Cell wall permeability screening identifies benzothiazole BT-08

It has recently been identified that BTP15 and ethoxzolamide (Fig 1A) could reduce mycobacterial virulence by affecting the ESX-1 secretion system (Rybniker et al, 2014; Johnson et al, 2015). Exploiting their structural similarities, we synthesized a set of benzothiazole derivatives (Table S1) to explore the possibility of developing active compounds based on this scaffold. These compounds could improve the permeability of the mycobacterial cell envelope, ultimately enhancing the effectiveness of traditional antibiotics. The model organism M. marinum was used for an ethidium bromide (EtBr) uptake assay, assessing membrane permeability. EtBr enters the cell and binds to bacterial DNA, resulting in higher fluorescence, indicating increased cell wall permeability (Rodrigues et al, 2008). As a control strain, we employed M. marinum overproducing MspA (+mspA), a porin known to increase outer membrane permeability (Niederweis et al, 1999; Stahl et al, 2001). M. marinum expressing mspA showed, as expected, higher uptake of EtBr compared with the WT strain (Fig 1A). Likewise, some of the WT M. marinum cultures incubated with the test compounds showed higher EtBr uptake compared with the non-treated WT strain (Fig 1A, Table S1). Compound BT-08 was the most prominent, showing a more than 26-fold increase in the fluorescent signal compared with the non-treated strain (Fig 1A). Note that none of the tested compounds inhibited the growth of M. marinum at the tested concentration of 10 μM, as determined by optical density (OD₆₀₀) measurements in 7H9 medium supplemented with ADS (albumin–dextrose–saline) and tyloxapol. To confirm these results, we repeated the EtBr uptake assay using various concentrations of BT-08 (Fig 1B). The increase in EtBr uptake was concentration-dependent, and even the lowest tested concentration of compound BT-08 (1.3 μM) showed a higher signal as compared to the non-treated WT strain, whereas the maximal activity of BT-08 was seen at 10 μM.

BT-08 increases mycobacterial cell envelope permeability in a dose-dependent manner

To study the effects of compound BT-08 in more detail, we adapted the previously described resazurin-based microtiter plate assay (REMA) to investigate the transport of the dye resazurin across the mycobacterial cell wall (Palomino et al, 2002). M. marinum was cultured in the presence of several concentrations of BT-08 and afterward transferred into a 96-well plate before the resazurin dye was added. The dye resazurin is imported into the cells and reduced during aerobic respiration into a fluorescent product resorufin (Palomino et al, 2002). Generally, the fluorescent signal of resorufin can be detected after 6–12 h of incubation with slow-growing mycobacteria. However, when M. marinum was grown in the presence of BT-08, we already detected the resorufin signal after 1 h of incubation. This effect followed a dose–response and was stronger with increasing concentrations of BT-08 over time (Fig 1C). Next, we evaluated our set of 21 benzothiazole-based compounds in the same assay, which demonstrated that compound BT-08 shows the strongest phenotype (Fig S1), which is in line with the data of the EtBr assay (Fig 1). In addition, compound BT-05 showed a higher signal compared with the positive control strain (+mspA). Our results indicate that both EtBr and resazurin are able to access the mycobacterial cell much more efficiently when these cells are cultured in the presence of compound BT-08.

BT-08 exhibits anti-mycobacterial activity ex vivo and in vivo

To investigate the effect of the benzothiazoles on mycobacteria during infection, we used the previously described M. marinum–zebrafish embryo infection model (Habjan et al, 2021). Zebrafish embryos were infected via yolk injection with M. marinum, expressing a red fluorescent protein (tdTomato). 1 d post-infection, the infected embryos were incubated with 10 μM of the test compound for 3 d. The efficacy of the treatment was evaluated by measuring the red fluorescent signal, which corresponds to the mycobacterial burden within the infected zebrafish embryos (Stoop et al, 2011). When we tested our initial set, only compound BT-08 showed a statistically significant (P < 0.0001) reduction in mycobacterial load in the zebrafish, as compared to the DMSO-treated control group of embryos (Table S1). In a follow-up experiment, compound BT-08 showed a dose-dependent efficacy in the M. marinum–zebrafish infection model, significantly reducing the bacterial burden starting at 3 μM (P < 0.0001) (Fig 1D). It is of particular interest that only compound BT-08 demonstrated activity in EtBr and resazurin uptake assays, as well as efficacy in zebrafish infection studies. This suggests a potential correlation between increased membrane permeability and observed activity in the zebrafish model. Although not extensively diverse, our library (BT-01–BT-22) was comprised of closely related analogs to compound BT-08. However, other than BT-08, none showed significant activity during in vitro and in vivo assays (Figs 1A and S1, Table S1). This observation underscores high specificity and selectivity required for the compound’s activity. Because BT-08 was the only compound displaying in vivo efficacy and caused the strongest phenotype during the EtBr and resazurin uptake assays, we decided to further focus on BT-08.

As mentioned previously, BT-08 had no effect on the bacterial growth of M. marinum in culture (Fig 1E), which stands in contrast with the in vivo results. We hypothesized that the effect of the compound on the bacterial cell wall is only critical in vivo, where the innate immune system of the zebrafish is an additional anti-bacterial factor. Next, we investigated whether this discrepancy between in vitro and in vivo activity translates to clinically relevant mycobacterial species M. tuberculosis. Likewise, the compound BT-08 did not inhibit the M. tuberculosis H37Rv growth in culture (Fig 1E), whereas we could demonstrate a dose-dependent reduction in intracellular M. tuberculosis when we treated M. tuberculosis-infected THP-1 macrophages. Treatment with 30 μM of BT-08 partially protected THP-1 macrophages from bacterial-induced lysis (Fig 1F).

Compound BT-08 did not inhibit the in vitro growth of other tested fast-growing mycobacteria and some G+ and G- species (Table S2). Thus, the effect of compound BT-08 could only be observed during host infection experiments, and it seems specific to mycobacteria, although we cannot rule out the possibility that other species are sensitive to BT-08 during infection. Because we hypothesized that BT-08 acts as a membrane-perturbing agent, we further explored the compound’s safety profile. BT-08 did not show toxicity toward zebrafish embryos, murine-derived macrophage cell line RAW 264.7, or human-derived monocyte THP-1 cell line during cytotoxicity experiments and infection studies (Table S3). In addition, BT-08 was not hemolytic toward sheep red blood cells (Table S4), hence demonstrating specificity toward slow-growing mycobacterial cells.

BT-08 in vitro anti-microbial activity is detergent- and media-dependent

Even though we observed the activity of BT-08 during infection studies, we were puzzled by the fact that it showed no in vitro activity by itself. All in vitro experiments up to this point were performed in 7H9 medium supplemented with ADS and the detergent tyloxapol. Although this is a rich standard medium for slow-growing mycobacteria, we wondered whether a defined minimal medium with different detergents would affect the bacteria differently because the medium-dependent activity of anti-microbial compounds has been reported previously (Pethe et al, 2010; Franzblau et al, 2012). First, we replaced the detergent tyloxapol with Tween-80. Interestingly, compound BT-08 prevented bacterial growth of M. marinum in the 7H9 medium supplemented with ADS and Tween-80 (Fig S2). We next compared the standard 7H9 medium supplemented with ADS and Tween-80 with Sauton’s minimal medium supplemented with Tween-80 and Hartman’s de Bond (HdB) medium supplemented with Tween-80 (Fig S3). We observed that BT-08 is most active against M. marinum in the HdB medium with Tween-80 (Fig S3). Consequently, we continued the in vitro experiments using the HdB medium and Tween-80.

BT-08 potentiates the activity of high molecular weight antibiotics

Because compound BT-08 facilitates the uptake of EtBr and resazurin dye into M. marinum cells, we investigated whether this effect extends to antibiotics. To clarify this, we selected two high molecular weight antibiotics, rifampicin (RIF) and vancomycin, as alterations in the mycobacterial cell envelope can enhance the susceptibility of mycobacteria to these two antibiotics (Healy et al, 2020). Moreover, the targets of RIF and vancomycin have distinct locations in the cytoplasm or periplasm, respectively. Drug-to-drug interactions in vitro are generally reported as the fractional inhibitory concentration (FIC) index (FICI). A FICI below 0.5 indicates synergy, whereas the FICI value between 0.5 and 1 represents an additive effect. First, we investigated the effect of BT-08 on the activity of RIF or vancomycin using checkerboard assays and 7H9 medium supplemented with tyloxapol, because BT-08 increased uptake of EtBr and resazurin in this medium. The addition of BT-08 resulted in a marked change in M. marinum sensitivity to the two antibiotics; for both vancomycin (Fig S4A) and RIF (Fig S4B), the MIC₉₀ was improved fourfold compared with the single-drug treatment. The same assay was used for M. tuberculosis H37Rv, where the shift for vancomycin was fourfold (Fig S4C), and for RIF, twofold (Fig S4D) when BT-08 was added. Because compound BT-08 alone does not inhibit the growth of M. marinum or M. tuberculosis in this media, the calculation of the FICI value was not possible. Thus, we repeated experiments using the HdB medium with Tween-80. BT-08 did not show sufficient activity in the HdB medium with Tween-80 in M. tuberculosis; therefore, we could not determine the FIC index. However, for M. marinum, the calculated FIC index for BT-08 and vancomycin was 0.38, and for BT-08 and RIF, 0.5, indicative of synergistic interactions for both tested drug combinations (Table S5). Thus, the data suggest that compound BT-08 also facilitates the uptake of RIF and vancomycin inside the cell, increasing their activity.

Next, we investigated whether we could confirm a similar synergistic effect in vivo using M. marinum-infected zebrafish embryos. Indeed, the combination of RIF and BT-08 was over 100-fold more active than the single agent at the same concentration (Fig 2A and B). This remarkable in vivo synergy was confirmed using various concentrations of both agents (Fig 2A). Inspired by these findings, we then tested several antibiotics, varying in their molecular weight, mechanism of action, and activity against mycobacteria (Fig 2C, Table S5). Interestingly, several combinations tested in M. marinum in vitro were shown to be effective, with a FICI value indicating synergistic or additive interactions (Fig 2C, Table S5). Compounds with molecular weights above 800g/mol, such as RIF, vancomycin, and polymyxin B, showed synergistic interactions, whereas compounds with a molecular weight below or close to 800g/mol showed additive effects (Fig 2C, Table S5). Notably, the compounds with the lowest molecular weight from our test set showed no interaction with BT-08. The correlation between the synergistic effect and a higher molecular weight of tested antibiotics was significant (*P = 0.0315) (Fig 2D) and suggests improved delivery of antibiotics inside the cells, which is typically a bottleneck for bulky compounds.

BT-08 optimization resulted in benzothiazole BT-37 with improved activity against M. tuberculosis

To optimize the compound BT-08 further, we explored the chemical space of the benzothiazoles (Table S6). All synthesized derivatives were screened for their activity against M. marinum in the M. marinum-infected zebrafish model. In earlier experiments, we observed that with REMA in the HdB medium with Tween-80, we can observe and evaluate growth inhibition of compound BT-08 (Fig S3). Thus, we evaluated compounds from the benzothiazole library for their in vitro activity using the HdB medium with Tween-80 by REMA (Table S6).

We mainly investigated two parts of the BT-08 benzothiazole-core molecule—the six-position of the benzothiazole cycle and the substitution in the side benzyl group, named “A” and “B,” respectively (Fig 3A). The introduction of hydroxy- (BT-25), as well as methyl (BT-27), substituents at the six-position led to the toxicity of compounds in the zebrafish infection model. Only compounds with linear alkyl chains, BT-30 and BT-31, were active in the zebrafish model and demonstrated acceptable MIC₉₀ values against M. marinum, whereas other compounds with different functional substituents were inactive. For the next SAR evaluation, we decided to keep the 6-ethoxy group. We observed that adding an alkyl chain to the two-position of the phenyl ring (BT-38 and BT-39) resulted in a complete loss of potency in both assays. On the contrary, BT-40 with a prop-2-yn-1-yloxy substituent showed activity in the zebrafish model. We also found that introducing a fluorine atom at the two-position (BT-42) resulted in activity in both assays. However, when introducing a chlorine atom (BT-41), we noticed a loss of activity in the zebrafish model. Of the synthesized derivatives, only BT-37 with a 2-methoxy group demonstrated good activity in both models. The addition of two methoxy groups (BT-46 and BT-47), as well as a methoxy group and fluorine atom (BT-48 and BT-49), resulted in the acceptable activity, with BT-46 and BT-48 being the most interesting. However, introducing three methoxy groups led to a complete activity loss in both models, surprisingly, except for BT-52.

Overall, several derivatives showed significantly improved in vitro and in vivo activity in M. marinum as compared to the parent compound BT-08 (Table S6, Fig 3B). The three best compounds, BT-37, BT-46, and BT-48, were further tested in the macrophage–M. tuberculosis infection model (Fig 3C–E). Interestingly, all three compounds showed dose-dependent intracellular activity by reducing the amount of viable M. tuberculosis within macrophages (Fig 3C–E). In addition, the compound BT-37 was the most successful in protecting macrophages from bacterial-induced lysis (Fig 3C). Thus, we decided to continue with compound BT-37. This molecule showed no cytotoxicity toward RAW 264.7 macrophages, THP-1 monocytes, and zebrafish embryos (Table S3) and did not lyse sheep red blood cells (Table S4). Compound BT-37 did not inhibit the in vitro growth of selected fast-growing mycobacteria and some G+ and G- species (Table S2). Moreover, using the EtBr assay, we confirmed that the compound BT-37 increases membrane permeability in M. marinum, as in the case of BT-08 (Fig S5). In addition, BT-37 increased EtBr uptake in M. tuberculosis mc²6206 strain, suggesting a conserved phenotype between these two species (Fig S6). We further confirmed that BT-37 synergizes with RIF in M. marinum-infected zebrafish embryos (Fig S7); however, the effect was less prominent compared with the BT-08 (Fig 2A).

Mutations in MMAR_0407 (Rv0164) cause resistance to BT-37

To identify the molecular target of BT-37, we raised spontaneously resistant mutants of M. marinum. Our approach was based on culturing M. marinum in the presence of a sublethal concentration of BT-37 in the HdB medium and gradually increasing the BT-37 concentration with every passaging step. However, the M. marinum WT strain did not develop resistance, even after prolonged incubation. Therefore, we employed a strain missing the endonuclease nucS, previously shown to have an increased mutation rate in Mycobacterium smegmatis by 100-fold compared with WT (Castañeda-García et al, 2017). We observed the growth of single M. marinum ΔnucS isolates selected during several passages with more than 30-fold higher MIC₉₀ as compared to the parental ΔnucS strain (Fig 4A). Four isolates were chosen for confirmation experiments in dose–response assays. Although two isolates were completely resistant to the tested compound concentrations (ΔnucS-R2 and ΔnucS-R3), the isolate ΔnucS-R1 showed an intermediate phenotype. We also identified one sensitive isolate (ΔnucS-R4) (Fig 4A) that had been subjected to the same number of culturing steps and, therefore, was used as a control for mutations not involved in resistance. The genomes of the four isolates ΔnucS-R1-4, and the ΔnucS parental strain were sequenced.

The BT-37–resistant strains showed several mutated genes compared with the M. marinum ΔnucS reference strain (Table S7). Among them, isolates ΔnucS-R2 and ΔnucS-R3, which displayed high resistance to BT-37 (Fig 4A), carried mutations in mmar_2080 and mmar_4794. However, these mutations were absent in the resistant isolate ΔnucS-R1, leading us to exclude them from further investigation despite their potential impact on susceptibility. Instead, we focused on identifying mutations common to all three resistant isolates. In all three resistant strains (ΔnucS-R1, ΔnucS-R2, and ΔnucS-R3), we found mutations in genes mmar_1438, mmar_1347, and mmar_0407 (Table S7). Notably, the mutations in genes mmar_1438 and mmar_1347 were also present in the BT-37–sensitive control strain ΔnucS-R4. Thus, we hypothesize that mutations in mmar_1438 and mmar_1347 arose during the continuous bacterial passaging and are unrelated to the resistance to compound BT-37, which leaves mmar_0407 as our prime candidate.

Interestingly, the resistant isolates that showed complete resistance, that is, ΔnucS-R2 and ΔnucS-R3 (Fig 4A), shared an identical mutation in mmar_0407, resulting in an amino acid change H73Y (Table S7), whereas the isolate ΔnucS-R1 (Fig 4A) contains a different mutation at a nearby codon, resulting in G69S (Table S7). This could explain the differences in the resistance profile of these isolates (Fig 4A). The gene mmar_0407 encodes a small hypothetical protein that is conserved across mycobacteria and has an ortholog Rv0164 (TB 18.5) in M. tuberculosis (Marmiesse et al, 2004). The proteins of M. marinum and M. tuberculosis share 85.43% sequence identity. Transposon mutagenesis studies indicate that this gene is essential in both M. marinum and M. tuberculosis (Griffin et al, 2011; Dejesus et al, 2017), which would be in line with a potential drug target.

We next showed that the identified BT-37–resistant M. marinum mutants are cross-resistant to other benzothiazole derivatives from our library, such as compounds BT-08 and BT-46 (Fig S8A), which exhibit low MIC against M. marinum WT (Fig S8B). The resistant strain was also investigated in zebrafish embryo survival assays, where zebrafish embryos are infected with a high bacterial load of M. marinum and embryo survival is monitored over several days upon treatment with the test compound (Fig 4B and C). Comparison of the lethality of WT or resistant M. marinum strains using the Kaplan–Meier survival test revealed no significant difference in the embryo mortality rate. This observation suggests that the resistant mutants do no exhibit changes in virulence in this model. Although treatment with BT-37 significantly delayed the death of zebrafish embryos infected with the M. marinum WT strain compared with the non-treated group (Fig 4B), the compounds were ineffective against infection with the resistant mutant (Fig 4C), thus suggesting that the compound’s activity in vivo is anti-bacterial and not host-directed. Moreover, the resistant strain no longer exhibited synergistic activity between BT-37 and RIF or vancomycin when tested using in vitro checkerboard assays (Fig S9A–D), indicating that the synergistic effect is linked to the gene mmar_0407 (rv0164). In the EtBr uptake assay, both the resistant strain treated with compound BT-37 and the untreated resistant strain exhibited a comparable phenotype to the WT strain (Fig S10). This suggests that the presence of the compound alone may not suffice to permeabilize the membrane. Rather, it must effectively target MMAR_0407.

BT-37 is targeting MMAR_0407 (Rv0164)

To confirm that the mutations in mmar_0407 are linked to BT-37 resistance, we overexpressed the WT mmar_0407 gene, the mutated gene mmar_0407 H73Y, and the M. tuberculosis ortholog rv0164 using the replicative vector pMN016 in the M. marinum WT strain. We then assessed the susceptibility of these overexpressing strains to BT-37 (Fig 4D). The overexpression of mmar_0407 led to a threefold shift in the MIC₉₀ (Fig 4D), likely because of increased production of the target protein. On the contrary, the overexpression of mmar_0407 with the H73Y mutation resulted in complete resistance to BT-37, providing further evidence of its involvement in the compound’s resistance (Fig 4D). Interestingly, the overexpression of the M. tuberculosis ortholog rv0164 also resulted in complete resistance (Fig 4D).

To investigate this further, we reduced the overexpression of genes rv0164 and mmar_0407 by integrating the genes into the genome using the giles integration site targeted by vector pML1357. The resulting strains were both sensitive to the compound; however, even with reduced expression levels, the expression of the M. tuberculosis ortholog (giles::rv0164) still caused a higher level of resistance to compound BT-37 than mmar_0407 (giles::mmar_0407) (Fig 4E).

In an alternative approach, we used CRISPR/Cas9 interference (CRISPRi) to construct a conditional knock-down strain of mmar_0407 (mmar_0407_KD), regulated by the addition of anhydrotetracycline (ATc) (Meijers et al, 2020; Wong & Rock, 2021). We tested the strain in the checkerboard assay, using ATc to gradually reduce the mmar_0407 expression in a dose-dependent manner. The mmar_0407_KD strain was more sensitive in the presence of ATc to compound BT-37 as compared to the non-induced strain (Fig 4F).

To directly prove the binding between Rv0164 and the inhibitor, we used a thermal shift assay. We assessed Rv0164’s stability at various temperatures with and without the inhibitor. Proteins have characteristic denaturation temperatures that can change with substrate or inhibitor binding. Rv0164-HA was expressed in Escherichia coli, with DMSO or BT-37 (10 μM). We tested temperatures from 45 to 70°C, blotted lysates on a membrane, and stained with Ponceau dye (Fig 4G). In both BT-37 and DMSO samples, soluble protein decreased as temperature rose. However, with BT-37, Rv0164-HA stayed soluble up to 67°C, whereas the DMSO-treated protein denatured at 54°C. We also probed for DnaK, a chaperone protein, which showed similar heat resistance in both samples. These results confirm that BT-37 alters Rv0164’s heat stability, likely by binding and stabilizing the protein. This supports MMAR_0407 (Rv0164) as the target for these benzothiazole compounds.

Molecular docking simulations predict the binding of BT-37 to the hydrophobic pocket of Rv0164

Lastly, we sought a comprehensive understanding of how benzothiazole inhibitors interact with the target protein. To achieve this, we used the AlphaFold protein folding prediction system to predict the structures of Rv0164 and MMAR_0407 (Jumper et al, 2021; Varadi et al, 2022). The resulting structures exhibited remarkable similarity, with a sequence identity of 85%, demonstrated by an RMSD score of 0.158 Å upon structure alignment (Fig 5A and B). Notably, these structures consist of seven parallel beta sheets and three alpha helices, with very high similarity to the crystal structure solved for the ortholog in M. smegmatis (RMSD 0.722 Å) (Zheng et al, 2018). Next, we aimed to identify potential binding pockets in the Rv0164 structure using the software Fpocket (Le Guilloux et al, 2009; Schmidtke et al, 2010), which employs a geometry-based approach to locate suitable empty spaces. This analysis revealed a substantial pocket situated centrally within the protein’s structure (Fig 5C). This accessible pocket, originating from the protein’s exterior, presents a potential binding site for ligands. Subsequently, we performed docking simulations of Rv0164 and compound BT-37 using the docking software HADDOCK2.4 (Dominguez et al, 2003; Van Zundert et al, 2016). To this end, we generated 50 conformations of BT-37 and then docked these ligands into the predicted Rv0164 structure, resulting in 200 models clustered based on their RMSD score and ranked according to the HADDOCK score. All top 10 clusters contained models predicting BT-37 inside the hydrophobic pocket of Rv0164. Notably, the best-scoring model of the best cluster according to the HADDOCK score predicted the compound in close vicinity to the highlighted amino acids within the binding pocket, including H81 and G77, which correspond to amino acids H73 and G69 of MMAR_0407 (Fig 5A and D). These residues are of particular interest because we have shown that mutations G69S and H73Y in MMAR_0407 cause resistance to BT-37 (Table S7). Thus, our predicted model suggests that the corresponding residues in Rv0164 are important for BT-37 to bind the target. Taken together, we confirmed Rv0164 (MMAR_0407) to be the target of our new benzothiazole compounds by genetic, protein-based, and in silico binding studies.

Reagents and compounds

All commercial chemicals and reagents were solubilized and stored according to the manufacturer’s recommendations. Ampicillin sodium salt, d-cycloserine, erythromycin, ethambutol, ethidium bromide, ethionamide, fluorescein sodium salt, gentamycin, kanamycin sulfate, penicillin G sodium, polymyxin B, resazurin sodium salt, RIF, nisin from Lactococcus lactis, tetracycline, and vancomycin were all purchased from Sigma-Aldrich. Bedaquiline, linezolid, macozinone, pretomanid, spectinomycin, streptomycin sulfate salt, and SQ109 were all purchased from MedChem Express. Anhydrotetracycline was purchased from Thermo Fisher Scientific, hygromycin from Roche, and fusidic acid sodium salt from Merck. Compounds from the benzothiazole-core library were stored as stock solutions (10 mM) in DMSO at −20°C.

Bacteria and cell lines

The bacterial strains used in this study are listed in Table S8. M. marinum strains were grown on 7H10 agar (Difco) plates with 10% ADS (0.5% BSA, 0.2% dextrose, 0.085% sodium chloride) at 30°C or cultured in the following liquid media: (i) Middlebrook 7H9 medium supplemented with 10% ADS, 0.2% glycerol, and 0.05% Tween-80 or 0.02% tyloxapol; (ii) Hartman’s de Bond liquid medium (HdB) (Smeulders et al, 1999) supplemented with 0.5% glucose, 0.5% glycerol, and 0.05% Tween-80 or 0.02% tyloxapol; and (iii) Sauton’s medium (Atlas & Parks, 1993) supplemented with 0.05% Tween-80, at 30°C. M. tuberculosis H73Rv, Mycobacterium abscessus, and M. smegmatis were cultured in 7H9 medium supplemented with 10% ADS, 0.2% glycerol, and 0.02% tyloxapol, or 7H10 agar plates supplemented with 10% ADS at 37°C. The M. tuberculosis mc²6206 strain was cultured in HdB media supplemented with 0.5% glucose, 0.5% glycerol, 0.05% Tween-80, pantothenic acid 24 μg/ml and leucine 50 μg/ml to allow growth of the auxotrophic strain. E. coli, Bacillus subtilis, Klebsiella pneumoniae, and Acinetobacter baumannii were cultured in Luria–Bertani (LB; Difco) medium or LB agar plates at 37°C. When appropriate, the antibiotics hygromycin (50 μg/ml) or kanamycin (25 μg/ml) were added to the growth media. THP-1 human monocytes (ATCC TIB-202) were cultured in RPMI medium with GlutaMAX (Gibco) supplemented with 10% FBS at 37°C with 5% CO₂. RAW 264.7 murine macrophages (ATCC TIB-71) were cultured in DMEM with GlutaMAX (Gibco) supplemented with 10% FBS at 37°C with 5% CO₂.

Ethidium bromide uptake assay

The ethidium bromide (EtBr) uptake assay was performed as described previously (Boot et al, 2017). The growth medium used in experiments with M. marinum strains in Fig 1A and B was 7H9 medium supplemented with 10% ADS, 0.2% glycerol, and 0.02% tyloxapol, whereas in Figs S5 and S10, the growth medium used was the HdB medium supplemented with 0.5% glucose, 0.5% glycerol, and 0.05% Tween-80. The experiment with M. tuberculosis mc²6206 strain used in Fig S6 was performed in HdB media supplemented with 0.5% glucose, 0.5% glycerol, 0.05% Tween-80, pantothenic acid 24 μg/ml, and leucine 50 μg/ml to allow growth of the auxotrophic strain. The strains were pregrown until the mid-logarithmic phase, inoculated (OD₆₀₀ = 0.1) in specified media with the addition of the compound at indicated concentrations, and grown for an additional 72 h. The cells were harvested by centrifugation (3,000g, 10 min) and washed in PBS supplemented with 0.02% tyloxapol. Bacteria were resuspended in PBS with 0.02% tyloxapol and distributed with a final OD₆₀₀ of 0.8 in a transparent, round-bottom, 96-well microtiter plate. EtBr was added with a final concentration of 5 μg/ml per well. Fluorescence was measured every 3 min at 30°C, using a BioTek plate reader (Synergy H1), bottom reading mode, and excitation 300 nm/emission 605 nm. The measurements were taken in triplicates. As a positive control, WT M. marinum M^USA transformed with plasmid pSMT3-mspA (Ates et al, 2015) was used. The EtBr uptake was compared between the compound-treated and non-treated (WT) samples after 60 min. The ratio between compound-treated and non-treated samples was used as a readout.

Resazurin uptake assay

M. marinum was grown in the 7H9 medium supplemented with 10% ADS, 0.2% glycerol, and 0.02% tyloxapol for 3 d at 30°C in the presence of the compound at the indicated concentration. Bacterial cells were harvested by centrifugation (3,000g, 10 min) and washed in PBS supplemented with 0.02% tyloxapol. Next, 160 μl of washed bacteria with a final OD of 1 was added to each well of a 96-well plate. The resazurin solution was prepared by mixing resazurin sodium salt (0.025% [wt/vol] in Milli-Q) and 20% Tween-80 in a ratio of 3:1, and 20 μl of the mixture was added to each well. The resazurin dye conversion was measured every 3 min at 30°C, using a BioTek plate reader (Synergy H1), bottom reading mode, and excitation 560 nm/emission 590 nm.

Bacterial susceptibility assays

The resazurin microtiter plate assay (REMA) was used to determine minimal inhibitory concentrations (MIC) of the compound toward mycobacterial species (Palomino et al, 2002). Selected compounds or antibiotics were diluted in the bacterial growth medium as twofold serial dilutions in a 96-well plate. Bacteria were routinely grown until the mid-logarithmic phase, harvested by centrifugation (3,000g, 10 min), washed in PBS supplemented with 0.02% tyloxapol, resuspended in the growth medium, and added to the well plate containing compound dilutions, to achieve the final OD₆₀₀ of 0.001 per well. The growth media used in the assays using M. marinum were the HdB medium supplemented with 0.5% glucose, 0.5% glycerol, and 0.05% Tween-80, except for experiments in Fig 1E and experiments with other mycobacteria where 7H9 medium supplemented with 10% ADS, 0.2% glycerol, and 0.02% tyloxapol was used. The plates were sealed with parafilm and incubated for 4 d at 30°C for M. marinum, 1 d at 37°C for M. smegmatis, 2 d at 37°C for Mycobacterium abscessus, and 6 d at 37°C for M. tuberculosis. After incubation, the resazurin solution containing resazurin sodium salt (0.025% [wt/vol] in Milli-Q) and 20% Tween-80 (ratio 3:1) was added to each well, and plates were further incubated. When the color conversion of the dye was observed, the fluorescence was measured using a BioTek plate reader (Synergy H1), bottom reading mode, and excitation 560 nm/emission 590 nm. The MIC of compounds toward E. coli, Bacillus subtilis, K. pneumoniae, and A. baumannii was determined by tracking changes in the OD₆₀₀. Bacteria were grown overnight, then freshly diluted in LB medium, and grown to the mid-logarithmic phase. Selected compounds were diluted in LB medium as a twofold dilution in a 96-well plate. Bacterial cells were washed in PBS and added as an OD₆₀₀ of 0.001 per well in a 96-well plate containing compound dilutions. Plates were sealed and incubated at 37°C with 3 mm continuous linear shaking in a BioTek plate reader (Synergy H1), and the bacterial growth was tracked by measuring the OD₆₀₀ every 15 min. All compounds were tested in duplicates or triplicates. The data of each 96-well plate were normalized to DMSO-treated wells (100% viability) after the background subtraction (medium only). The MIC₉₀ values represent the lowest concentration of the compound that results in 90% growth inhibition.

Zebrafish (Danio rerio) husbandry

All zebrafish experiments in this study were performed using transparent casper (roy^a9/a9;nac^w2/w2) (White et al, 2008) zebrafish (D. rerio) embryos. Adult fish were kept in recirculating tank systems at the Amsterdam Animal Research Center of the Vrije Universiteit University according to standard protocols (zfin.org). All protocols followed the international guidelines on the protection of animals used for scientific purposes specified by the EU Directive 2010/63/EU, which allows zebrafish larvae to be used up to the moment of free-living.

Zebrafish (D. rerio) embryo infection studies

Injection stocks of M. marinum strains carrying a plasmid pMS2-tdTomato were prepared in PBS with 20% glycerol, aliquoted, and stored at −80°C. Before use, the injection stocks were diluted with the green fluorescent dye fluorescein (2.5 μg/ml in PBS) to visualize and control the injection process. The number of injected bacteria was determined by plating the injection volume of bacterial suspension on 7H10 agar plates containing appropriate antibiotics, followed by counting CFU. Infection of zebrafish embryos was performed using an automated microinjection system (Life Science Methods BV) as described previously (Spaink et al, 2013). Briefly, zebrafish embryos were infected 1 h post-fertilization at 2–32 cell stage with M. marinum mixed with fluorescein (2.5 μg/ml in PBS). Each embryo was infected with 1 nl containing 80–150 CFU of M. marinum, whereas for survival experiments, ∼1,000 CFU was injected into each embryo. Successfully infected embryos were incubated overnight at 31°C in E3 medium (5.0 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl·2H₂O, 0.33 mM MgCl_2·6H₂O) supplemented with 0.3 mg/liter methylene blue until treatment with test compounds. 1 d post-infection, embryos were divided into treatment groups of 12–15 embryos per well in 12-well plates and incubated with the test compounds that were diluted in fish water (60 μg/ml instant ocean sea salts) and incubated at 28°C. The survival rate was determined daily based on the functioning of the embryos’ heart and blood circulation. Determination of the bacterial load in infected zebrafish embryos was performed 3 d after the treatment. Embryos were anesthetized in 0.02% (wt/vol) buffered 3-aminobenzoic acid methyl ester (pH 7.0) (Tricaine; Sigma-Aldrich), and the bacterial load was monitored with an Olympus IX83 fluorescence microscope (4x objective magnification, Hamamatsu ORCA-Flash 4.0 camera) at specific wavelengths (excitation/emission: 470/519 nm; 550/610 nm). Obtained images were analyzed using CellProfiler 3.19 (Broad Institute, Cambridge, USA) with a custom-made pipeline to count and quantify pixel intensity within the embryos. Integrated red fluorescence intensity per embryo was used as a readout for bacterial burden. Image acquisition and image analysis were automated. The effect of drug treatment in infected zebrafish embryos was analyzed and depicted using GraphPad Prism version 9.0.0 (GraphPad Software Inc.). Each data point represents an integrated red fluorescence intensity signal from a single zebrafish embryo. The signal from the non-infected embryos was used to set the signal threshold to 1. Data were log₁₀-transformed to achieve normal distribution, and furthermore, statistical analysis was done as a one-way ANOVA, and Dunnett’s multiple comparison test was used, where the signal from the DMSO-treated control sample was compared with each treatment group. Significance is indicated as follows: *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Survival curves for zebrafish survival experiments were generated using Kaplan–Meier survival tests.

Zebrafish (D. rerio) embryotoxicity studies

Fertilized zebrafish embryos were collected and kept overnight at 28°C in E3 medium supplemented with 0.3 mg/liter methylene blue. 1 d post-fertilization (dpf), zebrafish embryos were distributed as 10–12 embryos per well in 12-well plates and treated with compounds diluted in fish water. Zebrafish embryos were incubated at 28°C for 5 d with the test compound at indicated concentrations, and the morphology and mortality of zebrafish embryos were monitored daily.

Generation of spontaneous compound-resistant M. marinum mutants

To generate spontaneously BT-37–resistant M. marinum mutants, we used the hyper-mutating M. marinum ΔnucS strain (Izquierdo Lafuente et al, 2024). We grew M. marinum ΔnucS in the HdB medium supplemented with 0.5% glucose, 0.5% glycerol, and 0.05% Tween-80 and increasing concentrations of compound BT-37, starting from 0.3x MIC to 10x MIC over six culturing passages. Single colonies were obtained by streaking cultures on solid plates. The resistance to BT-37 was determined by testing the susceptibility of strains to BT-37 using the REMA. Genomic DNA extraction of the parental M. marinum ΔnucS strains, three BT-37–resistant isolates, and one susceptible isolate was done using phenol/chloroform/isoamyl alcohol extraction as described previously (Mve-Obiang et al, 2001). Whole-genome sequencing of genomic DNA was outsourced to Beijing Novogene Bioinformatics Technology Co., Ltd. (Novogene) using Illumina sequencing technology. Generated reads were aligned to the reference genome of M. marinum M^USA (NC_010612.1) and compared with the parental M. marinum ΔnucS strain using the software QIAGEN CLC Genomics Workbench 12 (QIAGEN).

Cytotoxicity in macrophages

Compounds were prepared as twofold serial dilutions in the appropriate medium. For assays with THP-1 monocytes, the RPMI GlutaMAX with 10% FBS medium was used, whereas for RAW macrophages, the DMEM GlutaMAX with 10% FBS medium was used. Cells were seeded as 2.5 × 10⁴ cells/well in 96-well plates containing compound dilutions, and the plates were incubated for 3 d at 37°C with 5% CO₂. After incubation, resazurin sodium salt (0.025% [wt/vol] in PBS) was added to each well, and plates were incubated at 37°C with 5% CO₂. When color conversion was observed (after ∼4 h), fluorescence intensity corresponding to the metabolically active cells was measured using a BioTek plate reader (Synergy H1), bottom reading mode, and excitation 560 nm/emission 590 nm. All compounds were tested in duplicates. Non-treated samples represented 100% viability.

Macrophage infection studies

The compound’s intracellular activity was investigated in macrophage infection studies, as described previously (Habjan et al, 2021). Infection stocks of M. tuberculosis H37Rv transformed with pTetDuo were prepared in RPMI GlutaMAX with 10% FBS as the infection medium, supplemented with 20% glycerol, and stored at −80°C. All incubation steps were performed at 37°C, 5% CO₂. All solutions were prepared in RPMI GlutaMAX with 10% FBS. THP-1 human monocytes were seeded into black 96-well plates (Ibidi) as 10⁵ cells/well and incubated with phorbol 12-myristate 13-acetate (PMA) at 25 ng/ml for 48 h to induce differentiation into macrophage-like cells. Next, macrophages were washed and infected with M. tuberculosis–pTetDuo at a multiplicity of infection (MOI) of 5. After a 3-h incubation, gentamycin (50 μg/ml) was added for 1 h, to remove extracellular bacteria. Meanwhile, serial dilution of test compounds was prepared in separate 96-well plates. Then, gentamycin solution was removed from wells and replenished with compound’s solution. The plates were incubated for 4 d, and then, anhydrotetracycline (ATc) solution to a final concentration of 100 ng/ml was added to the wells followed by a 24-h incubation. Afterward, the medium was removed and the fixating solution (3.2% [wt/vol] PFA in PBS) was added to the wells for 30 min at room temperature and later replaced with quenching/staining solution (0.1 M glycine, 0.2% [wt/vol] Triton X-100, and Hoechst dye at 1:500 in PBS) for 1 h in the dark. All wells were washed twice with PBS before being imaged using an Olympus IX83 fluorescence microscope with a 20x objective magnification and a Hamamatsu ORCA-Flash 4.0 camera. Images of each well were taken at specific wavelengths (excitation/emission: 385/455 nm; 470/519 nm; 550/610 nm). Images were analyzed using a custom-made pipeline in CellProfiler 3.19 (Broad Institute, Cambridge, USA), as described previously (Habjan et al, 2021). Briefly, the fluorescent signal of the ATc-inducible GFP around each macrophage was used as a readout for viable intracellular bacteria in each macrophage, whereas the number of stained and detected nuclei was used as a readout for the number of macrophages in each treatment group.

Hemolysis assay

The hemolytic effect of a compound on the red blood cells was investigated (Schouten et al, 2022). Compounds were prepared as twofold serial dilutions in phenol red–free DMEM (Gibco, Life Technologies). Meanwhile, defibrinated sheep erythrocytes (BioTrading) were harvested by centrifugation (600g, 7 min, 4°C), gently washed five times in phenol red–free DMEM, and seeded as 4.2 × 10⁷ erythrocytes per well, in plates containing dilutions of the test compounds. The plate was centrifuged (610g, 5 min) and incubated at 37°C, 5% CO₂ for 3 h. After incubation, the cells were resuspended and centrifuged, and the supernatant was transferred to a flat-bottom 96-well plate. The hemoglobin release was measured as an absorbance of 405 nm using a BioTek plate reader (Synergy H1). After background subtraction, the data were normalized to Triton X-100–treated samples = 100% hemolysis.

Checkerboard assay

Drug-to-drug interactions between test compounds A and B were determined using an in vitro checkerboard assay, as described previously (Hsieh et al, 1993). The growth media used in the assay were the HdB medium supplemented with 0.5% glucose, 0.5% glycerol, and 0.05% Tween-80, except for experiments in Fig S4 where 7H9 medium supplemented with 10% ADS, 0.2% glycerol, and 0.02% tyloxapol was used. Compound A was prepared as a twofold serial dilution in the bacterial growth medium in a 96-well plate, with dilutions directed horizontally. Compound B was prepared as a twofold serial dilution in a separate 96-well plate, dilutions directed vertically. Then, the solution from the plate containing dilutions of Compound B was transferred to the plate containing dilutions of Compound A, resulting in a checkerboard titration of Compound A on the horizontal axis and Compound B on the vertical axis. The selected mycobacterial strain was routinely grown until the mid-logarithmic phase, harvested by centrifugation (3,000g, 5 min), washed in PBS with 0.02% tyloxapol, and diluted to an OD₆₀₀ of 0.001 per well in 96-well plates. Plates were sealed and incubated (M. marinum for 4 d at 30°C, M. tuberculosis for 6 d at 37°C). After incubation, a resazurin solution consisting of resazurin sodium salt (0.025% [wt/vol] in Milli-Q) and 20% Tween-80 (ratio 3:1) was added to each well and the plates were further incubated. When a color conversion of the dye was observed, the fluorescence was measured using a BioTek plate reader (Synergy H1), bottom reading mode, and excitation 560 nm/emission 590 nm. First, the MIC₉₀ values of compounds A and B alone were determined as the lowest concentration of the compound, which results in 90% growth inhibition. Next, the MIC₉₀ values of combinations between Compound A + Compound B and Compound B + Compound A were determined. Thus, the FIC for each compound can be calculated as the difference in MIC between the treatment of a single compound or in combination, following the formula (European Committee for Antimicrobial Susceptibility Testing EUCAST of the European Society of Clinical Microbiology and Infectious Dieases ESCMID, 2000):FICA=MICA+B/MICAFICB=MICB+A/MICB

Next, the FICI (FIC_index or ΣFICI) was calculated as∑FIC=FICA+FICB

Based on the obtained ΣFIC value, the drug-to-drug interactions can be determined as follows:

ΣFIC ≤ 0.5 represents synergism, 0.5 < ΣFIC ≤ 1 represents additive effect, 1 < ΣFIC ≤ 2 represents indifference, and ΣFIC > 2 shows antagonism.

Chemistry

All reagents and solvents were purchased from commercial suppliers (Sigma-Aldrich, Alfa Aesar, Acros, Chimmed) and used without further purification. The ¹H and ¹³C spectra were recorded on a Bruker AC-300 (200 MHz, ¹H) or a Bruker AC-200 (50 MHz, ¹³C) NMR spectrometer. Chemical shifts were measured in DMSO-d₆, using tetramethylsilane as an internal standard, and reported as ppm values. The following abbreviations indicate the multiplicity: s, singlet; d, doublet; t, triplet; m, multiplet; q, quartet; brs, broad singlet; brm, broad multiplet. Mass spectra were recorded on a Finnigan MAT INCOS 50 quadrupole mass spectrometer (EI, 70 eV) with direct injection (Thermo Finnigan). The purity of the final compounds was analyzed by analytical HPLC on an Elute HPLC system (Bruker Daltonik) equipped with an Azura UVD 2.1S UV detector (Knauer) with a wavelength at 254 nm and acquisition rate at 1 Hz. Chromatographic separation was carried out on an Acquity HSS T3 column (2.1 × 100 mm, 1.3 μm, 100 Å; Waters) at 30°C, with sample injection volume—2.0 μl. A mobile phase consisting of 0.1% formic acid in water (A), and 0.1% formic acid in acetonitrile (B) was programmed with a gradient elution of 30–95% over 10 min at a flow rate of 250 μl/min. Data were processed using Compass DataAnalysis 5.1 software (Bruker Daltonik). All final compounds were >95% pure. Elemental analysis (% C, H, N) was performed on a EURO EA elemental analyzer (HEKAtech). IR spectra were recorded on a Bruker ALPHA FT-IR spectrometer (Bruker) in KBr pellets in the range 4,000-400 cm⁻¹. The spectra were processed using OPUS software. Melting points were determined on an Electrothermal 9001 melting point apparatus (Electrothermal) (10°C per min) and were uncorrected. Merck KGaA silica gel 60 F254 plates were used for analytical thin-layer chromatography. Spots were detected by a UV lamp. Column chromatography was performed using silica gel Merck 60 (70–230 mesh). Yields refer to purified products and were not optimized.

The scheme and synthetic procedures, as well as ¹H and ¹³C spectra of the compounds, are presented in Supplemental Data 1.

Molecular docking studies

For the prediction of the protein–ligand complex, we used the program HADDOCK version 2.4 (Dominguez et al, 2003; Van Zundert et al, 2016). Before docking, we generated 3D conformations of the compound starting from its isomeric SMILES (Weininger, 1988; Weininger et al, 1989) with RDKit using the 2020 parameters (only the small aliphatic ring subset) with energy minimization and the ETKDG algorithm (Riniker & Landrum, 2015; Wang et al, 2020). We capped the maximum number of conformers to 50 and provided this ensemble of conformers to HADDOCK for docking analysis. Concurrently, we also employed the program Fpocket (Le Guilloux et al, 2009; Schmidtke et al, 2010), a protein pocket (cavity) detection algorithm based on Voronoi tessellation, and identified the binding pockets of our target protein Rv0164. The top-scored protein pocket of Rv0164 was used for our targeted ligand docking protocol, adapted from the HADDOCK2.4 small molecule binding site screening protocol (Van Zundert et al, 2016). More specifically, two sets of restraints were created to be used at different stages of docking: (a) for the rigid-body docking, we defined the entire binding pocket on the receptor and the ligand as active, and (b) for the subsequent flexible refinement stages, we defined only the binding pocket as passive and the ligand as active. The ligand-specific parameters were used to perform the docking simulation.

Thermal shift assays

Dh5α E. coli strain overexpressing HA-tagged Rv0164 under the lac promoter in a pSMT3 vector (pSMT3-rv0164HA) was inoculated from overnight culture to an OD₆₀₀ of 0.05 in the presence of 10 μM BT-37 or DMSO. Bacteria were grown until they reached an OD₆₀₀ of around 0.6, and protein expression was induced by adding 1 μM IPTG. Bacteria were harvested by centrifugation after 3 h of protein production, and whole cells were subjected to different temperatures ranging from 45 to 70°C in a PCR machine for 3 min. Subsequently, heat-treated cells were lysed by bead beating (100 μm silica beads; BioSpec) for 1 min and were spun at 20,000g for 20 min at 4°C to remove cell debris and aggregated proteins. Proteins remaining in the supernatant were denatured with SDS and separated by SDS–PAGE (12.5% polyacrylamide). Gels were transferred to a nitrocellulose membrane (GE Healthcare Life Sciences) and stained with Ponceau dye solution before the membranes were blocked with skim milk. The primary mouse monoclonal antibody anti-HA (1:5,000) and secondary antibody goat anti-mouse IgG conjugated with horseradish peroxidase (1:2,500; American Qualex) were used to detect HA-tagged protein. Visualization of Western blots was done using an ECL substrate (Amersham).

Construction of plasmids

All DNA manipulation procedures followed standard molecular cloning techniques. Primers were synthesized and purified by Sigma-Aldrich and are all listed in Table S10. Phusion polymerase, restriction enzymes, and T4 DNA ligase were obtained from New England Biolabs (NEB). iProof polymerase was obtained from Bio-Rad. Where indicated, plasmids were constructed using the ligation-independent cloning method In-Fusion (TaKaRa), following the manufacturer’s recommendations. All plasmids used within this study can be found in Table S9, and cloning experiments are summarized in Fig S11. PCR-amplified inserts were routinely subjected to DNA sequencing (Macrogen).