Research ArticleTuberculosis

Targeting redox heterogeneity to counteract drug tolerance in replicating Mycobacterium tuberculosis

See allHide authors and affiliations

Science Translational Medicine  13 Nov 2019:
Vol. 11, Issue 518, eaaw6635
DOI: 10.1126/scitranslmed.aaw6635

Antibiotic redox-based redux

Phagosomal pH and redox heterogeneity in Mycobacterium tuberculosis (Mtb) can promote tolerance of the bacterium to antibiotics. Mishra et al. found that the approved antimalarial drug chloroquine inhibited this acidification and resulted in altered redox metabolism and improved susceptibility of Mtb to first-line antituberculosis drugs, particularly isoniazid, in infected macrophages in vitro. Coadministration of chloroquine improved isoniazid treatment outcomes in both mouse and guinea pig models of Mtb infection. This work suggests the repurposing of chloroquine to potentiate and possibly shorten antibiotic treatment of tuberculosis.

Abstract

The capacity of Mycobacterium tuberculosis (Mtb) to tolerate multiple antibiotics represents a major problem in tuberculosis (TB) management. Heterogeneity in Mtb populations is one of the factors that drives antibiotic tolerance during infection. However, the mechanisms underpinning this variation in bacterial population remain poorly understood. Here, we show that phagosomal acidification alters the redox physiology of Mtb to generate a population of replicating bacteria that display drug tolerance during infection. RNA sequencing of this redox-altered population revealed the involvement of iron-sulfur (Fe-S) cluster biogenesis, hydrogen sulfide (H2S) gas, and drug efflux pumps in antibiotic tolerance. The fraction of the pH- and redox-dependent tolerant population increased when Mtb infected macrophages with actively replicating HIV-1, suggesting that redox heterogeneity could contribute to high rates of TB therapy failure during HIV-TB coinfection. Pharmacological inhibition of phagosomal acidification by the antimalarial drug chloroquine (CQ) eradicated drug-tolerant Mtb, ameliorated lung pathology, and reduced postchemotherapeutic relapse in in vivo models. The pharmacological profile of CQ (Cmax and AUClast) exhibited no major drug-drug interaction when coadministered with first line anti-TB drugs in mice. Our data establish a link between phagosomal pH, redox metabolism, and drug tolerance in replicating Mtb and suggest repositioning of CQ to shorten TB therapy and achieve a relapse-free cure.

INTRODUCTION

An unusually long-term (6 months) therapy involving multiple antibiotics is required to cure tuberculosis (TB) in humans. This protracted treatment is necessary to prevent relapses due to genetically drug-sensitive bacteria that become transiently resistant inside host cells and tissues, a phenomenon called phenotypic drug tolerance. Thus, the mechanistic basis of phenotypic drug tolerance needs to be studied to develop new drugs with treatment-shortening properties. Recent studies indicate that heterogeneity in both the host environment and the bacterial population can promote phenotypic drug tolerance. For example, variability in the activation status of macrophages distinctly modulates drug tolerance in Mycobacterium tuberculosis (Mtb) (1). Immune activation of macrophages leads to release of antibacterial effectors such as reactive nitrogen species and reactive oxygen species (ROS) (2, 3), leading to a quiescent drug-tolerant state of Mtb (4). In support of this theme, drug tolerance is diminished in mice and macrophages deficient in producing nitric oxide (NO) (1). Moreover, extracellular Mtb present in the cavity caseum derived from Mtb-infected rabbits show slow replication and extreme tolerance to several first- and second-line anti-TB drugs (5). Single-cell measurements have revealed that stress conditions (for example, starvation) in vitro and host immune pressures (interferon-γ, a cytokine critical for anti-TB host immunity) in vivo create phenotypic heterogeneity within the Mtb population, which allows for the selection of nongrowing metabolically active bacteria responsible for postchemotherapeutic relapse (4).

However, recent studies suggest that adoption of a nongrowing state is not a prerequisite for drug tolerance (610). A fraction of both replicating and nonreplicating bacteria shows regrowth after drug withdrawal (4, 7), emphasizing that growth-arrested bacteria do not solely mediate tolerance. Alternate mechanisms—such as induction of drug efflux pumps, asymmetric cell division, and increased mistranslation rates—can contribute to substantial drug tolerance in actively multiplying cells (6, 8, 9, 11). Induction of efflux pumps is, so far, the only mechanism known to confer drug tolerance in replicating Mtb inside macrophages (6). Despite their importance, we lack understanding of macrophage-specific cue(s) and associated changes in the physiology of replicating Mtb that drive drug tolerance. Filling this knowledge gap will help in developing strategies to target both bacterial and host determinants crucial for mobilizing a drug-tolerant phenotype in vivo. A detailed summary of our current understanding of phenotypic drug tolerance in Mtb is described in fig. S1.

Using a ratiometric fluorescence biosensor (Mrx1-roGFP2) of the major mycobacterial antioxidant mycothiol (MSH), we previously showed that the environment inside macrophages rapidly generates heterogeneity in the MSH redox potential (EMSH) of the Mtb population (12). Confocal and flow cytometry measurements categorized infected macrophages into two distinct populations, one predominantly harboring EMSH-reduced bacteria (−300 ± 6 mV) and the other predominantly harboring EMSH-basal bacteria (−275 ± 5 mV) (12). In addition, a minor fraction of infected macrophages was enriched for Mtb in the EMSH-oxidized state (−242 ± 6 mV) (12). These results are consistent with the heterogeneous and dynamic nature of both host and pathogen (1315), suggesting that their interaction is likely to result in bacterial populations with diverse phenotypes. The EMSH-reduced population was found to be refractory to anti-TB drugs compared to other populations (EMSH-oxidized and EMSH-basal) (12). Therefore, understanding the basis of redox heterogeneity could inform strategies that result in better targeting of drug-tolerant Mtb. In this study, we performed RNA sequencing (RNA-seq) of redox-altered intraphagosomal Mtb populations and identified bacterial factors and host cues associated with drug tolerance.

RESULTS

Transcriptional profiling of redox-diverse populations identifies determinants of drug tolerance

We followed our previously developed flow cytometry protocol that averages median fluorescence ratio (405 nm/488 nm) of the Mrx1-roGFP2 biosensor expressed by intraphagosomal Mtb to gate macrophages into fractions enriched with either EMSH-reduced (EMSH = −300 ± 6 mV) or EMSH-basal (EMSH = −275 ± 5 mV) bacteria (fig. S2) (12). Using this gating strategy, we sorted THP-1 macrophages infected with Mtb/Mrx1-roGFP2 at 24 hours postinfection (p.i.), treated them with isoniazid [Inh; threefold the in vitro minimal inhibitory concentration (MIC)] for 48 hours, and confirmed that the EMSH-reduced fraction is more tolerant to Inh than EMSH-basal fraction (fig. S3). As reduced susceptibility to Inh originated from a population of intraphagosomal Mtb, our findings align with a recent consensus statement defining tolerance as the general ability of a population to survive longer antibiotic exposure (16). To investigate the physiological basis of the differential tolerance of redox-altered Mtb, we performed global transcriptional profiling by RNA-seq of EMSH-reduced and EMSH-basal populations derived from THP-1 macrophages flow-sorted 24 hours after infection with Mtb/Mrx1-roGFP2 (Fig. 1).

Fig. 1 RNA-seq of intraphagosomal Mtb derived from EMSH-reduced and EMSH-basal fractions.

(A) Schematic depiction of flow sorting–coupled RNA-seq of intraphagosomal bacteria present in EMSH-basal and EMSH-reduced fractions of THP-1 macrophages infected with Mtb/Mrx1-roGFP2. Mtb cells (optical density at 600 nm, 0.4) harvested and resuspended in RPMI for 24 hours were used as an in vitro control. FACS, fluorescence-activated cell sorting; GTC, guanidinium thiocyanate. (B) Scatter plot indicates relative distribution of differentially expressed genes (DEGs) from the EMSH-reduced and EMSH-basal fractions on the basis of log2 fold changes (FC) (blue, DEGs specific to EMSH-reduced; red, DEGs unique to EMSH-basal; black, DEGs common to both; gray, nonsignificant genes). (C) The table summarizes the transcriptional overlap between this study and the response of Mtb under intramacrophage and pH stress conditions. Fisher’s exact test with P < 0.05 as a cutoff for significance. ns, no significant difference. (D) Heat maps indicate log2 fold changes of DEGs belonging to various functional categories (obtained from Mycobrowser, École polytechnique fédérale de Lausanne) in EMSH-reduced and EMSH-basal fractions. Genes were considered differentially expressed on the basis of the false discovery rate (FDR) of ≤0.05 and absolute fold change of ≥1.5 (tables S1 and S3).

We isolated total bacterial RNA, performed RNA-seq, and analyzed data using DESeq2 (Fig. 1A). Control RNA was isolated from logarithmically grown Mtb resuspended in complete RPMI for 24 hours (in vitro control). We compared the transcription profiles of macrophage-derived populations to in vitro control and to one another to identify responses that were induced in both populations and responses that were significantly induced in EMSH-reduced bacteria [false discovery rate (FDR), ≤0.05]. Principal components analysis and clustering of heat map plots showed that the three samples clustered by their biological replicates (fig. S4 and tables S1, A to C). As compared to the in vitro control, the expression of 560 and 617 genes was affected in the EMSH-reduced and EMSH-basal populations, respectively (FDR, ≤0.05; fold change, ≥1.5) (table S1, B and C). Of 295 genes showing overlap, 151 were more induced in the drug-tolerant EMSH-reduced population (Fig. 1B and table S1D). The transcriptome of both populations considerably overlapped with that of a previously reported transcriptional signature of intraphagosomal Mtb (P < 0.05, Fisher’s exact test; Fig. 1C) (17). Consistent with studies showing phagosomal acidification as the earliest cue that alters the transcriptome of Mtb inside unstimulated macrophages (18, 19), RNA-seq data of the EMSH-reduced fraction overlapped significantly with the transcriptome of Mtb grown in 7H9 broth at pH 5.5 (P = 1.05 × 10−2) and pH 4.5 (P = 1.5 × 10−15) (Fig. 1C) (17, 20). The EMSH-basal transcriptome showed little similarity to the genes down-regulated at pH 5.5 (Fig. 1C) (17).

Using Mtb/Mrx1-roGFP2 conjugated with pHrodo dye, which emits fluorescence only in acidic pH (21), we infected THP-1 macrophages and confirmed phagosomal pH to be 6.25 ± 0.14, as previously reported for unstimulated macrophages (fig. S5) (22). Likewise, we examined the pH of flow-sorted macrophages enriched with EMSH-basal or EMSH-reduced bacteria. Macrophages enriched in EMSH-reduced bacteria are more acidic (pH 5.79 ± 0.2) than EMSH-basal bacteria (pH 6.67 ± 0.08) (fig. S5D), indicating that subtle variations in phagosomal pH underlie heterogeneity in EMSH of Mtb during infection. We also compared the transcriptional profiles of EMSH-altered Mtb and a mutant of the redox-sensitive transcription factor WhiB3 (MtbΔwhiB3), which has been linked to Mtb’s transcriptional response to low pH (20). We isolated RNA from wild-type (WT) Mtb, MtbΔwhiB3, and whiB3-complement (whiB3-Comp) cultured in 7H9 broth at neutral pH (6.6) and acidic pH (4.5) and performed RNA-seq. The WhiB3-specific, low pH–induced gene set showed comparable expression in the EMSH-reduced population. In contrast, only a fraction of the WhiB3 regulon coincided with the transcriptome of EMSH-basal bacteria, with lesser induction of the regulon than that seen in the case of the EMSH-reduced Mtb (fig. S6 and table S2D). These findings link the transcriptome of EMSH-reduced bacteria with Mtb’s response to low pH.

Low pH increases the solubility of transition metals including iron and copper (23), thereby allowing these metals to cross biological membranes and participate in metal-catalyzed ROS generation via the Fenton reaction (24). Consistent with this phenomenon, transcriptional sensors of metal toxicity (furA, csoR, and cmtR), exporters of toxic metals (ctpV and ctpG), redox sensors (whiBs and sufR), and antioxidant systems (rubA/B, ahpC, and trxB2) were induced in the EMSH-reduced fraction (Fig. 1D and table S3). Because ROS damages DNA, proteins, and lipids, we observed that several genes implicated in DNA repair, protein quality control, and envelope stress were induced in EMSH-reduced fraction (Fig. 1D and table S3). A previous study linked the stochastic expression of catalase (katG) in mediating Inh tolerance (7). However, katG expression was not differentially regulated in the EMSH-reduced fraction, indicating that redox-mediated Inh tolerance is unrelated to katG expression. Genes coordinating glyoxylate and methyl citrate cycles (icl1 and prpD) and alternate respiration (cydAB) were up-regulated in the EMSH-reduced fraction (Fig. 1D and table S3); both icl1 and cydAB are vital for mitigating oxidative stress and promoting multidrug tolerance in Mtb (2527). The overlap between acidic pH and oxidative stress responses has been reported in several bacteria (2830), indicating a link between pH- and oxidative stress–driven adaptations.

A reductive shift in the EMSH of Mtb indicates an increase in the cytoplasmic pool of reduced MSH. Supporting this, genes associated with the biogenesis of cysteine (CySH), a component of MSH (31), were up-regulated in the EMSH-reduced fraction. For example, lat, metA, and metC involved in methionine (Met) biogenesis (32), and metB encoding a bifunctional enzyme (cystathionine-γ-lyase/cystathionine-γ-synthase) that incorporates sulfur from Met to CySH (reverse transsulfuration pathway) (33) were induced in the EMSH-reduced fraction (Fig. 1D). The enzyme MetB generates H2S gas as a by-product (33), which protects several bacterial species from antibiotics and oxidative stress (3436). We detected that Mtb cultured in 7H9 broth at pH 6.2 and pH 4.5 generated more H2S than at neutral pH (fig. S7), indicating a link between H2S biogenesis, acid stress, and reductive shift in EMSH. CySH also serves as a source of sulfide for the biogenesis of Fe-S clusters, which modulate bacterial response to antibiotics (37). Accordingly, genes involved in Fe-S cluster biogenesis [Rv1460 (sufR) and Rv1461 (sufB)] (38, 39) were up-regulated in the EMSH-reduced fraction (Fig. 1D and table S3). Other transcriptional changes in EMSH-reduced cells involving genes that are known to promote drug refractoriness include genes associated with S-adenosyl methionine (SAM) biosynthesis (metK) (32), methyl transferases (menH, Rv0560c, Rv1403c, and Rv1405c) (4042), and drug efflux pumps (mmr, Rv1258c, and Rv1250) (6, 43, 44) (Fig. 1D and table S3). In conclusion, the RNA-seq data suggest a major role of host acidification and bacterial mechanisms involved in alleviating metal toxicity, ROS remediation, and realignment of sulfur metabolism in the emergence of drug-tolerant EMSH-reduced population during infection.

CySH-disposal pathways coordinate redox-mediated drug tolerance in Mtb

CySH-dependent pathways such as biogenesis of H2S, low molecular–weight thiols, and Fe-S clusters protect bacteria against antibiotics and oxidative stress (37, 45). In Mtb, supplementation with H2S donor [sodium hydrosulfide NaHS)] restored the imbalance in the EMSH of MSH recycling mutants (46), and Fe-S cluster–dependent regulators (for example, WhiB3 and WhiB7) mediate a reductive shift in EMSH of Mtb in response to acidic pH and antibiotics (20, 47). On this basis, we reasoned that the induction of metB (H2S biogenesis) and sufR (regulator of Fe-S cluster biogenesis) could contribute to the emergence of a drug-tolerant EMSH-reduced population (Fig. 2A). To test this idea, we independently disrupted metB (MtbΔmetB) and sufR (MtbΔsufR) in Mtb H37Rv (fig. S8, A to E). As expected, MtbΔmetB displayed a reduced capability to produce H2S compared to WT Mtb (fig. S8F). Similarly, disruption of sufR abrogated the induction of the suf operon (Rv1461 to Rv1466) involved in Fe-S cluster biogenesis (fig. S8G). Because WhiB3 promotes the emergence of EMSH-reduced population inside macrophages (20), we used the MtbΔwhiB3 strain as a control. THP-1 macrophages infected with MtbΔmetB or MtbΔsufR expressing Mrx1-roGFP2 showed a significant decrease in the reductive-EMSH fraction compared to WT Mtb (P < 0.01) (Fig. 2B). We next examined the influence of these pathways on Inh tolerance during infection. THP-1 macrophages infected for 24 hours with Mtb strains were exposed to 3× in vitro MIC of Inh, and survival was determined at 48 hours of Inh treatment. Intramacrophage growth of the mutants was marginally reduced compared to WT Mtb (Fig. 2C). However, upon Inh treatment, MtbΔmetB, MtbΔsufR, and MtbΔwhiB3 displayed 8.75-, 7-, and 9-fold reductions in survival compared to WT Mtb, respectively. Decreased tolerance displayed by these mutants was restored in the complemented strains to a degree similar to WT (Fig. 2C).

Fig. 2 Cysteine utilization pathways promote redox heterogeneity and drug tolerance in Mtb.

(A) Various cysteine (CySH) utilization pathways in Mtb. Expression of genes (blue) coordinating CySH flux into pathways for mycothiol (MSH), Fe-S cluster, and H2S biogenesis is induced in the EMSH-reduced fraction. (B) THP-1 macrophages were infected for 24 hours with the indicated strains of Mtb expressing Mrx1-roGFP2, and the percent distribution of redox-diverse fractions was measured at 24 hours p.i. and depicted as a stacked bar plot. **P < 0.01, by Mann-Whitney test compares EMSH-reduced fraction in various strains of Mtb with WT Mtb. (C) THP-1 macrophages infected for 24 hours with the indicated strains of Mtb were exposed for an additional 48 hours to Inh (2.18 μM, 3× of in vitro MIC) or left untreated. Bacillary load was determined by CFU enumeration, and percent survival was quantified by normalizing the CFU in Inh-treated samples at 48 hours against untreated samples (UT) at 0 hours. *P < 0.05, **P < 0.01, #P < 0.05, and ##P < 0.01, by Mann-Whitney test. Number signs (#) and asterisks (*) compare survival between WT Mtb and other strains under UT and Inh-treated conditions, respectively. (D and E) Indicated strains of Mtb grown in 7H9-tyloxapol broth acidified to pH 6.2 or pH 4.5 were exposed to Inh (7.25 μM, 10× of in vitro MIC) or kept unexposed. Bacterial load was quantified after 5 days of treatment by CFU enumeration, and percent survival was quantified strain-wise by normalizing the bacterial load in Inh-treated samples at day 5 against untreated samples. *P < 0.05, **P < 0.001, by Mann-Whitney test. Asterisks compare survival between WT Mtb and other strains after 5 days of Inh treatment. Data shown in each panel are the results of three independent experiments performed in triplicate (means ± SD). ns, no significant difference (P > 0.05).

It can be argued that the loss of MetB and SufR functions can profoundly affect the normal growth and metabolism of Mtb, complicating any association with redox heterogeneity and drug tolerance. However, we found that the growth of MtbΔmetB and MtbΔsufR in 7H9 broth was not different from WT Mtb (fig. S9, A and B). Furthermore, EMSH of MtbΔmetB (−280 ± 3 mV) and MtbΔsufR (−276 ± 4 mV) remained comparable to WT Mtb (−275 ± 2 mV) in vitro. In addition, metB or sufR disruption did not perturb the oxygen consumption rate or extracellular acidification rate (fig. S9, C and D), quantifiable readouts of oxidative phosphorylation and glycolysis, respectively (48). Last, we examined whether metB and sufR influenced tolerance to Inh under acidic pH. The MIC of Inh remained comparable (0.06 to 0.1 μg/ml) for MtbΔmetB and MtbΔsufR at neutral pH. However, at pH 4.5, WT Mtb showed 79 ± 3.97% survival to 10× MIC of Inh as compared to 19.79 ± 0.58% and 18.8 ± 1.85% in the case of MtbΔmetB and MtbΔsufR, respectively (Fig. 2, D and E). These data indicate that metB and sufR influence drug tolerance in the context of acidic pH and the intramacrophage milieu. In sum, the genetic data support our RNA-seq findings indicating CySH flux as an important mechanism underlying redox diversity and drug tolerance in Mtb during infection.

Phagosomal acidification is required for the redox-dependent multidrug tolerance of Mtb

To clarify the link between phagosomal pH, redox heterogeneity, and drug tolerance during infection, we pretreated THP-1 macrophages with nontoxic doses of well-established inhibitors of phagosomal acidification [bafilomycin A1 (BafA1), ammonium chloride (NH4Cl), and chloroquine (CQ)], infected them with Mtb/Mrx1-roGFP2, and measured EMSH (12, 20, 4951). Pretreatment with BafA1/NH4Cl/CQ uniformly diminished the fraction of Mtb displaying reductive EMSH at 24 hours p.i. (Fig. 3A). Next, we examined whether phagosomal pH enhanced drug tolerance during infection. THP-1 macrophages with or without BafA1 pretreatment were infected with Mtb for 24 hours and exposed to 3× MIC of Inh for an additional 48 hours before lysis and enumeration of viable counts. The addition of BafA1 further increased Inh-mediated killing of Mtb by fivefold (Fig. 3B). A similar increase in killing was observed upon substitution of Inh with rifampicin (Rif) or BafA1 with CQ (Fig. 3, B and C). We noted that although CQ and BafA1 uniformly increased killing efficacy of Inh and Rif, the effect was more pronounced in the case of Inh (Fig. 3, B and C). Consistent with findings in THP-1 cells, infection of peritoneal macrophages from BALB/c mice also led to a pH-dependent increase in the EMSH-reduced fraction and Inh tolerance in Mtb (fig. S10, A and B). These results suggest that phagosomal pH is a potent enhancer of multidrug tolerance in Mtb.

Fig. 3 Phagosomal pH is required for the redox-dependent multidrug tolerance of Mtb.

(A) THP-1 macrophages—untreated or pretreated with 10 nM BafA1, 10 mM NH4Cl, or 10 μM CQ—were infected with Mtb/Mrx1-roGFP2, and percent distribution of redox-diverse fractions was measured at 24 hours p.i. **P < 0.01, by Mann-Whitney test to compare the EMSH-reduced fraction with untreated sample. (B and C) THP-1 macrophages, untreated or pretreated with 10 nM BafA1 or 10 μM CQ, were infected with WT Mtb for 24 hours and exposed to Inh (2.18 μM) or Rif (1 μM) or left unexposed for an additional 48 hours. Percent survival was quantified by normalizing the CFU in drug-treated samples at 48 hours against untreated samples at 0 hours. *P < 0.05, **P < 0.01, by Mann-Whitney test. (D) THP-1 macrophages were infected with Mtb/Mrx1-roGFP2, and EMSH was measured at 24 hours p.i. After this, intraphagosomal bacteria were released and incubated in 7H9-albumen-dextrose-sodium chloride for 2 hours, and EMSH was determined. The 7H9-ADS–adapted Mtb was used to reinfect fresh THP-1 macrophages, with or without pretreatment with 10 μM CQ, and EMSH was measured at 24 hours p.i. **P < 0.01, ###P < 0.001, by Mann-Whitney test. Number signs (#) compare EMSH-reduced fractions between intramacrophage and 7H9-ADS–adapted Mtb. Asterisks (*) compare EMSH-reduced fractions between untreated and CQ-treated samples. (E) THP-1 macrophages harboring EMSH-reduced and EMSH-basal bacteria were flow-sorted at 24 hours p.i. (cycle 1 infection), and bacteria were released into 7H9-ADS. At 24-hour incubation, 7H9-ADS–adapted Mtb were used to infect THP-1 macrophages, with or without pretreatment with 10 μM CQ, for 24 hours (cycle 2 infection), and Inh tolerance was determined as mentioned earlier. *P < 0.05, **P < 0.01, by Mann-Whitney test. Data shown in each panel are the results of three independent experiments performed in triplicate (means ± SD). ns, no significant difference (P > 0.05).

We next examined whether the drug tolerance and redox heterogeneity displayed by intraphagosomal Mtb were a reversible phenotypic change or a stable genetic variation. We flow-sorted THP-1 macrophages infected with Mtb/Mrx1-roGFP2 for 24 hours into EMSH-reduced and EMSH-basal fractions, lysed the macrophages in 7H9 broth, and measured EMSH of the released Mtb. Incubation in 7H9 broth resulted in the loss of redox heterogeneity within 2 hours, indicating that the macrophage environment supports the emergence of drug-tolerant EMSH-reduced population (Fig. 3D). When these redox-homogeneous bacteria were used to reinfect THP-1 macrophages with or without CQ pretreatment, both the heterogeneity in EMSH and tolerance to Inh returned in untreated macrophages but not in CQ-treated macrophages (Fig. 3, D and E). Last, the bacteria that survived Inh treatment inside THP-1 macrophages showed an MIC comparable to the parental strain in 7H9 broth (fig. S10C). We conclude that pH- and redox-dependent tolerance of Mtb inside macrophages was due to reversible phenotypic changes rather than stable genetic mutations.

Phagosomal pH and redox heterogeneity drive drug tolerance during HIV-TB coinfection

Limited acidification is one of the hallmarks of Mtb-containing alveolar macrophages derived from HIV-TB–coinfected patients (52). We reasoned that pH- and redox-driven tolerance to anti-TB drugs could contribute to the lower TB treatment success rates commonly observed in HIV-TB–coinfected patients (53). To test this idea, we used the U1 monocytic cell line model of HIV-TB coinfection (54). U1 cells are derived from U937 monocytes wherein two copies of the HIV-1 genome are integrated and viral replication can be induced by phorbol 12-myristate 13-acetate (PMA) and tumor necrosis factor–α (55, 56). We confirmed viral replication by monitoring the expression of the HIV-1 gag transcript by quantitative reverse transcription polymerase chain reaction (qRT-PCR). Treatment with PMA (5 ng/ml) induced HIV-1 replication in a time-dependent manner in U1 cells (Fig. 4A). Next, we infected PMA-treated U1 and U937 (uninfected HIV-1 control) with Mtb/Mrx1-roGFP2 and measured heterogeneity in EMSH. Both U1 and U937 macrophages showed the emergence of redox-diverse fractions upon infection. However, a marked increase in the EMSH-reduced fraction was clearly evident in U1 compared to U937 macrophages (Fig. 4, B and C), indicating that HIV-1 replication is accompanied with the rise of EMSH-reduced population. Furthermore, treatment with BafA1, NH4Cl, and CQ uniformly decreased the EMSH-reduced fraction in U1 (Fig. 4D), confirming the role of phagosomal pH in the emergence of redox heterogeneity. Last, we tested Inh tolerance in U1 and U937 cells, as described earlier. Consistent with the increased EMSH-reduced fraction, a significantly higher proportion of Mtb tolerates exposure to 3× MIC of Inh in U1 (65.9 ± 11.67%) versus U937 (21.93 ± 0.42%) (P = 0.0015) (Fig. 4, E and F). As expected, pretreatment with CQ/BafA1 increased Inh-mediated killing of Mtb in U1 and U937 (Fig. 4, E and F). These results suggest that the phagosomal acidification encountered by Mtb inside HIV-Mtb–coinfected macrophages facilitates the development of a redox-altered drug-tolerant population.

Fig. 4 Phagosomal pH and redox heterogeneity drive drug tolerance during HIV-TB coinfection.

(A) The course of HIV-1 replication upon stimulation of the U1 promonocytic cell line with PMA (5 ng/ml). Viral load was monitored by gag qRT-PCR. **P < 0.01, by Mann-Whitney test comparing gag expression with 0 hours. U937 (uninfected HIV-1 control) (B) and U1 macrophages (C) were stimulated with PMA and infected with Mtb/Mrx1-roGFP2, and percent distribution of redox-diverse fractions was measured over time. *P < 0.05, by Mann-Whitney test. Asterisks (*) compare EMSH-reduced fraction at various time points with 0 hours. (D) U1 macrophages—untreated or pretreated with 10 nM BafA1, 10 mM NH4Cl, and 10 μM CQ—were infected with Mtb/Mrx1-roGFP2, and percent distribution of redox-diverse fractions was measured at 12 hours p.i. *P < 0.05, **P < 0.01, by Mann-Whitney test. Asterisks (*) compare EMSH-reduced fractions between untreated and BafA1/NH4Cl/CQ-treated samples. U937 (E) and U1 macrophages (F), untreated or pretreated with 10 μM CQ or 10 nM BafA1, were infected with WT Mtb for 12 hours and exposed to Inh (2.18 μM) or left unexposed for an additional 48 hours. Bacillary load was determined by CFU enumeration, and percent survival was quantified by normalizing the CFU in drug-treated samples at 48 hours against untreated samples at 0 hours. *P < 0.05, **P < 0.01, by Mann-Whitney test. Data shown in each panel are the results of three independent experiments performed in triplicate (means ± SD).

The drug-tolerant EMSH-reduced population is replicative inside macrophages

We next examined whether drug tolerance exhibited by the EMSH-reduced population is associated with slow replication as shown in several bacteria, including Mtb (1, 4, 57). We used an unstable replication clock plasmid, pBP10, which is uniformly lost from replicating but not from nonreplicating Mtb (58). Because the plasmid is resistant to kanamycin (Kan), its retention or loss can be easily estimated by determining colony-forming units (CFUs) on Kan-containing medium. We infected THP-1 macrophages with pBP10-containing Mtb/Mrx1-roGFP2. At 0, 24, and 72 hours p.i., 0.5 × 106 macrophages harboring an EMSH-reduced or EMSH-basal population were flow-sorted, after which, the bacteria were released and differences in replication were measured by enumerating Kanr (Kan-resistant) and Kans (Kan-sensitive) colonies. Expression of the pBP10 plasmid in Mtb/Mrx1-roGFP2 did not influence redox heterogeneity during infection (fig. S11A). The pattern of pBP10 plasmid loss indicated that both populations were replicative; however, the plasmid loss was faster over time in the EMSH-reduced population than the EMSH-basal population (Fig. 5, B and C). For example, at 72 hours p.i., only 17.8 ± 0.2% of cells retained pBP10 in the EMSH-reduced population as opposed to 61.19 ± 0.02% in the EMSH-basal population (Fig. 5, B and C). The cumulative bacterial burden, which provides the total number of living, dead, or damaged Mtb based on a mathematical model established for the clock plasmid (58), also confirmed the comparatively higher replication rate in the EMSH-reduced population (Fig. 5, B and C).

Fig. 5 The drug-tolerant EMSH-reduced population is replicative and has high efflux pump activity.

(A) Graphical depiction of Mrx1-roGFP2–coupled flow-sorting strategy to determine replication dynamics, metabolic state, and drug efflux activity in intramacrophage EMSH-reduced and EMSH-basal populations. (B and C) THP-1 macrophages were infected with pBP10-containing Mtb/Mrx1-roGFP2. At indicated time points, macrophages harboring EMSH-reduced and EMSH-basal bacteria were flow-sorted, and bacteria were released and plated in the presence or absence of kanamycin (Kan). The frequency of pBP10 loss and increase in cumulative bacterial burden (CBB) were calculated. *P < 0.05, **P < 0.01, #P < 0.05, ##P < 0.01, by Kruskal-Wallis test with Dunn’s correction over time p.i. Asterisks (*) and number signs (#) compare CFU per milliliter and percentage of pBP10 + ve bacteria over time p.i., respectively. (D) THP-1 macrophages were infected with Mtb/Mrx1-roGFP2, and at 24 hours p.i., the redox state of intraphagosomal Mtb/Mrx1-roGFP2 thiols was fixed using N-ethylmaleimide. Bacteria were released from macrophages and stained with calcein violet–AM (CV-AM). The CV-AM staining and EMSH status of Mtb cells were determined using multiparameter flow cytometric analysis. As a control, we performed CV-AM staining of Mtb grown in 7H9 broth for 24 hours at 4° and 37°C. *P < 0.05, by Mann-Whitney test. (E) THP-1 macrophages harboring EMSH-reduced and EMSH-basal bacteria were flow-sorted at 24 hours p.i., bacterial RNA was isolated, and expression of efflux pumps was measured by qRT-PCR. Expression was compared with in vitro control Mtb, and fold change was quantified after normalizing by 16S ribosomal RNA. *P < 0.05, **P < 0.01, by Mann-Whitney test for comparison with in vitro control Mtb. (F) THP-1 macrophages were infected with Mtb/Mrx1-roGFP2 bacteria preloaded with [14C]-Inh. At 24 hours p.i., macrophages harboring EMSH-reduced (Red) and EMSH-basal (Bas) bacteria were sorted and bacteria released. The relative distribution of radioactive [14C]-Inh was measured in bacterial and macrophage (MΦ) fractions. *P < 0.05, **P < 0.01, by Mann-Whitney test. Data shown in each panel are the results of two independent experiments performed in triplicate (means ± SD). ns, no significant difference (P > 0.05).

We examined the health of Mtb in the EMSH-reduced and EMSH-basal fractions using a fluorogenic cell-permeable dye calcein violet–acetoxy-methyl ester (CV-AM), an established metabolic indicator (59, 60). THP-1 macrophages infected for 24 hours with Mtb/Mrx1-roGFP2 were flow-sorted, and then, bacteria were released from the EMSH-reduced and EMSH-basal fractions and stained with CV-AM. Bacilli in the EMSH-reduced (91.5 ± 0.07%) and EMSH-basal fractions (99.1 ± 0.14%) showed strong CV-AM fluorescence, indicating healthy metabolic activity (Fig. 5D). As expected, metabolically active Mtb cultured in 7H9 broth at 37°C showed 91.3 ± 0.99% CV-AM staining as compared to negligible staining in bacteriostatic cells incubated at 4°C (Fig. 5D). Treatment of Mtb/Mrx1-roGFP2 with 2 mM cell-permeable thiol-reductant dithiothreitol (DTT) induces a reductive shift in EMSH (−320 mV) without any influence on metabolism and viability in vitro (20). Consistent with this, 83.9 ± 2.97% of DTT-treated Mtb cells scored positive for CV-AM staining (Fig. 5D). Together, these results suggest that the drug-tolerant EMSH-reduced population is replicative and metabolically active inside macrophages.

Redox-diverse populations of Mtb show differential activation of efflux pumps

Induction of efflux pumps is associated with drug tolerance in replicating Mtb during infection (6). We investigated whether the drug-tolerant EMSH-reduced population exhibited variation in efflux pump activity relative to the EMSH-basal population. THP-1 macrophages infected for 24 hours with Mtb/Mrx1-roGFP2 were flow-sorted into EMSH-reduced and EMSH-basal populations, and bacterial RNA was isolated for qRT-PCR of efflux pump transcripts. As a control, we performed qRT-PCR of efflux pumps on Mtb grown in 7H9 broth. We selected efflux pumps (Rv0194, Rv1348, Rv1250, ctpV, mmr, and Rv1819c) that are induced in intraphagosomal Mtb upon exposure to anti-TB drugs (44, 6164). The transcripts of ctpV, mmr, Rv1348, and Rv1250c were enriched in the EMSH-reduced fraction (Fig. 5E). We tested pH- and redox-dependent expression of efflux pumps by examining transcripts in response to pH 6.2, pH 4.5, and 2 mM DTT in vitro. Each of these conditions uniformly induces reductive shift in EMSH of Mtb in vitro (12, 20). All of these treatments increased expression of the efflux pumps (fig. S11, B and C). As a control, we analyzed efflux pump expression in an Mtb strain lacking the antioxidant buffer MSH (MtbΔmshA); this strain maintains oxidative EMSH (>−240 mV) at both neutral and acidic pH (20, 65). The expression of pH-inducible efflux pumps was significantly down-regulated in MtbΔmshA relative to WT Mtb (P < 0.05) (fig. S11D), suggesting redox-dependent regulation of efflux pump expression in Mtb.

To clarify the association between efflux pump activity and EMSH of Mtb, we assessed the steady-state distribution of Inh in EMSH-reduced and EMSH-basal populations inside THP-1 macrophages using [14C]-labeled Inh. We infected THP-1 macrophages with Mtb/Mrx1-roGFP2 cells preloaded with [14C]-Inh (0.5 μCi/ml for 2 hours). At 24 hours p.i., equal numbers of macrophages harboring either EMSH-reduced or EMSH-basal populations were sorted using flow cytometry. We chose the 24-hour time point because bacterial load was comparable in both populations (106 CFU/ml). Infected macrophages were lysed, the bacterial (pellet) and macrophage (supernatant) fractions were separated, and [14C]-labeled Inh radioactivity was measured. The distribution of [14C]-Inh was different in macrophages containing EMSH-reduced Mtb versus EMSH-basal Mtb and in the bacteria themselves. Whereas macrophages harboring EMSH-reduced Mtb showed high counts for [14C]-Inh and corresponding lower counts remained in the bacteria, the EMSH-basal population showed an inverse drug distribution. These data indicate higher efflux from the EMSH-reduced population into macrophages (Fig. 5F). Direct comparison of [14C]-Inh counts in EMSH-reduced and EMSH-basal bacterial pellets confirmed lower accumulation of the drug in the former (Fig. 5F). In summary, our data indicate that variations in efflux pump activity can be one of the factors managing drug tolerance in the EMSH-reduced population during infection.

CQ counteracts drug tolerance and relapse in vivo

Given our findings that acidic pH promotes redox heterogeneity and enhances drug tolerance in vitro, we sought to determine the impact of pharmacological inhibition of phagosomal acidification on drug tolerance in vivo. We used the antimalarial drug CQ, which deacidifies endosomes and lysosomes (66, 67), to test Mtb’s response to Inh in a chronic murine model of infection (Fig. 6A) (68).

Fig. 6 CQ counteracts drug tolerance and reduces relapse in vivo.

(A) Strategy to investigate the efficacy of CQ in reducing tolerance against Inh and Rif and posttherapeutic relapse in vivo. BALB/c mice (n = 6) were given an aerosol challenge with WT Mtb. From 4 weeks p.i. onward, groups of mice were left untreated or treated with anti-TB drugs (Inh/Rif) alone or in combination with CQ (CQ + Inh/ CQ + Rif). (B and C) Bacterial CFUs were measured in the lungs at the indicated time points. **P < 0.01, ****P < 0.0001, by Kruskal-Wallis test with Dunn’s correction across experimental groups at 12 weeks p.i. (D) Gross pathology of the lungs of WT Mtb–infected mice at 8 weeks of treatment across experimental groups. (E) Hematoxylin and eosin–stained lung sections (8 weeks of treatment) from mice infected with WT Mtb across experimental groups. The pathology sections show granuloma (G), alveolar space (AS), and bronchiole lumen (BL). All images were taken at ×40 magnification. Scale bars, 200 μm. (F) Outbred Hartley guinea pigs (n = 6) were given aerosol challenge with WT Mtb, and efficacy of CQ in reducing Inh tolerance was assessed as described in (B) and (C). **P < 0.01, ****P < 0.0001, by Kruskal-Wallis test with Dunn’s correction across experimental groups at 12 weeks p.i. (G) Hematoxylin and eosin–stained lung sections (8 weeks of treatment) from guinea pigs infected with WT Mtb across experimental groups. The pathology sections show granuloma (G), alveolar space (AS), and necrotic core (N). All images were taken at ×40 magnification. Scale bars, 200 μm. (H) Dexamethasone-induced reactivation of Mtb from the lungs of BALB/c mice (n = 5) after treatment with Inh alone or a combination of CQ plus Inh. Mann-Whitney test was used to compare the relapse frequency (Inh alone versus CQ + Inh combination) for effectiveness of CQ therapy (P = 0.0069). Data shown in each panel are the results of two independent experiments (means ± SD). ns, no significant difference (P > 0.05).

We treated chronically infected BALB/c mice (4 weeks p.i.) with Inh (25 mg/kg body weight), CQ (10 mg/kg body weight), or Inh along with CQ. After 2 and 8 weeks of therapy, we harvested the lungs and quantified the recovered bacteria. As reported (68), Inh monotherapy reduced the bacterial load from 106 to 104 at 2 weeks (P = 0.00012) and 103 per lung at 8 weeks (P = 0.0022) of treatment (Fig. 6B). CQ treatment alone showed no effect on bacterial viability over time (Fig. 6B). Relative to the control regimen (Inh alone), the addition of CQ did not alter lung CFUs after 2 weeks of treatment (Fig. 6B). However, 8 weeks of treatment with a combination of CQ with Inh (CQ plus Inh) completely sterilized the lungs of mice compared to 103 CFUs in the animals treated with Inh alone (Fig. 6B). The gross and histopathological changes observed in the lungs after 8 weeks of therapy were proportionate with the bacillary load observed (Fig. 6, D and E, and fig. S12A). After 8 weeks of treatment, the extent of pulmonary tissue destruction was highest in the untreated (score, 4) and CQ-treated animals (score, 3), intermediate in the case of the Inh-treated animals (score, 2), and negligible in CQ plus Inh–treated animals (score, 1 or 0). We also examined whether adjunct therapy with CQ for 8 weeks increased the efficacy of Rif (10 mg/kg body weight). The addition of CQ substantially reduced the fraction of Rif-tolerant Mtb in animal lungs (P = 0.021 for Rif alone versus a combination of CQ with Rif) (Fig. 6, C to E, and fig. S12A). However, the influence of CQ in reducing tolerance was more notable in the case of Inh as compared to Rif.

Because the pathophysiology of human TB is more closely recapitulated in guinea pigs (69), we aerosol-infected outbred Hartley guinea pigs with Mtb, followed 4 weeks later by treatment with Inh, CQ, or CQ plus Inh for an additional 8 weeks, and then estimated the lung bacillary load. The bacterial burden in the lungs of guinea pigs was 103 CFUs in Inh-treated animals and 105 CFUs in CQ-treated animals, compared to 100 CFUs in CQ plus Inh–treated animals (Fig. 6F). The effectiveness of CQ plus Inh was also reflected in the lung histopathology of guinea pigs (Fig. 6G and fig. S12B). Studying relapse can be another predictor of therapeutic efficacy in TB. Therefore, we aerosol-infected mice with WT Mtb, followed 4 weeks later by treatment of infected animals with Inh or CQ plus Inh for 8 weeks. As shown earlier, 8 weeks of CQ plus Inh treatment completely sterilized mouse lungs. At 20 weeks p.i., which was 8 weeks after completion of therapy, mice received immunosuppressant dexamethasone (10 mg/kg body weight) for 2 weeks, and the lung bacillary load was determined at 22 weeks p.i. Relapse of disease was observed in five of five Inh-treated mice with bacterial loads of 2 × 104 CFUs in the lungs. Only three of five CQ plus Inh–treated mice relapsed, with only 30 CFUs in the lungs (P = 0.0069) (Fig. 6H).

Although in vitro studies indicate that CQ mainly exerts its influence on intracellular Mtb by raising the vacuolar pH (70), CQ can interfere with other cellular processes such as DNA synthesis, generation of ROS, and necrosis (71, 72). Therefore, we questioned whether the effect of CQ in reducing tolerance was associated with pH alkalization in vivo. Using Magic Red cathepsin B substrate that fluoresces only upon cleavage by cathepsin B protease inside acidic lysosomes (73), we confirmed that 6 weeks of CQ treatment raised the vacuolar pH of macrophages isolated from the lungs of mice chronically infected with Mtb (fig. S13, A and B). Other antibacterial mechanisms such as ROS production and necrosis were not stimulated in macrophages derived from the lungs of mice chronically infected with Mtb after 6 weeks of treatment with CQ plus Inh as compared to Inh or CQ alone (fig. S13, A, C, and D). Together, these results confirm that adjunct therapy with CQ counteracts drug tolerance and reduces disease relapse.

CQ exhibits no adverse interaction with anti-TB drugs

Excellent oral bioavailability, oral human pharmacokinetics (half-life of 10 to 15 days), high tissue penetration, and years of clinical use in humans (74) make CQ a good candidate for developing new therapeutic combinations for the treatment of TB. We investigated the pharmacological compatibility of CQ by measuring its potential drug-drug interactions with first-line anti-TB drugs (Inh or H, Rif or R, Emb or E, and Pza or Z) given as a combination. A single-dose pharmacokinetic interaction test was performed by administering anti-TB drugs with and without CQ [10 mg/kg body weight, intraperitoneally (i.p.)] in mice. Another group of mice was dosed with CQ (10 mg/kg body weight, i.p.) to compare the pharmacokinetic behavior of CQ in the presence of combination therapy (Fig. 7A). Plasma samples were analyzed for individual drugs using liquid chromatography–mass spectrometry, and key pharmacokinetic parameters such as maximum plasma concentration (Cmax) and area under the plasma concentration-time curve (AUClast) were calculated as a ratio for combination versus single-treatment groups. Pharmacokinetic profiles revealed no adverse drug-drug interactions when CQ was coadministered with HREZ (Fig. 7, B and G). Cmax and AUClast for CQ alone was 228.5 ng/ml and 1358.0 ng·hour/ml and 297.2 ng/ml and 1358.0 ng·hour/ml for HREZ, respectively (Fig. 7, B and G).

Fig. 7 CQ exhibits no adverse interactions with anti-TB drugs.

(A) The table indicates three groups of treatment in BALB/c mice used in the pharmacokinetic study: CQ alone, front line anti-TB combination therapy (HREZ), and combination (CQ + HREZ). (B to F) Line plots indicate pharmacokinetic profiles of CQ and individual drugs of the anti-TB therapy regimen analyzed individually and in the presence of each other in plasma of animals over 24 hours. No significant difference was observed between groups at each time point indicated in each panel by Mann-Whitney test (P > 0.05). (G) The table depicts ratios of Cmax and AUClast of individual drugs alone or in combination to analyze drug-drug interaction. Doses used are the following: CQ, 10 mg/kg body weight, i.p.; Inh/H, 25 mg/kg body weight, p.o.; Rif/R, 10 mg/kg body weight, p.o.; Emb/E, 200 mg/kg body weight, p.o.; Pza/Z, 150 mg/kg body weight, p.o. p.o., per os consumption; BDL, below detection limit. All data are means ± SD of concentrations at each time point of samples in triplicates (n = 3 animals per group).

The plasma pharmacokinetic profiles of anti-TB drugs remained largely unchanged in the presence of CQ. We observed no major interaction for Rif, Emb, or Pza in the presence or absence of CQ (Fig. 7, D to G) because Cmax and AUClast were within 80 to 125% criteria for equivalence (Fig. 7G) (75). Comparative ratios of Cmax and AUClast for Rif, Emb, and Pza, with and without CQ, were close to one except for Inh, which showed a minor interaction (Cmax ratio, 0.685), although AUClast was not affected (Fig. 7, C and G). This minor influence of CQ on the pharmacokinetics of Inh may be due to the effect of CQ in reducing Inh influx in the intestines (76). Overall, the pharmacokinetic results suggested no adverse drug-drug interactions between the HREZ combination regimen versus CQ and vice versa. In summary, our study shows an effect of CQ on drug susceptibility, no major drug-drug interaction with HREZ, and enhanced in vivo efficacy of CQ-based combinations. With years of safe clinical history for CQ, these findings suggest that CQ could be repurposed for developing new curative combinations for TB.

DISCUSSION

Generation of phenotypic heterogeneity and metabolic quiescence in response to stresses induced by immune activation is the most commonly invoked mechanism of antibiotic tolerance in Mtb (1, 4). However, clinical evidence in humans indicates that tolerance can also be associated with growing bacterial populations (77, 78). Consistent with this idea, early tolerance was documented in actively multiplying Mycobacterium marinum and Mtb in zebrafish larvae and in macrophages, respectively (6). In the present study, we showed a dominant role of phagosomal acidification in facilitating heterogeneity in the redox physiology of Mtb to generate an actively replicating, drug-tolerant population exhibiting higher antioxidant capacity (EMSH-reduced) during infection. Our study constitutes an important foundation linking the macrophage environment with the core redox physiology of Mtb to promote drug tolerance in replicating Mtb, a population not typically associated with tolerance.

Host-induced oxidative stress is a major environmental stress encountered by Mtb during infection and is further exacerbated by anti-TB drugs (12, 79). Because a modest decrease in vacuolar pH is the earliest cue Mtb encounters (18), a pH-dependent reductive shift in EMSH might offer cross-protection to oxidative stress generated by immune activation and drugs. This is supported by the identification of a small molecule (AC2P36) that interferes with thiol homeostasis at acidic pH and increases vulnerability to antibiotics in vitro (80). We also found that the expression of drug efflux pumps and consequent accumulation of antibiotics in Mtb populations are also dependent on the pH-induced remodeling of intramycobacterial EMSH. Because efflux pumps have recently been shown to export oxidatively damaged proteins in Mtb (81), the pH-responsive induction of efflux pumps may be an elegant adaptation strategy to maintain redox homeostasis and tolerance in the face of antibiotics and host immune pressures. Consistent with this idea, the Mtb Rv1258c efflux pump is required for both survival and drug efflux during infection (6).

Because acidic pH encountered inside macrophages does not perturb intramycobacterial pH homeostasis (82), it is more likely that the phagosomal pH-dependent selection of an EMSH-reduced population is part of a bacterial adaptation program during infection. In support of this notion, we have recently found that a redox-sensitive transcription factor, WhiB3, is required to generate an EMSH-reduced population in response to phagosomal acidification (20). As a consequence, a WhiB3-deficient strain showed a growth defect in macrophages and guinea pigs (34) and also exhibited increased susceptibility to Inh. Acidic pH-mediated changes in gene expression and redox potential of Mtb were also reported to be dependent on the PhoPR two-component system (83), indicating overlapping roles of WhiB3/PhoPR in modulating Mtb adaptation to pH stress. Data from our transcriptional profiling also align with the adaptation model wherein the drug-tolerant population (EMSH-reduced) showed enhanced expression of stress-responsive regulons relative to the drug-sensitive fraction (EMSH-basal) inside macrophages. However, although heterogeneity in Mtb populations has been reported during infection (4, 78), it was unclear whether there were variations in the expression of stress regulons or virulence factors in Mtb populations as seen with other pathogens including Salmonella typhimurium and Yersinia pseudotuberculosis (84, 85). Our data demonstrate heterogeneity in the expression of several regulators involved in sensing toxic metals and oxidative stress in redox-altered populations. We also observed a major realignment of sulfur metabolism and proposed that the flux of reduced sulfur metabolites such as CySH into Fe-S cluster assembly, reverse transsulfuration, SAM biosynthesis, and MSH biosynthesis is likely to coordinate Mtb’s defense against antibiotics. Deviations in CySH flux contribute to potentiation of the mycobactericidal efficacy of anti-TB drugs (86). Furthermore, increased expression of SAM-dependent methyl transferases in EMSH-reduced bacteria can promote drug tolerance by N-methylation of antibiotics (40). These transcriptional changes—along with the reduced tolerance shown by MtbΔmetB, MtbΔsufR, and MtbΔwhiB3—led us to propose a model of how phagosomal acidification and bacterial pathways integrate to reset Mtb’s redox physiology for successfully counteracting anti-TB drugs (fig. S14).

Our data also suggest that redox-dependent drug tolerance in replicating Mtb is multifactorial. Although the deletion of single redox-responsive pathways (WhiB3, SufR, and MetB) exhibited substantial influence on drug tolerance, complete clearance of Mtb was not achieved. Similar to our findings, drug tolerance in the growth-arrested Mtb population possibly requires activation of multiple stress regulons and toxin-antitoxin modules (DosR, PhoP, MprA, and MazEF) (1). These studies imply that targeting few bacterial genes or pathways is unlikely to severely affect the ability of Mtb to mobilize drug tolerance in response to host-induced pressures. Conversely, targeting host cues that alter the physiology of Mtb could be an effective mechanism to diminish phenotypic heterogeneity-driven drug tolerance. Our animal data showing the nearly sterilizing effect of CQ in combination with anti-TB drugs provide a compelling argument for targeting vacuolar pH to subvert early emergence of drug tolerance in vivo. Although CQ reinstates redox homogeneity and drug susceptibility mainly by increasing phagosomal pH, other immunomodulatory properties of CQ such as iron depletion (70), blocking phagosomal maturation and autophagy (66), and reversing inflammation-dependent efflux pumps (51) could also contribute to the effect of CQ on multidrug tolerance in vivo. It has been recently shown that CQ potentiates the antimycobacterial activity of Inh and Pza in immune-activated macrophages (51). Because shortening TB chemotherapy requires rapid sterilization of Mtb, our study provides empirical evidence that targeting phagosomal acidification by small molecules has the potential to provide relapse-free control by subverting redox heterogeneity. Because CQ is clinically used, stable, cost effective, and highly tolerable with few side effects, it can be conveniently repurposed to formulate new combinations with the current anti-TB regimen to reduce therapy duration. CQ also reduced redox heterogeneity and Inh tolerance in a HIV-TB coinfection background. Along with clinical evidence showing anti-HIV properties of CQ (87), this raises the possibility of potentiating current anti-TB and anti-HIV therapies by CQ. Last, redox-mediated multidrug tolerance may be relevant to other chronic pathogens. For example, heightened antioxidant capacity is linked to the acquisition of phenotypic antibiotic resistance in the human pathogen Pseudomonas aeruginosa (88). Thus, our findings may have broad relevance to several human pathogens where a sterilizing cure is therapeutically challenging.

Although CQ therapy reduced drug tolerance in vivo, several issues remain to be addressed before it can be combined with the standard anti-TB therapy. First, it needs to be investigated whether a combination of CQ with current therapeutic regimens reduces therapy duration and facilitates the development of immunological memory to prevent relapse. Second, the efficacy of CQ in shortening regimens for drug-resistant Mtb infections remains to be evaluated. Third, despite treatment with anti-TB drugs, Mtb cells persist in animal tissues and the sputum of patients with TB in a metabolically altered state and remain undetectable by viable counts (89, 90). It will be interesting to investigate the impact of CQ on these heterogeneous subpopulations of Mtb that are difficult to detect and retain persister phenotypes in animal models. All these issues require further experimentation using animal models and prospective clinical trials to properly test the efficacy of CQ as adjunct anti-TB therapy.

MATERIALS AND METHODS

Study design

The overall objective of this study was to evaluate host and bacterial mechanisms underlying the emergence of redox heterogeneity and drug tolerance in Mtb during infection. First, we characterized the transcriptome of redox-diverse Mtb fractions (EMSH-reduced and EMSH-basal), which led to the identification of CySH metabolism of Mtb and phagosomal acidification as important factors coordinating redox-mediated drug tolerance inside macrophages. Next, we performed detailed mechanistic studies on the role of phagosomal acidification in generating drug-tolerant EMSH-reduced bacteria in macrophages infected with Mtb alone or coinfected with HIV-1. We applied multiple approaches to studying replication dynamics, metabolic status, and efflux pump activity in EMSH-reduced and EMSH-basal Mtb inside macrophages. We then evaluated the licensed antimalarial drug CQ, which deacidifies phagosomes, in reducing tolerance to standard anti-TB drugs and relapse in animal models of Mtb infection. We assessed the pharmacological compatibility of CQ with first-line anti-TB drugs in mice. Animals were randomly allocated into groups and were identifiable with respect to their treatment. All studies were carried out as per guidelines prescribed by the Committee for the Purpose of Control and Supervision of Experiments on Animals, Government of India, with approval from the Institutional Animal Ethical Committee. Drug treatment and euthanasia were carried out in humane ways to minimize suffering for animals. All experiments were carried out in a biosafety level 3 containment facility and approved by the Institutional Biosafety Committee.

In vivo experiments

For the chronic model of infection (68), 4- to 6-week-old female BALB/c mice (n = 6 per group) were infected by the aerosol route with 100 Mtb H37Rv bacilli using a Madison chamber aerosol generation instrument, housed for 4 weeks for progression of infection, and then left untreated or started under various treatment conditions: (i) 10 mg/kg body weight intraperitoneal doses of CQ on alternate days, (ii) 25 mg/kg body weight of Inh in drinking water daily, (iii) 10 mg/kg body weight of Rif in drinking water daily, (iv) a combination of CQ and Inh (CQ plus Inh) at earlier mentioned doses, and (v) a combination of CQ and Rif (CQ plus Rif) at the mentioned doses (68). At indicated time points of treatment, mice were euthanized, and the lungs were harvested for bacterial burden, gross pathology, and tissue histopathology analysis. The upper right lobe of the lungs of animals from each group was fixed in 10% neutral-buffered formalin. Fixed tissues were prepared as 5-μm-thick sections, embedded in paraffin, and stained with hematoxylin and eosin. Tissue sections were coded, and coded sections were analyzed by a certified pathologist to assess for granuloma formation and lung damage (91). Remaining tissue samples were homogenized in 2 ml of sterile 1× phosphate-buffered saline (PBS), serially diluted, and plated on 7H11-OADC agar plates supplemented with lyophilized BBL MGIT PANTA antibiotic mixture (polymyxin B, amphotericin B, nalidixic acid, trimethoprim, and azlocillin, as supplied by BD). Plates were incubated at 37°C for 3 weeks before colonies were enumerated.

For mice receiving treatment with Inh alone or a combination of CQ and Inh, all treatments were stopped at 12 weeks p.i. (when animals were found to be culture negative for Mtb) for remaining animals (n = 5 per group). Animals were further housed for 8 weeks without treatment, after which four intraperitoneal doses of dexamethasone at 10 mg/kg body weight were administered over 2 weeks for pan-immunosuppression. In the 22nd week p.i., animals in both groups were euthanized, and lung burden of reactivated Mtb was determined by plating lung homogenates for CFUs, as mentioned earlier.

Outbred Hartley guinea pigs (n = 5 per group) were given an aerosol challenge of 100 Mtb H37Rv (92) using a Madison chamber aerosol generation instrument, housed for 4 weeks for progression of infection, and then left untreated or started on treatment in one of three groups: (i) 5 mg/kg body weight intraperitoneal doses of CQ on alternate days, (ii) 30 mg/kg body weight of Inh in drinking water daily, and (iii) a combination of CQ and Inh (CQ plus Inh) at earlier mentioned doses. At 8 weeks after commencement of treatment, guinea pigs were euthanized, and lung burden of Mtb was determined by homogenizing organs in 5 ml of sterile 1× PBS, serial dilution, and plating on 7H11-OADC agar plates supplemented with PANTA. Upper right lobes of the lungs from different treatment groups were fixed in neutral-buffered formalin and prepared, as mentioned earlier, for histopathological analysis (91).

Statistical analysis

All statistical analyses were performed using GraphPad Prism software (version 6.0). All data indicated are means ± SD except figs. S12 and S13, where median ± interquartile range was plotted for animal groups. The Mann-Whitney rank sum test was used for comparison of nonparametric data between two experimental groups. Nonparametric multiple group comparisons were analyzed using the Kruskal-Wallis test with Dunn’s post hoc correction. For overlap analysis of differentially expressed (DE) genes with other microarray studies, the significance of gene number overlap was determined by Fisher’s exact test on a two-by-two contingency table (93). Differences with P < 0.05 were considered significant.

SUPPLEMENTARY MATERIALS

stm.sciencemag.org/cgi/content/full/11/518/eaaw6635/DC1

Materials and Methods

Fig. S1. Phenotypic drug tolerance in Mtb during infection.

Fig. S2. Flow cytometry–based quantification of redox heterogeneity in Mtb using Mrx1-roGFP2.

Fig. S3. EMSH-reduced population is tolerant to Inh.

Fig. S4. Transcriptome of Mtb from EMSH-reduced and EMSH-basal fractions.

Fig. S5. Measuring phagosomal pH of THP-1 macrophages infected with Mtb/Mrx1-roGFP2.

Fig. S6. The transcriptome of Mtb from the EMSH-reduced fraction overlaps with low pH–specific WhiB3 regulon.

Fig. S7. WT Mtb generates H2S gas in a pH-dependent manner.

Fig. S8. Generation and characterization of MtbmetB and MtbsufR.

Fig. S9. Deletion of metB and sufR does not impair growth and metabolism of Mtb.

Fig. S10. Phagosomal acidification is required for the redox-dependent multidrug tolerance of Mtb.

Fig. S11. Drug-tolerant EMSH-reduced population is replicative and has high efflux pump activity.

Fig. S12. CQ counteracts drug tolerance in vivo to reduce lung tissue damage in chronic model of Mtb infection.

Fig. S13. Long-term CQ treatment of chronically infected BALB/c mice deacidifies macrophage pH without affecting oxidative stress and necrosis.

Fig. S14. Model depicting various mechanisms underlying redox-mediated drug tolerance in replicating Mtb.

Table S1. List of differentially expressed genes from DESeq2 for EMSH-reduced, EMSH-basal, and in vitro control samples.

Table S2. List of differentially expressed genes from DESeq2 for WT Mtb, MtbwhiB3, and whiB3-Comp strains at pH 6.6 and pH 4.5 used to specify the low pH–inducible WhiB3 regulon.

Table S3. List of EMSH-reduced and EMSH-basal differentially expressed genes used to generate custom heat maps.

Table S4. List of strains and primers used in this study.

References (94112)

REFERENCES AND NOTES

Acknowledgments: We are grateful to C. Grundner at the Seattle Children’s Research Institute for critical reading of the manuscript and valuable input. We acknowledge W. R. Jacobs Jr. at the Albert Einstein College of Medicine for the MtbmshA mutant and D. J. V. Beste at the University of Surrey for providing the zeocin-marked vector pANEE001. C. N. Naveen at the Foundation of Neglected Disease Research (FNDR), India is acknowledged for supporting pharmacokinetic and drug-drug interaction studies in mice. We thank S. Nayak for valuable discussions about the key findings of the study and designing of schematics and models. We acknowledge the CIDR BSL-3 facility, Indian Institute of Science (IISc) for carrying out experiments with Mtb and A. Pandit and the NGS facility, NCBS for RNA-seq. The mshA mutant of Mtb was obtained from W. R. Jacobs Jr. under a material transfer agreement (MTA) between the Albert Einstein College of Medicine of Yeshiva University and the International Centre for Genetic Engineering and Biotechnology. Funding: This work was supported by Wellcome Trust–DBT India Alliance grant IA/S/16/2/502700 (to A.S.); in part, by Department of Biotechnology (DBT) grants BT/PR11911/BRB/10/1327/2014 and BT/PR13522/COE/34/27/2015 (to A.S.); and by the DBT-IISc Partnership Program (22-0905-0006-05-987436). A.S. is a senior fellow of Wellcome Trust–DBT India Alliance. A.S.N.S. is supported by the Wellcome Trust–DBT India Alliance Grant IA/I/16/2/5-2711. R.M. is supported by the IISc, Bengaluru. S.K. and M. Mehta are national postdoctoral fellows supported by the Department of Science and Technology (DST), India. N.M. is supported by the National Centre for Biological Sciences—Tata Institute of Fundamental Research (NCBS-TIFR), Bengaluru. P.B. is supported by a fellowship provided by the Council of Scientific and Industrial Research (CSIR), India. Author contributions: R.M. and A.S. conceptualized the research and prepared the manuscript. R.M. and S.K. performed the experiments and analyzed data. N.M. and A.S.N.S. conducted the RNA-seq analysis. P.B., M. Mehta, and M. Munshi constructed the knockout and complemented strains and conducted the HIV-TB coinfection and H2S measurement assays. V.A. performed the flow sorting. V.K.A. and R.K.S. conducted the pharmacokinetic drug-drug interaction studies. R.S.R. conducted animal experiments. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials. RNA-seq data generated and analyzed in this study have been uploaded to the NCBI Gene Expression Omnibus under accession number GSE123267.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article