Mechanisms of Drug-Induced Tolerance in Mycobacterium tuberculosis

Successful treatment of tuberculosis (TB) can be hampered by Mycobacterium tuberculosis populations that are temporarily able to survive antibiotic pressure in the absence of drug resistance-conferring mutations, a phenomenon termed drug tolerance. We summarize findings on M. tuberculosis tolerance published in the past 20 years. Key M. tuberculosis responses to drug pressure are reduced growth rates, metabolic shifting, and the promotion of efflux pump activity.

SUMMARY

Successful treatment of tuberculosis (TB) can be hampered by Mycobacterium tuberculosis populations that are temporarily able to survive antibiotic pressure in the absence of drug resistance-conferring mutations, a phenomenon termed drug tolerance. We summarize findings on M. tuberculosis tolerance published in the past 20 years. Key M. tuberculosis responses to drug pressure are reduced growth rates, metabolic shifting, and the promotion of efflux pump activity. Metabolic shifts upon drug pressure mainly occur in M. tuberculosis’s lipid metabolism and redox homeostasis, with reduced tricarboxylic acid cycle activity in favor of lipid anabolism. Increased lipid anabolism plays a role in cell wall thickening, which reduces sensitivity to most TB drugs. In addition to these general mechanisms, drug-specific mechanisms have been described. Upon isoniazid exposure, M. tuberculosis reprograms several pathways associated with mycolic acid biosynthesis. Upon rifampicin exposure, M. tuberculosis upregulates the expression of its drug target rpoB. Upon bedaquiline exposure, ATP synthesis is stimulated, and the transcription factors Rv0324 and Rv0880 are activated. A better understanding of M. tuberculosis’s responses to drug pressure will be important for the development of novel agents that prevent the development of drug tolerance following treatment initiation. Such agents could then contribute to novel TB treatment-shortening strategies.

INTRODUCTION

Tuberculosis (TB), an infectious disease caused by Mycobacterium tuberculosis, remains an important global public health problem, with 10 million new cases and 1.5 million TB deaths in 2018 (1). Even though TB is curable, the treatment success rates in 2016 were only 82% for drug-susceptible TB and 55% for multidrug-resistant (MDR) TB (2).

The standard treatment for drug-susceptible TB consists of four drugs (isoniazid [INH], rifampicin [RIF], ethambutol [EMB], and pyrazinamide [PZA]) for 2 months followed by INH and RIF for 4 months (3). The response to this standardized treatment regimen shows interpatient variability, even among patients with confirmed drug-susceptible TB. An important measure of the response to treatment is the time to culture conversion (i.e., the time from treatment initiation to the first of two consecutive negative cultures), which varies between patients, with 95% confidence intervals ranging from 41 to 83 days (4–6). Furthermore, while most patients are cured and remain free of TB upon the completion of a 6-month treatment regimen, some patients relapse even if they were fully adherent to treatment (7). To date, this interpatient variability in treatment responses has primarily been attributed to differences in host factors such as baseline mycobacterial loads, the presence of cavitary lesions, HIV infection, and smoking (8–13).

In addition to these host factors, the heterogeneity of mycobacterial populations may also contribute to the observed interpatient variability in treatment responses. Several studies demonstrated that kill curves of M. tuberculosis under drug pressure are biphasic and possibly even multiphasic (14, 15). Similarly, a recent modeling study of the time to culture positivity (i.e., the time between inoculation of medium and detection of growth) in serial isolates showed that multiple bacterial populations with different growth dynamics and distinctive kill rates are present in a patient’s sputum specimen (16). Taken together, these data suggest that mycobacterial populations are heterogeneous, with more than one bacterial phenotype being present in a single patient (14, 15, 17–20).

Drug tolerance, the phenomenon of M. tuberculosis surviving antibiotic treatment for a prolonged period in the absence of resistance mechanisms (21), is one manifestation of this heterogeneity in mycobacterial populations. While drug tolerance has been extensively studied in other bacterial species such as Staphylococcus aureus, Salmonella spp., and Escherichia coli, it has received less attention in mycobacteria (22). For M. tuberculosis, studies have mainly focused on mycobacterial persisters, a subset of drug-tolerant mycobacteria that arise within inherently heterogeneous populations. Through growth arrest, these persisters survive environmental stress conditions, including stress induced by concentrations of bactericidal drugs that are otherwise lethal (21, 23). While antibiotic persistence and drug tolerance share several characteristics, both differ from drug resistance. Drug resistance, or the ability to replicate in the presence of a drug, is most often specific to a single drug or drug class, is inheritable, and increases the MIC, the lowest concentration of a specific antibiotic needed to prevent growth (21, 24). In contrast, drug tolerance often acts across drug classes, is rarely genetically encoded, and does not affect the MIC but changes the MDK₉₉, the minimum duration of treatment that kills 99% of the mycobacterial population (7, 21, 24). Upon exposure to bactericidal drugs, tolerant mycobacteria are thus killed at a lower rate than the fully susceptible population from which they arose (21). In addition to persistence, tolerance, and drug resistance, M. tuberculosis can also adapt to antibiotic stress by employing “intrinsic resistance” strategies, which include transient mechanisms such as thickening of the mycobacterial cell wall, activation of efflux pumps, and altered expression of transcriptional regulators.

In this review, we summarize and critically appraise the published data on the mechanisms that mycobacteria employ to transiently alter drug sensitivity upon drug pressure. While M. tuberculosis adapts to many stress conditions, including nutrient deficiency, acid stress, and exposure to immune cells, we review here what is known about mechanisms that are activated upon antibiotic drug pressure. Furthermore, while intrinsic resistance mechanisms sensu stricto can be classified as drug resistance, we include these phenomena in our review under the umbrella of the drug-tolerant M. tuberculosis state, as intrinsic resistance is clearly distinct from drug resistance caused by genomic variants in resistance-conferring genes. A better understanding of drug tolerance in M. tuberculosis is important as it may allow the development of novel treatment-shortening strategies aimed at targeting those M. tuberculosis populations that have become tolerant and would otherwise temporarily survive the presence of drugs.

TOLERANCE MECHANISMS

The survival of mycobacteria strongly depends on their ability to quickly respond to stresses. The published literature reveals that several interconnected biological pathways are involved in the emergence and establishment of a drug-tolerant state in response to the stress imposed by an effective drug (25–27). While some of these mechanisms are general, as they are nonspecific and occur upon exposure across drug classes (general tolerance), others are drug-specific tolerance mechanisms and occur only following exposure to one specific drug or drug class (drug-specific tolerance).

Under drug pressure, the expression of hundreds of M. tuberculosis genes changes rapidly. The resulting transcriptional and posttranscriptional gene regulation results in a tolerant phenotype of M. tuberculosis. As shown in Fig. 1, the main mechanisms conferring a drug-tolerant phenotype in M. tuberculosis include (i) metabolic slowdown by reducing the metabolism and growth rate, which is a successful strategy as antibiotics generally target active cellular processes; (ii) metabolic shifting, which consists of rerouting metabolite fluxes as an approach to retain homeostasis when specific pathways are targeted by antibiotic drugs; (iii) cell wall thickening, which reduces the drug concentration in mycobacterial cells; and (iv) the upregulation of efflux pumps, which increases the clearance of antibiotics out of the mycobacterial cell. Whether the observed changes in gene regulation reflect the development of a tolerant M. tuberculosis state in response to antibiotic stress, the selection of a preexisting drug-tolerant mycobacterial subpopulation upon exposure to antibiotics, or a combination of both is not known and difficult to disentangle using currently available techniques.

FIG 1.

Mechanisms for cross-tolerance in Mycobacterium tuberculosis. (A) Under nonstress conditions, cellular processes driving core metabolic pathways are unaffected. (B) Under drug pressure, homeostasis is affected by targeting core metabolic processes and toxic compounds, resulting in cellular damage and eventually cell death. (C) M. tuberculosis (Mtb) adapts to drug pressure through transcriptional and posttranscriptional regulatory mechanisms to develop a drug-tolerant state: metabolic slowdown reduces drug-induced cellular damage (1); metabolic processes are rerouted, and homeostasis is retained through the up- and downregulation of interweaved pathways (2); cell wall thickening limits drug entry into the cell and lowers drug activity (3); and the upregulation of efflux pumps lowers intracellular drug concentrations (4).

General Tolerance Mechanisms

Transcriptional and posttranscriptional gene regulation.

The processes that orchestrate the cellular developments leading to drug tolerance are complex and regulated by highly interconnected metabolic pathways. In this section, we focus on the most frequently reported regulatory mechanisms and describe important transcriptional and posttranscriptional regulatory mechanisms in M. tuberculosis. At the transcriptional level, selected WhiB transcription factors and sigma factors function as stress regulators. Posttranscriptionally, toxin-antitoxin (TA) systems and small regulatory RNAs (srRNAs) coordinate drug tolerance.

whiB genes encode transcriptional regulators involved in multiple cellular processes, including cell division, pathogenesis, and the response to diverse stresses, including exposure to antibiotics (28). In M. tuberculosis, whiB3 and whiB7, two of the seven identified whiB-like genes, have been associated with intrinsic drug resistance (29). For whiB7, studies have shown that its expression is induced upon antibiotic pressure and that whiB7-null mutants are hypersusceptible to antibiotics in vitro (30). In Mycobacterium smegmatis, pretreatment with whiB7 activators reduced drug susceptibility (31). The whiB7 transcriptional regulator may result in a drug-tolerant state by upregulation of drug efflux pumps (tap or Rv1258c) and by readjustment of cellular processes that compensate for the metabolic shifting induced under drug pressure (30). whiB3 expression has also been associated with reduced drug sensitivity, as mice infected with M. tuberculosis whiB3 mutant strains had longer survival times under drug pressure than mice infected with wild-type whiB3 M. tuberculosis strains (32). Because whiB3 encodes a 4Fe-4S redox sensor protein, WhiB3 likely plays a role in modulating drug sensitivity through regulation of the balance between redox and bioenergetic homeostasis, both processes known to affect the response to TB drugs (33–35). Other core metabolic pathways where WhiB3 plays a regulatory role are glycolysis, the pentose phosphate pathway, the tricarboxylic acid (TCA) cycle, and amino acid biosynthesis (35).

Sigma factors also play a crucial role in prokaryote transcription by binding RNA polymerase. The regulatory network of sigma factors is complex and involves multiple posttranslational modifications such as phosphorylation and acetylation via protein kinases and anti-sigma factors. In M. tuberculosis, 13 sigma factors (SigA to SigM) have been identified (36). Upon drug pressure, the expression of SigB, -E, -F, -G, -H, -I, and -J is increased, while the expression of SigA was found to be reduced (27, 37). Studies have shown that SigA, SigB, SigE, and SigF may be involved in M. tuberculosis’s adaptation to antibiotic stress (38–40).

TA systems are two-component systems that permit rapid adaptation of gene expression in bacteria following stress exposure (41). Toxin-antitoxin systems comprise a toxin that can induce growth arrest (or cell death) and an antitoxin that neutralizes the toxin. Under conditions where growth is favored, both the toxin and antitoxin are ubiquitously expressed so that the action of the toxin is neutralized by the antitoxin and normal growth occurs. Under antibiotic stress, growth inhibition can be obtained through the induction of specific proteases that degrade the antitoxin so that the action of the toxin is no longer restricted (42, 43). In M. tuberculosis, 79 TA loci have been identified; both the expression of various TA systems as well as the expression of proteases are altered upon antibiotic pressure and increased upon prolonged exposure to drugs (42–44). While some toxins (e.g., the MazF, Rv1577c, Rv2651c, and Rv0366c family) confer drug tolerance across drug classes, other toxins (e.g., RelE) generate drug-specific antibiotic tolerance, as discussed in the relevant sections below (45–47).

Small RNAs (sRNAs) regulate bacterial gene expression by binding mRNA, thereby repressing mRNA translation and increasing mRNA degradation (48). sRNAs were implicated in the regulation of antibiotic responses by readjusting the expression of genes associated with efflux pumps, metabolic enzymes, transport proteins, cell wall synthesis, and membrane proteins (49–51). Whereas modified in vitro expression of sRNA has been shown to alter antibiotic susceptibility in nonmycobacterial species (52, 53), sRNA regulation in M. tuberculosis is poorly understood, with validation of sRNA-target interactions being limited to a single study (54).

Metabolic slowdown.

Because most antibiotics target metabolically active M. tuberculosis, lowered metabolic activity under drug pressure can successfully generate a drug-tolerant state (Fig. 1C, panel 1) (55–57). For example, the observation that RIF kills significantly more log-phase bacilli than stationary-phase bacteria illustrates that metabolic slowdown via reducing replication is a successful strategy to produce RIF tolerance (58). Furthermore, the accumulation of mutations related to a decreased metabolic state has been associated with long-term treatment courses, further supporting metabolic slowdown as an efficient drug tolerance strategy (59).

In M. tuberculosis, metabolic slowdown occurs through the regulation of multiple interrelated pathways that control energy, carbon, and lipid metabolism. Upon treatment initiation, genes involved in the TCA cycle, aerobic respiration, and ATP synthesis, the main pathways involved in energy metabolism, are downregulated under antibiotic stress (27). Also, mRNA synthesis and protein synthesis, proxies for metabolic activity, are reduced, as evidenced by the lowered expression of genes involved in transcription (tuf, gyrA, and gyrB) and essential translation components (including 16S RNA and genes encoding ribosomal proteins) (27).

Metabolic shifting.

In addition to lowered metabolic activity, M. tuberculosis can also generate drug tolerance by shifting between pathways (Fig. 1C, panel 2). Carbon flux rerouting from energy-generating pathways to energy storage pathways has been associated with growth arrest and reduced drug susceptibility in M. tuberculosis (33, 60, 61). While most of the findings on metabolic shifting are based on gene expression studies, a few differential protein expression studies also found changes consistent with metabolic shifting of lipid metabolism and redox homeostasis in response to drug exposure (62–64). Under stress conditions (including drug pressure), M. tuberculosis can shift from carbon metabolism via the growth-promoting TCA cycle to carbon storage in fatty acids (FAs) via triacylglycerol (TAG) synthesis via the upregulation of tgs1 (gene encoding triacylglycerol synthase) (33, 65). When the TCA cycle is slowed, the turnover of alpha-ketoglutarate, oxaloacetate, and energy-rich reducing agents such as NADH is reduced, which results in reduced amino acid synthesis, protein translation, and metabolism in general. In addition, genes encoding enzymes that catalyze the glyoxylate bypass, such as isocitrate lyases, are found to be upregulated under RIF pressure, which is shown to result in cross-tolerance to INH, RIF, and streptomycin (34). The use of the glyoxylate bypass, an alternative route in the TCA cycle, shortcuts the production of alpha-ketoglutarate and a significant part of the production of energy-rich compounds in the TCA cycle. Alpha-ketoglutarate-derived amino acids can thus form a bottleneck for normal growth and metabolism, while the decreased production of highly energetic reducing agents could limit the production of radical oxygen species (ROS) induced by bactericidal drugs (Fig. 2). Interestingly, the metabolic shifts associated with drug tolerance overlap metabolic shifts observed in dormancy, where M. tuberculosis accumulates TAG lipid bodies that are hydrolyzed during periods of nutrient starvation (66, 67). It is unclear whether the metabolic shifts resulting in the accumulation of energy storage compounds should be regarded as being the result of reduced growth or a more universal and fundamental process that induces both dormancy and drug tolerance.

FIG 2.

Metabolic shifting of carbon fluxes away from the TCA cycle. The promotion of FA synthesis through the upregulation of tgs1 is one strategy adopted by M. tuberculosis to limit carbon uptake in the TCA cycle by competing for mutual carbon sources. Reduced TCA cycle activity results in lower rates of turnover of alpha-ketoglutarate and oxaloacetate, both of which are metabolites required for amino acid synthesis. The upregulation of the glyoxylate bypass further limits the turnover of alpha-ketoglutarate.

Mycobacterial cell wall thickening.

In mycobacteria, the cell wall acts as a natural barrier, and thickening of the cell wall structure in response to antibiotic stress could result in a drug-tolerant state (Fig. 1C, panel 3) (68–71). The thick and viscous bilipid layer of the cell wall slows the uptake of lipophilic drug agents and reduces the entry of hydrophilic drug agents through mycobacterial porins (Fig. 3). The reduced uptake of lipophilic drugs following cell wall thickening is supported by the observation of increased uptake and influx of lipophilic drugs (RIF) and, consequently, enhanced drug susceptibility in M. smegmatis mycolate-deficient mutants that were defective in crucial cell wall synthesis genes (72–77).

FIG 3.

Cell wall thickening. Cell wall thickening is a tolerance mechanism that results in the less efficient transport of lipophilic drugs across the mycobacterial cell wall. The transport of hydrophilic drugs is reduced as fewer porins now extend across the entire thickness of the cell wall.

Mycobacterial cell wall thickening also restricts the entry of hydrophilic drugs, as porins may no longer be able to bridge the increased cell wall width (Fig. 3) (68, 78, 79). Mycobacterial porins may indeed play an important role in regulating drug susceptibility in mycobacteria. For example, M. smegmatis mspA mutants that result in a nonfunctional MspA porin are less sensitive to hydrophilic drugs (ampicillin) and hydrophobic drugs (RIF). Conversely, the overexpression of the MspA porin in M. tuberculosis increases susceptibility to INH, EMB, streptomycin, and β-lactam antibiotics (80, 81), and the expression of Rv1698 and OmpA porins restores drug susceptibility in mspA mutants (81, 82).

As cell wall thickening occurs only under anaerobic or microaerobic stress conditions, cell wall thickening of M. tuberculosis most likely acts as a strategy complementary to the metabolic shift from aerobic energy-related pathways (such as aerobic respiration and ATP synthesis) to anaerobic respiration (27, 83, 84). Similarly, the various adaptations in M. tuberculosis’s lipid metabolism and the altered redox homeostasis upon drug pressure could influence cell wall permeability and contribute to a drug-tolerant state (65).

Upregulation of efflux pumps.

Efflux pumps are proteinaceous transporter channels that pump various compounds, including antibiotics, out of the cell. While specific mutations in efflux-associated genes have been associated with fixed resistance to several TB drugs, the upregulation of efflux pump expression also functions as an adaptive mechanism that can be activated upon drug pressure in genetically susceptible M. tuberculosis strains. Efflux pump expression is condition dependent, reversible, and transient (Fig. 1C, panel 4) (4, 85). Several efflux pumps have been found to be upregulated under antibiotic stress, either a specific drug or multiple structurally unrelated drugs. Similar to the phenomenon where a specific mutation for one drug can cause “cross-resistance” to another drug, the broad target range of efflux pumps can result in reduced sensitivity of M. tuberculosis to multiple TB drugs (86) (Table 1). For example, the upregulation of efflux pump activity and metabolic shifting induced by exposure to RIF may result in cross-tolerance to ofloxacin (87).

TABLE 1.

Efflux pumps upregulated under antibiotic stress

Efflux pump(s)/gene	Induced upon exposure to:	Reduces susceptibility to:	Reference(s)
Rv1258c (Tap)	RIF, capreomycin	RIF	27, 30, 31, 154, 136
iniBAC operon	INH, EMB	Multiple drugs	26, 65, 155, 156
Rv3065 (mmr)	INH, levofloxacin	Conflicting information	27, 157, 158
efpA	INH, ETH	INH (hypothesized)	26, 27, 65, 94, 155, 158
Rv0849	INH or RIF	RIF, amikacin	157, 158
Rv1218c	No literature found	Multiple drugs	157
Multiple efflux and transporter genes	RIF	RIF, ofloxacin	87

Genetic adaptations conferring tolerance.

Even though tolerance is a reversible phenotypic state of M. tuberculosis, recent evidence suggests that the presence of specific genetic variants can increase the propensity of M. tuberculosis to enter a drug-tolerant state under antibiotic stress. For example, the presence of mutations in the prpR gene that regulates propionate catabolism conferred drug tolerance when M. tuberculosis was exposed to RIF, INH, and ofloxacin (88). Drug tolerance occurred through altering propionyl-CoA metabolism when bacteria were grown in vitamin B₁₂-free propionate-supplemented medium. Tolerance was reversed upon the addition of vitamin B₁₂, as propionate was then metabolized in the vitamin B₁₂-dependent methylmalonyl-CoA pathway (88). Another example is “high-persisting” (hip) mutants of genes linked to lipid biosynthesis and carbon metabolism pathways that have been isolated under RIF and streptomycin pressure (89). In hip mutants, reduced FadE30 activity is likely to manifest as reduced lipid catabolism, consistent with metabolic shifting of carbon away from the TCA cycle in favor of lipid synthesis. The upregulation of icl and tgs1 in hip mutants is also of interest (89), as this promotes the glyoxylate bypass and the redirection of carbon sources into lipid synthesis.

While genetic changes are usually permanent, transient frameshift mutations in the glpK gene, which encodes a glycerol-3-kinase, are associated with tolerance to INH, RIF, PZA, and moxifloxacin, both in vitro and in clinical isolates (90, 91). It has been hypothesized that glpK frameshift mutations result in drug tolerance through the metabolic changes that result from reduced glycerol metabolism in glpK mutants, which could affect drug effectiveness by reducing growth, impacting the cellular structure, promoting the expression of stress response regulators (such as DosR and SigH), or inducing genes related to TAG synthesis (90, 91).

Drug-Specific Tolerance Mechanisms

Isoniazid.

Isoniazid (INH) is a prodrug that is converted intracellularly by the catalase-peroxidase KatG into isonicotinoyl, which binds NAD⁺ to form isonicotinoyl-NAD. Isonicotinoyl-NAD negatively affects cell wall integrity by inhibiting InhA, a key enzyme in the synthesis of the bacterial cell wall lipid mycolic acid (Fig. 4). KatG-mediated conversion of INH generates high levels of mutagenic ROS, which contributes to the high bactericidal activity of INH. During the first 2 days of treatment, INH kills about 90% of M. tuberculosis bacteria (65). Thereafter, the bactericidal activity decreases, a phenomenon attributed to INH tolerance (27, 65). The transcriptional response to INH stress occurs rapidly following INH exposure (25, 65, 92–94), with over 100 genes upregulated in the bacilli that remain alive after the initial killing by INH (55, 57, 95, 96). The majority of the INH-regulated genes cluster in metabolic pathways involved in lipid metabolism and cell wall synthesis, reflecting the two main INH tolerance mechanisms: a reduced need for lipid synthesis due to reduced growth and metabolic shifting of lipid metabolism and redox homeostasis pathways to circumvent INH-induced disruption of cell wall integrity (Fig. 4). The involvement of different pathways suggests that M. tuberculosis exploits a wide arsenal of tolerance mechanisms and is consistent with the discovery of cross-talking mechanisms within mycobacterial metabolism.

FIG 4.

Isoniazid (INH)-specific tolerance in Mycobacterium tuberculosis. (A) In the absence of tolerance, INH inhibits mycolic acid synthesis, resulting in insufficient mycolic acid production for growth. (B) Upon INH exposure, M. tuberculosis employs several tolerance mechanisms to overcome the INH-induced inhibition of mycolic acid synthesis: de novo-synthesized fatty acids and degraded high-molecular-weight TAGs fuel mycolic acid synthesis by acting as a source of mycolic acid precursors that enter mycolic acid synthesis (1), genes involved in mycolic acid synthesis (cmaA2, cmrA, mmaA2, mmaA3, mmaA4, umaA1, and fbpC) are upregulated (2), reduced carbon flow toward the TCA cycle leads to metabolic slowdown due to lowered ATP and protein synthesis (3), the promotion of mycolic acid synthesis and the reduced need for mycolic acids in light of reduced growth ensure sufficient mycolic acid production (4), and M. tuberculosis decreases the expression of NADH dehydrogenase to avoid the further accumulation of NAD⁺ when NAD⁺ is accumulated upon increased production of fatty acids and mycolic acids and reduced TCA cycle activity (5).

The first important INH tolerance mechanism is the reduced need for lipid synthesis due to reduced growth and metabolic shifting of lipid metabolism, as INH tolerance has been observed in slow-growing and nonreplicating M. tuberculosis (55, 57, 96, 97). Under INH stress, M. tuberculosis likely reduces its lipid dependency by reducing metabolic activity (65). The production of cell wall lipids then becomes less critical as fewer metabolites are required for growth. In vitro experiments further suggest that under INH stress, M. tuberculosis may sense the low concentrations of mature mycolic acids and increased levels of mycolic acid biosynthesis intermediates, resulting in M. tuberculosis adapting its lipid metabolism. For example, M. tuberculosis upregulates multiple genes involved in mycolic acid synthesis (fas, the kas operon, cmaA2, cmrA, mmaA2, mmaA3, mmaA4, umaA1, and fbpC), which could promote mycolic acid synthesis and circumvent the cell wall damage induced by INH (62, 65, 92–94). In addition, genes involved in lipid metabolism not directly related to mycolic acid synthesis are implicated in the INH stress response. For example, tgs1, which metabolically shifts carbon sources from the TCA cycle into TAG lipid synthesis, and genes involved in lipid degradation, such as fabG4, fadE5, fadA, fadB, prpC, fadE24, and fadE23, are upregulated upon treatment with INH (65, 94). The storage lipids produced by the TAG synthesis pathway can be used by M. tuberculosis as a reservoir for fatty acids required for mycolic acid biosynthesis (98). Exposure to INH also increases the TCA cycle’s glyoxylate bypass through upregulating isocitrate lyase, generating lower rates of amino acid precursor turnover and fewer high-energy reducing agents such as NADH, which in turn likely results in decreased metabolic activity and decreased ROS production (34).

Another important INH tolerance mechanism is related to redox homeostasis. Redox homeostasis is affected under INH stress due to the altered lipid metabolism and reduced production of NADH in the TCA cycle, which results in the accumulation of NAD⁺. Under INH pressure, M. tuberculosis downregulates the expression of NADH dehydrogenase-encoding genes to limit the production of NAD⁺ (95). This shifts the chemical equilibrium toward reduced NAD⁺ to isonicotinoyl binding, which reduces InhA inhibition. This reversible adaptation is mechanistically similar to one of the resistance mechanisms that M. tuberculosis employs, where a fixed mutation in the NADH dehydrogenase-encoding ndh gene confers irreversible resistance to INH (99). In addition to the downregulation of NADH catabolism, INH stress is likely to modify M. tuberculosis’s redox homeostasis by inducing the expression of oxidoreductases (fprB, Rv0085, and Rv1885c); Rv3049c, which is presumed to function as a monooxygenase that interacts with NADP; and the oxidative stress response gene ahpC that encodes an antioxidant enzyme that protects mycobacteria against radical oxygen species (27, 63, 65, 94). The role of redox homeostasis in INH tolerance is further supported by the observation that upon host infection, when M. tuberculosis is taken up by macrophages, M. tuberculosis senses the redox changes that result from phagosomal acidification and uses them as a trigger to promote drug tolerance (100). Interestingly, in vivo mouse and guinea pig models suggest that chloroquine, an antimalarial drug that inhibits phagosomal acidification, may result in increased susceptibility of M. tuberculosis to INH, as the coadministration of chloroquine with INH increased the killing of drug-tolerant bacteria, reduced lung damage, and decreased relapse rates (100).

A temporary upregulation of inhA or downregulation of katG may also result in INH tolerance, but studies report conflicting results. One study reported that the histone-like MDP1 DNA binding protein that inhibits katG expression results in phenotypic INH tolerance in M. smegmatis (101). Another study observed that subpopulations with a reduced frequency of katG pulsing are more likely to be INH tolerant (102). However, a third study failed to show any differential inhA or katG expression upon challenge with INH (27).

Finally, the sigma factors SigB and SigE were found to play a role in M. tuberculosis tolerance induced upon exposure to INH, with INH being more active in SigB and SigE loss-of-function M. tuberculosis mutants (38). The role of SigB in INH tolerance has been studied in M. smegmatis, where SigB interacts with RNA polymerase binding protein A (RbpA) to promote the transcription of polyphosphate kinase 1 (ppk1) (103). Increased Ppk1 activity results in the accumulation of polyphosphates, a process that has been associated with tolerance in response to heavy metal or osmotic stress in E. coli, Lactobacillus, and other bacteria (103–105). Furthermore, SigB and SigE may play a role in M. tuberculosis tolerance through the transcriptional repressor MprA, which is bound to conserved motifs in the upstream regions of both SigB and SigE.

Rifampicin.

Rifampicin (RIF) inhibits the transcription of RNA by binding the RNA polymerase B subunit. Mutations in the RNA polymerase B-encoding rpoB gene that result in a reduced binding affinity of RIF have been extensively studied with regard to the acquisition of RIF resistance (106). Transient phenotypic adaptations of M. tuberculosis in response to exposure to RIF have been reported, but the literature is scarce. Studies demonstrated that M. tuberculosis can induce tolerance to RIF exposure through mistranslation of rpoB, upregulation of rpoB, increased efflux pump activity, and metabolic shifting (27, 34, 87, 107, 108).

Mistranslation is a stress-sensitive phenomenon where the translation of mRNA results in a mutated or truncated protein product. In vitro assays have shown that mistranslation of rpoB is a non-genetically-encoded phenotypic phenomenon that reduces the binding affinity between the drug and its target (108). Whether RIF pressure itself promotes the mistranslation of rpoB and paradoxically results in RIF tolerance remains unknown.

In response to low RIF concentrations, the upregulation of rpoB through the transcriptional regulation of M. tuberculosis rpoB promoters I and II occurs exclusively in rpoB wild-type M. tuberculosis populations and decreases the pathogen’s sensitivity to RIF (107). Under normal conditions, rpoB expression from promoter I inhibits rpoB expression from promoter II. Upon exposure to RIF, inhibition of RNA polymerase reduces transcription from promoter I, resulting in reduced inhibition of promoter II and a net increase of rpoB transcription. Interestingly, this mechanism is observed only when M. tuberculosis is exposed to low concentrations of RIF, as exposure to high RIF concentrations leads to the simultaneous inhibition of RNA polymerase activity at both promoters I and II (107).

In addition to the overexpression of the drug target, metabolic slowdown via reducing replication can result in RIF tolerance, as RIF kills significantly more M. tuberculosis bacilli in log phase than M. tuberculosis bacilli in stationary phase (58).

Sigma factors may also play a role in modulating RIF susceptibility, as mutants lacking sigF showed increased RIF susceptibility (40). However, this finding was not replicated in a later study (109).

Ethambutol.

Ethambutol (EMB) exerts its bacteriostatic activity by inhibiting the production of arabinogalactan, a mycobacterial cell wall constituent (110, 111). EMB may further act synergistically with INH by targeting the mycolic acid biosynthesis pathway through inhA repression, suggesting that the M. tuberculosis tolerance mechanisms employed for INH may generate cross-tolerance to EMB (112). While there is limited research on tolerance to EMB, one study found that EMB is highly bactericidal when M. tuberculosis lacks the SigB or SigE sigma factor, suggesting that the sigma factors SigB and SigE could induce tolerance upon EMB exposure (38).

Pyrazinamide.

Pyrazinamide (PZA) is a prodrug that is converted by the amidase encoded by the pncA (Rv2043c) gene into the active compound pyrazinoic acid (POA), which affects the activity of critical cellular enzymes and energy production through acidification of the cytoplasm and destabilization of the proton motive force (113). To date, no studies have investigated PZA-induced tolerance. What has been shown is that while PZA has limited bactericidal activity against actively growing bacteria, PZA is effective against static and metabolically slowed M. tuberculosis bacteria that remain under INH and RIF pressure (113, 114). An explanation for this observation could lie in the fact that POA is pumped out of M. tuberculosis by an efflux pump, a process that is energy dependent. Indeed, conditions such as hypoxic stress that reduce energy production or cotreatment with energy inhibitors increase PZA activity (113). The improved bactericidal activity of PZA under conditions of reduced energy could also explain the strong synergism between PZA and bedaquiline (BDQ), an important TB drug inhibiting ATP synthesis, in mouse models (115) and clinical trials (116). Downregulation of the drug target pncA probably does not contribute to PZA-induced M. tuberculosis tolerance, as pncA downregulation is not observed under PZA exposure (27).

Ethionamide.

Under ethionamide (ETH) stress, M. tuberculosis induces a transcriptional response similar to the one observed under INH stress, which is not surprising as both drugs target mycolic acid synthesis. The TA responses following exposure to INH and ETH are clustered and different from the TA responses induced by RIF or streptomycin (44). Of the 70 genes differentially expressed upon ETH-induced stress, 39 genes behaved similarly under INH stress (92). Nevertheless, M. tuberculosis strains that cannot convert the INH prodrug and thus do not develop a stress response under INH exposure still develop a stress response under ETH exposure (94). The stress response observed under ETH exposure is thus, at least in part, unique to exposure to ETH. Of the genes induced by ETH, 15 were associated with lipid metabolism, and 18 were associated with cell wall synthesis and cell processes (92). Downregulation of ethA, the gene that encodes the ETH drug-activating enzyme, does not seem to play a role in ETH tolerance, as such downregulation is not observed in the first 14 days of ETH exposure (27).

Aminoglycosides.

Aminoglycosides (AGs) inhibit protein synthesis by binding rRNA. Upon AG pressure, whiB7 overexpression affects genes involved in drug efflux (tap) and ribosomal protection (including eis) to reduce sensitivity to all AGs used for the treatment of TB (amikacin, kanamycin, capreomycin, and streptomycin) (30, 39). In contrast to the irreversible mutations in the eis promoter regions that confer resistance to capreomycin, kanamycin, and amikacin (117–120), WhiB7 overexpression and eis upregulation can generate a temporary state of reduced sensitivity to AG (117–120). SigA may play a role in this process as it binds to WhiB7 to form a WhiB7-SigA promoter complex. SigA also interacts with the DevR/DosR dormancy regulon, resulting in decreased metabolic rates, growth arrest, and shifts in carbon, energy, and lipid metabolism (121). While these changes affect nonresponsiveness to drugs, they have to date been reported only under hypoxic and chemical stresses (21).

Another pathway for AG tolerance may be through the sigma factors SigE and SigB, as it has been shown that SigE and SigB loss-of-function mutants showed reduced survival in vitro upon exposure to streptomycin (38).

TA systems are reported to participate in streptomycin tolerance, with the TA response to streptomycin being similar to the TA response following starvation, indicating that M. tuberculosis could make use of several fundamental stress pathways to cope with a wide variety of stresses (44).

Bedaquiline.

Upon exposure to BDQ, which inhibits ATP synthase, M. tuberculosis rapidly adapts its metabolism. The low bactericidal activity of BDQ during the first days following BDQ exposure likely reflects the slow killing that is associated with ATP inhibition, further compromised by a BDQ-induced stress response that triggers metabolic shifting and metabolic slowdown (Fig. 5) (37, 122, 123). Of the 10 known universal stress proteins in M. tuberculosis, 7 are upregulated upon BDQ exposure. In addition, the overexpression of several transcription factors (including SigG, Rv0324, and Rv0880) generates BDQ tolerance by inducing a BDQ stress response (37, 123). Of these, Rv0324 is also overexpressed upon exposure to capreomycin and moxifloxacin, suggesting potential antagonism. Interestingly, Rv0880 expression is reduced upon exposure to pretomanid (PTM), suggesting synergism between pretomanid and BDQ (123).

FIG 5.

Bedaquiline (BDQ)-specific tolerance in Mycobacterium tuberculosis. The effects of BDQ exposure in the absence of tolerance (red) and the presence of tolerance (green) consist of multiple adaptations: efflux pump activity that occurs even in the nontolerant state and may be increased in the tolerant state (A), metabolic slowdown to reduce ATP dependency (B), and promoting ATP synthesis to overcome the ATP shortage induced by BDQ, either directly through the overexpression of genes encoding subunits of the ATP synthase machinery or indirectly by the stimulation of oxidative phosphorylation (C). The latter can be performed by increasing the synthesis of components used during oxidative phosphorylation (e.g., cytochromes) and/or by promoting trehalose metabolism and TCA cycle activity to increase the production of the energy-rich metabolites that power oxidative phosphorylation and generate the proton motive force that is required for ATP synthesis.

The metabolic shifts that drive BDQ tolerance are likely to involve multiple pathways. Reduced metabolic activity through increased expression of the dormancy regulon could contribute to BDQ tolerance, as it reduces the need for ATP synthesis. Increased ATP synthesis through the overexpression of genes encoding ATP synthase subunits, genes involved in oxidative phosphorylation (such as cytochrome-associated genes), and genes associated with the TCA cycle (which generates the reducing agents that enter oxidative phosphorylation) all illustrate how M. tuberculosis employs metabolic shifting to overcome the BDQ-driven pressure on ATP synthesis (37). In addition, M. tuberculosis shifts central carbon metabolism in response to BDQ by upregulating treS, which introduces a shift toward trehalose-fueled metabolism, thereby maintaining ATP and NADPH levels and further contributing to BDQ tolerance (124). The role of trehalose as an internal carbon source in BDQ tolerance is supported by the findings that treS deletion mutants incapable of generating a trehalose-fueled central carbon metabolism showed rapid depletion of ATP and were 144-fold more susceptible to BDQ (124).

Finally, efflux pump expression upon BDQ exposure may contribute to the development of a transient state of BDQ tolerance. This is not surprising given that mutations in Rv0678, a transcriptional repressor of the MmpS5-MmpL5 efflux pump-encoding genes, confer resistance to BDQ and cross-resistance to clofazimine (125). The observation that the efflux pump inhibitors verapamil and reserpine increase BDQ activity in wild-type H37Rv M. tuberculosis suggests that efflux pump activity introduces a basal level of intrinsic BDQ resistance in M. tuberculosis (125, 126). While BDQ-induced overexpression of efflux pump activity could result in BDQ tolerance, this has, to our knowledge, not yet been reported.

DISCUSSION

Two decades ago, Wallis et al. suggested that drug tolerance (defined as delayed in vitro killing of M. tuberculosis) may be of clinical importance, as the presence of drug-tolerant populations in pretreatment isolates was associated with relapse or treatment failure. They also stated that drug therapy can trigger drug tolerance in vivo, as clinical isolates showed increased drug tolerance upon prolonged drug exposure (7). A better understanding of the mechanism that M. tuberculosis uses to generate a state of drug tolerance to either multiple drugs or a specific drug or drug class could result in the identification of novel agents aimed at preventing the development of a drug-tolerant phenotypic state in M. tuberculosis, thus creating a new approach to treatment shortening for TB. Moreover, understanding drug-specific tolerance mechanisms could contribute to improved drug regimen designs where antagonistic drug combinations are avoided.

Since this seminal paper, several experimental studies have provided evidence for the existence of a drug-tolerant phenotype of M. tuberculosis and have explored the mechanisms of general tolerance to multiple drug classes and tolerance to specific drugs. Tolerance, a temporal and reversible phenotypic state of M. tuberculosis that develops in response to stress (including drug pressure), results from adaptive mechanisms that allow M. tuberculosis to survive for a longer period when exposed to stress. The main general mechanisms that mycobacteria use to acquire a temporary drug-tolerant state are slowdown of its metabolism, rearrangement of core metabolic pathways to circumvent the effects of the drugs, upregulation of efflux pumps, and cell wall restructuring. These mechanisms are intertwined through shifting of central carbon pathways, lipid metabolism, and redox homeostasis via transcriptional and posttranscriptional regulators such as sigma factors, WhiB regulons, and toxin-antitoxin systems. With regard to drug-specific tolerance responses, most studies have focused on INH and BDQ, while less attention has been paid to the other drugs used for the treatment of TB.

Potential Clinical Utility and Topics for Further Study

Several studies have made important observations that highlight the clinical relevance and the potential application of an improved understanding of tolerance. First, the presence of tolerance can increase the risk of development of resistance, as has been demonstrated in (in vitro and in vivo) experimental evolutionary studies in E. coli and Staphylococcus aureus (127, 128). In M. tuberculosis, the observation that cell wall thickness increases from pansusceptible to multidrug-resistant and extensively drug-resistant (XDR) M. tuberculosis strains suggests that transient drug tolerance can precede fixed drug resistance. Cell wall thickening results in phenotypically tolerant mycobacteria that are able to survive longer in the presence of antituberculosis drugs. Under these conditions, M. tuberculosis is exposed for prolonged periods to an environment that is mutagenic, thus promoting the acquisition of drug resistance (129). The hypothesis of tolerance preceding drug resistance is further supported by the observation of a trehalose shift in clinical XDR isolates compared to susceptible isolates and the finding that prpR mutations are enriched in resistant strains compared to susceptible strains (88, 124). Second, the discovery that supplementation of INH-containing regimens with drugs that target redox metabolism (such as the malarial drug chloroquine) increases the killing of INH-tolerant bacilli and reduces relapse rates highlights the opportunity to explore the repurposing of existing drugs to optimize the treatment of TB by reducing the development of tolerance (100). Third, the finding that the proportion of RIF-induced tolerant bacteria is inversely related to the dose of RIF used supports the use of RIF high-dose treatment regimens, as this may reduce the development of RIF tolerance (107). Fourth, PZA is highlighted as a complementary and synergistic companion drug in various treatment regimens (113, 115, 116, 130). An understanding of how PZA targets tolerant bacterial populations would be of great value for the development of drugs that target tolerant M. tuberculosis (131). Finally, the observation that the development of tolerance to BDQ may be inhibited by the administration of BDQ in combination with PTM (123) is of interest, especially given the recently approved Nix-TB regimen (PTM plus BDQ and linezolid) for the treatment of difficult-to-treat MDR and XDR strains (132, 133).

A better understanding of the tolerance mechanisms used by M. tuberculosis under drug pressure could thus result in the identification of novel or repurposed drugs that enable intelligent regimen design to avoid tolerance to core drugs used in the treatment of TB. The prevention of the development of drug tolerance could complement current approaches to treatment shortening, as the prevention of the acquisition of tolerance could reduce the duration of infectiousness, decrease relapse rates, and reduce the risk of the development of drug resistance. A number of interesting genes and hypothetical proteins have already been identified (65, 92–94, 134), and future studies could result in the identification of additional novel agents. For example, one could investigate the use of WhiB3 and WhiB7 inhibitors to increase the efficacy of drugs that are currently prone to the effects of WhiB3 and WhiB7, or one could employ antibiotics that cannot be detected by these WhiB regulators (30, 31). Cotreatment with mycobacterial efflux pump inhibitors (such as verapamil) could limit intrinsic resistance induced by efflux pumps (87, 126, 135–138), and drugs that target sigma factors may also be of interest for the optimization of TB treatment (139). Finally, the use of repurposed drugs that target slow-growing and dormant bacteria, such as nonsteroidal anti-inflammatory drugs and nitroimidazopyrans, should be investigated (140, 141).

Limitations of Current Studies

The experimental studies that we review share several limitations. First, while the hypothesis of the existence of a tolerant phenotypic state in M. tuberculosis is driven by the clinical observation of heterogeneous treatment response dynamics between patients, most studies have been performed on laboratory M. tuberculosis strains. The few studies that used clinical samples to study the adaptation of M. tuberculosis to antibiotic stress included a small number of patients (n < 12) (59, 142–145), limiting the ability to examine the diverse responses that M. tuberculosis exhibits in response to drug pressure and restricting assessments of the relative importance of pathogen and host factors such as HIV status, smoking status, and the presence of lung cavities in the response to treatment. Second, most studies performed transcriptomics experiments. Studying a transient phenomenon such as tolerance using transcriptomics is difficult, as the results of RNA expression profiling analyses likely depend on environmental and growth conditions, and transcriptional profiles observed under culture conditions might not be representative of what is truly happening in the host. Conflicting data between transcriptomic studies may thus be due to the conditions under which the experiments were performed. For example, the conflicting data on the differential expression of katG upon INH exposure could be due to the study of M. tuberculosis in the stationary phase in one study compared to the study of extracted RNA from sputum samples conserved at −80°C within 2 h after collection in another study (27, 101). Third, disentangling whether the observed mycobacterial transcriptional responses upon drug pressure are generated by drug pressure, immune pressure, or the selection of preexisting bacterial subpopulations is challenging. Furthermore, other environmental triggers such as biofilm formation, hypoxic stress, and nutrient starvation have been associated with drug tolerance and might share underlying mechanisms provoked by drug exposure (146, 147). Further experimental work investigating the role of different microenvironments in the selection versus the creation of slow-growing subpopulations upon drug pressure is thus of interest. Fourth, most studies applied a single omics approach, mainly transcriptomics. This limits our understanding of the relative contributions of genetics, epigenetics, and proteomics to tolerance. Methylation-mediated pathway regulation has been shown to play a role in antibiotic stress survival in E. coli (148). The contribution of epigenetic mechanisms to the acquisition of tolerance in M. tuberculosis has not yet been investigated, as most epigenetic research in the field of tuberculosis has focused on host epigenetics and has linked host epigenetics to M. tuberculosis survival in the host and drug resistance (149–151). Few studies have assessed the complement, i.e., how the M. tuberculosis epigenome is used as a mechanism to evade bacterial killing (149–151). In addition to epigenetics, proteomics could be important for the study of tolerance, as the functional changes in expressed proteins ultimately result in the phenotypically tolerant state (152). Nevertheless, only a few investigators have applied a proteomics approach to the study of drug tolerance in M. tuberculosis (62–64, 153). Finally, while mechanisms such as thickening of the cell wall and increased efflux pump activity are sensu stricto mechanisms of resistance as they increase the MIC, we include these types of innate or intrinsic resistance under mechanisms that generate a drug-tolerant phenotype as they are very different from the classical mechanisms of drug resistance conferred by genomic variants in resistance-conferring genes such as rpoB, katG, gyrA, and others.

CONCLUSION

When the bacillus is under drug pressure, a myriad of mechanisms is used by M. tuberculosis to adapt to the stress of exposure to drugs and acquire a drug-tolerant M. tuberculosis state across drug classes or specific to one drug. The main mechanisms are metabolic slowdown, metabolic shifting, cell wall thickening, and upregulation of efflux pumps, whereas drug-specific tolerance mechanisms aim to overcome or counteract the action of the drug. Future research, including multi-omics approaches and clinical translational studies of the fundamental processes of M. tuberculosis tolerance, is needed. The knowledge generated could then result in the identification of novel strategies that target or prevent the acquisition of drug-tolerant M. tuberculosis populations. This could accelerate bacillary clearance, which in turn would lead to a shorter duration of infectiousness, a lower risk of the acquisition of drug resistance, and reduced inflammation and lung tissue damage. Such novel tolerance-targeting approaches could complement the treatment-shortening strategies that are already being investigated.

Biographies

Sander N. Goossens is a Ph.D. student at the University of Antwerp under the supervision of his promotor Prof. Annelies Van Rie. He obtained his master’s in cell, genetic, and molecular biology at Vrije Universiteit Brussel (VUB), Belgium, with greatest distinction in 2014. His thesis was on the application of the then highly novel CRISPR/Cas technology for the generation of knockout mouse embryonic stem cell lines that could be used for studying early embryonal development. Before starting his Ph.D., he was a biology and physics teacher at the International Montessori School in Brussels. Since his childhood, he has been passionate about the natural processes that govern our world and is particularly interested in how they translate clinically.

Samantha L. Sampson studied Medical Biochemistry and obtained her Ph.D., titled “Mycobacterium tuberculosis: Genetic and Phenotypic Comparison,” at Stellenbosch University (South Africa). After being a postdoc at the Harvard School of Public Health, she became a research associate at the Centre for Molecular Microbiology and Infection, Imperial College London, United Kingdom, followed by a position as an associate professor and eventually a professor at the Faculty of Medicine and Health Sciences at the University of Stellenbosch. Prof. Sampson has been involved in tuberculosis research for over 20 years. Her interest in tuberculosis, an infectious disease that not only presents intriguing biological questions but also has major public health and socioeconomic impacts in resource-poor settings (including South Africa), is focused on advancing our understanding of the host-pathogen interface in tuberculosis disease, with a particular interest in drug tolerance and persistence.

Annelies Van Rie studied pediatrics at the University of Leuven, Belgium, and obtained her Ph.D. on the “Molecular Epidemiology of Drug-Resistant Tuberculosis” at the University of Stellenbosch, South Africa. She started as a Medical Epidemiologist at Global Medical Affairs, Aventis Pasteur, became Assistant Professor of Epidemiology (and eventually full professor) at the School of Public Health, University of North Carolina at Chapel Hill, and is now Professor of Epidemiology at the University of Antwerp, Belgium. For over 20 years, her work has focused on clinical, epidemiological, and translational tuberculosis research, with a special emphasis on the molecular epidemiology of drug-resistant TB, evaluation of new diagnostics, optimized use of existing and new anti-TB drugs, and implementation of research to measure the impact of public health interventions. The goal of her research activities is to improve TB management and control in resource-poor high-burden as well as low-TB-burden settings.