The Role of Fatty Acid Metabolism in Drug Tolerance of Mycobacterium tuberculosis

Multiple lines of evidence have shown that PEPCK is essential for in vitro growth in medium containing fatty acids but is dispensable for growth when carbohydrates are supplied as carbon sources (23). To confirm these findings, we characterized the in vitro growth phenotypes of the M. tuberculosis wild-type (WT), a ΔpckA mutant, and a pckA complemented (COM) strain using Middlebrook 7H9 liquid medium (m7H9) containing either glycerol or a model fatty acid, acetate, as the sole carbon source. The ΔpckA mutant failed to increase the biomass when cultured in m7H9 containing acetate (acetate m7H9), but the growth kinetics were almost identical to those of the WT or COM strain in m7H9 containing glycerol (glycerol m7H9) (see Fig. S1A and B in the supplemental material) (23, 25). Monitoring of viable CFU showed that the ΔpckA mutant underwent neither net replication nor death in acetate m7H9 (Fig. S1C) (26). The loss of replicating capacity of M. tuberculosis without discernible death has been phenotypically associated with an increased level of drug tolerance. We therefore assessed the effect of acetate-induced ΔpckA strain nonreplication on drug tolerance by monitoring the optical density (OD) or CFU in the presence of d-cycloserine (DCS), a cell-lysing antibiotic known to generate a drug-tolerant subpopulation, termed persisters (27). As expected, we observed that M. tuberculosis with a pckA deficiency exhibited an increased persister-forming rate after treatment with a bactericidal dose of DCS in acetate m7H9 compared to that in glycerol m7H9 (Fig. 1A; Fig. S1D and E). When cultured in glycerol m7H9, the ΔpckA mutant was initially less susceptible to the DCS effect (at 1 day) compared to that of the WT or COM strain. However, the ΔpckA strain gradually succumbed to the effect of DCS and ultimately, the level of sensitivity to DCS became similar to that of the other two strains (at days 7 and 14) (Fig. 1A, upper panel). In contrast, ΔpckA strain growth in acetate m7H9 monitored by OD initially revealed a decrease (day 1). However, growth remained static after that, especially during the phase between days 7 and 14, the period known to form the drug-tolerant persisters in response to treatment with DCS (Fig. 1A, lower panel) (27).

To identify the metabolic shifts behind the acetate-induced dormancy-like state and drug tolerance of the ΔpckA mutant, we used a filter culture-based metabolomics profiling method (11). A semi-untargeted metabolomics analysis of the WT and ΔpckA strains was conducted to identify the metabolites and pathways that were uniquely altered in the ΔpckA mutant when cultured in acetate medium (see Fig. S2A to C in the supplemental material). A clustered heat map and principal-component analysis (PCA) indicated that the acetate-induced metabolic alterations of the ΔpckA mutant were different from those of the WT in acetate medium or the ΔpckA mutant in glycerol medium (Fig. S2A and B). Of 231 metabolites detected, the volcano plot and pathway enrichment analyses pinpointed that upregulated metabolites included the intermediates of the reductive branch of the tricarboxylic acid (TCA) cycle, such as aspartate (a surrogate of OAA), malate, fumarate, and succinate, as well as those of the MCC, such as 2-methylcitrate (2MC) and 2-methylisocitrate (2MI) (collectively 2MIC) (Fig. 1B; Fig S2C and D). The metabolomics and CFU analysis of the WT and ΔpckA strains suggested that the defective acetate catabolism of the ΔpckA mutant led to a dormancy-like, drug-tolerant state, and this may be associated with the formation of persisters. Targeted metabolomics focusing on the intermediates in the central carbon metabolism showed that pckA deficiency caused an overaccumulation of intermediates in both the TCA cycle and MCC (Fig. 1B; Fig. S2D). The accumulation level of the reductive TCA cycle branch intermediates of the ΔpckA mutant was by far greater than those of oxidative TCA cycle branch intermediates α-ketoglutarate (α-KG) and citrate: ∼21-fold for aspartate, 17-fold for succinate, 58-fold for fumarate, 69-fold for malate, and 10-fold for the MCC intermediates (2-methylcitrate and 2-methylisocitrate) (Fig. 1B; Fig. S2D).

To examine whether the accumulation of reductive TCA cycle branch intermediates of the ΔpckA mutant in acetate m7H9 was attributable to the catalytic slowdown arising from the PEPCK deficiency-mediated obstruction of gluconeogenic carbon flux, we first monitored the mRNA levels of genes encoding the TCA cycle and MCC enzymes. The quantitative reverse transcription-PCR (qRT-PCR) results indicated that mRNA levels of ΔpckA TCA cycle and MCC genes were globally downregulated in acetate m7H9 compared to those in glycerol m7H9. In comparison, the expression levels of TCA cycle genes in the WT were almost identical between the two carbon conditions (Fig. 1C, left panel). The genes examined included those encoding MCC enzymes (prpC, prpD, acn, and icl1), those encoding the reductive TCA cycle branch enzymes (mdh, mqo, fum, sucC and -D, and sdhA to -D), those encoding oxidative TCA cycle branch enzymes (citA, acn, icd1, and icd2), and those encoding ATP synthases (atpB and atpF) (Fig. 1C, right panel). A downregulated expression of prpC, acn, and prpD, the genes encoding the first three steps of MCC with minor downregulation of icl1 expression, may explain the overaccumulation of the MCC intermediates. The targeted metabolomics profile of the ΔpckA mutant in acetate m7H9 explained that overaccumulation of the TCA cycle and MCC intermediates could be attributed to the obstructed gluconeogenic carbon flux (Fig. 1B). The acetate-induced slowdown of the ΔpckA mutant TCA cycle activity was further examined by the ¹³C labeling patterns of TCA cycle intermediates. We monitored TCA cycle intermediates by transferring the WT, ΔpckA, or COM strain to Middlebrook 7H10 agar medium (m7H10) containing uniformly ¹³C-labeled acetate ([U-¹³C]acetate), and the metabolome was sampled after a 24-h incubation. Normal TCA cycle activity results in the progressive assimilation of acetate-based C₂ units, manifest by the accumulation of higher-order ¹³C₂-based isotopologues when M. tuberculosis was cultured in [U-¹³C]acetate m7H10. The ¹³C labeling patterns showed the downshift of ¹³C-labeled α-KG, malate, and succinate isotopologues of the ΔpckA mutant from M + 4 (+4 ¹³C labeled) to M + 2 (+2 ¹³C labeled), revealing acetate-induced slowdown of TCA cycle activity in the ΔpckA mutant compared to the activity of the WT. Intriguingly, acetate-induced slowdown of TCA cycle activity in the COM strain did not fully restore to WT levels (see Fig. S3 in the supplemental material). The lack of a complete restoration in the COM strain suggests that the complementation was not fully analogous to the WT metabolic state, despite the use of the native promoter to drive the complementary pckA gene. ¹³C isotopologue analysis and qRT-PCR results confirmed that the acetate-induced metabolic remodeling of the ΔpckA mutant primarily arose from a catalytic slowdown, rather than an induction of the TCA cycle and MCC activities. We thought that the acetate-induced slowdown of ΔpckA TCA cycle activity would be associated with a decreased NADH/NAD ratio as the TCA cycle serves as the main NADH source. Unexpectedly, we detected an increased NADH/NAD ratio in the ΔpckA mutant compared to that of the WT under acetate growth conditions (see Fig. S4A and B in the supplemental material). This may be attributable to downregulation of the electron transport chain (ETC), an activity required to recycle NAD from NADH. Indeed, we found that the NAD level in the ΔpckA mutant was significantly lower than that of the WT, compared to the depletion rates of NADH, when acetate was provided as the single carbon source (Fig. S4C and D).

Given the carbon source-dependent metabolic remodeling and phenotypic changes of the ΔpckA mutant, we explored the drug tolerance of the ΔpckA mutant against the clinically relevant anti-TB drugs INH and BDQ. The WT, ΔpckA, or COM strain was treated with 10× the MIC of both INH and BDQ for 5 or 7 days, respectively, in glycerol or acetate media. The cells were serially diluted and plated on m7H10, and the colonies were counted after 3 weeks of incubation at 37°C. The ΔpckA mutant was able to maintain a greater level of viable colonies after treatment with first- and second-line TB drugs in both carbon sources, albeit the level was much greater in acetate medium (Fig. 2A to D). The percentage of survival rates of the ΔpckA mutant in acetate medium were significantly greater by ∼13-fold (0.26% versus 0.02%) under the INH treatment condition or ∼487-fold (4.4% versus 0.0009%) under the BDQ treatment condition compared to those of the WT (Fig. 2B and D). These findings demonstrated that the acetate-induced bacteriostatic phenotypes and accompanying metabolic remodeling of the ΔpckA mutant led to a greater level of drug tolerance against conventional TB drugs (Fig. 2B and D; see Fig. S5A and B in the supplemental material). These findings also confirmed that M. tuberculosis cells in a state of viable but reduced growth rate are normally less susceptible to anti-TB drugs. Intriguingly, WT cells, despite an intact PEPCK activity, became significantly more tolerant to DCS and INH when cultured in acetate medium compared to glycerol medium (Fig. 1A; Fig. S5A to C).

We conducted a metabolomics analysis to investigate the role of fatty acid-induced metabolic changes of the ΔpckA mutant in drug tolerance. To this end, we generated M. tuberculosis-laden filters containing WT, ΔpckA mutant, or COM strain cultures in the mid-logarithmic phase and exposed them to 10× the MIC of INH or BDQ. After incubation for 24 h, the cells were harvested before the loss of viability, and the metabolome was sampled by mechanical lysis to allow targeted LC-MS metabolomics analysis focusing on intermediates in the TCA cycle and MCC (28). This analysis showed that intermediates accumulated in the reductive branch of the TCA cycle, such as malate and aspartate, and the MCC, such as 2-methylcitrate and 2-methylisocitrate, of the ΔpckA mutant in both glycerol and acetate media were maintained even after treatment with INH or BDQ (Fig. 3A and B). Similar patterns were also observed after treatment with DCS (see Fig. S6 in the supplemental material). Intriguingly, BDQ treatment further enhanced the abundance of the TCA cycle and MCC intermediates of the ΔpckA mutant, while only minor changes or slight downregulation was detected in response to treatment with INH or DCS, which was corroborated by the finding that the ΔpckA mutant was tolerant to BDQ much greater than to INH (Fig. 2B and D). Unlike the reductive TCA cycle branch or MCC intermediates, the oxidative TCA cycle branch intermediates of the ΔpckA mutant, such as citrate and α-KG, were unaltered or even further downregulated after antibiotic treatment relative to those of the ΔpckA mutant without treatment (Fig. 3A and B; Fig. S6). Bypassing the oxidative branch of the TCA cycle is known to be associated with enhanced glyoxylate shunt activity—an adaptive strategy of M. tuberculosis used to survive hypoxia or antibiotic effects (11, 15). These results showed that the metabolic profile of the acetate-induced ΔpckA nonreplicating phenotype (Fig. 1B; Fig. S2) was also maintained even when exposed to first- and second-line TB drugs (Fig. 3; Fig. S6), resulting in the high level of drug tolerance.

A previous study has shown that the accumulation of MCC intermediates causes dysregulated respiratory activity and subsequent destabilized membrane bioenergetics (29, 30). A separate study confirmed that the enhanced respiration rendered M. tuberculosis in a dormancy-like state to be metabolically active and hypersensitive to INH via induced oxidative damage (31). Thus, to determine if the remodeled membrane bioenergetics results in ΔpckA mutant tolerance against INH, we measured the sensitivity to reactive oxygen species (ROS), membrane potential, and ATP levels of all three strains in acetate m7H9 upon exposure to INH. ROS, membrane potential, and ATP levels were quantified by flow cytometry using specific fluorescence dyes before and after treatment with 10× the MIC of INH (29). BacLight DiOC₂(3) membrane potential dye was used to detect membrane potential homeostasis. This dye accumulates inside bacteria and emits green fluorescence in direct proportion to size. Under active oxidative phosphorylation, this dye also emits red fluorescence. The red/green fluorescence ratio can be used to track changes in bacterial membrane potential. H₂O₂ and CCCP (carbonyl-cyanide 3-chlorophenylhydrazone) were used as controls for ROS and membrane potential, respectively. The results showed that the ΔpckA mutant produced less ATP, even under no antibiotic treatment (Fig. 4A), while maintaining ROS and membrane potential at levels similar to those of the WT or COM strain under acetate medium (Fig. 4B and C). Interestingly, the ΔpckA mutant already had significantly increased intrabacterial ROS content when cultured in acetate medium compared to when cultured in glycerol medium. The fold increase shown in ΔpckA strain ROS levels in acetate medium was significantly greater than that in glycerol medium demonstrated by the WT when comparing ROS levels in acetate to those in glycerol (Fig. 4D). Intriguingly, ATP levels of the ΔpckA strain were not as negatively impacted by INH treatment as those of the WT, and the ROS levels of the ΔpckA strain were also relatively unaffected by INH treatment (Fig. 4A and B). This may be largely due to the slowed TCA cycle activity and low levels of ETC activity. The membrane potential of the ΔpckA strain was also relatively unaltered by treatment with INH compared to that of WT (Fig. 4C), presumably due to the presence of an additional strategy to sustain the membrane potential, including the active secretion of succinate, as previously reported (11, 32). Notably, a greater amount of succinate was secreted by ΔpckA cells cultured in acetate m7H9 compared to that of either WT cells in acetate m7H9 or ΔpckA cells in glycerol m7H9 (see Fig. S7A in the supplemental material). Consistent with previous reports, the MCC intermediates, if accumulated, affect the respiratory activity and membrane bioenergetics, leading to the growth arrest and drug tolerance (29).

BDQ inhibits the c and ɛ subunits of M. tuberculosis ATP synthase (33–36). Therefore, we were interested in understanding if ATP levels of the ΔpckA mutant were affected in acetate medium when exposed to this drug. In the absence of BDQ treatment in acetate m7H9, the ATP abundance of the ΔpckA mutant was maintained at levels significantly lower than those of the other two strains (Fig. 4A and Fig. 5), which correlates with the dormancy-like state, as previously identified (27, 37). Intriguingly, when exposed to BDQ, the ATP levels were not significantly altered at least for 2 days, while the ATP levels of WT and COM rapidly decreased (Fig. 5). This rapid decrease in ATP levels of the WT and COM strains, not of the ΔpckA mutant, suggested that the ΔpckA mutant largely relies on metabolic sources other than ETC activity for ATP biosynthesis. Thus, the ΔpckA mutant in the acetate medium could maintain its ATP levels despite the treatment with high doses of BDQ. Collectively, the findings suggested that the accumulation of MCC intermediates downregulates the ETC activity, while other metabolic activities may be involved in maintaining the ATP pool, such as the activation of trehalose metabolism that feeds toward glycolysis (12, 13, 38) or possibly slowed kinetics of the ATP consumption rate. To interrogate the role of preexisting carbon sources to feed glycolysis, we proceeded to calculate ¹³C enrichment rates of ΔpckA mutant PEP by transferring cells from medium containing unlabeled acetate to medium containing [U-¹³C]acetate, as was done to study ¹³C labeling patterns of TCA cycle intermediates (Fig. S3). We observed that the ΔpckA mutant biosynthesized PEP, the end product of M. tuberculosis glycolysis, at levels similar to those of the WT or COM strain, and a significant portion of the PEP biosynthesis occurred within the unlabeled fraction (Fig. S7B and C). Biosynthesis of PEP in the WT or COM strain, however, occurred within the ¹³C-labeled fraction, implying that the WT or COM strain efficiently consumed [U-¹³C]acetate, but the ΔpckA mutant used unknown internal sources (Fig. S7B). These findings collectively suggested the existence of endogenous carbon sources and the role of substrate-level phosphorylation of M. tuberculosis glycolysis activity in maintaining ATP levels of the ΔpckA mutant upon treatment with BDQ.

To determine if the MCC intermediate accumulation may indeed be associated with drug tolerance, we used two mutants lacking enzymes in the MCC: an isocitrate lyase knockdown (ICL KD) strain and a 2-methylcitrate dehydratase knockout (ΔprpD) strain. ICL catalyzes the last step (2-methylisocitrate to succinate and pyruvate) of the MCC (Fig. 1C, right panel) (29, 39). The ICL KD strain is a conditional knockdown mutant generated in the Schnappinger laboratory by adapting the MultiSite Gateway recombinational cloning method to assemble a regulated expression plasmid adapted to expression in M. tuberculosis (40). The ICL KD strain is a TetON mutant, and its expression is induced by treatment with anhydrotetracycline (ATc). The 2-methylcitrate dehydratase (PrpD) catalytically converts 2-methylcitrate to 2-methyl cis-aconitate. Previously, ICL-deficient M. tuberculosis had been shown to have a growth defect in fatty acid medium, and thus, ICL was considered to be one of the best drug targets (24). As propionate is an initial substrate of the M. tuberculosis MCC, we assessed the growth kinetics of the ICL KD and ΔprpD strains together with their parental strains (Erdman and CDC1551, respectively) in glycerol and propionate media as the sole carbon sources. The ICL KD strain’s growth kinetics were almost identical to those of the WT when cultured in glycerol m7H9 due to the functional dispensability of ICL activity in consuming glycerol as a carbon source (Fig. 6A). In propionate m7H9, ICL KD strain growth was completely impaired in the absence of ATc but fully restored at a level similar to that of the WT by treatment with ATc (Fig. 6B; see Fig. S8A in the supplemental material). However, rather intriguingly, the ΔprpD mutant grew normally in both glycerol m7H9 and propionate m7H9 (Fig. 6A and B), which was somewhat different from the results reported by Muñoz-Elias et al., although the strain used in their study was ΔprpDC (41). These findings suggested that the buildup of either 2-methyl cis-aconitate or 2-methylisocitrate was toxic to M. tuberculosis growth. We then exposed these four strains (WT and mutants) to INH in either glycerol m7H9 or propionate m7H9 and monitored the viable colonies after a 5-day incubation. Surprisingly, the ICL KD strain had a better survival rate than the other strains, which were restored by treatment with ATc (Fig. 6C; Fig. S8B). Similar to the ΔpckA mutant, only the ICL KD strain showed a significant accumulation of MCC intermediates in both glycerol and propionate media, albeit much greater in propionate medium (Fig. 6D; Fig. S8C). Interestingly, unlike the metabolomics profile of the ΔpckA mutant, the ICL KD mutant did not accumulate TCA cycle intermediates, except for succinate (Fig. S8C); thus, the tolerance is likely to be mediated by the accumulation of MCC intermediates. To confirm the role of MCC intermediates in M. tuberculosis drug tolerance, we chemically synthesized 2-methylisocitrate as conducted in our previous report (30), treated the Erdman WT strain with various doses of authentic 2-methylisocitrate (1 to 10 mM), and monitored the drug sensitivity against INH or BDQ. The percentage of survival rates indicated that 5 mM 2-methylisocitrate was sufficient to partly protect M. tuberculosis from the antibiotic effects of both INH and BDQ by ∼2- to 3-fold compared to the rate without treatment with 2-methylisocitrate (Fig. 6E). Concentrations of 2-methylisocitrate at greater than 10 mM might be toxic to M. tuberculosis viability (Fig. 6E). These findings are also corroborated by the percentage of survival rate of the ICL KD strain without ATc against INH (Fig. 6C and D; Fig. S8B).

The Mycobacterium tuberculosis Erdman wild-type strain (WT), Erdman ΔpckA mutant, Erdman pckA complemented (COM) strain, Erdman icl knockdown (ICL KD) strain, and CDC1551 and CDC1551 ΔprpD strains were precultured in Middlebrook 7H9 (m7H9) broth (Difco, Detroit, MI) supplemented with 0.5% (wt/vol) fraction V bovine serum albumin (BSA), 0.085% (wt/vol) NaCl, and 0.04% (vol/vol) tyloxapol with 0.2% (vol/vol) glycerol and 0.2% (wt/vol) dextrose. For experiments, the strains were resuspended in fresh m7H9 containing BSA, NaCl, tyloxapol, and 0.2% (vol/vol) glycerol, 0.2% (wt/vol) acetate, or 0.05 to 0.1% (wt/vol) propionate. The following antibiotics were added when necessary: isoniazid and bedaquiline at a final concentration of 0.3 μg/mL (10×) or 3 μg/mL (100×) and d-cycloserine at a final concentration of 100 μg/mL. For metabolomic profiling, filters were generated as previously described (8, 11). The M. tuberculosis Erdman and CDC1551 strains were cultured under containment in a biosafety level 3 facility. The ΔprpD mutant was purchased from BEI Resources.

Bacterial growth was monitored by optical density at 595 nm (OD₅₉₅) by using a Genesys 20 spectrophotometer (Thermo Scientific). For CFU assays, cells in the mid-logarithmic growth phase of M. tuberculosis Erdman or CDC1551 were diluted to an OD₅₉₅ of 0.05 in m7H9 broth containing BSA, NaCl, and tyloxapol and supplemented with 0.2% (vol/vol) glycerol, 0.2% (wt/vol) acetate, or 0.05 to 0.1% (wt/vol) propionate and antibiotics (isoniazid, bedaquiline, or d-cycloserine) when applicable, in a 24- or 96-well plate. After 5 or 7 days of antibiotic treatment (isoniazid and bedaquiline, respectively, or 7 and 14 days for d-cycloserine), the cells were then serially diluted and plated on m7H10 agar with 0.2% (vol/vol) glycerol, 0.2% (wt/vol) dextrose, 0.5 g/L BSA, and 0.085% (wt/vol) NaCl for 3 weeks at 37°C until the colonies were formed and counted. The ICL KD strain was plated on m7H9 containing 0.5% (vol/vol) glycerol, 0.2% (wt/vol) dextrose, 0.5 g/L BSA, 0.085% (wt/vol) NaCl, and 20 g/L agar due to its inability to grow on m7H10 medium containing malachite green.

The filters containing the Erdman or CDC1551 strains were incubated at 37°C. After reaching the mid-logarithmic phase of growth, the filters were transferred to chemically identical m7H10 agar containing fresh carbon source(s) and antibiotics when applicable and incubated for 24 h at 37°C. The metabolites were harvested by transferring the filters into precooled −40°C LC-MS-grade acetonitrile-methanol-water (40:40:20) solution and mechanically lysed with 0.1-mm Zirconia beads in a Precellys tissue homogenizer (Bertin Technologies, France) at 6,800 rpm for 6 min in dry ice. The lysate was centrifuged and filtered using 0.22-μm Spin-X columns (Sigma-Aldrich). The protein concentration of metabolite extracts was measured with a bicinchoninic acid (BCA) protein assay kit (Thermo Scientific, Waltham, MA, USA) to normalize samples to cell biomass.

LC-MS differentiation and detection of Erdman, CDC1551, and mutant strains were performed with an Agilent Accurate mass 6230 time of flight (TOF) device coupled with an Agilent 1290 liquid chromatography system using solvents and configuration as previously described (11, 13). An isocratic pump was used for continuous infusion of a reference mass solution to allow mass axis calibration. Detected ions were classified as metabolites based on unique accurate mass-retention time identifiers for masses showing the expected distribution of accompanying isotopologues. Metabolites were analyzed using Agilent Qualitative Analysis B.07.00 and Profinder B.06.00 software (Agilent Technologies, Santa Clara, CA, USA) with a mass tolerance of <0.005 Da. Standards of authentic chemicals of known amounts were mixed with bacterial lysates and analyzed to generate the standard curves used to quantify metabolite levels.

The extent of isotopic labeling for metabolites was determined by dividing the sum of the peak height ion intensities of all labeled isotopologue species by the ion intensity of both labeled and unlabeled species, expressed as a percentage. Label-specific ion counts were corrected for naturally occurring ¹³C species (i.e., [M + 1] and [M + 2]). The relative abundance of each isotopic form was represented by the sum of the peak 8 ion intensity of all labeled species.

The ΔpckA mutant was grown on culture filter membranes as done in metabolite extraction and exposed to glycerol or acetate as the sole carbon source for 24 h. The total RNA was extracted using TRIzol solution (Sigma-Aldrich) and mechanically lysed with 0.1-mm Zirconia beads in a Precellys tissue homogenizer. Lysates were clarified by centrifugation, and the TRIzol supernatant was removed and used for RNA extraction. RNA was isolated using a Qiagen RNA extraction kit. Isolated RNA was treated with DNase I (Sigma-Aldrich) to remove DNA contamination (Sigma-Aldrich). RNA concentrations were determined using a Nanodrop spectrophotometer, and qRT-PCRs were conducted using an iQ SYBR green Supermix (Bio-Rad) and C1000 thermal cycler instrument (Bio-Rad). The primers used for amplification are listed in Table S1 in the supplemental material. Fold changes were calculated by values that were normalized to sigA transcript levels.

Intrabacterial ATP concentrations were measured by BacTiter-Glo microbial cell viability assay (Promega) according to the manufacturer’s instructions. Cells were grown until an OD₅₉₅ of 1.0 and diluted to 0.6 in fresh m7H9 medium containing the appropriate carbon source (glycerol or acetate) and antibiotic (bedaquiline or isoniazid). A 1-mL sample was taken at each time point, harvested, resuspended in 1 mL phosphate-buffered saline (PBS), and heat lysed at 100°C for 1 h. The lysate was centrifuged and filtered using a 0.22-μm Spin-X column, and the ATP was analyzed according to the manufacturer’s instructions (Thermo Fisher Scientific).

Cells were grown until an OD₅₉₅ of 1.0 and diluted to 0.6 in fresh m7H9 containing the appropriate carbon source (glycerol or acetate) and antibiotic (bedaquiline or isoniazid). A 1-mL sample was harvested and resuspended in 1 mL PBS with 0.04% (vol/vol) tyloxapol. For ROS quantification, the Total reactive oxygen species (ROS) assay kit, 520 nm (Thermo Fisher Scientific), was used. The cells were fixed with 2.5% (vol/vol) glutaraldehyde and stained with 10 μM (final) dihydroethidium for 1 h. The flow cytometry was set up according to the manufacturer’s instructions (Thermo Fisher Scientific). Positive (treated with 1 mM H₂O₂) and negative (untreated) controls were included. For membrane potential quantification, the instructions of the BacLight bacterial membrane potential kit (B34950; Molecular Probes) were followed. The cells were stained with 30 μM (final concentration) DiOC2(3) for 5 min, washed with PBS, and fixed with 2.5% (vol/vol) glutaraldehyde. A 5 μM concentration of carbonyl cyanide m-chlorophenyl hydrazine (CCCP) or unstained cells was included as a positive control and negative control, respectively.

Our filter culture system was modified by replacing the underlying m7H10 agar with a plastic inset containing chemically equivalent m7H9 in direct contact with the underside of M. tuberculosis-laden filters, as previously conducted (11). m7H10 and m7H9 contained glycerol or acetate as a single carbon source. A blank filter was used as a negative control. After a 1-day incubation, cell-free m7H9 was collected, and the metabolome was extracted by adding an LC-MS-grade acetonitrile-methanol-H₂O (40:40:20) solution, which was precooled to −40°C, and quantified by LC-MS. Total M. tuberculosis biomass was determined to normalize the succinate ion counts to biomass (BCA protein assay kit; Thermo Scientific).

Statistical analyses were performed by analysis of variance (ANOVA) and unpaired Student's t test. P values of <0.05 were considered statistically significant.