Hydrogen sulfide stimulates Mycobacterium tuberculosis respiration, growth and pathogenesis

Abstract

Hydrogen sulfide (H₂S) is involved in numerous pathophysiological processes and shares overlapping functions with CO and •NO. However, the importance of host-derived H₂S in microbial pathogenesis is unknown. Here we show that Mtb-infected mice deficient in the H₂S-producing enzyme cystathionine β-synthase (CBS) survive longer with reduced organ burden, and that pharmacological inhibition of CBS reduces Mtb bacillary load in mice. High-resolution respirometry, transcriptomics and mass spectrometry establish that H₂S stimulates Mtb respiration and bioenergetics predominantly via cytochrome bd oxidase, and that H₂S reverses •NO-mediated inhibition of Mtb respiration. Further, exposure of Mtb to H₂S regulates genes involved in sulfur and copper metabolism and the Dos regulon. Our results indicate that Mtb exploits host-derived H₂S to promote growth and disease, and suggest that host-directed therapies targeting H₂S production may be potentially useful for the management of tuberculosis and other microbial infections.

Introduction

Mycobacterium tuberculosis (Mtb), the etiological agent of TB, can persist in a clinically latent disease state for decades. Identifying host factors that influence Mtb survival is essential for understanding how Mtb persists. Physiologic gaseous messengers such as carbon monoxide (CO) and nitric oxide (•NO) are important to the outcome of Mtb infection. In particular, •NO produced by inducible nitric oxide synthase (iNOS) exerts a bacteriostatic effect on Mtb by inhibiting respiration¹ and is crucial for protection against TB^2,3. Notably, •NO¹, CO⁴ and lowoxygen (O₂) levels⁵ induce the expression of the 48-gene Dos dormancy regulon, which is associated with reduced respiration and bacterial replication. Hydrogen sulfide (H₂S) is an emerging signaling molecule that is produced enzymatically in mammals by cystathionine β-synthase (CBS), cystathionine γ-lyase (CSE) and 3-mercaptopyruvate sulfurtransferase (3-MST)^6,7. At physiological pH, H₂S exists mainly as a hydrosulfide anion (HS⁻, 75−80%), with the remainder in the form of the dissolved uncharged H₂S gas (~20−25%) and negligible amounts of S^2− 8. Here, unless specified, we refer to the sulfide pool (H₂S + HS⁻ + S²⁻) as H₂S.

H₂S modulates various physiological functions including inflammation, metabolism, neuroprotection, vasodilation⁹, and protection of microbes against antibiotics^10,11. Surprisingly, despite having several overlapping functions with CO and •NO^8,12, a role for host-generated H₂S in bacterial pathogenesis has not yet been described. H₂S can modulate mitochondrial respiration resulting in a suspended animation-like state in mice¹³. In mammals, H₂S elicits a biphasic, concentration-dependent mitochondrial response¹⁴, which can be cytotoxic or cytoprotective. For example, at high concentrations H₂S reversibly inhibits cytochrome c oxidase (Complex IV)^15,16,17. In contrast, at low concentrations H₂S can serve as bioenergetic fuel to stimulate mitochondrial respiration without uncoupling of respiration^16,18. Due to the complex chemistry of H₂S and lack of routine assays for accurate measurement, the physiological concentration of H₂S in vivo has been a matter of debate^19,20,21. Here we hypothesize that host-derived H₂S plays a role in TB through modulation of Mtb respiration and bioenergetics. To test this hypothesis, we performed Mtb infection studies in CBS-deficient (Cbs^+/−) mice, which have reduced H₂S levels. We also examined the role of CBS in inflammation, determined the total sulfide levels in infected mice, and examined how H₂S modulates Mtb growth, gene expression, metabolism and respiration.

Results

Mtb infection alters CBS levels in macrophages

Infection of RAW 264.7 macrophages with Mtb H37Rv resulted in a 34-fold increase in the level of CBS (Fig. 1a, b), a major enzymatic source of H₂S in mammalian cells^22,23. This suggested that increased H₂S production is a host response to Mtb infection. To investigate the role of H₂S in Mtb virulence and pathogenicity, we used mice deficient in H₂S production due to disruption of Cbs, the gene encoding CBS^22,23. Since the Cbs^−/− genotype is perinatally lethal, we used heterozygous (Cbs^+/−) mice that have a normal life span and show an ~50% reduction in CBS levels²². Infection of Cbs^+/+ (WT) peritoneal macrophages resulted in a 25-fold increase in the level of CBS, whereas the levels of CSE and 3-MST did not change appreciably (Fig. 1c, d). Bacterial recovery from Mtb-infected Cbs^+/− peritoneal macrophages was reduced compared to WT cells (Fig. 1e), which correlates with lower H₂S levels in these cells as measured by an H₂S-specific probe (Fig. 1f). Mtb recovery from Cbs^+/− macrophages treated with the H₂S donor GYY4137²⁴ was increased compared to Cbs^−/+ cells alone (Fig. 1e). GYY4137 slowly releases H₂S thereby mimicking physiological release. This result suggests that differences in Mtb recovery are due primarily to H₂S. Lastly, we observed no impact on bacterial recovery from infected peritoneal macrophages treated with homocysteine (10, 25 or 50 µM), ruling out mild hyperhomocysteinemia observed in Cbs^+/− mice (<15 µM)^22,25 as a contributor to reduced bacterial burden (Supplementary Fig. 1).

Fig. 1: Attenuation of Mtb virulence and pathogenicity in Cbs^+/− mice.

a Representative western blot and b densitometry analysis of CBS levels in RAW 264.7 macrophages following 6 h of Mtb infection (n = 3). c Representative western blot and d densitometry analysis of H₂S-producing enzymes in WT mouse peritoneal macrophages following 6 h of Mtb infection (n = 3 animals per group). e CFU recovery from Mtb-infected peritoneal macrophages isolated from WT and Cbs^+/− mice at 36 h post infection. f Relative quantitation of H₂S production from uninfected or infected (Inf) peritoneal macrophages using the H₂S-specific fluorescent probe WSP1²⁸ at 36 h post infection. The Cbs^+/− Inf GYY4137 group received 25 µM GYY4137. (e) and (f) represent pooled data from two or more independent experiments (n = 4−6). g Survival of mice following Mtb infection, representative of two independent survival studies (n = 10 animals/group). h Representative western blot and i densitometry analysis of CBS in the lungs of age-matched mice without infection or 8 weeks post Mtb-infection (n = 3 animals per group). j Bacillary burden in lungs and k spleens at 8 and 14 weeks post infection (n = 5 animals per group). l Representative gross images of infected lungs at 8-week post Mtb-infection. Arrows indicate TB granulomatous lesions. m H&E staining of representative infected lung sections at 8 weeks post Mtb-infection. Arrows indicate area of lymphocyte infiltration (scale bar = 100 μm). n Quantitative histochemical appraisal of lesions in infected lungs of WT and Cbs^+/− mice_. Granulomatous region analysis data from three independent animal experiments with 2−3 mice per experiment. Data are shown as mean ± SEM, except (d) and (n) (mean ± SD). Data (b, d, j, k and n) were analyzed by unpaired parametric t test. Survival data (g) were analyzed by Log-rank (Mantel−Cox) test. Data in (e) and (f) were analyzed by one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test. Data in (i) were analyzed by two-way ANOVA with Tukey’s multiple comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001. Source data are provided as a Source Data file.

Cbs ^+/− mice survive longer following Mtb infection

The macrophage infection studies above clearly show decreased bacillary growth in H₂S-deficient Cbs^+/− cells and that Mtb growth could be chemically complemented with H₂S. To more clearly establish the role of H₂S in Mtb virulence and pathogenicity, we infected WT and Cbs^+/− mice with Mtb and observed that Cbs^+/− mice survived significantly longer than WT mice (median time to death 286 days vs. 211 days for WT mice (Fig. 1g)). CBS levels in the lungs of uninfected Cbs^+/− mice were significantly lower than in WT, and CBS levels in the lungs of WT mice increased significantly at 8 weeks post infection (Fig. 1h, i).

At 8 and 14 weeks post infection, Cbs^+/− mice had ~6-fold and ~4-fold fewer Mtb bacilli in the lungs, respectively, compared to WT mice (Fig. 1j). Similarly, we observed ~4.5-fold and ~2.5-fold fewer bacilli in spleens of Cbs^+/− mice compared to WT mice at 8 and 14 weeks post infection, respectively (Fig. 1k). While significant, these modest differences in organ burden contrast with the striking difference in survival (Fig. 1g). Furthermore, gross pathology (Fig. 1l) and histopathology appraisal (Fig. 1m, n) of infected WT and Cbs^+/− mice were consistent with the observed organ burden. We noted moderate to large coalescence of multiple granulomatous lesions with infiltration of lymphocytes in the lungs of WT mice whereas the lung architecture in Cbs^+/− mice was relatively intact with fewer small granulomatous lesions (Fig. 1m). In sum, our data suggest that Mtb infection of host cells upregulates CBS production, leading to increased H₂S levels, which exacerbate disease as indicated by reduced host survival, and to a lesser extent by increased organ burden. These findings identified a new phenotypic category of host genes that exacerbates TB disease whereas their disruption prolongs survival during infection.

Immune responses in Mtb-infected mice

To determine whether reduced bacillary burden and increased survival of Cbs^+/− mice could be attributed to immunological differences between Cbs^+/+ and Cbs^+/− mice, we examined host immune responses in Cbs^+/− and WT mice pre- and post-Mtb infection. In uninfected mice (Supplementary Figs. 2A−J, 3B, C), except in the spleens of Cbs^+/− mice which showed fewer macrophages (Supplementary Fig. 2F), we observed no significant differences in the percentages of myeloid or lymphoid cell types in the lungs and spleens of Cbs^+/− and WT mice, even though the levels of proinflammatory IL-1α were elevated and anti-inflammatory IL-10 were reduced in Cbs^+/− compared to WT mice (Supplementary Fig. 3A). To distinguish whether the reduced inflammatory response in Cbs^+/− mice during chronic infection is due to lower CBS protein levels or reduced bacillary burden, we analyzed the immune response at an early time after infection (3 weeks) when the lung bacillary burden was similar (Supplementary Fig. 4A), but serum sulfide levels differed between WT, and Cbs^+/− mice (Supplementary Fig. 4B). As shown in Supplementary Figs. 2K-R and 3D and E, except for increased percentages of splenic CD4⁺ T cells (Supplementary Fig. 2P), reduced percentages of CD4⁺IFN-γ⁺ T cells (Supplementary Fig. 2R) and reduced percentages of CD11b⁺Gr-1⁻ myeloid cell in the spleens (Supplementary Fig. 3E) of Cbs^+/− mice, there were no significant differences in percentages of myeloid or CD4⁺ T cells in the lungs and spleens of Cbs^+/− and WT mice at 3 weeks post infection. During the chronic stage of infection (8 weeks), cytokine levels in Cbs^+/− and WT mice were similar except for a lower TNF-α and IL-10 levels in Cbs^+/− mice (Supplementary Fig. 3A), which might be due to reduced percentages of neutrophils and lymphoid cells mostly in the spleen (Supplementary Figs. 2S-Z, 3F and G). We observed no clear predominance of either a Th1 or Th2 response, as indicated by the levels of proinflammatory (IL1α, IL-6, IL-12, INF-γ and TNF-α), anti-inflammatory (IL-5, IL-10) and regulatory (IL-17) cytokines in the sera of Cbs^+/− and WT mice (Supplementary Fig. 3A). Overall, these data suggest that the differences in bacillary burden and increased survival in Cbs^+/− mice are likely not attributable to immunological differences alone, and are more likely due to direct action of H₂S on Mtb.

H₂S modulates mycobacterial growth and energy homeostasis

In eukaryotes, depending on the concentration, H₂S can either stimulate or inhibit the electron transport chain (ETC) resulting in altered oxidative phosphorylation (OXPHOS) and ATP production^16,26. Reported physiological levels of H₂S in tissue are controversial because of the assay methods employed and volatility of the gas²⁰. However, the current consensus is that nanomolar concentrations of H₂S seem most plausible^19,20,21. Unfortunately, this low concentration is below the limit of detection (1 µM) of routine assays such as the methylene blue, 5,5′-dithiobis-2-nitrobenzoic acid and bismuth chloride assays^19,20.

Since our in vivo data suggest that host-generated H₂S interacts with Mtb to modulate disease, we tested the hypothesis that H₂S directly modulates Mtb growth. For these in vitro studies, we employed GYY4137, a widely used H₂S donor that releases low amounts of H₂S over a sustained period^24,27. Unlike widely used •NO and CO donors such as diethylenetriamine NONOate (DETA NONOate) and CO-releasing molecules (CO-RMs), respectively, GYY4137 is an unusually slow H₂S releaser wherein the final molar concentration of H₂S released is significantly lower than the starting concentration of GYY4137. These aspects, along with the volatility of the gas²⁰, make detection of H₂S by standard assays challenging^24,27. Therefore, we used LC-MS/MS to first confirm spent (decomposed) GYY4137, and the decomposition of GYY4137 in the presence of Mtb cells under our conditions into the corresponding molecular ions with m/z of 272.05 (M − H)⁻, 187.02 (M − H)⁻ and 88.08 (morpholine M + H)⁺ (Supplementary Fig. 5), and carefully considered the complex physicochemical properties of H₂S in the design of our experiments (see Materials and methods).

To correlate the amount of H₂S generated in vivo with our in vitro experiments, we used a fluorescent H₂S-specific probe, WSP1, to measure levels of H₂S produced over time by macrophages following Mtb infection²⁸. We then determined which concentrations of GYY4137 generated similar levels of fluorescence with WSP1 over time due to H₂S release. We then exposed Mtb cells to this range of GYY4137 concentrations that mimics levels of H₂S produced physiologically by macrophages following infection (Supplementary Fig. 6A, Fig. 1f). As expected, 5, 10, and 25 µM GYY4137 significantly stimulated growth. At day 12, maximum bacterial growth was observed with 25 µM GYY4137 (Fig. 2a, b). In contrast, 50 and 100 µM GYY4137 reduced the growth stimulation in a dose-dependent manner (Fig. 2a, b). Notably, we observed no growth change in Mtb control cultures exposed to 25 µM spent GYY4137, comprised of components of the chemical backbone alone (Supplementary Figs. 5 and 6B). These data establish that low levels of sulfide exert a concentration-dependent bimodal effect on Mtb growth.

Fig. 2: GYY4137 modulates Mtb growth, bioenergetics and ATP production.

a CFU-based growth profile of Mtb following incubation with GYY4137. Inset shows early Mtb growth. b Mean CFU values at day 12 of the growth analysis. c ATP levels in Mtb following 24 h exposure to GYY4137. Data (a−c) are representative of at least two independent experiments. d OCR profile (n = 4−6) and e average OCR of Mtb exposed to GYY4137 (n = 20−30). Control (CT) groups in panels (a−e) received spent GYY4137. f LC-MS targeted metabolic analysis of Mtb with or without exposure to 10, 25 and 100 μM GYY4137 (for 1 or 24 h). Graphs show relative abundance (peak area) of metabolites in Mtb cells (n ≥ 3 per group). Data are shown as mean ± SD, except (b) and (f) (mean ± SEM). Data in (a) were analyzed using two-way ANOVA with Tukey’s multiple comparisons test, with day 12 data shown in (b), or one-way ANOVA (c−f) with Sidak’s multiple comparisons test. All statistical differences are relative to untreated controls. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Source data are provided as a Source Data file.

Since enhanced growth is coupled to increased OXPHOS and ATP levels, and low levels of H₂S can stimulate mitochondrial respiration^14,18, we hypothesized that H₂S targets the Mtb ETC to modulate energy metabolism. We exposed Mtb cultures to GYY4137 and observed a dose-dependent increase in ATP levels, which was lost in cells exposed to 100 µM GYY4137 (Fig. 2c). Next, we performed extracellular flux analysis, which we have adapted to analyze Mtb bioenergetics^29,30,31, to determine the effect of H₂S on Mtb respiration in real time in a noninvasive manner. Again, we observed a bimodal response to GYY4137, wherein low concentrations resulted in a significant dose-dependent increase in the oxygen consumption rate (OCR) and 100 µM GYY4137 mitigated the increased OCR (Fig. 2d, e). Importantly, spent GYY4137 did not alter the basal OCR of Mtb cells (Supplementary Fig. 6C). The effects of sulfide on mycobacterial OCR were independently validated using another donor, NaHS (Supplementary Fig. 6D) and by employing an Oroboros O-2k respirometer with the nonvirulent vaccine strain Mycobacterium bovis BCG. The bimodal effect of H₂S on growth was also recapitulated with NaHS in Mycobacterium bovis BCG (Supplementary Fig. 6E) and Mtb (Supplementary Fig. 6F).

In sum, our results demonstrate that low levels of H₂S target the ETC to increase respiration and ATP levels, leading to increased growth. The capacity of H₂S to positively regulate the ETC, a key component of central metabolism, to induce a hypermetabolic state is likely to benefit Mtb during early infection.

H₂S stimulates Mtb central metabolism

Since respiration is tightly linked to central energy metabolism, we investigated the effect of H₂S on Mtb metabolism in more detail. We performed LC-MS/MS-based targeted metabolomic analysis of Mtb cells at 1 and 24 h post-exposure to GYY4137 and observed increased levels of the glycolytic intermediates glucose-6-phosphate, fructose-1,6-bisphosphate and pyruvate. Increased levels of these metabolites in the presence of excess ATP demonstrates augmented glycolytic flux (Fig. 2f), since this pathway would be otherwise inhibited when excess ATP is present³². Levels of the TCA cycle intermediates citrate and isocitrate were also increased, suggesting that feedback inhibition due to elevated NADH levels is not occurring (Fig. 2f). TCA cycle intermediates such as citrate also augment fatty acid biosynthesis to support mycobacterial growth³². Increased NADH, which feeds into the ETC to support increased ATP production is consistent with the ability of H₂S to stimulate growth (Fig. 2a, b) and respiration (Fig. 2d, e). It should be noted that small changes in absolute levels of metabolite pools can have critical implications for metabolism, as this points to dysregulation of metabolic balance that must be maintained at all costs^33,34. Carbon rearrangement in metabolites is an early indicator of metabolic dysregulation prior to changes in abundance and can be measured using stable isotopes. Therefore, to further examine how H₂S reprograms Mtb metabolism, we performed a series of carbon tracing experiments. We first confirmed that the mass distribution vectors (MDV) of glutamate and α-ketoglutarate were identical, thereby indicating that isotopic steady state had been reached under our culture conditions (Supplementary Fig. 7A, B). To determine the fuel source (glucose, acetate, or other endogenous carbon sources (e.g., amino acids)) contributing to carbon fractions in glycolytic and TCA metabolites, we monitored the fate of the uniformly labeled stable isotope [¹³C₆] glucose and [¹³C₂] acetate, and calculated the fractional nutrient contribution (FNC; Fig. 3a)³³. We found that exposure to GYY4137 reroutes carbon away from downstream glycolytic intermediates towards cell wall synthesis and the pentose phosphate pathway (PPP) in a dose-dependent manner. For example, increasing concentrations of GYY4137 increased glucose carbon flow into glucose/fructose-6-phosphate (G/F6P) relative to the spent GYY4137 control and reciprocally reduced carbon flow from other endogenous supplementary carbon sources into this metabolite (Fig. 3b). Secondly, an increase in the FNC of glucose into glycerol-3-phosphate (G3P) and sedoheptulose-7-phosphate (S7P) suggests reprogramming of metabolism towards cell wall biosynthesis and the PPP, which produces NADPH and nucleosides necessary for lipid anabolism and DNA synthesis, respectively (Fig. 3c, e). Notably, the total FNC from other carbon sources (~84%) into G3P was significantly higher than the FNC from glucose (~12%) and acetate (~1%; Fig. 3c). Thirdly, increased levels of GYY4137 reduced entry of glucose or acetate carbons into the TCA cycle intermediate succinate, whereas the FNC of other carbon sources to succinate is increased (Fig. 3d). Interestingly, entry of glucose carbons into malate was unchanged (Fig. 3f), whereas entry of acetate carbons into malate and succinate was reduced with the reciprocal increase from other carbon sources (Fig. 3d, f).

Fig. 3: Fractional nutrient contribution (FNC) of glucose and acetate exposed to GYY4137.

a FNC was calculated according to a previous study³³; see also Materials and Methods. b, c Glycolysis is stimulated by sulfide as is evident by an increased FNC_glucose of G/F6P that corresponds with increased GYY4137 concentration, whereas the FNC_acetate and FNC_other decreases. The FNC_glucose of G-3-P (c) and S-7-P (e) increases with higher concentrations of GYY4137. d, f The utilization of glucose is skewed towards cell wall synthesis and the PPP leading to a decrease in FNC_glucose and FNC_acetate of succinate and malate as GYY4137 concentration increases, as well as a corresponding increase in FNC_other of these two metabolites. Data are representative of two independent experiments (n = 5) and are shown as mean ± SD. Data were analyzed using one-way ANOVA with Sidak’s multiple comparisons test. All P values are relative to untreated controls (0 µM GYY4137) for individual metabolites. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Source data are provided as a Source Data file.

Overall, the carbon tracing data indicate that enhanced energy flux through glycolysis, the PPP and TCA cycle following H₂S exposure increases Mtb energy metabolism and anabolic substrate intermediates for growth. Lastly, the FNC data demonstrate that H₂S rewires energy flux through these pathways by modulating the fraction of the metabolite’s carbon produced from the respective fuel source.

H₂S exerts it bioenergetic effect predominantly through CytBD

We next examined how H₂S stimulates Mtb respiration and growth. Mycobacteria possess two terminal respiratory oxidases that provide substantial respiratory flexibility as electrons from the menaquinone pool can be channeled either to an aa₃-type cytochrome c oxidase via the cytochrome bc₁ (cytochrome bc₁/aa₃) complex or directly to cytochrome bd-type menaquinol oxidase (CytBD)³⁵. CytBD can partially compensate for a loss in activity of the bc₁-aa₃ complex³⁶, and is required during the transition from acute to chronic Mtb infection in mouse lungs to promote bacterial virulence³⁷. Potassium cyanide (KCN) inhibits the aa₃-type cytochrome c oxidase, but not CytBD in Mycobacterium smegmatis (Msm)³⁵. Also, CytBD in E. coli has been shown to be less sensitive to cyanide and sulfide than aa₃-type cytochrome c oxidase³⁸. To determine whether H₂S exerts its effect through CytBD or the bc₁-aa₃ complex, we exposed Msm cells to KCN and/or NaHS in an Oroboros O-2k respirometer. We observed that addition of NaHS to Msm leads to an increase in basal OCR (Fig. 4a), consistent with the effects of H₂S on Mtb. In contrast, addition of KCN resulted in a steady decline in the basal OCR of Msm cells. Interestingly, addition of NaHS to KCN-treated cells triggered an increase in OCR (Fig. 4a). To define the effect of H₂S on Mtb growth and ATP production in more detail, we used Mtb rv1620c::aph (hereafter referred to as cydC::aph), which contains the kanamycin resistance gene (aph) inserted in cydC (rv1620c) as described³⁷. Mtb cydC encodes a membrane protein linked to the incorporation of a heme cofactor into CytBD. Mtb cydC::aph is defective in CytBD activity and shares the phenotype of Mtb cydAB mutants^30,39. Importantly, Mtb cydC::aph produces only a functional cytochrome bc₁/aa₃⁴⁰. As a control, we used a complemented strain, Mtb cydC::aph-Comp, obtained by expressing cydC from an integrated plasmid. Effective complementation was demonstrated by inhibiting cytochrome bc₁ using Q203 as described³⁰ (Supplementary Fig. 8A), which reroutes electrons to CytBD leading to an increase in OCR (Supplementary Fig. 8B and C). Complementation was also demonstrated by the marked increase in OCR of WT Mtb and Mtb cydC::aph-Comp cells compared to Mtb cydC::aph in the presence of 25 and 100 µM GYY4137 (Supplementary Fig. 8D and E). Although E. coli cydD mutants were shown to be susceptible to high (20 mM) concentrations of cysteine⁴¹, we found that Mtb cydC::aph is not susceptible to physiological levels of cysteine (~128 and ~250 µM in human cells and plasma, respectively)^42,43 (Supplementary Fig. 9). Next, we measured the OCR in cydC::aph cells following exposure to GYY4137 and observed a small, but significant increase in OCR at 10 and 25 μM GYY4137 compared to control cells receiving spent donor (Fig. 4b, c). Notably, at higher GYY4137 concentrations (≥50 μM), not only is the stimulatory effect on the OCR lost, but the OCR is repressed below the basal level, indicating an inhibition of cytochrome bc₁/aa₃ oxidase in cydC::aph cells (Fig. 4b, c). CydC::aph cells exposed to GYY4137 displayed a modest bimodal response as shown by a moderate increase in growth and ATP levels at 10 and 25 μM GYY4137, but significant suppression at higher GYY4137 levels (≥50 μM) (Fig. 4d–f). In contrast, the basal OCR and growth of GYY4137-treated WT Mtb cells remained significantly higher than cydC::aph cells at 50 and 100 μM GYY4137 (Figs. 2d, e, 4b, c). In summary, the significantly reduced stimulation of OCR and growth in cydC::aph cells (Fig. 4g, h), i.e., when respiration is sustained only by aa₃-type cytochrome c oxidase, indicates that H₂S exerts its respiratory and growth-stimulatory effects predominately via CytBD in Mtb.

Fig. 4: H₂S stimulates Mtb bioenergetics primarily through cytochrome bd-type quinol oxidase.

a Representative OCR profile of M. smegmatis (Msm) cells. Arrows indicate the time of addition of KCN (100 μM) or NaHS (25 μM) (n = 5). b Representative OCR profile and c average OCR of cydC::aph cells exposed to GYY4137 (n = 12−18, data pooled from at least two independent experiments). d CFU-based growth profile of cydC::aph cells following exposure to GYY4137. e Mean cydC::aph CFU values at day 10 of the growth analysis. f ATP levels in cydC::aph after 24 h exposure to GYY4137. Data (d−f) are representative of at least two independent experiments. g Comparison of the dose-dependent effect of GYY4137 on the OCR of Mtb and cydC::aph cells (from Figs. 2e and 4c). h Percent change in CFU of Mtb and cydC::aph (e) following exposure to increasing concentrations of GYY4137 (from Figs. 2b and 4e). The percent change was calculated against the CFU of control cells at day 10 following GYY4137 exposure. Data are shown as mean ± SD, except (e) (mean ± SEM). All comparisons are against control cells (CT) exposed to spent GYY4137 except in (g) and (h), which compare Mtb and cydC::aph at equivalent GYY4137 concentrations. Data were analyzed using two-way ANOVA (b, d) with Sidak’s multiple comparisons test, with average OCR (c) or CFU at day 10 (e). Other data were analyzed by one-way ANOVA with Sidak’s multiple comparisons test (f) or by using an unpaired parametric t test for comparisons at each GYY4137 concentration (g and h). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Source data are provided as a Source Data file.

H₂S modulates sulfur/copper metabolism and the Dos regulon

To assess the effect of H₂S on the Mtb transcriptome, we performed RNA-seq analysis on Mtb and cydC::aph cells exposed to 25 µM GYY4137 for 1 h. Culture conditions were carefully controlled to eliminate the possibility of respiration-induced hypoxia (see Materials and methods). At 25 µM GYY4137, we observed a distinct transcriptomic profile of 92 differentially regulated genes in Mtb compared to untreated cells including those belonging to metabolism and respiration (26/92), and cell wall and cell processes (19/92) (Supplementary Fig. 10A, B). Notably, several members of the CsoR⁴⁴ and RicR⁴⁵ copper regulons, and six genes involved in cysteine/sulfur metabolism were upregulated (Fig. 5a, b). RNAseq analysis showed no differentially regulated genes in cydC::aph cells exposed to 25 µM GYY4137 compared to untreated cydC::aph cells, consistent with our observations that H2S exerts its effect primarily through CytBD.

Fig. 5: H₂S induces transcriptome remodeling and mitigates oxidative stress in Mtb.

RNA-seq-based gene expression analysis shows induction of the Mtb a copper, b sulfur and c Dos regulons following exposure to 25 µM GYY4137 for 1 h. Genes exhibiting ≥1.5-fold (+/–) change in expression compared to the expression in untreated cells were considered differentially regulated (n = 3 independent biological replicates per group). d Proposed model detailing the relationship between H₂S and induction of the Mtb copper regulon. Exposure to H₂S leads to the formation of copper sulfides, which relieve repression of CsoR leading to the export of copper via the cso. e Representative results of qRT-PCR analysis of key dormancy regulon genes following 1 h exposure to 25 µM GYY4137 in Mtb H37Rv, ΔdosS, ΔdosT and ΔdosST (n = 5 independent biological replicates per Mtb strain). Fold change in gene expression following addition of GYY4137 is shown. f Survival of Mtb cells (% CFU) exposed to metronidazole (MN) alone, or MN and 25 µM GYY4137. g Recovery profile of Mtb cells following prolonged hypoxia (8 weeks). Prior to hypoxia, one group of cells received 25 µM GYY4137. h Mean CFUs following 1 week of recovery following prolonged hypoxia. (g) and (h) represent pooled data from two or more independent experiments (n = 4). i Percentage survival (CFU) of 25 µM GYY4137-treated Mtb cells following exposure to oxidative stressors CHP and MND (0.5 mM for 6 h) relative to untreated cells. j Percent of dihydroethidium (DHE)-positive ROS-producing Mtb cells following exposure to 10−50 µM GYY4137 (spent GYY4137 used as 0 µM control; n = 3 per group). CHP treatment (15 mM) was used as a positive control for detection of ROS-positive cells. Data in (f) and (i) are representative of two or more experiments. All data are shown as mean ± SD, except (e) (mean ± SEM). Data were analyzed using one-way ANOVA (e and j) with Sidak’s multiple comparisons test, or by two-way ANOVA with Sidak’s multiple comparisons test (f−h), or by unpaired parametric t test (i). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Source data are provided as a Source Data file.

An important finding was that 31 of the 48 genes comprising the Mtb Dos dormancy regulon¹, which includes the heme sensor kinase DosS, the response regulator DosR, as well as hspX, fdxA, narX, narK2, tgs1 and others were significantly upregulated (Fig. 5c). Interestingly, the lack of CsoR was shown to upregulate the Dos dormancy regulon⁴⁴. CsoR is a negative regulator of the copper sensitive operon cso whose repression is relieved upon copper binding that leads to the export of copper⁴⁶. We propose that binding of copper to sulfide relieves repression of CsoR, which leads to the export of copper via the cso (Fig. 5d).

To determine whether H₂S regulates the Dos regulon via the Dos sensor kinases, we performed qRT-PCR analysis of WT Mtb and Mtb mutants lacking dosS (ΔdosS), dosT (ΔdosT) or both (ΔdosST) after exposure to 25 µM GYY4137 for 1 h. Quantitative RT-PCR analysis of ΔdosS, ΔdosT and ΔdosST strains showed significantly reduced induction of key dormancy regulon genes such as hspX, fdxA, and rv1813c (Fig. 5e)¹. These findings are consistent with previous studies showing that the copper regulator, CsoR regulates the Dos regulon⁴⁴, and that cysteine stimulates Mtb respiration leading to the induction of the CsoR copper and Dos regulons⁴⁷.

In sum, our findings provide further credence to the proposed link between sulfide/cysteine-mediated stimulation of respiration, the Dos regulon, and copper homeostasis. The lack of differential gene expression in cydC::aph cells compared to cells treated with GYY4137 points to CytBD as an important component of the H₂S energy-transduction pathway. Our data suggest that upon exposure to sulfide, Mtb reprograms its metabolism and transcriptional machinery including the copper regulon to mitigate the formation of copper sulfide species. Finally, our data suggest that in addition to •NO, CO and low O₂, we have identified a fourth gas, H₂S, that can induce the Dos dormancy regulon.

Role of H₂S during Mtb growth under hypoxia

To determine whether growth-stimulatory levels of H₂S accelerate the development of hypoxia-induced, nonreplicating persistence in vitro, we used the anti-TB drug metronidazole (MN), which selectively kills Mtb cells under conditions of low O₂⁴⁸. We posited that if H₂S facilitates early entry into hypoxic dormancy, then GYY4137-exposed Mtb cells in the presence of MN would be killed more quickly than unexposed cells in the presence of MN. Indeed, we observed a ~20% and ~40% reduction in CFU of GYY4137-exposed Mtb cultures in the presence of MN at 24 and 72 h, respectively (Fig. 5f). In contrast, the viability of unexposed cells was not affected by MN treatment for 24 or 48 h and a slight reduction (~10%) was seen at 72 h. These results show that H₂S can facilitate early entry of Mtb cells into hypoxic dormancy. Next, we subjected Mtb cultures to hypoxia for 8 weeks with or without initial exposure to GYY4137. Following re-aeration, Mtb cells previously exposed to GYY4137 displayed enhanced recovery from hypoxia as shown by more robust growth compared to unexposed cells (Fig. 5g, h). This outcome is consistent with our observation that exposure to low levels of H₂S triggers an antioxidant response evident by upregulation of genes involved in cysteine metabolism (Fig. 5a, b) and antioxidant genes such as sodA and ahpC^49,50 (Supplementary Fig. 10A), which permit rapid emergence from hypoxia. We subsequently tested the ability of H₂S to protect Mtb under conditions of oxidative stress and found that survival of Mtb cells exposed to GYY4137 was significantly increased compared to unexposed cells in the presence of the oxidants cumene hydroperoxide (CHP) and menadione (MND) (Fig. 5i). Since H₂S can act as an antioxidant^51,52, H₂S exposure may help mitigate oxidative stress during growth or when Mtb emerges from hypoxia.

We next evaluated whether increased respiration following H₂S exposure may result in increased ROS production in Mtb cells. We examined the effect of GYY4137 on ROS production during respiration and observed a slight reduction in the percentage of ROS-producing cells following GYY4137 exposure (Fig. 5j), which is likely due to the known antioxidant effect of H₂S. These data are consistent with the growth-promoting effect of H₂S on Mtb cells, as excessive ROS production will inhibit growth. We also excluded the possibility that H₂S exerts a direct effect on the ETC via a sulfide quinone oxidoreductase (SQR) since the OCR of the Mtb sqr mutant (rv0331::aph; TARGET, Johns Hopkins University, USA) exposed to H₂S was similar to that of WT Mtb (Supplementary Fig. 11A). This contrasts with Mtb exposure to the reductant DTT, which increases the menaquinol pool to stimulate OCR (Supplementary Fig. 11B)⁵³, but not growth⁵⁴. Overall, our data demonstrate a beneficial role for H₂S during entry and emergence from low-oxygen conditions, as well as protection against oxidative stress.

Crosstalk between H₂S and •NO alters Mtb infection

Studies in mammalian systems have shown crosstalk between •NO and H₂S^55,56. We observed increased iNOS levels in the lungs of Mtb-infected Cbs^+/− mice compared to WT at 8 weeks post infection (Fig. 6a, b). Consistent with increased iNOS levels, infected Cbs^+/− mice had higher NO₂⁻ + NO₃⁻ levels in serum compared to WT mice, suggesting increased •NO production in Cbs^+/− mice (Fig. 6c). Increased NO₂⁻ + NO₃⁻ levels were also observed in the culture media of infected Cbs^+/− macrophages (Fig. 6d). Addition of GYY4137 significantly lowered NO₂⁻ + NO₃⁻ levels in WT and Cbs^+/− macrophages (Fig. 6d). These observations are consistent with the ability of H₂S to regulate •NO bioavailability in vivo, including during inflammation^57,58. By itself, •NO exerts bacteriostatic effects on Mtb by inhibiting respiration¹, and increased •NO may also contribute to inhibition of bacterial respiration and growth during infection. As expected, exposure of Mtb to •NO in vitro decreased basal OCR (Fig. 6e, f). Intriguingly, the OCR was restored following exposure to GYY4137 (Fig. 6e, f). ATP levels were also reduced in Mtb cells following exposure to •NO (Fig. 6g), suggesting inhibition of respiration¹. In contrast, the ATP levels of cells exposed first to •NO followed by GYY4137 were similar to those in unexposed cells, corroborating our observation that GYY4137 restored OCR following •NO exposure (Fig. 6e, f). We then performed growth assays using Mtb cells exposed to •NO alone, GYY4137 alone, or •NO and GYY4137. Mtb cells exposed to GYY4137 exhibited significantly increased growth whereas growth was reduced in cells exposed to •NO, confirming the bacteriostatic effect of •NO (Fig. 6h). When GYY4137 was added following •NO exposure, the CFU recovery was similar to that of unexposed cells (Fig. 6h). While the exact chemical nature of the crosstalk between H₂S and •NO remains poorly defined, these data suggest that exposure to H₂S reverses the bacteriostatic effect of •NO on Mtb respiration and growth.

Fig. 6: Interaction between •NO and H₂S and pharmacological modulation of sulfide levels in mice.

a Representative western blot and b densitometry analysis of iNOS in mouse lungs at 8 weeks post infection (n = 3 animals per group). c Nitrate and nitrite levels in the sera of infected (Inf) and uninfected mice and d the culture medium of infected mouse macrophages exposed to 25 µM GYY4137 for 36 h. Data in (c) and (d) are pooled from two independent experiments (n = 7−8 animals per group). e OCR profile and f average OCR of Mtb cells exposed to •NO (DETA/NO; 250 µM), •NO followed by GYY4137 (10 µM), or untreated (UT) in XF media (n = 6−8). Media control contained no cells. g ATP levels (shown as RLU normalized to CFU) in Mtb cells treated (•NO, GYY4137, or •NO and GYY4137) or untreated (UT) for 2 or 8 h. h CFUs after 8 and 24 h of exposure. T = 0 indicates CFU at the start of the experiment. Data in (g) and (h) are representative of at least two independent experiments. i CFU recovered from infected Cbs^+/+ peritoneal macrophages receiving one of the following: aminooxyacetic acid (AOAA), AOAA followed by GYY4137, GYY4137 alone, or untreated (UT) (n = 4). j Mtb CFU recovered from mouse peritoneal macrophages. k Bacillary burden in the lungs of mice that received: AOAA, AOAA followed by GYY4137, or GYY4137 alone. AOAA (13 mg/kg, 6 d/wk for 6 weeks) and GYY4137 (25 mg/kg 3 d/wk for 6 weeks) were administered intraperitoneally. Treatment began 2 weeks post-Mtb infection and continued for 6 wks. Data represent two independent experiments (n = 5−6 animals per group). l Total sulfide levels in serum of mice in (k) using the methylene blue method (n = 6−8). Data represent mean ± SEM except in (e, f and l) (mean ± SD). Comparisons are against WT mice or untreated (UT) controls. Data were analyzed by unpaired parametric t test (b, j), one-way ANOVA with Tukey’s multiple comparisons test (c, d, f, k and l) or two-way ANOVA with Sidak’s multiple comparisons test (g−i). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Source data are provided as a Source Data file.

Pharmacological inhibition of CBS in vitro and in vivo

To validate our in vitro findings that H₂S stimulates Mtb respiration and growth, we tested whether manipulating H₂S levels via pharmacological modulation of H₂S-producing enzymes could alter disease progression. First, we infected peritoneal macrophages from WT mice and treated the cells with the CBS inhibitor aminooxyacetic acid (AOAA)¹⁴. We observed ~17% and ~30% reductions in CFU from infected macrophages after exposure to AOAA for 6 and 36 h, respectively (Fig. 6i). In contrast, CFU recovery from AOAA-treated WT macrophages exposed to GYY4137 was comparable to recovery from untreated macrophages indicating that H₂S itself is primarily responsible for the reduced bacterial burden in AOAA-treated cells (Fig. 6i). Further, within Mtb-infected Cbs^+/− peritoneal macrophages, bacillary growth was reduced by ~35% and 40% versus WT cells at 6 and 36 h post infection, respectively (Fig. 6j).

Next, we show that administration of the CBS inhibitor AOAA to WT mice reduces Mtb burden in the lungs to the level of Cbs^+/− mice (Fig. 6k). Further, administration of GYY4137 to Cbs^+/− mice increased Mtb burden to that of WT mice. Of note, the bacillary burden in the lungs of WT mice treated with AOAA and GYY4137 was similar to the burden in untreated WT mice (Fig. 6k). Determination of sulfide levels in the sera of these same groups of mice revealed that Mtb infection increased sulfide levels (Fig. 6l). Further, we observed that the reduction in serum sulfide observed in Cbs^+/− mice and AOAA-treated WT mice was associated with reduced lung bacillary burden (Fig. 6k, l). CFU recovery in Cbs^+/− mice and AOAA-treated WT mice following GYY4137 treatment was comparable to that in untreated WT mice, confirming our previous findings that differences in CFU are due to H₂S, and not CBS. While it could be argued that AOAA targets Mtb directly, addition of AOAA to Mtb cultures at concentrations as high as 1 mM does not inhibit Mtb growth. However, 2 mM AOAA reduces Mtb proliferation by ~35% (Supplementary Fig. 12). Thus, while our in vitro and in vivo infection data demonstrate that AOAA reduces host H₂S production to modulate Mtb infection, we cannot completely exclude the possibility of a direct effect of AOAA on Mtb. Overall, our findings provide evidence that H₂S is a modulator of Mtb respiration, and that host-derived H₂S is a significant contributor to TB disease.

Discussion

Here we establish a previously unrecognized role for H₂S in promoting TB, wherein increased host H₂S levels following Mtb infection promote bacterial growth, thereby reducing host survival. In addition, we have identified a genetic locus, Cbs, which exacerbates TB disease. This was evident from our mouse studies that demonstrated significantly increased survival of infected Cbs^+/− mice, which produce reduced levels of H₂S. This is an unusual finding as studies to date have shown that disruption of host loci implicated in TB lead to increased susceptibility and not resistance. Our findings suggest that Mtb elicits a deleterious host response via CBS to promote progressive pulmonary disease. Our data show that H₂S stimulates Mtb respiration, predominately via CytBD, and reprograms central metabolism leading to enhanced growth. We observed that iNOS expression is higher in infected Cbs^+/− mice and that H₂S effectively reverses •NO-mediated suppression of Mtb respiration in vitro. Further, pharmacological inhibition of CBS by AOAA reduced the sulfide levels and bacillary load in mice, and this effect could be reversed by administration of GYY4137, a physiologically relevant H₂S donor. Since host H₂S biosynthesis does not require O₂ as a cofactor^59,60 and H₂S could function as an Mtb fuel source during hypoxia, our findings have implications for understanding how Mtb can survive for extended periods under hypoxic conditions. Overall, our findings suggest a feedback loop mechanism, wherein host H₂S production is triggered by Mtb to promote a hypermetabolic state that supports establishment of TB disease.

The role of host-generated H₂S in microbial pathogenesis is poorly understood. To the best of our knowledge, no study has yet reported a role for host H₂S in the control of bacterial disease. A biphasic response to H₂S is well established wherein H₂S, depending on concentration, can trigger a variety of physiological and immunological responses, which can be cytotoxic or cytoprotective^14,16,61. We initially found that CBS protein and sulfide levels were increased in Mtb-infected macrophages, suggesting a direct link between Mtb infection and H₂S. Further studies showed that Cbs^+/− mice exhibited lower sulfide levels that corresponded with reduced bacterial burden and increased survival. We then sought to determine whether differential immune responses accounted for the increased survival and reduced bacillary burden in Cbs^+/− mice. Given the relative minor differences in immune response between WT and Cbs^+/− mice, and that H₂S can act directly on bacteria, we posited that differences in bacillary burden and survival are due to the direct action of host H₂S on Mtb physiology. Indeed, we observed modest differences in organ burden between Cbs^+/− and WT mice, but a striking difference in survival (Fig. 1g). Such a disparity between organ burden and survival is not unusual and was previously reported for Mtb whiB3⁶² and sigH⁶³ mutants that showed similar organ burden, but substantially prolonged survival. Overall, our findings suggest that Mtb possesses yet-to-be identified mechanisms that trigger upregulation of CBS, ultimately leading to increased organ burden and mortality.

Similar to the effects of H₂S observed in mammalian cells¹⁴, we show that low levels of H₂S exert a dose-dependent bimodal effect on Mtb growth and promote growth by stimulating respiration and ATP production. We confirmed the ability of H₂S to stimulate respiration and growth of Mtb, M. bovis BCG and M. smegmatis using two independent respirometry techniques and two different H₂S donors. Our results are complementary to recent studies of E. coli, a facultative anaerobe and member of the gut microbiota that is continuously exposed to large quantities of H₂S^38,64. These studies showed that H₂S inhibits the E. coli heme-copper bo₃ oxidase whereas the two cytochrome BD oxidases are resistant to H₂S. An unusual feature of Mtb respiration and metabolism is the ability to be stimulated by H₂S. Using respirometry, RNAseq and growth assays, our data indicate that cytochrome bc₁-aa3 (fused complex III/IV) is the primary target of H₂S and that CytBD mediates the outcome of metabolic stimulation and growth.

Initial studies of the stimulatory effect of H₂S on mitochondrial respiration have shown that at low sulfide levels, H₂S acts as a substrate by binding and reducing cytochrome c oxidase (Cco)^65,66. The cytochrome c oxidase complex contains two redox active metal sites, heme a and Cu_A, that donate electrons to the binuclear center composed of heme a3 and Cu_B, which reduces O₂^67,68. The redox states of the heme prosthetic groups and copper centers play a critical role in their reactions with gases and can mediate crosstalk between gaseous ligands. For example, in contrast to •NO and CO, which bind primarily to the binuclear center in the fully reduced (Fe²⁺/Cu¹⁺) state, sulfide does not react with the fully reduced enzyme, but reacts with met (Fe³⁺) heme. Hence, reactions with sulfide occur in the oxidized or mixed valence states^55,69. A recent model for sulfide-mediated inhibition of cytochrome c oxidase proposed that both Fe³⁺ heme a3 and Cu_B (Cu¹⁺ or Cu²⁺) can bind H₂S or SH⁻⁶⁹. The final enzymatic product of sulfide inhibition was suggested to be a mixed valence form of the enzyme containing reduced heme a and Cu_A, whereas heme a3 is in a ferric sulfide-bound state, which blocks the reaction with O₂^66,70. A second sulfide molecule is also possibly bound to Cu_B^+ 15,69. It should be noted that both CN⁻ and H₂S bind heme iron in the Fe³⁺ state and not surprisingly, also compete for the heme a₃-Cu_B binuclear center^71,72. Regardless of the precise mechanism whereby H₂S regulates Mtb respiration and metabolism, our data suggest that Mtb cytochrome bc1/aa3 is a target for host-generated H₂S. Lastly, the mechanisms whereby H₂S stimulates Mtb respiration and growth are expected to be different as Mtb has the remarkable capacity to rapidly reroute electrons to cytochrome bc₁-aa3 or CytBD³⁰, and the ETC is not subjected to backpressure. While it is possible that H₂S stimulates Mtb respiration via an SQR, our data suggest this is unlikely since lack of rv0331 did not alter the OCR following H₂S exposure. However, more studies are needed to conclusively demonstrate whether Rv0331 is indeed an SQR.

Since respiration is tightly linked to metabolism, we expected that H₂S-induced changes in Mtb respiration would impact central metabolism. Indeed, our carbon tracing and FNC analyses show that H₂S stimulates the PPP, which provides nucleosides for DNA synthesis and NADPH for lipid anabolism, which is required for increased metabolism and growth and the antioxidant response. Of interest is the capacity of H₂S to rewire energy flux by increasing NADH and ATP levels, and to control the carbon fractions of glycolytic and TCA metabolites, pointing to the complexity of this gas’s interaction with central metabolism.

The stimulation of respiration and upregulation of the Dos regulon by H₂S highlight an unusual facet of Mtb physiology. For example, inhibition of Mtb Complex III/IV (cytochrome bc1-aa3 oxidase) and Complex V (ATP synthase) with the anti-TB drugs Q203 and bedaquiline (BDQ), respectively, dramatically enhances Mtb respiration³⁰. Furthermore, increased respiration³⁰ and induction of the Dos regulon by BDQ⁷³ or cysteine⁴⁷ provide evidence of a link between respiratory control and the Dos regulon. Paradoxically, NO, CO, and low O₂ inhibit respiration and bind to DosS/DosT heme irons to induce the Dos regulon^1,4,5, whereas H₂S, BDQ and cysteine stimulate respiration leading to induction of the Dos regulon. These findings raise the questions of how H₂S induces the Dos regulon and what the relationship is between sulfur and copper metabolism as identified in our RNAseq experiments. Since H₂S can coordinate with heme⁶⁵, it is possible that H₂S directly triggers upregulation of the Dos regulon via the heme-containing sensor kinases DosS and/or DosT. However, the DosT heme iron is in the oxy (O₂) state and unable to bind H₂S whereas DosS is in the Fe³⁺ state⁷⁴, which can bind H₂S. The precise mechanism of regulation of the Dos regulon appears to be more complex than previously anticipated and is confounded in part by the lack of knowledge of how in vitro biochemical data relate to the redox states of DosS and DosT in the Mtb cell.

By examining our expression data, it could be argued that induction of the Dos dormancy regulon by H₂S is less robust than that of •NO¹ or hypoxia⁵. However, a direct comparison is not possible since different concentrations of gases were used and their diverse physicochemical properties will influence binding to DosS/DosT heme irons (i.e., solubility, diffusion, on- and off-rates, etc.). Notwithstanding, the overall levels of induction of Dos dormancy regulon members by H₂S is comparable to⁴ and may exceed that of CO⁷⁵.

Upregulation of the Dos regulon in a Mtb csoR mutant has been reported⁴⁴. Based on the unusually strong affinity between H₂S and copper, we posit that copper is (i) exported via the copper responsive proteins CtpV⁷⁶ and CtpF, (ii) bound via the strong copper-binding protein metallothionein [MymT;⁷⁷], and (iii) oxidized by the copper oxidase (rv0846c). Further attempts by the cell to detoxify excess H₂S are evident by the upregulation of rv0848 that detoxifies sulfide to generate cysteine (O-acetyl-l-serine + H₂S → l-cysteine + acetate). We propose that accumulation of toxic copper sulfides relieve CsoR-mediated repression of the cso regulon to promote copper export.

It is difficult to ignore the critical role that the copper-containing cytochrome bc1/aa3 plays in the induction of the Dos dormancy regulon as it is targeted by •NO, CO, and O₂ (hypoxia) to inhibit respiration and induce the Dos regulon. H₂S generates a distinct response since it reroutes electrons away from cytochrome bc1/aa3 toward the non-copper-containing CytBD, but also induces the Dos dormancy regulon. Intriguingly, this model may not appear to agree with a study showing induction of the Dos dormancy regulon with the anti-TB drug BDQ⁷³, which targets Mtb Complex V and not cytochrome bc1/aa3. However, this drug also reroutes electrons to CytBD³⁰. Hence, it appears that the rerouting of electrons to CytBD, and therefore the Mtb ETC, plays a more prominent role in the induction of the Dos regulon that previously appreciated. One limitation of this study is that currently available pharmacological inhibitors of CBS (such as AOAA) have suboptimal bioavailability and a degree of nonspecificity⁷⁸, which limit potency and clinical application. Further, as is evident from the wide range of H₂S levels reported in the blood of mice (0.1−80 µM) and humans (2−110 µM)¹⁹, accurate measurement of H₂S concentrations in biological samples is problematic due to the complex sulfur chemistry and extreme volatility of H₂S. In fact, critical assessment of the literature and H₂S chemistry suggests that H₂S concentrations in vivo are in the nanomolar range^19,20,21. Not surprisingly, there is a lack of a universally accepted gold standard for the measurement of low concentrations of H₂S^19,20,21. A constraint of the commonly used methylene blue-based assay for H₂S quantitation is that it measures total sulfide, not H₂S alone, and can lead to overestimation of H₂S concentrations. In contrast, the H₂S-specific fluorogenic probe-based approach circumvents some of the limitations of the methylene blue method, but provides only relative quantitation²⁰. Clearly, methods to measure H₂S accurately and consistently in human tissue, fluids or biological cultures are required to further explore a role for H₂S in TB. Lastly, generating tissue- or cell-type-specific Cbs knockout mice should allow more precise functional studies of the role of CBS in TB.

We demonstrate that exploiting host H₂S is an important Mtb survival strategy to establish disease (Fig. 7). These findings add to our understanding of the complex interaction between host gasotransmitters and Mtb and suggest the potential of pharmacological targeting of H₂S production as a therapeutic strategy for TB treatment. The interplay between H₂S and •NO that likely modulates TB pathophysiology is reflected in our animal studies and points to complex interactions at the primary site of infection, the lung. Lastly, knowing the spatial distribution of H₂S-producing enzymes in human hypoxic TB lesions may have important implications as H₂S possibly modulates Mtb bioenergetics under hypoxic conditions. It is tempting to speculate that H₂S may function as an essential fuel for maintaining Mtb energy metabolism during long-term persistence in hypoxic human TB lesions.

Fig. 7: Proposed mechanism for H₂S-induced disease exacerbation.

a As a response to Mtb infection, the host increases CBS and H₂S production in the lungs and macrophages leading to increased bacillary burden, increased inflammation, and reduced survival. b Host-generated H₂S is sensed by Mtb to stimulate respiration and energy metabolism, thereby increasing bacterial growth and accelerating disease. During infection, increased H₂S suppresses iNOS levels and production of •NO. H₂S also mitigates the bacteriostatic effect of •NO on Mtb respiration. Green arrows represent increase; red arrows represent decrease.

Methods

Study design

In this study, we hypothesized that host-derived H₂S modulates Mtb pathogenesis via respiratory and bioenergetics changes. To test this hypothesis, we first used Cbs-deficient mice (Cbs^+/−) to study survival and Mtb disease progression, and determined immunological changes. To determine the effect of H₂S on Mtb physiology, we studied Mtb growth kinetics, gene expression, metabolism and respiration. All animal experiments were conducted according to the highest ethical standards and approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Alabama at Birmingham.

H₂S donor compounds and preparation

GYY4137 [morpholin-4-ium 4 methoxyphenyl (morpholino) phosphinodithioate] (Cayman Chemicals, USA) and NaHS (Alfa Aesar, USA) were used as sources of H₂S and were prepared according to the manufacturer’s instructions. While NaHS releases H₂S instantaneously, GYY4137 is a water soluble H₂S donor that releases H₂S slowly for weeks both in vitro and in vivo^24,27, and more closely parallels the biological effect of endogenous H₂S. For H₂S exposure experiments, a 10 mM GYY4137 stock solution was prepared in water in a 5 ml tube and was immediately diluted to make a working stock and added to cultures to achieve the desired final concentration. Experimental control samples received time-spent 25 µM GYY4137 (referred to as spent GYY) that was previously aerated for at least 120 days. An NaHS stock solution (100 mM) was prepared by equilibrating phosphate-buffered saline (PBS), made anoxic by bubbling with argon, with NaHS in a septum-sealed glass syringe. In order to avoid any trace metals catalyzing the oxidation of H₂S, the metal chelator DTPA (diethylenetriaminepentaacetic acid) was added to the buffer to a final concentration of 50 µM prior to addition of NaHS. Photo-oxidation was prevented by wrapping the solution-containing syringe with aluminum foil. Despite these precautions, the presence of polysulfide in stock solutions cannot be completely excluded.

Bacterial strains and growth conditions

M. tuberculosis H37Rv (obtained from BEI Resources (NR-123), Mtb cydC::aph (containing a transposon in cydC; kindly provided by Dr. Digby Warner, University of Cape Town), Mtb ΔdosS, ΔdosT and ΔdosS/dosT mutants (kindly provided by Dr. David Sherman, University of Washington), M. smegmatis mc²155 and M. bovis BCG were grown in Difco Middlebrook (MB), 7H9 (broth) or BBL7H11 agar (BD Biosciences, USA) supplemented with 0.2% glycerol, 1× ADS (albumin, dextrose, sodium chloride) and 0.02% tyloxapol at 37 °C. Complementation of Mtb cydC::aph was achieved by placing cydC under control of the hsp₆₀ promoter. Cloning the hsp₆₀-cydC fragment into a stably integrating E. coli-mycobacterial plasmid (hygromycin resistance) followed by transformation and selection yielded viable Mtb cydC::aph-Comp colonies that showed a mild growth delay. H₂S donor compounds were prepared as mentioned above and added to the required concentration for growth assays. Cultures were removed from sealed vials at specific time points, serially diluted and aliquots were plated on BBL 7H11 agar plates. All sealed culture vessels containing H₂S donors were only opened once. CFU were enumerated after 4−6 weeks of incubation at 37 °C.

Breeding and genotyping of mice

Original colonies of C57BL/6J (Cbs^+/+) and B6;129P2-Cbs^tm1Unc/J heterozygous mice (Cbs^+/−) were obtained from the Jackson Laboratory (stock# 000664 and 002461, respectively; Bar Harbor, ME) and subsequently backcrossed for at least seven generations before starting the experiment²². Genotypes of Cbs^+/+ and Cbs^+/− mice were confirmed by performing standard PCR from digests of tail snips. The following diagnostic primers were used for genotyping according to Jackson Laboratory guidelines: a common Fwd WT allele primer 5′-GATTGCTTGCCTCCCTACTG-3′ (Jackson primer ID 12998); WT allele Rev primer 5′-AGCCAACTTAGCCCTTACCC-3′ (Jackson primer ID 12999) and mutant allele Rev primer 5′-CGTGCAATCCATCTTGTTCA-3′ (Jackson primer ID 12400). Food (standard chow) and water were provided ad libitum.

Mtb infection of mice

Male and female age-matched mice aged 8−10 weeks were infected with Mtb H37Rv either by intratracheal instillation or by an aerosol inhalation exposure system (Glas-Col, USA). Lung CFU at 24 h post infection was used to determine initial dose of infection. For time course infection studies, mice were sacrificed at 3, 8 and 14 weeks post infection. The number of viable bacteria present in lungs and spleens was enumerated by plating serial dilutions of homogenates on BBL 7H11 agar plates supplemented with carbencillin (50 µg/ml) and cycloheximide (50 µg/ml). CFU were enumerated as described above. For survival studies, 10 mice/group were used, and time-to-death was observed and represented by a Kaplan−Meier graph. Mice were monitored for survival until 300 days whereas all Cbs^+/− mice died by day 270. No deaths were observed in uninfected WT or Cbs^+/− mice within 1 year of the end of the survival experiment.

AOAA/GYY4137 treatment in mice

Mice were infected with Mtb H37Rv as described earlier by an aerosol inhalation exposure system (Glas-Col, USA). The number of lung CFU at 24 h post infection was used to determine initial dose of infection. After 2 weeks of initial infection, mice were subjected to intraperitoneal administration of AOAA (13 mg/kg/day, 6 d/wk for 6 weeks), GYY4137 (25 mg/kg/day, 3 d/wk for 6 weeks) or both. Control mice received vehicle control (PBS). Treatment started 2 weeks post-Mtb infection and lasted for 6 weeks when mice were sacrificed (n = 6). The number of lung CFU was determined as described above.

Histological analysis of mouse lung lesions

Lung lobes were fixed in 10% neutral buffered formalin for a minimum of 48 h then embedded in paraffin by standard procedures. Tissues were processed and stained at the UAB Comparative Pathology Laboratory. Briefly, paraffin sections were cut at a thickness of 5 µm with a rotary microtome. Sections were stained with hemotoxylin and eosin to score for pathology. Microscopic images were taken and NIS Elements software (Nikon) was used for semiquantitative morphometric analysis. For each slide, multiple representative, but random microscopic fields were analyzed. An experienced pathologist calculated the percentage surface area covered by lesions between Cbs^+/+ and Cbs^−/− mice. Micrographs were taken at ×10 magnification.

Cytokine analysis

Mouse blood was obtained using protocols approved by the UAB IACUC and sera isolated by standard methods. Levels of Th-1 and Th-2 cytokines were determined using a Bio-Plex Pro mouse cytokine 17-plex kit (Bio-Rad, USA), following the manufacturer’s recommendations.

Isolation of macrophages

Peritoneal macrophages were isolated from mice using aseptic technique. Briefly, mice were injected with 1 ml of 3% Brewer thioglycollate medium into the peritoneal cavity. Three days after injection, elicited macrophages were collected from the peritoneal cavity by injecting and withdrawing 5 ml of sterile PBS. Red blood cells were removed by lysis with ACK lysing buffer followed by washing with PBS pH 7.4.

Macrophage infections and AOAA/GYY4137 treatment

Mtb cells were cultured to mid-log phase (OD₆₀₀ ~ 0.8) in MB 7H9 medium, pelleted by centrifugation, washed once in PBS Tween (PBST) and resuspended in PBST. Mtb cells were passed through 27-gauge needle ten times to insure a uniform suspension for infection. Mouse peritoneal macrophages (1.5 × 10⁵) were infected at a multiplicity of infection (MOI) of 1:1 (bacilli:macrophage). Following infection, extracellular bacteria were removed by treatment with amikacin (200 µg/ml) for 2 h followed by washing. The macrophages at this stage received: AOAA alone (2 mM), GYY4137 alone (25 µM) or AOAA + GYY4137. To determine the intracellular growth of Mtb, macrophages were harvested and lysed by the addition of 0.025% SDS (Sigma, USA) at specific time points post infection. Appropriate dilutions of the lysates then were plated on MB7H11 agar. Colonies were counted after 4 weeks of incubation at 37 °C and data expressed as CFU/ml.

Immunoblotting

RAW264.7 macrophages (ATCC, USA) or murine peritoneal macrophages were cultured in complete RPMI media and were infected at an MOI of 1:1 (bacilli:macrophage) as described²⁹. Uninfected cells were used as controls. Following 6 h of Mtb infection, RAW cell monolayers were washed twice with PBS and whole cell lysate was prepared in RIPA lysis buffer (50 mM Tris/HCl, 1% NP-40, 0.25% deoxycholic acid, 150 mM NaCl, 1 mM EGTA, 1 mM sodium orthovanadate, and 1 mM sodium fluoride) containing a protease inhibitor cocktail tablet (Roche, USA) as per the manufacturer’s instructions. Total protein was quantified using bicinchoninic acid protein assay (Thermo Fisher Scientific, USA). Twenty micrograms of total protein was used for analysis and was resolved on 4−20% gradient Tris-glycine SDS-polyacrylamide gels (Bio-Rad, USA) and transferred to a poly-vinylidene fluoride membrane (Bio-Rad). Membranes were blocked with 5% non-fat dry milk in Tris-buffered-saline (TBS) with Tween (0.1%) for 1 h and then incubated with primary antibodies directed against CBS (1:1000; Santa Cruz; sc-67154), 3-MST (1:500; Abcam; ab85377), CSE (1:2500; Proteintech; 12217-1-AP) or GAPDH (1:5000; Santa Cruz; sc-25778) followed by a peroxidase-conjugated goat anti-rabbit IgG secondary antibody (1:20,000; Abcam; ab6721). Horseradish peroxidase activity was detected using an enhanced chemiluminescence detection system (GE Healthcare, USA).

Lungs isolated from Mtb-infected and uninfected age-matched mice were independently processed and were homogenized in RIPA buffer. Twenty micrograms of total protein was resolved as described above. Membranes were blocked as described above and then incubated with primary antibodies directed against CBS (1:1000; Santa Cruz; sc-67154), iNOS (1:1000; Abcam; ab3523), or GAPDH (1:5000; Santa Cruz; sc-25778), followed by a peroxidase-conjugated goat anti-rabbit IgG secondary antibody (1:20,000; Abcam; ab6721). Horseradish peroxidase activity was detected as described above. Densitometry was performed using ImageJ software and results were normalized against GAPDH expression. Uncropped, unprocessed scans of immunoblots are included in the Source Data file.

Flow cytometry

Single-cell suspensions were prepared from lungs and spleens by finely dicing the tissue followed by digestion with liberase DL (Roche, USA) at 37 °C for 30 min. RBC were removed by treatment with ACK lysis buffer followed by washing with PBS pH 7.4. Isolated cells were divided in two groups for myeloid and lymphoid cell characterization. For surface staining, cells were treated with fixable viability dye (eBiosciences, USA) for 30 min at 4 °C in the dark. Fc receptors were blocked with anti-CD16/32 for 10 min at 4 °C in the dark. Immune cell populations were identified by immunostaining using fluorescence-conjugated antibodies directed against the cell-surface markers for 30 min at 4 °C. For intracellular staining, prior to surface stain, cells were cultured in RPMI-1640 containing 5 ng/ml PMA, 500 ng/ml Ionomycin (Sigma, USA) and 10 μg/ml GolgiPlug protein transport inhibitor (BD Biosciences, USA) at 37 °C for 4 h. Cells were then harvested, washed twice and stained for cell-surface markers followed by permeabilization using a BD cytofix/cytoperm permeabilization kit (BD Biosciences) and then stained for intracellular IFN-γ. The following fluorescence-conjugated antibodies were used: Ly-6G (Gr-1) APC (clone: 1A8-Ly6g; eBioscience # 17-9668-82; 10 μl/200 μl reaction), CD11b eVolve™ 605 (clone: M1/70; eBioscience # 83-0112-42; 5 μl/200 μl reaction), F4/80 Antigen APC-eFluor® 780 (clone: BM8; eBioscience #47-4801-82; 10 μl/200 μl reaction), CD4 eVolve™ 605 (clone: RM4-5; eBioscience # 83-0042-42); CD45.2-Bv650 (clone: 104; BioLegend #109836; 10 μl/200 μl reaction); IFN-γ-APC (clone: XMG1.2; BioLegend # 505810; 10 μl/200 μl reaction). Finally, cells were washed twice followed by fixation using IC-fixation buffer (eBiosciences) before analysis. Flow cytometry acquisitions and analyses were performed using Becton Dickinson LSR II with FACS Diva 8.0 software (BD Biosciences, USA). Data were further analyzed using FlowJo 10 (Tree Star, USA).

Bioenergetic analysis using the Seahorse XF96 analyzer

OCR profiles of mycobacterial strains adhered to Cell-Tak®-coated XF cell culture microplates at 2 × 10⁶ bacilli/well were measured using a XF96 Extracellular Flux Analyzer (Agilent Technologies, USA) as described previously^29,30,31. The basal oxygen consumption (OCR) rate of bacterial cells was measured with or without GYY4137 and the shift in the OCR profile was evaluated. In assays using spermine NONOate (Cayman Chemicals, USA) and GYY4137, cells were exposed overnight (or injected during the XF96 run depending on the experiment) to spermine NONOate or GYY4137 before loading the cartridge into the XF96 instrument. Spent GYY4137 was used as a control compound. GYY4137 (final concentrations of 0, 10, 25, 50 and 100 µM) or DTT (final concentrations of 0, 1.25, 2.5, 5 and 10 mM) were also introduced through the drug ports in the XF96 plate by serial injection. Since GYY4137-generated H₂S bleeds over to adjacent wells, control wells filled with medium were used to separate experimental wells. Untreated control cells received either distilled water (UT) or spent GYY4137 (CT). Heat-killed Mtb cells were used as additional controls to rule out enzyme-mediated changes in OCR.

Bioenergetic analysis using Oroboros Oxygraph-2k

The OCR of M. bovis BCG (nonvirulent mycobacterium genetically related to Mtb) was measured using an Oroboros Oxygraph-2k respirometer (Oroborus Instruments, Austria). M. bovis BCG cells were grown as outlined above and placed in the sample chamber. Freshly prepared NaHS was administered through the injection ports and readings were recorded.

Total ATP quantitation in mycobacterial cells

Total ATP was quantified with the bioluminescence-based Enliten ATP assay system (Promega, USA) as per the manufacturer’s instructions. Briefly, 100 μl aliquots of bacterial culture were collected at designated times and immediately heat inactivated and cell debris pelleted by centrifugation as described⁷⁹. Twenty-five microliter aliquots of cleared cell lysate were transferred into 96-well assay plates and analyzed according to the manufacturer’s instructions. Luminescence was detected using the luminescence mode in a Synergy Hybrid plate reader (Biotek Instruments, USA) and was expressed as relative luminescence units (RLU). Readings were normalized against CFU counts.

ROS production in the presence of H₂S

Mtb H37Rv was grown in MB 7H9 media supplemented with 0.2% glycerol, 1× ADS (albumin, dextrose and saline) and 0.01% tyloxapol at 37 °C. ROS production in Mtb was measured using dihydroethidium (DHE; Excitation/Emission of 500/605 nm). Mtb cells were filtered through a 10 µm filter to generate a single-cell suspension, and treated with GYY4137 (0, 10, 25, 50 µM) for 1 h. Cumene hydroperoxide at 15 mM was used as positive control. DHE (10 µM) was added to the treated bacilli and incubated for a further 30 min at 37 °C. Prior to analysis, treated bacilli were washed once using fresh media to remove any residual extracellular DHE. The fluorescence of DHE stained samples was acquired using a FACS Aria III cell sorter (488 nm laser excitation, and BP 610/20 nm for emission acquisition). Cells were acquired at a constant flow rate (setting 4) and the threshold rate of approximately 1000−2000 events per second; 100,000 total events were recorded per sample. The bacterial population was identified according to the forward and side scattering property of the population (FSC versus SSC). The doublet population of bacteria was discriminated from the analysis according to the relationship of the width and area of the SSC signal pulses of Mtb. Mean fluorescent intensity was calculated using FlowJo software package.

LC-MS/MS targeted metabolic analysis

Intracellular ATP, NADH, NAD⁺, and metabolite levels were measured as described elsewhere^30,80. Mtb H37Rv cultures (OD₆₀₀ ~ 0.5−0.6) grown in MB 7H9 broth supplemented with 0.2% glucose, 0.08% NaCl, 0.5% BSA and 0.02% tyloxapol were exposed to GYY4137 under carefully controlled conditions to final concentrations of 10, 25 and 100 µM. At 1 and 24 h after exposure, the OD₆₀₀ of each culture was recorded and 1.8 ml of each culture was pelleted, 800 µl of cold chloroform: methanol at a ratio of 2:1 (containing the internal standards, 6-amininicotinic acid and mono-methyl fumarate) was added and culture lysates were stored at −80 °C until further processing. The bacterial lysate in the chloroform:methanol mixture was filtered (0.22 µm filter) and the filtrate evaporated under reduced pressure during centrifugation. The dried pellet was reconstituted in 200 µl water overnight and submitted for LC-MS/MS analysis. All experiments were carried out with an Agilent 1200 series binary HPLC system (Agilent Technologies, Germany) coupled to an Applied Biosystems/MDS Sciex QTRAP^TM 5500 linear accelerator trap mass spectrometer (AB/MDS Sciex, Canada) fitted with an ESI source. The mass spectrometer was operated in negative ion mode and samples analyzed using multi reaction monitoring (MRM). Data were acquired and analyzed using the Analyst (Version 1.6.2) software package (AB/MDS Sciex, Canada). Analytes were separated on a Waters Xterra C18 column (2.1 × 100 mm, 3.5 µm) using a gradient elution program at a flow rate of 200 µl/min and an injection volume of 5 µl. Two mobile phases were used: mobile phase A (A; 10 mM tributylamine, 15 mM acetic acid in 3% acetonitrile, 97% water) and mobile phase B (B; 10 mM tributylamine, 15 mM acetic acid in 100% acetonitrile). The gradient program was as follows: 0−1 min, 0% of A; 1−10 min, a linear increase from 0 to 100% of B; 10−15 min, 100% of B; 15−15.1 min, 0% of B; and 15.1−25 min, 0% of B. The AUC were normalized to the OD₆₀₀ (10⁸ bacilli/ml) of the sample cultures before ratios were determined.

Fractional nutrient contribution (FNC)

Mtb H37Rv was cultured in MB 7H9 media (Difco) supplemented with 10% OADC (Difco) and 0.01% Tyloxapol (Sigma-Aldrich) at 37 °C. All ¹³C-tracer carbon sources ([¹³C₆] glucose and [¹³C₂] acetate) were purchased from Sigma-Aldrich. Mtb cultures were grown till an OD600 ~0.6, pelleted and washed twice using MB 7H9 media (Difco) containing 0.01% Tyloxapol. After the final centrifugation step, the pellets were resuspended in 4.5 ml experiment media, MB 7H9 containing the one ¹³C-tracer carbon source (final concentration 0.2%) and GYY4137 to a final concentration of 0, 10, 25, 50 or 100 μM respectively. Cultures were incubated overnight (~16 h) at 37 °C, samples (4 ml) were centrifuged and pellets were kept on dry ice until further processing. The frozen pellets were resuspended 1.8 ml extraction buffer (2:2:1 methanol:acetonitrile:water) and bacilli lysed via five rounds of bead beating (7000 rpm for 60 s.). Cell lysates were sterilized using 0.2 μm membrane spin filters. Protein concentrations of bacterial lysates were measured using a Micro BCA^TM Protein Assay Kit (Thermo Scientific). Cell lysates were dried under reduced pressure and the resultant pellets were re-dissolved in water and analyzed for organic acid analysis by LC-MS. For amino acid analysis by LC-MS, the re-dissolved cell lysates were diluted with acetonitrile (50% final concentration) before analysis.

Organic acids were separated on a Biorad Aminex HPX-87 column (300 × 7.8 mm), using an aqueous 0.1% formic acid isocratic mobile phase program at a flow rate of 400 µl/min and an injection volume of 20 µl. Amino acids were separated on a Waters BEH Amide column (2.1 × 100 mm, 1.7 µm) using a gradient elution program at a flow rate of 200 µl/min and an injection volume of 1 µl. Two mobile phases were used: mobile phase A (A; aqueous 0.1% formic acid) and mobile phase B (B; acetonitrile, 0.1% formic acid). The gradient program was as follows: 0–0.2 min, 99% of B; 0.2–24 min, a curvilinear decrease from 99 to 30% of B; 24–25 min, 30% of B; 25.1–35 min, 99% of B. Both columns were connected to a Dionex Ultimate 3000 UPLC, and the total negative (organic acids) and positive (for amino acids) ion chromatograms (50–750 m/z scan range) were collected using a Thermo Scientific Q Exactive mass spectrometer. The total ion chromatograms of all relevant metabolites, and their respective isotopologues, were visualized and analyzed using Skyline V 3.7 (freely available open source software developed by MacCoss Lab, University of Washington, Seattle, WA USA). Chromatographic data were analyzed using Windows Excel 2016 and statistical analysis and data representation performed using GraphPad Prism 7.02.

FNC was calculated as previously reported³³. For a given metabolite, the fraction of ¹³C isotopologues in a specific metabolite pool is calculated as the sum of all the labeled isotopologues divided by the total abundance of the specific metabolite (all labeled and unlabeled isotopologues). For instance, in the case of glucose/fructose-6-phosphate, FNC = [(M + 1) + ⋯ + (M + 6)]/[(M + 0) + ⋯ + (M + 6)]. Thus, independent culturing of Mtb using ¹³C₆-glucose and ¹³C₂-acetate as carbon sources (¹³C-tracers) yielded FNC values of 0.463 and 0.016 respectively (Fig. 3). The FNC of “other” carbon sources utilized by the bacteria is subsequently calculated by FNC_other = 1 − (FNC_glucose + FNC_acetate) = 0.548. The same method was used to calculate the FNC for the “Other” metabolite pools. Glutamate is a likely source of FNC_other as MB 7H9 medium used in these experiments contains a substantial amount of glutamate (0.5 g/l).

Mycobacterial growth and ATP quantitation with •NO donor

The •NO donor DETA/NO (Sigma, USA) was freshly prepared according to the manufacturer’s instructions. Cultures (10 ml) receiving either DETA/NO or GYY4137 alone were sealed in sterile 60 ml PETG bottles following addition of reagents. For cultures receiving both DETA/NO (250 µM) and GYY4137, GYY4137 was added 15 min after the addition of DETA/NO and the cultures immediately sealed. For each time point, aliquots were removed to perform ATP assays and CFU analysis. Cultures were plated and CFU enumerated as described above. ATP analysis was performed as described in the earlier section.

Nitrate/nitrite estimation

Nitrate/nitrite estimation was performed on mouse sera at 8 weeks post infection, or on culture media of infected macrophages using a nitrate/nitrite fluorometric assay kit (Cayman Chemicals, USA) according to the manufacturer’s instructions. Mice used for this experiment were age-matched (n = 4–5).

Methylene blue method for sulfide measurement

The concentration of H₂S (determined as a combination of free H₂S, HS⁻ and S²⁻) in serum was measured spectrophotometrically using the methylene blue method. Briefly, serum/cultures were mixed with 0.25 ml Zn acetate (1% w/v) and 0.45 ml water for 10 min at room temperature. Trichloroacetic acid (TCA) (10%; 0.25 µl) was then added and centrifuged (14,000 × g; 10 min; 4 °C). The supernatant was collected and mixed with N, N-dimethyl-p-phenylenediamine sulfate (20 µM) in 1.2 M HCl and FeCl₃ (30 µM) in 1.2 M HCl. Absorbance was subsequently monitored at 670 nm and the concentration of H₂S (defined as above) was determined using a standard curve of freshly prepared NaHS (0–200 µM; R² = 0.9987).

Detection of H₂S using the fluorescent probe WSP-1

WSP-1 (Cayman Chemicals, USA) is a reactive fluorescent probe designed to detect H₂S in biological samples and cells, and rapidly reacts with this gas to generate benzodithiolone and a fluorophore with excitation and emission maxima 465 and 515 nm, respectively. WSP-1 was prepared according to the manufacturer’s instructions.

Hypoxia studies in mycobacteria

For hypoxia studies, 30 µl of 100 mg/ml metronidazole (MN) was added to 30 ml of actively growing bacterial culture in 60 ml PETG bottles followed by addition of 25 µM GYY4137 or spent GYY4137, sealed and kept at 37 °C without agitation. Control Mtb cultures from the same inoculum were processed similarly without MN or GYY4137. For prolonged hypoxia experiments, cells were processed using a well-established hypoxia model in Mtb⁸¹. All cultures received either 25 µM GYY4134 or spent GYY4137 just prior to sealing the culture bottles. Following hypoxia reactivation studies, post-hypoxia cultures were pelleted, washed and resuspended in fresh media. Equal numbers of cells were inoculated in fresh media and monitored by absorbance (OD₆₀₀) and CFU plating of culture aliquots at different time points. Colonies were enumerated after 4−6 weeks of incubation at 37 °C.

Oxidative stress assays

Mtb cells were grown to 0.6−0.8 OD₆₀₀ and exposed to 25 µM GYY4137 for 6 h. 2 × 10⁶ cells were transferred into each well of a 96-well plate²⁹. The oxidative stress agents menadione (MND, 0.5 mM) or cumene hydrogen peroxide (CHP, 0.5 mM) were added to the cells followed by incubation at 37 °C for 6 h. Unexposed cells received only spent GYY4137 and were processed in a separate plate. Cells were subsequently plated, and colonies were enumerated after 4−6 weeks. The percentage of surviving cells was determined relative to untreated cells.

Alamar blue assays

Mtb (200 µl of cells; OD₆₀₀ = 0.4) was incubated with various concentrations of aminooxyacetic acid (AOAA) or cysteine for 18 h, followed by incubation with Alamar blue for 24 h. Absorbance was then read at 570/600 nm. Control wells contained growth medium and AOAA or cysteine (at each respective concentration) as well as Alamar blue.

Isolation of RNA and sequencing

Mtb cultures were exposed to 25 µM GYY4137 or spent GYY4137 for 1 h under carefully controlled conditions (n = 3 per group). Total RNA was isolated from Mtb (0.6−0.8 OD₆₀₀) using Trizol as described previously²⁹. RNA purity was determined by measuring the ratio of UV absorbance at 260 and 280 nm. A parallel set of culture vials containing methylene blue was used as an indicator to ensure that hypoxia was not generated under these conditions. Ribosomal RNA depletion was performed using a Ribo-Zero kit (Illumina, USA) and the integrity of the RNA was assessed using an Agilent 2100 Bioanalyzer. mRNA sequencing was performed using an Illumina HiSeq2500 with the latest sequencing reagents and flow cells providing up to 300 Gb of sequence information per cell. Briefly, cDNA libraries were prepared using TruSeq library generation kits (Illumina USA). Library construction consisted of random fragmentation of the mRNA, followed by cDNA production using random primers. The cDNA libraries were quantitated using qPCR in a Roche LightCycler 480 with a Kapa Biosystems kit for library quantitation (Kapa Biosystems, USA) prior to cluster generation. Clusters were generated to yield approximately 725−825K clusters/mm². For each sequencing run, >10 million paired end (2 × 50 bp) reads were generated and the raw RNA-Seq fastq reads were aligned to the Mtb H37Rv genome using the short-read aligner Bowtie2 version 2.2.3. EDGE-Pro (Estimated Degree of Gene Expression in PROkaryotes) version 1.3.1 was used to estimate gene expression levels. Finally, significance and differential transcript expression was ascertained using DESeq2 version 1.10.1. Genes meeting the desired criteria (fold change ≥ ± 1.50 and q value < 0.05) were considered differentially regulated.

Isolation of RNA and quantitative RT-PCR

All mycobacterial strains (Mtb H37Rv, ΔdosS, ΔdosT and ΔdosST) were grown to mid-log phase. Cultures received 25 µM GYY4137 or spent GYY4137 for 1 h and were processed under carefully controlled conditions similar to the cultures processed for RNA seq analysis. Total RNA was isolated using RNApro (MP Biomedicals, USA) according to the manufacturer’s instructions. Extracted RNA was further treated with DNase (Thermo Scientific, USA) to remove any trace of genomic DNA. The purity of the RNA was determined by measuring the ratio of UV absorbance at 260 and 280 nm. One hundred nanograms of total RNA was used for cDNA synthesis using the iScript cDNA synthesis kit (Bio-Rad, USA). Quantitative RT-PCR was performed using SsoAdvanced SYBR green supermix (Bio-Rad, USA) with the Bio-Rad CFX96 detection system according to the manufacturer’s instructions. Quantitative RT-PCR reactions were performed in duplicate using three independent biological replicates. Relative fold changes in gene expression were determined using the 2ΔC_t method⁸²,where C_t values of target genes were normalized to C_t values of Mtb 16sRNA as an internal control. Relative fold change was determined as the ratio of expression of respective genes between untreated and GYY4137-treated cultures. Primers used in this study are listed in Supplementary Table 1.

Statistical analysis

All measurements were taken from distinct samples. Statistical computations were performed with Prism 6.01 software (GraphPad, La Jolla, California). Pairwise comparisons were performed by parametric unpaired t test. Multiple comparisons were performed using one-way or two-way ANOVA followed by Tukey’s or Sidak’s multiple comparisons tests. The log-rank (Mantel–Cox) test was used to determine median survival time.