Mce3R Stress-Resistance Pathway Is Vulnerable to Small-Molecule Targeting That Improves Tuberculosis Drug Activities.

One-third of the world's population carries Mycobacterium tuberculosis (Mtb), the infectious agent that causes tuberculosis (TB), and every 17 s someone dies of TB. After infection, Mtb can live dormant for decades in a granuloma structure arising from the host immune response, and cholesterol is important for this persistence of Mtb. Current treatments require long-duration drug regimens with many associated toxicities, which are compounded by the high doses required. We phenotypically screened 35 6-azasteroid analogues against Mtb and found that, at low micromolar concentrations, a subset of the analogues sensitized Mtb to multiple TB drugs. Two analogues were selected for further study to characterize the bactericidal activity of bedaquiline and isoniazid under normoxic and low-oxygen conditions. These two 6-azasteroids showed strong synergy with bedaquiline (fractional inhibitory concentration index = 0.21, bedaquiline minimal inhibitory concentration = 16 nM at 1 μM 6-azasteroid). The rate at which spontaneous resistance to one of the 6-azasteroids arose in the presence of bedaquiline was approximately 10-9, and the 6-azasteroid-resistant mutants retained their isoniazid and bedaquiline sensitivity. Genes in the cholesterol-regulated Mce3R regulon were required for 6-azasteroid activity, whereas genes in the cholesterol catabolism pathway were not. Expression of a subset of Mce3R genes was down-regulated upon 6-azasteroid treatment. The Mce3R regulon is implicated in stress resistance and is absent in saprophytic mycobacteria. This regulon encodes a cholesterol-regulated stress-resistance pathway that we conclude is important for pathogenesis and contributes to drug tolerance, and this pathway is vulnerable to small-molecule targeting in live mycobacteria.

Cholesterol metabolism plays an important role in the persistence, virulence, and intracellular survival of Mycobacterium tuberculosis (Mtb).[1−5]Mtb is able to survive and replicate inside macrophages[6] by utilizing host-derived nutrients, including cholesterol.[1−3,7]Mtb catabolism of cholesterol provides a source of acetyl-coenzyme A (CoA), pyruvate, and propionyl-CoA, which can be utilized for energy production and as lipid precursors.[8] Catabolism proceeds through β-oxidation of the side chain and oxidative cleavage of the sterol rings.[9] Comparison of transcriptional profiles of Mtb cultured with and without cholesterol identified over 200 genes that are regulated by cholesterol.[2] At least 52 cholesterol-regulated genes are clustered within the Mtb genome and encode the enzymes needed for catabolism.[2,8] Two TetR family regulators, KstR1 and KstR2, control transcription of the majority of these catabolism genes[10,11] and are derepressed by CoA metabolites of the catabolism pathway.[12,13]

Owing to the large number of potential targets in the Mtb cholesterol metabolism pathway, the protein target that would be the most vulnerable to inhibition is not clear.[14] Therefore, we undertook whole-cell phenotypic screening to identify inhibitors of Mtb growth. We reasoned that use of a steroid scaffold would bias the screen to target cholesterol metabolism. We further sought a scaffold with pharmacokinetic properties that would be advantageous for future in vivo experimentation.

For this purpose, we chose the 6-azasteroid scaffold (Figure ). The 6-azasteroids were developed by Glaxo-Wellcome as part of a 5α-reductase inhibitor program for treatment of benign prostatic hyperplasia, but they were subsequently abandoned in favor of 4-azasteroids.[15−18] Several 6-azasteroids were found to be orally bioavailable in rats, dogs, or both and to have low in vivo toxicity.[16] Because of the pharmacokinetic properties of 6-azasteroids and their structural similarity to cholesterol, we chose to screen them as possible inhibitors of cholesterol metabolism in Mtb.

Figure 1

Structures and synthesis of 6-azasteroids. Key intermediate 1 was prepared as described in the literature[16−18] and subjected to further elaboration as shown.

Previously, we demonstrated that 6-azasteroids with a large, hydrophobic R1 side chain and an unsubstituted N6 atom (e.g., 4a, Figure ) are competitive inhibitors of 3β-hydroxysteroid dehydrogenase.[19] However, because this enzyme is not essential for Mtb survival in the mouse or guinea pig models of infection, its relevance as a drug target was questionable.[20] We reasoned that other enzymes in the Mtb cholesterol metabolism pathway might also be susceptible to inhibition by compounds with the 6-azasteroid scaffold. In the present study, we identified several 6-azasteroids with antimycobacterial activity that improve the activity of existing anti-TB drugs, and we established a relationship between activity and side-chain structure. By investigating the mechanism of action and the target of two of the active 6-azasteroids in Mtb, we identified a connection between stress resistance and cholesterol-regulated genes that reside outside the cholesterol catabolic pathway. The inhibitors described herein offer a strategy for combating innate drug tolerance in Mtb and highlight the complex role of cholesterol-regulated genes in Mtb infection.

Results

6-Azasteroids Were Prepared from Intermediate 1

Following a previously established synthetic route,[15−18] we prepared a library of 6-azasteroids from key intermediate 1 (Figure ), which was prepared on a 500 g scale. Briefly, hydrolysis of the methyl ester group of 1 followed by activation of the resulting carboxylic acid and substitution reactions with various primary amines provided amides 3 with more than 20 different substituents at R1. Subsequent deprotection of the ring nitrogen provided compounds 4a–4v. Alkylation of 4a or 4b at N6 provided compounds 5a–7a and 6b. Treatment of 4a with an excess of sodium hydride and iodomethane yielded 8a, which has a methyl group both on the ring nitrogen and at C2. All compounds were purified by column chromatography before testing. Yields from intermediate 1 ranged from 60% to 90%.

6-Azasteroids Inhibit Mtb Growth in Combination with Isoniazid

We initially screened the synthesized 6-azasteroids (Figure ) to determine their minimal inhibitory concentrations (MICs) in a Mtb growth assay. As a carbon source, we used cholesterol solubilized in tyloxapol detergent micelles to ensure that we could detect activity against the cholesterol catabolism pathway. The 6-azasteroid concentrations were varied from 0.15 to 40 μM. We were unable to obtain accurate MICs, but we estimated the values to exceed 40 μM for the compounds that showed at least some activity.

Despite the lack of MICs for most of the 6-azasteroids, we recognized that they might show synergistic activity when combined with isoniazid (INH), a frontline TB drug. Therefore, we screened for inhibition of Mtb growth at a fixed concentration of 6-azasteroid (20 μM) while varying the INH concentration. The potentiation potency was recorded as a fold change of the INH MIC. Of the 35 6-azasteroids tested in combination with INH, seven (4a, 4d, 5a–8a, and 3c) improved the INH MIC at least 16-fold (Figure ). In addition, compound 6b showed moderate potentiation with a 4-fold decrease in the INH MIC. We also tested two 4-azasteroids, finasteride and dutasteride, which inhibit human 5α-reductase and are FDA-approved for treatment of benign prostatic hyperplasia. Neither of the 4-azasteroids showed inhibitory activity, either alone or in combination with INH.

Figure 2

6-Azasteroids improve the efficacy of INH. CDC1551 Mtb in 7H9 medium with cholesterol as a carbon source was incubated with 20 μM 6-azasteroid, finasteride, dutasteride, or DMSO control and varying concentrations of INH (2-fold dilutions) for 14 days at 37 °C. The INH concentration at which no growth occurred was recorded as the MIC. Results are shown as the ratio of the INH MIC in the presence of azasteroid or DMSO to the MIC of INH alone (0.2–0.4 μM). A ratio of 0.5–1 is considered to indicate no change in the INH MIC. The data shown are representative of at least two experiments.

Clear Structure–Activity Relationship Was Observed for the 6-Azasteroids

Among the compounds with R2 = H (4), an aniline side chain at R1 was required. Compounds with side chains derived from aliphatic amines (4n, 4o, 4p, 4r, 4s) or aliphatic amines linked to aromatic groups (4m, 4q, 4t) were inactive. Furthermore, compounds with aniline side chains bearing a single substituent (4e, 4f, 4g), compounds with aniline side chains bearing polar substituents (4b, 4c, 4h, 4i, 4j, 4k, 4l), and compounds with pyridyl-based side chains (4u and 4v) were inactive. Only 4a and 4d, the (2,5-di-t-butyl)phenyl and the (3,5-di-t-butyl)phenyl compounds, improved INH activity. The activity of parent compound 4a was retained upon installation of an alkyl R2 substituent on the ring nitrogen (5a, 6a, 7a). Interestingly, moderate potentiation was observed upon conversion of the inactive 2-t-butyl-5-trifluoromethylaniline-derived compound 4b to N-propyl derivative 6b. Likewise, compound 3c, which has a t-butyloxycarbamate-protected ring nitrogen, was active, whereas the corresponding unprotected compound (4c) was not. Thus, derivatization of the aza ring nitrogen was well tolerated and, in some cases, could convert inactive compounds into potentiators of INH activity.

6-Azasteroid Activity Is Not Specific for Cholesterol Catabolism

Next, we assessed the specificity of the 6-azasteroid scaffold for cholesterol catabolism. We found that when glycerol was used as the carbon source, the seven active 6-azasteroids retained their ability to improve the efficacy of INH. As a negative control, we tested compound 4g, which was inactive with cholesterol as the carbon source, and found it remained inactive in glycerol (Figure S1). Thus, the activity of 6-azasteroids was not carbon-source specific.

Two of the active compounds, 4a and 6a, were selected for further study. Their effect on the rates of CDC1551 Mtb growth was monitored by optical density (OD). As expected from our initial MIC screen, 10 μM 4a or 6a did not inhibit the growth of Mtb on glycerol as a carbon source over the course of 15 days (Figure A).

Figure 3

Azasteroids 4a and 6a inhibit Mtb growth and are bactericidal in combination with INH or BDQ. (A) CDC1551 Mtb was cultured in 7H9 medium with glycerol as a carbon source and was diluted to an OD (600 nm) of 0.2. These cultures were incubated with 10 μM 4a or 6a or with DMSO as a control (two replicates), and the growth of Mtb was monitored by OD (600 nm) for 15 days. Data are representative of at least two experiments. (B) H37Rv(mlux) Mtb was cultured in 7H9/glycerol medium and diluted to approximately 5–10 RLU per well. These cultures were incubated with INH, BDQ, and/or 4a at the indicated concentrations or with DMSO as a control, and the growth of Mtb was monitored by autoluminescence for 9 days. (C) H37Rv(mlux) Mtb was cultured in 7H9/glycerol medium and diluted to approximately 5–10 RLU per well. These cultures were incubated with INH, BDQ, and/or 6a at the indicated concentrations or with DMSO as a control, and the growth of Mtb was monitored by autoluminescence for 9 days. (D) CDC1551 Mtb was cultured in 7H9/glycerol medium and diluted to an OD (600 nm) of approximately 0.002. These cultures were incubated with BDQ and/or 6a at the indicated concentrations or with DMSO as a control. After 2 or 7 days, the cultures were diluted, and the CFUs were determined by serial dilution onto agar plates. Error bars in panels B–D indicate SDs (n = 2). The limit of detection is 10 CFU/mL.

6-Azasteroids Improve the Activity of Other TB Drugs and Drug Candidates

Compounds 4a and 6a were tested as potentiators of the following additional TB drugs and drug candidates: rifampin, pretomanid, bedaquiline (BDQ), clofazimine, pyrazinamide, moxifloxacin, linezolid, and ethionamide. We found that these two 6-azasteroids improved the activity of at least five of the TB drugs in vitro: rifampin, pretomanid, BDQ, pyrazinamide, and ethionamide (Tables and S1A). The 6-azasteroids did not appear to improve the activity of clofazimine, linezolid, or moxifloxacin. BDQ, in addition to INH, was selected for further study because it is used for the treatment of multidrug-resistant Mtb infection.

Table 1

6-Azasteroids 4a and 6a Improve the Efficacy of Multiple TB Drugsa

CDC1551 Mtb or H37Rv(mlux) Mtb in 7H9 medium with glycerol as a carbon source was incubated at 37 °C with 20 μM 4a or 6a or with DMSO control and one of the listed drugs (2-fold dilution) at various concentrations for 14 days (CDC1551) or 6 days (H37Rv(mlux)). The drug concentration at which no growth occurred was recorded as the drug MIC. The ratio of the drug MIC in the presence of azasteroid to the MIC of the drug alone was determined. Ratios of 0.0625–0.125 are indicated by green shading and are considered to reflect potentiation. Ratios of 0.5–1 are indicated by red shading and are considered to reflect no change in MIC. Data are representative of at least two independent biological replicate experiments, which were performed in technical triplicate. MIC values are listed in Table S1A.

Values in this column are MICs for CDC1551 Mtb except for clofazimine.

Experiment performed only with CDC1551 Mtb.

Experiment performed only with H37Rv(mlux) Mtb.

6-Azasteroids Show Synergism with INH or BDQ

Using the H37Rv(mlux) strain, we carried out a checkerboard growth-inhibition assay at various 4a or 6a and INH or BDQ concentrations. As expected, neither 10 nor 20 μM 4a inhibited the growth of H37Rv(mlux) (Figure B). A sub-MIC concentration of INH (0.032 μM) also failed to inhibit growth. However, when Mtb was treated simultaneously with 20 μM 4a and 0.032 μM INH, growth was inhibited. Similarly, at a low concentration (4 μM), BDQ alone inhibited growth by approximately 50%. Importantly, when BDQ was used in combination with 4a, complete inhibition was observed. Likewise, combinations of 6a and 0.032 μM INH or 4 μM BDQ resulted in complete inhibition (Figure C). Growth readings at day 9 were used to prepare an isobolographic plot for INH and 6-azasteroid (Figure S2). Although an accurate 6-azasteroid MIC could not be determined from the plots, the combinations of 4a or 6a and INH were clearly synergistic.

6-Azasteroids Improve Bactericidal Activity

Combining 6a with BDQ (1.8 μM) improved the bactericidal activity of BDQ. Specifically, the combination of 10 μM 6a and BDQ resulted in a 2 log10 CFU improvement in bacterial kill at day 7 of treatment. However, the rate of bacterial kill by BDQ was not increased, as evidence by the 2-day time points (Figures D and S3A). The combination of 4a and INH was also bactericidal (Figure S3B).

6-Azasteroids Are More Potent under Low-Oxygen Conditions than under Normoxic Conditions

We also conducted a checkerboard low-oxygen recovery assay (LORA) of H37Rv(mlux) growth inhibition by 4a or 6a and INH or BDQ at various concentrations (Figure ). Both 4a and 6a were active in the LORA, showing MICs of 11 and 13 μM, respectively. Although INH was not active under low-oxygen conditions (MIC ≈ 400 μM), treatment with INH plus 4a or 6a reduced the MIC at least 2-fold (Figure A). The LORA fractional inhibitory concentration (FIC) indices for INH/4a and INH/6a were 0.75 and 0.69, respectively. BDQ had an MIC of 160 nM in the LORA. Compound 4a or 6a improved BDQ activity in the LORA (Figure B). At 4.4 μM, 4a reduced the BDQ MIC approximately 50-fold (to 3 nM), and 6a reduced the BDQ MIC to 23 nM. The FIC index for both BDQ/4a and BDQ/6a was 0.21, indicating strong synergy.

Figure 4

6-Azasteroids and INH or BDQ interact under low-oxygen conditions. Isobolographic analysis of codrug LORA MIC in H37Rv(mlux) Mtb: (A) INH and 4a or 6a and (B) BDQ and 4a or 6a. The lines connecting 400 μM INH or 0.21 μM BDQ and 13 μM 4a or 6a denote lines of additivity. H37Rv(mlux) Mtb was cultured and diluted to approximately 5–10 RLU per well. These cultures were incubated with INH or BDQ and 4a or 6a at a checkerboard of concentrations (3-fold dilutions) or with DMSO as a control under hypoxic conditions for 10 days. The RLU values for H37Rv(mlux) Mtb after 28 h of normoxic recovery were recorded. The drug concentrations in wells with at least 90% inhibition of growth relative to that in the no-drug control are plotted. Plotted values are averages (n = 3). (C) H37Rv(mlux) Mtb was cultured at approximately 1 × 105 CFU/mL. These cultures were incubated with BDQ, 4a (10 μM), or 6a (10 μM), alone or in combination, at the indicated concentrations under hypoxic conditions for 10 days. The cultures were diluted, and CFUs were determined by serial dilution onto agar plates. Plotted values are averages with SDs (n = 3); *p < 0.05. The limit of detection is 10 CFU/mL.

6-Azasteroids Improve the Bactericidal Activity of BDQ under Low-Oxygen Conditions

The bactericidal activities of 4a and 6a were determined under low-oxygen conditions. The minimum bactericidal concentration of 4a was 80 μM and that of 6a was higher than 80 μM (80% kill). To assess the bactericidal activity of combinations of 6-azasteroids and BDQ, we performed a time-kill assay with 10 μM 4a or 6a in combination with BDQ at concentrations below the minimum bactericidal concentration (Figure C). At day 10 of treatment, the combination of 10 μM 4a and 0.15 μM BDQ resulted in a 3.2 log10 CFU reduction in bacterial kill relative to that in a no-drug control group. In addition, this same combination (10 μM 4a plus 0.15 μM BDQ) resulted in a 1.4 log10 CFU reduction in bacterial kill at day 10 relative to that in the BDQ-only group at a slightly higher BDQ concentration (0.25 μM); that is, 10 μM 4a plus 0.15 μM BDQ was more potent than 1–0.25 μM BDQ alone (Figure C).

6-Azasteroids Retain Potentiation Activity in Cholesterol Catabolism Mutants

Because 6-azasteroids are cholesterol analogues, we determined whether the cholesterol catabolism pathway was required for azasteroid activity. For this purpose, we tested the activity of 6-azasteroids in combination with INH in mutant Mtb strains with disrupted cholesterol catabolism genes. Specifically, we selected strains with mutations in fadA5, chsE4 (fadE26), fadE31, and fadE33 (the transcription of which is repressed by the two main transcriptional regulators of cholesterol catabolism, KstR1 and KstR2)[10,11] and in hsd (which encodes the first enzyme in the pathway).[21] KstR1 regulates the transcription of genes encoding side-chain- and A/B-ring-degrading enzymes, whereas KstR2 regulates the transcription of genes encoding C/D-ring-degrading enzymes. These two regulators are derepressed by CoA metabolites in the pathway.[12,13] We found that none of these five genes were required for 4a to improve INH activity against Mtb grown with either cholesterol or glycerol as the carbon source (Tables and S1B).

Table 2

Mce3R Regulon Is Required for Azasteroid 4a Activitya

The data shown are representative of at least two independent biological replicates and were performed in technical triplicate.

With cholesterol or glycerol as a carbon source, the indicated strain of Mtb in 7H9 medium was incubated with 20 μM 4a or DMSO control and INH at various concentrations (2-fold dilutions) for 14 days at 37 °C.

With glycerol as a carbon source, the indicated species of mycobacteria in 7H9 medium was incubated with 20 μM 4a or DMSO control and INH or BDQ at various concentrations (2-fold dilutions) for 4–7 days at 37 °C.

The drug concentration at which no growth occurred was recorded as the drug MIC. MIC values are shown in Table S1B. The ratio of the drug MIC in the presence of 4a to the MIC of INH or BDQ alone was determined. Ratios of 0.0625–0.125 are indicated with green shading and reflect potentiation. Ratios of >0.125 and ≤0.5 are indicated with orange shading and reflect low potentiation due to low activity of INH against M. marinum. Ratios >0.5 are indicated with red shading and reflect no change in MIC.

Mce3R Regulon Is Required for Potentiation of INH by 6-Azasteroids

Previously, we discovered that the Mce3R regulon encodes FadE17-FadE18, an unusual heterotetrameric acyl-CoA dehydrogenase that is associated with cholesterol metabolism.[22] Moreover, Mce3R regulates the mel2 operon, which is implicated in Mtb persistence in macrophages[23] and in resistance to oxidative stress[24] that is associated with lipid metabolism.[25,26] The regulon is derepressed upon treatment with cholesterol, most likely by a cholesterol metabolite,[2] and expression of the regulon is up-regulated during hypoxia and in the stationary phase.[27] We found that azasteroid 4a no longer improved INH activity when tested against Mtb Mce3R regulon mutants fadE18, melF, and melH (Table ), and this loss of activity was independent of carbon source.

The cholesterol catabolism pathway regulated by KstR1 and KstR2 is found in both saprophytic and pathogenic mycobacteria.[28] In contrast, the Mce3R regulon is conserved in M. marinum and Mtb but not in other pathogens (e.g., M. avium) or in saprophytic mycobacteria (e.g., M. smegmatis). In results consistent with our mutant potentiation data, we found that 6-azasteroid 4a did not improve INH or BDQ activity when tested against M. smegmatis or M. avium grown with glycerol as a carbon source (Table ). However, inhibition of M. marinum growth by BDQ and INH was increased by addition of 4a.

Spontaneous Resistance to 6a Is Mce3R Independent

We examined the frequency with which spontaneous resistance to one of the azasteroids arose (Table ). Specifically, resistant mutants were raised against 6a alone and in combination with BDQ. The reported frequency of in vitro resistance to 20 μM BDQ ranges from 9 × 10–9 to 5 × 10–7 mutations per cell division, and mutations arise predominately in the atpE gene.[29] We found similar frequencies of BDQ resistance in this study. The frequencies of resistant mutant formation upon treatment with 6a in combination with BDQ were comparable to the frequencies of BDQ resistance. Treatment with 6a alone resulted in an approximately 5–10-fold higher frequency of resistance. All the resistant mutants that formed upon treatment of 6a alone or in combination with BDQ had significant alterations in colony morphology. The wild-type colonies were thick, rough, irregular, and yellow, whereas the mutant colonies were thin, small, round, and almost transparent (Figure S4). Twenty mutants were selected, and six of these grew in liquid culture. The mutant colonies unable to grow in liquid culture were restreaked on agar plates, and one more mutant grew. All the resistant mutants grew slowly in liquid culture or on agar plates. The 6a, BDQ, and INH MICs against these seven mutants were determined (Figure S1C). Against three of the mutants (ASR1, ASR4, and ASR5), the MIC of 6a increased more than 4-fold with respect to the MIC against WT-CDC1551, whereas a 2–4-fold increase was observed for the remainder of the mutants. The INH and BDQ MICs against all the 6a-selected mutants were unchanged (1–2-fold increase) compared to their MICs against WT-CDC1551.

Table 3

Frequencies of in Vitro Spontaneous Resistance to 6a and BDQa

	frequency of resistanceb
	BDQ (20 μM)	BDQ (10 μM)	BDQ (0 μM)
6a (400 μM)	7.8 × 10–8	ND	1.1 × 10–7
6a (200 μM)	ND	2 × 10–7	3.6 × 10–6
6a (0 μM)	5.5 × 10–9	3.4 × 10–7

CDC1551 Mtb was plated on 7H11 medium supplemented with 10% OADC containing the indicated concentration of drug(s). Resistant mutants were counted after 3–4 weeks. ND = not detected.

Mutations per cell division.

DNA from six of the resistant mutants was sequenced. Only three mutants (ASR1, ASR2, and ASR3) produced high-quality data; the DNA from the other three mutants was highly fragmented. Twenty-nine nonsynonymous single nucleotide polymorphisms (SNPs), occurring in 25 genes, were identified and confirmed by at least 10 Illumina reads (Figure S5). Of the 25 genes, 6 belong to the PE/PPE family and 3 encode enzymes involved in PDIM biosynthesis.

Because mutant ASR2 showed no significant increase in the 6a MIC, we focused on analyzing mutants ASR1 and ASR3, in which we identified, respectively, 21 and 11 nonsynonymous SNPs, with 3 SNPs that were identical in both mutants. A stop gain occurred at S154 in Rv2940c, which encodes a mycocerosic acid synthase involved in PDIM biosynthesis. In addition, we identified a frameshift insertion at P188 in Rv3467, a nonessential gene, and a recombination of PPE19. No Mce3R gene mutations were identified in the resistant mutants.

As for individual nonsynonymous mutations, we determined that ASR1 carried a nonsynonymous SNP in guaB3 (Rv3410c), which is an essential gene annotated as an inosine-5′-monophosphate dehydrogenase (Figure S5). However, GuaB3 does not perform this dehydrogenase enzymatic function.[30] The guaB3 gene resides in a three-gene operon Rv3409c-guaB3-guaB2, and Rv3409c is required for bacterial survival and growth in macrophages.[31] We also identified frameshift deletions in Rv0678, Rv2552c, and Rv3267 from ASR1. These genes are involved in the regulation of a cell efflux system, aromatic amino acid synthesis, and cell-wall synthesis, respectively. A frameshift deletion was also identified in Rv0204c from ASR3. The disruptions in genes involved in mycobacterial cell-wall biosynthesis are consistent with the phenotypic changes in colony morphology that we observed.

6-Azasteroids Suppress the Transcription of Mtb Stress-Response Genes Including the Mce3R Regulon

RNA-seq gene expression profiles of the wild-type CDC1551 Mtb and the resistant mutants ASR1 and ASR2 were generated after 6 h of exposure to 4a or 6a. Upon treatment of the wild-type with 4a or 6a, genes involved in lipid metabolism, cell-wall synthesis and cell processes, PE/PPE, and information pathways (TubercuList functional classes) were generally down-regulated (>2-fold). The global expression profiles for 4a and 6a treatment were like the profile for BDQ treatment, with >80% of the genes that were down-regulated by BDQ (>1.5-fold) also being down-regulated (>1.5-fold) by 4a.(32) In the absence of any of the test compounds, resistant mutants ASR1 and ASR2 had a common set of 49 up-regulated (>2-fold) genes, including stress-response genes regulated by DosR, PhoP/R, MprA, Crp, and WhiB3.[33]

In addition to the observed general response to drug treatment, six sets of genes that appeared to be specific to the 6-azasteroid mechanism of action were up- or down-regulated by treatment of wild-type CDC1551 Mtb with 4a or 6a (Figures and S6). The magnitude of differential expression of these specific genes in response to treatment with 4a or 6a was much lower in ASR1 than in the wild-type, a result that is consistent with the increased drug MIC.

Figure 5

Differentially expressed transcripts as analyzed by RNA-seq are visualized in heatmaps. The transcriptional profiles of Mtb strains CDC1551 (WT), ASR1, and ASR2 exposed for 6 h to 4a (30 μM) or 6a (30 μM) were compared with the profile of the corresponding untreated group. Gene names are indicated to the right of the heatmap, and strains and treatment conditions are shown on the bottom. The color scales represent log2-fold changes in gene expression in treated groups relative to expression in untreated groups; each colored cell represents the mean of biological triplicates. Red, blue, and black lines denote the echA13-fadE17-fadE18, mel2, and mce3 operons, respectively.

In the wild-type CDC1551 Mtb, the KstR1 and KstR2 regulon genes that encode cholesterol catabolism enzymes were either up-regulated or unaffected (Figure ). In contrast, KstR1 genes not directly related to cholesterol catabolism were greatly down-regulated (up to 25-fold, not shown). Additionally, KstR1-regulated genes in the mce4 operon, which encodes the cholesterol transport system, were down-regulated by 4a or 6a. Genes in the mce1R operon, which is associated with mycolic acid transport, and in the mel1 operon, which encodes putative membrane protein and transglutaminases thought to be important for infection,[23] were greatly down-regulated (>5-fold). In the wild-type, the Mce3R-regulated echA13-fadE17-fadE18 operon was down-regulated by treatment with 4a or 6a, whereas the other two Mce3R-regulated operons were unaffected. In addition, guaB2, which is in the Rv3409c-guaB3-guaB2 operon, was up-regulated by 4a or 6a (Table S4). Gene expression of representative genes from Mce3R, KstR1, and KstR2 regulons and mel1 operon in Mtb CDC1551 exposed to 4a was confirmed by qRT-PCR (Figure S6).

Discussion

We tested 37 azasteroids in MIC assays and identified a set of eight 6-azasteroids that clearly sensitized Mtb to existing TB drugs (Figure ). We observed a clear structure–activity relationship, indicating that the compounds acted on a specific target or targets. The TB agents that exhibited synergy with these 6-azasteroids utilize a wide array of killing mechanisms, disrupting processes ranging from cell-wall biosynthesis to ATP biosynthesis (Table ).

We found that genes involved in cholesterol catabolism were not required for the activity of 6-azasteroid 4a (Table ), despite our previous work demonstrating that 6-azasteroids, including 4a, inhibit the first enzyme in the catabolic pathway, 3β-hydroxysteroid dehydrogenase (hsd),[19] and a body of evidence indicating that cholesterol catabolism is important for Mtb persistence and survival in a mouse model of infection.[1,2] Moreover, the 6-azasteroids were active against Mtb grown on a sugar carbon source, glycerol, as well as against Mtb grown on a cholesterol carbon source (Figure ). Importantly, 6-azasteroids 4a and 6a were active under low-oxygen growth conditions. Taken together, the evidence suggests that 6-azasteroids act on a target that contributes to drug tolerance regardless of carbon source or oxygen level.

Transcriptional,[2] phenotypic,[34] and biochemical profiling[22] have identified genes outside the Mtb cholesterol catabolism gene cluster that are regulated by cholesterol, that are important for growth on cholesterol in vitro, or that have cholesterol-specific structural motifs. Several of these genes are found within a TetR-like transcriptional regulon controlled by the Mce3R repressor.[26,27] Transposon disruption of mce3R significantly inhibits growth of Mtb on cholesterol.[34] However, the genes within the Mce3R regulon are not required for catabolism of cholesterol,[8,9] suggesting that these genes have an alternate role in Mtb survival. Therefore, we investigated the importance of four genes in the Mce3R regulon (fadE18, echA13, melF, and melH) in 6-azasteroid activity. Intriguingly, fadE18, melF, and melH were required for 6-azasteroid potentiation of INH activity. Although echA13, annotated to encode an enoyl-CoA hydratase, was not required, the Mtb genome encodes many other enoyl-CoA hydratases that might compensate for disruption of this gene.

The precise biochemical functions of the proteins encoded in the Mce3R regulon have yet to be established. However, bioinformatics and phenotypic assays allow for tentative assignment of their function. The fadE17, fadE18, and mel2 genes are important contributors to Mtb resistance to various cellular stresses. The entire Mce3R locus is present in M. marinum and provides resistance to reactive oxygen species and reactive nitrogen species.[24] Significantly, the Mce3R regulon is absent in saprophytic mycobacteria,[28] which suggests that the primary role of the regulon is mediating host-derived stresses, such as those encountered by Mtb in activated macrophages. Consistent with this observation, the mel2 operon is required for persistence and dissemination of Mtb infection in mice.[35]

The fadE17 and fadE18 genes encode an acyl-CoA dehydrogenase with an unusual α2β2 heterotetrameric that is characteristic of acyl-CoA dehydrogenases that oxidize cholesterol-derived substrates.[22] The melF and melH genes reside in a single operon (Rv1936-Rv1941) of the mel2 locus, which encodes homologues of the Lux luminescence system. This system catalyzes formation of a fatty acid aldehyde by means of an ATP-dependent process.

The mel2 operon is thought to provide resistance to oxidative stress, which is consistent with the idea that the operon encodes catalytic machinery to generate a fatty acid aldehyde that can remove oxidizing species from the cellular milieu.[24] When Mtb production of ergothioneine, which is a redox couple in Mtb, is disrupted, fadE18 is up-regulated.[36] When Mtb is exposed to a lysosomal soluble fraction prepared from activated macrophages, both fadE17 and fadE18 are induced.[37] In addition, a recent study has shown that cysteine in combination with INH or rifampin can enhance oxygen consumption by Mtb, thereby leading to increased production of reactive oxygen species and sterilization of Mtb cultures.[38] Upon addition of cysteine to INH-treated Mtb cultures, the transcription level of fadE18 increases nearly 4-fold. Treatment of Mtb with BDQ for 48–96 h results in up-regulation of mel2 genes, especially melF, which is up-regulated about 8-fold.[32] These responses indicate that the Mce3R regulon plays a critical role in the Mtb response to oxidative, host-induced, or drug-induced stress.

In contrast to treatment with INH, rifampin, or BDQ, treatment with 6-azasteroid 4a or 6a for 6 h resulted in 3- to 4-fold down-regulation of the Mce3R-regulated echA13-fadE17-fadE18 operon. Although the transcription levels of mel2 genes were unchanged or only marginally changed after 6 h of 6-azasteroid treatment, the response of the mel2 gene to BDQ treatment is likely to be indirect, given the long post-BDQ treatment time (48–96 h) at which up-regulation is observed.[32] Together with our mutant experiment in which fadE18 was required for 6-azasteroid potentiation, the evidence supports a mechanism in which the 6-azasteroids suppress fadE18/Mce3R-dependent activation of a drug-induced stress-resistance pathway.

Mtb spontaneous resistant mutants that were raised against 6a showed a thinner, almost transparent cell-wall phenotype and were difficult to culture. Importantly, we found that Mce3R genes were not directly involved in spontaneous resistance to the 6-azasteroid. Among the genes where nonsynonymous mutations occurred in the resistant mutants, PE/PPE genes were not investigated further as candidates for involvement in the azasteroid-resistance phenotype, because of their high variability in Mtb. In addition, it is unlikely that genes involved in PDIM biosynthesis are directly involved in the azasteroid-resistance phenotype, because the loss of PDIM production is commonly observed during in vitro propagation of Mtb cultures.[39] The remainder of the mutations identified were in genes involved in mycobacterial cell-wall synthesis and efflux systems, which is consistent with a mechanism of resistance that decreases 6-azasteroid uptake, and simultaneously, the growth fitness of Mtb is reduced.

In addition to suppressing Mce3R genes, 6-azasteroids also suppressed the expression of genes in the DosR,[40] PhoP/R,[33] and SigB regulons[41] and in the icl1 operon.[42] All these genes are involved in Mtb stress resistance and the establishment of persistence. There is evidence that inhibitors of DosR genes also increase the activity of INH,[43] although the increase is not as dramatic as that observed for 6-azasteroids. This enhancement of INH activity by the 6-azasteroids indicates that they inhibit a network of Mtb stress resistance and that the disruption of this network leads to the potentiation of TB drugs.

Codrug sensitization is emerging as a useful strategy for treating TB, including multidrug-resistant TB.[44−47] Variations on this strategy include targeting specific drug-activation pathways[46] as well as more general drug-desensitization targets such as the DosRST regulon[43] and thiol stress.[47] Thus, these drug-sensitization strategies are a powerful tool for killing resilient and/or nonreplicating Mtb.

In summary, our work further supports the idea that Mce3R-regulated genes are important for managing the Mtb cellular response to drug-induced stress (Figure ). Moreover, our findings suggest that Mtb depends on cholesterol or its metabolites for activation of stress resistance through the Mce3R regulon (Figure ). Our discoveries that 6-azasteroid suppression of drug tolerance depends on Mce3R-regulated genes and occurs under low-oxygen conditions present an intriguing avenue for further development of codrugs that can improve the efficacy of existing TB drugs in vivo.

Figure 6

Efficacy of 6-azasteroids requires Mce3R-regulated genes. Genes in the Mce3R regulon are required for 6-azasteroid antimycobacterial activity. Cholesterol transcriptionally regulates the Mce3R genes,[2] and this regulon is repressed by Mce3R, a tetR-like repressor protein.

Materials and Methods

Synthesis of 6-Azasteroid Precursor 1 and the Corresponding 17β-Carboxylic Acid

17β-Carbomethoxy-6-t-butoxycarbonyl-6-azaandrost-4-en-3-one (1) was synthesized as reported previously[16−18] and then converted to 17β-carboxy-6-t-butoxycarbonyl-6-azaandrost-4-en-3-one as described in the literature.[17] Both compounds were assessed to be greater than 95% pure by 1H NMR spectroscopy.

General Procedure for Coupling Amines to the 17β-Carboxylic Acid and Subsequent Deprotection of the Ring Nitrogen[17]

17β-Carboxy-6-t-butoxycarbonyl-6-azaandrost-4-en-3-one (30 mg) was suspended in 1 mL of toluene. One drop of DMF and 15 μL of pyridine were added to the solution, which was then cooled in an ice bath, treated with 10 μL of thionyl chloride, and stirred for 1 h. The solution was filtered, and the filtrate was concentrated. The resulting acid chloride residue was dissolved in DCM (5 mL) and treated with the desired amine to give amide 3. Amide 3 was dissolved in DCM (1 mL) and treated with 2 mL of TFA at 20 °C to remove the BOC protecting group from the ring nitrogen. After 2 h, the reaction mixture was poured into saturated NaHCO3 solution; the layers were separated, and the organic layer was washed with brine and dried over Na2SO4. Chromatography using 5% (v/v) MeOH in DCM provided deprotected amide 4.

Procedure for N6 Alkylation of Deprotected Amides 4a and 4b(8,18)

Amides 4a and 4b (30 mg) were dissolved separately in THF (5 mL) and treated with 1.2 equiv of NaH. After stirring for 30 min at 20 °C, the reaction mixture was treated with the desired iodoalkane (1 equiv) for a further 30 min. When the reaction was complete as judged by TLC (approximately 5 h), 30 mL of EtOAc was added, and the resulting solution was washed with water and brine (3× each) and dried over Na2SO4. Chromatography using 5% (v/v) MeOH in DCM provided N6-alkylated amides 5a, 6a, 6b, and 7a. For the synthesis of 8a, amide 4a (30 mg) was dissolved in DMF and treated with 3 equiv of NaH. After 30 min at 20 °C, 5 equiv of iodomethane was added into the reaction, followed by stirring at 20 °C for 2 h to give compound 8a. Active azasteroids 4a, 5a, 6a, 8a, and 3c were resynthesized for verification and were determined to be greater than 99% pure by LC-MS.

Bacterial Strains and Culture Conditions

Mutant and complemented strains used in this study and their sources are listed in Table S2. Mtb strains CDC1551, H37Rv, and H37Rv(mlux), also known as H37Rv(pFCA-luxABCDE), were used as wild-type strains for this study. Mtb strains were cultured at 37 °C either in Middlebrook 7H9 medium (broth) supplemented with 0.2% glycerol or 0.1 mM cholesterol, 0.5% BSA, 0.08% NaCl, and 0.05% (v/v) tyloxapol (7H9/glycerol or 7H9/cholesterol medium) or on 7H11 medium (agar) supplemented with 10% oleate-albumin-dextrose-NaCl (OADC), 0.5% glycerol, and 0.05% Tween 80.

Determination of MICs

MICs were determined by means of a broth microdilution assay.[48] Briefly, cells were grown to mid log phase and then diluted 1000-fold in defined media. The cell suspension was added to a 96-well plate containing compound dilutions to obtain a final volume of 100 μL. Plates were incubated at 37 °C for 14 days, and MICs were determined as the lowest concentration that resulted in complete inhibition of growth by visual inspection or luminescence readouts (mlux). For isobolograms, 5-fold serial dilutions of INH or BDQ were used so that the concentration range in a single plate was sufficient. If the 90% reduction in RLU could not be determined directly, a linear interpolation between two RLU values was used to determine the concentration at 90% inhibition.

LORA[49]

A low-oxygen-adapted culture of H37Rv(mlux) Mtb expressing a luciferase was used for the LORA. H37Rv(mlux) Mtb stored at −80 °C (1 × 105 CFU/mL) was thawed and exposed to 2-fold or 3-fold serial dilutions of test compound diluted in 7H12 broth in black 96-well plates, which were incubated for 10 days at 37 °C in an anaerobic jar under hypoxic conditions created with an Anoxomat system (MART Microbiology). Luminescence readouts were obtained after 28 h of normoxic recovery in 5% CO2. To calculate the MIC, the dose response curve was plotted as percentage growth and fitted to the Gompertz model. The MIC was defined as the lowest concentration inhibiting recovery of the luminescence signal by 90% relative to bacteria-only controls. FIC is defined as the ratio of the MIC of the inhibitor when used in combination to the MIC of the inhibitor alone. The FIC index is the sum of the FICs for the two drugs used. The minimum concentrations that worked in combination were used. The determination of drug MICs for rifampin, INH, linezolid, pretomanid, and BDQ were conducted as positive controls.

LORA Time-Kill Assay

The low-oxygen-adapted Mtb strain H37Rv(mlux) was cultured as described above for the LORA. Compounds 4a and 6a were tested at 10 μM in combination with BDQ at 0.5, 0.25, 0.15, 0.0625, and 0.05 μM. At 0, 7, and 10 days, cultures were recovered in an aerobic environment (5% CO2) for 28 h and then were serially diluted in 7H9 broth and plated on 7H11/OADC agar plates. Drug concentration is diluted at least 300-fold. Plates were incubated at 37 °C, and CFUs were counted after 3–4 weeks. The limit of detection is 10 CFU/mL. Statistical analysis was done using an unpaired Student’s t-test (GraphPad Prism 4).

Aerobic Time-Kill Assay

Mtb was grown at 37 °C to mid log phase and then diluted in fresh medium to 5 × 105 CFU/mL. Test compounds were added at defined concentrations. Aliquots of cultures were withdrawn at specified time points, and either OD (600 nm) (CDC1551) or luminescence (mlux) was recorded. At 0, 2, and 7 days, aliquots of cultures were serially diluted in 7H9 broth and plated on 7H11/OADC agar plates. Drug concentration is diluted at least 300-fold. Plates were incubated at 37 °C, and CFUs were counted after 3–4 weeks. The limit of detection is 10 CFU/mL.

Frequency of Resistance

Mtb mutants resistant to 6a were isolated by means of the procedure reported by Ioerger et al.[50] H37Rv Mtb bacteria were grown at 37 °C to mid log phase and then diluted in fresh Middlebrook 7H9 medium containing ADC-Tween 80 to 5 × 108 CFU/mL. Middlebrook 7H11/OADC agar plates with 6a at 200 or 400 μM with or without 10 or 20 μM BDQ were inoculated with 108, 107, 106, and 105 CFU/plate, and the plates were incubated at 37 °C for 3–4 weeks. Resistance was tested by measuring the MICs of 6a, BDQ, and INH. The frequency of the appearance of resistant mutants was calculated.

Whole Genome Sequencing

Mtb strains CDC1551, ASR1, ASR2, and ASR3 were grown to log phase, and their genomes were extracted with cetyltrimethylammonium bromide and lysozyme as described in the literature.[51] Whole genome preparation, sequencing, assembly, and pairwise analysis were performed as previously described.[52] Briefly, DNA was sheared into ∼20 000 bp fragments using Covaris G-tube spin columns and was end-repaired before being ligated to SMRTbell adapters (Pacific Biosciences). The resulting library was treated with an exonuclease cocktail to remove unligated DNA fragments and was size-selected on a Sage Science BluePippin system to obtain fragments of ≥7000 bp. The P5–C3 sequencing enzyme and chemistry were used to sequence the resulting libraries on the Pacific Biosciences (PacBio) RS II platform. Resulting PacBio sequencing data were assembled using HGAP3 (ver. 2.2.0). For Illumina sequencing, genomic DNA (1 μg) was sheared to achieve ∼200 bp fragments using a Bioruptor Pico sonicator (Diagenode). Library preparation was performed using the end repair, A-tailing, and adaptor ligation NEBNext DNA library prep modules for Illumina from New England Biolabs, according to the manufacturer’s protocol. The resulting libraries were sequenced on the Illumina HiSeq 2500 platform. Illumina reads were then mapped to the curated PacBio assemblies using samtools mpileup[53] to correct PacBio sequencing errors. Genome circularization, curation, and annotation were performed with a custom postassembly pipeline (https://github.com/powerpak/pathogendb-pipeline).[52] Finally, NUCmer (ver. 3.1)[54] was used for aligning mutant genome strains to the PS00103 reference genome to identify genetic variants. Whole genome sequencing data are available in the NCBI BioProject database under accession number PRJNA482894.

RNA-seq Transcriptional Profiling and qRT-PCR

Mtb strains CDC1551, ASR1, and ASR2 were grown to OD ∼ 0.6 and treated with 30 μM 4a or 6a or with vehicle control for 6 h in Middlebrook 7H9 medium supplemented with 0.2% glycerol, 0.5% BSA, 0.08% NaCl, and 0.05% (v/v) tyloxapol. Total RNA was extracted using TriZol, with chloroform back extraction and 70% ethanol precipitation. RNA was purified with an RNeasy kit (Qiagen) with DNase treatment. Total RNA was processed for ribosomal reduction library construction. Libraries were sequenced as single-end 75 bp reads on an Illumina NextSeq500 sequencer following the manufacturer’s protocols (Cofactor Genomics, Inc.). The sequence reads were aligned to the CDC1551 Mtb complete genome (The National Center for Biotechnology Information Database) using STAR (https://github.com/alexdobin/STAR). For each transcript or patch, the reads per kilobase million (RPKM) expression value was calculated for each sample. For each replicate group, the mean and coefficient of variation for each transcript or patch were calculated across the expression values for the samples in that group. These means were considered to be the expression values for the replicate group. P-values were calculated for comparisons between the means of each pair of replicate groups using a Welch’s t-test corrected for false discovery rate by means of the Benjamini–Hochberg procedure. For each comparison, differentially expressed genes were identified as genes with an average normalized count of >100, differential gene expression of >2-fold, and a P-value of <0.05. Three biological replicates were performed. RNA-seq data are available in the NCBI GEO database under accession number GSE118482. Total RNA was reverse-transcribed using PrimeScript RT Master Mix (TaKaRa Bio). The resulting cDNA was used for PCR amplification using iTaq Universal SYBR Green Supermix (Bio-Rad). Relative level of gene expression was calculated by the ΔΔCq method with 16s rRNA as an internal control. Primers are listed in Table S3.