Novel Antimicrobials from Uncultured Bacteria Acting against Mycobacterium tuberculosis

Decreasing discovery rates and increasing resistance have underscored the need for novel therapeutic options to treat Mycobacterium tuberculosis infection. Here, we screen extracts from previously uncultured soil microbes for specific activity against Mycobacterium tuberculosis, identifying three novel compounds. We further define the mechanism of action of one compound, amycobactin, and demonstrate that it inhibits protein secretion through the Sec translocation machinery.

their derivatives (4,5). However, the inability to culture most bacteria under laboratory conditions, as well as the continual rediscovery of known compounds, led to diminishing returns, and natural product screening efforts were largely abandoned by the 1960s (6). Improved methods of growing previously uncultured bacteria now provide access to untapped biological and chemical diversity. A previously uncultured bacterium, Eleftheria terrae, produces teixobactin, a novel cell wall-acting inhibitor (7). We also introduced a selective screening approach that is aimed at eliminating the large background of toxic and known compounds. Screening of extracts from uncultured bacteria against M. tuberculosis, and counterscreening against Staphylococcus aureus, led to the discovery of lassomycin, produced by a Lentzea sp., which targets the C1 subunit of the essential P1P2C1 protease of mycobacteria (8).
Here, we report the identification of three novel antimicrobials from previously uncultured bacteria with selective activity against M. tuberculosis. Streptomycobactin is a cationic depsipeptide, kitamycobactin is a lasso peptide that requires native C1 for activity, and amycobactin is a novel antimicrobial that targets the SecY protein of the mycobacterial secretion system.

RESULTS
Identification of M. tuberculosis selective compounds. Extracts prepared from the fermentation broth of 10,241 previously uncultured bacteria were screened for the ability to inhibit the growth of M. tuberculosis and S. aureus. Extracts with activity against M. tuberculosis but not S. aureus were selected for follow-up studies. From this screen, four extracts were selected based on reproducibility of activity. The extracts were fractionated by high-performance liquid chromatography (HPLC) until a single active fraction was identified by bioactivity-guided purification. The structures of the four compounds were determined by nuclear magnetic resonance (NMR) analysis. Three compounds were determined to be novel, since they did not have matches in AntiBase or SciFinder. The fourth was determined to be the previously identified compound marfomycin D, purified from a deep-sea Streptomyces species. However, its activity against mycobacteria had not been reported (9). The target of marfomycin D is unknown. We named the novel compounds amycobactin, kitamycobactin, and streptomycobactin (Fig. 1). The compounds were named by combining the name of the genus of the producing species with the mycobactin suffix to indicate activity against mycobacteria. The structures of amycobactin (Table S1, Fig. S1, Fig. S2, and Text S1) and kitamycobactin (Table S2, Fig. S3, and Text S2) were assigned by 1 H, 13 C, correlation spectroscopy (COSY), total correlation spectroscopy (TOCSY), 1 H-13 C/ 15 N heteronuclear single quantum coherence (HSQC), 1 H-13 C heteronuclear multiple-bond correlation (HMBC), and nuclear Overhauser effect spectroscopy (NOESY)/rotating-frame nuclear Overhauser effect spectroscopy (ROESY) experiments. For elucidation of the structure of streptomycobactin, additional triple-resonance experiments on a universally 13 C-and 15 N-labeled sample were used. 13 C chemical shifts of the peptidyl backbone were mapped using HNCACB and HNCOCACB experiments, while the side-chain 13 C and 1 H chemical shifts were assigned using (H)CCCONH and H(CCCO)NH experiments, respectively (Table S3, Fig. S4, and Text S3).
Amycobactin is produced by an Amycolatopsis sp., has an exact mass of 762 Da, and is the smallest of the compounds identified. It has an unusual structure, featuring a ketal moiety within a macrolactone backbone. Kitamycobactin and streptomycobactin are semicyclic peptides with exact masses of 2,259 and 1,735 Da, respectively. Streptomycobactin is a 20-amino-acid peptide (VIITVLLVRAVLATVRIERV) produced by a Streptomyces sp. Kitamycobactin, produced by a Kitasatospora sp., is a lasso peptide (GFGRIKADELVGRLIP) and belongs to the same class as lassomycin. Antimicrobials have been identified from these genera previously. However, our in situ cultivation method utilizing the iChip increases recovery 50-fold (10). This means that, using our methods, the probability that an isolate represents a conventional cultivable species is 1/50; for the three isolates described here, the probability is 8 ϫ 10 Ϫ6 . This makes it highly unlikely that the producing species have been cultivated previously.
Compound bioactivity. The compounds were tested for inhibition of the growth of M. tuberculosis, as well as inhibition of the growth of other mycobacterial pathogens and some gut symbionts (Table 1). Rifampin was used as a positive control in all MIC testing against M. tuberculosis H37Rv (MIC, 0.5 g/ml) and mc 2 6020 (MIC, 0.125 g/ml). We included susceptibility data for marfomycin D as well, since this compound had not been tested against mycobacteria (9). Streptomycobactin and kitamycobactin were the most potent compounds, with MICs of 0.03 g/ml and 0.06 g/ml, respectively. While efforts to identify new antimycobacterial compounds focus mainly on M. tuberculosis, there has been increasing interest recently in nontuberculous mycobacteria (NTM). The NTM family of organisms encompasses Ͼ150 different species of opportunistic patho-  gens that cause various diseases ranging from pulmonary infections similar to tuberculosis to soft-tissue and skin infections (11). Increasing rates of drug resistance have led to a need for new therapeutic options to treat NTM infections. With this in mind, we tested the compounds against the major NTM pathogens: Mycobacterium avium, Mycobacterium abscessus, and Mycobacterium paratuberculosis. Clarithromycin was used as a positive control, with MICs of 0.5 g/ml, 0.25 g/ml, and 0.25 to 0.125 g/ml against M. avium, M. abscessus, and M. paratuberculosis, respectively. Overall, the compounds proved to be effective against the NTM species at a level similar to their effectiveness against M. tuberculosis (Table 1). The detrimental effects of antibiotic treatment on the gut microbiome are a well-recognized phenomenon (12). With this in mind, we tested all the compounds against the gut commensal bacteria Bacteroides fragilis and Lactobacillus reuteri (Table 1). The limited activity of the compounds against commensal bacteria confirms their selectivity against M. tuberculosis. We also determined cytotoxicity against the liver cell line HepG2 and the mouse embryonic fibroblast cell line NIH/3T3. Kitamycobactin had no activity at the highest concentration tested, 100 g/ml. Amycobactin and streptomycobactin both had 50% toxic concentrations (TC 50 ) of 16 to 32 g/ml against HepG2 and NIH/3T3 cells. Given the streptomycobactin MIC of 0.03 g/ml against M. tuberculosis, this creates a significant therapeutic window. However, the therapeutic window for amycobactin is considerably smaller, given its MIC of 4 to 8 g/ml against M. tuberculosis. Overall, the compounds displayed good selectivity against M. tuberculosis and mycobacteria, with limited cytotoxicity.
The killing kinetics were determined for all four compounds against M. tuberculosis strain mc 2 6020 in both the exponential-and stationary-growth phases. Amycobactin was bacteriostatic in exponential phase ( Fig. 2a) but, interestingly, was moderately bactericidal (Ͻ1-log killing) in stationary phase (Fig. 2b). Streptomycobactin and kitamycobactin were both bactericidal to exponentially growing cultures (Fig. 2a), while kitamycobactin displayed bactericidal activity against stationary-phase cultures (Fig. 2b). Marfomycin D was bacteriostatic against exponential-phase cultures until day 7, after which efficacy was lost and the cultures began to grow (Fig. 2a). The ability of M. tuberculosis to enter nonreplicating or slowly replicating states, such as stationary phase, is considered a primary reason underlying treatment failure (13,14). Compounds with effectiveness against M. tuberculosis in these nonreplicating or slowly replicating states, such as amycobactin and kitamycobactin, are essential to reducing treatment duration and increasing treatment success.
Target identification. We next sought to generate mutants in order to identify the targets of the compounds. Using single-step or sequential selection, we were unable to generate mutants against streptomycobactin. However, given the sensitivity of M. tuberculosis to this compound and the large therapeutic window (Table 1), continued efforts to understand the mechanism of action are warranted. As mentioned above, the general architecture of kitamycobactin resembles that of the M. tuberculosis-selective compound lassomycin (8). We therefore tested a ClpC1 Q17R mutant resistant to lassomycin (8) and found that it was resistant to kitamycobactin as well. Kitamycobactin had a MIC of 4 g/ml against the ClpC1 Q17R mutant, while its MIC against the parental strain, mc 2 6020, is 0.06 g/ml (Table 1). This suggests that kitamycobactin acts against ClpP1P2C1, the essential protease of mycobacteria. Action against this target explains the selectivity of kitamycobactin against mycobacteria. The independent evolution of two compounds, lassomycin and kitamycobactin, that interfere with the same essential target is noteworthy. This would suggest that the ClpP1P2C1 protease is an attractive target for the inhibition of mycobacteria by soil microbes.
Attempts to obtain mutants of M. tuberculosis resistant to amycobactin failed, but we were able to generate two independent mutants (designated N28R1 and N28R2) in the related species Mycobacterium smegmatis. The amycobactin MIC against wild-type (WT) M. smegmatis was 1.5 g/ml, while the MICs against N28R1 and N28R2 were 12.5 g/ml and Ͼ100 g/ml, respectively. Sequencing revealed that both mutants contained in-frame deletions in the general protein secretion translocase SecY (MSMEG1483), which, together with SecE and SecA, forms the core of the Sec protein translocation machinery (15,16). Sec secretion is essential in all bacteria and is responsible for the translocation of a majority of secreted and integral membrane proteins (17,18). N28R1 contained an 18-bp deletion in the secY gene, resulting in the deletion of amino acids 222 to 227 (VIAALV). N28R2 contained a 9-bp deletion in secY, resulting in the deletion of amino acids 407 to 409 (FGG) (Fig. 3a). Given the novelty of SecY as a putative target, we pursued the mechanism of action further.
Deletions corresponding to those in N28R1 and N28R2 were made in a clean background of WT M. smegmatis using site-directed recombineering (19,20). The deletions were confirmed by sequencing, and the mutants were renamed Msm⌬secY 6AA and Msm⌬secY 3AA. The MIC for Msm⌬secY 6AA was 12.5 g/ml, and the MIC for Msm⌬secY 3AA was Ͼ100 g/ml, confirming that these deletions confer resistance to amycobactin. Both mutants grew as well as the wild type, indicating no apparent fitness cost of the secY deletions (Fig. 3b). We next sought to re-create these mutations in M. tuberculosis in order to confirm a conserved target in both species. The sequences of the SecY proteins in M. smegmatis and M. tuberculosis are very similar, including both the 3-and 6-amino-acid deletion regions. Site-directed mutagenesis targeting M. tuberculosis secY (Rv0732) was used to re-create these deletions in M. tuberculosis. The deletion of amino acids 407 to 409 was successful; it was confirmed by sequencing, and the mutant was named Mtb⌬secY 3AA. We were unsuccessful in deleting amino acids 222 to 227 in M. tuberculosis despite several attempts. Mtb⌬secY 3AA grew similarly to WT M. tuberculosis (Fig. 3c) and had a MIC of Ͼ100 g/ml. Such large deletions are rare and may impose significant restrictions on protein function. Phyre2 (21) was used to predict the structure of M. tuberculosis SecY with 100% confidence to the highestscoring template (PDB ID 2ZQP). The 6-and 3-amino-acid deletions were mapped to the predicted structure and are shown in red in Fig. 3d and e. The secretion of peptides through SecY in Escherichia coli is restricted by a hydrophobic core ring and a channel plug requiring ATP hydrolysis via SecA for secretion (22). The predicted channel plug is shown in orange in Fig. 3d and e. The lateral gate of SecY, formed by helices 2 and 7 (23), is shown in blue. The SecY mutations conferring resistance to amycobactin are near the channel plug and lie within the hydrophobic core ring (Fig. 3d and e). This  We next sought to determine whether amycobactin inhibited protein secretion by the Sec translocase. For this purpose, we turned to a strain of M. smegmatis engineered to secrete the E. coli =BlaTEM-1 ␤-lactamase via the Sec translocation machinery (24). This strain was constructed by the fusion of E. coli =BlaTEM-1 to the peptide secretion signal from MPT63, a protein known to be secreted through Sec. Additionally, the endogenous copy of ␤-lactamase in M. smegmatis was deleted (24). By using this strain, we can easily monitor protein secretion through the Sec translocase by monitoring the activity of the exported ␤-lactamase. A microtiter assay was used to determine the amount of ␤-lactamase in the culture filtrate (CF) and whole-cell lysate (WCL) by monitoring the hydrolysis of the chromogenic substrate nitrocefin over time. The cleavage of nitrocefin by ␤-lactamase results in a shift in the absorbance of nitrocefin from 490 nm to 390 nm (25,26). Here, we monitored the hydrolysis of nitrocefin by the ratio of the absorbance at 490 nm to the absorbance at 390 nm. As shown in Fig. 4a, treatment of the reporter strain with amycobactin resulted in a lower level of nitrocefin hydrolysis in the culture filtrate of the amycobactin-treated sample than in the untreated control, indicating a decrease in the amount of ␤-lactamase present. Additionally, the maximum rate of the change in absorbance at 490 nm (V max490 ) was significantly greater in the untreated sample than in the amycobactin-treated sample CF (Fig. 4b), indicating a decrease in the amount of ␤-lactamase present after amycobactin treatment. Importantly, amycobactin treatment only moderately decreased the hydrolysis of nitrocefin in the WCL (Fig. 4a).
To confirm that amycobactin treatment inhibits the secretion of ␤-lactamase by the reporter strain, we used Western blotting to directly quantify the amounts of ␤-lactamase in the CF and WCL. Treatment of the reporter strain with amycobactin resulted in a significant decrease in the amount of ␤-lactamase in the CF from that in untreated cells (Fig. 4c and d). Importantly, the amount of ␤-lactamase in the WCL was unchanged after treatment (Fig. 4c). Taken together, these findings confirm that amycobactin acts to inhibit protein secretion through the Sec translocation machinery.

DISCUSSION
Recent work has reiterated the power of natural product discovery to deliver promising new antimicrobials (27)(28)(29). Here, we turned our attention to previously uncultured microbes as a novel source of antimycobacterials. Coupling this method with differential screening for compounds acting selectively against M. tuberculosis, we identified three new antimicrobials with activity against mycobacteria and identified the targets for two of them, amycobactin and kitamycobactin.
Amycobactin directly targets protein secretion by the Sec translocation machinery and, to our knowledge, represents the first natural product targeting the protein secretion machinery. Amycobactin is bacteriostatic against exponentially growing M. tuberculosis but bactericidal against stationary-phase M. tuberculosis, which is puzzling, since antibiotics are generally more effective against actively growing cells. While amycobactin likely blocks the secretion of vital proteins in both states, its bactericidal activity in stationary phase may be due to the phenomenon of Sec translocon "jamming" (30). When secretion through the Sec translocation machinery fails, the SecY translocase becomes "jammed" with a linear peptide. The cell's response is to target the entire apparatus for degradation by cellular proteases. This process itself can be fatal to the cell (30). Exponentially growing cells can easily replace the Sec machinery that was degraded. However, the decreased metabolic state of stationary-phase cells may limit the production of new Sec secretion components, resulting in killing by amycobactin.
The essential components of the Sec translocation machinery in M. tuberculosis and M. smegmatis are SecY, SecE, and SecA (17). SecE forms a "clamp" near the lateral gate of SecY to stabilize the structure (22). SecA provides the recognition of secreted proteins as well as the energy for secretion by hydrolyzing ATP (31). Amycobactin may act to disrupt the interaction of SecY with SecE or SecA, inhibiting secretion. Alternatively, amycobactin may act directly on SecY. Secretion through SecY is restricted by a hydrophobic core ring and a channel plug; the latter is displaced during secretion (22). Amycobactin may act to restrict the displacement of the channel plug in WT SecY, resulting in blocked secretion. Deletion of amino acids 407 to 409 may alleviate this blockage by providing more flexibility in the channel plug. The high MIC and narrow therapeutic window of amycobactin suggest that this is an early lead that could be further optimized. Its inhibition of protein secretion also makes it an attractive tool for study of the Sec translocation machinery. The essentiality of the Sec secretion machinery makes studying its contribution to virulence challenging. Considering that several virulence-associated proteins are secreted via the Sec machinery (17,32), a tool for studying this aspect of M. tuberculosis physiology would be invaluable. Taking these considerations together, this work underscores the power of natural product discovery to deliver promising new antibiotics with novel mechanisms of action against M. tuberculosis.

MATERIALS AND METHODS
Bacterial strains and growth conditions. The M. tuberculosis strains used were H37Rv and H37Rv mc 2 6020 where noted. M. smegmatis strain mc 2 155 was used for all M. smegmatis work. M. avium strain ATCC 700898, M. abscessus strain ATCC 19977, and M. paratuberculosis strain ATCC 43544 were used for antibiotic susceptibility. All mycobacterial strains were grown in Difco 7H9 medium supplemented with 10% oleic acid-albumin-dextrose-catalase (OADC) and 5% glycerol. Lysine (80 g/ml) and pantothenate (24 g/ml) were added to mc 2 6020 medium. Tyloxapol was added to liquid cultures at a final concentration of 0.05%. Mycobactin J (Allied Monitor) was added to M. paratuberculosis cultures at a final concentration of 10 mg/liter. Bacteroides fragilis ATCC 25282D-5 and Lactobacillus reuteri ATCC 23272 were grown in brain heart infusion broth supplemented (per liter) with 5 g yeast extract, 10 ml 10% (wt/vol) L-cysteine HCl, 15 mg/liter hemin, and 66 ml 1.5 M 3-(N-morpholino)propanesulfonic acid [MOPS; pH 7.0] in an anaerobic chamber. Kanamycin was used at a final concentration of 50 mg/liter where appropriate.
Isolation and cultivation of soil bacteria. Soil samples (1 g) were vigorously agitated in 10 ml of deionized water for 10 min and were then allowed to sit for 10 min to allow large soil particles to settle. The supernatant was diluted into molten SMS medium ((0.125 g casein, 0.1 g potato starch, 1 g Casamino Acids, 20 g Bacto agar in 1 liter of water). Aliquots were then dispensed into the wells of 96-well microtiter plates or iChips (10). The microtiter plates were incubated at room temperature in humidified chambers for as long as 12 weeks, and the appearance of colonies was monitored weekly. At weekly intervals starting after 4 weeks, colonies were picked onto SMS medium. The iChips were placed in direct contact with the soil. After 4 weeks of incubation, the iChips were disassembled and the colonies picked onto SMS medium in order to test for the ability to propagate outside the iChip and to purify colonies. Glycerol stocks (15% glycerol) were made for isolates with robust growth on SMS medium.
Extract preparation and screening. Isolates from the NovoBiotic collection were transferred to seed broth (15 g glucose, 10 g malt extract, 10 g glycerol, 2.5 g yeast extract, 5 g Casamino Acids, and 0.2 g calcium carbonate chips·2H 2 O per liter of deionized H 2 O [pH 7.0]) and incubated with vigorous agitation at 28°C for 5 to 12 days until the culture was turbid (by visual inspection). The time to turbidity depends on the isolate. The seed culture was then diluted 1:20 into three different fermentation media. After 11 days of growth at 28°C with agitation, the cultures were dried down. Dimethyl sulfoxide (DMSO) was added to the dried biomass and mixed, and the crude extracts were tested for activity against S. aureus. A 5-l aliquot of each extract was applied to a lawn of S. aureus growing on Mueller-Hinton agar. After overnight incubation at 37°C, the presence of a clearing zone indicated hit activity.
The extracts were tested against M. tuberculosis by transferring 1.5 l of extracts to the wells of a 96-well microtiter plate. A culture of M. tuberculosis mc 2 6020 expressing mCherry was grown in Middlebrook 7H9 broth at 37°C with agitation to an optical density at 600 nm (OD 600 ) of 0.4 to 0.5. The culture was diluted to an OD 600 of 0.003 and 148.5 l of culture added to the wells (1:100 dilution). After 7 days of incubation with agitation at 37°C, growth was monitored by measuring the OD 600 and by measuring fluorescence with excitation at 580 nm and emission at 610 nm. Extracts that demonstrated Յ25% growth relative to that of the DMSO positive-growth controls were cherry picked and retested under the same assay conditions. These extracts were also tested at a 1:100 dilution in a broth microdilution assay against S. aureus. Extracts without any effect on the growth of S. aureus that repeated the inhibition of the growth of M. tuberculosis were considered confirmed hits.
Three strains were chosen for additional fermentation to provide material. IS019924, IS019923, and IS008612 are the producer strains for amycobactin, streptomycobactin, and kitamycobactin, respectively. Grown biomass of each isolate was inoculated into 50 ml seed broth and was grown with shaking at 200 rpm for 3 to 8 days. The cultures were monitored visually daily for robust growth. A 2% inoculum of the seed culture was transferred into 500 ml fermentation medium in 2-liter, 6-baffle flasks at 28°C, shaken at 200 rpm, and harvested after 7 days. IS019924 was fermented in BPM fermentation medium [20 g glucose, 10 g organic soy flour (Bob's Red Mill organic whole ground soy flour), 10 g Pharmamedia (Traders protein), 1 g (NH 4 ) 2 SO 4 , 10 g CaCO 3 , and 20 g glycerol in 1 liter] and required a total of 40 liters of production for the isolation of sufficient material to characterize amycobactin. IS019923 and IS008612 were fermented in R4 fermentation medium (described previously by Ling et al. [7]) and required 40 liters for the isolation of sufficient material.

ACKNOWLEDGMENTS
This work was supported by Gates Foundation grant OPP1132809 to D.H. and K.L., NIH grant AI118058 to L.L.L., and NIH grant P01AI118687 to K.L. The