The Role of ESX-1 in Mycobacterium tuberculosis Pathogenesis

ABSTRACT

In this article, we have described several cellular pathological effects caused by the Mycobacterium tuberculosis ESX-1. The effects include induction of necrosis, NOD2 signaling, type I interferon production, and autophagy. We then attempted to suggest that these pathological effects are mediated by the cytosolic access of M. tuberculosis-derived materials as a result of the phagosome-disrupting activity of the major ESX-1 substrate ESAT-6. Such activity of ESAT-6 is most likely due to its pore-forming activity at the membrane. The amyloidogenic characteristic of ESAT-6 is reviewed here as a potential mechanism of membrane pore formation. In addition to ESAT-6, the ESX-1 substrate EspB interferes with membrane-mediated innate immune mechanisms such as efferocytosis and autophagy, most likely through its ability to bind phospholipids. Overall, the M. tuberculosis ESX-1 secretion system appears to be a specialized system for the deployment of host membrane-targeting proteins, whose primary function is to interrupt key steps in innate immune mechanisms against pathogens. Inhibitors that block the ESX-1 system or block host factors critical for ESX-1 toxicity have been identified and should represent attractive potential new antituberculosis drugs.

BACKGROUND

The identification of ESAT-6 secretion system-1 (ESX-1) as a virulence determinant of Mycobacterium tuberculosis is a major discovery in the history of tuberculosis research. ESX-1 is encoded by a genetic locus known as RD1, which stands for “region of difference” and is one of the deleted regions in the vaccine strain Mycobacterium bovis bacille Calmette-Guérin (BCG) for humans (1). The first evidence emerges from the finding that the absence of RD1 is responsible for the attenuation of BCG’s virulence (2–4). Introduction of RD1 into BCG is sufficient to induce BCG growth in lung and spleen, granuloma formation in lung, splenomegaly, and inflammation and abscesses in liver and kidney in mice (4). Conversely, deletion of RD1 in the virulent H37Rv M. tuberculosis strain inactivates the ability of H37Rv to enable rapid bacterial replication in lung and spleen, to cause lung histopathology and death in mice (2, 3). Lung sections from infected mice show evidence of macrophage lysis, which is a RD1-dependent process (2). Consistent with this observation, Lewis et al. describe the requirement of RD1 for H37Rv to grow within and kill human macrophages (3).

Macrophage is the primary cell harboring M. tuberculosis during tuberculosis infection in humans (5). Accordingly, the discovery that RD1 mediates the induction of macrophage death by M. tuberculosis suggests that the outcome of the interaction between RD1 and macrophage host cell will be a key determinant of the virulence of M. tuberculosis pathogenesis. Significant progress toward understanding how ESX-1/RD1 facilitates tuberculosis pathogenesis has been made recently. In this article, we aim to provide an updated overview about the progress and put forward emerging concepts for improving our understanding of the nature of ESX-1-induced pathogenesis. In doing so, we also highlight some recent developments in designing intervention strategies based on targeting the RD1-encoded proteins.

ESX-1 SECRETION SYSTEM

The RD1 locus encodes a type VII secretion system known as ESAT-6 secretion system-1 (ESX-1) in Mycobacterium species (6). RD1 comprises nine genes, two of which are the secreted protein ESAT-6 (early secreted antigenic target of 6 kDa) and CFP-10 (culture filtrate protein of 10 kDa). Both of these proteins are immunodominant antigens and major virulence factors of M. tuberculosis. ESAT-6 and CFP-10 are cosecreted as a heterodimeric complex and are also essential components of the ESX-1 secretion system. Other RD1 genes encode the ATPase Rv3871, the transmembrane protein Rv3870, and the channel protein Rv3877. The ESAT-6-CFP-10 complex is recognized by the ATPase Rv3871, which is localized to the inner membrane through its interaction with Rv3870. The complex is then translocated across the inner membrane, most likely through the channel protein Rv3877. Other RD-1 genes might facilitate the biogenesis of the secretion system. Non-RD-1 genes are also required for proper functioning of ESX-1. EspR is a transcriptional regulator of ESX-1. EspA is an ESX-1-secreted substrate, and its secretion is required for the secretion of ESAT-6 and CFP-10. Additional detail about how the ESX-1 secretion system works is beyond the scope of this article. Interested readers should consult the comprehensive review by Abdallah et al. (6).

KEY OBSERVATIONS LEADING TO RECENT EVIDENCE OF PHAGOSOME DISRUPTION BY ESX-1

M. tuberculosis is phagocytized by alveolar macrophage after it is inhaled into the lungs of an individual. The internalized bacteria then reside within a membrane-bound compartment called phagosome. Armstrong and Hart used electron microscopy to examine these compartments in detail (7). Phagosomes containing nonpathogenic BCG normally undergo maturation, which is then followed by fusion with multiple lysosomes. The fusion events bring lysosomal digestive enzymes into the lumen of the phagosome for degradation. Indeed, damaged BCG is frequently observed. Using a sophisticated electron microscopy technique, Peters’ group revisited the pioneer work of Armstrong and Hart (8). To their surprise, they found evidence of phagosome maturation and lysosome fusion in phagosomes containing H37Rv that contain ESX-1 (8). These are transient events and are only observed during early infections. In contrast, phagosome maturation and lysosome fusion are uninterrupted and occur continuously in macrophages infected with BCG. The authors show that, after the early engagement with lysosome, M. tuberculosis translocates from phagosome into cytosol and this process requires ESX-1 (8). This translocation has been proposed as a virulence mechanism (9). Similar observations were made earlier by McDonough et al. (10).

How does M. tuberculosis impair phagosome-lysosome fusion? The fusion requires membrane localizations of lysosome-associated proteins (LAMP)-1 and LAMP-2 on phagosomes (11). Membrane disruption should therefore affect phagosome-lysosome fusion. Importantly, membrane disruption by M. tuberculosis was first observed in 1984 by Leake et al. and Myrvik et al. using electron microscopy (12, 13). The disruption is observed only in macrophages infected with virulent M. tuberculosis H37Rv but not with attenuated BCG or H37Ra, consistent with the results of Armstrong and Hart. This finding has recently been confirmed by other methods. Damaged phagosomes can be marked by galectin-3 and are indeed observed in M. tuberculosis-infected macrophages in an-ESX-1-dependent fashion (14, 15). Simeone et al. have developed a single-cell fluorescence resonance energy transfer assay to monitor phagosome disruption (16). Using this assay, the authors demonstrate phagosome disruption in mouse spleen and lungs as revealed by the assay based on fluorescence resonance energy transfer (17). In vitro, the assay reveals evidence of phagosome disruption within 24 h of macrophage infections (17). Interestingly, most of the cytosolic translocation event for M. tuberculosis requires 2 or 3 days to occur (9).

It is possible that the appearance of cytosolic localization of M. tuberculosis arises from collapse of phagosomal membrane as a result of continuous disruption of the phagosomal membrane. According to this scenario, cytosolic translocation of M. tuberculosis and impairment of phagosome-lysosome fusion can be seen as downstream effects of phagosome disruption. This has an important implication in M. tuberculosis pathogenesis. Phagosome disruption is a proposed mechanism of “patterns of pathogenesis.” This concept was proposed by Vance, Isberg, and Portnoy to help in understanding pathogenic host responses to virulent microorganisms (18). In the next section, we will present new evidence that fits the idea that phagosome disruption is responsible for all major pathogenic responses as a result of tuberculosis infections. Therapeutic implication of this idea will then be discussed.

DAMAGE OF THE M. TUBERCULOSIS-CONTAINING PHAGOSOME: CONSEQUENCES OF CYTOSOLIC ACCESS BY M. TUBERCULOSIS

Phagosome disruption compromises the integrity of the phagosomal membrane and can lead to cytosolic access of the lumen content within the phagosome compartments. There are at least four major cellular responses as a result of such cytosolic access of M. tuberculosis-derived materials. And for each of the responses, the element of pathogenesis can be readily identified. First, phagosome disruption can lead to activation of the cytosolic inflammasome receptor NLRP3 (15). NLRP3 is activated by M. tuberculosis ESX-1 and the activation triggers downstream activation of caspase-1, which then facilitates secretion of cytokines interleukin 1 beta (IL-1beta) and interleukin 18 (IL-18) (19). These cytokine productions are protective against M. tuberculosis infections (20, 21). NLRP3 activation can also lead to the induction of a programmed necrotic death in M. tuberculosis-infected macrophages (15). M. tuberculosis mutants carrying a defective ESAT-6 that are unable to disrupt the phagosome cannot activate NLRP3 (as measured by IL-18 production) and cannot induce necrosis (15). This programmed cell death is not known to have any negative impact on the viability of M. tuberculosis. This kind of death is necrotic and highly inflammatory, which contributes to significant tissue damage. Indeed, macrophage necrosis is suggested to play a key role in all major stages of tuberculosis (22). It is presently unclear how NLRP3 senses phagosome disruption. NLRP3 is known to sense cell swelling and hyperosmotic stress (23, 24). It remains to be shown whether ESAT-6 promotes cell swelling and alters cellular osmolarity through its phagosome-disrupting activity. This scenario is possible since it has been shown that ESX-1 can cause residual effects on plasma membrane upon contact with macrophages (25).

Second, cytosolic access can activate a cytosolic surveillance pathway that triggers induction of type I interferon production. This induction by M. tuberculosis is dependent on ESX-1 (26). The phagosome-disrupting effect of ESX-1 provides cytosolic access of M. tuberculosis materials such as extracellular DNA, which acts as a specific activating ligand for the STING-TBK1-IRF3 signaling axis for type I interferon production (27). The cytosolic DNA sensor that is responsible for direct recognition of M. tuberculosis DNA and for activating STING has recently been identified by three independent groups to be cyclic GMP-AMP synthase (28–30). Interestingly, mice deficient in IRF3 are resistant to long-term M. tuberculosis infection (27). This observation is consistent with the idea that phagosome disruption is a pattern of pathogenesis. According to the pattern of pathogenesis model, phagosome disruption is sensed by a host mechanism that initiates a pathogenic response. In our case of M. tuberculosis infection in macrophage, cytosolic localization of M. tuberculosis DNA as a result of phagosome disruption triggers the STING-IRF3-type I interferon pathway responsible for pathogenesis. Indeed, prolonged induction of type I interferon by intranasal poly-l-lysine and carboxymethyl cellulose exacerbates tuberculosis in mice (31). Limiting excessive type I induction by IL-1beta confers host resistance against tuberculosis in mice (32). But M. tuberculosis has an ability to inhibit IL-1beta production, through a yet-to-be-defined ESX-1-dependent mechanism to suppress the AIM2 inflammasome that normally activates caspase-1 and triggers IL-1beta production upon recognizing cytosolic DNA (33). Consistent with these observations, excessive induction of type I interferon and reduced IL-1beta responses in tuberculosis patients are associated with tuberculosis exacerbation (32). Furthermore, transcriptional signature dominated by type I interferon signaling correlates strongly with the radiological extent of disease in patients with active tuberculosis and the signature reverts back to the levels seen in healthy controls after anti-M. tuberculosis treatment (34). Taken together, phagosome disruption as a result of ESX-1 is responsible for the pathogenesis of tuberculosis, at least in part through induction of type I interferon production by cytosolic M. tuberculosis DNA.

Third, cytosolic access enables peptidoglycan recognition by the cytosolic receptor NOD2 (35). M. tuberculosis induces the NOD2 pathway in an ESX-1-dependent fashion (36). This induction triggers type I interferon expression (36). Additionally, the NOD2 pathway also plays an important role in apoptosis (35). A role for NOD2 in apoptosis is consistent with the demonstration of a key role for ESX-1 in apoptosis induction by Aguilo et al. (37). Although the authors do not show that the ESAT-6-induced apoptosis is mediated by the NOD2 pathway, they do show that induction of apoptosis requires p38 MAPK activity, which is known to mediate signaling from the NOD2 pathway (38).

The fourth consequence of ESX-1-induced phagosome disruption is impairment of autophagic flux and fusions with lysosome. Autophagy is a cellular mechanism for degradation of intracellular protein aggregate. A selective form of autophagy known as xenophagy targets damaged organelles or intracellular pathogens. Indeed, autophagy has been shown to target M. tuberculosis for degradation in mouse macrophages (39). The initial step of autophagy involves the processing of microtubule-associated protein 1 light chain 3 (LC3) into the smaller isoform 2 (LC3-II). The processed LC3-II is then associated physically with the so-called isolation membrane, which is used eventually to engulf its targeting intracellular materials to create a closed membrane compartment known as autophagosome. Processed LC3-II recognizes polyubiquitin chains that mark intracellular contents destined for autophagy. In the case of damaged mitochondria or M. tuberculosis-disrupted phagosomes, the polyubiquitin chain is produced by the ubiquitin ligase parkin and then is recognized by the autophagy receptors NDP52 and p62 for LC3-II binding (40, 41). ESX-1 is necessary for polyubiquitin chain labeling and recruitments of NDP52, p62, and LC3-II on M. tuberculosis-containing phagosomes (42). It has been further shown that recognition of cytosolic DNA by the STING-dependent pathway is required for the polyubiquitination of M. tuberculosis-containing phagosome for autophagy (42). To complete the autophagy process, the autophagosome must fuse with lysosome for degradation of the phagosomal content. It is this step where autophagy is interrupted by ESX-1 (43). Indeed, ESX-1 genes are found to be critical for blocking phagosome maturation (44, 45). It remains unclear whether ESX-1 can directly permeabilize the outer autophagosomal double membrane to block fusion with lysosomes. M. tuberculosis-secreted effectors previously shown to block phagosome acidification and fusions with endosome or lysosome include PtpA, SapM, and EsxH (46–48). It is probable that ESX-1 could provide access of these effectors into the cytosol and impair the autophagosome-lysosome fusion through its membrane-permeabilizing effect.

IS M. TUBERCULOSIS AN AMYLOID DISEASE MEDIATED BY ESX-1?

In their investigation about the role for ESX-1 in tuberculosis pathogenesis, Hsu et al. study how an ESX-1-secreted product mediates the lysis of macrophage. The authors first notice an earlier report from Ralph Isberg’s group, which shows that host cell lysis caused by Legionella pneumophila is mediated by lethal ion fluxes as a result of the pore-forming ability of a protein secretion system and that exogenous glycine can prevent such ion fluxes as well as host lysis (49). To test whether M. tuberculosis may process similar pore-forming ability, Hsu et al. find that pretreatment of glycine prevents the M. tuberculosis-promoted macrophage lysis (2). Hsu et al. then show further that purified ESAT-6 induces ion fluxes across artificial bilayer membrane and eventual total destruction of the artificial membrane (2). Therefore, their results strongly indicate that ESAT-6 can form pores on membrane.

To further understand how ESAT-6 forms pores, Hsu et al. also observed that ESAT-6 forms structures similarly observed in amyloidogenic intermediate soluble protein (2). This intermediate structure is believed to be the precursor of insoluble amyloid fibril. In humans, amyloid aggregate is associated with several pathological conditions, including Alzheimer’s disease, Parkinson diseases, and type 2 diabetes mellitus. Current consensus on amyloid diseases suggests that amyloidogenic prefibril intermediate is the most toxic entity and that aggregation of the amyloidogenic intermediate into insoluble form is a detoxification mechanism (50). The toxic species in amyloid diseases are small oligomers with pore-forming activity composed by the amyloidogenic intermediate (51). By analogy, it is possible that the amyloidogenic nature of ESAT-6 can similarly form small oligomer with pore-forming capability. According to this possibility, M. tuberculosis expressing an ESAT-6 mutant defective for amyloid aggregate should not possess pore-forming activity and, as a result, should not produce phenotypes associated with phagosome disruption. Such ESAT-6 mutants that are defective for forming amyloid aggregate have been described biochemically by the laboratory of David Eisenberg (52). When introduced in M. tuberculosis, these ESAT-6 mutants can no longer induce macrophage lysis, induce NLRP3 inflammasome, and damage phagosome (15). These observations are therefore consistent with the idea that the amyloidogenic nature of ESAT-6 is critical for disrupting the phagosomal membrane during macrophage infections. Nevertheless, this idea must be confirmed further by a demonstration of the existence of amyloid-like ESAT-6 aggregate, as well as by identification of ESAT-6 amyloid-forming inhibitors that block ESAT-6’s cellular effects. In addition, ESAT-6 will not be able to form amyloid within macrophages if it cannot be present without binding to CFP-10. Thus, the amyloid hypothesis is critically dependent on showing that ESAT-6 exists in macrophages without CFP-10, which remains to be shown.

There has been some new exciting progress in Alzheimer’s disease research using the approach of targeting amyloid in treating Alzheimer’s disease (53). If ESAT-6 does indeed form toxic amyloid-like intermediate, the opportunity to disrupt such behavior may represent new opportunities for therapeutic intervention.

REGULATIONS OF ESX-1

ESX-1 secretion is regulated by at least two two-component sensing regulatory systems. The first one is PhoPR. A point mutation in PhoP causes drastic reduction of ESAT-6 secretion (54, 55). ESAT-6 secretion is dependent on EspA secretion (56). Expression of espA in turn is regulated positively by PhoPR as well as the PhoPR-regulated transcriptional factor EspR (57). Accordingly, ESAT-6 secretion can be triggered by PhoPR, since it induces espR and espA expressions. Interestingly, the transcriptional factor EspR is also a substrate of ESX-1 and is secreted by ESX-1. Raghavan et al. suggest that the secretion of the positive regulator EspR represents a negative feedback loop for the ESX-1 secretion (58). EspR is exported when ESX-1 secretion is activated. If its intracellular level falls below a threshold necessary to drive espA expression to replace the secreted EspA, there will not be enough EspA and ESX-1 secretion will shut down (58). However, ESX-1 secretion can resume when EspA accumulates as a result of signaling from PhoPR.

Tan et al. identify acidic pH and increase in chloride concentration as two synergistic cues sensed by PhoPR (59). These two cues correlate with phagosome maturation and immune activation. In response to these immune stresses PhoPR signals M. tuberculosis to turn on ESX-1 secretion. Phagosome maturation is promoted by interferon gamma (IFN-γ) (60). In mice macrophages, immune activation by IFN-γ induces effective M. tuberculosis-killing activities (61, 62). But IFN-γ fails to do so in human macrophages infected with M. tuberculosis. Instead, it promotes necrosis in M. tuberculosis-infected human macrophages without harming M. tuberculosis (63). We have recently reported that IFN-γ promotes necrosis in M. tuberculosis-infected human macrophages and this process is dependent on the ESX-1 system (64). It seems plausible, although not demonstrated, that the necrosis promoted by IFN-γ is mediated by the ability of IFN-γ to promote phagosome maturation, which then activates PhoPR to promote ESX-1 secretion and ultimately cause necrosis.

Another two-component system that regulates ESX-1 is MprAB, which senses envelope stress such as protein misfolding in extracytoplasmic space. The absence of MprAB blocks secretions of ESAT-6 and EspA, but not secretion of CFP-10 (57). In the mprAB mutant, protein levels of EspA and ESAT-6 are unchanged although transcription of espA increases (57). It has been proposed that the absence of MprAB causes increased degradation resulting in defective secretion (57).

ESX-1 secretion is also regulated by ATP levels in M. tuberculosis. Zhang et al. show that low ATP levels shut down ESX-1 secretion system (65). Bedquiline inhibits ATP production in Mycobacterium species and completely blocks ESX-1 secretion (65). Sensing of low ATP levels is mediated by EspI, which is encoded by a gene within the RD1/ESX-1 locus. Mutations within the ATP-binding motif in EspI prevent turning off ESX-1 secretion, even in the presence of bedquiline (65).

ESPB: AN ESX-1 SUBSTRATE THAT TARGETS INNATE IMMUNE MECHANISMS

ESX-1 also secretes other proteins in addition to ESAT-6. One example is the 60-kDa protein EspB encoded by a gene within the RD1 locus. Chen et al. report that the espB mutant of M. tuberculosis is defective for cytotoxicity against the human macrophage cell line THP-1 (66). Interestingly, purified EspB is capable of binding to phosphatidic acid and phosphatidylserine (66). Phosphatidylserine is an “eat me” signal on the cell surface of apoptotic cells for clearance by macrophages through a mechanism known as efferocytosis. This mechanism is very effective in killing M. tuberculosis that resides within apoptotic cells (67). EspB could therefore block efferocytosis through disrupting phosphatidylserine signaling, although this idea awaits confirmation (66). In addition, EspB can inhibit autophagosome formation induced by INF-γ in murine macrophages (68). This result is consistent with the phosphatidic acid binding ability of EspB, because phosphatidic acid is an important intracellular signal for autophagy (69). Taken together, EspB can affect membrane-mediated innate immune mechanisms through binding of phosphatidic acid and phosphatidylserine. Two recent structural studies suggest that exported EspB forms an oligomeric structure with a central pore (70, 71). Since EspB can interact with membrane, it is very possible that the oligomeric EspB can form a membrane pore that facilitates phagosome permeabilization within infected macrophages.

INTERVENTIONS BY TARGET ESX-1 SECRETION

Since M. tuberculosis mutants carrying a defect in ESX-1 lose the ability to cause diseases in multiple animal models, drugs that target ESX-1 may lead to potential antituberculosis drugs. Several groups have performed compound library screens and have succeeded in identifying several ESX-1 inhibitors (72). Using a M. tuberculosis strain carrying a fluorescence reporter under an acid-inducible PhoPR promoter, Johnson et al. screen for inhibitors that block the PhoPR regulation (73). They identify ethozolamide as an ESX-1 inhibitor. Interestingly, the compound inhibits M. tuberculosis carbonic anhydrase activity, which has not previously been linked to PhoPR signaling (73).

Rybniker et al. screen for inhibitors that prevent fibroblast killing by M. tuberculosis (72). Their identified inhibitors were able to block ESAT-6 secretion, avert the block in phagosome maturation, reduce bacterial loads, and prevent M. tuberculosis-induced lysis. All these effects are consistent with the known roles for ESX-1 during macrophage interactions. Additionally, the screen identifies an inhibitor that targets the histidine kinase MprB and another one that affects M. tuberculosis metal-ion homeostasis. This observation has led the authors to uncover zinc stress as a previously unrecognized activating signal for ESX-1 (72). Just as in the previous study, Rybniker’s study identifies inhibitors that target bacterial activities that have not previously been linked to regulating ESX-1.

SUMMARY

In this article, we have described several cellular pathological effects caused by the M. tuberculosis ESX-1. The effects include induction of necrosis, NOD2 signaling, type I interferon production, and autophagy. We then attempted to suggest that these pathological effects are mediated by the cytosolic access of M. tuberculosis-derived materials as a result of the phagosome-disrupting activity of the major ESX-1 substrate ESAT-6. Such activity of ESAT-6 is most likely due to its pore-forming activity at the membrane. The amyloidogenic characteristic of ESAT-6, first observed by Hsu et al., is reviewed here as a potential mechanism of membrane pore formation. In addition to ESAT-6, the ESX-1 substrate EspB interferes with membrane-mediated innate immune mechanisms such as efferocytosis and autophagy, most likely through its ability to bind phospholipids. Overall, the M. tuberculosis ESX-1 secretion system appears to be a specialized system for the deployment of host membrane-target proteins, whose primary function is to interrupt key steps in innate immune mechanisms against pathogens. Inhibitors that block the ESX-1 system or block host factors critical for ESX-1 toxicity have been identified and should represent attractive potential new antituberculosis drugs.