Essential roles of methionine and S-adenosylmethionine in the autarkic lifestyle of Mycobacterium tuberculosis

Significance

Mycobacterium tuberculosis (Mtb) is the most deadly bacterial pathogen in the world, killing 1.5 million people in 2013. Very little is known about the way this pathogen interacts metabolically with its host to achieve long-term persistence and antibiotic tolerance. We uncovered a previously unknown metabolic vulnerability of Mtb, the absolute requirement for methionine and S-adenosylmethionine for successful host infection and virulence. Inactivation of methionine biosynthesis in Mtb leads to unusually rapid cell death, a highly desired feature for chemotherapy. Bactericidal auxotrophies are rare in Mtb, and so far their killing mechanisms have not been characterized systematically. Our study unravels a killing mechanism induced by amino acid starvation that leads to multitarget inhibition, opening new avenues for antimycobacterial interventions.

Abstract

Multidrug resistance, strong side effects, and compliance problems in TB chemotherapy mandate new ways to kill Mycobacterium tuberculosis (Mtb). Here we show that deletion of the gene encoding homoserine transacetylase (metA) inactivates methionine and S-adenosylmethionine (SAM) biosynthesis in Mtb and renders this pathogen exquisitely sensitive to killing in immunocompetent or immunocompromised mice, leading to rapid clearance from host tissues. Mtb ΔmetA is unable to proliferate in primary human macrophages, and in vitro starvation leads to extraordinarily rapid killing with no appearance of suppressor mutants. Cell death of Mtb ΔmetA is faster than that of other auxotrophic mutants (i.e., tryptophan, pantothenate, leucine, biotin), suggesting a particularly potent mechanism of killing. Time-course metabolomics showed complete depletion of intracellular methionine and SAM. SAM depletion was consistent with a significant decrease in methylation at the DNA level (measured by single-molecule real-time sequencing) and with the induction of several essential methyltransferases involved in biotin and menaquinone biosynthesis, both of which are vital biological processes and validated targets of antimycobacterial drugs. Mtb ΔmetA could be partially rescued by biotin supplementation, confirming a multitarget cell death mechanism. The work presented here uncovers a previously unidentified vulnerability of Mtb—the incapacity to scavenge intermediates of SAM and methionine biosynthesis from the host. This vulnerability unveils an entirely new drug target space with the promise of rapid killing of the tubercle bacillus by a new mechanism of action.

Understanding the metabolic interactions between an invading microbe and its host is becoming a new cornerstone of host–pathogen research (1–3). Many intracellular pathogens modulate the host response to satisfy their nutritional needs and as a result have become auxotrophic for several essential amino acids and cofactors (4–6). Mycobacterium tuberculosis (Mtb), arguably the most deadly bacterial pathogen in the world (7), adopted a different strategy. This ultra-slow-growing bacterium is prototrophic for all essential cofactors and amino acids, suggesting that it either dwells in host compartments where such metabolites are unavailable or actively chooses this autarkic lifestyle to retain metabolic flexibility and remain invisible to the host. Indeed, much of Mtb’s long-term success as a human pathogen is ascribed to its extraordinary stealth in the face of host immunity (8, 9); Mtb’s ability to evade detection by the host might explain why devising an efficient vaccine has failed thus far and why drug therapy is difficult. Therefore, understanding Mtb’s in vivo metabolic requirements could help in the development of much-needed new strategies for antimycobacterial therapy.

Methionine and S-adenosylmethionine (SAM) are essential metabolites that have gained considerable scientific attention because of their recently discovered roles as sentinel metabolites in the control of the eukaryotic cell cycle (10), autophagy (11), and differentiation of human pluripotent stem cells (12). Methionine takes a leading role in translation initiation (13) and is the precursor of SAM, a cofactor for one-carbon metabolism, the process responsible for the methylation of DNA, RNA, proteins, and lipids by SAM-dependent methyltransferases (14). Methylation can have direct biochemical relevance in metabolic reactions and is involved in regulatory mechanisms at the epigenetic level (14). The many essential functions of methionine and SAM prompted us to investigate their roles in the context of Mtb’s cellular metabolism and pathogen–host interaction. To do so, we used an integrated approach of mycobacterial genetics, in vivo (mouse) and ex vivo (human macrophages) experiments, ultrasensitive metabolite analysis, transcriptomics, and biochemical assays.

Results

Methionine and SAM biosynthesis in Mtb involves up to six enzymatic reactions starting with homoserine (Fig. 1A). To construct a mutant deficient in producing methionine, we surveyed bioinformatics databases and literature from genetic screens (8, 15). Given the premise that the ideal target should be essential for growth and have no human homolog, homoserine transacetylase (MetA), the first dedicated enzyme in methionine biosynthesis, was the only candidate that fulfilled both these requirements (Fig. 1A). First, we sought clarity about the biochemical function of MetA_TB. The gene encoding MetA_TB was cloned and expressed in Escherichia coli. Purified recombinant MetA_TB is a disulfide bond-linked homodimer (Fig. S1) that is specifically active on the substrate acetyl-CoA at pH 6–10 (K_m, 50 μM; k_cat, 2.4/s) (Fig. 1B); no enzymatic activity was observed with succinyl-CoA (Fig. 1C), nor are Acetyl-CoA analogs with a longer carbon chain, such as myristoyl-CoA, substrates of MetA_TB. We observed no feedback inhibition of MetA_TB by SAM or methionine at 3.2 mM and 12.8 mM, respectively, indicating that MetA_TB is not allosterically regulated by these compounds. After gaining biochemical evidence that metA_TB encodes a homoserine transacetylase, we constructed unmarked deletion strains (∆metA) and genetically complemented strains (∆metAcomp) in the background of Mtb H37Rv, Mtb CDC1551, and Mycobacterium bovis bacillus Calmette–Guérin (Tables S1–S3). Neither mutant could grow in unsupplemented medium, but genetic complementation or the addition of methionine (50 µg/mL) restored growth fully (Fig. 1 D and E and Fig. S2). Remarkably, ∆metA, regardless of the genetic background, died rapidly in unsupplemented medium (Fig. 1F and G and Fig. S2), and no revertants could be isolated from three independent pools, each containing 10⁹ mutants. Mtb ∆metA displayed extraordinarily rapid cell death compared with other, previously published auxotrophic mutants of Mtb [i.e., ∆trpD (16), ∆panCD (17), and ∆leuCD (18)] (Fig. 1G). Importantly, ∆metA was chemically complemented by all pathway intermediates except homoserine (Fig. 1H).

Fig. 1.

MetA encodes a homoserine transacetylase that is essential for growth in unsupplemented 7H9 medium. (A) Methionine and SAM biosynthetic pathway in Mtb. MetA, homoserine O-acetyltransferase; MetB, cystathionine gamma-synthase; MetC, O-acetylhomoserine sulfhydrylase; MetE/MetH/MmUM, methionine synthase; MetK, S-adenosylmethionine synthetase; Rv0075/Rv2294, cystathionine beta-lyase. (B) Lineweaver–Burk plot of the inverses of initial enzymatic reaction rates versus the inverses of multiple concentrations of homoserine. The Michaelis–Menten constant (K_m) and k_cat were calculated as 5 × 10⁻² mM and 2.4 s⁻¹, respectively. (C) Time courses of the enzymatic reaction with acetyl CoA (Ac-CoA) and succinyl CoA (Succ-CoA). The formation of CoA was monitored by the Ellman’s reagent with absorption at 412 nm. (D and E) Growth of WT (blue), ΔmetA (red), and ΔmetAcomp (black) strains in 7H9 medium (D) or supplemented with 50 µg/mL methionine (E). (F) Survival of ΔmetA in the presence or absence of 3 μg/mL methionine. Cultures were serially diluted and plated on 7H10 plates containing 50 µg/mL methionine. Detection limit, 100 cfu/mL. (G) Survival curves of different auxotrophic mutants in the Mtb CDC1551 background starved in unsupplemented medium. Detection limit, 10⁻⁶. ΔleuCD, leucine auxotroph; ΔpanCD, pantothenate auxotroph; ΔtrpD, tryptophan auxotroph. (H) Rescue of growth of ΔmetA with 125 µM of different pathway intermediates of methionine biosynthesis. Ac-HoSe, O-acetyl homoserine; HoCy, homocysteine; HoSe, homoserine; Met, methionine; SAM, S-adenosylmethionine.

With the metA mutant in hand, we could probe for the availability of those pathway intermediates in the host environment. Typically, human serum contains around 3 μg/mL methionine (mouse serum contains 9 μg/mL) (19, 20), which is enough to allow in vitro growth of ∆metA (Fig. 1E) (minimal growth-promoting concentration: 200 ng/uL). However, ∆metA was unable to establish an infection in immunocompetent (C57BL/6) or immunocompromised (SCID) mice via aerosol or i.v. infection. WT and complemented strains elicited typical disease progression in low-dose aerosol infection of C57BL/6 mice with a lung burden of 10⁵–10⁶ bacteria 3 wk postchallenge (Fig. 2A) and granulomatous foci on lungs 8 wk after infection (Fig. 2B). In contrast, ∆metA was cleared from mouse lungs 3 wk after challenge with no signs of granulomatous lesions on lungs at 8 wk postinfection (Fig. 2B), and no bacteria were recovered from spleens (Fig. 2C). This severely attenuated nature of ∆metA also was observed in SCID mice (Fig. 2 D–G). In contrast, WT and ∆metAcomp grew unrestrictedly in the SCID lungs and spleens (Fig. 2 D and F), resulting in 100% mortality 7 wk postinfection (Fig. 2H). Lung histopathology data and gross histological appearance corroborated these results (Fig. 2 E and J), showing exuberant tissue inflammatory infiltrates and conspicuous acid-fast bacilli (AFB) in mice infected with WT and ∆metAcomp (Fig. S3 and Table S4), whereas virtually no inflammation, gross pathology, or bacteria was detected in ∆metA-infected mice 6 wk postchallenge (Fig. 2 E and J and Fig. S3 and Table S4). To achieve a more systemic infection, we challenged SCID mice i.v. with a high bacterial dose (10⁶ cfu). Mice succumbed to the WT or ∆metAcomp infection after 22 d, but those infected with ∆metA lived more than 300 d with no signs of sickness (Fig. 2I). These results indicated that ∆metA resides in a host niche that does not support rescue of the mutant. To test this hypothesis, we studied ∆metA’s interaction with human macrophages, which are regarded as Mtb’s preferred niche in human tissues.

Fig. 2.

Mtb ΔmetA is avirulent in immunocompetent (C57BL/6) mice, immunocompromised (SCID) mice, and human macrophages. Mice were infected with Mtb H37Rv (blue), ΔmetA (red), or ΔmetAcomp (black) via a low-dose aerosol (100 bacilli), and bacterial burdens in lungs and spleens were measured at 1, 7, 21, 56, and 112 d postinfection in C57BL/6 mice and at 1, 7, 21, and 42 d postinfection in SCID mice. (A–C) C57BL/6 mice. (A) The cfu in lung. (B) Gross appearance of lungs demonstrating pathology induced in infected mice on day 56. (C) The cfu in spleen. (D–F) SCID mice. (D) The cfu in lung. (E) Gross lung pathology on day 42. (F) The cfu in spleen. (G) Spleen pathology in SCID mice on day 42. Spleens cross-sectioned at the point of entry of the splenic vasculatures demonstrate varying degrees of splenomegaly as a readout for the severity of infection. (H) Survival of SCID mice (n = 6) after low-dose aerosol infection. (I) Survival of SCID mice (n = 6) after high-dose (10⁶ bacilli) infection via tail vein injection. (J) Lung pathology highlighted by H&E staining. No signs of inflammation were detected with ΔmetA strain infection. (K) Primary human macrophages infected at an MOI of 0.02 were incubated for 2 wk in RPMI medium supplemented with four different methionine concentrations (0, 3, 15, and 50 µg/mL). Macrophages were lysed, serially diluted, and plated for cfu at 0, 7, and 14 d postinfection. Error bars represent the SD of three biological replicates. ***P < 0.001; two-tailed t-test.

Infection of macrophages differentiated from human peripheral blood-derived mononuclear cells (PBMCs) at a multiplicity of infection (MOI) of 0.02 confirmed that ∆metA is incapable of scavenging sufficient intermediates within these phagocytic cells to support proliferation or survival. Surprisingly, even in the presence of physiological methionine concentrations (3 µg/mL), ∆metA died in the macrophages (Fig. 2K). Only the presence of 50 µg/mL methionine in the extracellular medium fully restored growth in macrophages to WT levels. These macrophage results argue strongly that access to O-acetylhomoserine, homocysteine, methionine, and SAM is severely restricted in human macrophages.

Rapid cell death, the inability to establish a productive infection, and the absence of histopathological manifestations in tissue pointed to a unique mode of killing. To elucidate this killing mechanism, we searched for correlates of death using metabolomics and transcriptomics. Metabolite and transcript levels in Mtb were measured during a time-course starvation experiment in unsupplemented medium, using untargeted high-sensitivity ultra performance liquid chromatography (UPLC)-MS and microarray analysis.

Metabolites in pathways downstream of MetA, including methionine and SAM, were rapidly depleted to undetectable levels upon starvation (Fig. 3A). Metabolic block at MetA created backpressure on the homoserine biosynthesis pathway with a transient increase in aspartate and possibly a redirection of metabolites into lysine metabolism as evidenced by the increased size of lysine and aminoadipate pools (Fig. 3A). These data provided biochemical proof that MetA is essential for methionine and SAM biosynthesis and suggested that the previously proposed methionine/SAM salvage pathway (21) is insufficient to rescue ∆metA. Interrogation of the metabolome for changes in carbon metabolism revealed that succinate, malate, and glyoxylate pools decreased significantly in WT and complemented strains but were unchanged in the mutant (Fig. 3B). The marked decrease of these central metabolites was paralleled by a 100-fold increase in the secondary messenger cyclic-AMP in ∆metA (Fig. 3B).

Fig. 3.

Time-course metabolic profile of ∆metA compared with WT and ∆metAcomp during 6 d of starvation. Samples from three biologically independent replicates were harvested on days 0, 2, 4, and 6, and metabolites were extracted. Aqueous-phase metabolites were measured by UPLC-MS. Fold changes in metabolite abundance were calculated relative to time 0. (A) Methionine biosynthetic pathway including SAM metabolism and lysine metabolism. In the ∆metA strain metabolites downstream of MetA are depleted, whereas those upstream accumulate. (B) Changes in metabolite abundance of glyoxylate shunt intermediates and cAMP. MetA, homoserine transacetylase; MetC, O-acetylhomoserine sulfhydrylase; MetE/MetH, methionine synthase; MetK, S-adenosylmethionine synthetase; ThrA, homoserine dehydrogenase. Error bars represent the SD of three biologically independent replicates. *P < 0.05; **P < 0.01; ***P < 0.001; two-tailed t-test.

We found significant induction of 16 of 25 SAM-dependent methyltransferases as a consequence of SAM depletion (Fig. 4A). To acquire direct evidence of inhibition of methylation, we assessed the global DNA methylome by single-molecule real-time (SMRT) sequencing using the Pacific Biosciences (PacBio) RSII platform. SMRT-sequencing records the replication of individual DNA molecules and can detect methylation through a characteristic slowdown of the polymerase at methylated sites (22). The results supported previous data showing that only a single DNA methylase is active in H37Rv [Rv3263, which methylates the adenines in CTGGAG/CACCAG to form N6-methyladenine (23)]. The interpulse duration (IPD) ratio (Methods) represents the proportion of molecules in a population that are methylated at each site (24), and the mean IPD ratio summarizes this value across the genome. We measured the mean IPD ratio of triplicate samples of the H37Rv control, the ∆metA mutant grown in the presence of methionine, and the mutant after 5 d of methionine starvation. The mean IPD ratio of the H37Rv control was 7.53 (SD 0.07), and that of the ∆metA mutant in the presence of methionine was slightly higher at 8.21 (SD 0.10) (P = 0.0007, two-tailed t-test). The mean IPD ratio of the mutant after 5 d of starvation was 6.03 (SD 0.16; P < 0.0001, two-tailed t-test) (see also Fig. S4). The differences between the mean IPD ratios indicate that, for any particular motif, only 70% of the DNA molecules are methylated after starvation. This reduction in methylation seems to occur evenly across all motifs in the genome, because 99.88% of CTGGAG sites are methylated in the mutant in the presence of methionine, and 99.73% of CTGGAG sites are methylated after methionine starvation (P = 0.152, two-tailed t-test). Because SAM is completely depleted in Mtb ∆metA after 4 d (Fig. 3A), we estimated a maximal 50% reduction in DNA methylation after 5 d of starvation (assuming a 24-h doubling time), and this estimate compares well with the measured 30% decrease in methylation across the whole genome (Fig. S4). This reduction in methylation was consistent with our hypothesis of a global shutdown of methyltransferase reactions in the Mtb ∆metA strain. Interrogation of the transcriptome data revealed several methyltransferase candidates that potentially contribute to the rapid-killing phenotype.

Fig. 4.

Time-course transcriptomic profile of the ∆metA strain during methionine starvation. Samples from three biologically independent replicates were harvested on days 0, 2, 4, and 6, and RNA was extracted. (A) Genes involved in respiration are mainly down-regulated during methionine starvation of ∆metA. As is consistent with SAM depletion (Fig. 3A), the cells up-regulate SAM-dependent methyltransferases, including the essential genes bioA, bioB, and menH. (B) Depletion of SAM (Fig. 3A) leads to the induction of metC, whose expression is controlled by a SAM-IV–dependent riboswitch. Consequently, metC expression increases 30-fold. (C) Genes associated with antioxidant defense during antibiotic treatment (28) are strongly induced. (D) As is consistent with the depletion of methionine (Fig. 3A), we observed a gradual increase in the expression of genes involved in translation initiation. Error bars represent the SD of three biologically independent replicates. *P < 0.05; **P < 0.01; ***P < 0.001; two-tailed t-test. P values for heat maps can be found in the GEO database (accession no. GSE67843).

Among the up-regulated SAM-methyltransferases were essential genes such as menH (Rv0558), bioA (Rv1568), and bioB (Rv1589) (Fig. 4A). MenH catalyzes the methylation of demethylmenaquinol to menaquinol, the major electron carrier in Mtb’s electron transport chain (25). We detected a significant down-regulation of genes involved in respiration (Fig. 4A), as was consistent with the implicated inhibition of menaquinone biosynthesis and the observed growth arrest. Inactivation of BioA in Mtb leads to a bactericidal phenotype in mice and in vitro upon biotin starvation (26). To test if inhibition of biotin biosynthesis contributes to the rapid killing of ∆metA, we starved the mutant in a biotin-free minimal medium and tested the effect of biotin addition on viability (Fig. S6). Biotin supplementation (1 µg/mL) slowed the killing of ∆metA, as demonstrated by a fivefold difference (P = 0.0002, two-tailed t-test) in viable counts after 13 d of starvation. A ∆bioA strain displayed even slower killing kinetics (a 1.5-log difference in cfu after 13 d of starvation, as compared with ∆metA). These data show that Mtb ∆metA is killed by the inhibition of multiple targets, one of which is biotin biosynthesis.

SAM depletion also led to derepression of the metC-metA-Rv3342 operon that is controlled by a recently identified SAM-IV riboswitch (27), and was paralleled by significant induction of the SAM-producing enzyme methionine adenosyltransferase (metK) (Fig. 4B). This derepression of the metA operon indicates that MetA_TB is regulated mainly at the transcriptional level, because we did not observe feedback inhibition of the purified enzyme in the presence of SAM (12.8 mM) or methionine (3.2 mM).

The induction of six adenylate cyclases (Fig. S5) corroborated the accumulation of cAMP, and the increased expression of anaplerotic genes icl, prpD, and prpC (Fig. 4C) is in line with activation of the glyoxylate shunt and methylcitrate cycle. This pattern suggests the recently described antioxidant defense role of anaplerosis in Mtb (28) and is consistent with the increased expression of the oxidative stress response genes furA, ahpC, ahpD, trxB, trxB2, and whiB6 (Fig. 4C).

In line with methionine’s role as the leader amino acid in the majority of proteins, we found strong induction of the translation initiation machinery upon methionine depletion. Specifically, four genes were induced up to 12 fold (Fig. 4D): Methionyl-tRNA formyltransferase (fmt), rRNA cytosine-C5-methyltransferase (fmu), initiation factor IF2 (infB), and ribosomal protein L10 (rplJ), all of which are involved in the preparation and stabilization of the 30S preinitiation complex before recognition of the start codon (29–31). Consistent with an increased need for translation initiation factors, we observed gradual induction of methionyl-tRNA synthetase (metS) (Fig. 4A). Collectively, these data strongly support the notion that inactivation of methionine biosynthesis in Mtb leads to the inhibition of translation.

Discussion

Mtb has evolved with humans for several thousand years and as a consequence has found niches to stay silent within the host until recrudescence arises and transmission ensues. Therefore understanding the physiology, nutrient availability, and nutrient acquisition in the host is of paramount importance to identify Mtb’s weaknesses and potential targets for intervention. Our work shows that Mtb is completely dependent on a functional methionine/SAM biosynthetic pathway for successful infection and survival in host tissues; this finding is in direct contrast to the common belief that Mtb can acquire methionine from the host (32–34). The inability of ∆metA to scavenge sufficient O-acetylhomoserine, homocysteine, methionine, and SAM in vivo argues that these metabolites, although plentiful in serum, are not present in sufficient amounts in Mtb’s preferred host niches or that the bacterium’s transport mechanisms for these metabolites are not activated in vivo. It appears that Mtb seems to follow strategies unlike those used by many other intracellular pathogens (e.g., Legionella pneumophila, Francisella tularensis, or Chlamydia trachomatis) (4–6) that exploit host amino acid pools for proliferation. Subverting nutritional restrictions by the host (9, 35) via an autarkic metabolic lifestyle is likely an evolutionary advantage that contributes greatly to Mtb’s unmatched persistence in humans. It turns out that this important feature of Mtb’s metabolism also is a substantial vulnerability, because several auxotrophic strains of Mtb show significant levels of attenuation in mice (8, 16, 17, 26, 36–42). However, most of these auxotrophies are less bactericidal than the one presented here, and all these studies fall short of providing a biochemical understanding of the bacteriostatic or bactericidal mechanisms in these mutants. Given our results presented here, a systematic investigation of the physiochemical consequences of such auxotrophies could prove extremely valuable for the development of new antimycobacterial strategies.

Our approach revealed an unprecedented and remarkably potent killing mechanism for Mtb directed at multiple vital targets. Depletion of methionine and SAM leads to a pervasive metabolic shutdown caused by the inhibition of essential methyltransferase-dependent pathways such as biotin biosynthesis and possibly through stalling of translation initiation. Also biosynthesis of mycolic acid, an essential component of the mycobacterial cell wall, or menaquinone, the electron carrier in the electron transport chain, have an absolute requirement for SAM (25, 43). Many of these processes are validated drug targets in Mtb (25, 26, 44–46). The possibility that the upstream accumulation of a toxic metabolite might contribute to the killing cannot be ruled out, but so far such a compound could not be identified. Characterization of the role of other enzymes (e.g., MetK) in the SAM biosynthesis pathway will help to dissect this mechanism further. Some aspects of the metabolic trauma described here are similar to the oxidative stress elicited by many front-line TB drugs (28). The induction of glyoxylate and methylcitrate pathways and of oxidative stress response genes is consistent with a recently proposed role for anaplerosis in antioxidant defense in Mtb (28). In fact, our results suggest that this response is not confined to antibiotics but is a general adaptation in dying Mtb cells. Targeting methionine/SAM biosynthesis by a drug could elicit the same rapid cell death as observed in our experiments. Failure of ΔmetA to acquire suppressor mutation is a highly desirable feature for drug and vaccine development. In support of the utility of methionine/SAM pathway-targeting agents in the treatment of tuberculosis, homoserine transacetylase was shown to be a potential drug target in the fungal pathogen Cryptococcus neoformans (47), and a recent report on inhibitors of folate metabolism suggests that interfering with SAM biosynthesis contributes to their efficacy (44).

Taken together, our results show that methionine and SAM depletion in Mtb leads to complete loss of viability in vivo and to an inability to proliferate in the host’s macrophages. Blockade of methionine/SAM biosynthesis results in disruption of numerous functions required for life and to rapid death, indicating that this pathway provides excellent targets for the development of effective antituberculosis chemotherapy.

Methods

Please see SI Methods for the suppressor mutant screen and details of RNA extraction.

Protein Purification and Enzyme Assays.

His-tagged MetA_TB was cloned using plasmid pET28a, expressed in E. coli BL21(DE3) and purified with Ni-NTA agarose slurry and size-exclusion chromatography (Fig. S1). Enzyme kinetics were determined by monitoring the formation of CoA-SH by Ellman's reagent (DTNB). More details can be found in SI Methods.

Mycobacterial Strains and Growth Conditions.

All bacterial strains, plasmids, and primers used in this study are listed in Tables S1–S3. Mycobacterial strains were grown in Middlebrook 7H9 medium (Difco) supplemented with 10% (vol/vol) OADC enrichment (0.5 g oleic acid, 50 g albumin, 20 g dextrose, 0.04 g catalase, 8.5 g sodium chloride in 1 L water), 0.2% (vol/vol) glycerol, and 0.05% (vol/vol) tylaxopol (Sigma). Selective medium contained 75 µg/mL hygromycin B and/or 20 µg/mL kanamycin. For the time-course starvation experiments, WT, Mtb ∆metA, and Mtb ∆metAcomp strains were grown in the presence of 50 µg/mL methionine to an OD₆₀₀ of 0.5 and then were switched to methionine-free medium after three washings with PBS tylaxopol. Samples from three biological replicates were harvested on days 0, 2, 4, and 6 and were processed for metabolite and RNA extraction (see below).

Gene Knockout and Complementation.

The gene metA (Rv3341) was deleted in Mtb H37Rv and M. bovis bacillus Calmette–Guérin by specialized transduction as described previously (48). Transductants were recovered on selective medium containing hygromycin and methionine (50 µg/mL). Mutations were confirmed by three-primer PCR using primers Rv3341L, Rv3341R, and Universal_uptag, listed in Table S3. Mutants were unmarked using phage phAE280 as previously described (48). The Mtb ΔmetA strain was complemented using pMV306 harboring a copy of the operon Rv3340–Rv3342 including 300 bp of the upstream region of Rv3340 to include the native promoter (Table S1). This DNA fragment was PCR-amplified using primers Rv3341_fw_EcoRI and Rv3341_re_HindIII (Table S1) and was cloned into pMV306 using EcoRI and HindIII restriction sites, resulting in plasmid pYUB1939. This construct was electroporated into Mtb and M. bovis bacillus Calmette–Guérin. The nucleotide sequences of all constructs were verified by Sanger sequencing. Gene deletions of trpD, panCD, leuCD, and bioA were constructed and confirmed by the same method. Details about strains, primers, and plasmids used can be found in Tables S1–S3.

Metabolite Extraction.

Samples of triplicate bacterial cultures were harvested at given time points. An equivalent of 5 mL culture at an OD₆₀₀ of 0.5 was quenched rapidly in 10 mL 100% methanol (MeOH) at −20 °C and centrifuged at 3300 × g (Sorvall Legend RT) for 10 min at −9 °C. Cell pellets were resuspended in extraction solvent containing 40% (vol/vol) acetonitrile, 40% (vol/vol) MeOH, and 20% (vol/vol) H₂O, were transferred to 2-mL screw-cap tubes containing silica beads, and then were agitated twice in a FastPrep-24 (MP Biomedicals) for 45 s at a speed of 6 m/s, with 5 min on ice between beatings. Samples were spun briefly, and 750 µL of extract was filtered through 0.22-µm Spin-X centrifuge tube filters (Corning, Life Technologies) and then frozen at −80 °C until time of analysis.

Metabolomics.

Metabolomics analysis was performed using an Acquity UPLC system (Waters) coupled with a Synapt G2 quadrupole–time of flight hybrid mass spectrometer. More details can be found in SI Methods.

Microarray.

Microarray analysis was performed as described previously (49) using arrays purchased from Microarrays Inc., Gene Expression Omnibus (GEO) platform GPL19545, following protocols SOP M007 and M008 from the Institute of Genomic Research (TIGR) (50). All data have been deposited in the GEO database (accession no. GSE67843). More details can be found in SI Methods.

Mouse Experiments.

Mouse studies were performed in accordance with National Institutes of Health guidelines following the recommendations in the Guide for the Care and Use of Laboratory Animals (51). The protocols used in this study were approved by the Institutional Animal Care and Use Committee of Albert Einstein College of Medicine (Protocol #20120114).

Female SCID mice and female C57BL/6 mice (Jackson Laboratories) were infected via the aerosol route using a 1 × 10⁷ cfu/mL mycobacterial suspension in PBS containing 0.05% tylaxopol and 0.004% antifoam. Infection yielded ∼100 bacilli per lung as determined by quantification of lung bacterial loads at 24 h postinfection (four mice per group). Subsequently, four mice from each group were killed at days 1, 7, 21, and 56 (C57BL/6) or 1, 7, 21, and 42 (SCID) to determine the bacterial burden in the lung and spleen. Six mice per group were kept for survival experiments. High-dose i.v. infection of SCID mice was conducted via tail vein injection of 200 µL PBS containing 0.05% tylaxopol and ∼5 × 10⁶ cfu/mL mycobacteria, yielding an initial load of 1 × 10⁶ bacteria per mouse. All mice infected with Mtb were maintained under appropriate conditions in an animal biosafety level 3 laboratory. All experiments were repeated to confirm reproducibility. Pathological analysis and histological staining of organ sections were done on tissues fixed in buffered 10% (vol/vol) formalin.

Pathology.

Lung and spleen samples were fixed in 10% (vol/vol) neutral buffered formalin for 48–72 h and then were subjected to paraffin embedment. Tissues were sectioned at 5 µm; one section was stained with H&E, and a second section was stained for AFB using the Kinyoun method. Samples were evaluated histologically by a board-certified veterinary pathologist. Lesion severity was graded using a scale of 0–5: a score of 0 = no findings; 1 = minimal findings; 2 = mild findings; 3 = moderate findings; 4 = marked findings; and 5 = severe findings. See also Table S4.

Infection of Primary Human Macrophages.

Peripheral blood monocytes were isolated from human peripheral blood (New York Blood Center) by density centrifugation on Ficoll-PaquePLUS (GE Healthcare), followed by positive selection using CD14 magnetic beads (Miltenyi Biotec). Infection experiments with primary human macrophages were conducted according to published protocols (52). For the methionine titration experiment the macrophages were maintained in the standard RPMI 1640 macrophage medium containing 15 μg/mL methionine; for the final methionine concentrations of 50 μg/mL, additional methionine was added at the time of infection. For a methionine-free medium we used R7513 (Sigma) and added back l-glutamine and cystine. For the final concentration of 3 μg/mL methionine, we added methionine to the methionine-free medium R7513 (Sigma) at the time of infection after washing the cells. Mycobacterial growth was assessed on 7H10, 10% (vol/vol) OADC, 0.2% glycerol plates with methionine supplementation, after lysis of the macrophage monolayer with 300 uL 0.5% Triton X-100 per well. Results were expressed as the means and SEs of triplicate samples from one well.

PacBio Sequencing.

DNA was prepared and sequenced on the Pacific Biosciences RSII machine according to standard manufacturer’s conditions. More details can be found in SI Methods and Table S5.