Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
ACS Chem Biol. Author manuscript; available in PMC 2013 August 17.
Published in final edited form as:
PMCID: PMC3560293

Identification of novel inhibitors of M. tuberculosis growth using whole cell based high-throughput screening


Despite the urgent need for new antitubercular drugs, few are on the horizon. To combat the problem of emerging drug resistance, structurally unique chemical entities that inhibit new targets will be required. Here we describe our investigations using whole cell screening of a diverse collection of small molecules as a methodology for identifying novel inhibitors that target new pathways for Mycobacterium tuberculosis drug discovery. We find that conducting primary screens using model mycobacterial species may limit the potential for identifying new inhibitors with efficacy against M. tuberculosis. In addition, we confirm the importance of developing in vitro assay conditions that are reflective of in vivo biology for maximizing the proportion of hits from whole cell screening that are likely to have activity in vivo. Finally, we describe the identification and characterization of two novel inhibitors that target steps in M. tuberculosis cell wall biosynthesis. The first is a novel benzimidazole that targets mycobacterial membrane protein large 3 (MmpL3), a proposed transporter for cell wall mycolic acids. The second is a nitro-triazole that inhibits decaprenylphosphoryl-β-d-ribose 2′-epimerase (DprE1), an epimerase required for cell wall biosynthesis. These proteins are both among the small number of new targets that have been identified by forward chemical genetics using resistance generation coupled with genome sequencing. This suggests that methodologies currently employed for screening and target identification may lead to a bias in target discovery, and that alternative methods should be explored.


The alarming rise in resistance to drugs currently used to treat tuberculosis has spurred a significant increase in efforts geared towards identifying new drugs effective against Mycobacterium tuberculosis. Currently there are only six new compounds in the clinical development pipeline that were specifically developed for treating tuberculosis (1). Of these potential drugs, only two function by inhibiting a completely novel target (2). Two of the compounds are inhibitors of ribosomal function, and two have complex mechanisms with multiple targets (2). In order to meet future therapeutic needs, the discovery of new chemotypes, and the chemical and biological validation of new targets are required. Although several promising inhibitors have been identified that encompass both novel chemical scaffolds and novel targets, more work is needed to ensure that a sufficient number of new drugs will be available to combat the continual emergence of drug resistance.

There is as yet no clear consensus on the best methods for identifying novel inhibitors of M. tuberculosis that will be effective in vivo. Approaches range from target based discovery to whole cell screening efforts against M. tuberculosis or related mycobacteria (3). Target based approaches rely on knowledge of protein essentiality and function, either through direct studies in mycobacterial species, or through homology to proteins in other species. One challenge with target-based approaches is that knowledge of protein essentiality is not sufficient to guarantee “druggability” (4). In addition, compounds identified in target based screens often lack activity against whole cells. Whole cell screening approaches circumvent these issues, but also have inherent challenges. Screening directly against M. tuberculosis is extremely challenging as a BSL3 level containment facility equipped with high-throughput screening equipment is required, and procedures required for handling this organism are laborious. Many groups have therefore relied upon the use of model mycobacterial species such as Mycobacterium smegmatis. However, it is unclear to what degree these organisms can be relied upon as adequate surrogates for M. tuberculosis. In addition, despite the fact that it is becoming clear that the conditions used for culturing M. tuberculosis during whole cell screening are critical to identifying inhibitors that will be effective in vivo (5), there is no clear consensus on the conditions that best reflect in vivo biology. Finally, the elucidation of the target of a promising small molecule identified in a whole cell screen can be extremely challenging.

Herein we report our experience in whole cell screening using a library of ~20,000 diverse small molecules from the Broad Institute collection against M. tuberculosis and the related mycobacterial species M. smegmatis and Mycobacterium bovis BCG. We found a surprisingly small overlap in the small molecules that are effective against these three species, highlighting the importance for directly screening against M. tuberculosis. In addition, we describe the characterization of 91 hits derived from this screen. We also describe testing of 1113 possible leads obtained from a 341,808 compound screen conducted by the NIAID and Southern Research Institute (68). From these combined screening efforts we identified novel inhibitors that target decaprenylphosphoryl-β-d-ribose 2′-epimerase (DprE1), an enzyme required for cell wall biosynthesis, and mycobacterial membrane protein large 3 (MmpL3), a proposed exporter of newly synthesized cell wall mycolic acids. Interestingly, these two proteins now appear to be the targets of several chemically unrelated small molecules. Notably, these targets were all identified using a similar method of resistance generation and whole genome sequencing of mutants (911). This repeated identification of the same targets by the same methodology, albeit using multiple structurally unrelated inhibitors, may imply either that these targets are highly “druggable” and/or the prioritization of molecules whose targets are identified by resistance generation leads to a bias in identified targets.


Screening to identify M. tuberculosis growth inhibitors

To identify compounds with activity against M. tuberculosis strain H37Rv, we employed a high-throughput screen using constitutive GFP fluorescence as a reporter of bacterial number (12) to screen a total of 20,000 compounds selected from the larger screening collection of the Broad Institute for maximal chemical diversity, and consisting of commercially available compounds, natural products, compounds generated by Diversity Oriented Synthesis (13) and other synthetic compounds. The complete data from the screen has been deposited in the publicly accessible database Chembank ( Cells were seeded into 384 well plates at a low density, and were incubated in the presence of ~25μM small molecules for a total of three days at which time GFP fluorescence was used as a readout of bacterial growth (Z′ factor = 0.8 for rifampicin and Z′ factor = 0.7 for clofazimine, coefficient of variation = 0.05). Dose responses of the known anti-tubercular drugs clofazimine and rifampicin were used as positive controls for selection of a z-score cutoff to define hits from the screen (see Materials and Methods). Although the absolute magnitude of the z-scores obtained from day to day using this assay varied, the relative performance of a given compound from day to day was highly reproducible. Independent screening of the same subset of 4608 compounds on different days showed extremely highly correlated z-scores with R2 = 0.95 (Figure 1a). Using a z-score cutoff of −4, we identified 91 hits that were cherry-picked for further analysis.

Figure 1
Screening validation and cross-species comparison of hits identified in primary screens against M. tuberculosis, M. smegmatis, and M. bovis BCG

The 91 hits obtained from the primary screen were cherry picked and retested in a dose response assay. 79 of the 91 compounds (87%) retested at concentrations at or below the screening concentration (Table S1, Table S2). 45 compounds retested with an IC90 of 10μM or less, and 6 compounds retested with an IC90 of less than 1μM. The active hits were subjected to a number of secondary screens to assess mammalian cell toxicity and activity against M. tuberculosis in infected macrophages (Table S1, Table S2). The majority of the hit compounds are unique structures that have not been reported to have antimicrobial activity. We thus sought to further characterize these small molecules to assess their potential for development as tool compounds for the identification and validation of novel targets for drug development.

Cross species comparison of growth inhibitors

To determine whether the fast-growing mycobacterial species M. smegmatis could be used in subsequent follow-up studies, the active hits identified against M. tuberculosis were tested for activity against M. smegmatis. Approximately 20% of these hits exhibited activity against M. smegmatis at the highest concentration tested against M. tuberculosis using a low stringency cutoff of a z-score < −1.5 as criteria for activity against M. smegmatis. That relatively few of the hits from our M. tuberculosis primary screen had activity against M. smegmatis suggests that the spectrum of compounds that kill M. tuberculosis and M. smegmatis are potentially quite different, either because of differences in essential targets, in target affinity for an inhibitor, or in cell entry. This is surprising, as M. smegmatis whole cell screening was recently used to identify a novel inhibitor of mycobacterial ATP synthase with activity against both M. smegmatis and M. tuberculosis (2). However, a recent study comparing a set of 5000 compounds screened across M. smegmatis, M. bovis BCG, and M. tuberculosis found that only 50% of the compounds that kill M. tuberculosis were detected using M. smegmatis (14). A significant proportion of the compounds screened in this study were known pharmacologically active compounds, which could create a potential bias. In our study comparing compounds that were rigorously selected for good activity against M. tuberculosis, we found an even smaller degree of overlap in the compounds that are effective against both organisms. This finding has potential implications for drug discovery efforts using M. smegmatis as a surrogate organism for whole cell screening. To extend this analysis to a larger number of compounds with less likelihood of inherent structural bias, 11,147 compounds from the primary screen against M. tuberculosis were tested in a large-scale primary screen against M. smegmatis (Z′ factor = 0.8 using hygromycin). These compounds were randomly selected from the full set used against M. tuberculosis in the primary screen. Consistent with the previous observation that few of the 91 primary M. tuberculosis hits had M. smegmatis activity, this unbiased screen revealed little overlap in the compounds with activity against M. tuberculosis and those with activity against M. smegmatis (Figure 1b) using composite z-scores from the primary screens as the basis for comparison (R2 = 0.0001). The majority of active compounds have species-specific activity.

The surprising lack of concordance in activity of small molecules observed against M. tuberculosis and M. smegmatis led us to test the degree of overlapping hits when M. bovis BCG is used as the screening organism. M. bovis BCG is the M. tuberculosis vaccine strain that is often used as a surrogate organism for biological studies due to its high degree of relatedness but relative lack of pathogenicity. To this end, 13,715 compounds used in the primary screen against M. tuberculosis were randomly selected and were screened in a GFP based screen using M. bovis BCG (Z′ factor = 0.8 using hygromycin). Comparison of the hits from the primary screen against M. tuberculosis and M. bovis BCG by composite z-score (Figure 1c) revealed a higher degree of correlation (R2= 0.24) than observed in the comparison with M. smegmatis. However, the majority of the hits from the individual primary screens still appear to be specific for either M. tuberculosis or M. bovis BCG. These results are surprising as most effective anti-tubercular drugs are effective against all three mycobacterial species, although MICs range for these organisms. It is noteworthy that our studies revealed less overlap between mycobacterial species than that identified in the previous study, perhaps due to the larger and less biased composition of the compound collection that we used for cross-species comparison. While the previous study of ~5000 compounds a significant proportion of compounds were known bioactives with the inherent potential for structural bias, our larger compound collection consists of uncharacterized structurally diverse and unbiased small molecules. Finally, while it is possible that performing the primary screens with a higher average concentration of inhibitor would result in enhanced correlation of hits, these results suggest a note of caution in the use of model mycobacterial species for conducting primary screens to identify novel drug-like molecules that are effective against M. tuberculosis.

Characterization of gliotoxin analogs

One of the most potent inhibitors identified in our screen against M. tuberculosis was an acetylated analog of the known fungal toxin gliotoxin (Table S1, Table S2, compound 3). Gliotoxin is an epidithiodiketopiperazine toxin characterized by a bridged disulfide ring (Figure 2a) (15, 16). Gliotoxin is known to have toxicity against a variety of cell types from a wide range of species, including mycobacteria (17). The toxic effects of gliotoxin on mammalian cells are dependent on the presence of the strained disulfide bond that reacts with cysteine residues on proteins, resulting in protein inactivation (18). We determined that gliotoxin has an IC90 of 0.14μM against M. tuberculosis. Gliotoxin is therefore extremely active against M. tuberculosis, with an IC90 that is comparable to drugs currently used for treating tuberculosis. However, as expected, we found that gliotoxin was toxic to mammalian cells, with an IC90 of 8.2μM. Interestingly, among bacterial species, M. tuberculosis appears particularly sensitive to gliotoxin, as we found that the IC90 of gliotoxin against other bacterial species including M. smegmatis, Pseudomonas aeruginosa, Staphylococcus aureus, and Enterococcus faecalis were >25μM (data not shown). We were unable to generate mutants of M. tuberculosis that were resistant to gliotoxin, indirectly supporting the hypothesis that gliotoxin may have a non-specific mechanism of action. To determine whether the disulfide bond is required for gliotoxin activity against M. tuberculosis we tested an analog of gliotoxin where formation of the disulfide bond is prevented by methylation, resulting in dimethyl thioethers at the 1 and 4 positions (Figure 2b). This gliotoxin analog was significantly less potent than gliotoxin itself, with an IC90 of > 50μM. While the disulfide bond may serve as a structural scaffolding requirement for specific target inhibition, the reducing potential of the disulfide bond in formation of mixed disulfides with M. tuberculosis proteins may also account for gliotoxin activity against M. tuberculosis.

Figure 2
Structures of gliotoxin analogs

Strategies for target identification

One of the major hurdles of drug development is the identification of the target of small molecules identified in whole cell screens. Recent successes in the field of anti-tubercular therapeutics (2, 9, 11) utilize resistance generation coupled with next generation sequencing for target identification. Using this approach to focus on compounds with a high likelihood of a specific and novel target in M. tuberculosis, we eliminated compounds that had significant toxicity against mammalian cells, that had structures suggestive of detergents or ionophores, or that were structurally related to known bioactive agents. We selected a set of 25 compounds to generate resistant mutants and were successful for three structural classes of compounds described below.

Identification of glycerol dependent growth inhibitors

Compound A039 (1-(4-ethoxyphenyl)-2-(4-methoxyphenyl)-6,7,8,9-tetrahydro-5H-imidazo[1,2-a]azepin-1-ium chloride) (Figure 3a) is a highly active bactericidal substituted imidazole with an IC90 of 1.5μM and limited toxicity against mammalian cells (Table S1, Table S2, compound 7). High-level resistance (> 25x the MIC) was observed in mutants generated to compound A039 (Figure 3b). Four mutants derived from resistance generated in parallel using four independent parent clones were subjected to whole genome sequencing using Illumina based technology. In each of the four mutants analyzed, we identified a frame shift-mutation in the glpK gene resulting from a single base-pair insertion at codon 191 (191GGT > 191GGGT) in each of the four mutants. Frame shift mutations in glpK have been previously associated with resistance to inhibitors of M. tuberculosis replication identified in whole cell screens, and it has been demonstrated that the activity of these compounds is linked to metabolism of glycerol contained in standard 7H9 mycobacterial media (5). To confirm that the observed activity of compound A039 against M. tuberculosis is dependent upon growth in glycerol containing media, we compared the activity of A039 against M. tuberculosis grown on glycerol or acetate as the sole carbon source. Activity of A039 was dependent on the presence of glycerol (Figure 3c). Compounds with glycerol specific activity have been demonstrated to induce “self-poisoning” of M. tuberculosis by promoting accumulation of glycerol phosphate and rapid ATP depletion, and have no efficacy in animal models of tuberculosis (5). These hits are therefore artifacts of the standard laboratory growth conditions used to culture M. tuberculosis.

Figure 3
Identification and characterization of compounds with glycerol dependent killing mechanisms

While the prior study reported the identification of a compound series with glycerol dependent activity from screening of M. tuberculosis, we sought to determine the relative frequency with which this phenomenon of identifying glycerol dependent inhibitors occurs. To determine the number of hits from our primary screen with a glycerol dependent mechanism of action, we subjected the hits from the M. tuberculosis primary screen to secondary screens using either glycerol or acetate as a sole carbon source. Approximately ~10% of the hits were shifted towards greater activity in glycerol containing media than acetate containing media, including compound A039 (Figure 3d). These results indicate that the artifact resulting from growth on glycerol can result in large numbers of spurious hits from screens conducted on glycerol containing media, and reinforce the need for conducting primary screens on alternate carbon sources, or performing an early secondary screen to rapidly identify hits with a mechanism of action (MOA) that is glycerol independent.

Identification of a novel inhibitor of MmpL3

C215 (N-(2,4-dichlorobenzyl)-1-propyl-1H-benzo[d]imidazol-5-amine) is an inhibitor (Figure 4a) that we identified in HTS with glycerol independent activity against M. tuberculosis, limited non-specific toxicity against mammalian cells, an IC90 of 16μM against M. tuberculosis, and efficacy against M. tuberculosis growing in macrophages (Table S1, Table S2, compound 72). Four independent resistant mutants were generated to C215. These mutants had IC90 values that were shifted only ~2 fold relative to the wild-type parents used for resistance generation (Figure 4b). Sequencing of the genomes of the four resistant mutants identified Rv0206c as the only gene with mutations identified in all four resistant mutants, suggesting that mutation of Rv0206c confers resistance to C215. Rv0206c MmpL3, a member of the resistance, nodulation and cell division (RND) family of proteins that are thought to encode lipid transporters in M. tuberculosis (19). AU12345, a recently identified adamantyl urea compound with activity against M. tuberculosis, targets MmpL3, resulting in inhibition of the transport of mycolic acids across the inner membrane (11). In addition, SQ109, a 1,2-diamine related to ethambutol, was also shown to target MmpL3 (20). These compounds were used to demonstrate that MmpL3 encodes a transporter of trehalose-monomycolate (TDM), an essential precursor to myolic acid containing lipid species in the outer mycolate membrane. The target of AU12345 was identified using the method of resistance generation coupled to genome sequencing, and resistant mutants were found to harbor a single mutation in mmpL3 (G758A). In contrast, in our analysis of C215 resistant mutants, a different mutation in mmpL3 was identified in each of the resistant mutants we isolated (L320P, T667A, V684A, V51A). To demonstrate that C215 kills M. tuberculosis by inhibiting MmpL3 and transport of mycolic acids across the inner membrane, we treated cells with 1x, 5x, and 10x the MIC of C215, and examined cell wall mycolic acids. We observed a dose dependent decrease in cell wall bound mycolic acids (α, methoxy, and keto) upon treatment with C215, similar to the effect previously reported for AU12345, confirming that MmpL3 is the likely target of C215. This represents the fourth identification of a small molecule that kills M. tuberculosis by inhibiting MmpL3 (11, 20, 21).

Figure 4
The benzimidazole C215 kills M. tuberculosis by targeting MmpL3

Identification of a novel DprE1 inhibitor

Recent screens of 341,808 compounds conducted at the Southern Research Institute under the auspices of the National Institutes of Health resulted in a selected set of 1113 hits that were made available from NIAID (68). This screen was conducted using 7H12 media that utilizes palmitate as the major carbon source. We obtained these compounds and screened them in two independent screens using the GFP based assay with either glycerol or acetate as the primary carbon source. The compounds were screened at an average concentration of 20μg mL−1, slightly higher than the 15μg mL−1 used in the original NIAID/SRI screen. We observed a high degree of correlation between the composite z-scores obtained from the glycerol and acetate based screens (Figure S1). Using a z-score cutoff of −4 for this screen yielded 751 (67%) hits on glycerol and 698 (63%) hits on acetate. While this set of compounds was selected for their activity against M. tuberculosis, the relatively low proportion of active compounds again suggests that activity can be dependent on growth conditions and screening methodologies. The GFP assay we used depends upon bacterial replication and compounds that score as hits in this assay necessarily inhibit growth. For this reason, our assay utilizes culture conditions that support robust growth of the bacteria. In contrast, the resazurin (alamar blue) based assay used by the NIAID/SRI for their M. tuberculosis screening measure the reductive potential of living cells and does not require active growth conditions (22). Thus, their screen using resazurin as a reporter of bacterial viability utilized a long chain fatty acid as the primary carbon source, conditions that do not support robust growth of the bacteria. The differences in the conditions employed for screening thus likely account for the variation in activity observed.

From the set of compounds obtained from the NIAID we confirmed that compound 377790 (1-(4-(tert-butyl)benzyl)-3-nitro-1H-1,2,4-triazole), a novel nitro-substituted triazole (Figure 5a, Table S3), has good activity against M. tuberculosis with an IC90 of 0.5μM. Of note, several classes of antibiotics that are active against M. tuberculosis are characterized by the presence of a nitro-substituted heterocycle, including the nitroimidazole metronidazole (23), various nitrofurans (24), and the promising new agent PA-824, a bicyclic nitroimidazole (25, 26). These compounds all have activity against non-replicating, dormant M. tuberculosis, suggesting the exciting possibility that they may be more effective in treating latent M. tuberculosis infection than currently used first-line drugs. Although the MOA of these compounds is still under investigation, the reduction of the nitro-group and subsequent formation of reactive intermediates is critical for activity against non-replicating populations of cells. Resistant mutants generated against such compounds often have acquired mutations in activating enzymes. For example, FGD1, a specific glucose-6-phosphate dehydrogenase, and its deazaflavin cofactor F420 provide electrons for reductive activation of PA-824, and mutation in either of these enzymes confers resistance to this drug (25).

Figure 5
Nitro-triazole inhibitors kill M. tuberculosis by targeting DprE1

SAR analysis conducted around 377790 suggests that the nitro-group is essential for activity, as analogs lacking the nitro-group, or with various substitutions in place of the nitro group are significantly less active than the original nitro-containing hit (Table S3). The requirement for the nitro group could indicate that the mechanism of action of these compounds requires reduction of the nitro group to a reactive species, similar to PA-824, or that the nitro group is required for binding of the inhibitors to its target.

To elucidate the mechanism of action of these compounds, we generated resistance to compound 377790. Mutants generated from selections were resistant at >60x the IC90 (Figure 5b). The genomes of two independently generated resistant mutants to 377790 were sequenced. All four resistant mutants contained a single common mutation in dprE1, a gene that encodes the epimerase DprE1, resulting in the amino acid change C387S. Recently, a novel class of benzothiazinones (BTZs) that have activity against M. tuberculosis via inhibition of DprE1 were identified (9). These compounds, though structurally dissimilar to the triazole compounds described here, also possess a nitro group functionality, and are thought to inhibit DprE1 by formation of a covalent bond via reduction of the nitro group to a nitroso-derivative that reacts specifically with Cys387 (27, 28). Mutation of Cys387 also confers a high degree of resistance to BTZs. The requirement for the nitro-group in the triazoles is consistent with a mechanism in which this functionality is required for covalent modification of the DprE1 enzyme. As observed with BTZs, overexpression of DprE1 confers resistance to compound 377790 and other nitro-triazole analogs (Figure 5c, Table S3) further confirming that these compounds likely target DprE1. This is the third reported discovery of a small molecule that targets this enzyme (10, 27).

In this report we describe the identification of two novel series of inhibitors with clearly defined molecular targets. To identify the targets of these inhibitors we employed the method of resistance generation coupled with whole genome re-sequencing. Using this method, we identified the target of the novel benzimidazole inhibitor C215 as the cell wall mycolic acid transporter MmpL3, and the target of the novel nitro-triazole inhibitor 377790 as the epimerase DprE1. In recent years, three new targets have been identified from inhibitors discovered in whole cell screens. These targets are DprE1, MmpL3, and ATP synthase. It is interesting to note that several structurally unrelated inhibitors have now been reported for MmpL3. Prior to our study, the two reported DprE1 inhibitors were structurally related, suggesting that the apparent promiscuity of MmpL3 could be an isolated case. However, our identification of a completely structurally distinct inhibitor of DprE1 argues that both MmpL3 and DprE1 are highly druggable and can be inhibited by a wide variety of chemical structures. In addition, these findings suggest that whole cell screens, coupled with whole genome sequencing of resistant mutants, appear to be revealing the same targets repeatedly despite the fact that there are likely many essential and druggable proteins in M. tuberculosis that have yet to be uncovered. Alternatively, the repeated identification of the same targets may suggest that the use of resistance generation for target identification may bias the nature of targets identified to those that easily accommodate mutations in their active site while preserving target function. Thus, selecting compounds for prioritization based on the ability to identify a target using resistance generation may result in lead compounds with biased mechanisms of action.

Most target identification efforts in the field of M. tuberculosis anti-infectives currently rely on resistance generation, resulting in a possible quandary. By focusing on small molecules with an identified target based on this method, such as inhibitors of DprE1 or MmpL3, candidates may be prioritized to which clinical resistance may be easily and rapidly engendered. This bias highlights the need for multiple approaches to identify targets of promising inhibitors of M. tuberculosis. In fact, one might argue that prioritization should be given to those candidates for which we are unable to generate resistance, as this may be a desirable property of a new antibiotic. Thus, the development and use of parallel strategies for target identification are of vital importance to the discovery efforts geared towards the identification of the next generation M. tuberculosis drugs. Nonetheless, the identification of novel inhibitors with clearly defined targets is still relatively rare in the field of M. tuberculosis drug discovery, and the two novel inhibitors described in this report will be useful in future investigations related to drug discovery and M. tuberculosis cell wall biosynthesis.

All currently used antibiotics were discovered using whole cell screening, highlighting the importance of this approach. In this report we describe our experiences with whole cell high-throughput screening to identify novel inhibitors of M. tuberculosis replication using both an unbiased collection of >20K compounds from the Broad Institute and a set of 1113 hit compounds made available to the M. tuberculosis screening community by the NIAID. This latter data set provides independent verification and testing of the NIAID compounds to help define the importance of this collection to M. tuberculosis drug discovery efforts. In addition, by analyzing the primary and secondary screening data from the Broad collection, we were able to expand upon and quantify observations related to whole cell screening. This study using a large compound collection with no predicted potential for structural bias found an even smaller degree of overlap in compounds then previously reported (14) that are active against three mycobacterial species, clearly demonstrating the potential pitfalls of using surrogate mycobacterial species for identifying inhibitors of M. tuberculosis growth. In addition, this work describes the surprisingly high rate at which such hits are identified from whole cell screening that bear the recognized characteristic of dependence on growth on standard glycerol containing M. tuberculosis media. This is particularly important, as inhibitors of this nature have no potential for in vivo activity, suggesting that a major shift away from the standard growth media used for the study of M. tuberculosis may be warranted, particularly with respect to chemical biological investigations. Finally, we note the interesting phenomenon of repeated identification of the same targets DprE1 and Mmpl3 by the methods utilized, which may suggest that they are highly targetable but also raise the caveat that they may be targets that are easily mutated to confer resistance. In summary, this work provides insights and perspectives that should help guide future M. tuberculosis screening efforts and antibiotic discovery in general.


Bacterial strains and growth conditions

Bacterial strains used were M. tuberculosis H37Rv, M. bovis BCG strain Pasteur, and M. smegmatis MC2155. GFP was expressed using a pUV15tetORm derivative (29) with tetR deleted for constitutive expression. Mycobacterial strains were cultured in Middlebrook 7H9 medium with 0.05%(v/v) Tween–80 and 10%(v/v) OADC with either 0.25%(v/v) glycerol or 10μM acetate.

Compound Library

The 20,502 compounds screened against M. tuberculosis were obtained from the Broad Institute small molecule library. The categories of compounds were as follows: 30% commercially available compounds purchased predominantly from ChemDiv, Maybridge (Thermo Fisher Scientific) and TimTec; 40% compounds synthesized by Broad chemists using Diversity Oriented Synthesis (DOS) or other strategies for organic synthesis; 30% natural products consisting of natural product extracts and purified natural products.

Screening, hit selection, and IC90 determinations

For M. tuberculosis and BCG screening assays, bacteria expressing GFP was grown to mid-log phase (OD600=0.6–0.8), diluted, and plated into 384 well plates into which compounds had previously been pinned for a final OD600 of 0.025 and compound concentration of 25μM (Broad Library) or 20μg mL−1 (NIH inhibitors). Plates were incubated for a period of 72h, at which time GFP fluorescence was read. Each compound was screened in duplicate, and composite z-scores were calculated using DMSO controls as reference. Hits from the M. tuberculosis screen were defined as compounds with a composite z-score of less than −4. This z-score cutoff was selected using average of the z-scores of the concentrations of the control antibiotics clofazimine and rifampicin that gave a Z′-factor of 0, meaning that the distance separating the positive and negative controls is 3x the sum of the standard deviations of the two populations. For dose response curves and IC90 determination of hits from the screens, bacteria were grown to mid-log phase and plated in 96 well plates at OD600=0.05 in the presence of small molecule inhibitors for 7 days unless otherwise indicated, and growth was assessed by reading OD600. The IC90 was defined as the minimum concentration that inhibited growth by 90% relative to the DMSO control. For the M. smegmatis screen, OD600 was used as the readout and the plates were read after 2 days of incubation.

Secondary Screening Assays

Hits from the primary screen against M. tuberculosis were cherry picked from the library and arrayed in 384 well plates as a dose response curve using 2 fold serial dilutions from the primary screening concentration. Preliminary cherrypick plate MIC values were determined using the GFP based assay and are reported as the minimum concentration that gives the maximum inhibition observed using rifampicin controls. Macrophage activity was assessed by infecting J774 macrophages plated into 96 well plates with a strain of M. tuberculosis constitutively expressing firefly luciferase. After three days of infection in the presence of inhibitors, the infected monolayers were washed, lysed, and luminescence was used as a reporter of bacterial viability. To determine macrophage toxicity, J774 macrophages were incubated with small molecules for a period of 3 days upon which time CellTiter-Glo (Promega Corporation) was used as a readout for macrophage viability.

Generation of resistant mutants

The MIC of each compound on solid media was identified by plating 107 bacteria on agar containing a dose response in 96 well plate format. The MIC was defined as the lowest concentration resulting in inhibition of bacterial growth. Resistant mutants were generated by plating M. tuberculosis cells onto agar pads containing 2x and 10x the agar MIC of each compound using four independently derived wild-type clones. Colonies that arose on inhibitor containing plates were inoculated into liquid media containing 1x the liquid MIC of the inhibitor. These cultures were grown to mid-log and samples were retested in a liquid MIC assay to confirm that a shift relative to the wild-type MIC was observed.

Metabolic labeling of mycolic acids and compound treatment

10mL cultures of M. tuberculosis were grown to OD600=0.6 in inkwell bottles. 1x, 5x, or 10x the MIC of C215 added to the cultures simultaneously with 10μCi of 14C acetate. After 20h of labeling, the cultures were washed once with water, saponified with tetrabutylammonium hydroxide, and methyl esters of fatty acids and mycolic acids created by addition of methyl iodide. Fatty acid methyl esters (FAMEs) and mycolic acid methyl esters (MAMEs) were extracted from the aqueous layer using methylene chloride, and run out on TLC plates using 95:5 hexanes:ethyl acetate. TLC plates were loading was normalized according to the OD600 of the cells after the 20h treatment period. FAMEs and MAMEs were visualized by phosphorimaging.

Supplementary Material



We thank L. Verplank, N. Tolliday, J. Bittker, and the Broad Institute Chemical Biology platform for technical assistance, data management, and advice on assay development; B. Birren, C. Nussbaum, C. Russ and the Broad Institute Genome Sequencing and Analysis Platform for technical assistance and data analysis for genome sequencing; R. Goldman, B. Laughon, T. Parker, the NIAID and TAACF/SRI for providing the 1113 compound set; E. Rubin, J. Gomez and A. Stern for helpful discussion and advice throughout the project; K. Gordon for technical assistance. SAS gratefully acknowledges Novartis and the Helen Hay Whitney Foundation for funding. SSG gratefully acknowledges the NIH for funding K08AI085033. This work was supported through funding from the Chinese Academy of Science to SAS and the Bill and Melinda Gates Foundation to DTH. This work was also supported by the NIH Molecular Libraries Production Centers Network (MLPCN;, which funded primary screening of a 215,110-compound library against M. tuberculosis strain H37Rv at Southern Research Institute.


Supporting Information. Figure S1, Table S1, Table S2, Table S3, and materials and methods describing the synthesis of triazole compounds NT19-NT27 is available. This material is available free of charge via the Internet at


1. Ma Z, Lienhardt C, McIlleron H, Nunn AJ, Wang X. Global tuberculosis drug development pipeline: the need and the reality. Lancet. 2010;375:2100–2109. [PubMed]
2. Andries K, Verhasselt P, Guillemont J, Göhlmann HWH, Neefs JM, Winkler H, Van Gestel J, Timmerman P, Zhu M, Lee E, Williams P, de Chaffoy D, Huitric E, Hoffner S, Cambau E, Truffot-Pernot C, Lounis N, Jarlier V. A diarylquinoline drug active on the ATP synthase of Mycobacterium tuberculosis. Science. 2005;307:223–227. [PubMed]
3. Sacchettini JC, Rubin EJ, Freundlich JS. Drugs versus bugs: in pursuit of the persistent predator Mycobacterium tuberculosis. Nat Rev Microbiol. 2008;6:41–52. [PubMed]
4. Keller TH, Pichota A, Yin Z. A practical view of ‘druggability’ Curr Opin Chem Biol. 2006;10:357–361. [PubMed]
5. Pethe K, Sequeira PC, Agarwalla S, Rhee K, Kuhen K, Phong WY, Patel V, Beer D, Walker JR, Duraiswamy J, Jiricek J, Keller TH, Chatterjee A, Tan MP, Ujjini M, Rao SPS, Camacho L, Bifani P, Mak PA, Ma I, Barnes SW, Chen Z, Plouffe D, Thayalan P, Ng SH, Au M, Lee BH, Tan BH, Ravindran S, Nanjundappa M, Lin X, Goh A, Lakshminarayana SB, Shoen C, Cynamon M, Kreiswirth B, Dartois V, Peters EC, Glynne R, Brenner S, Dick T. Nat Comms. Vol. 1. Nature Publishing Group; 2010. A chemical genetic screen in Mycobacterium tuberculosis identifies carbon-source-dependent growth inhibitors devoid of in vivo efficacy; pp. 1–8. [PMC free article] [PubMed]
6. Ananthan S, Faaleolea ER, Goldman RC, Hobrath JV, Kwong CD, Laughon BE, Maddry JA, Mehta A, Rasmussen L, Reynolds RC, Secrist JA, III, Shindo N, Showe DN, Sosa MI, Suling WJ, White EL. High-throughput screening for inhibitors of Mycobacterium tuberculosis H37Rv. Tuberculosis. 2009;89:334–353. [PMC free article] [PubMed]
7. Goldman RC, Laughon BE. Discovery and validation of new antitubercular compounds as potential drug leads and probes. Tuberculosis. 2009;89:331–333. [PMC free article] [PubMed]
8. Maddry JA, Ananthan S, Goldman RC, Hobrath JV, Kwong CD, Maddox C, Rasmussen L, Reynolds RC, Secrist JA, Sosa MI, White EL, Zhang W. Antituberculosis activity of the molecular libraries screening center network library. Tuberculosis. 2009;89:354–363. [PMC free article] [PubMed]
9. Makarov V, Manina G, Mikusova K, Mollmann U, Ryabova O, Saint-Joanis B, Dhar N, Pasca MR, Buroni S, Lucarelli AP, Milano A, De Rossi E, Belanova M, Bobovska A, Dianiskova P, Kordulakova J, Sala C, Fullam E, Schneider P, McKinney JD, Brodin P, Christophe T, Waddell S, Butcher P, Albrethsen J, Rosenkrands I, Brosch R, Nandi V, Bharath S, Gaonkar S, Shandil RK, Balasubramanian V, Balganesh T, Tyagi S, Grosset J, Riccardi G, Cole ST. Benzothiazinones Kill Mycobacterium tuberculosis by Blocking Arabinan Synthesis. Science. 2009;324:801–804. [PMC free article] [PubMed]
10. Christophe T, Jackson M, Jeon HK, Fenistein D, Contreras-Dominguez M, Kim J, Genovesio A, Carralot JP, Ewann F, Kim EH, Lee SY, Kang S, Seo MJ, Park EJ, Skovierová H, Pham H, Riccardi G, Nam JY, Marsollier L, Kempf M, Joly-Guillou ML, Oh T, Shin WK, No Z, Nehrbass U, Brosch R, Cole ST, Brodin P. High content screening identifies decaprenyl-phosphoribose 2′ epimerase as a target for intracellular antimycobacterial inhibitors. PLoS Pathog. 2009;5:e1000645. [PMC free article] [PubMed]
11. Grzegorzewicz AE, Pham H, Gundi VAKB, Scherman MS, North EJ, Hess T, Jones V, Gruppo V, Born SEM, Korduláková J, Chavadi SS, Morisseau C, Lenaerts AJ, Lee RE, McNeil MR, Jackson M. Inhibition of mycolic acid transport across the Mycobacterium tuberculosis plasma membrane. Nat Chem Biol. 2012;8:334–341. [PMC free article] [PubMed]
12. Collins LA, Torrero MN, Franzblau SG. Green fluorescent protein reporter microplate assay for high-throughput screening of compounds against Mycobacterium tuberculosis. Antimicrob Agents Chemother. 1998;42:344–347. [PMC free article] [PubMed]
13. Dandapani S, Marcaurelle LA. Current strategies for diversity-oriented synthesis. Curr Opin Chem Biol. 2010;14:362–370. [PubMed]
14. Altaf M, Miller CH, Bellows DS, O’Toole R. Evaluation of the Mycobacterium smegmatis and BCG models for the discovery of Mycobacterium tuberculosis inhibitors. Tuberculosis. 2010;90:333–337. [PubMed]
15. Waring P, Beaver J. Gliotoxin and related epipolythiodioxopiperazines. Gen Pharmacol. 1996;27:1311–1316. [PubMed]
16. Müllbacher A, Waring P, Tiwari-Palni U, Eichner RD. Structural relationship of epipolythiodioxopiperazines and their immunomodulating activity. Mol Immunol. 1986;23:231–235. [PubMed]
17. Tompsett R, McDermott TW, Kidd JG. Tuberculostatic activity of blood and urine from animals given gliotoxin. J Immunol. 1950;65:59–63. [PubMed]
18. Kwon-Chung KJ, Sugui JA. What do we know about the role of gliotoxin in the pathobiology of Aspergillus fumigatus? Med Mycol. 2009;47(Suppl 1):S97–103. [PMC free article] [PubMed]
19. Cox JS, Chen B, McNeil M, Jacobs WR. Complex lipid determines tissue-specific replication of Mycobacterium tuberculosis in mice. Nature. 1999;402:79–83. [PubMed]
20. Tahlan K, Wilson R, Kastrinsky DB, Arora K, Nair V, Fischer E, Barnes SW, Walker JR, Alland D, Barry CE, Boshoff HI. SQ109 Targets MmpL3, a Membrane Transporter of Trehalose Monomycolate Involved in Mycolic Acid Donation to the Cell Wall Core of Mycobacterium tuberculosis. Antimicrob Agents Chemother. 2012;56:1797–1809. [PMC free article] [PubMed]
21. La Rosa V, Poce G, Canseco JO, Buroni S, Pasca MR, Biava M, Raju RM, Porretta GC, Alfonso S, Battilocchio C, Javid B, Sorrentino F, Ioerger TR, Sacchettini JC, Manetti F, Botta M, De Logu A, Rubin EJ, De Rossi E. MmpL3 is the cellular target of the antitubercular pyrrole derivative BM212. Antimicrob Agents Chemother. 2012;56:324–331. [PMC free article] [PubMed]
22. Taneja NK, Tyagi JS. Resazurin reduction assays for screening of anti-tubercular compounds against dormant and actively growing Mycobacterium tuberculosis, Mycobacterium bovis BCG and Mycobacterium smegmatis. J Antimicrob Chemother. 2007;60:288–293. [PubMed]
23. Brooks JV, Furney SK, Orme IM. Metronidazole therapy in mice infected with tuberculosis. Antimicrob Agents Chemother. 1999;43:1285–1288. [PMC free article] [PubMed]
24. Murugasu-Oei B, Dick T. Bactericidal activity of nitrofurans against growing and dormant Mycobacterium bovis BCG. J Antimicrob Chemother. 2000;46:917–919. [PubMed]
25. Stover CK, Warrener P, VanDevanter DR, Sherman DR, Arain TM, Langhorne MH, Anderson SW, Towell JA, Yuan Y, McMurray DN, Kreiswirth BN, Barry CE, Baker WR. A small-molecule nitroimidazopyran drug candidate for the treatment of tuberculosis. Nature. 2000;405:962–966. [PubMed]
26. Singh R, Manjunatha U, Boshoff HIM, Ha YH, Niyomrattanakit P, Ledwidge R, Dowd CS, Lee IY, Kim P, Zhang L, Kang S, Keller TH, Jiricek J, Barry CE. PA-824 kills nonreplicating Mycobacterium tuberculosis by intracellular NO release. Science. 2008;322:1392–1395. [PMC free article] [PubMed]
27. Trefzer C, Skovierová H, Buroni S, Bobovská A, Nenci S, Molteni E, Pojer F, Pasca MR, Makarov V, Cole ST, Riccardi G, Mikušová K, Johnsson K. Benzothiazinones are suicide inhibitors of mycobacterial decaprenylphosphoryl-β-D-ribofuranose 2′-oxidase DprE1. J Am Chem Soc. 2012;134:912–915. [PubMed]
28. Trefzer C, Rengifo-Gonzalez M, Hinner MJ, Schneider P, Makarov V, Cole ST, Johnsson K. Benzothiazinones: prodrugs that covalently modify the decaprenylphosphoryl-β-D-ribose 2′-epimerase DprE1 of Mycobacterium tuberculosis. J Am Chem Soc. 2010;132:13663–13665. [PubMed]
29. Ehrt S. Controlling gene expression in mycobacteria with anhydrotetracycline and Tet repressor. Nucleic Acids Research. 2005;33:e21–e21. [PMC free article] [PubMed]