Protein kinase G confers survival advantage to Mycobacterium tuberculosis during latency-like conditions

Protein kinase G (PknG), a thioredoxin-fold–containing eukaryotic-like serine/threonine protein kinase, is a virulence factor in Mycobacterium tuberculosis, required for inhibition of phagolysosomal fusion. Here, we unraveled novel functional facets of PknG during latency-like conditions. We found that PknG mediates persistence under stressful conditions like hypoxia and abets drug tolerance. PknG mutant displayed minimal growth in nutrient-limited conditions, suggesting its role in modulating cellular metabolism. Intracellular metabolic profiling revealed that PknG is necessary for efficient metabolic adaptation during hypoxia. Notably, the PknG mutant exhibited a reductive shift in mycothiol redox potential and compromised stress response. Exposure to antibiotics and hypoxic environment resulted in higher oxidative shift in mycothiol redox potential of PknG mutant compared with the wild type. Persistence during latency-like conditions required kinase activity and thioredoxin motifs of PknG and is mediated through phosphorylation of a central metabolic regulator GarA. Finally, using a guinea pig model of infection, we assessed the in vivo role of PknG in manifestation of disease pathology and established a role for PknG in the formation of stable granuloma, hallmark structures of latent tuberculosis. Taken together, PknG-mediated GarA phosphorylation is important for maintenance of both mycobacterial physiology and redox poise, an axis that is dispensable for survival under normoxic conditions but is critical for non-replicating persistence of mycobacteria. In conclusion, we propose that PknG probably acts as a modulator of latency-associated signals.

Tuberculosis (TB) 5 although curable, is the leading cause of death from bacterial diseases. In response to Mycobacterium tuberculosis infection, host mounts a robust immune response, which contains the bacterium in ϳ90% of the cases, leading to latent infection. During latency, bacilli become non-responsive to nearly all known anti-TB antibiotics (1), serving as the major reservoir of the pathogen. Granuloma formation represents the tissue hallmark of the latent mycobacterium. This bunch of immune aggregates, previously thought to be beneficial for host by curtailing bacteria in a nutritionally starved, hypoxic environment, is now being reexamined as a niche for survival and dissemination during latency and reactivation (2). M. tuberculosis is a facultative aerobe, which can survive during hypoxia only if oxygen is gradually depleted (3,4). However, sudden anaerobiosis is lethal for the bacilli (5), highlighting the importance of sensory mechanisms necessary for the adaptation to a hypoxic environment. M. tuberculosis down-regulates transcription of key complexes of the electron transport chain in response to hypoxia, and ATP levels are reported to be approximately 5 times lower as compared with the replicating counterparts (6). Despite being metabolically slow, de novo ATP production, NAD ϩ /NADH balance, and maintenance of energized membrane is crucial for the persistence of hypoxic mycobacteria (6).
The extracellular cues during bouts of active disease, dormancy, and resuscitation probably contribute to M. tuberculosis adaptation during various stages of infection. DosS and DosT are redox and hypoxia sensor kinases, respectively, which are up-regulated upon hypoxia and nitric oxide exposure and in turn up-regulate ϳ50 genes through response regulator DosR (7)(8)(9). Eukaryotic-like serine/threonine protein kinases (STPKs) are regulatory elements associated with signal recognition and adaptive responses. M. tuberculosis encodes 11 STPKs, which are involved in a myriad of functions, such as cell division, virulence, transport of metabolites, and regulation of metabolic processes (10,11). Protein kinase G (PknG), the sole soluble STPK in M. tuberculosis, has three distinct domains: an N-terminal rubredoxin domain containing two thioredoxin motifs, a central kinase domain, and a C-terminal tetratricopeptide repeat motif (12). PknG was shown to be secreted into host cytosol during Mycobacterium bovis BCG infection, wherein it inhibits phagolysosomal fusion by an unknown mechanism (13). However, recent reports suggests a more definitive role of PknG in regulating bacterial metabolism (14). This is supported by its ubiquitous presence across pathogenic and non-pathogenic species. PknG-mediated in vitro phosphorylation of GarA relieves its interaction with ␣-ketoglutarate decarboxylase and glutamate dehydrogenase (15). Recently, amino acid availability was shown to act as a stimulus for PknG-mediated GarA phosphorylation (14). Furthermore, upon oxidative stress, PknGmediated phosphorylation of L13 promotes its interaction with RenU, which in turn regulates the NADH levels (16). Disruption of PknG enhances the susceptibility of mycobacteria toward antibiotics (17), attenuates growth during stationary phase (18), and impairs the biofilm formation (16), suggesting a plausible role in latency. However, to date, the role of PknG in latency and maintenance of redox poise has not been elucidated. In this study, we demonstrate the role of PknG in mycobacterial survival through in vitro models of latency like hypoxia, persisters, and nutrition starvation. Real-time redox measurements demonstrated an enhanced oxidative shift in the presence of antibiotics and hypoxia, resulting in the lower survival of the pknG mutant. The absence of PknG not only limits utilization of host-derived carbon sources but also results in lower metabolic activity compared with the wild type. Our study demonstrates that by regulating the TCA cycle through GarA phosphorylation and maintenance of the precarious redox status of bacilli, PknG augments the survival of mycobacteria during latency-like conditions. Moreover, we establish that PknG also plays a critical role in the formation of stable granulomas in guinea pigs, repositories of dormant mycobacteria.

PknG subverts host-induced stresses
In vivo survival of the pknG mutant in mice is more attenuated during the early phase of infection (18), indicating a role in subverting host-inflicted stresses, such as reactive oxygen species (ROS), reactive nitrogen intermediates, and low pH. Thioredoxin motifs, present at the N terminus of PknG, are known to protect mycobacteria against oxidative defense of host innate cells. To elucidate the role of PknG in combating host-imposed defenses, mycobacterial strains were exposed to various stresses in vitro. Survival of Rv⌬G was 100-and 10-fold lower as compared with Rv and Rv⌬G::G in the presence of ROS generators like cumene hydroperoxide (CHP) and menadione, respectively ( Fig. 1, a and b). Deletion of PknG also leads to ϳ10-fold higher susceptibility of mycobacteria in acidic media (Fig. 1c). Similarly, pknG mutant had reduced survival in the presence of 0.1% SDS (Fig. 1d). The lack of restoration in the complemented strain might be the result of overexpression of Figure 1. PknG plays a crucial role in subverting host-induced stresses. a, 50 M CHP was added to log-phase cultures of Rv, Rv⌬G, and Rv⌬G::G, and survival was monitored after 24 h. b, early log-phase cultures of Rv, Rv⌬G, and Rv⌬G::G were spotted on OADC-7H11 containing 50 M menadione. cfu obtained in OADC-7H11 plates without menadione was normalized to 100%, and survival on menadione-containing plates was calculated with respect to 100%. c, to assay the sensitivity of M. tuberculosis strains to acidic stress, single cell suspensions of bacterial strains were inoculated in sodium citrate buffer (pH 4.5), and cfu were enumerated on days 0 and 7. d, for surfactant stress, mid-log phase cultures were grown in 7H9 broth containing 0.1% SDS for 3 h. The y axis (percent survival) represents survival of bacterial strains in the presence of SDS/absence of SDS ϫ 100. The data depict mean Ϯ S.D. (error bars) of two independent experiments performed in triplicates (n ϭ 6). **, p Ͻ 0.005; ***, p Ͻ 0.0005.

Role of PknG during latency-like conditions
PknG (Fig. 1d). Together, these results suggest that PknG plays a critical role in surmounting a plethora of insults imposed by the host.

PknG plays an essential role in survival of mycobacteria during hypoxia
The success of M. tuberculosis as a pathogen relies on its immense ability to persist inside hypoxic granulomas for decades by undergoing transcriptional and translational remodeling of redox-responsive genes. Expression of pknG has been shown to be up-regulated during hypoxia in both Mycobacterium smegmatis (19) and M. tuberculosis (20). Moreover, PknG has been shown to be important for biofilm formation (16), whose core is hypoxic (21). To comprehensively understand the role of PknG in hypoxia, we examined the survival of the pknG mutant in a modified Wayne's model (22). Deletion of PknG was resulted in ϳ100-fold lower survival at 40 days posthypoxia in M. tuberculosis (Fig. 2a). Expression blots demonstrated that PknG levels reduced significantly at day 20 of hypoxia with a sharp increase at day 40 ( Fig. 2b). To elucidate the functional conservation across pathogenic (M. tuberculosis) and non-pathogenic (M. smegmatis) mycobacteria with respect to the hypoxia phenotype, we generated a pknG deletion mutant in M. smegmatis (Fig. 2, c-e). Similar results were obtained with M. smegmatis strains, wherein Msm⌬G mutant showed significantly attenuated survival during hypoxia in comparison with M. smegmatis and Msm⌬G::G (complementation strain) (Fig. 2f).

PknG plays a role in phenotypic drug tolerance
Anaerobic cells have been shown to be refractory to antibiotics (23). Such phenotypically drug-tolerant cells are called persisters and are also used as an in vitro model of latency (24). To examine the contribution of PknG in the survival of persisters, we first determined the minimum inhibitory concentration (MIC) of isoniazid and found it to be similar for both wild-type and pknG mutant strains (data not shown). Survival of persisters was scored in the presence of isoniazid (50ϫ MIC) for M. tuberculosis and M. smegmatis and was found to be drastically lower (ϳ10 -30-fold) for pknG mutant (Fig. 3, a and b) in comparison with the parental and complementation strains. Based on our observations that PknG promotes mycobacterial survival during hypoxia and abets drug tolerance, we suggest that PknG ameliorates survival of mycobacteria during latencylike conditions. Dormancy is tenuously linked with nutrient starvation inside granuloma of infected animals (25). Thus, we examined the role of PknG in the survival of M. tuberculosis during its growth in nutrient-rich and limiting media. Whereas pknG mutant in M. tuberculosis had no survival defect in 7H9ϩADC medium, its growth was severely compromised in nutrient-limited Dubos medium, which could be restored upon in trans complementation (Fig. 3c). Whereas supplementation with glucose partially restored the growth of the pknG mutant, the addition of acetate, succinate, or fumarate failed to rescue the growth defect (Fig. 3c). These results suggest that the levels of multiple metabolites of the TCA cycle may be lower in the pknG mutant.

PknG aids in metabolic adaptation
M. tuberculosis adapts to oxygen limitation and nutrient depletion by remodeling its TCA cycle (26 -28). PknG has been shown to in vitro phosphorylate TCA cycle enzyme, malate dehydrogenase (29), and central carbon metabolic regulator GarA (15); thus, it may directly or indirectly regulate the TCA cycle. Therefore, we examined the metabolic profile of the central carbon metabolites in M. smegmatis and Msm⌬G in normoxia and hypoxia ( Fig. 4 and supplemental Fig. S1). Interestingly, the levels of most of the glycolysis and TCA cycle metabolites were significantly lower in Msm⌬G compared with M. smegmatis during normoxia, suggesting that mycobacteria are metabolically not as active in the absence of PknG (Fig. 4). In corroboration with the previous studies (28,30), we observed that the wild-type mycobacterial strain effectively adapts to micro-aerophilic conditions by significantly waning its central carbon metabolism (ϳ3-4-fold; Fig. 4 and supplemental Table S1). Interestingly, the decrease in metabolite levels in hypoxic conditions was not as perceptible in the Msm⌬G mutant (ϳ2 fold; Fig. 4 and supplemental Table S1). Taken together, we propose that PknG aids in efficient metabolic remodeling during hypoxia, and its absence results in inept metabolic adaptation, eventually compromising the survival.

PknG is crucial for maintaining intramycobacterial redox homeostasis
PknG expression is induced by NADH, a key player in metabolism and redox homeostasis (16). Thioredoxin motifs of PknG render its kinase activity sensitive to variations in the redox state of the microenvironment (31). We hypothesized that lower survival of pknG mutant under latency-like conditions is probably due to its inability to maintain intramycobacterial redox balance. To test our hypothesis, we measured the precise perturbations in the mycothiol redox potential (E MSH ) upon deletion of PknG, using a highly sensitive, noninvasive redox biosensor specific to mycobacteria (Mrx1-roGFP2) (32). An increase in the fluorescence excitation ratio at 405/488 nm indicates oxidative stress, whereas a decrease corresponds to reductive stress (32). We found that the intracellular redox state of bacteria was significantly shifted toward reductive E MSH in Msm⌬G and Rv⌬G as compared with the parental strains (Fig.  5a). In trans complementation of PknG restored the redox state in Msm⌬G::G (Fig. 5a). In addition, we examined the antioxidant capacity with increasing dose of H 2 O 2 and found that deletion of PknG results in oxidized E MSH in M. smegmatis (Fig. 5b) and M. tuberculosis (data not shown). Analysis of redox status after low-dose H 2 O 2 exposure revealed that, whereas M. smegmatis was able to restore its resting redox state within ϳ15 min, the ratiometric -fold change was higher in Msm⌬G, and it failed to return to its ambient redox state even after 25 min (Fig. 5c). These results suggest that enhanced sensitivity of the pknG mutant to oxidative stress ( Fig. 1, a and b) is due to its inability to maintain redox poise. Furthermore, exposure to isoniazid resulted in enhanced oxidative stress in Rv⌬G (Fig. 5d) and Msm⌬G (Fig. 5e). This phenomenon was not isoniazid-specific, as exposure to other anti-mycobacterial antibiotics like rifam-

Role of PknG during latency-like conditions
picin and clofazimine also lead to 3-4-fold higher oxidized environment in Msm⌬G, as early as 6 h after antibiotic treatment (Fig. 5, f and g). Most importantly, even under hypoxia, Msm⌬G was ineffective in maintaining redox homeostasis (Fig.  5h), which corroborated with its reduced survival (Fig. 2, a and f). Together, these results demonstrate that PknG is crucial for

Active kinase and thioredoxin domain of PknG aids in survival of latent mycobacteria
PknG harbors three domains: an N-terminal rubredoxin domain with two CXXC thioredoxin motifs, a central kinase domain, and a C-terminal tetratricopeptide domain (Fig. 6a). The kinase activity of PknG is essential for its established role in virulence and metabolism (13,31). The presence of thioredoxin motifs is necessary for modulation of its kinase activity in response to changes in redox environment (31). Thus, we investigated the contributions of kinase activity and thioredoxin domain in abetting survival of mycobacteria during latencylike conditions. The Msm⌬G mutant was complemented with either wild-type (Msm⌬G::G) or kinase-inactive PknG (Msm⌬G::G K ) or thioredoxin mutant, wherein cysteine residues were mutated to alanines (Msm⌬G:: TT ) and checked for survival in latency models (Fig. 6b). In agreement with our previous results (Fig. 2, a and f), deletion of PknG led to reduced survival during hypoxia (Fig. 6b). Interestingly, whereas in trans expression of PknG was able to rescue the survival defect, complementation with kinase-inactive PknG or thioredoxin mutant failed to do so, indicating the importance of kinase activity as well as the functional thioredoxin motifs of PknG (Fig. 6b). Sim-ilar observations were found with the persister model of latency (Fig. 6c).

PknG ameliorates survival of latent mycobacteria through GarA
PknG and PknB have been shown previously to in vitro phosphorylate GarA at residues Thr-21 and Thr-22 (15,33), respectively (Fig. 7a). GarA phosphorylation at Thr-21 alleviates the inhibition of ␣-ketoglutarate decarboxylase and glutamate dehydrogenase, leading to derepression of the TCA cycle (15). Amino acids such as glutamate and aspartate were recently shown to be stimuli for this phosphorylation (14). As NADH, an inducer of PknG expression (16), acts at the crossroad of the TCA cycle and redox homeostasis, we hypothesized that PknG may regulate redox balance through GarA. We performed metabolic labeling to examine whether GarA is indeed phosphorylated in vivo. Results suggest that, although GarA is in vivo phosphorylated on both the Thr-21 and Thr-22 residues, Thr-21 appears to be the major site (Fig. 7b). This result is in agreement with a recent report by O'Hare's group (14). Similarly, immunoblotting with anti-pThr antibody indicated that Thr-21 and Thr-22 residues are in vivo phosphorylated in M. tuberculosis (Fig. 7c). Further, to study the role of GarA during latency-like conditions, we generated the GarA mutant in M. smegmatis (Msm⌬garA) (Fig. 7, d-f). The recombination at

Role of PknG during latency-like conditions
the native locus was confirmed with the help of PCR and Western blotting (Fig. 7, e and f). As reported earlier (34), deletion of GarA did not impact growth and cellular morphology in 7H9 medium but led to severely compromised growth in nutrientlimited Sauton's medium (data not shown). Next, we electroporated Msm⌬garA with wild-type and mutant GarA constructs and probed for threonine phosphorylation (Fig. 7g). Results showed that GarA is phosphorylated at both Thr-21 and Thr-22 residues, and the mutation of both residues (T21A/ T22A) resulted in complete abrogation of phosphorylation (Fig. 7g).
To elucidate the biological implications of PknG-mediated phosphorylation of GarA, we mutated the Thr-21 residue to phosphoablative (GarA-T21A) and phosphomimetic (GarA-T21E) forms and electroporated these constructs into Msm⌬G and Msm⌬garA. Western blot analysis confirmed the expression of wild-type and mutant GarA-His proteins (Fig. 8a). Importantly, Msm⌬G electroporated with wild-type GarA and GarA-T21A failed to alleviate the hypoxia phenotype, but GarA-T21E successfully restored the survival (Fig. 8b). We also observed that Msm⌬garA and Msm⌬garA::GarA (overexpression strain) showed reduced survival during hypoxia, suggesting that steady-state levels of GarA are critical for mycobacterial survival under stress (Fig. 8b). Interestingly, expression of GarA during hypoxia correlated with the expression pattern of PknG (Fig. 2b). Although in trans complementation of Msm⌬garA with GarA-T21A could not rescue the survival phenotype, complementation with GarA-T21E, reminiscent of phosphorylation by PknG, could functionally complement the sur-vival defects (Fig. 8b). These data were further strengthened by the persister assay (Fig. 8c). Taken together, PknG-mediated phosphorylation of GarA is important to maintain redox homeostasis, resulting in survival of mycobacteria during latency.

PknG spurs stable granuloma formation in guinea pigs
In vitro dormancy models, such as multistress, hypoxia, nutrient starvation, and drug tolerance, are based on hostile conditions in granuloma. PknG mutant has been shown to exhibit reduced survival in mice (18). However, a major deficit of murine TB model is lack of hypoxia and caseation necrosis in lesions, which not only represent the pathological hallmark of human TB granulomas but also serve as repositories of dormant mycobacteria (35). Aerosol infection of M. tuberculosis in guinea pigs creates a diseased state, which morphologically and clinically resembles human TB (36). To delineate the in vivo function of PknG, guinea pigs were challenged with Rv, Rv⌬G, or Rv⌬G::G. Bacterial depositions inside lungs determined at day 1 revealed that Rv⌬G had ϳ40 -60% lower implantation as compared with Rv and Rv⌬G::G (Fig. 9a). This variation was not due to a difference in inoculum, as aliquots of in vitro culture plated before infection gave similar cfu (data not shown). The bacillary load in the lungs of a guinea pig infected with Rv⌬G was lower as compared with Rv and Rv⌬G::G after 4 and 8 weeks (Fig. 9b). Furthermore, dissemination of Rv⌬G to spleen was also attenuated (Fig. 9, c and d). Taken together, PknG is important for bacterial survival inside lungs of guinea pigs.
The ability of M. tuberculosis to cause classical TB does not solely correlate with its propensity to grow and survive inside

Role of PknG during latency-like conditions
infected animals, but is also due to its ability to cause pathogenesis and disease. Hence, we assessed the impact of PknG disruption on the immunopathology of infected guinea pigs. Gross evaluation of lungs after 4 and 8 weeks revealed discrete tubercles distributed throughout the tissue of guinea pigs infected with Rv and Rv⌬G::G. However, lungs of animals infected with Rv⌬G had markedly reduced inflammation and an absence of tubercles (data not shown). Histopathological observations also revealed numerous granulomas with central necrosis in the lung tissues of Rv-infected animals. In contrast, guinea pigs infected with Rv⌬G exhibited significantly reduced pathology with normal lung parenchyma and clear alveolar spaces (Fig. 9, e and f). Importantly, whereas five Rv infected animals showed granulomas in the lung sections, only one Rv⌬G-infected lung section showed a small granuloma with no necrosis (Fig. 9 (e and f) and Table 1). Similarly, most of the spleen sections of Rv-infected animals (Fig. 9g) revealed the presence of granulomas, whereas spleen sections of animals infected with Rv⌬G had intact white pulp without any granulomas. Taken together, these data suggest that in guinea pigs, the primary function of PknG is maintenance of inflammatory tubercules, histological correlates of dormancy.

Discussion
Eradication of tuberculosis has been a challenge due to the vast reservoirs of latent carriers. To survive in hostile conditions during latency, bacteria readjust their redox regulatory circuits and metabolism, especially the TCA cycle (37). Expression of pknG has been reported to be up-regulated under acidnitrosative multistress conditions (38) and in host macrophages (39). The first report on the role of PknG as a virulence factor inside the host was established by the group of J. Pieters (13). However, most of the subsequent reports focused on the role of PknG as a modulator of bacterial metabolism (15,18). In this study, we found that deletion of PknG results in significantly reduced mycobacterial survival in in vitro latency-like conditions, such as Wayne's model of hypoxia, persisters, and nutrient starvation (Figs. 2 and 3). Results suggest that the ability of M. tuberculosis to withstand oxidative and acidic stress was compromised in the absence of PknG (Fig. 1). Disruption of PknG also led to lower persister survival (Fig. 3), potentially suggesting PknG to be a target for adjunct therapy to eliminate latent TB.
The mycothiol (MSH) levels have been shown to be higher in the pknG mutant due to altered glutamine metabolism (18). Redox measurements indicated that the pknG mutant displays more reductive E MSH as compared with the parental strain (Fig.  5), which could be due to higher mycothiol. Despite higher levels of MSH and reductive E MSH , the PknG mutant has lower capacity to maintain MSH redox balance upon oxidative challenge (Fig. 5). This may be due to inefficient reduction of MSSM (oxidized form of MSH) by mycothione disulfide reductase (Mtr). There may also be reduced levels of NADPH in the

Role of PknG during latency-like conditions
mutant, which serves as an electron donor in Mtr-mediated reduction of MSSM. Furthermore, in the presence of antibiotics, mycobacteria were unable to maintain their redox equilibrium in the absence of PknG (Fig. 5). Stimulation of hydroxyl radicals has been proposed as a means to manage persisters (40). Enhanced oxidative shift in the presence of antibiotics and hypoxia probably contributes to the lower survival of the pknG mutant. The intracellular redox environment of mycobacteria upon antibiotic treatment is more oxidized compared with untreated cells (32,41). However, to date, the intracellular redox environment of cells during hypoxia has not been investigated. We found that hypoxic cells undergo oxidative shift (Fig. 5), which was also confirmed with the help of CellROX dye (data not shown). Notably, the ratiometric response of Mrx1-roGFP2 in the pknG mutant subjected to hypoxia is close to the maximum oxidation limit of the biosensor (Fig. 5) to be detected by this biosensor, signifying importance of PknG as a redox rheostat.
In addition to rewiring of the redox environment, the other well-known factor contributing to induction of non-replicating persisters is nutrient starvation, analogous to the environment of granuloma and phagosomes. Persistent bacteria isolated from lung lesions have similar staining properties as nutrientstarved bacteria (25). The absence of PknG not only limits hostderived carbon source utilization but also results in lower metabolic activity compared with the wild type (Fig. 4). Previously, glutamine levels have been reported to be higher in the M. tuberculosis PknG mutant (18) but not in the M. bovis PknG mutant (42). We did not observe increased glutamine levels in Msm⌬G, which could be due to a difference in species, as suggested earlier (42) or different experimental conditions, such as growth media. We also observed enhanced glucose levels in the Msm⌬G mutant under anaerobic conditions, which could probably be due to its lower utilization or direct/indirect regulation of glucose transporters during hypoxia by PknG, as suggested earlier for PknF (43). The ability of mycobacteria to adapt to hypoxia by waning its cellular metabolism (37), is compromised upon deletion of PknG ( Fig. 4 and supplemental Table S1), which may be responsible for compromised survival. Taken together, PknG seems to regulate metabolic adaptations of central carbon metabolism in response to changing environments.
Reversible phosphorylation is regarded as one of the fastest ways to transcribe changes in environmental signals. Both hypoxic and antibiotic-induced non-replicating persistence result in metabolic perturbations, especially through modulation of enzymes of the TCA cycle. Phosphorylation of GarA by PknG abrogates its interaction with the TCA cycle enzymes glutamate dehydrogenase and ␣-ketoglutarate decarboxylase, thus alleviating GarA-mediated inhibition (15). The trigger for this phosphorylation is amino acid availability (14). We demonstrate that GarA is phosphorylated in vivo on residues Thr-21 and Thr-22 (Fig. 7). It is important to note that most of our findings were obtained both with M. tuberculosis and M. smegmatis strains, highlighting functional conversation of PknG across pathogenic and non-pathogenic mycobacterial strains. Because GarA is essential in M. tuberculosis, and phosphorylated GarA is found both in M. smegmatis and M. tuberculosis (14), we used Msm⌬garA mutant and complemented strains to elucidate the mechanism. Whereas phosphoablative GarA-T21A mutant could not rescue the survival defect of the pknG mutant, the phosphomimetic mutant, GarA-T21E, could (Fig. 8). Together, these results suggest that the survival of mycobacteria in latency-like conditions occurs through PknG-

Role of PknG during latency-like conditions
mediated phosphorylation of GarA on residue Thr-21. Attenuation of the pknG mutant during latency-like conditions might ensue from loss of multiple functions. For example, anaerobic cultures have been reported to be unusually acidified due to higher amounts of secreted acids (28). As depicted in Fig. 1, PknG helps the bacterium to survive during acidic conditions. Moreover, PknG broadens the utilization of various carbon sources (Fig. 3). Based on our data, we propose that PknG most likely senses host-associated signals, such as oxidative stress, low pH, and nutrient deprivation. Redox changes during persistence, nutrition starvation, and hypoxia result in an oxidized environment, which may act as a trigger for PknG activation (Fig 10). This hypothesis is in line with our previous observations, which show enhanced activity of PknG in an oxidized environment (31).
Last, we performed animal infection studies using virulent M. tuberculosis strains to determine the role of PknG in vivo.
Similar to humans, guinea pig granulomas are caseating and hypoxic (35). To investigate the role of PknG in bacillary survival and pathology in a pertinent animal model, we choose guinea pigs. Infection with the pknG mutant resulted in lower deposition and hence may have resulted in reduced early survival. Attenuated survival during the early phase of infection might have also emanated from host-inflicted stresses, such as ROS, reactive nitrogen intermediates, burst, and low pH of phagosomes. Importantly, we found severe reduction in granuloma formation in guinea pigs infected with the pknG mutant (Fig. 9). M. tuberculosis promotes granuloma formation for better survival and dissemination (2). PknG is thought to be a secretory protein and hence may foster active recruitment of immune cells for granuloma formation to enhance long-term mycobacterial survival. Furthermore, because granuloma formation requires constant chronic stimuli, it is tempting to speculate that in the absence of , and pVV16-GarA-T22A (Msm::GarA-T22A) was metabolically labeled as described previously (50). C-terminal hexahistidine (His)-tagged GarA and GarA mutant proteins were pulled down using Ni 2ϩ -nitrilotriacetic acid affinity beads, resolved on SDS-PAGE, transferred to nitrocellulose membrane, and autoradiographed (top). The membrane was later probed with anti-GarA antibodies (bottom). c, GarA or GarA mutants from 1 mg of whole-cell lysate (WCLs) of Rv electroporated with pVV16-GarA (Rv::GarA), pVV16-GarA-T21A (Rv::GarA-T21A), and pVV16-GarA-T22A (Rv::GarA-T22A) were pulled down using Ni 2ϩ -nitrilotriacetic acid affinity beads. Four-fifths of pulldown was probed with anti-pThr antibody (top), and the remaining one-fifth was probed with anti-GarA antibody (bottom). d, schematic outline of the methodology used for the disruption of garA in M. smegmatis. The primers used for PCR amplification are represented by arrows. e, agarose gel showing PCR amplifications using F1-R1 and F2-R2 primers as depicted in d. A PCR amplicon of 1.1 kb with the F1-R1 pair is expected only in M. smegmatis, whereas a 1.2-kb product is expected with the F2-R2 pair only in the Msm⌬garA. f, to confirm GarA knock-out, 25 g of WCLs prepared from M. smegmatis, Msm⌬garA, and Msm⌬garA::GarA were resolved on 12% SDS-PAGE, transferred onto nitrocellulose membrane, and probed with anti-GarA and anti-GroEL1 antibodies. g, His-tagged GarA or GarA mutants were pulled down from 1 mg of WCLs prepared from different strains. Four-fifths of the pulldown was probed with anti-pThr antibody, and the remaining one-fifth was probed with anti-GarA antibody.

Role of PknG during latency-like conditions
PknG, mycobacteria are unable to survive in the hostile environment of granuloma.
Based on our results, we propose that PknG acts as a "stress regulator" by combating different cellular and redox stress experienced by mycobacteria, through phosphorylation of substrates, such as GarA (15) or L13 (16) (Fig 10). Lack of phosphorylation of substrates in the absence of PknG results in redox imbalance, leading to compromised survival under multiple stress conditions, including latency (Fig 10). In line with this notion, we observed that whereas the wild-type PknG can complement survival defects of the PknG mutant during latency-like conditions, the kinase-inactive mutant and thioredoxin domain mutant fail to do so (Fig. 6). Because dormant bacteria inside granuloma are inaccessible to drugs, we propose PknG to be a unique target for eradicating latent TB. Using a PknG inhibitor for adjunct therapy would probably result in lower granuloma formation, lower persister, and antibiotic-resistant cells. Moreover, it would be interesting to test the pknG mutant in the Cornell model of latency. In summary, our data establish that M. tuberculosis requires PknG for long-term survival during non-replicating persistence.

Bacterial strains, growth conditions, reagents, and plasmid constructs
Bacterial strains used in the study are listed in Table 2. Escherichia coli strain DH5␣ (Invitrogen) used for cloning was grown in LB broth or agar. Mycobacterial strains were grown in Middlebrook 7H9 broth supplemented with 10% ADS (albumin, dextrose, NaCl) or ADC (ADS, catalase) along with 0.2% glycerol (Sigma) and 0.05% Tween 80 (Sigma) or 7H11 agar supplemented with 10% OADC (oleic acid, ADC) and 0.2% glycerol (Sigma) at 37°C with shaking at 200 rpm. Hygromycin was used at 50 g/ml for mycobacteria and at 150 g/ml for E. coli. Kanamycin and chloramphenicol were used at 25 and 40 g/ml, respectively, for mycobacteria and at 50 and 34 g/ml, respectively, for E. coli. Restriction enzymes were purchased from New England Biolabs. Analytical grade chemicals, antibiotics, and oligonucleotide primers were procured from Sigma. Electron microscopy reagents were purchased from Electron Microscopy Sciences. Medium components were purchased from BD Biosciences. pVV16-PknG was generated by subclon-

Role of PknG during latency-like conditions
ing the NdeI-HindIII fragment from pQEII-PknG/PknG-K181M (31) into the corresponding sites in pVV16 vector. The mrx1-roGFP2 gene was PCR-amplified from the Mrx1-roGFP2 (32) construct with the help of specific primers, and the amplicon was cloned into pST-K (44) to generate the pST-K-Mrx1-roGFP2 construct.

Susceptibility of M. tuberculosis strains to in vitro stresses
To test the susceptibility to oxidative stress, mycobacterial strains were grown to mid-log phase (A 600 ϳ0.6 -0.8). Cultures were washed with PBS containing 0.05% Tween 80 (PBST) and inoculated at A 600 of 0.05 in 7H9 medium containing 0 or 50 M cumene hydroperoxide for 24 h, and serial dilutions were plated on 7H11 agar. For assessing the impact of superoxide stress, mid-log phase cultures of bacterial strains were diluted and plated on 7H11 agar containing 0 or 50 M menadione. For acidic stress, mid-log phase cultures were washed with PBST, and single cell suspension was inoculated into sodium citrate buffer (200 mM sodium phosphate, 100 mM citric acid, 0.02% tyloxapol, pH 4.5) at A 600 of 0.05.
Serial dilutions of cultures were plated on 7H11 agar after 7 days of incubation. The viability of strains was also enumerated in the presence and absence of 0.1% SDS (Sigma).

Hypoxia experiments
Viability of mycobacterial strains were assessed under hypoxic conditions as described earlier (45). Briefly, strains were inoculated at A 600 ϳ0.1 in ADS-supplemented 7H9 or Dubos medium containing 1.5 g/ml methylene blue (visual indicator of oxygen depletion) in glass culture tubes with a headspace of 15% at 37°C. At the indicated time points, bacteria were plated onto 7H11 plates for cfu enumeration.  (24). The depicted structure of PknG was adapted from Sherr et al. (12). Oxygen depletion in cells renders them drug-tolerant (23), indicated by the two-way arrow. Upon oxygen limitation, the NADH/NAD ϩ ratio increases (6), which has been shown to enhance PknG expression (16). Antibiotic exposure may similarly enhance expression or activity of PknG. Recently, amino acids such as glutamate and aspartate have been shown to induce PknG activity (14). In wild-type mycobacteria (Rv/Msm), PknG combats oxidative, acidic, nitrosative, and surfactant stress and hence may act as a "stress regulator" (this study). PknG-mediated phosphorylation of GarA at Thr-21 results in derepression of the TCA cycle (15). PknG, most importantly, regulates precarious redox homeostasis both in hypoxic and antibiotic-treated cells (this study). Redox balance may be aided by modulation of the TCA cycle (shown by a dotted arrow) or through the L13-RenU axis (16) previously shown to maintain NADH levels. Taken together, PknG helps mycobacteria remodel metabolism in response to latency-like conditions resulting in the survival of mycobacteria.

Metabolomics with NMR spectroscopy
M. smegmatis and Msm⌬G were cultured in liquid Dubos medium containing uniformly labeled [ 13 C]glucose ([U- 13 C] glucose; Isotec Sigma-Aldrich) for 16 h (aerobic) or 5 days (microaerobic, to avoid metabolic consequences ensuing from the death in pknG mutant). 2 mg of WCLs were lyophilized, resuspended in 99.9% D 2 O, and centrifuged to remove any debris. The supernatant was transferred to 3-mm NMR tubes. All NMR experiments were carried out on a Bruker AvanceIII spectrometer equipped with a 5-mm TCI cryogenic triple-resonance probe, operating at field strengths of 500 MHz. 2D 13 C, 1 H heteronuclear single-quantum coherence (HSQC) experiments were measured at 310 K. Temperature calibration was performed using a 100% [ 2 H]methanol sample (46). For 1D 1 H NMR spectra, a 1 H 90°pulse of 7.5 s was obtained. The spectrum was measured for a spectral width of 7002.8 Hz and total acquisition time of 2.34 s. A total of four dummy scans and 128 scans were used with a relaxation delay of 2 s. Data processing was performed using a 90°shifted sine-square bell window function with a zero filling of 65,536 points. The 2D 13 C, 1 H HSQC spectra were measured with spectral width of 7002.8 Hz along the 1 H dimension and 25,155.08 Hz along the 13 C dimension. A total of 16 dummy scans and 96 scans with a relaxation delay of 1.5 s were used for a total acquisition time of 145 ms (t 2max ) along the 1 H dimension and 2.5 ms (t 1max ) along the 13 C dimension. Processing was performed using a 90°shifted sinesquare bell window function for both of the dimensions, and zero filling of 4096 and 2048 was used in the 1 H and 13 C dimension, respectively. Topspin version 2.1 (Bruker AG) was used for acquisition, Fourier transformation, and processing of data. Referencing of all of the spectra was done using DSS. Batch volume integration of the peaks was performed and corre-sponding peak lists were generated using CARA (Computer Aided Resonance Assignment) software (47). Chemical shift values were assigned to specific metabolites using the Biological Magnetic Resonance Bank (BMRB) (http://www.bmrb.wisc. edu) and Human Metabolome Database (HMDB) (http://www. hmdb.ca). A chemical shift error tolerance of 0.05 and 0.5 ppm was used for 1 H and 13 C chemical shifts, respectively. The y axis represents normalized intensities of metabolites (Fig. 4). Individual intensities of metabolites were divided by total intensity of all of the metabolites of a sample to calculate normalized intensity. Mean and S.D. of multiple peaks of individual metabolites were added to represent mean intensity of each metabolite. Negative peak values were considered as noise and numerically zero.

Persisters
The MIC of isoniazid for different mycobacterial strains was determined by an alamarBlue assay, as described previously (32). For enumerating mycobacterial persisters, log phase cultures of bacterial strains were diluted to an A 600 of ϳ0.05 in 7H9-ADS containing 50 and 5 g/ml isoniazid for 48 h and 7 days for M. smegmatis and M. tuberculosis strains, respectively, and incubated at 37°C. At the indicated time points, cfu were enumerated.

Construction of pknG and garA gene replacement mutants in M. smegmatis
Approximately 1-kb regions upstream and downstream of pknG loci were PCR-amplified using specific primers. The amplicons were digested with PflMI and ligated with two compatible fragments (OriEϩcos and Hyg res ϩSacB) from pYUB1474 (48) to generate allelic exchange substrate (AES). PknG-AES was linearized with PacI and ligated with the temperature-sensitive PacI-digested phAE159 phasmid. The M. smegma- Table 2 List of strains used in the study

Strains Description
Antibiotic resistance Source tis pknG mutant (Msm⌬G) was generated as described previously using specialized transduction (48). Hygromycin-resistant colonies obtained were screened by PCR and Western blotting to confirm replacement of pknG at its genomic locus. The M. smegmatis garA mutant (Msm⌬garA) was generated as described above. Hygromycin-resistant colonies obtained were screened by PCR and Western blotting to confirm replacement of garA at its genomic locus. For complementation studies, garA was cloned into NdeI-HindIII sites of pVV16 vector, and the pVV16-GarA construct obtained was transformed into Msm⌬garA to generate the Msm⌬garA::GarA strain. Point mutants of GarA (T21A, T21E, T22A, and T2122A) were generated with the help of overlapping extension PCR mutagenesis with appropriate mutagenesis primers and confirmed by sequencing. The mutants GarA, GarA-T21A, GarA-T21E, GarA-T22A, and Gar-T21T22A were cloned into pVV16 vector and electroporated into Msm⌬garA to generate Msm⌬garA:: GarA-T21A, Msm⌬garA::GarA-T21E, Msm⌬garA::GarA-T22A, and Msm⌬garA::GarA-T2122A, respectively.

In vitro measurement of E MSH
The intracellular redox state of different mycobacterial strains was measured as described previously (32). Briefly, bacilli expressing Mrx1-roGFP2 bioprobe were treated with 10 mM N-ethylmaleimide (to block the redox state of roGFP2 (32)) for 5 min followed by fixation with 4% paraformaldehyde, and cells were analyzed by flow cytometry using FITC (495-nm excitation, 519-nm emission) and V500 (415-nm excitation, 500-nm emission) channels. The observed 405/488 ratios were normalized by treating bacilli with 10 mM CHP (100% oxidation; 405/488 ratio set to 1) and 20 mM DTT (100% reduction; 405/488 ratio set to 0.1). To study the effect of antibiotics on redox stress, MIC values for isoniazid for wild-type and mutant strains were determined, as described previously (32). Exponentially growing M. smegmatis and Msm⌬G cells were treated with 50 g/ml isoniazid, 10 g/ml rifampicin, and 10 g/ml clofazimine for 6 h, followed by redox measurements. Similarly, redox status of M. tuberculosis strains treated with 5 g/ml isoniazid was determined. To elucidate the effect of hypoxia on redox status, M. smegmatis strains were incubated in sealed long tubes for 5 days (to avoid redox consequences ensuing from death), and biosensor response was analyzed.

Growth rate and nutrient utilization
To analyze the growth patterns of M. smegmatis, Msm⌬garA, and Msm⌬garA::GarA, each strain was grown in 7H9 to A 600 of 0.8. The cultures were washed, inoculated at A 600 of 0.025 in the required medium at 37°C, 200 rpm, and A 600 was measured every 3 h. To investigate role of GarA in differential carbon source utilization, fresh cultures were seeded at an initial A 600 of 0.05 in salt-based minimal Sauton's medium containing 10 mM NH 4 Cl and 0.05% tyloxapol with either glycerol (1%) or sodium succinate (10 mM) as the sole carbon source, and growth was monitored every 6 h. To evaluate the growth of Rv and Rv⌬G in 7H9 and Sauton's medium, cultures were seeded at A 600 of 0.1, and growth was monitored. For spot analysis on different carbon substrates, Rv and Rv⌬G were cultured in liquid Dubos medium until mid-log phase. Equal numbers of cells (3 ϫ 10 5 ) in equal volume were spotted on Dubos agar alone (basal) or supplemented with 50 mM glucose, 25 mM acetate, 50 mM succinate, or 50 mM fumarate, and growth was analyzed after 4 weeks.

M. tuberculosis infection of guinea pigs
Out-bred female guinea pigs of the Duncan-Hartley strain (weight 200 -300 g) were housed in individually ventilated cages at the animal house facility (Tuberculosis Aerosol Challenge Facility) of ICGEB, New Delhi, India. Guinea pigs (n ϭ 6/group/time point) were infected with M. tuberculosis wild type (Rv), pknG mutant strain (Rv⌬G), and complemented strain (Rv⌬G::G) by an aerosol route to implant nearly 100 cfu/ lung, as described previously (45). Deposition of the bacterium inside the lung was assessed by sacrificing two guinea pigs 24 h p.i. and plating the lung homogenates in triplicates on 7H11 agar plates. Bacillary loads in lung and spleen were evaluated at different time points (4 and 8 weeks) after aerosol infection to follow the course of infection. Organs of infected animals were harvested and fixed in 10% buffered formalin, and H&E staining was performed on 5-m-thick paraffin-embedded tissues, as described previously (49). Granulomas were categorized as necrotic, non-necrotic, or fibrotic and were given different pathological scores, as described earlier (49).

Ethics
Animal experiment protocols were reviewed and approved by the institutional animal ethics committee of the National Institute of Immunology, New Delhi, India (approval number IAEC 251/ 10). The experiments were carried out as per the guidelines issued by Committee for the Purpose of Control and Supervision of Experiments on Animals (CPCSEA), Government of India.

Statistical analysis
Student's t-test (two-tailed, unpaired) was used to analyze the significance of results, unless otherwise specified. Graph-Pad Prism version 5.0 was used for plotting the results and modified using Adobe Illustrator CS5.1. p values Ͻ0.05 were considered as statistically significant.