All Escherichia coli strains were grown in Luria-Bertani (LB) broth and agar (Difco). When appropriate, hygromycin and kanamycin were added at 150 and 50 μg/ml into the medium, respectively. Chemically competent E. coli TOP10 strain (Invitrogen) was used for propagation of all plasmids in this study.

Mycobacterium bovis BCG was purchased from ATCC (Pasteur strain ATCC 35734), while MTB Erdman, H37Rv and CDC1551 strains were a kind gift from Novartis Institute for Tropical Diseases (NITD), Singapore. All these strains and their derivative were grown at 37°C in Middlebrook 7H9 liquid media (Difco) supplemented with ADS (0.5% bovine serum albumin-fraction V, 0.2% dextrose, 0.085% saline) or with OADC (0.5% bovine serum albumin-fraction V, 0.2% dextrose, 0.085% saline, 0.005% oleic acid, and 0.0004% catalase) enrichment as indicated. For growth on agar, 7H11 (Difco) containing 0.05% Tween 80 (Tw) and 0.2% glycerol supplemented with OADC was used. Appropriate antibiotics [80 μg/ml hygromycin (Roche), 20 μg/ml kanamycin (Sigma)] were added when required.

The ethA/R locus was deleted by double homologous recombination as described previously (Bardarov et al., 2002). Briefly, ∼800 bp-long DNA regions flanking the ethA/R locus in M. bovis BCG and MTB strains were amplified by polymerase chain reaction (PCR) and cloned directionally into vector pYUB854 (Bardarov et al., 2002) with the hygromycin-resistance cassette (hyg) lying in between the flanking regions. The lacZ ORF and promoter region from the pGoal17 plasmid (Parish and Stoker, 2000) were then cloned into the unique PacI site of the pYUB construct. To prepare electrocompetent mycobacteria, 10 mL cultures were grown to mid-log phase [optical density at 600 nm (OD₆₀₀) of 0.4–0.6] in 7H9-ADS in the absence of glycerol, followed by culture in fresh 7H9-ADS supplemented with 1.5% glycine (Sigma) 1 day prior to electroporation. On the day of electroporation, mycobacteria cells were washed thrice with 0.05% Tween80 before final resuspension in 1 ml of the same medium. 200 μl of electrocompetent mycobacteria suspension were electroporated (2.5 kV, 800 Q, 25 mF) with 2 μg of treated recombinant plasmid subjected to ultraviolet (UV) treatment as previously reported (Parish and Stoker, 2000) and plated onto Hygromycin-containing 7H11 medium supplemented with 40 μg/ml X-gal. White hygromycin-resistant clones were selected after 16 days incubation at 37°C and screened by PCR with a set of internal ethA/R primers (5′-TCC AGC GGT TTT CCG CGG TC-3′ and 5′-TCC CGG TGC GCC ACA TGT TC-3′).

To complement the ethA/R KO mutants, the 2.2 kb ethA/R full-length locus was PCR amplified (Supplementary Table S1), cloned into the multiple cloning site of the integrative vector pMV306 (Stover et al., 1991) and introduced into the genome of all ethA/R KO mutants via electroporation as described above. The resulting transformants were plated onto kanamycin-containing 7H11 agar. After 16 days incubation at 37°C, kanamycin-resistant colonies were PCR screened using the internal ethA/R primers as mentioned above.

The ethA KO and mshA KO mutants were obtained by double homologous recombination as described above, whilst complement ethR and complement mshA mutants were constructed in selected MTB strains using vector pMV306. Overexpressing ethR mutants were constructed using the multicopy replicative plasmid pMV262 (Stover et al., 1991). All plasmid constructs were synthesized and validated using oligonucleotides listed in Supplementary Table S1.

Construction of the double mshA/ethA/R (m/e) double KO mutants was obtained by first unmarking the ethA/R KO mutants in order to remove the hygromycin cassette which was inserted in place of the ethA/R locus. To do so, the ethA/R KO mutants were transformed with plasmid pYUB870 which harbors a χδ-resolvase (tnpR)-encoding gene (Bardarov et al., 2002), thereby allowing resolvase-mediated cleavage of the hygromycin cassette. Gentamicin-resistant clones were first selected after incubation at 31°C for the resolvase activity and selected again on 7H11 containing 2% sucrose after incubation at 39°C. Loss of the hygromycin cassette and retained deletion of the ethA/R locus were verified by PCR. Successfully unmarked ethA/R KO mutants were then used for deletion of mshA by classical double homologous recombination as described above, followed by complementation of mshA under the hsp60 promoter and using pMV306 plasmid (Supplementary Table S1).

1–3 μg of genomic DNA (gDNA) was digested with SacI (Promega), separated on a 1.5% agarose gel and treated as previously reported (Sambrook and Russell, 2001). DNA was transferred onto a Millipore Immobilon-Ny+ Transfer membrane and UV cross-linked. For detection of ethA/R and ethA KO mutants, a 415 bp DIG-labeled probe was amplified using a set of primers that bind approximately 1.5 kb downstream of ethR, 5′-TGA GTT TAG TTG GGA CCT AGG CC-3′and 5′-CTA GAG TCA CAT CAG AAA CAT TTG A-3′. For detection of mshA and double mshA/ethA/R KO mutants, a 600 bp DIG-labeled probe was amplified using a set of primers that bind immediately upstream of mshA, 5′-CCC GTC CAC TCT GAA ATG CTC G-3′ and 5′-ATC AAC CCT GAA CCG TCA TCG TGT-3′. Probe amplifications were done via PCR according to the manufacturer’s instructions (DIG-labeling kit, Roche). Hybridization and signal detection were performed using a detection kit (Roche) according to the manufacturer’s protocol. EasyHyb (Roche) was used as the pre-hybridization and hybridization solutions, and CSPD (Roche) was used as the detection substrate for chemical luminescence.

Bacterial drug susceptibility assays were performed in 7H9 media supplemented with either ADS or OADC (as indicated in the figure legends) as described previously (Kurabachew et al., 2008). ETH (Sigma), ISO (NITD), and TAC (NITD) were dissolved in 90% DMSO, whilst INH (Sigma) was dissolved in ultrapure water for stock solutions. Using a broth microdilution method, INH and ETH were twofold serially diluted (0.02–5 and 0.3–80 μM respectively) in 7H9 medium in 96-well flat bottom plates. Log phase MTB cultures were diluted in 7H9-OADC or 7H9-ADS medium to obtain ∼2 × 10⁵ colony forming units (CFUs)/ml. 100 μl of the bacterial inoculum were added to each well-containing an equal volume of drug suspension. The plates were then incubated for 5–7 days at 37°C. On the 5th or 7th day (as indicated), OD₆₀₀ was measured using a Biorad iMark Microplate absorbance reader and curves were plotted using PRISM to determine the minimal inhibitory concentration (MIC₅₀) values. The MIC₅₀ is defined as the lowest concentration of drug that is required to inhibit 50% growth of the MTB strain compared to growth obtained in drug-free 7H9 medium. Drug assays were performed thrice independently. After determining the MIC₅₀ values, 50 μl of the bacterial suspensions incubated at 1x, 2x, and 4x MIC₅₀ were plated onto 7H11 agar plates. CFU were enumerated after 16 days incubation at 37°C and the CFU-based MIC₉₀ range (in μM) was defined as the range of drug concentrations within which the number of CFUs compared to the drug free control was reduced by 90% (1 log reduction).

To generate spontaneous ETH^R clones from the Erdman ethA/R KO mutant, a protocol was adapted from both Luria and Delbrück (1943) and Mathys et al. (2009). Since the Erdman ethA/R KO mutant displays low resistance to ETH, optimization was performed to determine the appropriate ETH concentration range that would minimize background and false positive. Eventually, 360, 420, and 480 μM ETH were selected as appropriate ETH concentrations.

To generate spontaneous ETH^R mutants, an exponential phase 7H9-ADS liquid culture (OD₆₀₀ of 0.6–0.8) of the ethA/R KO mutant was used to inoculate three individual flasks of 7H9-ADS medium at an initial OD₆₀₀ of 0.005–0.01. The cultures were incubated for 1–2 weeks. An estimated 10⁸ and 10⁷ bacteria were then plated onto 7H11 plates containing selected ETH concentrations. ETH^R colonies were picked and grown in liquid medium for gDNA extraction and glycerol stocks. The MIC of ETH was determined for these clones, and extracted gDNA was used for whole genome sequencing (WGS).

Unless otherwise stated, bars represent means + standard deviations (SD) and averages were compared using a bidirectional unpaired Student’s t-test with a 5% significance level (^∗p ≤ 0.05).

The ethA/R locus was removed and substituted in place by the hygromycin-resistant cassette hyg in M. bovis BCG and three MTB backgrounds, namely Erdman, H37Rv and CDC1551 (Supplementary Figure S1). The minimum inhibitory concentration of drug required to inhibit 50% growth of mycobacteria compared to drug-free control (MIC₅₀) was measured for the respective ethA/R KO mutant strains in order to determine their level of resistance to ETH and to the two other thiocarbamide-containing drugs ISO and TAC. INH was included as well as a negative control since bio-activation of INH is not EthA/R-dependent. Consistently, all ethA/R KO mutants displayed parental susceptibility to INH (Table 1). In contrast, and as expected, removal of the ethA/R locus in M. bovis BCG led to complete resistance to ETH (Table 1 and Supplementary Figure S2). However, complete resistance to ETH was not seen with the MTB ethA/R KO mutants whereby a dose-dependent killing could still be observed as evidenced by the sigmoidal MIC curves obtained (Supplementary Figure S2). The MIC₅₀ values measured with the ethA/R KO MTB mutants were increased by 2–3X compared to their respective parental strains (Table 1). Furthermore, all the ethA/R KO mutant strains displayed increased resistance to TAC, with MIC₅₀ values increasing by 2–8X (Table 1). As for ISO, with the exception of Erdman ethA/R KO mutant which was found slightly more resistant with a twofold increase in the MIC₅₀ value compared to its parental counterpart, no significant changes in MIC₅₀ were observed with the other mutant strains. Importantly, parental susceptibility to the three drugs was restored in each strain upon re-introduction of the ethA/R locus (Table 1 and Supplementary Figure S2).

To confirm the MIC₅₀ data obtained, the bacterial suspensions incubated with ETH concentrations that corresponded to 1x, 2x, and 4x MIC₅₀ were plated for CFU enumeration. The ETH concentration range within which a 90% reduction of the bacteria load (also equivalent to 1 log) compared to drug-free control was determined and called CFU-based MIC₉₀. Consistent with the MIC₅₀ data, while BCG ethA/R KO displayed full resistance to ETH, dose-dependent killing was observed with all three MTB KO strains over the range of ETH concentrations assayed (Supplementary Figure S3). The MIC₉₀ range of ETH on CDC1551 ethA/R KO mutant was increased by 8–16 fold compared to its parental and complemented counterparts (Table 2). In contrast, a mild twofold increase in ETH MIC₉₀ was observed with ethA/R KO H37Rv and Erdman strains compared to their parental and complemented counterparts. Although these three MTB strains belong to the same Lineage 4 (Euro-American), H37Rv and Erdman are considered as laboratory strains whereas CDC1551 is regarded as a “clinical” strain. The extensive passages in vitro are likely to result in genetic changes that may lead to distinct phenotypes both in vitro and in vivo when compared to CDC1551, including the drug susceptibility profile (Betts et al., 2000; Coscolla and Gagneux, 2010).

Altogether, these data show that although the ETH MICs values obtained for ethA/R KO MTB strains were higher than those measured with the corresponding parental and complemented strains, ETH susceptibility and dose-dependent drug response to ETH were retained. In particular both MTB Erdman and H37Rv strains remained very susceptible to ETH upon deletion of ethA/R locus with 3- and 2-fold increases in their MIC values respectively. The retained susceptibility to ETH despite removal of ethA/R in the three MTB strains suggests that the pro-drug ETH still gets activated into its bactericidal form in an EthA-independent manner, thus supporting the existence of an alternative bio-activation pathway for ETH in MTB.

To further investigate the possible existence of an alternative pathway of ETH bio-activation in MTB, we questioned whether the transcriptional repressor EthR which negatively modulates the ethA/R locus, would also modulate this alternative pathway. Indeed, using programs available online², EthR was predicted to bind to a number of promoter regions in addition to the ethA/R intergenic region. Thus it is conceivable that in WT MTB, EthR may repress the expression of another gene that is involved in ETH bio-activation. To address this hypothesis, the ethR open reading frame (ORF) was over-expressed in all three WT MTB strains under the control of the constitutive strong promoter hsp60 and using the multicopy replicative plasmid pMV262. Real-time PCR analysis confirmed the over-expression of ethR (8–16 fold increase) in comparison to the parental strains. Over-expression of ethR in these three strains was expected to lead to the strong repression of ethA as well as any other genes that may be negatively regulated by EthR. Therefore, should an alternative EthR-dependent pathway of ETH bio-activation exist in MTB, susceptibility to ETH would be affected when ethR is over-expressed. However, the ethR over-expressing MTB strains retained ETH susceptibility in a dose-dependent manner (Supplementary Figures S4A–C) and displayed CFU-based MIC₉₀ values similar to those obtained with the ethA/R KO mutants (Table 2). These results thus support the existence of an alternative pathway of ETH bio-activation in MTB and indicate that this pathway is likely to be EthR-independent.

To confirm this observation, ethA single KO mutants were generated in the three MTB backgrounds. We reasoned that should EthR negatively repress the alternative ETH bio-activation pathway, ethA KO mutants would display full or increased resistance to ETH compared to their corresponding ethA/R KO counterparts. Results indicated that the CFU-based MIC₉₀ ranges obtained with ethA single KO mutants were similar to the values obtained with the ethA/R KO mutants (Table 2), thus further supporting that the alternative pathway of ETH bio-activation is not negatively regulated by EthR.

Finally, ethR was re-introduced into the Erdman ethA/R KO mutant using integrative plasmid pMV306. We postulated that expression of ethR in the ethA/R KO mutant would only impact the susceptibility to ETH if the alternative pathway of ETH bio-activation is negatively regulated by EthR. However, ETH susceptibility was retained in this strain (Supplementary Figure S4D) and the MIC₉₀ range was similar to that obtained with the ethA/R KO counterpart (Table 2).

Altogether, these findings strongly support the existence of an EthA/R-independent alternative pathway of ETH bio-activation in MTB strains.

In order to identify the molecular players that are involved in the alternative pathway of ETH bio-activation in MTB, spontaneous mutants that were highly resistant to ETH were generated from the Erdman ethA/R KO mutant. Briefly, Erdman ethA/R KO bacteria were plated onto agar plates containing a range of ETH concentrations (360–480 μM). Two independent rounds of ETH^R mutants were generated. Individual ETH-resistant (ETH^R) colonies were picked and sub-cultured. Drug susceptibility assays were then carried out for these ETH^R clones to confirm ETH resistance. No MIC₅₀ values could be derived for these mutants as all mutants grew uninhibitedly even at the highest concentration of ETH (80 μM) tested (data not shown).

Since InhA is the downstream target of both ETH and INH (Banerjee et al., 1994; Vilchèze et al., 2005), a pre-screen was first conducted to exclude all the ETH^R spontaneous mutants harboring mutation(s) in the inhA gene. Previous literature has shown that the INH MICs obtained with most MTB isolates harboring mutations in inhA promoter region and ORF were usually low (>1 μg/ml) (Ramaswamy et al., 2003); hence, susceptibility to INH was assayed by growing the ETH^R clones in the presence of INH at 1 μg/ml (or 7.29 μM) and turbidity was visually assessed after 10 days incubation. For the great majority of the clones (∼90%) obvious turbidity was observed, indicating that these clones were INH^R thereby suggesting the presence of a mutation in their inhA locus. However, seven clones displayed impaired growth in the presence of 7.29 μM INH and further independent drug susceptibility assays indicated that these clones displayed varying levels of resistance to INH with MIC₅₀ values ranging between 0.625 and 1.25 μM, with the exception of one clone displaying an unusually high level of INH resistance (5 μM). Sequencing revealed that there was no mutation in the inhA ORF and promoter region for each of the seven clones analyzed (data not shown).

The seven ETH^R clones were then subjected to WGS and compared against the Erdman ethA/R KO parental genome. Numerous mutations came up upon WGS, but the list was restricted to insertion/deletions (INDELs) and non-synonymous SNPs (NS-SNPs), and further refined by eliminating NS-SNPs that resulted in conservative amino acid changes (i.e., change to an amino acid with similar physiochemical properties). Strikingly, NS-SNPs were found in mshA for six out of the seven clones analyzed (Supplementary Table S2). Apart from mshA, mutations were found in other genes candidates as well, a large proportion of which are involved in metabolism pathways (galE3, cobD, plcB, and pks5). Another group of gene candidates included gltx, recD, and topA involved in the transcriptional, translational, and nucleotide assembly pathways (Supplementary Table S2). The remaining identified genes Erdman_1484 and Erdman_0263 could not be categorized under any pathways and these genes remained unclassified.

Since six out of the seven clones analyzed by WGS displayed a mutation in the mshA locus, we decided to further investigate the contribution of this gene in ETH bio-activation. MshA is a glycosyltransferase involved in mycothiol biosynthesis, an equivalent to the antioxidant glutathione in mammals, and was found to be essential for ETH susceptibility in MTB (Vilchèze et al., 2008). Furthermore, mshA mutations were shown to confer varying levels of co-resistance to INH and ETH in M. bovis (Vilchèze et al., 2011). Consistently, a recent study reported that 45.6% of ETH^R MDR-TB isolates harbored a mutation in mshA (Rueda et al., 2015). However, the role of the glycosyltransferase MshA in ETH-mediated killing has yet to be deciphered. The identification of ETH^R ethA/R KO Erdman mutants with mutations in mshA here further supports the role of MshA in ETH killing efficacy and suggests that it may be in an ethA/R-independent fashion.

To validate the role of MshA in ETH bio-activation, the mshA locus was deleted from WT MTB Erdman, H37Rv and CDC1551 strains, thus generating mshA single KO mutants. Furthermore, mshA was also deleted in the corresponding ethA/R KO mutants, giving rise to mshA ethA/R (m/e) double KO mutants. All clones were verified by Southern blot (Supplementary Figure S5) and subsequently complemented by introducing the hsp60-mshA construct into the mycobacterial genome using the integrative plasmid pMV306. Our choice to express mshA under hsp60 promoter was driven by the fact that the mshA promoter has yet to be characterized and delineated.

Growth defect was previously reported for MshA-deficient MTB mutants (Vilchèze et al., 2008). Consistently, mshA and m/e KO mutants displayed a significant growth defect in 7H9 supplemented with ADS (Figures 2A–C). In contrast, all the KO strains grew well in 7H9-OADC (Figures 2D–F). However, it is worth to note that mshA and m/e KO mutants in the CDC1551 background still displayed a slight but significant and reproducible growth defect compared to their respective parental counterpart (Figure 2F).

Given that mshA and m/e KO mutants are unable to grow on 7H9-ADS, drug susceptibility assays involving these mutants were performed in 7H9-OADC. However, since the initial MIC values obtained in this work were generated in 7H9-ADS (Supplementary Figure S2), we re-established their MIC in 7H9-OADC. The MIC₅₀ values obtained for INH were comparable regardless of the medium supplement (OADC or ADS) for the three MTB backgrounds (WT and ethA/R KO) (Supplementary Table S3). Similar MIC values for ETH were also observed with the three WT MTB strains in both types of medium. However, the ETH MICs increased by twofold for all ethA/R KO mutants upon replacing ADS with OADC (Supplementary Table S3). Despite the greater resistance to ETH in the presence of OADC, dose-dependent drug susceptibility to ETH was nevertheless retained (data not shown). Therefore, our data indicated that although some variation exists between the MIC values when using ADS or OADC as supplement, ETH susceptibility was still retained in the ethA/R KO MTB mutants.

Drug susceptibility assays were conducted with the mshA single KO and mshA ethA/R (m/e) double KO mutants. The slower growth rates observed with the CDC1551 mshA and m/e KO mutants (Figure 2F) were taken into consideration by reading the plates at day 7 post-setup.

Remarkably, the combined absence of both ethA/R and mshA loci in all three MTB backgrounds abrogated ETH susceptibility, rendering the m/e double KO mutants completely resistant to ETH (Supplementary Figure S6 and Table 3). MIC₅₀ values could not be obtained for these mutants since mycobacteria grew uninhibitedly even at the highest concentration of ETH (80 μM) tested (Table 3). These observations were supported by the lack of a dose-response curve with all three m/e KO mutant strains (Supplementary Figure S6). Re-introduction of mshA in the m/e KO mutants restored ETH susceptibility to levels similar to those observed with their respective ethA/R KO counterparts (Table 3). Thus, these data indicate that the combined removal of mshA and ethA/R loci in MTB leads to complete resistance to ETH, and further confirms the involvement of mshA in ETH killing efficacy.

Interestingly, deletion of mshA alone led to MIC₅₀ values either comparable (Erdman background) or greater (H37Rv and CDC1551 backgrounds) than those obtained with their ethA/R KO counterparts (Supplementary Figure S6 and Table 3). This observation thus suggests that mshA is at least as critical as ethA/R for ETH killing efficacy. Complementation with mshA only partially restored the levels of ETH susceptibility, possibly due to the usage of hsp60 promoter in place of its native promoter.

Additionally, and consistent with previous report (Vilchèze et al., 2008, 2011), the mshA KO mutants displayed mild increased resistance to INH with a twofold (Erdman) and fourfold (H37Rv and CDC1551) increase of the MIC₅₀ values compared to the parental strains (Table 3).

Together, these data confirm the contribution of mshA in ETH and (to a lower extent) INH killing efficacy in MTB. In addition, the complete resistance to ETH upon deletion of mshA from ethA/R KO mutants suggests that the role of MshA in ETH killing efficacy is independent on EthA/R-mediated ETH bio-activation.