Gut Microbiota Metabolite Indole Propionic Acid Targets Tryptophan Biosynthesis in Mycobacterium tuberculosis

New drugs against tuberculosis are urgently needed. The tryptophan (Trp) analog indole propionic acid (IPA) is the first antitubercular metabolite produced by human gut bacteria. Here, we show that this antibiotic blocks Trp synthesis, an in vivo essential biosynthetic pathway in M. tuberculosis. Intriguingly, IPA acts by decoupling a bacterial feedback regulatory mechanism: it mimics Trp as allosteric inhibitor of anthranilate synthase, thereby switching off Trp synthesis regardless of intracellular Trp levels. The identification of IPA’s target paves the way for the discovery of more potent TrpE ligands employing rational, target-based lead optimization.

acid L-tryptophan (Trp) (8). M. tuberculosis is capable of synthesizing Trp de novo. Zhang et al. (12) and Wellington et al. (13) have demonstrated that the Trp biosynthetic pathway is essential for growth and viability of M. tuberculosis in standard (Trp-free) medium as well as in macrophages and mice. In M. tuberculosis, Trp regulates its own synthesis by acting as a cooperative allosteric inhibitor of anthranilate synthase TrpE, which catalyzes the first committing step in the pathway (14,15).
The homodimeric M. tuberculosis protein TrpE (Rv1609), in conjunction with glutamine amidotransferase (TrpG) providing ammonia, catalyzes the formation of anthranilate from chorismate (15,16). The N-terminal region of TrpE contains the allosteric Trp binding motif 53 LLESX 10 S 67 (15). The 10-amino-acid residue X 10 loops of the two TrpE subunits approach each other at the interface, providing part of the structural framework for the cooperative allosteric inhibitory effect of Trp binding. In addition, TrpE residues 170 HHEGT 174 are involved in the cooperative allosteric inhibition mechanism by making interactions with their subunit's Trp binding motif as well as with the Trp binding motif and the 170 HHEGT 174 residues of the partner subunit (15).
In view of the structural relationship between IPA and Trp, we hypothesized that IPA may exert its antibacterial activity by mimicking Trp. Here we present metabolic, chemical rescue, genetic, and biochemical analyses showing that IPA exerts its wholecell anti-M. tuberculosis activity by blocking Trp biosynthesis through allosteric inhibition of TrpE activity.

RESULTS
IPA docks into the allosteric Trp binding pocket of TrpE. To determine whether IPA can theoretically bind to the allosteric Trp binding pocket of M. tuberculosis TrpE, docking experiments were carried out using Autodock4 and Vina software (17,18). The allosteric binding site for M. tuberculosis Trp was determined by superposition using the Trp-bound TrpE structures of Serratia marcescens and Salmonella enterica serovar Typhimurium with the Trp-free structure of M. tuberculosis TrpE (14,15,19) (Fig. 1A). Figure 1B shows that IPA can be docked into the Trp binding site of M. tuberculosis TrpE with a conformation similar to Trp, as shown in Fig. 1C. The indole moieties of IPA and Trp overlap similarly with the Trp binding site, forming hydrogen bonds with the carbonyl groups of Glu55 and/or Met285 and hydrophobic interactions with Leu54 and Tyr284. The positively charged amino group of Trp forms a hydrogen bond with the carboxylate group of Asp453, whereas the carboxylate of Trp is not involved in hydrogen bonding (Fig. 1C). In contrast, the carboxylate of IPA forms a hydrogen bond  (19), and Salmonella Typhimurium with bound Trp (PDB no. 1I1Q [blue]) (14). The location of Trp is indicated in red. The superposition was performed using the MOE software (44). IPA (B) and Trp (C) were docked into the structure of M. tuberculosis TrpE (5CWA) (15) using the AutoDockTools 4 (ADT) and Vina programs (17,18). The ADT and the Vina poses are colored green and blue, respectively. The calculated binding energies were Ϫ5.79 (ADT) and Ϫ8.2 (Vina) kcal/mol for IPA and Ϫ6.73 (ADT) and Ϫ8.2 (Vina) kcal/mol for Trp. Amino acid residues shown in panels B and C have at least one atom within a 5-Å radius of IPA or Trp. Leu54, Glu55, Ser56, Ser 64, and Trp66 are part of the allosteric Trp binding motif 53 LLESX 10 S 67 (Fig. 5) reported by Bashiri et al. (15).
with the hydroxyl group of Tyr445 (Fig. 1B). The finding that IPA docks into the Trp binding pocket lends support to our hypothesis that IPA mimics Trp as an allosteric inhibitor of TrpE (Fig. 2).
IPA treatment reduces intrabacterial Trp levels. If the mechanism of action of IPA involves interference with Trp biosynthesis, IPA treatment of M. tuberculosis cultures should cause a reduction of the intrabacterial concentration of Trp (Fig. 2). To measure the effect of IPA treatment on intrabacterial Trp levels, we used the biosafety level 2 (BSL2)-compatible surrogate for M. tuberculosis, Mycobacterium bovis BCG. The sequences of the TrpE orthologs in both bacteria are identical, and as demonstrated in M. tuberculosis, growth of M. bovis BCG is inhibited by IPA (Table 1). Fluoro-anthranilate (FA) was used as a positive control (12). This analog of the Trp pathway substrate anthranilate is converted to fluoro-tryptophan (F-Trp) by TrpD and subsequent pathway enzymes (20), and F-Trp has been shown to allosterically inhibit TrpE (21-23) (Fig. 2). Exponentially growing M. bovis BCG cultures were treated with increasing IPA concentrations, and Trp levels were measured using liquid chromatography-mass spectrometry (LC-MS). Figure 3 shows that IPA reduced Trp levels in a dose-dependent manner. Dose-dependent reduction of Trp was also observed with the positive control fluoro-anthranilate. Reduction of Trp levels by IPA and fluoro-anthranilate at growthinhibitory concentrations appears to be specific (as opposed to an epiphenomenon of growth inhibition) as these effects were not observed with the antitubercular agent ethambutol (EMB) targeting cell wall synthesis (Fig. 3). This result suggests that IPA interferes with Trp biosynthesis.
Supplementation of medium with Trp abrogates the inhibitory activity of IPA. As the standard mycobacterium broth Middlebrook 7H9 does not contain Trp, M. , low-level IPA resistance.
d Nuc, location and nature of single nucleotide polymorphism in coding sequence of respective gene; AA, location and nature of amino acid substitution or other effects on coding sequence (stop/nonsense codon, frameshift) associated with respective DNA polymorphism. MTB_IPA R _M1 harbored a C-to-G substitution 7 bp upstream of the start codon of Rv0880. Mutations in other genes represent genes other than trpE, Rv0880, and Rv0948c for which whole-genome sequencing revealed DNA polymorphisms. All strains shown were subjected to whole-genome sequencing, with the exception of the BCG strains labeled "ND." All strains were subjected to targeted Sanger sequencing for trpE, Rv0880, and Rv0948.
tuberculosis needs to synthesize this amino acid to grow. If IPA acts via blocking Trp synthesis by inhibiting TrpE, exogenous supply of anthranilate or Trp should eliminate the inhibitory effect of IPA (12,13,24,25). M. tuberculosis cultures grown in medium without any supplement or with exogenously supplied anthranilate or Trp were treated with increasing IPA concentrations, and the optical density at 600 nm (OD 600 ) was measured. Figure 4A shows that the presence of anthranilate or Trp completely suppressed the inhibitory activity of IPA. Figure 4B depicts as a positive control the expected suppressive effect of anthranilate and Trp on the inhibitory effect of fluoroanthranilate (12). Figure 4C shows that exogenous Trp or anthranilate did not affect the growth-inhibitory activity of the non-Trp pathway targeting ethambutol, suggesting that the effect of the Trp pathway metabolites was specific to inhibitors of the Trp pathway. The results of the medium supplementation experiments were confirmed with a different readout by plating aliquots of culture samples from each treatment condition on 7H11 agar and observing growth on solid medium (see Fig. S1 in the supplemental material). The finding that supplementation of medium with anthranilate or Trp abrogates the inhibitory activity of IPA supports the hypothesis that IPA inhibits synthesis of Trp. Mutations at the allosteric Trp binding site of TrpE cause loss of Trp feedback inhibition and resistance to IPA. If IPA mimics Trp as an allosteric TrpE inhibitor, mutations in the Trp binding site that result in loss of feedback inhibition by Trp should  To generate trpE mutants that have lost Trp binding, we employed a chemical genetic approach and selected mutants with spontaneous resistance to fluoroanthranilate. M. bovis BCG and M. tuberculosis cultures were plated on agar supplemented with fluoro-anthranilate, and three resistant colonies, two from M. bovis BCG and one from M. tuberculosis, emerging at a frequency of about 10 Ϫ8 /CFU (12), were restreaked on fluoro-anthranilate agar to confirm resistance. Growth inhibition experiments in broth revealed that all three mutants exhibited high-level resistance to fluoro-anthranilate ( Table 1). As expected, targeted sequencing of fluoro-anthranilateresistant M. bovis BCG revealed missense mutations in the N-terminal region of TrpE (Table 1). One strain (BCG_FA R _M1) harbored a mutation in the Trp allosteric binding motif 53 LLESX 10 S 67 (Fig. 5A), where Ser67 was replaced by Ala, likely abrogating Trp binding (14,19). The second fluoro-anthranilate-resistant M. bovis BCG mutant (BCG-_FA R _M2) harbored a Val130Phe mutation (Fig. 5A). Targeted sequencing of the fluoro-anthranilate-resistant M. tuberculosis strain (MTB_FA R _M1) ( Table 1) revealed a His170Asn missense mutation in the 170 HHEGT 174 motif, which interacts with the allosteric binding motifs of both TrpE subunits as well as with its counterpart of the partner subunit (15) (Fig. 5A). In the crystal structure of homodimeric TrpE, His170 is located next to the allosteric X 10 loop of the 53 LLESX 10 S 67 Trp binding motif of its own subunit, where it interacts via hydrogen bonding with Arg65 located in the X 10 loop (AENGRSWSR) (Fig. 5A) and via stacking with its symmetry mate, His170=, located in the To confirm that the fluoro-anthranilate resistance mutations in the allosteric Trp pocket of TrpE indeed prevent Trp binding, we generated the respective recombinant TrpE proteins TrpE S67A and TrpE H170N (see Fig. S2 and S3 in the supplemental material) harboring the Ser67Ala substitution found in BCG_FA R _M1 and the His170Asn mutation present in strain MTB_FA R _M1, and carried out enzyme inhibition experiments as described previously (15) (see Fig. S4 in the supplemental material). As expected, the activity of recombinant wild-type TrpE was inhibited by the physiological inhibitor Trp, with a 50% inhibitory concentration (IC 50 ) of 1.6 M, consistent with previous reports (12,15). This was also true for F-Trp, the transformation product of fluoro-anthranilate, which showed an IC 50 of 4.1 M. In contrast, the IC 50 values for Trp and F-Trp against the recombinant TrpE harboring Ser67Ala or His170Asn mutations increased by more than 100-fold (Fig. S4). These results show that fluoro-anthranilate-associated resistance mutations in the allosteric pocket of TrpE indeed abrogate Trp (and F-Trp) binding and are consistent with data generated for the fluoro-anthranilate-resistant Phe67Ile mutation in the Trp binding pocket of TrpE reported by Zhang and colleagues (12).
Next, we determined whether fluoro-anthranilate-resistant mutations conferred resistance to IPA as predicted. Table 1 shows that all three strains harboring fluoroanthranilate-resistant TrpE mutations indeed showed high-level cross-resistance to IPA. Direct biochemical IPA inhibition experiments with recombinant proteins could not be carried out because of strong autofluorescence of IPA and the resulting assay interference (12) (see Fig. S5 in the supplemental material).
Expression of trpE alleles with altered TrpE binding site in the wild-type background confer IPA resistance. If the mutant trpE alleles Ser67Ala and His170Asn cause IPA resistance and do so by preventing allosteric inhibition of TrpE by IPA, these alleles should be dominant, i.e., they should confer IPA resistance when expressed in trpE wild-type background. To test this prediction, we isolated the trpE Ser67Ala and His170Asn alleles via PCR (see Table S1 in the supplemental material) and expressed these genes in M. bovis BCG wild type under the control of the phsp60-based constitutive expression system supplied by the episomal plasmid pMV262 (26). We then determined the effect of trpE heterozygosity on the growth-inhibitory activity of IPA. Figure 6A shows that expression of both mutant trpE alleles caused IPA resistance. As expected, expression of the mutant trpE alleles also caused decreased susceptibility to the positive control, fluoro-anthranilate (Fig. 6B), but did not affect susceptibility of BCG to the negative control, ethambutol (Fig. 6C), suggesting that expression of mutant trpE alleles does not cause general drug resistance. Expression of the wild-type trpE allele IPA Inhibits TrpE ® did not affect susceptibility to IPA or fluoro-anthranilate as expected (13) (Fig. 6A and  B). Together, these results show that TrpE amino acid substitutions Ser67Ala and His170Asn in M. bovis BCG and M. tuberculosis are responsible for the observed resistance to IPA and fluoro-anthranilate and confirm the expected dominant phenotype of these mutations.
Selection of spontaneous IPA-resistant mutants identifies two trpE-independent mechanisms of IPA resistance. To determine whether allosteric inhibition of TrpE is the major and/or only mechanism of action of IPA, M. bovis BCG and M. tuberculosis cultures were plated on IPA agar, and resistant colonies were restreaked on IPA agar to confirm resistance. Interestingly, the spontaneous IPA resistance mutation frequency was about 5 ϫ 10 Ϫ6 /CFU, more than 100-fold higher than the frequency of resistance observed for fluoro-anthranilate, suggesting that additional, non-trpE-related resistance mechanisms may exist.
Indeed, MIC determinations of six IPA-resistant M. bovis BCG strains (BCG_IPA R _M1 to -M6) showed low-level IPA resistance compared to the high-level resistance displayed by trpE mutants BCG_FA R _M1/M2 and MTB_FA R _M1 selected on fluoroanthranilate (Table 1). Furthermore, these low-level IPA-resistant mutants were fluoroanthranilate sensitive. Interestingly, one IPA-resistant strain (BCG_IPA R _M7) displayed a third drug susceptibility pattern: high-level IPA resistance (similar to BCG_FA R _M1, M2 and MTB_FA R _M1 selected on fluoro-anthranilate) combined with a low level of resistance to fluoro-anthranilate. As expected, targeted sequencing showed no polymorphisms in trpE.
MIC determination of six IPA-resistant M. tuberculosis strains (MTB_IPA R _M1 to -M6) showed a similar result. Four out of the six mutants (MTB_IPA R _M1 to -M4) showed low-level IPA resistance and were fluoro-anthranilate sensitive (Table 1). Two out of the six strains (MTB_IPA R -M5 and -M6) showed high-level IPA resistance and low-level resistance to fluoro-anthranilate, similar to BCG_IPA R _M7 (Table 1). Again, targeted sequencing showed no polymorphisms in trpE. These phenotypic characterization results suggested two additional, trpE-independent resistance mechanisms against IPA. Whereas trpE-dependent resistance mechanisms cause high-level IPA and fluoroanthranilate resistance, the trpE-independent mechanisms confer either low-level IPA resistance with no fluoro-anthranilate resistance or high-level IPA resistance associated with low-level fluoro-anthranilate resistance.
To explore the genetic basis for the two trpE-independent resistance phenotypes, we subjected the six IPA-resistant M. tuberculosis strains (MTB_IPA R _M1 to -M6) to whole-genome sequencing. The four M. tuberculosis strains (MTB_IPA R _M1 to -M4) displaying low-level IPA and no fluoro-anthranilate resistance harbored various nonsense, deletion, and missense mutations in Rv0880 (Table 1). The Rv0880 gene encodes a nonessential transcription factor belonging to the family of multiple antibiotic resistance regulators (MarR) (27,28), regulating expression of a 23-gene regulon (29) and influencing bedaquiline susceptibility (30). How Rv0880 mutations affect IPA susceptibility remains to be determined and is likely to involve an indirect mechanism (30).
The two M. tuberculosis strains MTB_IPA R -M5 and -M6 displaying high-level IPA and low-level fluoro-anthranilate resistance harbored missense mutations (Arg100Ser and Ser64Pro) in Rv0948c (Table 1). Rv0948c encodes a cytoplasmic chorismate mutase catalyzing the conversion of chorismate, the substrate of TrpE, to prephenate, the precursor for the biosynthesis of phenylalanine and tyrosine (31) (Fig. 2). Both missense mutations are located in the catalytic domain of the enzyme, and Arg100 was shown to be involved in its catalytic activity (32). Hence, it is conceivable that IPA-resistant mutations in Rv0948c reduce the catalytic activity of chorismate mutase, resulting in an increase in the concentration of the starter substrate of Trp biosynthesis, chorismate. This in turn may influence the vulnerability of the Trp pathway to inhibition of TrpE by IPA and fluoro-anthranilate (32,33). Rather than being a direct target of IPA, Rv0948c appears to rewire the anabolism of aromatic amino acids, impacting vulnerability of the Trp biosynthetic pathway to inhibitors.
Counterscreening of spontaneous IPA-resistant mutants for high-level fluoroanthranilate resistance identifies trpE mutations. The ϳ100-fold difference in resistance frequency to IPA and fluoro-anthranilate (Table 1) led us to hypothesize that high-level fluoro-anthranilate cross-resistant mutants harboring lesions in trpE could be detected in a collection of spontaneous IPA-resistant mutants at a frequency of about 0.1 to 1%. To identify such high-level IPA-fluoro-anthranilate double-resistant mutants, we carried out a second IPA-resistant mutant selection experiment with M. tuberculosis, in which we isolated and confirmed 200 IPA-resistant colonies (Table 1). Then we determined cross-resistance to fluoro-anthranilate and identified five IPA-fluoroanthranilate double-resistant colonies. Broth MIC determinations of these five strains revealed four high-level IPA-high-level fluoro-anthranilate-resistant strains (MTB_I-PA R _M7 to -M10), suggesting that these strains may harbor trpE mutations. Targeted sequencing of these four high-level double-resistant strains showed indeed missense mutations at various locations in the N-terminal part of TrpE around the Trp allosteric binding motif (Fig. 5 and Table 1). These mutations included again a substitution at His170, this time with Arg (MTB-IPA R _M8 to -M10) ( Fig. 5 and Table 1). These results confirmed that high-level IPA-fluoro-anthranilate resistance is associated with missense mutations in the allosteric Trp binding site of TrpE. These mutations can be obtained via selection on fluoro-anthranilate-or IPA-containing medium.
Whereas four out of the five isolated IPA-fluoro-anthranilate double-resistant strains showed high-level IPA and high-level fluoro-anthranilate resistance, one strain (MTB_IPA R _M11) showed high-level IPA and low-level fluoro-anthranilate resistance ( Table 1), suggesting that this strain may harbor a mutation in the chorismate mutase gene, Rv0948c. Indeed, sequencing showed that this strain harbored a trpE wild-type gene and a missense mutation in the chorismate mutase gene (Table 1). This time the polymorphism was located at codon 33, resulting in a substitution of Arg by Leu. Arg33 of chorismate mutase is conserved across different species of bacteria and predicted to be involved in substrate binding (32). Thus, this mutation may affect activity of the enzyme.
Among the 200 resistant mutants isolated on IPA, 195 had shown no fluoroanthranilate resistance. To characterize the resistance phenotype of these fluoroanthranilate-sensitive strains, we randomly selected seven strains (MTB_IPA R _M12 to -M18) and determined their MICs. All seven strains showed low-level IPA resistance and no fluoro-anthranilate resistance in liquid medium, suggesting that their resistance is due to mutation in the transcription factor gene Rv0880. Sequencing indeed confirmed that all 7 strains harbored polymorphisms in Rv0880 (Table 1). We also subjected the M. bovis BCG mutants BCG_IPA R _M1 to -M7 (Table 1) to targeted sequencing of the Rv0948c and Rv0880 orthologues. As expected, the high-level IPA-low-level fluoroanthranilate-resistant strain (BCG_IPA R _M7) harbored a mutation in the chorismate mutase gene (Rv0948c) ( Table 1). Consistent with the corresponding M. tuberculosis observations, the remaining low-level IPA-resistant-fluoro-anthranilate-susceptible strains displayed polymorphism in the orthologue of the transcription factor Rv0880 ( Table 1).
Taken together, these results show that IPA resistance can emerge via at least three different mechanisms (Table 2): (i) mutations in TrpE abrogating feedback inhibition, (ii) mutations likely to reduce chorismate mutase activity, thereby reducing vulnerability of the Trp pathway, and (iii) mutations in transcription factor Rv0880, which appears to be involved in general drug resistance.
IPA is active against drug-resistant clinical M. tuberculosis isolates and nontuberculous mycobacteria. The identification of the mechanism of action of IPA paves the way for a chemistry driven optimization program to generate more potent analogs. We showed that IPA is active against the drug-susceptible laboratory M. tuberculosis strain H37Rv. To ensure that anti-M. tuberculosis activity is not a laboratory strain artifact, we tested and confirmed activity of IPA against a collection of clinical M. tuberculosis isolates, including drug-resistant strains (Table S2).
In addition to M. tuberculosis, environmental mycobacteria (also called nontuberculous mycobacteria [NTM]) are increasingly causing difficult-to-cure lung disease (34). To determine whether IPA or its future analogs hold the potential for treating NTM infections in addition to tuberculosis, we determined whether IPA shows activity against these relatives of the tubercle bacillus. Growth inhibition experiments showed a spectrum of IPA susceptibilities in this group of mycobacteria, with Mycobacterium avium, Mycobacterium kansasii, and Mycobacterium chelonae showing susceptibility similar to M. tuberculosis and Mycobacterium fortuitum, Mycobacterium abscessus, and the mycobacterial model organism Mycobacterium smegmatis showing weaker inhibition by IPA (see Fig. S6 in the supplemental material). IPA showed no activity against selected Gram-positive and -negative bacteria (Fig. S6). Taken together, IPA appears to display broad-spectrum antimycobacterial-specific activity.

DISCUSSION
IPA is a host microbiota-derived metabolite that modulates mucosal immunity and exhibits neuroprotective effects (8,(35)(36)(37)(38)(39)(40). IPA is produced from the aromatic amino acid tryptophan by several species residing in the gut, including Clostridium sporogenes (8). This low-molecular-weight metabolite can readily be detected in the human bloodstream and cerebrospinal fluid (7,41). We recently discovered that IPA is active against M. tuberculosis in vitro and in a mouse model of TB infection (9). However, its bacterial target is unknown. Here, we have identified anthranilate synthase TrpE as the target of IPA in M. tuberculosis.
Since Trp is an allosteric inhibitor of mycobacterial TrpE required for Trp biosynthesis, we hypothesized that IPA, a structural analog of Trp, may exert its anti-M. tuberculosis activity by mimicking the Trp-mediated negative feedback on TrpE. Structural modeling and metabolic and chemical rescue, as well as genetic evidence, indicated that IPA indeed blocks Trp biosynthesis at the TrpE-catalyzed step and does so by binding to the allosteric pocket of Trp on TrpE. Resistance studies revealed three operating mechanisms. High-level resistance to IPA emerges by two distinct mechanisms in M. tuberculosis. Mutations in the allosteric binding site of Trp on TrpE eliminate not only negative-feedback control of Trp synthesis but also IPA inhibition of Trp synthesis. A second high-level IPA resistance mechanism involves metabolic rewiring. Missense mutations in the chorismate mutase, which utilizes the substrate of TrpE for the synthesis of other aromatic acids, appear to reduce the vulnerability of the Trp biosynthetic pathway to inhibition by small molecules. Finally, we uncovered a lowlevel IPA, likely an indirect resistance mechanism caused by mutations in the transcriptional regulator Rv0880 via an unknown mechanism.
We show that IPA is active not only against M. tuberculosis, including drug-resistant clinical isolates, but also against nontuberculous mycobacteria (NTM), such as Mycobacterium avium. Similar to multidrug-resistant tuberculosis, NTM lung disease is difficult to cure since first-line anti-TB drugs are largely inactive against common NTM species (34). Interestingly, IPA appears to display selective broad-spectrum antimycobacterial activity since it is inactive against major Gram-positive and Gram-negative pathogens, many of which harbor TrpE with a Trp allosteric binding site. The reasons behind this puzzling observation remain to be determined and could include drug uptake, efflux or metabolism (42), or differences in Trp binding motifs (Fig. S6). Although IPA did not show activity against representative nonmycobacterial microbes in our study, it is interesting to note that Mandelbaum-Shavit et al. reported antibacterial activity of the molecule against Legionella pneumophila causing pneumonia (Legionnaires' disease) (43).
In conclusion, we find that IPA blocks tryptophan biosynthesis in M. tuberculosis. Intriguingly, IPA blocks this pathway by mimicking tryptophan, the physiological allosteric inhibitor of TrpE, the enzyme catalyzing the first committed step. Thus, IPA exploits a bacterial regulatory feedback mechanism to exert its antibacterial whole-cell activity. The present results, together with IPA's proven in vivo activity, validate TrpE as a vulnerable target and reveal IPA as an attractive starting point for the discovery of a broad-spectrum antimycobacterial. The determination of TrpE as the target of IPA will enable the rational, target-based generation of a semisynthetic, next-generation IPA with improved potency.

MATERIALS AND METHODS
Docking studies. The docking studies were performed using the Autodock 4.2.6 (AD4) and AutoDock Vina 1.1.2 (Vina) software (17,18). The crystal structure of anthranilate synthase component I (TrpE) from M. tuberculosis was retrieved from the Protein Data Bank (PDB no. 5CWA) (15). The TrpE protein was prepared using the MOE2016.0802 (MOE) software (44). Issues with alternate or missing side chains, chain breaks, and terminal residues were corrected. Cocrystallized ions, water, and ligands were deleted. Hydrogens were added to the target structure using the Protonate3D module in MOE to predict the rotamers and protonation and tautomeric states of each residue at pH 7. The protein structure was saved in the PDB format. The 3D structures of IPA and L-Trp were retrieved from PubChem database in the SDF format. The structure of L-Trp was further optimized using the MMFF94x force field in MOE, and both structures were saved in the mol2 format. The protein, ligand, grid parameter, and docking parameter files were prepared using AutoDockTools 1.5.6. Kollman and Gasteiger charges were assigned to the protein and ligand residues, respectively. The grid box was set to cover most of the protein spaces, including the chorismate and Trp binding sites, with dimensions of 48.6 by 48.6 by 44.8 Å. For AD4, the docking was performed by setting the number of evaluations and number of Lamarckian genetic algorithm runs to 2,500,000 and 20, respectively. For Vina, the docking was performed by setting both the exhaustiveness and maximum number of binding modes to 20. The simulation was repeated 5 times for each ligand using AD4 or Vina. After the docking was completed, the output files were converted to SDF and TXT files and were analyzed using MOE. were obtained from the American Type Culture Collection. Mycobacterium species were grown in Middlebrook 7H9 (Becton Dickinson, Difco) complete broth medium supplemented with 0.05% Tween 80, 0.2% glycerol, and 10% albumin-dextrose-catalase (ADC) at 37°C. All other bacteria were grown in LB broth at 37°C with 200-rpm shaking. Clinical M. tuberculosis isolates (cMtb_1, cMtb_2, cMtb_3, and cMtb_4) were provided by the strain collection of the BSL3 core facility of the National University of Singapore. All experiments involving M. tuberculosis were performed in the BSL3 core facility of the National University of Singapore.
Quantification of Trp metabolite. To extract the intrabacterial Trp, we followed a previously published protocol (45). Briefly, exponentially grown M. bovis BCG cultures were adjusted to a final optical density at 600 nm (OD 600 ) of 0.4. The cultures were treated with IPA, fluoro-anthranilate, or ethambutol with 0ϫ, 0.5ϫ, 1ϫ, and 2ϫ MICs and incubated at 37°C with 80-rpm shaking for 24 h. The IPA Inhibits TrpE ® following MICs were used under these culture conditions: IPA, 140 M; fluoro-anthranilate, 6 M; EMB, 3 M. To measure growth of samples after 24 h of treatment, we transferred 100 l of aliquots into a clear-bottom 96-well plate and measured the OD using a Tecan Infinite 200 Pro microplate reader. OD 600 measurements after 24 h showed no significant differences. We collected pellets (ϳ10 9 CFU/ml) of cultures by centrifugation at 3,200 rpm for 10 min and washing with 1ϫ phosphate-buffered saline (PBS). The pellet was resuspended with 500 l of 1ϫ PBS and transferred into lysing matrix B tubes (MP Biomedicals). The samples were subjected to bead beating (Precellys 24 homogenizer) at 6,500 rpm three times for 30 s each as described earlier (45). The lysate was pelleted by centrifugation at 13,000 rpm for 10 min, and the supernatant was carefully transferred to new tubes for Trp quantification. Trp quantification was performed using liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS [details are provided in Text S1 in the supplemental material]).
MIC determination. MIC determinations were performed in 96-well plates (Corning catalog no. 3527) as described previously (9). Briefly, an inoculum of exponentially growing M. tuberculosis was adjusted to a final density of ϳ10 6 CFU/ml in Middlebrook 7H9 (Becton Dickinson) complete broth medium. Serially diluted drug concentrations were mixed with the inoculum in a 96-well plate and incubated at 37°C with orbital shaking at 80 rpm for 7 days. Turbidity at 600 nm was measured using a Tecan Infinite 200 Pro microplate reader. Percentage of inhibition was calculated relative to the drug-free samples. Regarding non-M. tuberculosis mycobacterial species, exponentially growing M. smegmatis, M. avium, M. abscessus, M. chelonae, and M. fortuitum cells were diluted in Middlebrook 7H9 complete medium to a final inoculum density of ϳ10 6 CFU/ml and incubated with IPA in 96-well flat-bottom microtiter plates for 2, 5, 3, 4, and 4 days, respectively. Exponentially growing S. aureus, E. coli, Pseudomonas aeruginosa, and Acinetobacter baumannii cells were diluted in LB medium and incubated with IPA overnight. Ten-point 2-fold serial dilutions of IPA starting from a concentration of 1 mM were used as described previously (9).
Chemical supplementation studies. To evaluate the effect of anthranilate and Trp supplementation on the activity of IPA, we tested IPA against M. tuberculosis in 7H9 complete medium with or without supplementation with 1 mM Trp or 0.2 mM anthranilate. The assay was performed in 96-well plates as described for MIC determination. The OD measurements were complemented by an agar growth assay. M. tuberculosis culture samples were transferred from 96-well plates onto single-well 7H11 complete agar microplate (Stem Corporation, Japan) using a 96-well replicator (Stem Corporation, Japan). The plates were incubated for 10 days at 37°C prior to visual examination and photo taking.
Spontaneous resistant mutant selection. Mutant generation was performed as described previously (46). Briefly, 10 6 to 10 9 CFU from exponentially growing M. tuberculosis or M. bovis BCG cultures were plated on complete 7H11 agar containing 500 or 750 M IPA or 150 M fluoro-anthranilate. The plates were incubated for 4 weeks at 37°C. Mutant colonies were colony purified and confirmed for resistance by restreaking on agar containing the same concentration of drug used for selection. MIC determinations were performed to determine the level of resistance. To select fluoro-anthranilate cross-resistant IPA mutants, we inoculated 200 IPA-resistant colonies in 96-well plates containing 100 l of 7H9 broth medium with 150 M fluoro-anthranilate. Bacteria in wells showing growth were restreaked on 7H11 agar medium containing 150 M fluoro-anthranilate. The plates were incubated for 4 weeks at 37°C. Isolated single colonies were inoculated into 10 ml of 7H9 complete broth medium, and crossresistance was confirmed by performing MIC determinations with the wild-type strain as the control.
Targeted and whole-genome sequencing. Extraction of genomic DNA was performed as described previously using the phenol-chloroform method (45,47). To perform targeted sequencing, PCR amplification was performed using custom-designed primers (Table S1) following instructions provided by a high-fidelity PCR kit (Toyobo no. F1066K). Sanger sequencing was performed for two independent PCR products by Bio Basic Asia Pacific, Singapore. For whole-genome sequencing (WGS), library construction, whole-genome sequencing, and bioinformatics analyses were performed by NovogeneAIT Genomics, Singapore (Details of WGS and bioinformatics analysis are described in Text S1.) Following wholegenome sequencing, Sanger sequencing was performed to confirm mutations in trpE, Rv0880, and Rv0948c amplified by custom-designed primers (Table S1).
TrpE expression studies. To overexpress trpE (Rv1609), we cloned the gene into the pMV262 episomal plasmid under the control of the hsp60 constitutive promoter (26). First, the gene was amplified from the wild-type or mutant genomic DNA of M. tuberculosis/M. bovis BCG using custom primers (Table S1). The amplification was performed according to the KOD-plus-neo high-fidelity PCR kit protocol (Toyobo no. F1066K). The PCR products were purified by the QIAquick PCR purification kit (Qiagen, Hilden, Germany) and digested along with pMV262 vector using 3 l of HindIII-HF (New England Biolabs no. R3104S) for 3 h at 37°C. The digested products were purified by agarose gel electrophoresis. Prior to ligation, the vector was treated with 0.5 l of alkaline phosphatase (CLP; New England Biolabs no. M0690S) for 30 min to prevent religation. Following overnight ligation at 16 0 C, the ligation products were transformed into competent E. coli DH5␣ cells. Transformants were confirmed by colony PCR, restriction digestion, and Sanger sequencing (Bio Basic Asia Pacific, Singapore). These constructs were electroporated into M. bovis BCG strains, which were plated on kanamycin (25 g/ml)-containing 7H11 agar plates and incubated at 37°C for 3 weeks. Positive colonies were confirmed by restreaking on kanamycincontaining agar plates. Verified single colonies were inoculated into 7H9 complete broth medium, and stocks were made in 10% glycerol and stored at Ϫ80°C until use. Transformation of mutant constructs pMV262-phsp60-trpE (Ser67Ala or His170Asn) in the E. coli DH5␣ strain consistently failed. Thus, we directly electroporated the ligation products into M. bovis BCG, where transformants could be obtained. Positive clones were verified by colony PCR, restriction digestion, and Sanger sequencing (Bio Basic Asia Pacific, Singapore).

Recombinant His tag TrpE generation.
Cloning, expression, and purification of TrpE were performed as described previously (16). Briefly, the trpE gene (Rv1609) was amplified from the genomic DNA of wild-type or mutant M. tuberculosis/M. bovis BCG using custom primers (Table S1). The amplification was performed according to the KOD-plus-neo high-fidelity protocol (Toyobo no. F1066K). The PCR products were purified by QIAquick PCR purification kit (Qiagen, Hilden, Germany). The purified PCR products and pET30aϩ vector were digested using 2 l of NcoI-HF (20,000 U/ml) and HindIII-HF (20,000 U/ml) in 1ϫ CutSmart buffer (New England BioLabs no. B7204S) for 3 h. Following overnight ligation at 16°C using T4 DNA ligase, the products were transformed into the competent E. coli BL21(DE3) strain and plated on kanamycin (25 g/ml)-containing LB agar overnight at 37°C. Positive colonies were confirmed by colony PCR, restriction digestion, and Sanger sequencing (Bio Basic Asia Pacific, Singapore).
To express the recombinant protein, a single colony was inoculated in 10 ml of LB medium containing 25 g/ml kanamycin overnight at 37°C. The following day, 5 ml of overnight culture was inoculated into 1 liter LB medium containing 25 g/ml kanamycin and incubated until the OD 600 reached ϳ0.5 at 37°C. TrpE expression was induced by the addition of 0.02 mM isopropyl-␤-D-thiogalactopyranoside (IPTG) as described previously (16). The cultures were incubated for an additional 16 h at 20°C. The protein purification was performed according to the Qiagen Ni-nitrilotriacetic acid (NTA) Superflow column protocol (Qiagen, Hilden, Germany). The purified proteins were subjected to buffer exchange with 5 mM Tris-HCl (2.5% glycerol) and concentrated using an Amicon Ultra-15 filter device by centrifugation at 4,000 ϫ g for 20 min at 4°C (Millipore Corporation, Bedford, MA). The proteins were stored at Ϫ80°C until use. The molecular weight of His-TrpE was estimated by Coomassie brilliant blue R250-stained sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) with molecular weight standards (calculated mass of TrpE including the 6ϫ His tag, 56.6 kDa). In-gel digestion was performed to confirm the protein using an Applied Biosystems 4800 proteomics analyzer for matrix-assisted laser desorption ionization-time of flight tandem mass spectrometry (MALDI-TOF/TOF) (Applied Biosystems, Framingham, MA), which provides the highest score, 1,310, confirming the protein is indeed TrpE (48,49) (Fig. S3 and Text S1). This analysis was performed by the Protein and Proteomics Center, Department of Biological Sciences, National University of Singapore. The concentration of the proteins was measured following a colorimetric detection method using the bicinchoninic acid (BCA) kit (Thermo Scientific no. 23225).
Biochemical TrpE studies. To perform enzymatic assays, we used wild-type and Ser67Ala and His170Asn mutant versions of recombinant TrpE. To measure the activity of the enzymes, the rate of the anthranilate formation was monitored fluorometrically with excitation of 320 nm and emission of 460 nm using a coupled enzymatic assay. The concentration of anthranilate formed was determined from a standard curve established by the concentration of anthranilic acid versus its fluorescence intensity measured by a Tecan Infinite 200 Pro microplate reader. To determine the K m and V max of the enzymes, various concentrations of chorismate were added to the reaction mixture (20 mM Tris [pH 9], 100 mM NH 4 Cl, 10 mM MgCl 2 , 0.1 mM EDTA, 0.075 mg/ml wild-type or mutant enzymes) in 96-well plates as described previously (15,16). The K m and V max for chorsimate were calculated by nonlinear curve fitting to the Michaelis-Menten equation using GraphPad Prism 6 as described previously (15). The in vitro enzymatic inhibition assay was performed in a final volume of 200 l containing 100 mM NH 4 Cl, 10 mM MgCl 2 , 0.1 mM EDTA, 20 mM Tris (pH 9), 0.075 mg/ml wild-type and mutant TrpE proteins, and 50 M chorismate in black 96-well plates as described previously (15). Trp and F-Trp were dissolved in water at a concentration of 10 mM and tested in the assay with concentrations ranging from 0.1 to 100 M. Prior to the addition of the substrate chorismate, the mixture was incubated for 10 min at 25°C. The IC 50 values were calculated as described previously (50). All assays were performed in triplicate.
Drug susceptibility testing using Bactec MGIT 960. To determine the activity of IPA against multidrug-resistant M. tuberculosis strains, the Bactec MGIT 960 system (Becton Dickinson, Sparks, MD) was used according to the manufacturer's instructions and the MGIT manual by FIND (Geneva, Switzerland). We used four clinical M. tuberculosis isolates from the collection of the National University of Singapore BSL3 core facility and the laboratory strain H37Rv. Susceptibility testing for streptomycin (SM