Baeyer-Villager monooxygenases EthA and MymA are required for activation of inhibitors against replicating and non-replicating Mycobacterium tuberculosis

Summary

Successful treatment of Mycobacterium tuberculosis infection typically requires a complex regimen administered over at least six months. Interestingly, many of the antibiotics used to treat M. tuberculosis are prodrugs that require intracellular activation. Here, we describe three small molecules, active against both replicating and non-replicating M. tuberculosis, that require activation by Baeyer-Villiger monooxygenases (BVMOs). Two molecules require BVMO EthA (Rv3854c) for activation and the third molecule requires the BVMO MymA (Rv3083). While EthA is known to activate the antitubercular drug ethionamide, this is the first description of MymA as an activating enzyme of a prodrug. Further, we found that MymA also plays a role in activating ethionamide, with loss of MymA function resulting in ethionamide resistant M. tuberculosis. These findings suggest overlap in function and specificity of the BVMOs in M. tuberculosis.

Graphical Abstract

INTRODUCTION

Infection with the bacterium Mycobacterium tuberculosis remains a major global health problem, with an estimated 9.0 million new infections and 1.5 million deaths occurring in 2013 (WHO, 2014). While active tuberculosis (TB) infection is curable, successful treatment requires a complex treatment regimen over at least six months. In the case of infection with a drug-resistant M. tuberculosis strain, the treatment course is typically longer and less likely to be successful (Zumla et al., 2013). Treating clinically asymptomatic or latent infection, in which the host immune system restricts M. tuberculosis growth but does not eradicate infection, also requires a prolonged course of antibiotic therapy: either multi-drug therapy for 3 months or mono-therapy for as long as 9 months (Zumla et al., 2013).

During TB infection, a subpopulation of M. tuberculosis bacteria likely exists in a non-replicating or slowly replicating state, refractory to many antibiotics (Gomez and McKinney, 2004; McCune et al., 1966). The presence of non-replicating bacteria within the host is thought to contribute to the need for prolonged drug therapy for active and latent TB infection (Connolly, 2007). The majority of anti-tuberculosis drugs used clinically are ineffective against carbon-starved non-replicating bacteria, with the notable exceptions of bedaquiline, which targets ATP synthase, and clofazimine, an antibiotic approved for the treatment of leprosy, which is reduced by the enzyme NDH-2 within the electron transport chain and produces toxic ROS upon oxidation (Yano et al., 2011; Zhang and Yew, 2009). Recent studies demonstrate that including bedaquiline and clofazimine in anti-tubercular regimens in a murine model of tuberculosis improves treatment outcomes and in addition, a clofazimine-containing 12-month regimen successfully treated 89% of MDR-TB cases (Williams et al., 2012; Kuaban et al., 2015). These data suggest that including antibiotics with activity against non-replicating bacteria may shorten the duration of antibiotic therapy for both active and latent tuberculosis infection.

We recently identified novel inhibitors of the non-replicating state, using carbon starvation as a model for non-replicating M. tuberculosis (Grant et al., 2013). In this report, we describe 3 molecules that act like prodrugs in that they require oxidation by monooxygenases for activity. EthA, a Baeyer-Villiger monooxygenase (BVMO) required for the activation of the anti-tubercular drug ethionamide and the second-line antibiotics thiacetazone and isoxyl (DeBarber et al., 2000; Dover et al., 2007; Fraaije et al., 2004), is required for the activity of two of the small molecules. MymA (Rv3083), also a Baeyer-Villiger monooxygenase (Bonsor et al., 2006), is required to activate the third molecule. Further, we show that MymA can additionally activate ethionamide and that loss of MymA function confers some resistance to ethionamide. We suggest that EthA and MymA both function as activating enzymes within the cell, oxidizing ethionamide as well as newly identified sulfur-containing compounds. Our experiences suggest that the abundance of oxidizing enzymes in M. tuberculosis may favor the discovery of prodrugs that require oxidation.

RESULTS

Sequencing of resistant mutants for target identification

We had previously identified 3 small molecules with replicating and non-replicating activity against M. tuberculosis: 3-(m-tolyl)-5-((1-piperidinyl)carbonylmethyl)thio-1,2,4-thiadiazole (1), N-(4-chlorophenyl)-5-(1-methyl-5-(trifluoromethyl)-1H-pyrazol-3-yl)thiophene-2-carboxamide (2), and 2-((2-(1,3-dioxan-2-yl)ethyl)thio)-5-phenyl-1,3,4-oxadiazole (3) (Fig. 1A). We sought to understand their mechanisms of action by generating resistant mutants followed by whole genome sequencing (Andries et al., 2005; Grzegorzewicz et al., 2012; Loerger et al., 2010 Makarov et al., 2009; Stanley et al., 2012). Independent resistant mutants were generated in parallel using 4 parent clones and resistance was confirmed prior to whole genome sequencing using Illumina based technology (Fig. S1A-C).

Figure 1. Resistance to compounds 1 and 2 is conferred by mutations in ethA.

A) Structures for compounds 1, 2 and 3. B-D) Dose response curves for 1-RM1 (orange), 1-RM3 (red), 2-RM1 (light blue), 2-RM3 (dark blue) and wild-type M. tuberculosis (open black circles) against ethionamide (B), 1 (C) and 2 (D). E-G) Dose response curves for wild-type M. tuberculosis (open black circles circles), 1-RM3 with (red open squares) and without (red closed squares) episomal complementation with ethA, and a 2-RM1 with (blue open squares) and without (blue closed squares) episomal complementation with ethA against ethionamide (E), 1 (F) and 2 (G). See also Figure S1.

EthA is required for activity of compounds 1 and 2

For two compounds, thiadiazole 1 and thiophene-pyrazole 2, mutations in the ethA gene were identified in each of the four independent resistant clones sequenced (Fig. S1A-C and Table 1). All contained frameshift or missense mutations leading to loss of function mutations in ethA. Consistent with there being loss of function mutations in ethA, the mutant strains displayed varying levels of ethionamide resistance compared to the wild-type strain (Fig. 1B and Fig. S1D,E). The resistant mutant strains generated against one compound were cross-resistant to the other (Fig. 1C,D). When complemented with an episomal copy of wild-type ethA under the control of a constitutive promoter, the resistant strains regained sensitivity to ethionamide, 1 and 2 (Fig. 1E-G). While EthA has previously been reported to oxidize the thioamide group in small molecules such as ethionamide and thioureas in thiacetazone and isoxyl, and was recently reported to oxidize a thiophene that inhibits the CTP Synthetase PyrG, the role of EthA in oxidizing other sulfur containing compounds such as thioether 1 suggests a wider substrate preference than previously reported (Nishida and Ortiz de Montellano, 2011; Mori, et al., 2015).

Table 1.

Strains and plasmids used in this work.

Strains
Strain	Genotype	Allele characterization	Source
Wild-type (WT)	H37Rv
1-RM1	ethA-	ethA E36Q	WT selected with 1
1-RM2	ethA-	ethA Frameshift +g AA436	WT selected with 1
1-RM3	ethA-	ethA W69R	WT selected with 1
1-RM4	ethA-	ethA W240stop	WT selected with 1
2-RM1	ethA-	ethA Frameshift –c AA156	WT selected with 2
2-RM2	ethA-	ethA P447T	WT selected with 2
2-RM3	ethA-	ethA Frameshift –a AA133	WT selected with 2
2-RM4	ethA-	ethA C131stop	WT selected with 2
3-RM1	mymA-	13kB deletion RV3079c-Rv3089	WT selected with 3
3-RM2	mymA-	virS E285Q	WT selected with 3
3-RM3	mymA-	virS E285Q	WT selected with 3
3-RM4	mymA-	Frameshift +cAA318	WT selected with 3
Rv3083::TN		3448753	This study
RV3084::TN		3450033	This study
Rv3085::TN		3451750	This study
Rv3086::TN		3452413	This study
Rv3087::TN		3452961	This study
Rv3088::TN		3454690	This study
Rv3089::TN		3455999	This study
Rv3854c::TN		4327379	This study
Rv1393c::TN		1569366	This study
Rv0565c::TN		657361	This study
Rv3049c::TN		3411023	This study
SG1	ethA- mymA-	ethA S138A mymA 13kbΔ Rv3079c-Rv3089	3-RM1 selected with ethionamide
SG2	ethA- mymA-	ethA W45C mymA 13kbΔ Rv3079c-Rv3089	3-RM1 selected with ethionamide
SG3	ethA- mymA-	ethA -cAA302 mymA 13kbΔ Rv3079c-Rv3089	3-RM1 selected with ethionamide
1-RM3_OE_ethA		1-RM3 harboring pRS3Δtet_Rv3854	This study
2-RM1_ethA		2-RM1 harboring pRS3Δtet Rv3854	This study
3-RM1 VirS		3-RM1 harboring pMV762_VirS	This study
3-RM4 VirS		3-RM4 harboring pMV762_VirS	This study
WT_mymA		H37Rv harboring pUVRS3_Rv3083	This study
Rv3083::TN_mymA		Rv3083::TN harboring pUVRS3_Rv3083	This study
3-RM1_mymA		3-RM1 harboring pUVRS3_Rv3083	This study
3-RM_ethA		3-RM1 harboring pUVRS3Δtet_Rv3854	This study
SG1_ethA		SG1 harboring pUVRS3Δtet_Rv3854	This study
SG1_ethA		SG1 strain harboring pUVRS3_Rv3083	This study
Plasmids
pUVRS3	pUV15tetORm derivative with HA tag
pMV762_VirS	pMV762 derivative expressing VirS
pUVRS3_Rv3083	pUVRS3 derivative expression Rv3083 (inducible)
pUVRS3Δtet_Rv3854	pUVRS3 derivative with tetR deleted and subsequent constitutive expression of Rv3854

To determine whether EthA is also required for the activity of these two compounds against non-replicating cells, we tested the activity of 1 and 2 against carbon starved wild-type M. tuberculosis and one of the ethA mutants (2-RM3). We found that both small molecules exhibited significantly less activity against the ethA mutant strain compared to wild-type under carbon starvation conditions, suggesting that activation by EthA is also required for the non-replicating activity (Fig. 2A,B). The ethA mutation does not provide complete protection against these molecules, however, suggesting that there may be an additional mechanism for activation, or alternatively, that the reduced parent molecule may have some activity on non-replicating bacteria.

Figure 2. Resistance under non-replicating conditions is also conferred by loss of function mutations in ethA.

A,B) Wild-type M. tuberculosis (black bars) or an ethA mutant (2-RM3, hatched bars) was starved for 5 weeks in 7H9/Tx media and then treated for 14 days with compound at the indicated doses. Dashed lines indicate limit of detection. C) Structure of 4. D) Dose response curves for 1 and 4 versus wild-type M. tuberculosis. E) Dose response curves for 4 versus wild-type M. tuberculosis (open black circles) and two ethA mutants (purple, red squares). F) The analog 4 was tested against carbon starved wild-type M. tuberculosis (black bars) or an ethA mutant (2-RM3, hatched bars). Treatment duration was 2 weeks.

For thiadiazole 1 we hypothesized that EthA may oxidize the sulfide to a sulfoxide. To test this hypothesis we synthesized the sulfoxide analog, 4 (Fig. 2C). When tested against replicating wild-type M. tuberculosis, the MIC₉₀ (minimum inhibitory concentration, measured by outgrowth of bacteria) of 4 is the same as that of the parent molecule 1 (Fig. 2D). When tested against two of the ethA mutants that are resistant to 1, the MICs of the sulfoxide 4 were the same for both the ethA resistant mutant and the wild-type strain, consistent with the hypothesis that EthA is no longer required for activation of 4 (Fig. 2E). We then tested the sulfoxide 4 against a carbon starved ethA mutant (2-RM3) and found that the starvation MBC₉₀ (minimal bactericidal concentration, measured by plating for colony forming units following treatment with compound) was the same for the wild-type and ethA mutant strain, in contrast to 1 which had reduced starvation activity in the ethA mutant strain (Fig. 2F). The similar efficacies of 1 and 4 against non-replicating cells suggest that the reduced parent molecule does not have significant activity under non-replicating conditions. Thus the residual 1 mediated killing observed in the carbon starved ethA mutant (Fig. 2A) likely reflects the presence of an additional mechanism for compound activation within the cell. Together these data suggest that EthA oxidizes compound 1 to the sulfoxide analog 4, which is the active form of this small molecule under replicating and non-replicating conditions, and further, that there may be multiple mechanisms to activate the molecule within the cell.

MymA is required for thiooxadiazole 3 activity

For the thiooxadiazole 3, we isolated four independent resistant clones which all contained mutations involving the gene virS (Fig. S1c, Table 1). VirS, an AraC/XylS transcription factor, regulates an operon (mymA operon) located downstream from virS, in a divergent orientation (Fig. 3A) (Singh et al., 2003). The mymA operon has been reported to be upregulated under acidic stress and virS mutants exhibit altered mycolic acid composition and cell wall ultrastructure (Singh et al., 2005; Singh et al., 2003). One of the resistant mutants (3-RM1) was found to have a 13 kB deletion spanning the entire mymA operon as well as virS and pknK, the gene encoding a serine/threonine protein kinase located upstream of virS. Two resistant mutants contained an identical SNP at position 285 (3-RM2 and 3-RM3) in virS, and the fourth mutation contained a frameshift mutation in the 3’-end of virS (3-RM4) (Table 1, Fig. S1). The fact that resistance is conferred when virS and the mymA operonare deleted indicates that VirS cannot be the direct target of the thiooxadiazole.

Figure 3. The monooxynase encoded by mymA activates compound 3.

A) The structure of the mymA operon. VirS is a transcription factor located upstream from the mymA operon in a divergent orientation. B) Dose response curves for wild-type M. tuberculosis (open black circles), 3-RM1 (pink triangles), 3-RM4 (green squares) and 3-RM1 and 3-RM4 complemented with an episomal copy of virS (open triangles and squares) versus 3. C) Dose response curves for transposon mutants in the mymA operon versus 3. D) Dose response curves for wild-type M. tuberculosis containing an inducible episomal copy of mymA, either induced (gray circle) with anhydrotetracycline (AHT) or uninduced (open circle), versus 3. E) Dose response curves for the Rv3083::TN_mymA, either induced with AHT (open triangles) or uninduced (closed triangles) versus 3. F) Dose response curves for the virS deletion mutant containing an inducible copy of mymA, either induced with AHT (open squares) or uninduced (closed squares), versus 3.

We complemented two of the resistant mutants, the deletion mutant (3-RM1) and the frameshift mutant (3-RM4), with a wild-type episomal copy of virS under the control of the groEL promoter. In the frameshift mutant 3-RM4, complementation with wild-type virS restored sensitivity to thiooxadiazole 3. However, in the deletion mutant 3-RM1, complementation of virS failed to restore sensitivity (Fig. 3B). These results led us to hypothesize that the loss of function mutations in virS may confer resistance via its impact on expression of genes in the mymA operon.

To test this hypothesis, we measured the sensitivity to thiooxadiazole 3 of a set of mutants in which each of the genes of the mymA operon (Rv3083-3089) was disrupted by a transposon insertion (Fig. 3C). We found that disruption of mymA (Rv3083) uniquely confers resistance to the thiooxadiazole 3, suggesting that mymA is required for thiooxadiazole 3 activity (Fig. 3C). When we overexpressed mymA in three strains: wild-type, the mymA transposon mutant (Rv3083::Tn) and 3-RM1, we observed the restoration of sensitivity to 3 in each of these strains when mymA expression was induced with anhydrotetracycline (AHT) compared to uninduced controls (Fig. 3D-F). This result suggests that resistance to 3 in all of the mutants is due to loss of function of MymA.

Optimization of the thiooxadiazole scaffold for replicating and non-replicating activity

We next performed chemistry on thiooxadiazole 3 to optimize both replicating and non-replicating activity. We found that compounds with S-2-pyridylmethyl substitution improve activity against non-replicating bacteria (3, 3A vs 3B~3G, Table S1) and the presence of a halogen or small substituent (methyl or methoxy) at the para position of phenyl group at 5-position of oxadiazole improves activity against replicating bacteria (3B vs 3C~3G). The three most active molecules, 3F, 3E, and 3C, have an MIC₉₀ against replicating bacteria of 16μM (Fig. S2A). When tested against non-replicating bacteria the compounds remain active; 3F has an MBC₉₀ against non-replicating, carbon-starved bacteria of 8μM (Fig. S2B). Thus this molecule is active against replicating M. tuberculosis as well as non-replicating M. tuberculosis. We confirmed that loss of function mutations in MymA confer resistance to 3F under both replicating and non-replicating conditions, similar to the original thiooxadiazole 3 (Fig. S2A,B). MymA is therefore required for 3 and 3F activity under both replicating and non-replicating conditions. 3F was selected for these experiments it has the greatest dependence on MymA for activity.

MymA activates thiooxadiazole 3

Given that MymA is required for thiooxadiazole 3 activity, we hypothesized that MymA, previously described as a Baeyer-Villiger monooxygenase similar to EthA, may be functioning as an activating enzyme of a prodrug by oxidizing thiooxadiazole 3 (Bonsor et al., 2006; DeBarber et al., 2000). As oxadiazolones have previously been reported to have activity against replicating wild-type M. tuberculosis, as determined by an Alamar blue inhibition assay (Macaev et al., 2005; Zampieri et al., 2009), we hypothesized that MymA may be converting the thiooxadiazole 3 into the corresponding oxadiazolone 5 (Fig. 4A, Fig. S3). When we tested oxadiazolone 5 as well as an additional oxadiazolone 6 against wild-type M. tuberculosis H37Rv, however, we found that the oxadiazolones had significantly less activity against replicating bacteria than the parent thiooxadiazole 3 (Fig. 4B). Alternatively, MymA could oxidize the thioether to its corresponding sulfoxide. We thus synthesized the corresponding sulfoxide, 7, (Fig. 4A) for one of the thiooxadiazoles (3A) and found that the sulfoxide 7 displayed similar efficacy against wild-type H37Rv as the parent 3A (Fig. 4C). In addition, it was equally active against the virS deletion mutant and the mymA transposon mutant, consistent with the hypothesis that MymA is required for activation of 3A by oxidizing it to the sulfoxide (Fig. 4C).

Figure 4. Activity of oxidized 3 analogs.

A) The structures of analogs 5-7. See also Figure S3 B) Dose response curves for analogs 3, 5, and 6 against wild-type replicating M. tuberculosis. C) Dose response curves for analog 7 against wild-type (open black circles), the mymA transposon mutant (red triangles) and the mymA deletion mutant (pink triangles). D) The activity of analog 7 (hatched) and analog 3A (solid) versus carbon-starved wild-type M. tuberculosis at the indicated doses. The bacteria were treated for 14 days, then serially diluted and enumerated by CFU. See Figure S4 for data on the activity of recombinant MymA on 3A and 7 in vitro.

To confirm MymA's ability to oxidize 3A, we directly tested the activity of recombinant MymA on 3A in vitro and observed, by LC-MS, the formation of products with molecular weight and retention time consistent with that of the corresponding sulfoxide 7 and sulfone (Fig. S4). The oxidized product, 7 is also a substrate for MymA in vitro as we observed conversion to the sulfone in a MymA dependent manner (Fig. S4). The sulfone then degrades to lower molecular weight products. Though 7 is a substrate for MymA in vitro, its activity in whole cells does not depend on MymA (Fig. 4C, D), suggesting that 7 must act on its target in vivo before MymA can oxidize it further.

We next tested the thiooxadiazole 3A and the sulfoxide 7 in wild-type H37Rv carbon starved M. tuberculosis to compare the non-replicating activity of the thioether and sulfoxide analogs (Fig. 4D). We found that the sulfoxide 7 demonstrates greater activity than the thioether 3A under non-replicating conditions. The superior activity of 7 against non-replicating bacteria compared to 3A could arise if MymA is less active in carbon-starved cells than in replicating cells, thereby preventing the conversion of 3A to the active entity 7. Together these data suggest that the sulfoxide is the active form of the compound under both replicating and non-replicating conditions and that it might be a preferable candidate to the parent thioether for any further development against TB.

Evidence for MymA activation of ethionamide

Previous work has suggested the presence of 6 Baeyer-Villiger monooxygenases in M. tuberculosis, with EthA being the most characterized because of its known role in activating the prodrug ethionamide. Of these 6 monooxygenases, mymA and ethA share the greatest sequence homology (Bonsor et al., 2006; DeBarber et al., 2000). In light of the finding that MymA activates 3 through oxidation, we sought to test whether MymA might also activate ethionamide. We first tested whether loss of function mutations in mymA confer resistance to ethionamide. Indeed, we found that loss of MymA function in 3-RM1 or Rv3083::Tn confers comparable levels of resistance to ethionamide as loss of EthA function in 2-RM3 or a mutant containing transposon disruption of ethA (Rv3854c::Tn) (Fig. 5A). We found that loss of function of MymA (3-RM1) also confers low-level resistance to 1 and 2, though not as high level as observed with EthA loss of function mutations (Fig. 5B,C). We then overexpressed mymA using an inducible promoter in wild-type M. tuberculosis H37Rv and observed a two-fold decrease in the MIC₉₀ for ethionamide, further suggesting that MymA can activate ethionamide (Fig. 5D). When we overexpressed mymA using an inducible promoter in 3-RM1, we similarly observed a decrease in the MIC₉₀ for ethionamide (Fig 5E). Given this finding, we tested transposon mutants in three other BVMOs, Rv1393c, Rv0565c, and Rv3049c, and found that loss of any of these genes does not confer resistance to ethionamide in vitro (Fig. 5F).

Figure 5. MymA and EthA can both activate ethionamide.

A) Dose response curves for ethionamide against wild-type (black open circles), the mymA transposon mutant (red triangles), the 3 deletion mutant (pink triangles), the ethA transposon mutant (purple squares) and 2-RM3 (blue circles) strains. B,C) Dose response curves for the 3 deletion mutant (pink triangles) and 2-RM3 (blue circles) strains versus 2 (B) and 1 (C). D) Dose response curve for wild-type M. tuberculosis uninduced (open black circles) and induced (black circles) mymA overexpression versus ethionamide. E) Dose response curve for wild-type (open black circles), 3-RM1 (pink triangles), 3-RM1_mymA induced with AHT (red triangles) or uninduced (purples triangles) versus ethionamide. F) Activity of ethionamide against replicating M. tuberculosis Baeyer-Villiger monooxygenase transposon mutants. Dose response curves for wild type, Rv1393c::TN, Rv0565c::TN, Rv3049c::TN against ethionamide. Points represent the average of quadruplicates and error bars represent standard deviation with the experiment repeated twice.

In order to test the complementary roles of EthA and MymA in oxidizing the prodrug ethionamide, we attempted to select for resistant mutants with mutations in both mymA and ethA. We plated the mymA deletion mutant (3-RM1) on high concentrations of ethionamide (500μM) with a selection frequency for resistance of approximately 1 in 10⁸ colonies. Three independent clones (SG1-3) were selected from different plates for confirmation of resistance. When tested against ethionamide, these strains exhibited significantly higher levels of resistance to ethionamide (IC₅₀ > 50μM; wild-type IC₅₀ = 2.2μM) than the original strains carrying only a mymA deletion (IC₅₀ = 10.1μM) or only a loss of function mutation in ethA mutant strains (IC₅₀ = 12.5μM) (Fig. 6A,B). In addition, while the parent mymA mutant 3-RM1 did not exhibit resistance to thiacetazone, the 3-RM1 clones selected for high level ethionamide resistance also exhibited resistance to thiacetazone (Fig. 6C). As thiacetazone is an antitubercular drug that requires activation by EthA, this observed resistance suggested loss of EthA function in the mutants. Subsequent whole genome sequencing indeed confirmed new mutations in ethA in these strains with high-level resistance to ethionamide, in addition to the parental mymA deletion (Table 1). We then complemented these mutants with episomal copies of mymA or ethA and found that episomal copies of mymA or ethA restored sensitivity to ethionamide (Fig. 6D,E). Together these data suggest that both EthA and MymA can activate ethionamide in an additive manner, thus representing a novel mechanism for ethionamide activation and resistance.

Figure 6. The SG1 double mutant exhibits high levels of ethionamide resistance.

A) Dose response curves for SG1 (purple), SG2 (brown), SG3 (green), 2-RM3 (blue), 3-RM1 (pink), and wild-type M. tuberculosis (black open circles) versus ethionamide. B) Activity of ethionamide against wild-type M. tuberculosis, the original resistant mutants and the mymA/ethA double mutants. IC₅₀ is reported as a more relevant metric of activity due to the long tail of slight growth at higher concentrations of ethionamide that would make accurate determination of an MIC₉₀ challenging. C) Dose response curves for thiacetazone against M. tuberculosis. Points represent the average of quadruplicates with the experiment repeated twice. D, E) mymA/ethA double mutants complemented with ethA (D) and mymA (E).

Selection for loss of function of mymA in clinical strains

A review of recently published whole genome sequencing data for clinical M. tuberculosis strains (Casali et al., 2014) suggests a potential role for MymA in vivo. The landscape of resistance to ethionamide in vivo is very complex. Many resistance loci have been identified for ethionamide, including ethA, ndh, mshA, and the inhA promoter, and yet there are still clinical isolates resistant to ethionamide that do not harbor mutations in any of these known loci (Eldholm et al., 2014). Furthermore, not all strains containing ethA loss of function mutations achieve levels of clinical resistance. For example, in the Casali et al. dataset, 29 of 83 strains harboring ethA loss of function mutations (35%) are classified as sensitive to ethionamide (Table S2), which would be consistent with our laboratory findings that an ethA mutation alone, is not sufficient to confer high level ethionamide resistance. In contrast, the two strains in their data set that carry both ethA and mymA mutations are indeed classified as clinically ethionamide resistant.

Analysis of this dataset identified 105 strains with 15 unique mymA loss of function mutations (nonsense or frameshift mutations). Of this set of >1000 clinical strains, mymA had more unique loss of function mutations than nearly all other TB genes, ranking 10th; ethA, pncA, and gidB loss of function, which are associated with resistance to ethionamide, pyrazinamide, and streptomycin, ranked 2nd, 6th, and 15th on the list, respectively (Table S3). Unique mymA loss of function mutations are, therefore, as frequent as other known loss of function resistance mechanisms, and are more frequent than 99.7% of all M. tuberculosis genes, suggesting that there is indeed selective pressure to inactivate mymA. In this list, another monooxygenase, Rv0565c ranks 25th (Table S3). In fact, others have also observed that this gene is frequently mutated in clinical samples, but it has yet to be linked to a specific phenotype in vitro (Merker et al., 2013).

DISCUSSION

Given the challenges in finding novel antibiotics that can penetrate the cell wall of bacteria, recent efforts have shifted toward taking a whole cell screening approach to antibiotic discovery. With this shift has come the challenge of target identification of active candidates. In recent years the method of resistance generation coupled with whole genome resequencing has become a dominant approach to identify the targets of M. tuberculosis inhibitors identified using whole-cell screens. Using this approach to try to determine the targets of replicating and non-replicating M. tuberculosis inhibitors that we had previously identified, we have encountered a high frequency of cases in which mutations in activating enzymes were found. It has recently been suggested that resistance generation coupled with whole genome resequencing may bias towards identifying targets that can easily accommodate mutations while preserving protein function (Stanley et al., 2012). This work illustrates a second common outcome in using resistance generation to identify targets in M. tuberculosis, specifically the frequency of mutations encountered in activating enzymes rather than in the actual mechanistic target.

Unlike most of the antibiotics against conventional, rapidly growing bacterial species, a disproportionate number of the antibiotics currently used to treat TB infection are prodrugs, which require metabolism by a bacterial enzyme before having a bactericidal effect. Examples include isoniazid, oxidized by the catalase-peroxidase KatG; pyrazinamide, activated by pncA encoded pyrazinamidase; and ethionamide, activated by the BVMO EthA (DeBarber et al., 2000). It is unclear why so many antibiotics used to treat tuberculosis are prodrugs. One hypothesis is that this phenomenon may be linked to the large number of oxidases encoded in the M. tuberculosis genome (Barry et al., 2000) with at least 49 genes annotated as monooxygenases, oxidoreducatases or cytochrome P450 oxygenases (DeBarber et al., 2000) that may be required for the bacterium to survive in an environment of increased oxidative stress. Among the 49 oxygenases, M. tuberculosis encodes 6 monooxygenases that contain the BVMO motif (Bonsor et al., 2006). While the presence of BVMOs is not uncommon among prokaryotes, with approximately 50% of sequenced microbial genomes containing a BVMO gene, only a few microorganisms contain multiple BVMOs, primarily actinomyces and mycobacteria (Torres Pazmino, 2007). Each BVMO typically has a preference for a certain class of compounds and the multiplicity of BVMOs within mycobacteria may increase the diversity of molecules that can be oxidized within the cell (Torres Pazmino et al., 2010).

In this work we characterized 3 new molecules, that behave like prodrugs in requiring a monooxygenase for activity, that we had previously identified as having activity against both replicating and non-replicating M. tuberculosis. For two small molecules, thiadiazole 1 and thiophene-pyrazole 2, we found evidence that activation by EthA, a BVMO previously described as activating the antitubercular prodrugs ethionamide, isoxyl and thiacetazone, is required for both replicating and non-replicating activity. These molecules do not contain the thioamide or thiourea groups previously described as EthA substrates (Nishida and Ortiz de Montellano, 2011), suggesting that the diversity of molecules that can be activated by EthA is greater than previously described. For the third small molecule thiooxadiazole 3, we found that MymA, a BVMO not previously described as an activating enzyme, is required to oxidize 3 to the corresponding sulfoxide for its replicating and non-replicating activity. We confirmed its enzymatic activity using purified MymA and LC-MS to detect the oxidized products. This is the first example of a compound with antitubercular activity activated by the MymA BVMO.

Surprisingly, our results indicate that MymA can also activate ethionamide and that loss of MymA function confers resistance comparable to loss of EthA function. Loss of MymA function therefore represents a novel ethionamide resistance mechanism observed in vitro. ethA was originally determined to be the ethionamide resistance locus due to the observation that overexpression of EthR, a repressor of EthA, confers resistance to ethionamide in vitro while overexpression of EthA sensitizes to ethionamide (Baulard et al., 2000; DeBarber et al., 2000). Furthermore, recombinant EthA has been shown to activate ethionamide in vitro (Vannelli et al., 2002) and EthA loss of function has been associated with clinical resistance to ethionamide (DeBarber et al., 2000; Morlock et al., 2003). To our knowledge, however, there is no study on ethionamide sensitivity in a mutant with EthA loss of function compared directly to a parent strain with the same genetic background. In this work, we find that compared to the levels of resistance reported for clinical EthA loss of function mutants (Morlock et al., 2003), loss of function of EthA alone, or MymA alone, only confers lower levels of resistance. Loss of function in both EthA and MymA were required for high levels of ethionamide resistance. This observation indicates some redundancy in function and substrate specificity for these two BVMOs and suggests that other mutations, in addition to those in ethA, are required for high-level resistance to ethionamide.

Analysis of whole genome sequencing data from clinical isolates (Casali et al., 2014) shows that unique mymA loss of function mutations are more frequent than in 99.7% of all M. tuberculosis genes and at least as common as other loss of function mutations associated with resistance to drugs. These findings suggest that mymA loss of function mutations are under selective pressure in vivo. We were unable to definitively link mymA mutations with an ethionamide resistant phenotype in the absence of access to these particular strains, potentially because mymA loss of function mutations confer only low-level resistance to ethionamide while loss of function of both MymA and EthA is required to confer the much higher levels of resistance that are detected by conventional, clinical MIC testing. Such additional experiments however may be warranted, given recent interest in developing genotype-based diagnostics for drug resistance in tuberculosis.

In addition to highlighting the complex mechanisms of compound activation present in M. tuberculosis, this work demonstrates that three molecules that we have identified with antitubercular activity and that contain a sulfur (thioether or thiophene) group require activation by a BVMO. At a time when whole-cell high-throughput screening is being utilized to identify new leads against M. tuberculosis, this finding suggests that the selection of sulfur containing compounds may bias lead selection towards a prodrug mechanism and a relatively high frequency of resistance development. Considerations for further development of these compounds should prioritize synthesis and testing of the oxidized forms, with preferential focus on the oxidized forms of such compounds if proven active.

In summary, in this work we highlight the high propensity for thiol-containing substrates to be oxidized in a promiscuous manner by the large number of M. tuberculosis BVMOs, thereby activating the prodrugs. Not only do we show the ability of EthA to activate a broader range of substrates than previously described, but we also identify a new BVMO, MymA, that is capable of activating prodrugs. Further, we demonstrate that MymA plays a role in activating the known tubercular drug ethionamide. Surprisingly, loss of function mutations in MymA or EthA confer comparable levels of resistance. EthA and MymA thus have previously unrecognized overlap in their substrates, including the redundant ability to activate ethionamide. Loss of function of MymA represents a novel mechanism for ethionamide resistance and together, loss of function of MymA and EthA confer higher levels of resistance to ethionamide. The greater significance of mymA mutations in affording clinical resistance to ethionamide, in conjunction with ethA mutations, will need to be evaluated as greater numbers of M. tuberculosis genomes become available with rigorous ethionamide resistance testing and MIC measurements.

Significance

Infection with the bacterium M. tuberculosis remains a major global health problem, with an estimated 9.0 million new infections and 1.5 millions deaths occurring in 2013. While active tuberculosis (TB) infection is curable, successful treatment requires a complex treatment regimen over at least six months. Unlike most of the antibiotics against conventional, rapidly growing bacterial species, a disproportionate number of the antibiotics currently used to treat TB infection are prodrugs, which require metabolism by a bacterial enzyme before having a bactericidal effect. In this work we characterized 3 molecules previously identified as having activity against both replicating and non-replicating M. tuberculosis and found that all three molecules act like prodrugs, requiring activation by Bayer Villiger monooxygenases (BVMOs). For two small molecules, 1 and 2, we found that EthA, a BVMO previously described as activating several antitubercular prodrugs including ethionamide and thiacetazone, is required for both replicating and non-replicating activity. For the third small molecule, 3, we found that MymA, a BVMO not previously described as an activating enzyme, is required to oxidize 3 to the corresponding sulfoxide for its replicating and non-replicating activity. Surprisingly, we found that MymA also activates ethionamide and further, that loss of MymA function confers levels of resistance comparable to loss of EthA function. Loss of MymA function therefore represents a previously unknown mechanism for etshionamide resistance. In addition, this work demonstrates that EthA and MymA have previously unrecognized overlap in substrates and raises the possibility that there may be additional overlap between the six M. tuberculosis BVMOs during in vivo infection.

EXPERIMENTAL PROCEDURES

Bacterial strains and growth conditions

The strain M. tuberculosis H37Rv was used for all experiments. M. tuberculosis was grown at 37°C in Middlebrook 7H9 broth supplem ented with 10% OADC (oleic acid-albumin-dextrose complex), 0.2% glycerol and 0.05% Tween-80 or on Middlebrook 7H10 plates supplemented with 10% OADC, 0.5% glycerol, and 0.05% Tween-80. Carbon starved bacteria were prepared as previously described(Grant et al., 2013). The plasmid used to overexpress ethA in M. tuberculosis was pUVRS3Δtet (a derivative of pUV15tetORm with the tet repressor deleted for constitutive expression). The inducible plasmid pUVRS3 (a derivative of pUV15tetORm) was used for mymA overexpression and pMV762 was used to overexpress virS under the control of the groEL promoter (Ehrt et al., 2005; Lougheed et al., 2014). Inducible constructs were pre-induced with 50ng/mL anhydrotetracycline for 24 hours prior to determination of dose response curves. Additional information regarding the identification of transposon mutant strains is provided in the supplemental methods.

Chemistry

Unless otherwise noted, reagents and solvents and were obtained from commercial suppliers and were used without further purification. Column chromatography was performed using EMD silica gel 60 (230-400 mesh). ¹H-NMR (300 MHz) and ¹³C-NMR (75 MHz) spectra were recorded on a Bruker instrument (XWIN-NMR) and chemical shifts are reported in parts per million (ppm, δ) downfield from tetramethylsilane (TMS) or residual solvent peak. Coupling constants (J) are reported in Hz. Spin multiplicities are described as s (singlet), brs (broad singlet), d (doublet), t (triplet), q (quartet), and m (multiplet). Liquid Chromatography-Mass Spectrometry analysis was conducted on a Agilent Poroshell 120 EC-C18 column (30 mm × 3.0 mm i.d.) eluting with 0.01% formic acid in water (solvent A) and 0.01% in acetonitrile (solvent B) at a flow rate of 1.75 mL/min, using the following gradient: 0.0 min ~ 0.03 min, 5%B; 0.03 min ~ 1.78 min, 5~95%B; 1.78 min ~ 2.28 min, 95%B; 2.28 min ~ 2.30 min, 95~5%B; 2.30 min ~ 2.50 min, 5%B. The mass spectra were recorded in electrospray positive and/or negative ion modes (ESI+ and ESI−) on a Waters ZQ mass spectrometer. Purity of the compounds was assessed from chromatogram (UV₂₅₄) and was >95%. Ethionamide, compounds 1, 2, 3, 3A and 5 were obtained from commercial suppliers. Synthesis for additional compounds is described in supplemental materials.

Generation of resistant mutants

The agar MIC was determined for each compound by plating bacteria on agar containing a dose response of compound in a 96 well plate format. The MIC was defined as the lowest concentration resulting in inhibition of bacterial growth. Resistant mutants were generated by plating 1×10⁹ bacteria on agar containing 2X, 4X and 10X the agar MIC using four independently derived wild-type clones. Colonies were inoculated into liquid media containing 2X the liquid MIC of the compound. These cultures were retested in a liquid MIC assay to confirm resistance and then used to generate genomic DNA for whole genomic sequencing.

Determination of dose response curves, MIC₉₀s and MBC₉₀s

Bacteria were grown to mid-log phase and plated in 96 well plates at OD₆₀₀ = 0.025 in the presence of a dose response of small molecule inhibitors for 2 weeks, and growth was assessed by reading the OD₆₀₀. The minimal inhibitory concentration, MIC₉₀, was defined as the minimum concentration that inhibited growth by 90% relative to the DMSO control. For all dose response curves the points represent the average and error bars represent the standard deviation of technical quadruplicates. All dose response curves were repeated 2 (minimum) to 4 times.

To determine starvation MBC₉₀, carbon-starved bacteria were diluted to OD₆₀₀ = .05 in starvation media as previously described and incubated in the presence of small molecule inhibitors for 2 weeks (Grant 2013). Surviving bacteria were enumerated by plating for CFUs and minimal bactericidal concentration, MBC₉₀ was defined as the minimum concentration at which CFU of treated bacteria was 10% (or less) that of untreated bacteria. CFU experiments were conducted using three technical replicates and repeated a minimum of 2 times.

Analysis of clinical strain sequencing data

Sequence data for clinical strains was downloaded from NCBI (BioProject PRJEB2138). Reads were mapped onto a reference strain of H37Rv (GenBank accession number CP003248.2) using BWA version 0.5.9 (PMID: 19451168). Variants were identified using Pilon version 1.5 (PMID: 25409509) and variant effects were determined using VCF annotator (http://sourceforge.net/projects/vcfannotator/). Gene sequences were then evaluated for nonsense mutations and small indels (<= 10 bp) causing frameshifts.

Highlights.

• Three novel inhibitors of non-replicating M. tuberculosis are characterized.

• Two inhibitors require activation by EthA, a Baeyer-Villiger monooxygenase (BVMO).

• One inhibitor requires activation by the BVMO MymA, a novel activating enzyme.

• Loss of MymA function confers ethionamide resistance, suggesting overlap with EthA.