The bacterial species used in this study are M. smegmatis mc²155 (ATCC 700084), and M. abscessus (NCTC 13031). Mycobacterial strains were cultured in Middlebrook 7H9 broth (MB7H9; BD Difco™) supplemented with albumin/dextrose/catalase (ADC, Remel™) enrichment or Glycerol-Alanine-Salt Tween80 (GAST) medium.

REMA is a rapid colorimetric redox MIC indicator, when reduced to resorufin, the solution can be visibly seen as pink, indicating aerobically active bacterial cells. We performed REMA following our earlier protocol (Shyam et al., 2022b). A 0.01% (w/v) resazurin (Alfa Aesar) solution was used. M. abscessus was grown in a rotating incubator at 37°C in MB7H9 until OD₆₀₀ ~1.0 then diluted to ~1x10³ cells/mL, per well. Chellex-100 defferrated GAST medium (iron-deficient media), was prepared and used on a triplicate set of plates as well as M. abscessus grown in MB7H9. The plates were grown in a static incubator (Genlab) at 37°C for 48 h, 30 μL of resazurin was added, then left in a static incubator at 37°C for 48 h before being read. All plates were covered and parafilmed.

The assay was performed following our earlier reported methodology (Shyam et al., 2022b) by using early log phase cells of M. smegmatis and M. abscessus (OD₆₀₀ ~ 0.4) and compounds were tested at sub-MIC concentration (0.25 x MIC) to ensure unaltered cell viability. Cells were collected by centrifugation and resuspended in an equivalent volume of 1 x PBS. The test samples comprised bacterial culture 10⁷ bacteria/mL in PBS along with 0.4% glucose as the source of energy for efflux pump activity. A stock of 0.5 mg/L ethidium bromide (EtBr) was used as the substrate for efflux pumps, and verapamil was used as the positive control at a concentration of 125 μg/mL for M. smegmatis and 256 μg/mL for M. abscessus. The experiment was performed in a 96-well plate that was placed in a fluorimeter (Biotek Synergy 2) and programmed with following parameters: excitation 540 nm, emission 620 nm, fluorescence gain 80, cycle of measurement every minute for a total period of 60 min at 37°C.

Individual MICs were determined by REMA and used for the checkerboard assay (Shyam et al., 2022b). A 2-fold dilution was conducted of BDQ and compound Il across a 96-well plate. M. abscessus was grown and resazurin added as per the REMA. The fractional inhibitory concentration index (FICI) value was calculated with the following formula:

where, A and B are the MICs of each antibiotic in combination and MIC_A and MIC_B are the MICs of the antibiotic individually. A FICI < 0.75 is considered synergistic, between 0.75 and 4.0 is considered additive or indifferent, and a FICI > 4.0 is considered antagonistic.

Starting Cartesian coordinates for the MmpL4 and MmpL5 receptors were constructed using both homology modelling (Waterhouse et al., 2018) and AlphaFold (Jumper et al., 2021). In this study, the homology model constructed for the MmpL4 system was created using SWISS-MODEL (Waterhouse et al., 2018). A 3D model of the given target protein sequence (MmpL4) is based on its alignment to a protein of known structure (template). In the current work, the MmpL4 3D structure was built upon a sub-family protein MmpL3 crystal structure (PDB ID: 6AJG (Zhang et al., 2019) of resolution 2.6 Å with an identity of 24.5%. In contrast to homology modelling, Google’s deep learning-based AlphaFold (Jumper et al., 2021) was used to derive the 3D structure of an entire sequence of MmpL4 and MmpL5 for the M. tuberculosis strain (ATCC 25618/H37Rv) (Uniprot Id: P9WJV3 and P9WJV1).

AutoDock 4.2 program (Morris et al., 2009) was used in this study to predict ligand binding modes by using a combination of the Lamarckian Genetic Algorithm and empirical force field parameters. The MmpL4/5 protein structure (apo state) predicted by AlphaFold was subjected to minimization and equilibration as described in the molecular dynamics section. The equilibrated MmpL4/5 receptor then provided the starting Cartesian coordinates for the calculations. A grid box was generated at the center of the protein and was large enough to encompass the entire folded structure [region where the per-residue confidence score predicted by AlphaFold was very high] with a size of 90 grid points at each axis (x, y, and z). All AutoDock default parameters were utilized, with the exception that the exhaustiveness of the global search was set to 100.

Molecular dynamics (MD) simulations were performed on the MmpL4 and MmpL5 protein-ligand complexes. The PACKMOL-Memgen software (Schott-Verdugo and Gohlke, 2019) distributed with AmberTools23 was used to embed the enzyme complexes into a lipid bilayer consisting of 1-stearoyl-2-palmitoyl-sn-glycero-3-phopshoethanolamine (SPPE), 1,3-bis(1-oleoyl-2-palmitoyl-sn-glycero-3-phospho)-sn-glycerol (OPOPCL), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoinositol (POPI), and 1-palmitoyl-2-myristoyl-sn-glycero-3-phosphoinositol (PMPI) lipids, in a 27:38:2:33 ratio ( Supplementary Figure S2 ), respectively (Li et al., 2023; Vasyankin et al., 2024). These enzyme-ligand-lipid-bilayer systems, referred to as a complex, were solvated explicitly using a TIP3P cuboid box that enclosed the system based on its van der Waals boundary, and the overall charge of the complex was neutralized by adding a suitable number of potassium cations. Each system was subsequently solvated in a 0.15 M solution of KCl. The generalized Amber force field (GAFF2) (Jorgensen et al., 1983; Wang et al., 2004; Vassetti et al., 2019; He et al., 2020) was used to parameterize the inhibitor, the protein topology file was created using the ff14SB force field (Maier et al., 2015), and the lipid-bilayer was generated with Lipid21 parameters (Dickson et al., 2022). All systems were subjected to a five-step minimization protocol performed using the sander Amber22 engine. The water molecules and ions were initially minimized while the complex residues were harmonically restrained with 25 kcal mol^-1Å^-2 using the steepest descent (SD) algorithm for 5,000 steps, followed by 5,000 steps of conjugate gradient (CG) minimization. Next, water molecules and ions were minimized while the complex residues were harmonically restrained with 5 kcal mol^-1Å^-2 using the using the SD algorithm for 5,000 steps, followed by 5,000 steps of CG minimization. Subsequently, solvent molecules and lipid bilayer residues were minimized while the enzyme-ligand residues were harmonically restrained with 5 kcal mol^-1Å^-2 using the SD algorithm for 5,000 steps, followed by 5,000 steps of CG minimization. This was followed by harmonically restraining only the enzyme with 5 kcal mol^-1Å^-2 and allowing the rest of the system to undergo 5,000 steps of SD minimization, followed by 5,000 steps of CG minimization. Lastly, the entire system was minimized without restraints using the SD algorithm for 5,000 steps, followed by 5,000 steps of CG minimization. Thereafter, the system underwent a three-step equilibration protocol using mixed CPU and GPU-enabled Amber22 engines. To prime the system velocities, the system was gradually heated from 0 to 100 K, with harmonic restraints of 5 kcal mol^-1Å^-2 on the enzyme-ligand residues, in four 25 K increments using a constant NVT ensemble over 5 ps each with a Langevin thermostat, a temperature coupling value of 1.0 ps, and a collision frequency (γ) of 10.0 ps^-1, 7.5 ps^-1, 5.0 ps^-1, and 1.0 ps^-1, respectively. An additional heating step from 100 K to 310 K, with harmonic restraints of 5 kcal mol^-1Å^-2 on the enzyme-ligand residues, using a constant NPT ensemble over 115 ps with a Langevin thermostat, a temperature coupling of 1.0 ps, and a γ of 1.0 ps^-1 was performed. Finally, to equilibrate periodic boundary conditions (PBC) 5 ns stepwise simulations were performed for over 50 ns using a constant NPT ensemble at 310 K and 1 atm with the temperature and pressure coupling values set to 1.0 ps with identical Langevin parameters as the second equilibration phase. The last 5 ns of the equilibration phase were used to compute boost potentials for the subsequent 100 ns accelerated molecular dynamics (aMD) simulations using a constant NVT ensemble at 310 K (Gangadevi et al., 2021; Jumper et al., 2021; Gangireddy et al., 2022). Degrees of freedom boosted per system and their respective values are provided in Supplementary Table S1 . Long-range electrostatics were accounted for by using the particle mesh Ewald, all covalent bonds involving hydrogen atoms were constrained with the SHAKE algorithm, periodic boundary conditions were enforced using a nonbonded cutoff distance of 12 Å, and a time step of 1.0 fs was utilized. All computational analysis was performed with the cpptraj and ptraj programs available in the AmberTools23 suite and custom python scripts (Roe and Cheatham, 2013). The root-mean-square deviations (rmsd’s) and fluctuations (rmsf’s) were calculated to monitor the structural stability of each simulation, and are provided in the Supplementary Figure S3 .

These analogues are termed mycobactin analogues as they partially mimic the chemical framework of the iron-scavenging siderophore mycobactin and are identified as potential inhibitors of mycobactin biosynthesis (Shyam et al., 2022b). In our recently published research article, the steps of chemical synthesis and structural analysis (including ¹H NMR, ¹³C NMR, ESI-MS, and HR-MS), and antimycobacterial profiles against M. tuberculosis H37Rv, M. smegmatis, Mycobacterium bovis BCG and Mycobacterium aurum are reported for these analogues (Shyam et al., 2022b). We have used M. smegmatis to examine the suppression of the whole-cell efflux pump by mycobactin analogues against mycobacteria. A non-virulent mycobacterial strain, M. smegmatis mc²155, is frequently employed as a surrogate model for the screening of recently identified efflux pump inhibitors to avoid the complexity associated with M. tuberculosis in the early drug discovery stage (Johnson et al., 2020). However, pathogenic, non-pathogenic, and NTM categories all-share siderophore biosynthesis and recycling pathways (Shyam et al., 2021; Shyam et al., 2022a). In the fluorometric EtBr experiment, EtBr serves as the efflux substrate and fluoresces when in combination with DNA (Lepecq and Paoletti, 1967) (EtBr is approximately 30-fold more luminous intracellularly than in the extracellular environment). Therefore, as EtBr builds up inside the cell, fluorescence will increase over time if the substance being tested is an efflux inhibitor. In a discussion of comparative efficacy, the prevalent efflux-pump inhibitor (EPI) verapamil (VP) is employed as the positive control. Minimum inhibitory concentration (MIC) values were calculated against M. smegmatis as reported in our earlier study (Shyam et al., 2022b). We investigated efflux-pump inhibition by determining the relative fluorescence unit (RFU) at the sub-MIC dose concentration over a 60 min time-window for individual candidates ( Table 1 , Figure 2 , Supplementary Figure S4 ).

In the EtBr whole-cell efflux pump assay, it has been noted that the substitution of fluorine at the para position of the benzene ring on the 3-(2-hydroxyphenyl)-5-(aryl)-pyrazoline scaffold (compound Il), yielded high efflux pump inhibitory activity in comparison to verapamil (a known efflux pump inhibitor) ( Table 1 ). During lead optimization of the inhibitor, bioisosteric replacement of fluorine with oxygenated functional groups (hydroxy and alkoxy) is a common path adopted in structure-based drug design. As oxygen and fluorine have similar sizes and share many properties, the addition of a hydroxy or alkoxy group means incorporating a hydrogen bond donor group and increasing the polarizability (as fluorine is highly electronegative). However, such bioisosteric replacement of fluorine at the para position of a benzene ring with a hydroxy or methoxy group (compounds Ic and Id, respectively) did not increase the potency of this pyrazoline scaffold as efflux pump inhibitors. As we know, fluorine is more electronegative than oxygen, so it has a stronger electron-withdrawing effect and increases the lipophilicity of the compound. This clarifies that the fluorine atom replacement is the most plausible reason for the increased inhibitory activity. In addition, the fluorine atom replacement prohibits the scaffold from developing a metabolic oxidation reaction that is advantageous over other bioisosteric substitutions.

The intrinsic and acquired resistance of M. abscessus to commonly used antibiotics limits the therapeutic options to treat infection caused by this pathogen (Johansen et al., 2020). Moreover, a lackluster effort has been noted for rationale design inhibitors development against NTMs. We investigated HIT candidate Il against M. abscessus in our search for novel antibiotics with novel mechanisms of action i.e., by disrupting iron-homeostasis (Bythrow et al., 2022; Mudde et al., 2022) against NTM. Both MB7H9 broth, a traditional mycobacteria growth medium that is iron-enriching, and GAST, an iron-deficient environment, were used to assess compound Il’s MIC against M. abscessus in a comparative manner to understand putative target preferentiality (Shyam et al., 2022b). The MICs in MB7H9 and GAST medium were noted 256 g/mL and 8 g/mL, respectively, showing a 32-fold higher inhibition in an iron-deficient condition over an iron-supplemented medium and emphasizing antimycobacterial efficacy associated with the suppression of mycobactin synthesis. The primary hypothesis behind investigating the designed mycobactin analogues in iron-rich and iron-deprived medium to understand the target preferentiality of these analogues. We were more interested to identify a HIT candidate which could potentially inhibit the growth of mycobacteria in iron-deprived medium and showing less/no lethal effect in iron-rich medium as siderophore biosynthesis occurs at iron-deprived condition.

Excellent efflux-pump inhibition was observed following examination of whole-cell efflux-pump inhibition in M. abscessus at a sub-MIC ( Table 1 and Figure 3 ). Findings indicate compound Il showed greater efficacy against M. abscessus relative to VP over M. smegmatis. Interestingly, both verapamil and compound Il produced lower RFUs in M. abscessus than M. smegmatis, and we hypothesized that this occurrence could potentially be due to the relative concentration of positive controls and test compounds in relation to the increased number of MmpL proteins found in M. abscessus in comparison to M. smegmatis (31 and 17 mmpL genes, respectively (Deshayes et al., 2010).

As MmpL4 and MmpL5 oversee siderophore recycling, we hypothesized that a pocket for mycobactin binding should exist in the MmpL4/MmpL5 complex present in mycobacteria that possesses the spatial geometry required to accommodate the signature hydroxyphenyloxazoline chemical block of mycobactin. Utilizing this insight, we designed compounds with a similar hydroxyphenyloxazoline layout, hypothesizing that they could potentially bind to and interfere with the MmpL4/5 efflux-pump proteins. Through cell-based experiments, we identified compound Il as a promising inhibitor of M. abscessus. However, while these assays provided valuable initial insights, they did not offer a detailed explanation for the superior performance of compound Il. Consequently, an in-depth computational investigation was performed to elucidate compound’s Il inhibitory profile. To date, no crystal structures have been reported for the MmpL4/5 proteins. As a result, the initial structural coordinates of the MmpL4/5 complex were derived using the neural network-based model AlphaFold. Developed by Google DeepMind, AlphaFold has been shown to predict protein structures from amino acid sequences with extraordinary atomistic accuracy, even in the absence of a homologous structure (Jumper et al., 2021).

To validate the AlphaFold-predicted structure, an additional homology model of MmpL4 was constructed using a reported experimental structure of MmpL3 (Zhang et al., 2019) and the SWISS-MODEL software (Waterhouse et al., 2018). Visual comparison of the two generated structures suggested that multiple regions within the SWISS-MODEL generated homology model did not properly fold ( Supplementary Figure S1 ). However, both methods did generate a similar active site region where the AlphaFold per-residue confidence score was greater than 90. Nearby regions also appeared to be properly folded with AlphaFold giving per-residue confidence scores ranging from 70 to 90 for both the predicted MmpL4 and MmpL5 structures. As such, the AlphaFold generated MmpL4/5 models were adopted in this study to identify the inhibitor binding site and predict the ligand binding pose. It should be noted that two AlphaFold predicted regions were not properly folded, residues 1-18,477-710 and 931-967, but were located far away from the putative ligand binding site; consequently, these residues were truncated and capped during the subsequent MD simulations ( Supplementary Figure S1 ).

Molecular docking was then applied to predict the location of the ligand binding pocket and the potential ligand poses using the unbound (i.e., apo) AlphaFold-predicted structures of the transporters. In this study, blind docking was carried out using compound Il and the putative binding pocket identified in the MmpL4 protein was homologous to the MmpL3 active site ( Figure 4A ). The agreement between the binding site of the co-crystallized structure of the MmpL3 inhibitor (SQ109) and docked pose of compound Il provided high confidence in the docking protocol. The docking study identified three different binding poses within the same binding pocket with each pose possessing different electrostatic interactions with the active site residues ( Figures 4B–D ). Given the positive agreement between the experimental MmpL3 and the predicted MmpL4 structure of the ligand binding site, one may assume that family-member MmpL5 would possess a similar binding site. To verify, compound Il was investigated in MmpL5 using an identical docking procedure. The high structural similarity between the MmpL isoforms resulted in the same active site residues interacting with the ligand in a nearly identical fashion ( Figure 5 ).

Given the known limitations of docking scores for estimating binding affinities and the lack of an encompassing microenvironment, a more computationally demanding in silico approach was required to compute more precise binding affinities. Accordingly, the most favorable binding orientations (i.e., 1, 2, and 3 given in Figures 4 , 5 ) for compound Il were examined in a complex of MmpL4 and MmpL5 transporters embedded within a lipid bilayer composed of 1-stearoyl-2-palmitoyl-sn-glycero-3-phopshoethanolamine (SPPE), 1,3-bis(1-oleoyl-2-palmitoyl-sn-glycero-3-phospho)-sn-glycerol (OPOPCL), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoinositol (POPI), and 1-palmitoyl-2-myristoyl-sn-glycero-3-phosphoinositol (PMPI) lipids, in a 27:38:2:33 ratio respectively (discussed in detail in the method section). The model strived to mimic the transporters’ native lipophilic environment within the inner membrane ( Supplementary Figure S2 ). Instead of relying on rigid-body docking methods, the embedded complexes were simulated using biased accelerated molecular dynamics (aMD) (Hamelberg et al., 2004; Gedeon et al., 2015) to account for any structural changes of the transporters that could occur as a result of both the protein-ligand binding event and the encapsulating lipid bilayer. The structural stabilities of the MmpL4/5-Il complexes were monitored by computing the root-mean-square deviation (RMSD) and root-mean-square fluctuation (RMSF) using the backbone protein atoms from each aMD trajectory ( Supplementary Figure S3 ). A steady RMSD plot was observed after ~10 ns of aMD simulations of the 100 ns trajectories for all complexes, suggesting stable complexes ( Supplementary Figure S3A ).

To better estimate binding affinities for compound Il, the resulting trajectories were leveraged to compute binding affinities using the MM/PBSA method (Miller et al., 2012; Greene et al., 2016; Wang et al., 2016; Xiao et al., 2017). Comparison of the computed binding free energies, ΔG _bind , for all three orientations of compound Il bound to the MmpL4 and MmpL5 complexes found that orientation-3 bound to MmpL4 possessed the most favorable ΔG _bind of -12.5 kcal/mol ( Table 2 ). Interestingly, decomposing the MM/PBSA energetic contributions into van der Waals and electrostatics interactions and polar and nonpolar contributions to the solvation free energies (E _vdw , E _el , G _pol and G _np , respectively) found the largest differences in the total ΔG _bind came from the E _el and G _pol contributions as the E _vdw and G _np gave comparable energetic values for all three binding orientations within both receptors.

To get further structural insight from the MD simulations, the lowest energy binding poses for the MmpL4/5-Il complexes were examined, and it was found that the ligand adopted a different conformation when compared to the initial binding pose predicted by molecular docking methods ( Figures 4B–D , Figure 6 ). Of particular importance, as previously indicated in the literature, were two particular Asp-Tyr residue pairs, i.e., Asp272-Tyr845 and Asp844-Tyr273 in MmpL4, as these dyads are located in the active site and conserved across the transmembrane domains of the MmpL transporter family, e.g., Asp276-Tyr850 and Asp849-Tyr277 in MmpL5, with the exception of MmpL7 (Bernut et al., 2016; Zhang et al., 2019)

Studies employing mutational analysis have linked mutation of Asp-Tyr dyads in MmpL3 to the inhibition of mycobacteria growth, postulating that these Asp-Tyr residue pairs play a critical role in facilitating proton translocation across the MmpL family (Bernut et al., 2016). Within MmpL3, two sets of Asp-Tyr dyads are located at the center of the transmembrane domain, specifically in the middle region of α-helices 4 and 10, prompting structural stabilization of the helices via formation of hydrogen bonds (Bolla, 2020). Interestingly, the location of the Asp-Tyr dyads in MmpL3 is analogous to that of the Asp-Asp-Lys triad in AcrB, which is known to be involved in the proton relay network, and with whom MmpLs share homologous sequences. AcrB, an acridine resistance complex is a resistance-nodulation-cell division transporter multidrug efflux pump that is involved in the virulence and pathogenesis of various bacterial pathogens (Du et al., 2018; Fay et al., 2019) Similar to MmpLs, AcrB is hypothesized to be localized to the inner membrane (Fanelli et al., 2023). Considering the homology between MmpLs and AcrB, and the high sequence and structural similarity amongst MmpL isoforms, the dynamics of the binding active site pocket of MmpL4/5 are expected to be similar to that of MmpL3. These interactions are crucial for understanding the binding mechanism and potential inhibitory effects of compound Il within MmpL4/5. As a result, probing molecular interactions native to the protein-ligand binding event may aid in rationalizing the preference of compound Il for MmpL4.

The disruption of the Asp-Tyr dyads interaction network not only detrimentally affect the structure and stability of the protein, but also compromise the function of the transporter. Consequently, ligand orientations that fail to maximize interactions amongst Asp-Tyr dyads are energetically destabilizing and predicted to be poor binders. Energetically, compound Il positioned in orientation-3 within MmpL4 yielded the most favorable binding affinity, suggesting this specific orientation facilitates the ligand in maintaining stabilizing interactions with adjacent active site residues ( Table 2 ). To identify the favorable electrostatic interactions that compound Il maintained within the active site of MmpL4, hydrogen bonding occupancies were monitored for the full trajectories, with special consideration placed on ligand interactions with Asp-Tyr dyads ( Table 3 ). Detailed examination of the binding site within MmpL4 revealed compound Il, when in orientation-3, is positioned between the Asp-Tyr dyads, i.e., D272-Y845 and D844-Y273, which allows the compound to interact with the residues of both Asp-Tyr dyads ( Figure 6B ). As a result, the positioning of compound Il allows it to act as a pseudo replacement for an Asp-Tyr dyad interaction, which contributes to the more energetically favorable electrostatic contributions that are reflected in the E _el and G _pol terms from the MM/PBSA calculations. Unlike orientations 1 and 2, orientation-3 positioned the pyrazoline ring of compound Il near Y845(64.5%) while its hydroxyphenyl moiety was in close contact with D844 (72.4%) ( Figure 6B ), maximizing hydrogen bonding interactions within the Asp-Tyr network ( Table 3 ). To a lesser extent, compound Il in orientation-3 maintained hydrogen bonding interactions with D272 (15%), with no hydrogen bonding interactions determined between the ligand and Y273, which is the complementary tyrosine residue of the D844-Y273 dyad ( Table 3 ).

Considering the positioning of compound Il’s hydroxyphenyl ring is vastly different in orientation-3 than alternative orientations, additional analysis was performed to elucidate if dispersion interactions, such as π-π stacking, was a contributing factor to the energetic favorability of compound Il when in orientation-3. Within proteins, π-π interactions generally have intermolecular distances ranging from ≤ 5.6 Å and dihedral angles, γ, ranging from 50° to 90° for non-substituted aromatic rings such as phenyls while substituted aromatic rings, like hydroxyphenyls slightly shift these ranges (Zhao et al., 2015). To account for π-π stacking interactions, normalized vectors centered on the mass of the hydroxyphenyl rings of the ligand and Y273 were used to compute the angle between them while centers of mass were leveraged to compute intermolecular distances ( Figure 7 ) (Zhao et al., 2015; Hayatshahi et al., 2018; Somai et al., 2023). To estimate for π-π stacking interactions, a distance cutoff of ≤ 5.0 Å denoted by R_cen and an angle cut-off of 0° to 180° was utilized, resulting in compound Il when in orientation-3 exhibiting stacking behavior for 81.8% of the trajectory while orientations 1 and 2 yielded dismal results of 9.5% and 0.0%, respectively. Notably, analysis of the lowest energy structure of the MmpL4-Il complex in orientation-3 revealed the hydroxyphenyls of the compound and Y273 maintained an intermolecular distance of 4.5 Å and a dihedral angle of 45.6°, supporting the notion of π-π stacking interactions, particularly T-shaped, being prominent within the protein-ligand binding event. Given the stacking behavior of compound Il in orientation-3 is almost an order of magnitude greater than alternative positions, the stacking analysis suggests only orientation-3 allows the compound to maintain π-π stacking interactions that further aid in promoting the binding event. Unsurprisingly, investigation into alternative positions revealed orientations 1 and 2 in MmpL4/5 failed to maintain notable hydrogen bonds with both Asp-Tyr dyads, mostly engaging with a sole aspartate residue from an Asp-Tyr dyad, i.e., D272-Y845 in MmpL4 and D276-Y850 in MmpL5 ( Table 3 ). Thus, only compound Il when in orientation-3 maximized all Asp-Tyr dyad interactions through a combination of electrostatic and van der Waals forces, suggesting the ligand’s chemical framework is capable of effectively targeting MmpL4 over MmpL5.

It has already been highlighted that the newly discovered anti-TB drug BDQ for the therapeutic intervention of multi- and extreme-drug-resistant TB patients confers efflux-mediated resistance, indicating the grim scenario of an ever-emerging AMR burden (Willyard, 2017). Mutations to the regulator of the mmpL5-mmpS5 genes (Rv0676c and Rv0677c) is responsible for BDQ resistance, as noted in previous studies. In this context, we examined compound Il’s efficiency in combination drug therapy with BDQ in a checkerboard assay, and fractional inhibitory concentrations were calculated. First, the MIC of BDQ against M. abscessus in MB7H9 and GAST media was determined to be between 0.43 μg/mL - 0.22 μg/mL and between 0.22 μg/mL - 0.05 μg/mL respectively. Following this, we investigated this new drug combination (BDQ and compound Il) against M. abscessus in both iron-rich and iron-deprived conditions. BDQ and compound Il combination showed additive profiles in both the medium against M. abscessus at their respective MICs. This newly identified drug combination by targeting essential (BDQ targets ATP synthase) and conditionally essential mechanism (mycobactin biosynthesis inhibition of compound Il) might pave the way to eradicate AMR challenges associated with superbugs in the coming days.