Tuberculosis (TB) remains a major health threat, killing nearly 2 million individuals around this globe, annually. The only vaccine, developed almost a century ago, provides limited protection only during childhood. After decades without the introduction of new antibiotics, several candidates are currently undergoing clinical investigation. Curing TB requires prolonged combination of chemotherapy with several drugs. Moreover, monitoring the success of therapy is questionable owing to the lack of reliable biomarkers. To substantially improve the situation, a detailed understanding of the cross-talk between human host and the pathogen Mycobacterium tuberculosis (Mtb) is vital. Principally, the enormous success of Mtb is based on three capacities: first, reprogramming of macrophages after primary infection/phagocytosis to prevent its own destruction; second, initiating the formation of well-organized granulomas, comprising different immune cells to create a confined environment for the host–pathogen standoff; third, the capability to shut down its own central metabolism, terminate replication, and thereby transit into a stage of dormancy rendering itself extremely resistant to host defense and drug treatment. Here, we review the molecular mechanisms underlying these processes, draw conclusions in a working model of mycobacterial dormancy, and highlight gaps in our understanding to be addressed in future research.
The etiology of tuberculosis (TB) – one of the most devastating diseases of humankind – was first elucidated by Robert Koch (1843–1910) in 1882 (Koch, , ; Kaufmann & Winau, ). Koch developed a specific staining method based on methylene blue combined with brown counterstaining of host tissues with vesuvin for the causative agent Mycobacterium tuberculosis (Mtb), which allowed the visualization of bacteria not only in cultures, but also in tissues (Box 1). Precisely, 130 years down the road, his diagnostic method is still in use virtually unchanged. Potent antibiotics have been discovered, and public health systems improved significantly in many parts of the world since Koch's times. Nevertheless, Mtb remains as deadly as it was, claiming nearly 2 million lives annually and exploiting an estimated 2 billion as reservoir of latently Mtb-infected (LTBI) individuals (Dye et al., ; Yew & Leung, ). These figures were calculated when our globe hosted about 6 billion people. In 2011, when we reached the 7 billion mark, about 2.3 billion LTBI individuals appear more likely. Currently, up to 9 million new cases of TB arise each year, more than ever before (Dye & Williams, ). Most cases are not because of new infections but through the reactivation of dormant Mtb residing in LTBI hosts (Fig. 0001).
The bacillus causing TB in humans belongs to the genus Mycobacterium that includes several other obligate human pathogens, most importantly Mycobacterium leprae (leprosy), Mycobacterium africanum (TB-like symptoms, lower pathogenicity), and Mycobacterium bovis (primarily TB in cattle). The family of pathogenic mycobacteria arose from soil-dwelling ancestors. They most likely became pathogens to animal and human hosts during domestication of animals about 10 000 years ago (Smith et al., ). At 37 °C and under optimal availability of oxygen and nutrients, a single Mtb organism has a generation time of 18–24 h and forms a white to light-yellow colony on agar within 3–4 weeks. The aerobic-to-facultative anaerobe, Gram-positive pathogen is surrounded by an impermeable and thick cell wall/capsule that is made of peptidoglycans, polysaccharides, unusual glycolipids, and lipids mainly consisting of long-chain fatty acids, such as mycolic acid. Unlike many other bacteria, Mtb does not form spores but has the capacity to become dormant – a nonreplicating state characterized by low metabolic activity and phenotypic drug resistance. Note that phenotypic drug resistance is related to a specific physiologic state and independent from genetic mutations. Mtb is typically visualized by Ziehl–Neelsen (acid-fast) staining and appears as a rod-shaped red bacillus. The GC-rich (65.6%) 4.4-Mbp genome of Mtb is one of the biggest among the bacteria and encodes about 4000 predicted proteins (http://genolist.pasteur.fr/TubercuList/, Cole et al., ; integrated platform for TB research: http://www.tbdb.org/, Reddy et al., ). M. bovis Bacillus Calmette–Guérin (BCG) (attenuated form of M. bovis) and Mycobacterium smegmatis are nonpathogenic and therefore common surrogates for Mtb in research. The former is used as vaccine in children with partial success. Mtb is typically diagnosed by microscopy in the sputum of active TB patients. A regime of several drugs is available to effectively cure the disease by 6–9 months of combination therapy. Incomplete treatment or noncompliance of patients often leads to drug-resistant Mtb, which is conferred by genetic mutations.
Primary infection can: (1) progress toward active disease; (2) be contained as latent infection; (3) be eradicated by the host's immune system. Less than 10% of infected individuals develop active TB during their lifetime. It is impossible to predict who will contain latent infection throughout their lifetime and remain healthy, and who will develop active TB at some point. However, the risk of active disease is increased in immunocompromising situations such as during antitumor necrosis factor therapy of patients with chronic inflammatory diseases, by diabetes/obesity or by co-infection with human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS) (Barry et al., ). Other risk factors include alcoholism and poor nutrition. Multiple factors are likely involved in defining overall risk of TB, and the genetic makeup of both host and pathogen plays a decisive role. Thus, biomarkers that would allow prognosis of TB reactivation in healthy individuals with LTBI would be of tremendous value.
Today, intervention measures for controlling TB are available. We have at hand numerous drugs to cure the disease, diagnostics to identify patients, and a vaccine to prevent severe forms of childhood TB. Yet, these measures are insufficient for the following reasons:
Diagnosis of TB in low-income countries is correct in only an estimated half of all TB cases.
Treatment for TB requires 3–4 drugs given for 6 months or longer. Frequently, compliance is poor and premature termination of drug therapy often results in emergence of resistant strains. Today, an estimated 50 million individuals harbor multidrug-resistant Mtb of whom 500 000 fall ill annually. Even worse, strains of virtually untreatable extensively drug-resistant (XDR) Mtb are on the rise, and XDR-TB has already been notified in 58 countries. Even totally drug-resistant-TB has been described (Velayati et al., ).
The current vaccine, M. bovis BCG, protects against severe forms of childhood TB, but fails to protect against adult pulmonary TB, which has become the most prevalent form of the disease today. Hence, BCG does not impact the transmission of Mtb.
Many definitions in infectious disease research are difficult to apply in TB. Needless to say, the etiologic agent Mtb is a true pathogen and not an opportunistic microorganism, even though active disease only develops in the minority of infections. Virulence describes the capability of the pathogen to cause disease in quantitative terms. High virulence, therefore, is often related to marked severity of disease and vice versa. Yet, the decisive survival factor of Mtb is its capacity to persist in the host for long periods of time, both during noncontagious LTBI and contagious active TB before it is spread. To further complicate the situation, pathogenicity of TB is largely influenced by the host immune response. Hence, a discussion of the disease without regard for the host would remain incomplete. Although our review of the molecular mechanisms of TB is oriented toward the pathogen's perspective, we will consider host influences where appropriate.
Infection of the alveolar macrophage
Macrophages operate as prime defense cells against microbial intruders (Nathan & Shiloh, ; Liu & Modlin, ; Deretic et al., ). These microorganisms are ingested by phagocytosis, a process consisting of membrane invaginations finally culminating in phagosome formation (Aderem & Underhill, ). This organelle is a part of the intracellular trafficking and transport system and the site to which the entire arsenal of host defense is targeted (Schekman, ; Rothman & Wieland, ; Gruenberg & Stenmark, ). Microorganisms captured in the phagosome experience increasing acidification, reactive oxygen and nitrogen species (ROS and RNS), hydrolytic enzymes, and cationic antimicrobial peptides (CAMPs). Acidic pH inside the maturing phagosome activates enzymes that degrade bacterial lipids and proteins (Huynh & Grinstein, ). Simultaneously microbial metabolism is suppressed by such conditions. ROS and RNS generated by the phagosomal enzymes NADPH phagocyte oxidase and inducible nitric oxide synthase (iNOS) damage captured microorganisms by modification of their DNA, lipids, thiols, tyrosine side chains, and active centers of metal-dependent proteins (Fang, ). Further damage of ingested pathogens is incurred by CAMPs via permeabilization of their cell membrane (Purdy & Russell, ). It must be kept in mind that the responses in mice, one of the most common model organisms, might be different from humans with respect to iNOS, ROS, and RNS. The final steps of bacterial destruction and clearance require phagolysosome fusion. All of the described destruction pathways are influenced by the host's immune status. Macrophage activation via cytokines, notably, interferon-gamma (IFN-γ), for instance, allows these host cells to control their intracellular predators (Cooper et al., ; North & Jung, ).
Phagosomal content can then be further processed toward the antigen presentation pathway (Wolf & Ploegh, ; Pieters, ). Components of Mtb, notably secreted proteins, are processed by macrophages and dendritic cells (DCs). Resulting peptides are loaded on the gene products of the major histocompatibility complex, and in this way, T cells are instructed to allow for an appropriate adaptive immune response (Amigorena et al., ; Tulp et al., ; West et al., ). Thus, T lymphocytes are critical for the control of Mtb during latent infection. Failure of T cells to maintain protective immunity promotes reactivation of TB.
Humans become infected with Mtb by inhaling minute aerosol droplets carrying a small number of bacteria (Kaufmann, ) (Fig. 0001). At the site of infection, the lung, Mtb bacilli are phagocytosed by alveolar macrophages. These cells are programmed to combat microbial intruders and to ultimately destroy them. However, Mtb manages to escape eradication by macrophages and survives within these cells (Armstrong & Hart, ; Kaufmann, ; Russell, ). The unique composition of the mycobacterial cell wall and envelope likely enables the tubercle bacillus to enter macrophages by employing multiple receptors such as Fc-, complement-, or mannose receptors and the DC-specific intercellular adhesion molecule-3-grabbing nonintegrin (DC-SIGN) (Brennan & Nikaido, ; Ernst, ; Greenberg, ; Cambi et al., ). Some receptors allow silent entry (CR), and others induce defense mechanisms (FcR). Once inside the resting macrophage, Mtb impairs phagosome maturation (Vergne et al., , ; Walburger et al., ; Robinson et al., ; Axelrod et al., ; Katti et al., ; Russell et al., ). Principally, phagosome maturation is a highly complex process in which the phagosome harboring a particle other than Mtb constantly interacts with the recycling endosome, secretory organelles, multivesicular bodies, and the endoplasmic reticulum (Desjardins, ). Following the oxidative burst, the phagosome gets acidified, a process lasting between 10 and 20 min. Acidification is mediated by proton pumps that reduce the neutral pH to an acidic pH of ca. 5.0 (Yates et al., ). Mtb arrests this process and therefore also downstream events (Sturgill-Koszycki et al., ). In contrast to Mtb-containing phagosomes, those harboring inert particles go on to fuse with the lysosome within an hour to 1.5 h. Acidification of the organelle is arrested by Mtb at pH 6.4, which is near the neutral pH and is significantly higher than in the terminal phagolysosome at pH 4.5–5.0 (Sturgill-Koszycki et al., ; MacMicking et al., ; Yates et al., ). IFN-γ activation of macrophages promotes delivery of Mtb into the mature phagolysosome, as shown by colocalization with green fluorescent bacilli, and probably remains viable in this extremely hostile environment (Via et al., ). Some Mtb mutants that fail to prevent phagosome maturation do survive in the resting macrophage, while others are attenuated (Pethe et al., ; MacGurn & Cox, ). An alternative strategy of Mtb has been established more recently: its egress from the phagosome into the cytosol of macrophages (van der Wel et al., ; Behar et al., ). Additional host defense mechanisms include apoptosis and autophagy (Behar et al., ; Deretic, ; Levine et al., ). Apoptosis is a highly regulated process mediated by host mechanisms which likely contributes to host protection. In contrast, necrosis is driven by exogenous insults which might benefit the pathogen rather than the host. Studies on experimental TB in mice have provided evidence of such an association (Pan et al., ). Virulent Mtb inhibit apoptosis by a number of antiapoptotic genes, and more recent evidence suggests a role of prostaglandin E2 in this mechanism (Behar et al., ). Autophagy is an essential mechanism for host cell integrity that can also serve as defense mechanism against bacterial pathogens (Deretic, ; Levine et al., ). In TB, contribution of autophagy to protection has been described (Alonso et al., ). In sum, the intracellular survival stratagem of tubercle bacilli comprises not only active manipulation of host defense mechanisms to neutralize and counteract a highly aggressive armamentarium of activated macrophages but also robust resistance against assault.
Adaptation of Mtb to the intracellular environment of macrophages
Mtb is shielded from the environment by a robust cell wall (Box 1). Upon phagocytosis by host cells, Mtb experiences drastic environmental changes and therefore has to realign its metabolism to assure survival. Genome-wide microarray techniques to study Mtb's transcriptional response to this transition have provided deeper insights into the nature of the phagosomal environment, which was suggested to be nitrosative, oxidative, low in oxygen tension, and limited in nutrients (Schnappinger et al., ). Additionally, the pathogen upregulated genes involved in lipid metabolism confirming previous evidence that lipids are critical for virulence of Mtb (McKinney et al., ; Movahedzadeh et al., ; Brzostek et al., ; Chang et al., ; Nesbitt et al., ). Isocitrate lyase (Icl) was identified as gate enzyme of the glyoxylate shunt, a short-cut of the tricarboxylic acid (TCA) cycle, bypassing the steps of carbon loss by CO2 formation. The glyoxylate cycle is mobilized in Mtb growing on fatty acids as exclusive carbon source and also during chronic infection in mice, suggesting that lipids are accessible as nutrients in vivo (McKinney et al., ). Metabolic pathways that are relevant during infection are shown in Fig. 0002.
Although the exact nature of carbon sources utilized during infection remains elusive, Mtb has been shown to metabolize host-derived cholesterol (Pandey & Sassetti, ). Disruption of mce4 encoding a cholesterol transporter results in the failure of Mtb to maintain chronic infection in mice while retaining full virulence during the acute phase comparable to observations made for an Icl-deficient mutant (McKinney et al., ; Mohn et al., ; Pandey & Sassetti, ). Recent data demonstrate that Mtb residing in phagosomes of macrophages utilizes triacylglycerol from the host cell to be stored in the form of intracellular lipid droplets (Daniel et al., ). Catabolism of cholesterol, odd-chain fatty acids, methyl-branched fatty acids, and amino acids funnels into propionyl-CoA, a C3 intermediate, which is toxic in excess (Savvi et al., ; Yang et al., ). However, propionyl-CoA toxicity can be avoided by condensing the C3 body with oxaloacetate to form succinate and pyruvate by the 2-methylcitrate cycle (Fig. 0002). Intriguingly, both mycobacterial enzymes Icl-1 and Icl-2 can act as 2-methylcitrate lyase (Munoz-Elias et al., ). Thus, mycobacterial Icl plays a critical dual role during infection (1) to bypass carbon loss using the glyoxylate shunt under limited nutrient availability and (2) to prevent the excessive accumulation of toxic propionyl-CoA. Recent studies carried out with steady-state chemostat cultures have demonstrated the importance of Icl in slow-growing Mtb metabolizing glycerol as main carbon source. Under such conditions, anapleuric reactions prevailed while the pathogen was also capable of carbon dioxide fixation as demonstrated by isotope flux (Beste et al., ). Propionyl-CoA metabolization can also be performed by the methylmalonyl-CoA pathway ending up in methylmalonyl-CoA, which can either be converted into succinyl-CoA by a vitamin B12-dependent mutase or directly incorporated into methyl-branched fatty acids (Savvi et al., ) (Fig. 0002). These fatty acids are found among the large family of unique mycobacterial lipids which build up the pathogen's cell envelope (Brennan & Nikaido, ; Jackson et al., ). Typically, the loss of cell wall components leads to decreased virulence (Glickman et al., ; Makinoshima & Glickman, ). In conclusion, propionyl-CoA detoxification is extremely critical for Mtb in vivo. The asymmetric cleavage of isocitrate by Icl produces glyoxylate, which is converted into malate and succinate. The TCA cycle intermediate malate can be used to generate pyruvate and further to replenish the pool of glycolytic intermediates by gluconeogenesis (Fig. 0002). Such intermediates are required to produce the essential building blocks of proteins, DNA, and the cell wall. Gluconeogenesis is critical throughout TB infection in mice and, thus, might be relevant in dormancy (Marrero et al., ).
Mtb tolerates low pH (i.e. 4.5–5.4) inside the phagolysosome of INF-γ-activated macrophages (Schaible et al., ; Via et al., ). The lack of complete acidification in Mtb-infected resting macrophages is likely – at least in part – caused by the exclusion of the phagosomal proton ATPase and by the secretion of mycobacterial urease, an enzyme producing neutralizing ammonia from urea (Reyrat et al., ; Sturgill-Koszycki et al., ; Grode et al., ). Yet, Mtb is exposed to acidic conditions in vivo as evidenced by the fact that the first-line TB drug pyrazinamide kills the pathogen only at low pH in vitro (Zhang & Mitchison, ). In further support of this notion, acid-sensitive Mtb mutants are attenuated in mice, and transcriptional analysis of the bacillus residing inside activated macrophages revealed upregulation of pH-responsive genes (Buchmeier et al., ; Raynaud et al., ; Rohde et al., ; Vandal et al., ). Interestingly, most acid-sensitive Mtb mutants show defects in genes associated with cell wall biogenesis (Vandal et al., ). This unique lipid-rich permeability barrier has been suggested to provide effective protection to protons more than 100 years ago (Metchnikoff, ). It is therefore not surprising that acid-sensitive mutants also confer hypersensitivity to detergents or lipophilic antibiotics owing to increased cell wall permeability (Vandal et al., ). Although a pH sensing adenylate cyclase (Rv1264) has been described in Mtb, it turned out to be nonessential in vivo (Tews et al., ; Dittrich et al., ). More recently, the membrane-associated serine protease Rv3671c has been characterized. Even though the precise functions of the protein remain elusive, the authors elegantly showed that an Rv3671c mutant failed to maintain a neutral intrabacterial pH in acidic culture medium as well as in the late (acidic) phagolysosome of activated macrophages (Vandal et al., ). Moreover, the mutant was attenuated in mice suggesting that acid resistance is critical for the virulence of the tubercle bacillus. A recent study demonstrated that aprABC, a locus unique to pathogenic mycobacteria, is involved in adaptation of Mtb to the phagosomal low pH environment. Disruption of aprABC conferred a defect in intracellular growth of the pathogen and influenced lipid abundance of intracellular stores and cell wall (Abramovitch et al., ). In sum, Mtb is resistant to elevated acidic stress in the late phagolysosome compartment of macrophages at conditions that are lethal for many other microbial pathogens (Huynh & Grinstein, ).
The defense repertoire that any bacterial intruder experiences in the phagosome of an activated macrophage is not limited to low pH but also includes ROS and RNS. The enzyme phagocyte oxidase (NOX2) transfers electrons from cytosolic NADPH to phagosomal oxygen to form superoxide anions. These highly reactive anions dismutate into hydrogen peroxide and finally produce toxic hydroxyl radicals, which are members of the ROS family (Bedard & Krause, ). Accordingly, this reaction is often referred to as ‘oxidative burst'. In monocytes and neutrophils, chlorination further adds to the toxicity of ROS (Bedard & Krause, ). RNS are produced by iNOS, an enzyme that generates nitrate and nitrite. The latter intermediate reacts at low pH to nitrous acid that forms nitric oxide and nitrogen dioxide, two highly reactive radicals (Nathan & Shiloh, ). Finally, the two molecules, nitric oxide and superoxide, form the toxic peroxynitrite (Beckman et al., ; Bogdan, ; Nathan & Ehrt, ). To avert toxicity caused by ROS and NOS, Mtb follows a dual strategy of detoxification and damage repair.
The mycobacterial enzyme catalase peroxidase encoded by katG converts hydrogen peroxide into water and oxygen. Accordingly, a katG loss-of-function Mtb mutant showed hypersensitivity to hydrogen peroxide in vitro (Ng et al., ). In NOX2-deficient mice, the mutant was fully virulent while a unique phenotype of transient attenuation was reported in iNOS−/− mice: Initial growth was followed by rapid decline and a lag phase of about 6–8 weeks characterized by stable bacterial burden in lungs, after which replicative activity was resumed. However, the molecular mechanisms of the observed phenomena remain unclear. Several other mycobacterial genes have been implicated in ROS and RNS detoxification (in-depth reviewed by Ehrt & Schnappinger, ):
Although the underlying molecular mechanisms require further elucidation, it is clear that Mtb employs a large variety of counterstrategies to detoxify ROS and RNS.
The second principal strategy to oppose the effects of highly reactive intermediates includes repair of the damage caused or degradation of affected biomolecules and their rapid replacement via de novo synthesis. Proteasomes are multiprotein complexes that degrade peptides by multiple proteolyic activities (Etlinger & Goldberg, ; Lowe et al., ). In Mtb, two putative accessory factors PafA and Mpa are involved in coupling the prokaryotic ubiquitin-like protein Pup to peptides that are destined for turnover and recognition of Pup-tagged proteins, respectively (Pearce et al., ; Striebel et al., ; Sutter et al., ). Mpa is an ATPase likely involved in unfolding and delivery of proteins into the proteolytic core of proteasomes (Wang et al., ). Mutations in the respective genes mpa and pafA increase the sensitivity of the tubercle bacillus to RNS (Darwin, ). Disruption of the mpa gene reduced virulence of Mtb in wild-type mice, which was less profound in iNOS−/− mice. Conditional silencing of the essential genes prcB and prcA, which encode proteolytic core proteins of the proteasome, not only conferred sensitivity to RNS but also impaired the survival of Mtb during chronic infection in mice (Gandotra et al., ). In sum, proteasome-mediated protein turnover becomes critical during RNS-related stress. Two different but not mutually exclusive scenarios or combinations could explain these results: (1) degradation of a transcriptional repressor that controls expression of proteins required for the synthesis of antioxidants; (2) removal of irreversibly damaged proteins with toxic potential (Darwin, ).
More recently, transcriptional profiling of Mtb and human macrophages during infection provided evidence for heavy metal poisoning (Tailleux et al., ; Botella et al., ). In particular, ctpC encoding a putative zinc efflux pump in mycobacteria was strongly induced upon infection, suggesting exposure of the bacillus to zinc ions inside the phagosomal compartment. Indeed, disruption of the ctpC gene rendered Mtb hypersensitive to zinc. While infection progresses, zinc ions are quickly released from stores inside host cells and translocated to the phagosome. Thus, zinc poisoning is likely exploited by macrophages to destroy Mtb. Indeed, zinc is known to be of particular importance for the immune system playing multiple roles, for instance, in defense and signaling (Rink & Gabriel, ; Haase & Rink, ). Although not yet directly shown, accumulating evidence suggests that Mtb counteracts the stressor zinc by expression of an appropriate efflux pump (Botella et al., ).
Inside the mature phagolysosome, Mtb experiences the permeabilizing properties of CAMPs (Purdy & Russell, ). The positively charged CAMPs – cathelicidin, hepcidin, and ubiquitin-related peptides – gain their bactericidal activity by disrupting the negatively charged bacterial cell wall (Alonso et al., ; Liu et al., ; Sow et al., ). Microorganisms lower their affinity to CAMPs by reduction of their negative surface charge (Peschel & Sahl, ). The Mtb plasma membrane contains a positively charged lipid consisting of phosphatidyl glycerol linked to lysine moieties. Its generation requires the lysine transferase LysX (Maloney et al., ). Sensitivity to positively charged antibiotics and a specific CAMP of neutrophils was increased in an Mtb lysX loss-of-function mutant. Virulence of this mutant was lowered in mice and guinea pigs, demonstrating a critical role of LysX in vivo (Maloney et al., ). The unique impermeable cell envelope of Mtb represents a physical barrier for CAMPs. Expression of the major porin of M. smegmatis, MspA, in Mtb, which does not encode an ortholog, increases membrane permeability (Mailaender et al., ). Concomitantly such strains become more sensitive to ubiquitin-derived CAMPs (Purdy et al., ). Probably, the most obvious strategy of pathogenic mycobacteria to protect themselves from phagosomal assault is to translocate from the hostile organelle to the cytosol of host cells. Such behavior has been suggested for Mtb and M. leprae (van der Wel et al., ). If true, this could have deep implications, not only for the fate of bacilli in host cells, but also for granuloma formation. Although the host macrophage has developed a remarkable arsenal of bactericidal mechanisms over millennia of coevolution, Mtb has learned to survive in this hostile intracellular environment. As a result, the coevolutionary tie shifted to the tissue level, namely to the formation of a remarkable self-organizing capsular structure for bacterial containment, the granuloma (Reece & Kaufmann, ).
Existence of Mtb inside granulomas
Infected macrophages migrate and thereby transport Mtb from the airways into pulmonary tissue sites. There, an inflammatory focus is formed comprised of infected macrophages and freshly immigrant monocytes (Fig. 0001). The primary lesion matures into a granuloma, the hallmark of TB, while some infected cells disseminate to seed secondary lesions in the lung. In the majority of individuals, the pathogen is controlled at this stage by the immune system and does not spread further: LTBI is established. The solid granuloma is not only the site of Mtb containment during latency but also the source of tissue damage at the early stage of disease (Reece & Kaufmann, ). In other words, it is the histological correlate of both protection and pathology. In humans, the granuloma shows high plasticity and three major types can be distinguished. These are (1) solid granulomas which contain Mtb; (2) necrotic granulomas typical for early stages of active TB; (3) caseous granulomas during end-stage or severe TB. These different stages are not distinct entities but form a continuum.
Solid granulomas prevail during LTBI. These highly structured tissue reactions are comprised of mononuclear phagocytes of different developmental stages, DCs, as well as T and B lymphocytes. Although T lymphocytes are the critical mediators of protection in TB, B lymphocytes are also abundant. Often, lymphocytes form an outer ring whereas in the central parts, mononuclear phagocytes, fibroblasts, and DCs predominate. The solid granuloma is typically encircled by a fibrotic wall that separates it from surrounding tissue. The burden of Mtb inside solid granulomas is low. Most likely these bacilli are in a stage of dormancy, characterized by low metabolic activity with non- to low-replicating persistence. Often, such bacteria are difficult to grow under normal culture conditions and therefore have been termed viable but not culturable (VBNC).
The necrotic granuloma remains well structured, but the center becomes increasingly necrotic, that is, composed of solid cell detritus which is often hypoxic. Later, Mtb organisms can be resuscitated: they start replicating and become metabolically active.
In caseous granlomas, the center becomes liquefied leading to cavity formation. The structure of these granulomas wanes. Evidence has been presented for a harmful role of polymorphic neutrophilic granuloyctes (Lowe et al., ). In cavitary lesions, high oxygen content is reestablished. Moreover, the caseous material provides a fertile source of nutrient-promoting growth of the pathogen up to some trillion organisms. Finally, Mtb finds access to blood capillaries and the alveolar space paving the way not only for dissemination to other organs but also for transmission to other individuals.
During active TB, different stages of granulomas coexist and provide a multitude of diverse microenvironments to which the pathogen has to adapt. Hence, the bacillus is found in different areas of the granulomas, intracellularly within the rim of host mononuclear phagocytes, some DCs, and perhaps fibroblasts as well as extracellularly in the caseous center consisting of host cell debris. It has been proposed that unfavorable conditions inside the granuloma, such as nutrient limitation and low oxygen tension, trigger the metabolic downshift of subpopulations of Mtb to dormancy. As most TB drugs target functions essential for growth, they fail to eradicate nonreplicating bacilli. This could explain the prolonged treatment time required to cure disease. In vitro models of dormancy have been developed to study nonreplicating persistence of Mtb. As early as , Loebel et al. observed that carbon starvation terminates the growth of the tubercle bacillus and causes a drastic drop in respiration, indicating a low metabolic rate. More recent work confirmed that nutrient-starved nonreplicating bacilli undergo a global downregulation of metabolic genes, including those involved in respiration (Betts et al., ). Such bacilli are extremely tolerant to TB drugs (Xie et al., ). However, nutrient-starved Mtb remains sensitive to inhibition of NADH dehydrogenase 2, a single protein enzyme that can serve as alternate entry point of electrons into the electron transport chain but lacks proton translocation capacity (Xie et al., ; Yano et al., ; Teh et al., ; Gengenbacher et al., ). Other characteristics of the pathogen analyzed in this in vitro dormancy model include significantly reduced intracellular ATP levels and very low but continuous respiration. The glyoxylate shunt enzyme Icl is essential for the survival of nutrient-starved nonreplicating Mtb (Gengenbacher et al., ). Promotion of the glyoxylate cycle under limited nutrient access is in the best interest of the bacillus, as no carbon is lost by CO2-formation in contrast to the citrate cycle. Icl is required for the maintenance of a chronic infection in the mouse model of TB, suggesting a relevant role of nutrient limitation in vivo (McKinney et al., ).
The influence of oxygen shortage on Mtb has been extensively studied. In 1996, Wayne and coworkers introduced an Mtb-in vitro dormancy model based on gradual oxygen depletion (Wayne & Hayes, ). The pathogen passes through two phases of declining metabolic activity to dormancy and phenotypic drug resistance. More detailed physiological characterization revealed reduction of intracellular ATP in hypoxic nonreplicating bacilli and sensitivity of the pathogen to further depletion, observations which later on were also made for nutrient-starved nonreplicating Mtb (Rao et al., ; Gengenbacher et al., ). Even though both models generate quiescent organisms by contrary conditions – carbon depletion in an oxygen-rich environment vs. oxygen starvation in a nutrient-rich medium – physiological overlaps identified could qualify as core features of dormancy. In line with the general upregulation of lipid metabolism genes during oxygen starvation of mycobacteria, hypoxic nonreplicating BCG accumulates neutral lipid triacylglycerol to form visible intracellular droplets. Importantly, such lipid stores are required for regrowth of the hypoxic nonreplicating organism in nutrient-rich medium (Low et al., ). Accumulation of triacylglycerol droplets might therefore be important during dormancy and could be useful for the identification of dormant mycobacteria.
On the genetic level, the adaption to changes in oxygen availability is mediated by the DosS/DosT-DosR regulatory complex that controls roughly 50 genes (Boon & Dick, ; Park et al., ). In other words, the DosR regulon governs metabolic shift of Mtb from aerobic to anaerobic functioning, ensures survival of the bacillus during hypoxia-induced in vitro dormancy, and controls reversal to replication upon reexposure to oxygen (Rustad et al., ; Leistikow et al., ). Furthermore, dosR responds to nitric oxide and carbon monoxide (Kumar et al., ). Yet, disruption of the dosR gene only slightly affects the survival of the pathogen in different animal models, such as mouse, guinea pig, or rabbit; the molecular basis of this finding remains unclear (Converse et al., ). The role of hypoxia in vivo was impressively analyzed by Via et al. (), who introduced the hypoxia-activated compound pimonidazole to different experimental animal species and directly measured oxygen tension in granulomas of guinea pigs, rabbits, and non-human primates. This study revealed that low oxygen pressure could restrict the growth of aerobic to microaerophilic Mtb in the hypoxic core of necrotic and solid granulomas. Note that TB histopathology of non-human primates most closely resembles active TB in humans (Leong et al., ). The characteristic continuum of granulomatous lesions in human TB is rarely reflected in small animal models. Guinea pigs show lesions of different types. As they are extraordinarily susceptible to Mtb, caseous lesions predominate. Mice, the most widely used experimental animals for research on immunology and infection, develop nonhypoxic ill-structured lesions, and distinct stages are not observed. To capitalize on the wealth of information available from the mouse model, mice that develop human-like TB pathology would be of great value. Recently, a mouse model that mimics the different stages of granulomas similar to human TB has been introduced. The iNOS-deficient mouse mutant infected with Mtb developed well-structured solid granulomas, which controlled the pathogen at low to intermediate load. Neutralization of IFN-γ led to granuloma necrosis with hypoxic areas, followed by massive caseation. In this model, cathepsin G (CatG) was identified as critical effector molecule of both protection and pathology. CatG activity, in turn, was controlled by serpin b3 and fine-tuning of this control mechanism seems to decide whether pathology or protection prevails (Reece et al., ). Production of RNS by the iNOS system represents a vital antimicrobial defense mechanism, but because of its strict oxygen dependence, it is likely insufficiently active in hypoxic areas of granulomas. In another murine model of TB, the sst1 locus of mice has been demonstrated to prevent formation of necrotic lesions. The intracellular pathogen resistance 1-protein encoded within this locus possesses the ability to direct infected macrophages to undergo apoptosis rather than necrosis (Pan et al., ). Such necrotic lesions have recently been shown to be hypoxic (Harper et al., ). Whether ‘human-like' mouse models have potential for broad application in TB research has yet to be determined. The widespread assumption that chronic TB infection is caused by a rather static equilibrium of slow or nonreplicating bacilli has recently been questioned. Authors engineered Mtb to harbor an instable plasmid that was lost during cell division. This replication clock used to study TB infection in mice has provided evidence for active replication of Mtb, not only during the acute stage, but also throughout the chronic phase of infection (Gill et al., ).
In vitro models aim at reproducing impacts of the host environment on Mtb. Thus, the pathogen has been debarred from iron or phosphate and exposed to low concentrations of nitric oxide to mimic either the phagosome of host macrophages or the necrotic center of granulomas (Fisher et al., ; Ohno et al., ; Rifat et al., ). Other studies in M. bovis BCG have combined different stressors in one model system to better represent the microenvironment of mycobacteria in vivo (Bryk et al., ). Most recently, drug-tolerant persister bacilli have been isolated from in vitro cultures of Mtb by D-cycloserine treatment. A few persisters were found during lag and early exponential phases, while they made up to 1% in late exponential and stationary phases. The global transcriptome of such persisters was then profiled and compared to the transcriptomes obtained from various in vitro dormancy models. Authors identified a set of five genes upregulated in all models that probably represents a core dormancy response (Keren et al., ). These were: acr2, encoding a heat shock protein; the transcriptional regulator gene Rv1152; pdhA, the gene of a putative pyruvate dehydrogenase subunit; a hypothetical protein encoded by Rv2517c; and lat, encoding an L-lysine-epsilon aminotransferase. To date, an unusually high number of 65 toxin–antitoxin (TA) loci were identified in the genome of Mtb. A TA module produces a toxin that is detoxified by its respective antitoxin. In E. coli, a number of TA modules are relevant in dormancy (Keren et al., ; Vazquez-Laslop et al., ). Interestingly, in the recent transcriptome study of Keren et al., 10 TA loci were overexpressed in Mtb persisters, suggesting their importance in survival without replication. Altogether, in vitro dormancy models of Mtb suggest that stress-related genes and alternative pathways are upregulated, while the genes of central metabolic routes, including glycolysis, TCA cycle, energy production and respiration, are downregulated. The specific roles of distinct genes such as TA modules remain to be elucidated.
Reactivation and resuscitation
In the human host, Mtb persisting in a dormant stage causes LTBI without clinical disease. While Mtb is well equipped for persistence in the host, the term ‘persister' is used for those Mtb organisms that are phenotypically resistant to drugs although they are in fact genetically susceptible to these antibiotics. The reason for this is probably transformation of bacteria into a nonreplicating stage with low-to-absent metabolic activity – the precise conditions of dormancy. Principally, this feature underlies the so-called Cornell model of Mtb persistence in mice. In this model, animals infected with Mtb are treated with the drugs pyrazinamide and isoniazid to reduce bacterial load to a level, which is nondetectable by culture. At first sight, drug treatment achieves sterile eradication of the pathogen, but after termination of drug treatment, some bacilli recover and grow to high abundance causing reactivation of TB (McCune et al., ). Prolonged culture of stationary-phase Mtb can generate bacteria that fail to grow to visible colonies on agar. Regrowth of these VBNC organisms was only supported in spent media taken from exponentially growing Mtb (Shleeva et al., ). VBNC bacilli in sputum of patients with TB could lead to false-negative results of diagnostics, as those organisms are ‘invisible' in standard cultures (Mukamolova et al., ). A very recent report showed that M. smegmatis (Box 1) generates cell-to-cell heterogeneity by asymmetric growth in combination with time-controlled cell division. Most importantly, distinct subpopulations showed different susceptibility to antibiotics (Aldridge et al., ).
Accumulating evidence suggests that regrowth of dormant Mtb is initiated by resuscitation, which is the reestablishment of metabolic and replicative activity. Resuscitation has been studied at the molecular level in Micrococcus luteus where a so-called resuscitation-promoting factor (Rpf) has been shown to induce resuscitation. Rpf orthologs of Mtb possess a conserved domain with putative lysozyme activity and therefore might cleave the peptidoglycan network that makes up the cell wall (Cohen-Gonsaud et al., ). Similarly, germination of spores of Bacillus anthracis begins with hydrolysis of the cell wall (Giebel et al., ). Recent studies in Bacillus subtilis have revealed a signaling cascade that is initiated by peptidoglycan degradation products (Shah et al., ). Altogether, regrowth could be initiated through a cell wall hydrolysis step, but further steps involved are thus far unknown. The genome of Mtb comprises five rpf genes, and it has been claimed that such genes can facilitate recovery of Mtb in sputum of patients with active TB or in freeze-dried M. bovis BCG (Wu et al., ; Mukamolova et al., ). As Mtb possesses several rpf genes, redundancy likely exists. As a result, only multideletion mutants, not single knockouts of rpf genes, show impaired resuscitation in vitro and attenuation in mice (Kana et al., ). In the mouse, chronic progression of disease characterized by high bacterial burden is observed, while reactivation in humans develops from minute bacterial load of LTBI. Hence, mice are probably not a suitable model to study reactivation of TB.
Evidence for coexistence of different Mtb stages in infected individuals is increasing. Thus, probably only a few dormant bacilli coexist in face of a large number of metabolically active replicating organisms during active TB. The reciprocal is also true: during latent infection, in addition to nonreplicating metabolically inactive (i.e. dormant) Mtb, some actively replicating Mtb are present. In other words, the equilibrium of dormant/replicating Mtb represents a distinguishing factor between LTBI and active TB.
Isoniazid only targets replicating Mtb and yet has been used widely and successfully for the chemoprophylaxis of LTBI where there is considered to be elevated risk, suggesting that during several months treatment time, bacteria transform into an isoniazid-susceptible stage (Fox et al., ). A more recent hypothesis describing latent infection as a dynamic process of constant reinfection could explain the efficacy of isoniazid therapy during latency (Cardona, ). Although formal proof is lacking, the clinical observation is taken as evidence for sporadic emergence of some replicating Mtb during LTBI. Reciprocally, addition of Rpf significantly increases recovery of Mtb in sputum from active patients with TB providing circumstantial evidence that these bacteria are resuscitated from dormancy (Mukamolova et al., ). It has been argued that the long treatment time of six or more months to cure TB depends, at least in part, on the coexistence of both replicating and nonreplicating Mtb as current drugs preferentially, if not exclusively, target metabolically active replicating Mtb. On the one hand, replicating bacilli are the main culprits causing active disease; on the other hand, they are the target of current chemotherapy. In contrast, the dormant pathogen is likely a ‘bystander' in disease but mainly contributes to phenotypic drug resistance. Hence, the dormant pathogen serves as a reservoir of renascent active Mtb to sustain pathology and disease. It is therefore assumed that treatment with available drugs has to eradicate replicating tubercle bacilli in several waves as they are resuscitated from dormancy.
Currently, the scenario of equilibrium between dormant and replicating Mtb during the course of infection from LTBI to active TB remains speculative. In one attractive model, a few dormant Mtb resuscitate either stochastically or through signals such as those emitted by Rpf (Fig. 0003). These few Mtb organisms have been termed ‘scouts' as they sense the environment for its appeal to awake and replicate. Under adverse conditions, the scouts die and the bulk of organisms remain dormant. In an appealing environment, scouts send activation signals to the dormant majority of bacteria which then resuscitate (Epstein, ; Chao & Rubin, ). The biochemical nature of these signals could involve Rpfs but has to be further investigated. In an environment attractive for growth, the bacterial population starts replicating, causes pathology, and leads to reactivation of TB. A small population will stay in or fall into dormancy thus showing tolerance to drug treatment (Lewis, ). Active disease caused by reinfection could involve similar factors as Mtb newly entering the host is actively growing and hence can provide wake-up signals to dormant bacilli in an LTBI host.
Deletion of Rv2623, a gene of unknown function under the control of the dosR regulon, for instance, confers hypervirulence to the tubercle bacillus in mice and guinea pigs, while its overexpression leads to growth delay in vitro (Drumm et al., ). Thus, Rv2623 could be involved in restricting replication. Reciprocally, regrowth of dormant bacilli could be promoted by the downregulation of Rv2623. The dosR regulon provides an instructive example of how environmental changes can trigger a complex adaptation program in Mtb. It is well understood how DosR responds to decreasing oxygen tension (Park et al., ). But, how a change in oxygen availability is sensed on a molecular level and ultimately translated into an impulse for the dos regulatory system is not yet understood. Simply turning off the ‘dos program' will most likely not be sufficient to allow for resuscitation. Although the molecular mechanisms of resuscitation are largely unknown, evidence for several aspects is unfolding.
The last decade has witnessed a remarkable increase in our knowledge about Mtb, not only on the bacillus itself but also on the communication with its host. We have learned that Mtb has invented complex mechanisms to survive in the intracellular environment, that it counteracts or evades the numerous defense mechanisms of macrophages, and that the cross-talk between pathogen and host immune system is focused on granulomas, which serve as both, habitat and containment for Mtb. However, many interesting observations remain fragmented in our current view of Mtb. Remarkable efforts are ongoing to integrate the relatively new science of systems biology into mycobacterial research (http://www.systemtb.org/ and http://www.broadinstitute.org/annotation/tbsysbio/home.html). Ideally, this will tie up loose ends and improve our understanding, allowing for a broad network-oriented view of Mtb and its host.
After decades of dormancy, TB drug discovery has reawakened. Although the boost that was expected from whole genome sequencing of Mtb and application of high-throughput screening to target-based approaches has not paid off so far, nine TB drug candidates are currently undergoing clinical investigation – more than ever before (Cole et al., ; Payne et al., ; Ma et al., ). Most of these compounds belong to existing classes of antibiotics and thus are derivates of the respective parental molecule: oxazolidinones target protein synthesis; fluoroquinolones are well-known DNA gyrase inhibitors; 1,2-ethylene diamines interfere with cell wall biosynthesis pathways; nitroimidazoles cause multiple damage to Mtb by RNS. The target of sudoterb has yet to be identified. The ATP synthase inhibitor TMC207 is a diarylquinoline and represents a relatively new antimycobacterial class (current TB drug discovery reviewed by Koul et al., ). Which new entity will complete the stony road to drug application is yet to be seen. Applying our increasing knowledge of host–pathogen interactions to TB drug discovery will most likely result in more new candidate compounds (Nathan et al., ).
Most impressive are the dynamics and the plasticity of the infection process, which unfolds as a continuum in which different populations of Mtb, as well as different pathologic forms of granulomas, coexist. As a corollary, different stages of infection are characterized by different ratios in the abundance of pathogen populations rather than by distinct periods of life cycle. Accordingly, analysis of single cells ‘frozen' in a distinct stage, rather than of populations will be required in the future.
Development of next-generation vaccines will benefit from our deeper insights into the adaption of Mtb to different host environments, as well. Currently, over a dozen vaccine candidates are at different stages of clinical trial development (Kaufmann et al., ; Kaufmann, ). All these candidates are preexposure vaccines, which do not prevent or eradicate infection with Mtb but rather aim at precluding emergence of active TB. Accordingly, these vaccine candidates stimulate an immune response that targets the pathogen promptly after infection, presumably by means of antigens expressed by the metabolically active, replicating pathogen. However, during LTBI, Mtb changes its genetic program and antigens expressed by the dormant pathogen prevail. To sustain efficacious control of dormant Mtb, the vaccine-induced immune response needs to target antigens expressed by dormant bacilli, the so-called latency antigens (Kaufmann, ). This stratagem becomes even more valid for postexposure vaccines that are administered during LTBI – more than 2 billion individuals with elevated risk of developing TB. A first example of a postexposure vaccine comprising an antigen selectively expressed during nutrient starvation in addition to canonical preexposure vaccine antigens has been described recently (Aagaard et al., ; Kaufmann, ). This vaccine was highly successful in containing Mtb in the mouse model. Hence, better understanding of the tubercle bacillus' life cycle will facilitate rational design of next-generation vaccine candidates.
Definition of terms
Active TB: characterized by the presence of clinical symptoms caused by a high bacterial burden in the lung and sometimes in other organs; patients spread the disease.
Latent TB infection (LTBI): asymptomatic infection with a low number of Mtb in the absence of clinical signs; infected individuals do not spread the disease.
Reactivation of TB: transition from latency to full-blown active disease.
Relapse of TB: redevelopment of active disease after incomplete or false treatment with antibiotics; single- or multiple-drug resistance is often observed.
Active Mtb: metabolically active replicating bacilli; susceptible to drug-inhibiting processes essential for growth (i.e. DNA replication, RNA synthesis, cell wall biogenesis).
Dormant Mtb: nonreplicating bacilli maintaining full viability at a very low metabolic rate; organisms show minor susceptibility or phenotypic drug resistance to antibiotics targeting functions required for growth.
Resuscitation of Mtb: transition of the pathogen from dormancy to growth.
Drug resistance: inheritable resistance to a drug conferred by genetic mutation.
Phenotypic drug resistance: noninheritable resistance to a drug conferred by a specific metabolic state (usually dormancy).
Viable but not culturable (VBNC): refers to a state of viable Mtb that is not capable of colonizing on nutrient-rich solid media without being resuscitated.
We would like to thank Olivier Neyrolles and colleagues for providing most recent experimental data prior to publication. We are grateful for outstanding editorial support of M.L. Grossman and excellent graphic design of D. Schad. This work received financial support from the European 7th Framework Program SYSTEMTB (HEALTH-2009-2.1.2-1-241587) and the National Institutes of Health (SysBio, 745090 HHSN272200800059C).