Mycobacterium tuberculosis Metabolism

ABSTRACT

Mycobacterium tuberculosis is the cause of tuberculosis (TB), a disease which continues to overwhelm health systems in endemic regions despite the existence of effective combination chemotherapy and the widespread use of a neonatal anti-TB vaccine. For a professional pathogen, M. tuberculosis retains a surprisingly large proportion of the metabolic repertoire found in nonpathogenic mycobacteria with very different lifestyles. Moreover, evidence that additional functions were acquired during the early evolution of the M. tuberculosis complex suggests the organism has adapted (and augmented) the metabolic pathways of its environmental ancestor to persistence and propagation within its obligate human host. A better understanding of M. tuberculosis pathogenicity, however, requires the elucidation of metabolic functions under disease-relevant conditions, a challenge complicated by limited knowledge of the microenvironments occupied and nutrients accessed by bacilli during host infection, as well as the reliance in experimental mycobacteriology on a restricted number of experimental models with variable relevance to clinical disease. Here, we consider M. tuberculosis metabolism within the framework of an intimate host-pathogen coevolution. Focusing on recent advances in our understanding of mycobacterial metabolic function, we highlight unusual adaptations or departures from the better-characterized model intracellular pathogens. We also discuss the impact of these mycobacterial “innovations” on the susceptibility of M. tuberculosis to existing and experimental anti-TB drugs, as well as strategies for targeting metabolic pathways. Finally, we offer some perspectives on the key gaps in the current knowledge of fundamental mycobacterial metabolism and the lessons which might be learned from other systems.

INTRODUCTION

The notion that metabolic capacity might correlate with lifestyle has largely been confirmed by comparative genomics analyses of different bacterial species: free-living bacteria tend to have larger, more complex genomes than pathogens and symbionts (1), with obligate intracellular organisms often representing the extreme limit of small genome size (2). This correlation exists also in Mycobacterium, a genus which includes environmental organisms as well as a select group of animal and human pathogens—primarily of the Mycobacterium tuberculosis complex (MTBC) (3)—whose genome sizes and compositions relate broadly to habitat and host range (4, 5). Most notable among these from a public health perspective is M. tuberculosis, the causative agent of tuberculosis (TB), an airborne infectious disease which continues to result in over 10 million new cases and more than 1.5 million deaths globally each year (6). This is despite the widespread use of a neonatal anti-TB vaccine (7) and the existence of an effective combination chemotherapy (8), the insufficiency of which points to systemic failures in TB diagnosis and treatment (9, 10), as well as the complexities encountered in eliminating an organism that has coevolved with its human host (11) and, throughout modern history, has proven adept at exploiting urbanization (12), economic disparity (13, 14), societal upheaval (15), and incarceration (16–18) in maintaining a global prevalence characterized by high-burden endemic regions (6).

M. tuberculosis is an obligate human pathogen; the bacillus can infect other animals, but evidence for zoonotic transmission (back) to humans from these “spillover” hosts (11) is scarce (19). Consistent with its near exclusive association with a single “maintenance host” (11), the M. tuberculosis genome has undergone significant downsizing during its evolution from a generalist ancestor. Nevertheless, M. tuberculosis retains a large metabolic repertoire, much of which is shared with other mycobacteria with very different lifestyles (20–22); moreover, horizontal gene transfer during the early evolution of the MTBC has enabled the acquisition of additional functions that are critical for pathogenicity (23–25). In other words, the organism has harnessed, adapted, and even augmented the metabolic pathways of its environmental ancestor in evolving the capacity to persist and propagate within its obligate human host (21, 22, 26).

The development of specialist M. tuberculosis sublineages from more generalist precursors (11, 27)—in some cases through the inactivation of certain genes (28)—appears, therefore, to signal natural evolution driven by key structural changes in human population demographics and geographies. These include the shift to urbanization, the introduction and widespread use of anti-TB drugs and, possibly, BCG vaccination (29). Greater host population density and the emergence of comorbidities such as diabetes and HIV might additionally have favored the selection of more virulent strains (30, 31), though this inference is not always clear-cut (32). A common theme, however, is that gain or loss of different metabolic functions appears to be a major factor in the modern evolution of the M. tuberculosis bacillus (21, 22, 28, 33–36).

We noted previously (37) that the topic “M. tuberculosis metabolism” is very broad, encompassing all the dynamic processes and chemical transformations which enable bacillary cellular function and physiology, including those involved in maintaining structural integrity, energy production, generation of biomass, and elimination of waste. It is not possible, therefore, to cover these in any detail in a single review. Instead, this article will attempt to situate M. tuberculosis metabolism within the framework of an intimate host-pathogen coevolution while providing some insights into the special challenges faced in elucidating metabolic function (as distinct from metabolic capacity) in a pathogen with no known reservoir outside its obligate human host. Where relevant, we will highlight recent advances in our understanding of different aspects of the metabolism of M. tuberculosis, focusing primarily on those which represent unusual adaptations or departures from the model intracellular pathogens; that is, the key observations implicating specific metabolic functions in mycobacterial pathogenicity and the host-pathogen interaction. In addition, given the strong interest in the development of new (shorter) combination regimens for both drug-susceptible and drug-resistant TB, we will consider the impact of metabolism on the susceptibility of M. tuberculosis to existing and experimental anti-TB drugs. As with other summaries of this subject, the reader is advised not to consider this a standalone piece. Rather, any of a number of specialist reviews should be consulted for further details on those aspects covered only superficially here, as well as for different perspectives on the key gaps in current knowledge of fundamental mycobacterial metabolism and the implications thereof for TB prevention and new TB drug development (for a list of suggested readings, please see Table 1).

TABLE 1.

Selected readings

Metabolic pathway	Recommended reviews
Cell wall biosynthesis and maintenance	253, 254, 329
Energy metabolism and respiration	230, 231
Central carbon metabolism	26, 66
Nitrogen metabolism	330–333
Sulfur metabolism	334, 335
Metallobiology	275, 336
Cofactors (molybdenum, coenzyme A, riboflavin, coenzyme F420, cytochrome P450, vitamin B12, etc.)	145, 337–340
Nucleic acid metabolism	341–343
PE/PPE proteins	77

MYCOBACTERIAL METABOLISM AND HUMAN TB DISEASE

Although the precise duration of the M. tuberculosis-human coevolution remains subject to contention (11)—estimates range from fewer than 6,000 years (38) to around 70,000 years (39)—there is general acceptance that the interaction extends across millennia. It was striking, therefore, when the first whole-genome sequence of M. tuberculosis revealed the genetic potential for synthesis of all essential amino acids, vitamins, and cofactors, with a disproportionate number of pathways dedicated to lipid and fatty acid metabolism (40). This large metabolic capacity (often conflated with “metabolic flexibility” [41–43]) represents a key departure from many other obligate pathogens which tend to rely on the host to complement their limited (and often diminishing) metabolic repertoires (44) and is critical to understanding the defining characteristics of this organism and the disease it causes (26, 45–48). Such characteristics include (i) long-term coevolution of M. tuberculosis with humans (38, 39), leading to sympatric associations of specific lineages with geographically distinct human populations (11, 49), (ii) the ability to persist for months to decades in a subclinical state (in some cases characterized by undetected disease [50]) which might develop later to cause active TB (26, 47, 51–53), (iii) the inertness to clearance by the host immune response or eradication by antibiotics (54, 55), (iv) the requirement for extended-duration combination therapy to effect symptom-free cure (8, 56) and, linked to that, (v) the difficulties inherent in developing new drugs and combination regimens to accelerate therapy (57, 58), and (vi) the capacity to acquire—and adapt to—drug resistance mutations that might otherwise be expected to incur a crippling fitness cost, particularly where sequential mutations confer multidrug resistance (59–61). Each of these characteristics reinforces the durability of M. tuberculosis in preserving its special niche within humans and, by implication, the challenges inherent in disrupting this association, whether by vaccines, chemotherapy, or other interventions (Fig. 1).

FIGURE 1.

The metabolic capacity of M. tuberculosis is critical to understanding key features of mycobacterial pathogenicity, which include (i) long-term coevolution with humans, (ii) the ability to persist for months to decades in a subclinical state, (iii) inertness to clearance by the host immune response or eradication by antibiotics, (iv) the requirement for extended duration combination therapy to effect symptom-free cure and, linked to that, (v) the difficulties inherent in developing new drugs and combination regimens to accelerate therapy, as well as (vi) the capacity to acquire—and adapt to—multiple drug resistance mutations without incurring a crippling fitness cost.

Pathogenicity and Metabolism

That physiology and pathogenicity cannot be separated in an obligate pathogen such as M. tuberculosis seems axiomatic (26, 62). However, the relevance of this precept to efforts to counter TB might be less evident and so bears restating: an understanding of mycobacterial metabolic capacity supports the inference that M. tuberculosis is a “survival machine,” able to exploit its human host as both a source of nutrients for increased biomass and a reservoir for long-term persistence (26, 37, 63). Reinforcing this premise is the observation that the infecting population does not necessarily drive a metabolic program aimed at maximizing growth rate (64, 65) or biomass (26, 37, 63, 66). Instead, as observed by others (66), the frequent requirement for bacilli to exit the standard bacterial cell cycle owing to the pressures imposed by the host immune response and, in the modern history of M. tuberculosis, antimycobacterial chemotherapy, demands the capacity to maintain viability in a state of rapidly reactive and reversible metabolic quiescence for extended periods.

While evidence exists for decades-long survival of bacilli in an asymptomatic host (67, 68), most cases of clinical disease are thought to arise within 1 to 2 years of infection (53, 69). During this period, those bacilli which survive the attempts at clearance by the host’s innate immunity generally maintain a state of paucibacillary, asymptomatic infection in which host lipids and mycobacterial antigens are accumulated within alveoli in the lung apices. There is a spread of bacilli which, depending on the preferred model, is thought to occur via bronchi (69) or via the lymphatics (70). And, although most new lesions regress spontaneously, necrosis can result in caseous pneumonia, a heterogenous outcome which can be paucibacillary or multibacillary, with or without significant neutrophil involvement. Importantly, this manifestation—known as post-primary TB disease—depends on the acquisition of systemic immunity by the infected host. Its sequelae are considered the key determinant of disease outcome: softening and fragmentation of the caseous pneumonia triggers its expulsion by coughing, in turn leaving a hole which can become the site of active bacillary growth and the primary source of organisms for transmission; in contrast, those sites which are not removed by coughing are sources of ongoing inflammation, ultimately leading to fibrocaseous disease (69). Though central to disease pathogenesis, these events in post-primary TB, in particular, the development during latent infection of an asymptomatic obstructive lobular pneumonia, have not been adequately studied, owing to the unavailability of suitable clinical specimens and the failure of animal models to recapitulate many of the key features. Inevitably, mycobacterial metabolism under these conditions, too, remains very poorly understood.

The quiescent state adopted by M. tuberculosis is distinct from “true dormancy,” a very specific physiological and morphological condition often involving sporulation, for which the bacillus appears to lack the necessary genetic machinery (71). Physiologically, though, it is interesting to note that a deficiency in α-crystallin—a major TB antigen (72), bacterial protectant, and chaperone protein which has been implicated in long-term bacillary survival (55)—is associated with hypervirulence in vivo (73). This seems to imply a disturbance of the natural host-pathogen interaction where a genetic lesion has deprived the bacillus of its natural capacity for persistence. The observation, noted previously (37), that mutations in regulatory genes are often associated with hypervirulence (74–76) appears to buttress this conclusion. So, too, does recent evidence (28) implicating disrupted secretion of the large class of Mycobacterium-specific PPE-MPTR and PE_PGRS proteins (whose functions remain largely unknown [77]) in enhanced growth and virulence of modern Beijing lineage strains in vivo. The inferred centrality of a conserved host-pathogen interaction in TB pathogenesis is reinforced in the “hyper conservation” of T cell epitopes (39, 78), a feature which might be integral to the development of post-primary TB (69, 79).

At an epidemiological level, the inference that M. tuberculosis is hard-wired for persistence is supported by the fact that the number of active TB cases globally (∼10 million) (6) is dwarfed by the estimated annual incidence (thought to exceed tens of millions [53]) and global prevalence of infections (estimated at ∼25% of the world’s population [80] but subject to contention [53]). As Rhee and colleagues have cogently argued, this entails that the bacillus must occupy a range of physiological states book-ended by “unrestrained replication” and “perpetual quiescence” (26), extremes which are incompatible with the natural history of the species (11, 49). In turn, the notion that M. tuberculosis cellular function defaults to persistence immediately suggests the corollary: host status, whether genetic (81), immunological (82), nutritional (83, 84), environmental (85, 86), or a combination of some or all of these, is critical in determining who develops, and succumbs to, TB disease (a seemingly obvious conclusion, perhaps, but one which reinforces the potential pitfalls in attempting to define—and quantify—mycobacterial “virulence” as a solely bacterial property).

The intimate, dynamic, and durable interaction between M. tuberculosis and humans defines the framework within which any new anti-TB interventions must be considered. That is, tackling any single aspect—for example, drug regimens to treat active cases, novel preventative chemotherapies or vaccines to forestall new disease, or strategies to block transmission—is unlikely to prove sufficient on its own (87). Importantly, it also underscores an acknowledged limitation of many studies which aim to elucidate fundamental aspects of M. tuberculosis metabolism: namely, that the potential disconnect separating bacillary metabolic functions active during host infection (whether in the presence or absence of manifest TB disease) versus those actuated—and therefore described—in vitro and/or in vivo in the various disease models risks a growing phenomenology of metabolic capacity distinct from direct evidence of metabolic function in the context of the complete biological system which defines TB (37, 88).

Investigating Metabolism in Vitro

While the majority of genome-wide mutagenesis studies recognize the importance of, and even exploit, the applied in vitro conditions (or infection model) in inferring conditional essentiality (89–93), the impact of growth conditions on experimental observations has often been ignored in fundamental mycobacteriology. This is despite an increasing number of reports of the profound impact of the growth medium and various additives on the described phenotypes (94–97, 344). Aside from auxotrophy (92, 98, 99), perhaps the most illustrative example of how even apparently absolute concepts such as essentiality (100) can be condition-dependent is provided by the paradoxical tolerance of cell wall-targeting antibiotics in cell wall-deficient L-form bacteria (101), a phenotype which demonstrates the potential for wholesale loss of a fundamental macromolecular structure under the appropriate conditions—in that case, in vivo residence in a suitably osmoprotective intracellular environment.

Similarly, practical considerations tend to favor the use of culture conditions that enable optimal (that is, the most rapid) growth. For M. tuberculosis, this generally means aerobic incubation in standard medium (usually Middlebrook 7H9 base or equivalent) comprising glycerol and glucose as the primary carbon sources (102). The disconnect between in vitro and in vivo models used in experimental investigations of mycobacterial metabolism and those environments which are likely to prevail in the natural human host is not unique to M. tuberculosis (103). Therefore, in addition to appreciating that logarithmic aerobic growth represents just another condition and that mycobacterial physiology in this state will necessarily influence assessments of gene essentiality or metabolic pathway vulnerability, there is a pressing need to develop experimental systems that target slow-growing and nongrowing states, perhaps in the context of more advanced cellular scaffolds (104). Though often performed in the context of TB drug discovery, the application of advanced analytical technologies to determine the precise molecular content (proteins, lipids, metabolites) of a number of environments, including macrophage organelles (105), lipid droplets (106), and tissue samples from animal models and patients (107, 108), represents an exciting area of development and might be complemented by recent advances in generating more complex tissue culture models (109).

Many researchers have made pioneering efforts to develop other complex models of M. tuberculosis growth and survival, and the strategies adopted to model environments that are far less favorable for unchecked generation of biomass have yielded important insights into metabolic function under the applied conditions, as well as revealing condition-selective antimycobacterial compounds. These include manipulating the pH of growth media, the inclusion of peroxide and/or nitric oxide stresses, oxygen and nutrient deprivation, and combinations thereof (110–115), many of which have been applied in screening programs aimed at identifying compounds with condition-specific activities or those able to inhibit survival of nonreplicating bacilli (116, 117). Again, while each is intuitively attractive, the difficulties inherent in translating inferences derived from these in vitro systems to human TB disease remain daunting; the field urgently requires a proof-of-concept demonstration of enhanced therapeutic efficacy achieved using a novel small molecule whose development has been prioritized owing to its condition-selective activity in one of more in vitro models.

Moving to More Complex Models

Like for many other pathogens, investigations of cellular functions in the context of the host-pathogen interface of M. tuberculosis have tended to focus primarily on the effectors of immunity (host) or virulence (bacillus), with the metabolic consequences of infection often ignored (44). Systems biology and its tools, including the analysis of labeled metabolites (118–122) and other mass spectrometry-based technologies (123–127), have eroded this bias, enabling key insights into the biochemical compositions and metabolic fluxes which characterize M. tuberculosis bacilli under different conditions and in different infection models (128). These and related approaches have also been used to elucidate the regulators of metabolic adaptation and function (129–131), the mechanisms of action of existing and experimental anti-TB drugs (43, 131–134), in assigning functions to proteins of unknown or predicted function (133) and in functional genomics of clinical M. tuberculosis isolates (36). Except for sputum bacilli, though, the number of studies directly investigating M. tuberculosis metabolic function and physiological state(s) in situ during host infection remains very small (135–137).

In some cases, results from experimental systems in vitro and in vivo have been combined with analyses of human TB pathology to construct sophisticated models of mycobacterial metabolism (138, 139). Though compelling, the in vivo relevance of these is largely untested; for this reason, determining the metabolic and physiological states adopted by the tubercle bacillus in the infected human host represents the ultimate, though daunting, scientific goal. Knowledge that the functions of approximately 40% of the M. tuberculosis proteome remains predicted or obscure (140, 141)—so, too, the speculative reasons the bacillus possesses numerous examples of apparently redundant or duplicate pathways (142–149)—exacerbates the size of the task. Moreover, growing evidence indicating key functional differences between (and even among different laboratory stocks of) the laboratory strain, M. tuberculosis H37Rv, clinical M. tuberculosis isolates, and other species of the MTBC (33–35, 150, 151), implies that diverse phenotypes will inevitably be recorded when different species, and even strains (152), are tested in the same assay. That is, while prevailing approaches might enable some generalizable insights into core M. tuberculosis metabolism, the next phase demands innovative approaches to study bacilli of different genotypes in the context of natural host infection (153). Here, understanding the features which differentiate infection outcomes in sympatric versus allopatric hosts (154) might provide some important insights (151).

Of course, the fact that the diversity which characterizes M. tuberculosis lineages and strains is matched by the genetic and phenotypic variety of host immune cells further complicates these analyses. For example, different macrophages and macrophage-like cell lines are commonly explored to understand mycobacterial function in the intracellular environment (155); for TB, these are usually immortalized human and mouse cell lines (e.g., THP-1s, J774s), mouse bone marrow-derived macrophages, and human peripheral blood mononuclear cells (156), with some recent reports proposing alternative, more complex, systems (109, 157). Macrophages themselves are functionally diverse, however, and follow different developmental trajectories and respond to a variety of activation signals (158). Consequently, there are significant risks inherent in a rigid dependence on standard models, a pitfall underscored by recent observations that alveolar macrophages might offer a more favorable intracellular environment for M. tuberculosis survival and dissemination than interstitial macrophages (159, 160). Moreover, evidence that a shift to aerobic glycolysis in infected cells can enable proliferation of intracellular bacilli (161, 162) requires that caution should be exercised in drawing conclusions about mycobacterial metabolic function from experiments utilizing the different macrophage lines. To enable rapid anabolism, proliferating cells (as is the case for the various cell lines) naturally drive a strong Warburg metabolism which is independent of intracellular infection but must inevitably impact the metabolic response of the infecting bacillus (161). A further complication is that this effect can be species-specific, with recent evidence highlighting differential metabolic responses in human versus mouse macrophages (163). In this light, the proposal that intracellular pathogens might have evolved to take advantage of specific immune cell subsets locked into promicrobial metabolic programs (44) seems apposite. Therefore, special emphasis must be placed on the need for much greater functional and ontogenic information about the host cells in which bacilli preferentially reside during infection. And this should be coupled with a better understanding of the extent to which immunometabolic responses of host cells to M. tuberculosis are cell-type specific, host species specific, and/or strain specific (45, 161).

M. tuberculosis Metabolism and the TB Disease Life Cycle

Other than during transient periods of transmission, when bacilli are released into the environment in droplet aerosols, the entire M. tuberculosis life cycle occurs within the infected host (26, 164). This concept—encapsulated in the notion of humans as the “maintenance host” (11)—is critical to understanding M. tuberculosis pathogenicity: the ability to infect, persist, cause disease, and trigger production of infectious aerosols demands the capacity for survival and propagation in different environments and under a variety of stresses. Current knowledge of M. tuberculosis infection reinforces this notion. Most experimental models, ranging from intramacrophage systems in vitro (165, 166) through small animals (159, 167, 168) to nonhuman primates (NHPs) (169, 170), indicate that bacilli are able to replicate rapidly immediately after infection, with evidence of early dissemination in NHPs (170) until host-mediated killing results in a decline in replication rates and total bacillary load. Even then, it appears from the mouse model (167) that replication might continue in discrete microenvironments throughout infection (68), a possibility recapitulated in genomic data from bacilli isolated from NHPs which revealed the acquisition of chromosomal mutations during subclinical disease (171). Moreover, although its relevance to in vivo infection remains to be established, the observation that M. tuberculosis replicates rapidly inside dead macrophages (172) further argues that the transient growth rates (and fates) of infecting bacilli in different lesions in the same host are likely to be dynamic and unlinked.

The heterogeneity which characterizes host infection supports a model in which infecting bacilli must establish an early foothold within the new host before seeding a number of discrete lesions, the fate of which appears to be determined for each lesion at a local level, independent of events in other microenvironments (173). For M. tuberculosis as a species, each new infection represents a potential dead-end if transmission is prevented. Therefore, long-term survival within a host—and by any mechanism—appears to represent the best hedge against this risk. Loss of immune control at any stage during the host’s lifetime might then allow for stimulation of the destructive pathology that could enable (or enhance) infectious aerosol release, renewing the infection cycle.

This is the lens through which mycobacterial metabolism must be viewed. Transient opportunity for generation of biomass means that preservation of viability under varied and multiple stresses represents the driving metabolic imperative. While contrary to popular views which conflate bacterial virulence with rapid growth, the enforced (and sustained) quiescence which likely characterizes human M. tuberculosis infections is a feature of most bacteria in their preferred niches, whether pathogens or not (103). The apparently “surprising” retention of a multifunctional metabolism in a successful obligate pathogen appears then to be just the opposite: metabolic flexibility is, instead, precisely the attribute which enables M. tuberculosis to occupy its preferred environment—the human host—without recourse to a separate (environmental) reservoir. For M. tuberculosis, the absence of any competing microorganisms within its “ultra-narrow ecologic niche” (66) additionally ensures a privileged, albeit transiently hostile, space in which replication can occur subject to prevailing conditions, largely determined by the host immune response. Again, this highlights the need for greater understanding of the metabolic functions which ensure prolonged viability under nonreplicating persistence. Here, it is interesting to consider recent work with other bacteria which has demonstrated the conditional essentiality of many genes, including a large number of proteins of unknown function, in long-term fitness under growth arrest (174, 175), as well as the potential for selective targeting of pathogens under these conditions (176).

The example of pat mutants, engineered strains of M. tuberculosis which achieve population numbers similar to wild-type bacilli in vivo but do not cause the same degree of pathology (177), has been cited as evidence of the fact that disease outcome is not necessarily linked to the bacillary population size (37). Rather, it is the interplay between the host response and the infecting M. tuberculosis population—and the antigens it produces (69)—which determines infection outcomes (178). This tenet is perhaps best captured in the aphorism that M. tuberculosis is necessary, but not sufficient, for TB disease (82, 179).

Again, evidence from pioneering studies in NHPs reinforces this conclusion. For both diseased and clinically latent monkeys, the population of culturable organisms commonly numbers around 2 × 10⁵ bacilli per lesion prior to the onset of the adaptive immune response and stabilizes at a median of ∼10² bacilli per lesion during active disease (169). That is, the total numbers of bacilli present in active versus latent disease can overlap. Moreover, while there can be microenvironments in which the mycobacterial population achieves large numbers during active disease, the total bacillary burden generally seems much lower than might be expected given the extent of pathology (180). This observation is reiterated in human studies (69, 82) and contrasts with recent conclusions from an unrelated infectious disease model which demonstrated that infection outcome was directly attributable to bacterial load (181).

It is important, though, to differentiate TB disease in HIV-infected versus HIV-uninfected individuals. The enhanced propensity for extrapulmonary disease and TB bacteremia in HIV-TB cases argues that infection outcomes in the context of HIV-mediated immunodeficiency might correlate much more closely with the mycobacterial burden (182). Moreover, while the precise cause of death in disseminated TB disease in HIV-infected individuals remains speculative (183, 345), for pulmonary TB in individuals who are not immune-suppressed, it is clear that most patients ultimately succumb to advanced pathology (signaling a systemic failure in TB diagnosis and treatment [184]), a distinction which reinforces the intimate interplay between host and pathogen in the context of a functional immune response (185). Therefore, to consider TB disease as a single entity and, by extension, to categorize M. tuberculosis metabolic states (and, by implication, metabolic vulnerabilities) according to a single framework that assumes a handful of possible infection outcomes (51, 52) imposes a conceptual constraint that, albeit practically appealing, is potentially crippling in its failure to accommodate the complexity inherent in an infectious organism adapted to persistence, not virulence.

TB as Polymicrobial Infection of Humans

The continual acquisition of new sequence data from different mycobacterial species and strains has enabled comparative genomics analyses to determine the evolutionary histories of the different members of the MTBC with greater resolution (3). Importantly, when key genetic differences are considered against the backdrop of the very specific host tropisms of the different MTBC members, it becomes increasingly evident that laboratory-determined virulence is not only insufficient to understand pathogenicity, but likely inappropriate. As noted elsewhere (186), each species has diverged from a common ancestor through coevolution with its own cognate host, resulting in genetically related but phenotypically distinct pathogens specifically suited to hosts with disparate social and biological features.

M. tuberculosis causes TB, a disease which manifests primarily as damage of the lungs, yet extrapulmonary forms are also known which affect other organs and sites, including the pleura, abdomen, genitourinary tract, meninges, skin, joints, and bones (187). Lymph nodes, too, are often cited among lists of “extrapulmonary” sites, but this modifier is likely inaccurate given significant evidence that infection of lymph nodes constitutes a critical event in TB pathogenesis (70, 188, 189). Also distinct from extrapulmonary TB is the observation that M. tuberculosis can occupy diverse anatomical and cellular niches during latent host infection (190). Even in “simple” cases of pulmonary TB, infecting M. tuberculosis bacilli are not restricted to specific environments but can be found in multiple different host compartments, including cellular and necrotic granulomas and the inner surfaces of open cavities (58, 191–193). In each of these, bacilli can be sequestered within host cells or located in extracellular niches (194), reinforcing the suggestion that TB disease might usefully be considered a polymicrobial infection (195).

The occupation of diverse host microenvironments—in many cases, by paucibacillary populations—with independent disease trajectories constitutes a major obstacle to elucidating the metabolic adaptations demanded of bacilli located within (or transitioning between) the different niches (26). From a pharmacological perspective, this uncertainty presents a clear challenge to developing approaches that will ensure optimal drug delivery to bacilli in dynamic and disparate physical loci and metabolic states (57, 58, 196) and to understanding which (sub)populations must be controlled (or eliminated) for favorable therapeutic outcomes (55, 189, 197–199). For drug discovery, such uncertainty undermines efforts to identify unequivocally the pathways and functions which are critical to mycobacterial survival and so might represent viable antibiotic targets, that is, the organism’s “metabolic vulnerabilities.” In addition, it renders ambiguous the potential utility of compounds known to be selectively active against either proliferating or quiescent organisms, growing or subsisting on different nutrient sources (conversely, such molecules are of unquestionable utility as probes of the distinct metabolic states adopted by M. tuberculosis in those environments). Therefore, for those wanting (simply) to understand the link between bacillary metabolism and host disease, the lack of granularity informing available descriptions of in situ metabolism poses unresolved challenges regarding the assays and approaches—available and required—that might best elucidate mycobacterial metabolic function in vitro, in infection models, and in clinical disease.

Immunometabolism and Pathometabolism at the Host-Pathogen Interface

Like all intracellular pathogens, the bacillus must adapt its metabolism to the host cell, whose own metabolic status might correspondingly be altered as a consequence of infection (200). Though poorly understood, this interplay between immunometabolism (201) and pathometabolism (202) is critical to infection outcomes (45). There is increasing evidence that successful intracellular pathogens such as M. tuberculosis have evolved to exploit the reciprocal determinism linking host cellular function and metabolic status (161). Ergo, this interplay might offer a solution to the apparent paradox that the very immune cells designed to eliminate the invading pathogen can function as a preferred niche (44). Consistent with their role as early responders, innate immune cells appear much more vulnerable to infection than adaptive immune cells, reinforcing the inference that organisms such as M. tuberculosis have adapted to the tightly regulated responses of innate immune cells to enable long-term intracellular survival. The observation that a core transcriptional profile unites species and lineages across the MTBC in response to intracellular infection (151) strengthens this conclusion and is supported by the presence of metabolic enzymes among the hyperconserved T-cell epitopes in M. tuberculosis (203). That is, through millennia of coevolution, the bacillus is likely to have adapted to the complex network of metabolic and immune alterations and signals triggered in its host.

The corollary also appears to hold true: in response to M. tuberculosis infection, some host macrophages shift to a “Warburg-like” (161) program of aerobic glycolysis (162, 204–207) in which oxidative phosphorylation is reduced and cytosolic glycolysis and the mitochondrial tricarboxylic acid cycle are redirected primarily to the biosynthesis of fatty acids. In an example of “bipartite metabolism” (44), these are utilized as energy sources by the intracellular bacillus (120) without imposing excessive metabolic stress on the host macrophage, which is able to drive an anabolic program utilizing the available glucose. Notably, a metabolic shift in the infected macrophage has been implicated in the accumulation of intracellular lipid bodies characteristic of foamy macrophages. In turn, this phenotype has been closely linked with mycobacterial persistence and disease progression (208–211), though recent observations suggest that it might be more nuanced (212). There is also evidence that altered metabolism at the site of M. tuberculosis infection influences cytokine production (213), reinforcing the connection between cellular metabolic state and immune function. Here, it is useful to consider the key observation that, at a metabolic level, the host-pathogen interaction is likely to possess emergent properties: specifically, the metabolic responses of host cells to living bacteria are likely to be more subtle and complex than the sum of metabolic responses to individual bacterial components (161).

Given the major recent impact of positron emission tomography and computed tomography imaging in providing unprecedented insights into TB disease pathogenesis and heterogeneity, both within and between individuals in different disease models and human TB patients (50, 169, 180, 189), it is worth noting here that the technique utilizes a [¹⁸F]fluoro-2-deoxy-2-d-glucose ([¹⁸F]-FDG) probe which exploits the enhanced glucose utilization of inflammatory cells. Recent work investigating the potential for alternative probes to provide cell-specific information ([¹⁸F]-FDG is taken up by all glycolytically active cells and does not distinguish subsets) has revealed the dynamic nature of granuloma composition and metabolism over the course of infection (214), again reinforcing the critical concept that lesional progression with an individual does not follow a fixed temporal program (173).

M. TUBERCULOSIS METABOLISM

As noted at the outset, it is not possible to cover all aspects of mycobacterial metabolism within a single review. Instead, this and subsequent sections will provide a very brief update on research supporting four key concepts which are critical to understanding M. tuberculosis metabolic function and regulation, namely, (i) metabolic and respiratory flexibility, (ii) the capacity for cocatabolism of different nutrient sources, (iii) the potential for metabolic anticipation, and (iv) the heavy investment in cell wall metabolism.

Central Metabolism

“Not a picky eater, at least not for carbon and nitrogen sources” (Luiz Pedro De Carvalho, Twitter, 1 March 2019). Despite its levity, this declaration, made recently in promotion of research demonstrating the utilization by M. tuberculosis of amino acids as nitrogen sources (215), perhaps best summarizes a sizeable literature on (mostly in vitro) mycobacterial metabolic function (see Table 1 for selected references). Unsurprisingly, the predominant focus over the past 2 decades has been on carbon utilization, and consistent with this bias, this aspect of mycobacterial metabolism has been extensively reviewed (26, 45). Among the central tenets emerging from this work are that M. tuberculosis can cocatabolize different carbon sources (216), an observation which signals a key departure from classic bacterial metabolism in which diauxic growth regulated via carbon catabolite repression constitutes the dominant model.

A number of subsequent studies have reinforced carbon cocatabolism as critical to M. tuberculosis cellular function and pathogenicity, identifying specific roles for fructose-1,6-bisphosphate aldolase (217) and pyruvate kinase (218) in enabling this capacity. Moreover, the persistence defect observed in an M. tuberculosis mutant lacking both glucokinase homologues—in other words, functionally depleted of glycolytic potential—supports the conclusion that both carbohydrates and fatty acids are required for long-term survival in vivo (219). More recent research has extended the capacity for cocatabolism to nitrogen sources in vitro (215) and in an intracellular model (220), strongly suggesting the generalizability of this principle to mycobacterial metabolism. In this context, recent evidence of the power of techniques such as cryo-electron microscopy in revealing cytoplasmic organization and the presence of microcompartments in different prokaryotes (221, 222) suggests the intriguing possibility of investigating the potential that cocatabolism is spatially regulated.

At least one codicil to the inferred absence of catabolite repression in M. tuberculosis might exist in the observation that cholesterol-dependent growth inhibition cannot be alleviated by the provision of glucose as an alternative carbon source (223). This implies that preferential utilization of cholesterol—predicted to be a feature of M. tuberculosis pathogenicity (224)—might represent a critical constraint to mycobacterial metabolic flexibility (223). In some ways, this phenomenon is reminiscent of the very well-described differential (often poor) growth of clinical M. tuberculosis isolates in specialist laboratory medium following isolation from patient samples, including sputum (54, 225). In turn, it highlights the fact that, while M. tuberculosis H37Rv is the workhorse of experimental mycobacteriology, this strain has become adapted to growth in laboratory media through decades of serial passage in vitro (152).

The concept that growth arrest represents a key stress response in M. tuberculosis and, importantly, might be triggered even when the organism retains the metabolic capacity to grow (65) is also central to understanding mycobacterial pathogenicity. This preservation strategy appears to be enabled, at least partially, by the capacity for “anticipation” of future events, an example of which was provided by the observation that M. tuberculosis exposed to hypoxia remodels its own cell envelope to generate building blocks for later reentry into the cell cycle (226). Notably, this capacity is enabled by the catabolism of trehalose mycolates, from which the bacillus generates pentose phosphate intermediates that are incorporated into peptidoglycan biosynthesis much later, when environmental conditions improve, that is, contingent on re-aeration at an indeterminate future time. This illustrates an inherent capacity for preservation that might be hard-wired through long-term adaptation to predictable host environments and stimuli (227). Reinforcing this idea, very recent work has demonstrated that multiple stresses induce expression of the small regulatory RNA, MrsI, which functions to preserve intrabacillary iron reserves, perhaps in anticipation of iron deprivation in the infected macrophage (228). Although not involving metabolic anticipation per se, the observation that utilization of nitrate as the sole nitrogen source requires a functional proteasome (229) is analogous in demonstrating the very tight connection between metabolic pathway utilization and regulation of cellular macromolecular synthesis and recycling, perhaps offering insights into the organism’s intrinsic responses to in vivo nutrient availabilities.

Respiration

Like all mycobacteria, M. tuberculosis is an obligate aerobe. Therefore, survival in growth-inhibitory hypoxic environments demands sufficient energy generation for maintenance functions, while aerobic conditions enable full metabolic activity for bacillary replication (230). From comparative genomics of different mycobacteria, it is apparent that, whereas major biosynthetic pathways are reasonably conserved, occurring within core orthologous clusters, the respective complements of genes involved in energy metabolism are more diverse. This observation is consistent with the adaptation of individual species to specific environmental and host niches and, perhaps, replication rates (21, 22).

M. tuberculosis possesses multiple branched pathways for electron transfer from a variety of low-potential reductants to oxygen during aerobic growth (230) and to alternative electron acceptors under hypoxic conditions (231). The potential flexibility suggested by the expanded complement of primary dehydrogenases and oxidoreductases (232)—the function and regulation of which remain poorly understood (230)—is, however, tempered by the observation that aerobic respiration in M. tuberculosis appears to depend on a single lipoquinone, menaquinone, and two terminal respiratory oxidases, the aa₃-type cytochrome c oxidase and the cytochrome bd menaquinol oxidase (231). Moreover, evidence suggests the obligatory coupling of ATP generation to the electron transport chain and the membrane-bound F₁F₀-ATP synthase, irrespective of the oxygen concentration or the proton motive force (230). And, whereas some bacteria are able to utilize ATP to energize the cytomembrane via coupling of ATP hydrolysis to proton pumping, the M. tuberculosis F₁F₀ ATP synthase lacks ATP hydrolysis activity (233). These limitations are unusual for mycobacteria and have been exploited in the development of the diarylquinoline bedaquiline (Sirturo), the first new FDA-approved anti-TB drug in over 40 years, whose initial use has been restricted to multidrug-resistant TB pending the results of ongoing phase 2 and phase 3 trials (234, 235).

As noted elsewhere (230), the M. tuberculosis genome encodes expanded complements of both electron donors and electron acceptors whose roles in pathogenicity are largely unexplored. It is perhaps no exaggeration to suggest that, of all the core metabolic functions, respiration and energy metabolism are those which have been most prone to very selective investigation given the practical difficulties associated with manipulating the bacillus under variable conditions of hypoxia and anoxia, the default reliance on standard, glucose-based growth media in vitro, and the general use of mouse models which recapitulate only certain aspects of TB pathology (236). A salient example is provided by the likely contribution of the assimilatory/respiratory nitrate reductase to bacillary survival under various stresses, a notion supported by strong functional and comparative genomics evidence that the acquisition of molybdopterin biosynthetic and salvage genes (molybdenum cofactor is essential for multiple enzymes in carbon, sulfur, and nitrogen metabolism, including nitrate reductase [145, 237]) differentiates the pathogenic M. tuberculosis from precursor (nonpathogenic) mycobacteria (22).

The predicted capacity of infecting bacilli to subsist under different conditions and in various microenvironments offering very different substrates for energy generation argues for further research into the consequences of disrupted mycobacterial bioenergetics for host infection outcomes (238, 239). Certainly, early clinical reports of the benefits of bedaquiline for multidrug-resistant and extensively drug-resistant TB are very encouraging (234, 235). Moreover, in addition to highlighting the possible benefits of targeting components of the oxidative phosphorylation machinery for new TB drug development, knowledge about the precise mechanism of action of this compound has proved very instructive. By binding the atpE-encoded F₁F₀ oligomeric c-ring subunits, bedaquiline effectively functions as a nonclassic uncoupler: exposure of bacilli to the drug stimulates respiration owing to the electroneutral uncoupling of proton transport and ATP synthesis (240, 241). This is important because it suggests the potential to target the proton motive force, a property that appears to be shared by known antimycobacterial compounds (242), as well as verapamil (243). However, caution is required given recent observations that inhibition of energy metabolism might have unintended consequences in some combinations (244).

Cell Wall Homeostasis

Consistent with its dominant role in the biology of M. tuberculosis, the cell wall has been the subject of intensive study for decades (Table 1): as a primary site of the host-pathogen interface (245–247), as a target of—and impermeability barrier to—many current and experimental antimycobacterial drugs (248–251), and as a major determinant of the robustness and durability of the bacillus throughout its life cycle (252). The molecular composition and arrangement of the mycobacterial cell envelope, classified as Gram positive but comprising an outer membrane and a periplasmic space, is quite well understood and broadly comprises three distinct layers (251, 253, 254). Exterior to the plasma membrane, a peptidoglycan layer links covalently to a mycolyl-arabinogalactan layer which is decorated with noncovalently linked lipids and lipoglycans such as the phosphatidylinositol mannosides, phthiocerol dimycocerosates, phenolic glycolipids, and lipoarabinomannan, which have been investigated for their roles in mycobacterial pathogenicity (255) and as diagnostic markers (256, 257). This outer membrane or “mycomembrane” is, in turn, surrounded by a loosely attached capsular layer predominantly comprising polysaccharides and proteins (95, 253).

Less well understood is how the construction, maintenance, and remodeling of the cell envelope are regulated and accomplished in concert with the mycobacterial cell cycle, particularly in vivo in different host microenvironments (258, 259). Here, an emerging theme is the application of advanced, live-cell and time-lapse imaging in combination with elegant fluorescent bioreporters, in most cases fluorophore-linked analogs of natural cell wall precursors which are incorporated into the cell wall without significant impairment of mycobacterial growth (260–265). Moreover, advances in mass spectrometry have enabled the cataloguing of cell wall lipids, fatty acids, and proteins with ever greater resolution (266), including under different growth conditions (267), in response to drug treatment (268), and in drug resistance and drug tolerance (94, 269).

Accessing Nutrients from the Host

The number of M. tuberculosis organisms required to initiate successful host infection remains unknown. Though deep sequencing suggests the possibility for transmission of multiple genotypes (270), most historical estimates are that a single bacillus might be adequate (271). This prediction has been reinforced most convincingly in NHP models in which the elegant use of genetic “barcodes” has shown that individual lesions can arise from a single seeding M. tuberculosis genotype (169, 170). Whatever the number, proliferation depends on the ability of the invading pathogen to assimilate host nutrients for energy consumption and macromolecular synthesis. Precisely which host-derived nutrients M. tuberculosis accesses, where this occurs, and how are subjects of ongoing investigations (41, 272–277). A core paradox in M. tuberculosis research is that, whereas the promiscuity in the nutrients and cofactors potentially assimilated from the host seems to complement M. tuberculosis’s autarkic metabolism (41), the same promiscuity is difficult to reconcile with the inferred impermeability of the mycobacterial cell wall. This apparent incongruity has driven the search for potential transport and porin proteins in the mycobacterial outer membrane (278, 279).

An intriguing mechanism for nutrient acquisition was proposed in a series of studies characterizing the tuberculosis necrotizing toxin (TNT), an enzyme which intracellular M. tuberculosis utilizes to kill the host macrophage and gain access to cytoplasmic nutrients (280–283). Synthesizing multiple lines of experimental evidence, the prevailing model holds that bacilli contained within the macrophage phagolysosome deploy CpnT, a bifunctional protein comprising an N-terminal pore-forming domain and C-terminal NAD⁺ glycohydrolase (TNT), via the mycobacterial ESX-1 type VII secretion system (284) to effect macrophage death. Notably, release of TNT into the macrophage cytoplasm (the glycohydrolase is pH sensitive, and so, inactive in the acidified phagolysosome) is dependent on the ability of the other ESX-1 substrate, EsxB (also known as ESAT-6/CFP-10), to puncture the phagolysosomal membrane, therefore requiring the neat coordination of mycobacterial effectors to ensure that intraphagosomal bacilli (which, from the inserted CpnT N-terminal domain, now possess the membrane pores necessary for maximal nutrient uptake) are poised to take advantage of the inactivated host cell. In an intriguing twist, parallel work has suggested that, contrary to the connotations of its explosive moniker, the mycobacterial toxin might deliver a “blast without power” (285), resulting in necrotic cell death that might deliberately limit immunogenicity. Given the implied importance of this mechanism to the ability of M. tuberculosis to exploit the host macrophage, perhaps while minimizing host immune detection, it would be interesting to know whether the enhanced growth in dead macrophages described recently (172) also depends on functional TNT and, moreover, whether loss of TNT impacts pathogenicity in vivo.

MYCOBACTERIAL METABOLISM AND NEW TB DRUG DISCOVERY

We noted previously (286) that investigations of core metabolic processes in M. tuberculosis are generally framed within the context of drug discovery—whether looking specifically for new active molecules or vulnerable targets or attempting to elucidate mechanisms of action or to reveal reasons for compound failure or bypass. Examining the current TB drug pipeline, it is evident that significant progress has been made in expanding not only the number of different mechanistic classes represented in the most promising drug-target pairs, but also in validating new high-value targets for future discovery programs (134, 287). Nevertheless, the challenges for new TB drug discovery remain significant. Compound permeation (288), xenobiotic metabolism (289), and efflux (290–292) are three major innate mechanisms of drug resistance, all of which are poorly understood and correspondingly difficult to demonstrate or quantify. Impermeability of the complex mycobacterial cell envelope is often cited as a major confounder to new TB drug development. Though intuitively attractive, the evidence to support this conclusion remains scarce. In contrast, the formidable capacity of M. tuberculosis for biotransformation of xenobiotics is much better established (289) and, with the mostly theoretical contribution of efflux (293), might represent the major innate barrier to antibiotic activity (294).

The increasing appreciation of the organism’s capacity for biotransformation has exposed the lack of rapid, medium-throughput strategies to measure metabolism of xenobiotics (289). Similarly, despite the frequency with which the inactivity of many compounds against M. tuberculosis is assumed to be a function of poor or failed permeation, proving so remains enormously difficult. The inability to determine quickly and reliably whether a compound—and its analogs in a structure-activity relationship analysis—has lost activity owing to differential (lost) target engagement, inactivation via metabolic transformation, extrusion by efflux, or failure to penetrate severely undermines medicinal chemistry efforts. Without complete information about the fate of the applied molecule, efforts to optimize physicochemical properties toward a drug-like lead are based almost entirely on experience, intuition, and empiricism (i.e., that which works is clearly evident in the inhibition of bacillary growth in a specific assay, usually an MIC determination in vitro [295]). The scarcity of techniques to enable rapid determination of the intrabacillary accumulation and metabolism of compounds, perhaps even independent of inhibitory activity, therefore presents a significant obstacle to rational drug development and suggests that strategies employed in other model systems (296, 297) might be usefully adapted to this end.

Recent work elucidating the impact of cell wall antibiotics on membrane fluidity and biogenesis (298, 299), as well as the observation that many compounds with antimycobacterial activity function as protonophores (242, 243)—in some cases, in addition to interacting directly with their cellular target activity (300)—and so operate at the cytoplasmic membrane (the site of many so-called promiscuous targets, including DprE1, QcrB, and MmpL3 [250, 301]), reinforces the sense that outer membrane impermeability might not present the major obstacle to drug efficacy. This impression will, however, require confirmation via the use of advanced metabolomics approaches to elucidate the (differential) penetration of different compounds into the mycobacterial cytoplasm (302).

Efflux, too, presents a potentially sizeable but largely undefined obstacle. The M. tuberculosis genome encodes more than 40 predicted efflux pumps, only a handful of which have been implicated in innate drug resistance (232, 292, 303, 304). Recent evidence that the antihypertensive verapamil does not function as an efflux pump inhibitor but instead disrupts membrane energetics in M. tuberculosis (243) reinforces the perils inherent in making inferences from phenotypic observations in the absence of compelling mechanistic insight. Even ostensibly clear-cut drug-target interactions can have surprising locations (305), while the mechanism of resistance does not always correlate with the mechanism of action (306, 307).

From a metabolic perspective, incomplete understanding of M. tuberculosis metabolism in its human host can have very practical implications for TB drug discovery. For example, two separate studies elucidating the guaB2-encoded mycobacterial inosine monophosphate dehydrogenase (IMPDH) as the target of related compound series took very different views of the potential of their respective molecules for advancement as candidate drugs and, by implication, the attractiveness of de novo purine biosynthesis as an anti-TB target. The first (149) propounded IMPDH as “vulnerable” and, consequently, worth pursuing. In contrast, challenging pharmacokinetics, coupled with the observation that concentrations of guanine in resected lung tissue from TB patients achieve levels in excess of those required to allow bypass of IMPDH through salvage, prompted the authors of the second study to conclude the opposite. They argued that IMPDH might be “essential” but is “not vulnerable” (107), thus consigning GuaB2 and its chemical inhibitors to red flag status for new TB drug development, at least in their estimation (270). These conflicting views reinforce the fact that, even where a molecular catalogue of individual tissues, lesions, or microcompartments is available, it might not necessarily enable an unequivocal prediction of what the bacillus is able to access and assimilate from its surroundings. That is, the availability of any nutrient must be distinguished from its accessibility to the infecting M. tuberculosis population. Of course, when prioritizing resources for new drug candidates, it is prudent to assume that all nutrients, if present, are (potentially) accessible. The risk that salvage might overcome antibiotic-mediated nutrient starvation therefore ensures that the bar is set very high for new TB drug development (308).

Harnessing Metabolism to Potentiate Drug Activity

Earlier sections highlighted some of the ways in which different intrinsic metabolic functions contribute to the innate resistance of M. tuberculosis to antibiotics, reinforcing a theme which has been discussed extensively elsewhere (65, 309, 310). Increasingly, however, knowledge of metabolic function is being exploited not only to circumvent innate resistance mechanisms, but also to understand the metabolic functions which might subvert antibiotic efficacy so as to identify novel countermeasures (64, 311, 312). Recent work highlighting the contribution of altered propionate metabolism to the emergence of drug resistance in M. tuberculosis (313) provides a very important example of the potential for metabolic deficiency to confer a tolerant phenotype that enables the acquisition and fixation of genetic drug resistance. Moreover, the growing number of studies demonstrating altered metabolic function in drug-resistant organisms, including M. tuberculosis (269, 314), suggests that this phenomenon might even be more widespread.

A corollary to the above is that metabolism might also be harnessed to improve drug action. For example, the inferred mechanism of action of the repurposed agent, clofazimine, which triggers a redox cycling mechanism that generates toxic reactive oxygen species (315), reinforces the notion that bacilli might be trapped in a futile and ultimately destructive metabolism. Notably, by combining clofazimine with other compounds targeting components of the mycobacterial electron transport chain, Steyn and colleagues demonstrated the potential to enhance killing of M. tuberculosis in an intracellular model (43). In an analogous example of this approach, inactivation of the essential maltosyltransferase, GlgE, was shown to trigger an amplification loop which resulted in rapid mycobacterial death owing to self-poisoning by maltose 1-phosphate accumulation (316). Though the precise mechanism differs in each case, these observations suggest the potential for combination regimens based on drug targets and/or mechanisms of action which incorporate inhibition of growth (target/drug A) and prevention of ordered metabolic shutdown (target/drug B); that is, combinations designed to prevent metabolic escape. This strategy is distinct from both metabolic bypass (in which an alternative, perhaps less efficient or autonomous, mechanism exists to enable generation of the same essential metabolite, e.g., via scavenging from the host) and metabolic buffering (in which partially redundant mechanisms enable the organism to maintain homeostasis). Instead, it entails eliminating the organism’s inherent capacity for surviving through replicative and metabolic quiescence and so might add a further element to the three models for antimycobacterial drug synergy proposed previously (65). A recent report describing the identification of chemical inhibitors of the stringent response enzyme, Rel_Mtb, a (p)ppGpp synthetase, provides a cogent demonstration of the potential of this therapeutic approach (317).

To some extent, this strategy overlaps with the idea, summarized in a recent review, of developing “metabolism-targeted adjuvant therapies” to prevent antibiotic tolerance (311). In support of that proposal, the authors cite the targeted disruption of regulatory mechanisms, for example, through the identification of artemisinin analogs as inhibitors of the mycobacterial DosRST system (318), an ∼50-gene regulon that has been implicated in antibiotic tolerance in infection models in vitro and in vivo (319). In addition, they provide several illustrations of the potential to achieve or enhance antimycobacterial activity by interfering with bacillary metabolism. These include the direct inhibition of specific “tolerance/persistence” effectors such as isocitrate lyase (320) or nicotinamide adenine dinucleotide synthetase, NadE (321). Of particular interest, though, is the proposition that stimulating bacillary respiration and energy metabolism, for example, by providing a glycolytic substrate together with an alternative terminal electron acceptor such as fumarate (322), can increase efficacy of known antibacterial agents against apparently recalcitrant subpopulations. This concept not only reinforces the impact of metabolic state on antibiotic efficacy (an obvious conclusion), but it also suggests the potential to manipulate bacillary metabolism artificially through the deliberate provision of nutrients which might drive a defined metabolic program that ensures susceptibility to the applied agent. That is, it opens the door for a combination comprising an (inhibitory) antibiotic and a (stimulatory) metabolite.

FUTURE PERSPECTIVES

This review has highlighted the complications inherent in working with an infectious pathogen whose preferred anatomical target (i.e., the human lung) is difficult to access. The result is that most studies, rather than revealing mycobacterial metabolic function in the context of TB disease, have tended to elucidate the metabolic capacity of the bacillus, either in vitro in different growth media and under various stresses thought to mimic the intrahost environment or in vivo in experimental models of infection (37). Those attempts, much fewer in number, to investigate M. tuberculosis pathogenicity (including metabolism) in situ or closer to the site of disease have predominantly relied on bacilli captured in sputum or bronchoalveolar lavage fluid, lymph node or other biopsies, or lung resections of treated patients. One consequence is that the potential distinction between primary and post-primary TB as separate disease entities (and therefore likely invoking different mycobacterial metabolic functions) has been obscured (69, 79).

As noted above, determining the metabolic and physiological states adopted by the tubercle bacillus in the infected human host represents the ultimate goal and, though still a significant challenge, seems increasingly attainable with the development of new technologies. Recent advances such as the ability to obtain transcriptional data from very-low-abundance RNA (258, 323) and at the single-cell level (324) promise to provide much greater insight into the exchange of stimuli and responses that occurs between M. tuberculosis and its host cell during intracellular infection and which shapes the metabolic adaptation of the invading bacilli to the available nutrients in this and other microenvironments. Exciting advances in other bacterial models, including the combined application of RNA profiling with stable isotope labelling (325, 326), reinforce the sense that elucidating the growth rates—and metabolic states—of even slow-growing pathogens such as M. tuberculosis will become feasible (327).

A recent systems pharmacology study (198) suggests another approach to this problem, namely, the potential to combine empirical data derived from in vitro and in vivo experiments—ideally including clinical samples (197, 328)—to generate a complex mathematical model that simulates disease progression within a single locus. In that case, a shift was made from the M. tuberculosis-infected macrophage as the “unit of infection” (155) to the granuloma in modeling the efficacies of different fluoroquinolone antibiotics as anti-TB agents. Their approach enabled a compelling demonstration of the capacity to incorporate both host and bacillary factors in building the model and, at the same time, raised the possibility that this strategy could be adapted to (or informed by) metabolomic, proteomic, and lipidomic data. Notably, the study also highlighted the emergent properties of a complex system comprising infecting bacilli and different types of host immune cells (the variety was necessarily limited to a handful of T cell subsets and macrophage phenotypes owing to the demands of computational modeling). Again, this suggests the potential to enable key insights into the dominant metabolic pathways active in different lesions, how these might influence infection trajectories and pathological outcomes, and how they might be exploited for improved chemotherapy.