Depletion of M. tuberculosis GlmU from Infected Murine Lungs Effects the Clearance of the Pathogen

M. tuberculosis N-acetyl-glucosamine-1-phosphate uridyltransferase (GlmUMtb) is a bi-functional enzyme engaged in the synthesis of two metabolic intermediates N-acetylglucosamine-1-phosphate (GlcNAc-1-P) and UDP-GlcNAc, catalyzed by the C- and N-terminal domains respectively. UDP-GlcNAc is a key metabolite essential for the synthesis of peptidoglycan, disaccharide linker, arabinogalactan and mycothiols. While glmU Mtb was predicted to be an essential gene, till date the role of GlmUMtb in modulating the in vitro growth of Mtb or its role in survival of pathogen ex vivo / in vivo have not been deciphered. Here we present the results of a comprehensive study dissecting the role of GlmUMtb in arbitrating the survival of the pathogen both in vitro and in vivo. We find that absence of GlmUMtb leads to extensive perturbation of bacterial morphology and substantial reduction in cell wall thickness under normoxic as well as hypoxic conditions. Complementation studies show that the acetyl- and uridyl- transferase activities of GlmUMtb are independently essential for bacterial survival in vitro, and GlmUMtb is also found to be essential for mycobacterial survival in THP-1 cells as well as in guinea pigs. Depletion of GlmUMtb from infected murine lungs, four weeks post infection, led to significant reduction in the bacillary load. The administration of Oxa33, a novel oxazolidine derivative that specifically inhibits GlmUMtb, to infected mice resulted in significant decrease in the bacillary load. Thus our study establishes GlmUMtb as a strong candidate for intervention measures against established tuberculosis infections.


Introduction
The cell wall, which contains a number of virulence determinants, is the first line of defence for survival of the pathogen in the hostile host environment [1]. The mycobacterial cell envelope includes three layers of cell membrane and a cell wall made up of peptidoglycan, mycolic acid, arabinogalactan and lipoarabinomannan (LAM) [2][3][4]. Most existing first line and second line drugs used to treat TB such as isoniazid, ethambutol, ethionamide and cycloserine, act on enzymes engaged in the synthesis of different cell wall components [5]. The current high mortality rates of infected individuals as well as increasing incidence of multidrug-resistant (MDR) and extensively drug-resistant (XDR) tuberculosis (TB) among patients underscore the importance of finding new targets for therapeutic intervention.
GlmU Mtb is a bi-functional enzyme, with acetyltransferase and uridyltransferase activities catalyzed by the C-and N-terminal domains respectively ( Fig 1A) [6,7]. The carboxy-terminal domain of GlmU Mtb transfers the acetyl moiety from acetyl CoA onto glucosamine-1-phosphate to generate N-acetylglucosamine-1-phosphate (GlcNAc-1-P). The N-terminal uridyltransferase domain of GlmU Mtb then catalyzes the transfer of UMP (from UTP) to GlcNAc-1-P to form UDP-GlcNAc ( Fig 1A) [6]. The UDP-GlcNAc thus produced is among the central metabolites that is required for the synthesis of peptidoglycan, lipid A of LAM, arabinogalactan, Rha-GlcNAc linkers, mycothiol (required for maintaining redox homeostasis) [8][9][10][11][12][13][14]. The crystal structure of M. tuberculosis GlmU (GlmU Mtb ) displays two-domain architecture with an N-terminal α/βlike fold and a C-terminal left-handed parallel-β-helix structure [15,16]. Unlike its orthologs, GlmU Mtb has a long carboxy-terminal tail which displays little secondary structure [17]. Results from transposon mutagenesis experiments have indicated glmU Mtb to be an essential gene, supported by the fact that M. smegmatis is unable to grow in the absence of glmU smeg [18][19][20]. However, no studies have addressed the question of whether both the activities of GlmU Mtb are independently essential for the growth or survival of the bacterium.
While the enzymes required for the synthesis of UDP-GlcNAc are well conserved among prokaryotes, they are very different from those found in eukaryotes, making GlmU Mtb an attractive putative drug target [21,22]. Researchers have developed compounds that inhibit the activities of the orthologs of GlmU Mtb (GlmU from T. brucei, P. aeruginosa, E. coli, H. influenza and X. oryzae) in vitro [23][24][25][26][27][28][29][30]. Bioinformatic analyses and kinetic modelling data advocate GlmU Mtb to be a potential target for the development of suitable inhibitors [31]. In concurrence with these predictions, effective inhibitors have been developed against, the acetyltransferase and uridyltransferase domains of GlmU Mtb [32,33]. However, the precise sites of inhibitor-protein interactions and the efficacy of the inhibitors ex vivo or in vivo have not been investigated. Subjecting Mtb cultures in vitro to gradual decrease of oxygen (hypoxic stress) results in reprogramming of metabolic pathways and up-regulation of stress response genes, and is considered to be an in vitro model for the dormancy [34,35]. The importance of GlmU Mtb for growth under hypoxic conditions and in an in vivo infection model is yet to be investigated. In the present study we have generated a conditional gene replacement mutant of glmU Mtb and used this mutant to investigate any role GlmU Mtb may play in modulating the growth of the bacterium in vitro, ex vivo and in vivo. The data presented here demonstrate that GlmU Mtb is a viable and promising target for therapeutic intervention against tuberculosis.

Results
GlmU Mtb depletion perturbs cell wall structure and affects the bacterial survival in normoxia As the tetracycline-inducible system is an effective means to regulate gene expression [36], we introduced the integration-proficient pST-KirT-glmU construct (wherein glmU Mtb gene was cloned under a promoter that shuts down upon ATc addition; S1A Fig) into Mtb H37Rv ( Fig  1B). Whereas the expression of GlmU Mtb from its native locus remained unaltered, the expression of FLAG-GlmU Mtb in Rv::glmU strain was drastically compromised in the presence of ATc (Western blot inset, Fig 1B). This merodiploid strain was transduced with temperature sensitive phage, and the fidelity of homologous recombination at the native locus was confirmed by amplification across the replacement junctions using appropriate primers (Fig 1C). A comparison of GlmU Mtb expression in the presence and absence of ATc revealed that the protein was not detectable by western blot analysis after 6 days of growth in the presence of ATc (Fig 1D). While the growth of RvΔglmU in the absence of ATc was similar to Rv, in the presence of ATc the growth was drastically compromised (Fig 2A). A comparative analysis of growth by spotting of serially diluted cultures of Rv and RvΔglmU grown in the presence versus absence of ATc showed that GlmU Mtb depletion by addition of ATc led to complete inhibition of growth, with no growth detected after 6 days ( Fig 2B). Interestingly, analysis of GlmU Mtb expression every 24 hours post-ATc addition uncovered significant reduction in GlmU Mtb expression by the third day itself (Fig 2C). These results indicate that GlmU Mtb is required for the Mtb survival. To determine the impact of GlmU Mtb depletion on cellular morphology we carried out SEM and TEM imaging analysis of Rv and RvΔglmU cells grown for three days in the absence or presence of ATc. SEM analysis revealed severe morphological perturbations in the absence of GlmU Mtb , with the bacilli showing wrinkled surface and fused cells ( Fig 2D). TEM analysis showed that whereas in Rv and RvΔglmU cell wall structure and thickness are comparable, there was a marked decrease in cell wall thickness in RvΔglmU cells where GlmU Mtb was not expressed (cells grown in the presence of ATc; Fig 2E and 2F).

Impact of GlmU Mtb depletion on dormant bacteria
Next we used the Wayne model to investigate the consequence of GlmU Mtb depletion on the dormant bacteria under hypoxic conditions [35]. Accordingly, hypoxia was established and maintained for 42 days with depletion of GlmU Mtb or addition of INH for either 22 days, or for 2 days (Fig 3A, line diagram). In agreement with previous reports, we observed that bacteria were tolerant to INH under hypoxic conditions [37] (Fig 3C), with a thicker cell wall being observed under hypoxic conditions compared with the normoxic cultures (Fig 3D and 3E). Depletion of GlmU Mtb for 22 days resulted in complete clearance of growth (Fig 3B), which was also reflected in severe morphological perturbations and drastic reduction in cell wall thickness (Fig 3D and 3E). Significantly, GlmU Mtb depletion for as less as 2 days decreased cell RvΔglmU cultures grown to an A 600 of 0.8 were used to seed fresh cultures at an initial A 600 of 0.1 in the presence or absence of ATc as indicated and the day wise growth was monitored for six days. Inset shows the growth in the culture tubes after six days. The experiment was performed in triplicates (n = 3). Error bars indicate standard error of mean (s.e.m). (B) Day 0 and Day 6 cultures grown in the presence or absence of ATc were serially diluted and spotted on 7H11 agar plates. (C) Large scale cultures were inoculated at A 600 of 0.1 in the presence or absence of ATc and the WCLs were prepared on different days post ATc addition (as indicated). WCLs were resolved and probed with anti-GlmU Mtb and anti-GroEL1 antibodies. (D) Scanning electron microscopy of Rv and RvΔglmU grown for 72 h with or without ATc as indicated. Scale bars in upper panel were 10 nm and for the inset was 200 nm. The experiment was repeated thrice. (E) Transmission electron micrographs results at 50,000X of Rv and RvΔglmU cultures grown with or without ATc. Red lines viability by three orders of magnitude ( Fig 3B) and decrease in cell wall thickness (~18%; Fig  3D and 3E). Taken together, the data suggests that the absence of GlmU Mtb in hypoxic condition leads to aberrant cell wall thickness and morphology, eventually leading to the death of the cell.
Acetyl and uridyltransferase activities of GlmU Mtb are independently essential Biochemical investigations have shown that the N-terminal fragment (1-352 amino acids) and C-terminal fragment (150-495 amino acids) of GlmU Mtb can independently undertake uridyltransferase and acetyltransferase activities respectively (Fig 4A and 4B) [15,17]. The active site residues that are necessary for these activities have also been identified (Fig 4A and 4B) [17]. To investigate if both activities are essential for cell survival, we have generated previously reported truncation mutants GlmU-N and GlmU-C [38]. We also generated GlmU K26A and GlmU H374A, the uridyltransferase and acetyltransferase active site mutants, and GlmU DM wherein both the active site residues were concomitantly mutated. GlmU Mtb wild type and mutant proteins were purified ( Fig 4C) and their uridyltransferase and acetyltransferase activities were assayed. While GlmU-C and GlmU K26A mutants showed acetyltransferase activity, as expected they did not show any uridyltransferase activity (Fig 4D). On the other hand GlmU-N and GlmU H374A had uridyltransferase activity but not the acetyltransferase activity ( Fig 4D). As expected the double mutant did not have either uridyl or acetyltransferase activity ( Fig 4D). Next complementation assays using one or other truncations / active site mutants were carried out. The FLAG-GlmU Mtb and the complemented untagged wt-GlmU Mtb proteins were found to be expressed at similar levels ( Fig 4E). The episomally expressed wt-GlmU Mtb could rescue the RvΔglmU phenotype in the presence of ATc ( Fig 4F). Contrastingly, while the various GlmU Mtb mutant proteins were expressed at levels comparable to that of FLAG-Gl-mU Mtb ( Fig 4E); none of them rescued the growth defects of the RvΔglmU strain in the presence of ATc ( Fig 4F). These results indicate that both uridyltransferase and acetyltransferase activities of GlmU Mtb are essential for pathogen survival and imply that the only source of the metabolites GlcNAc-1-P and UDP-GlcNAc is through the GlmU Mtb mediated synthesis pathway.

Presence of GlmU Mtb is obligatory for the survival of Mtb in the host
Mtb cells devoid of an intact cell wall have been found to be capable of surviving inside the host [39,40]. Some pathogens have been reported to resort to cell wall "recycling" for the synthesis of UDP-GlcNAc, and others have been known to utilize GlcNAc from the host for this purpose [41][42][43][44]. However, such mechanisms have not yet been reported in Mtb. To investigate these possibilities we examined the impact of GlmU Mtb depletion on survival of the pathogen in the host. Using an ex vivo THP-1 infection model we observed~80% phagolysosome fusion in the absence of GlmU Mtb (Fig 5A and 5B; compare RvΔglmU with RvΔglmU +ATc). This was also reflected in the survival pattern of the pathogen upon depletion of GlmU Mtb (Fig 5C), with survival being strongly compromised in absence of GlmU Mtb    (C) GlmU Mtb and GlmU Mtb -mutants were purified as described earlier [38] and the 1 μg of the purified proteins were resolved on 10% SDS-PAGE and stained with coomassie. (D) Uridyltransferase (left panel) and acetyltransferase (right panel) activities were carried out as describe in Methods using 0.5 to 20 pmoles of wild type or mutant GlmU Mtb proteins. Activity was defined as μM product formed / min / pmole of enzyme. Relative activities of the mutants were calculated with respect to the activity of GlmU Mtb , which was normalized to 100%. The experiment was repeated three times and the error bars indicate s.e.m. (E) Wild type and mutated GlmU Mtb genes were cloned into pNit vector without any Nor C-terminal tag. pNit-glmU wt or pNit-glmU mutant constructs were electroporated into RvΔglmU, and the WCLs prepared from RvΔglmU and RvΔglmU:: glmU mutant cultures were resolved and probed with anti-GlmU and anti-GroEL1 antibodies. Bands corresponding to FLAG-GlmU Mtb , complemented GlmU wt/ mutant and the deletion fragments of GlmU are indicated. (F) RvΔglmU and RvΔglmU::glmU mutant cultures were seeded at an initial A 600 of 0.1 and grown for five days in the absence or presence of ATc. The experiment was performed in triplicates and the error bars represent s.e.m.  efficient and equivalent implantation of both wild type and mutant bacilli in the lungs of guinea pigs ( Fig 5D). Discrete bacilli were observed in the lungs of guinea pigs infected with Rv and RvΔglmU 28 days post-infection (Fig 5E and 5F). In contrast, the lungs of the guinea pigs infected with RvΔglmU in the presence of doxycycline were clear (Fig 5E and 5F). In addition splenomegaly was significantly reduced upon depletion of GlmU Mtb (RvΔglmU + Dox; S2A and S2B Fig). Whereas the bacillary load in the lungs and spleen of guinea pigs infected with Rv and RvΔglmU were comparable, we did not detect any bacilli when the RvΔglmU infected guinea pigs were administered Dox ( Fig 5D). In accordance with these observations, while the gross pathology of lungs infected with Rv and RvΔglmU displayed considerable granulomatous architecture, normal lung parenchyma was observed upon GlmU Mtb depletion ( Fig 5E). These results suggest that the presence of GlmU Mtb is obligatory for mycobacteria to survive in the host.

Depletion of GlmU Mtb from infected lungs leads to clearance of pathogen
It was apparent from the data presented above that the addition of ATc or Dox at the time of inoculation or at the time of infection does not allow mycobacterial cell growth or survival in the host. In the ideal candidate for therapeutic intervention, inhibiting the activity of/ depleting the enzyme at any stage of the infection should result in pathogen clearance. We assessed this parameter of GlmU Mtb by providing ATc at different stages of bacterial growth (early, log and stationary phases) and investigating its influence on cell survival in liquid cultures. Addition of ATc to RvΔglmU cultures on the 2 nd , 4 th or 6 th day after inoculation significantly thwarted growth ( Fig 6A). A similar analysis of bacterial growth by serial dilution of cultures followed by spotting on solid medium also revealed that viability was compromised by~2 log fold 48 h after the addition of ATc, indicating that GlmU Mtb depletion negatively impacted cell survival regardless of which stage of cell growth it was depleted at (S3A Fig).
The influence of GlmU Mtb depletion on an established ex vivo infection was estimated by providing ATc 24 h post-infection in a THP-1 infection model. As expected the bacillary load in THP-1 cells infected with Rv and RvΔglmU were similar at 0 and 24 h after infection ( Fig  6B). In contrast, while at 96 h post-infection the bacillary load for Rv and RvΔglmU-infected THP-1 cells remained the same, the addition of ATc to RvΔglmU-infected THP-1 cells 24 h after infection decreased the pathogen load by~2.5 log fold, indicating that the reduction of GlmU Mtb levels impacts pathogen survival even in an established ex vivo infection ( Fig 6B). We extended this investigation to analyze the effect of GlmU Mtb depletion from a fully-infected lung using murine infection model. As anticipated, the bacillary load in the lungs of mice infected with Rv and RvΔglmU were comparable both on Day 1 and on Day 28. Administration of Dox to RvΔglmU infected mice for the next 56 days (Day 28 to Day 84) drastically decreased the CFUs in the lungs ( Fig 6C) and the pathogen was completely cleared from the spleen (S3B Fig). Unlike the lungs of mice infected with Rv and RvΔglmU, mice infected with RvΔglmU to whom Dox was administered displayed a total absence of lesions and granulomas in the lungs (Fig 6D and 6E). Collectively, these data suggest a fundamental role for UDP-GlcNAc, the end product of the GlmU Mtb -mediated enzymatic reaction, in modulating the persistence of Mtb infection.

Oxa33: A novel allosteric GlmU Mtb inhibitor
In addition to the acetyltransferase and uridyltransferase active site pockets, GlmU Mtb also contains an allosteric site. Binding of any suitable molecule/inhibitor to the allosteric site would prevent the conformational change essential for GlmU Mtb uridyltransferase catalytic activity. To target the allosteric site on GlmU Mtb we drew on crystal structure data of H. influenza  [27]. Alignment of the GlmU Mtb and GlmU HI allosteric pocket residues suggested that the interacting residues were conserved between the two proteins (S4C and S4D Fig). The Asinex database was screened against shape as described (S5A Fig) and the resulting 43 hits were biochemically characterized for their ability to inhibit GlmU Mtb uridyltransferase activity. One of the promising molecules was used for further structural optimization (S5A Fig). Of the 53 structurally optimized compounds one molecule, namely (4Z)-4-(4-benzyloxybenzylidene)-2-(naphthalen-2-yl)-1,3-oxazol-5(4H)-one (Oxa33; Synthesis scheme provided in Figs 7A and S5B), was found to be an efficient inhibitor of GlmU Mtb activity with an IC 50 of 9.96±1.1 μM (Fig 7B). Isothermal titration analysis suggested an adequately high affinity binding for the compound (K a = 2.35×10 6 M -1 ), with a binding stoichiometry of 0.7 (S6A Fig). We sought to identify the residues in GlmU Mtb that are critical for interacting with Oxa33. Docking and MD simulation studies revealed polar, non-polar and hydrophobic interactions between Oxa33 and the allosteric site residues (Figs 7C, S6B and S6C). Based on the obtained data a panel of GlmU Mtb proteins each carrying a single mutation was created, the mutant proteins were purified (S6D Fig), and their uridyltransferase activity assayed. While all the mutants had similar levels of uridyltransferase activity there was a substantial increase in their IC 50 values, suggesting a loss of interaction with Oxa33 (Fig 7D and 7E).
To decipher the mechanism of Oxa33 mediated inhibition of uridyltransferase activity, we superimposed the GlmU Mtb -Oxa33 complex with the unbound GlmU Mtb structure. Upon Oxa33 binding, the loop regions (in the range of 3-6 Å) at the uridyltransferase active site undergo significant conformational changes, decreasing the active site volume, which results in occlusion of the substrates (Fig 7F). Differential scanning fluorimetry (DSF) analysis of GlmU Mtb in the presence of Oxa33 showed a 3°C shift in protein melting temperature (T m ) validating the conformational changes (S5E Fig). Interestingly we also observed much higher relative fluorescence units (~10000 vs 2500) in the presence of Oxa33, which is likely due to the compound induced structural changes facilitating increased binding of the dye (S6E Fig).
Together, these data demonstrate that Oxa33 binds to the allosteric site at N-terminal domain of GlmU Mtb and inhibits its uridyltransferase activity by causing structural changes.

Specificity and efficacy of Oxa33
Subsequently we investigated the ability of Oxa33 to inhibit the in vitro growth of Mtb H37Rv. Oxa33 inhibited the in vitro growth of Mtb H37Rv with a minimum inhibitory concentration (MIC) of~75 μM (~30 μg / ml) and a maxium bacteriocidal concentration (MBC) of~150 μM (~60 μg / ml). To ascertain if this inhibitory effect was due to inhibition of GlmU Mtb activity we overexpressed GlmU Mtb in the cells prior to drug treatment and determined the effect of this on the MIC value (Figs 8A and S6A). Whereas the inhibition of growth in the presence of INH was similar with or without GlmU Mtb overexpression in the cells (Fig 8A, lower panel), Oxa33 failed to inhibit cell growth even at concentrations as high as 150 μM (60 μg/ ml) (Fig 8A,  upper panel). Interestingly when sub lethal concentration of Oxa33 was provided, the MIC of INH decreased from 32 to 16 ng/ml (S7B Fig). The impact of Oxa33 on THP1 cells 24 h after infection with either Rv or Rv::glmU tet-on was also investigated. In concurrence with the in vitro growth data, overexpression of GlmU Mtb alleviated Oxa33-mediated clearance of Mtb from THP-1 cells (Figs 8B, S6D and S6E). These results suggest that the inhibition of mycobacterial growth by Oxa33 is specifically due to inhibition of endogenous GlmU Mtb . Finally, we analysed the efficacy of Oxa33 in clearing bacilli from infected lungs using a murine infection model. Oxa33 compound is highly hydrophobic in nature. After trying many solvents, we could successfully resuspend it in 2.5% Tween-80. Prior to performing the experiments we examined the maximum dose tolerance and survival analysis to determine the toxicity (S8A and S8B Fig). Based on the data obtained we chose 50 mg / kg as the appropriate dose.  Since it was difficult to predict the fate of Oxa33 during the process of digestion, we avoided using the oral administration route. We chose intra peritoneal route for administering the compound as the intravenous (I/V) injection of Tween 80 (solvent) in the animals was known to cause hypersensitivity and anaphylactic shock [45,46].
Groups of mice were infected with Rv and were treated with vehicle, INH, or Oxa33 at 28 days post-infection, for a duration of 56 days (Fig 8C, line diagram). Compared with the vehicle-treated group where we observed a marginal increase in bacillary load, a significant reduction in the bacillary load was observed in the lungs and spleen of both, INH-and Oxa33-treated groups (~4 and 2.5 log fold, respectively for lungs) (Figs 8C and S9A). This was also reflected in the gross pathology and histopathology of lungs (Fig 8D and 8E). Although in vitro MBCvalues of Oxa33 was~150 μM (60 μg/ml), it seems to be a relatively more efficacious in vivo, which could be due to its accumulation in the lungs of the infected mice. To investigate this possibility uninfected mice were treated with 50 mg/kg Oxa33 for a period of 3 weeks or 8 weeks. In order to estimate the concentration of Oxa33 in the lung, we first determined the absorbance spectra for Oxa33, which gave a clear peak at 401 nm (S10A Fig). We determined the A 401 at different concentrations of Oxa33 and the standard curve was plotted (S10B and S10C Fig). Oxa33 was extracted from the lungs and its concentration was determined. The concentrations of Oxa33 in the lungs were in the range of~200-300 μg /lung at 3 weeks and~800-1300 μg/lung at 8 weeks (S10D Fig). The accumulation of Oxa33 in the lungs is~13 to 18 fold higher than the MBC values, which may be the reason for higher potency of Oxa33 in vivo compared with the in vitro experiments. Taken together, the results presented in this study establish GlmU Mtb to be an effective target against which new sets of inhibitors may be developed.

Discussion
Cell wall provides the structural rigidity and protects bacteria from various environmental and physiological insults. Biosynthesis of the cell wall of bacteria is a complex process requiring enzymes localized to different cellular compartments [47]. Due to the essentiality of the enzymes involved they are considered attractive targets for anti-microbial therapies. The majority of the first line and second line anti-tuberculosis drugs from the existing regimen target enzymes involved in cell wall synthesis [5]. These include Isoniazid and Ethionamide targeting enoyl-[acyl-carrier-protein] reductase and inhibiting mycolic acid synthesis, Ethambutol targeting arabinosyl transferase and inhibiting arabinogalactan biosynthesis, and Cycloserine targeting D-alanine racemase and ligase, which inhibits peptidoglycan synthesis [5]. However most of these drugs are not very effective against dormant/ latent Mtb [48].
UDP-GlcNAc is a critical metabolite both in prokaryotes and eukaryotes. In eukaryotes it is mainly utilized for O-or N-GlcNAcylation, sialic acid biosynthesis and hylauronic acid biosynthesis [49][50][51]. In addition to the peptidoglycan synthesis [10], in gram negative bacteria UDP-GlcNAc is required for the synthesis Lipid A subunit of lipopolysaccharide [52] and in gram positive bacteria it is required for Rha-GlcNAc linker [53], arabinogalactan [54], teichioc acid synthesis [55]. In few prokaryotes, UDP-GlcNAc has also been shown to be required for sialic acid [56], N-GlcNAcylation [57] and poly (-GlcNAc-) n [58]. GlmU Mtb, an enzyme with dual activity, synthesizes a core metabolite necessary for the synthesis of cell wall peptidoglycan, UDP-GlcNAc [6]. Interestingly, we found that depleting GlmU Mtb during both normoxic and hypoxic growth resulted in substantial decrease in cell viability (Fig 3). This may be due to the requirement of UDP-GlcNAc, which in addition to participating in cell wall synthesis is also required for other cellular processes such as mycothiol biosynthesis (to maintain redox homeostasis) [14,59]. However, the TEM data clearly shows decreased cell envelop thickness even in hypoxic conditions (Fig 3). Although the CFUs do not change significantly the cells may be undergoing significant replication, which might be balanced by death [60]. Alternatively, new cell envelop may be required even if the bacteria are not replicating. Thus one can rule out the possibility that decreased viability may well be due to requirement of UDP-GlcNAc for the cell envelop synthesis.
While UDP-GlcNAc is a critical metabolite for both prokaryotes and eukaryotes, the enzymes involved in its de novo synthesis are significantly different [10]. In addition, both prokaryotes and eukaryotes can utilize GlcNAc from different sources to synthesize UDP-GlcNAc through salvage pathways [61][62][63] (S11 Fig). Capnocytophaga canimorsus, a member bacteria from Bacteroidetes phylum lacks endogenous GlmM and GlmU required for the synthesis of GlcNAc and it instead relies on GlcNAc obtained from forages glycans from the host mucin and N-linked glycoproteins [42]. Depending on the enzymes of the salvage pathway present in the bacterial system, it would require either both the activities or only the uridyltransferase activity of GlmU Mtb for UDP-GlcNAc synthesis. Till date the presence of alternate salvage pathway in Mtb has not been demonstrated. However, even with an operating salvage pathway GlmU Mtb is essential for the utilization of host GlcNAc to form UDP-GlcNAc (S11 Fig). In line with this, we find that depletion of GlmU Mtb during ex vivo or in vivo infection either at the start or after infection has been definitively established leads to clearance of pathogen.
GlmU Mtb and the acetyltransferase and uridyltransferase enzymes found in eukaryotes share very little sequence similarity. Although efforts have been made by different groups to target bacterial GlmU proteins, the specificity of these inhibitors for GlmU in vivo have not been established [23][24][25][26][27][28][29][30]. Most GlmU inhibitors characterized till date target either the acetylor uridyltransferase active sites. In contrast, inhibitors of GlmU HI target the allosteric site near the uridyltransferase active site [27]. The interaction of the inhibitor with the enzyme via this allosteric site perturbs the active site conformation of the protein, thus inhibiting uridyltransferase activity [27]. In the present study we have used shape based designing and developed a novel oxazolidine molecule, Oxa33, and characterized its ability to bind to the GlmU Mtb allosteric site. MD simulation and mutation of critical interacting residues to defined the possible allosteric site residues required for Oxa33 binding (Fig 7). DSF (S6E Fig) and structural superimposition (Fig 7) supports that inhibition of uridyltransferase activity is due to structural changes in the N-terminal domain of GlmU Mtb . Further in order to determine the specificity of Oxa33, GlmU Mtb over expressing strains of Rv was used to determine the MIC. Both in vitro and ex vivo results (increased MIC or MBC) validate that Oxa33 specifically binds to GlmU Mtb inside the bacteria. Administrating the Oxa33 to fully infected (28 days) mice resulted in partial ablation of pathogen load in the lungs. Taken together results presented here demonstrates that GlmU Mtb is a viable and promising target for therapeutic intervention and Oxa33 can be pursued as a lead molecule, which needs to be developed further to improve its efficacy.

Chemicals and reagents
Restriction enzymes and Phu DNA polymerase were purchased from New England Biolabs. pENTR/directional TOPO cloning kit (Invitrogen), pQE2 (Qiagen), were procured from the respective sources. Analytical grade chemicals and oligonucleotide primers were procured from Sigma. Malachite green phosphate assay kit (POMG-25H) was purchased from BioAssay System (Gentaur). Electron microscopy reagents were purchased from Electron Microscopy Sciences. Media components were purchased from BD Biosciences. Doxycycline hydrochloride was purchased from Biochem pharmaceutical.

Generation of glmU conditional gene mutant in M. tuberculosis
The hexa-His tag in the pST-KiT construct [15] was replaced with an N-terminal FLAG tag, and the tetracycline repressor gene (tetR) was replaced with a reverse tetR (r-tetR) from pTC28S15-OX [64] to create plasmid pST-KirT. To generate the integrating shuttle plasmid pST-KirT-glmU Mtb , the glmU Mtb gene was excised from pQE2-glmU Mtb using NdeI-HindIII digestion and was subcloned into the corresponding sites on pST-KirT. The resulting pST-KirT-glmU construct expresses GlmU Mtb in the absence of inducer ATc. Upon addition of ATc, ATc binds to the r-TetR repressor resulting in the conformational change that would allow it to bind to the operator seqeunces in P myc tetO (S1 Fig) [64]. The integration-proficient plasmid containing the inducible glmU Mtb gene was electroporated into mycobacterial cells to create the merodiploid strain Rv::glmU Mtb . 5' and 3' genomic flank sequences of glmU Mtb (approximately 1 kb on either side) were amplified, the amplicons digested with PflMI, and ligated with the antibiotic resistance cassette along with the oriE and cosλ fragments generated from pYUB1474 construct, to generate the allelic exchange substrate (AES) [65]. The AES was linearized using the unique PacI site and then cloned into temperature sensitive shuttle phagemid phAE159 at the PacI site. A conditional gene replacement mutant of RvΔglmU was created from the merodiploids with the help of specialized transduction methodology (S1A Fig) [66]. RvΔglmU recombinants obtained were analyzed by PCR amplification to verify the fidelity of the recombination event.

Analysis of growth patterns
H37Rv (Rv) and RvΔglmU cultures were grown in Middlebrook 7H9 medium supplemented with 10% ADC (albumin, dextrose and catalase), or in 7H11 medium supplemented with 10% OADC (oleic acid, ADC). To analyze bacterial growth in vitro, Rv and RvΔglmU mutant bacterial cultures were inoculated at A 600 of 0.1, in the presence or absence of anhydrotetracycline (ATc), and A 600 was measured every 24 h for 6 or 8 days. For spotting analysis, cells were harvested by centrifugation, washed twice with PBST (0.05% Tween 80) to remove ATc, resuspended in 7H9 medium, and serially diluted in the same medium, followed by spotting 10 μl aliquots of the various cell dilutions on 7H11 agar plates to assess cell viability. To determine the impact of GlmU Mtb depletion during hypoxia in Rv and RvΔglmU strains, we established hypoxia in 1.5 ml HPLC tubes or 500 ml flasks with penetrable caps, following modified Wayne model [35]. ATc (2 μg/ml) or isoniazid (INH) (50 ng/ml) were injected into the cultures at different time points and the number of CFUs were determined after 42 days. Scanning and transmission electron microscopy (SEM & TEM) analysis of Rv and RvΔglmU mutant grown in the presence or absence of ATc were performed as described earlier [67]. Transmission electron microscopy was performed using standard protocols. Briefly, bacteria was fixed in 2.5% gluteraldehyde and 4% paraformaldehyde, dehydrated in graded series of alcohol and embedded in Epon 812 resin. Ultrathin sections were cut and stained with uranyl acetate and lead citrate [68]. SEM images were procured using Carl Zeiss Evo LS scanning electron microscope, and TEM images were captured using Tecnai G2 20 twin (FEI) transmission electron microscope.

Generation of glmU Mtb mutant constructs and western blotting analysis
Site directed mutations of glmU Mtb were generated with the help of overlapping PCR and the amplicons were cloned into NdeI-HindIII sites of pQE-2, pNit and pST-KT vectors [15,69]. The tetracylin repressor (TetR) expressed from the plasmids binds to the operator sequence in the promoter P myc tetO in the absence of ATc (S1B Fig) [70]. Addition of ATc to TetR alleviates the repression thus inducing the expression of GlmU. pST-KT-glmU was electroporated into Rv to generate Rv::glmU tet-on strain. pNit-glmU (wild type and mutated) constructs were electroporated into RvΔglmU to generate RvΔglmU::glmU wt/mutant strains. Rv and RvΔglmU:: glmU wt/mutant strains were grown in the presence or absence of ATc as described above. GlmU Mtb was expressed and purified using plasmid pQE2-GlmU Mtb , as described earlier [15]. Whole cell lysates (WCL) isolated from Rv, RvΔglmU and RvΔglmU::glmU wt/mutant strains that had been grown for 5 days in presence or absence of Atc, were resolved on 10% SDS-PAGE, transferred to nitrocellulose membrane, and probed with anti-GlmU Mtb and anti-GroEL1 antibodies as described earlier [15,67].

Ex vivo and in vivo infections
THP1 infection experiments were carried out with either unlabelled or FITC-labelled Rv and RvΔglmU strains at 1:10 MOI, as described earlier [71]. For examination of cells under a fluorescence microscope, infected cells (48 h post-infection) were labelled with Lyso Tracker red DND 99 dye (50 nM) and mounted with Antifade (Invitrogen) mounting agent. To determine CFUs per infected cell, the infected cells were lysed in PBS containing 0.1% TritonX-100 for 15 min and different dilutions were plated on OADC-containing 7H11 agar plates. For animal infection experiments, Rv and RvΔglmU strains grown till A 600 of 0.6 were used to infect 3 to 4 week old guinea pigs or~2 month old mice as described previously [72,73]. We initially used guinea pig model system as it has robust immune response. However, for the remaining experiments we chose to use Balb/C mice model of infection as the cost associated with performing the experiments and the amount of Oxa33 required for guinea pig experiments was prohibitive.
To determine the implantation dosage, the bacillary load in the lungs of guinea pigs or mice was determined 24 h post-infection. To investigate the impact of glmU Mtb depletion on survival of the pathogen, doxycycline hydrochloride (Dox, 1 mg/ kg with 5% dextrose in drinking water) was provided to Rv and RvΔglmU-infected animals as indicated, either from the time of the infection (guinea pig experiment), or 4 weeks post-infection (mice infection experiments). To assess the impact of INH or Oxa33 treatment on pathogen survival, Rv-infected mice (4 weeks post-infection) were supplied with INH (25 mg/ kg body weight, with 5% dextrose in drinking water) or Oxa33 (50 mg/ kg body weight, with 2.5% Tween 80, injected intra peritoneally) every third day for 8 weeks. Bacillary loads in the lungs and spleens of infected guinea pigs and mice were determined 4 weeks and 12 weeks post-infection. Histopathological evaluation of the harvested organs was performed as described earlier [67,72,73].
Shape based screening and molecular docking studies ROCS (Rapid Overlay of Chemical Structures), a shape based technique for rapid similarity analysis was used to assess the compounds. Gaussians and shape tanimoto were used to assess the volume and shape overlaps of the compounds, respectively. As the chemical functionality is critical, the chemical feature based similarity was also considered using ROCS colour score whose force field was composed of SMARTS patterns of the chemical functions [74,75]. The shape tanimoto score and scaled color score were considered during the selection of the compounds for further virtual screening. The compounds selected were subjected to molecular docking studies using Glide v5.8 of Schrödinger molecular modelling suite 2012 (Glide v5.8, Schrödinger). The compounds were subjected to a series of docking protocols-high throughput virtual screening (HTVS), standard precision (SP) and extra precision (XP) docking. As the docking progresses from HTVS to XP, the algorithm differs, which starts from a simple docking of compounds and ends with docking protocol with high precision and parameterization while cutting off the number of compounds.

Isothermal titration calorimetry
To investigate the binding of Oxa33 to GlmU Mtb , we performed Isothermal Titration Calorimetry (MicroCal 2000 VP-ITC, GE Healthcare) [28]. Oxa33 was re-suspended in dialysis buffer (25 mM Tris pH 7.4, NaCl 140 mM and 15% glycerol), 100 μM of MgCl 2 containing 2% DMSO. 625 μM of Oxa33 was injected for titrations from syringe (rotating at 307 rpm) into ITC cell containing 25 μM of GlmU or blank buffer at 25°C. Each injection lasted for 20 sec with 300 sec interval between every step. The quantity of heat associated by every injection was calculated by combining the area beneath every heat burst curve (microcalories/second vs. seconds). Data was corrected for the buffer signal and fitting was done by one-site binding model. Origin software (version 7.0) was used to obtain different thermodynamic binding parameters.

In vitro cytotoxicity
Oxa33 was evaluated for its cytotoxic activity in THP1 cells with the help of alamar blue assay. Serially diluted inhibitors (in 2.5% DMSO) incubated with 5 x 10 3 differentiated THP-1 cells in 96 well plates for 3 days. After 3 days cells were incubated for 5 h with 10 μl of alamar blue and color development was measured using micro-plate reader at 570 nm.

Docking and molecular dynamics simulations studies of GlmU Mtb with Oxa33
Molecular dynamics (MD) simulation for the protein-ligand complex was carried out for a time scale of 20 ns so as to analyze the stability of molecular interactions between ligand and protein employing Newton's Laws of Motions. Desmond molecular dynamics system v3.1 was used for carrying out the simulations employing OPLS-AA force field [77]. The protein-ligand complex was solvated using TIP3P water model which was setup as an orthorhombic solvent box, keeping a cut-off of 10 Å from any solute atom in all directions [78]. Na + counter ions were added in order to neutralize the system. A cut-off of 14 Å was maintained for calculating the solvent-solvent and solute-solvent non-bonded interactions. Initially, the system was minimized keeping the convergence threshold criteria of 1.0 kcal.mol -1 .Å -1 so as to allow the adjustment of atoms to the system environment. A simulation for each system was performed using isothermal-isobaric ensemble (NPT) including a relaxation process. Under NPT, the system was simulated for 12 ps using a Berendsen thermostat and a Berendsen barostat with temperature of 10K and a pressure of 1 atm. The later step of relaxation protocol included the simulation of the system for 24 ps with a temperature of 300 K and 1 atm pressure with and without restraints on solute heavy atoms. M-SHAKE algorithm was used with an integration time step of 2 fs for rearranging the hydrogen bonds in the simulation [79]. The temperature and pressure of the system were maintained at 300 K and 1.013 atm respectively. The molecular dynamics simulation was run for 20 ns recording the trajectory frames at an interval of every 4.8 ps and the trajectory analysis was carried out using the Simulation Event Analysis of Desmond.

Determination of percentage inhibition, IC 50 and MIC
Uridyltranferase assays were performed using malachite green phosphate detection kit as described previously [17]. Acetyltransferase activity of GlmU Mtb was carried out in the presence of 500 μM each of GlcN-1-P and acetyl-CoA in a 30 μl reaction volume for 30 min at 30°C as described earlier [80]. To determine the percent inhibition by different compounds the enzyme was preincubated with either 5% DMSO or 100 μM compounds for 30 min prior to performing uridyltransferase activity assays. In order to determine the IC 50 values, GlmU wt/ mutant proteins were preincubated with different concentrations of Oxa33 compound for 30 min followed by the uridyltransferase assay. To determine minimum inhibitory concentration (MIC), 5x10 5 bacteria of Rv or Rv::glmU tet-on (overexpressing GlmU Mtb ) cultures (grown in the presence or absence of 2 μg/ml ATc) were mixed with 100 μl of 2.5% DMSO or different concentrations of Oxa33/ INH in 96-well plates, and incubated at 37°C for 6 days. After 6 days, 40 μl of resazurin dye (0.02% in 5% Tween-80) was added to each well and the colour change was observed after 12 h.

Ethical statement
Experimental protocol for the animal experiments was approved by the Institutional Animal Ethics Committee of National Institute of Immunology, New Delhi, India (the approval number is IAEC# 315/13). The approval is as per the guidelines issued by Committee for the Purpose of Control and Supervision of Experiments on Animals (CPCSEA), Govt. of India.

Statistical analysis
Student's t-test (two tailed non parametric) was used to analyze the significance of cell wall thickness, THP1 and animal infection experimental results. SigmaPlot version 10.0 and Graph-Pad Prism version 5.0 was used for the statistical analysis and for plotting the results. Virtual screening workflow used towards the identification of allosteric inhibitors for GlmU Mtb . Shape model of allosteric inhibitor of GlmU HI was generated using ROCS software. In the initial set of screening, a total of 43 compounds were identified and tested for their ability to inhibit uridyltransferase reaction. 10 among the 43 compounds showed~90% inhibition at 100 μM. Of these, an oxozolone derivative was identified and further chemically derivatized to a library of 52 compounds. One compound from these, Oxa33, which showed~90% inhibition was considered for further studies. (B) Nuclear magnetic resonance spectra of purified Oxa33 (Yellow solid). 1  Model shows De novo pathway for UDP-GlcNAc synthesis is mediate by GlmS, GlmM and GlmU enzymes. Shaded pathway is conserved in Mtb. UPD-GlcNAc can inhibit uridyltransferase activity by feedback inhibition mechanism. Also GlcNAc from host resources or from cell wall recycling can be transported inside the bacteria and further metabolized and feeded into the de novo pathways through GlmS/ GlmU mediated reactions. Question marks show that these pathways are still not characterized in Mtb. Dashed lines shows possible input of substrates or unknown pathway while complete lines shows established and known pathways for UDP-GlcNAc synthesis. (TIF) S1 Table. List of bacterial strains, plasmids and phages used in the study.

Supporting Information
(DOC)