased 26-fold, from 0.8610 28 in the exponential-phase cells to 20.5610 28 in the stationaryphase cells ( Table 1). These results clearly suggest that mycobacterial MazG is involved in safeguarding genomic DNA by preventing CG to TA mutation under adverse growth conditions.

5-OH-dCTP is a preferred substrate for mycobacterial MazG
CG to TA transition, the most common base substitution occurring in aerobic organisms [38,39], can be induced by incorporation of oxidatively damaged nucleotides into DNA, especially oxidized pyrimidine nucleotides [9,10,22]. We measured the MazG NTP-PPase activity towards 5-OH-dCTP, 5-CHO-dUTP and 2-OH-dATP, all of which are known to induce CG to TA mutation when incorporated into DNA [9,22,40].
Each substrate was mixed with mycobacterial MazG proteins of different origin and was incubated at 37uC for 10 minutes. The hydrolyzed product, pyrophosphate, was quantified by an enzyme coupled colorimetric method [28]. It was found that all of these substrates were hydrolyzed into monophosphate derivatives and pyrophosphate in a time-and enzyme concentration-dependent manner ( Figure 2A). Of the nucleotides examined, 5-OH-dCTP and 2-OH-dATP were the most preferred substrates for the Mtb MazG, with K m values of 1.9 and 2.4 mM, respectively, approximately 26 times lower than that of their canonical nucleotides ( Table 2 and Figure 2B). It appears that 5-CHO-dUTP is unlikely to be the natural substrate of Mtb MazG, shown by its extremely high K m value (,500 mM, Table 2). The Msm MazG exhibited similar kinetic constants compared to its Mtb counterpart, except for 2-OH-dATP, which showed a K m of 311 mM (Table S4), much higher than that of the Mtb MazG ( Table 2). Based on the kinetic constants and the same antimutator role of MazG in Msm and Mtb, we conclude that 5-OH-dCTP is the most likely natural substrate of mycobacterial MazG.

5-OH-dCTP is an in vivo substrate of mycobacterial MazG
To further characterize the natural substrate of mycobacterial MazG under cellular physiological conditions, we compared the in vivo mutagenic effects of these oxidized nucleotides in wild-type and mazG-null Msm strains using an established in vivo incorporation assay [22,36].
Of the nucleotides tested, only 5-OH-dCTP exhibited a mutagenic effect upon the mazG-null Msm in a dose dependent manner ( Figure 3A-C). When treated with 100 mM 5-OH-dCTP, the mazG-null mutant showed a ,2 fold increase (P,0.01) in rifampicin-resistance mutation frequency compared to that of the wild-type Msm ( Figure 3B). The increased mutation frequency of the mazG-null mutant can be restored to normal by complementation with a single copy of the wild-type mazG from Table 1. mazG-null Msm exhibited elevated CG to TA mutation under oxidative stress conditions and in the stationary phase of growth.
Growth phase Strain (n) Mutation frequency 610 28 (n) Spontaneous rifampicin-resistant colonies were collected from 3 independent experiments (see Methods). Cluster I region of rpoB were PCR-amplified using pfu DNA polymerase and sequenced bi-directionally. All of the sequenced colonies contain single non-synonymous mutations (see also  either Msm or Mtb, indicating that MazG plays the same role in these two mycobacteria species ( Figure 3B). Therefore, the antimutator role of mycobacterial MazG, particularly related to 5-OH-dCTP induced mutagenesis, is inferred. Furthermore, expression of the loss-of-function A219E MazG variant [28,41] in mazG-null Msm failed to restore the mutator phenotype ( Figure 3B), implying that the in vivo antimutator role of MazG requires NTP-PPase activity. We also found that the mazG-null Msm was susceptible to killing by 5-OH-dCTP treatment ( Figure 3D), suggesting that mycobacterial MazG was involved in the defense against both the cytotoxic and the mutagenic effects of 5-OH-dCTP. Based on these biochemical and chemical genetic results, we conclude that 5-OH-dCTP is one of the in vivo substrates of mycobacterial MazG. However, we cannot exclude the possibility that other substrates may also exist.
The mazG-null Mtb is hypersensitive to RNS During intracellular infection, Mtb is exposed to genetic assaults elicited by both ROS and RNS produced by host macrophages [5,42]. We tested whether the MazG housecleaning function is involved in Mtb resistance to ROS and RNS. The mazG-null Mtb was found to be more susceptible to killing by acidified nitrite treatment in vitro than the wild-type Mtb, showing a 0.8-log 10 lower CFU. The reduced survival ability of the mazG mutant can be fully restored by expression of a single copy of the wild-type mazG in the mutant ( Figure 4A). Accordingly, transcription of the mazG gene in the wild-type Mtb was upregulated 3,5 fold by the treatment of acid nitrite or DETA/NO (2,2-(hydroxynitrosohydrazino)-bisethanamine), which liberates nitric oxide ( Figure 4B), indicating that mazG is involved in the genetic response to RNS. However, unlike the mazG-null Msm [28], the mazG-null Mtb was just as susceptible to H 2 O 2 as the wild type ( Figure 1C), a property shared by other Mtb mutants which are more sensitive to RNS in vitro [43,44].
We further compared the intracellular survival ability between wild-type Mtb and the mazG-null mutant. No difference was observed between the growth of these two strains in resting macrophages ( Figure 4C). However, when infected with activated macrophages, the titer of the mazG-null Mtb declined from 2  days post-infection and onward ( Figure 4D), showing 1 log 10 lower CFU than that of wild-type Mtb by 5 days post-infection. This suggests that MazG is required for Mtb resistance to intracellular RNS. Consistent with this finding, the attenuated survival of mazG-null Mtb in activated macrophages was partially rescued by addition of NMMA (N G -Methyl-L-arginine acetate salt), a specific inhibitor for macrophage inducible NO synthase ( Figure 4E) [45]. Introduction of a wild-type mazG into the mutant strain restored the attenuated phenotype of the mazG-null Mtb ( Figure 4D-E).

MazG is required for Mtb survival in vivo
Our results demonstrate that Mtb MazG is required for maintenance of genetic stability and resistance to RNS both in vitro and ex vivo, indicating that MazG may function as a virulence factor during Mtb infection. To investigate whether mazG is involved in survival of Mtb in vivo, immune-competent mice were infected by a low-dose aerosol challenge with Mtb strains. Compared to wild-type Mtb, the mazG-null Mtb exhibited 1.1log 10 lower CFU in mice spleens by 4 weeks post-infection, and 0.7-log 10 lower CFU by 8 and 12 weeks post-infection, indicating an attenuation at the stage of persistent infection ( Figure 5A). No significant difference between the growth of wild-type and the mazG-null Mtb strains was observed in mice lung ( Figure 5B). However, histological stained sections of the infected lung tissue (8 weeks after infection) showed that the mazG-null Mtb caused minimal pneumonitis, with a 3-fold reduction in lung inflammation compared to wild-type Mtb ( Figure 5C). The attenuated phenotype of the mazG-null mutant can be fully restored by complementing the mutant with a single copy of the wild-type mazG ( Figure 5). Taken together, these results suggest that the housecleaning role of MazG is required for Mtb survival and pathogenesis in vivo.

Discussion
Growing evidence suggests that elimination of oxidized nucleotides from the cellular dNTP pool is an important safeguarding mechanism for maintenance of genetic stability [2,3,6,7]. However, most of the knowledge about this housecleaning role has focused on oxidized purine nucleotides and the MutTlike NTP-PPase. Due to lack of knowledge of pyrimidine specific housecleaning NTP-PPase, the contribution of oxidized pyrimidine nucleotides to DNA mutagenesis and its related mechanism remains unclear. Here we characterized a NTP-PPase that specifically degrades 5-OH-dCTP in vivo and prevents CG to TA mutation. To our knowledge, MazG is the first oxidized pyrimidine-specific housecleaning enzyme to which an antimutator function can be assigned.
Our previous study suggested that 8-oxo-dGTP and dUTP are unlikely to be the natural substrates of mycobacterial MazG, as the K m values for these nucleotides are substantially high [28]. In this study, based on the observation that deletion of mazG leads to the increase of CG to TA mutation frequency, we considered 5-OH-dCTP, 5-CHO-dUTP and 2-OH-dATP as putative substrates of MazG, and all of these oxidized nucleotides are known to induce CG to TA mutations upon incorporated into DNA by polymerase [9,22,40]. Among the substrates tested, mycobacterial MazG showed high affinity towards 5-OH-dCTP (K m = 1.9 mM) and 2-OH-dATP (K m = 2.4 mM), values comparable to that of E. coli MutT towards its natural substrate 8-oxo-dGTP (K m ,0.5 mM) [8]. Through direct incorporation of these oxidized nucleotides into Msm cells, we observed a dose-dependent 5-OH-dCTP-specific mutagenic effect in mazG-null Msm, indicating 5-OH-dCTP is an in vivo substrate of mycobacterial MazG (Figure 3). Moreover, we also found that the mycobacterial MazG was involved in the defense against the cytotoxic effect of 5-OH-dCTP ( Figure 3D). This cytotoxic effect is likely to be caused by lethal DNA strand breaks or replication block [11,19] induced by incorporation of 5-OH-dCTP into DNA. Taken together, these data demonstrate that the mycobacterial MazG is a 5-OH-dCTP-specific housecleaning enzyme involved in preventing CG to TA mutation.
In mazG-null Msm with H 2 O 2 treatment, we did not observe an increase of GC-TA mutations that should be induced by incorporation of 2-OH-dATP [46]. On the other hand, MazG does exhibit a high affinity to 2-OH-dATP, similar to its affinity to 5-OH-dCTP ( Table 2). It is still unclear whether MazG is the only 2-OH-dATPase existing in mycobacteria. Therefore, whether 2-OH-dATP is a natural substrate of mycobacterial MazG remains unclear. It is worth noting that Mtb encodes four MutT proteins (MutT1 to 4) [47]. A recent study showed that the MutT1 carried out the physiological role of MutT (8-oxo-dGTPase) in Mtb [48], while the MutT2 did not function as an 8-oxo-dGTPase [49]. Therefore, to date, the natural substrate of MutT2-4 is still unidentified.
CG to TA transition can be induced by oxidative deamination of cytosines on DNA [50], misincorporation of oxidized pyrimidine nucleotides into DNA by DNA polymerase [9,10,22] and mismatch induced by keto-enol transitions of guanine [51]. Based on the biochemical and chemical genetic results described above, and the fact that MazG is unlikely to perform a DNA repair role, as this protein family does not contain any DNA binding/repair signatures [27,28], we conclude that the increased CG to TA mutation in mazG-null Msm is mainly due to incorporation of oxidized nucleotide 5-OH-dCTP.
Deletion of mazG in mycobacteria did not lead to a mutator phenotype under the exponential phase of growth. However, lack of MazG activity in mycobacteria resulted in higher CG to TA mutation under both oxidative stress and the stationary growth phase, compared to that of the parental strains ( Table 1). The likely mechanism underlying this stress-related mutagenesis is that under stress conditions, mycobacterial cells may accumulate 5-OH-dCTP and lacking MazG, more 5-OH-dCTP is misincorporated into DNA. Moreover, down-regulated DNA repair activity under these stress conditions [34,52] may also contribute to the stress-related mutagenesis observed in the mazG-null strains. Nonetheless, although the related molecular mechanism of 5-OH-dCTP induced stress-related mutagenesis remains to be determined, as the host environment for Mtb parasitism is always adverse, 5-OH-dCTP induced mutagenesis may be hypothesized to play an important role in the microevolution process of the infected Mtb under stress conditions, i.e. emergence of drug resistant mutations during bacterial infection. In this connection, it is worth noting that CG to TA transition is a dominant mutation in Mtb isolated from either macaques with latent/reactivated infection or humans [53,54].
It is still unclear why the lack of the 5-OH-dCTP sanitization function in Mtb results in hypersusceptibility to RNS (Figure 4). RNS is a group of radicals derived from nitric oxide (NO N ) which are produced by macrophage as antimicrobial effector molecules [5]. An important antimicrobial action of RNS is inhibition of DNA replication and repair. It was found that NO N can inhibit DNA synthesis by zinc mobilization from DNA-binding metalloproteins [55]. RNS can also inhibit ribonucleotide reductase [56], and thus, limit the availability of precursors for the synthesis and repair of DNA. Based on the fact that lack of MazG activity leads to increased incorporation of 5-OH-dCTP into DNA (Table 1), a possible explanation for the hypersusceptibility of mazG-null Mtb to RNS is that inhibition of DNA repair activities and lack of DNA precursors caused by RNS mediated enzyme inactivation result in higher levels of genetic instability (such as DNA strand breakage) in mazG-null Mtb than that in the wild-type Mtb.
During infection, Mtb is exposed to an oxidative environment of host macrophages rich in DNA-damaging ROS and RNS. Therefore, safeguarding of the genetic information is essential for mycobacterial survival, especially during the non-replicating dormancy stage, as slow or non-replicated genomic DNA and diminished DNA repair activities are likely lead to more genetic assaults than that during fast growing phase [52,57]. Our results demonstrated that deletion of mazG leads to attenuated survival of Mtb in mice spleen during the persistent infection phase, suggesting that oxidative damage to nucleotides and the subsequent genetic assault is one of the bactericidal effects of the adaptive immune response (corresponding to the bacterial persistent infection stage). This is consistent with the data indicating that genes involved in removal of oxidized pyrimidines are essential for Mtb survival during primates' infection [58]. Although the difference between the lung and spleen microenvironments exposed to Mtb is unclear, it is conceivable that the immune responses and metabolic constraints are different between the two tissues. Interestingly, tissue specific attenuation have been demonstrated for several Mtb mutants, including the dosR, fadD26, mptpB and narG mutants [59][60][61][62].
Recent studies have proven that bactericidal antibiotics-induced ROS production within bacterial cells is a common mechanism for cell death [12,[63][64][65], predominantly elicited by incorporation of 8-oxo-dGTP into DNA [11]. Therefore, it is not surprising that 5-OH-dCTP and other oxidized nucleotides have a similar bactericidal effect, as shown in our results ( Figure 3D and Figure 5A). These findings suggest that clinical treatment of tuberculosis with specific inhibitors of housecleaning enzymes might facilitate Mtb elimination, especially when combined with bactericidal antibiotics which are known to induce oxidative stress.

Ethics statement
Six-to-eight week old female C57BL/6 mice were purchased from the Shanghai SLAC Laboratory Animal Company. The mice were housed and cared for in a specific pathogen-free (

Bacterial strains and culture conditions
Bacterial strains used in this study are list in Table S1. Bacterial culturing was performed as described [66]. Msm strains were grown at 37uC in 7H9 broth (BD Difco), or on Luria-Bertani agar supplemented with 0.5% glycerol (LBG agar). Mtb strains were grown at 37uC in 7H9 broth supplemented with 10% OADC (7H9-OADC), or on 7H11 plates supplemented with 10% OADC (7H11-OADC). When required, the following antibiotics were used at the specified concentrations: kanamycin (15 mg/ml), hygromycin B (150 mg/ml for Msm and 50 mg/ml for Mtb) and rifampicin (250 mg/ml for Msm and 10 mg/ml for Mtb). For treatment with acid NO, Mtb stains grown to OD 600 ,0.5 were pelleted and re-suspended in 7H9-OADC pH 4.5 (adjusted by 1 M citrate) with or without 2.5 mM NaNO 2 [45]. After 20 h treatment, bacteria were plated on 7H11-OADC, CFUs were counted after 3,4 weeks culture at 37uC

Generation of mazG mutants and complemented strains
The mazG-null mutant was generated by the phage transduction method [66]. mazG-null Msm and the complemented strains were generated as described [28]. To construct a transducing phage for Mtb mazG knockout, the left homologue arm was PCR amplified using primers KOP1 and KOP2 (primers used in this study are listed in Table S5). The right homologue arm was PCR amplified using primers KOP3 and KOP4. The PCR products were ligated into the AflII/XbaI and HindIII/XhoI sites of pYUB854. The recombinant transducing phage was used to construct the mazGnull Mtb as described [28]. The mazG-null mutant was verified by southern blot and PCR ( Figure S2). The probe for Southern blot was PCR amplified using primers SB1and SB2. A dUTP-biotin labeled probe (Fermentas) was used for Southern blot analysis of the PstI/KpnI digested chromosomal DNA on the Hybond-N + nylon membrane (GE Amersham), according to the standard method [67]. Primers used for genotyping PCR were P1, P2 and P3. The complementation plasmid for mazG-null Mtb was generated by ligating the PCR product amplified using primers C1 and C2 into the BamHI and HindIII sites of pMV306. Expression of Mtb mazG was controlled by its own promoter (1142671-1143646).

Determination of rifampicin-resistance mutation frequency
Single colonies of various Msm strains from the 7H11 agar plate were inoculated in 5 ml media and cultured at 37uC for 48 h (2 weeks for Mtb strains). For determination of rifampicin-resistance mutation frequency, the cultures were inoculated with 1% of primary culture in 20 ml 7H9 media (in a 100-ml flask) without antibiotics and grown at 37uC with rolling (150 rpm) to exponential phase (OD 600 ,0.5). Then 10 ml of the cultures were treated with 10 mM H 2 O 2 , and another 10 ml cultures were untreated. After incubation at 37uC, 150 rpm, for 5 h, CFU per ml was determined by plating; the cell pellet from 3 ml culture was plated on LBG agar (3 plates of each sample) containing 250 mg/ml rifampicin (Sigma-Aldrich). The CFU and rifampicin-resistant colonies were counted after culturing at 37uC for 4 days (28 days for Mtb strains). The rifampicin-resistance mutant frequency was calculated by dividing the number of rifampicin-resistant colonies on each plate by the counts of the total viable cells plated. Rifampicin-resistance mutation frequencies of Mtb strains were determined by the same method, except that the oxidative stress was elicited by resuspending the exponential-phase cell pellet in 7H9 media containing 10 mM H 2 O 2 , followed with incubating at 37uC for 24 h. Mtb strains were plated on 7H11-OADC with or without 10 mg/ml rifampicin. For determination of rifampicin-resistance mutation frequency of the stationary phase cultures, cells were cultured in liquid media for 5 days (for Msm) or 28 days (for Mtb) and plated as described above. Three independent experiments were performed with totaling 15 cultures of each Msm strains and 6 of Mtb.

Analysis of mutation spectra
Rifampicin-resistant colonies were collected from three independent experiments. The isolated colonies were grown in 1 ml 7H9 at 37uC for 1 week. Cells were pelleted and suspended in 50 ml TE buffer (10 mM Tris-HCl, pH 8.0, and 1 mM EDTA) and incubated at 95uC for 10 min to extract the genome DNA [37]. The lysate was centrifuged at 12000 g for 5 min. The supernatant was used as template to amplify (using pfu DNA polymerase) the fragment containing the cluster I region of rpoB using primers Rpo1 and Rpo2. All PCR products were sequenced by bi-directionally. Mutation spectra of the sequenced region were analyzed by BioEdit software.

NTP-PPase assay
Protein expression and purification was performed as described [28]. Protein was purified to nearly SDS-PAGE homogeneity. Protein concentration was determined by the bicinchoninic acid (BCA) method [68]. The oxidized nucleotides used as substrates for MazG were purchased from TriLink Biotechnologies Inc. (5-OH-dCTP) or Hongene Biotechnologies Inc. (2-OH-dATP and 5-CHO-dUTP). The NTP-PPase activity of MazG was assayed as described [28]. The NTP-PPase assay was carried out in 20 ml reaction buffer (20 mM Tris-HCl, pH 7.5, 5 mM MgSO 4 , 100 mM NaCl) containing 1 mg mycobacterial MazG and substrate nucleoside triphosphates at 37uC for 10,20 min. The reaction was stopped by heating at 65uC for 5 min, and 10-20 ml products were applied for pyrophosphate assay (Molecular Probes) according to the manufacturer's instructions. Reactions with heat inactivated (95uC for 20 min) MazG protein were set up as a background controls. GraphPad Prism 5.0 (GraphPad Software, Inc.) was used for enzyme kinetic constants analysis.

Incorporation of nucleotides into Msm competent cells
The Msm competent cells were prepared from 400-ml cultures (OD 600 = 0.8,1.0) as previously described [69]. Incorporation of nucleotides into Msm competent cells was performed as described [22,36]. Briefly, Nucleotide solution (100 mM final concentration) was added to 150 ml competent cells suspension and the mixture was placed on ice for 10 min. After heat shock treatment (42uC for 90 sec and then on ice for 30 min), 2 ml 7H9 was added and the cells were incubated at 37uC with rolling (150 rpm) for 5 h. After treatment, 2 ml of culture was centrifuged at 4000 g for 5 min and plated on LBG agar containing 250 mg/ml rifampicin. The remaining culture was diluted and plated onto LBG agar for CFU determination. Rifampicin-resistance mutation frequencies were calculated as described above.

RNA extraction and quantitative real-time PCR
Wild-type Mtb (OD 600 ,0.5) was treated with acid NO or 2.5 mM DETA/NO for 1 h. Total RNA was extracted with TRIzol-Reagent (Invitrogen) and further purified with RiboPure-Bacteria kit (Ambion). Briefly, cell pellet was resuspended in 1 ml TRIzol reagent, mixed with 400 ml 0.1 mm Zirconia Beads (BioSpec Products) and lysed in a mini-beadbeater (Biospec) for three cycles (40 s at maximal speed) with cooling on ice for 1 min between pulses. RNA was extracted according to the protocol of TRIzol-Reagent. The extracted RNA was further purified using the RiboPure-Bacteria kit followed by DNase I treatment to eliminate DNA contamination. cDNA was synthesized using the SuperScript III First Strand kit (Invitrogen) with random hexamer primer. Target gene transcript levels were measured by real-time PCR using SYBRH Premix Ex Taq GC (TaKaRa) on Mastercycler ep realplex thermal cyclers: 95uC 60 sec, 40 cycles of 95uC 5 sec, 62uC 8 sec and 72uC 20 sec, followed by melting curve analysis. Data were normalized to sigA and expressed as fold change compared to the untreated samples. PCR primers for sigA, mazG, dosR, recA and dnaE2 are listed in Table S5.

Macrophage infection
The murine macrophage cell line RAW264.7 was grown in DMEM medium (GIBCO) supplemented with 10% fetal calf serum (FCS) and incubated at 37uC with 5% CO 2 . For Mtb infection, cells were plated at a density of 2.0610 5 cells per well in 24-well plates without antibiotics and activated with 200 U/ml murine IFN-c (Peprotech) for 16 h [45]. Cells were primed with 1 mg/ml lipopolysaccharides (LPS, Sigma) for 1 h and then infected at a multiplicity of infection (MOI) of 2:1 (bacteria:cells). After 4 h incubation at 37uC with 5% CO 2 , cells were washed three times with DMEM to remove extracellular bacteria and cultured with complete DMEM medium. To inhibit macrophage NO production during the infection cause, NMMA (Sigma) was added to the culture medium at a final concentration of 400 mM. At indicated time points, bacteria were released with PBS solution containing 0.05% Tween-80 and 0.025% SDS, and plated onto 7H11-OADC plates. CFUs were counted after 3,4 weeks culture at 37uC.

Mice infection
Mice were infected with wild-type Mtb, mazG-null mutant or complemented strain at an inhaled aerosol dose of 100-200 CFU per lung by an inhalation exposure system (Glas-Col, Terre Haute, IN). At indicated time points, mice were sacrificed and lung and spleen homogenates (four mice per group) were plated onto 7H11-OADC followed by incubation at 37uC for 4 weeks.

Histopathology
Lung sections stained with hematoxylin and eosin were photographed using a Nikon Optiphot 2 microscope fitted with a camera which was connected to a computer. The Image Pro Plus program (Media Cybernetics) was utilized to objectively assess the level of inflammation present in each image. To quantify the percent area inflamed, we determined the mean percent inflamed area from three to five lung sections of each mouse.

Statistical analysis
Statistical significance was determined with the unpaired twotailed Student's t test at P,0.05 level of significance using GraphPad Prism 5.0 software. Figure S1 mazG-null Msm exhibited higher level of SOS response under oxidative stress. Expression level of recA and dnaE2 from exponential phase bacteria and oxidative stressed samples (treated with 10 mM H 2 O 2 for 1 h) were measured by quantitative real-time PCR and normalized to sigA. Shown are fold change compared to the untreated samples. wt, wild-type Msm; DmazG, mazG-null Msm. Mean6 S.E. of three independent repeats. (TIF)