Molecular insights into Mmpl3 lead to the development of novel indole-2-carboxamides as antitubercular agents

Abstract

Tuberculosis (TB) is an air-borne infectious disease and is the leading cause of death among all infectious diseases globally. The current treatment regimen for TB is overtly long and patient non-compliance often leads to drug resistant TB resulting in a need to develop new drugs that will act via novel mechanisms. In this research work, we selected Mycobacterium membrane protein large (MmpL3) as the drug target and indole-2-carboximide as our molecule of interest for further designing new molecules. A homology model was prepared for the Mycobacterium tuberculosis MmpL3 from the crystal structure of Mycobacterium smegmatis MmpL3. A series of indoles which are known to be MmpL3 inhibitors were docked in the prepared protein and the binding site properties were identified. Based on that, 10 molecules were designed and synthesized and their antitubercular activities evaluated. We identified four hits among which the highest potency candidate possessed a minimum inhibitory concentration (MIC) of 1.56 μM at 2-weeks. Finally, molecular dynamics simulation studies were done with 3b and a previously reported MmpL3 inhibitor to understand the intricacies of their binding in real time and to correlate the experimental findings with the simulation data.

Graphical Abstract

This work utilises high-end molecular modelling techniques to analyse a drug target with known inhibitors and design molecules based on it, and subsequently leading to the development of a novel antitubercular agent.

Introduction

Tuberculosis (TB) is caused by the pathogen Mycobacterium tuberculosis and mostly manifests as a lung disease. It can also occur in other organs like kidney, spine, and brain. It is an air-borne disease and needs a very small bacterial load to spread from one person to another [1], As of 2019, 10 million people were infected with TB worldwide with 1.4 million deaths. It has now become the leading cause of death due to a single infectious agent, standing higher than HIV/AIDS [2]. Currently, the treatment regimen consists of four drugs over a period of six months which includes intensive treatment with isoniazid (H), rifampicin (R), pyrazinamide (Z) and ethambutol (E) for two months followed by the continuation of HRE for the next four months [3]. Often, this long treatment regime results in severe hepatic damage [4, 5]. The lengthy duration of treatment often associated with toxic side-effect leads to poor patient compliance and treatment adherence.

Treatment interruption or discontinuation has led to the emergence and spread of multiple and extensively drug resistant TB (MDR-TB and XDR-TB) [6]. The treatment options for MDR and XDR tuberculosis include multiple second line drugs. Usually, the treatment regimen is long and expensive with the problem of unavailability in many countries [6]. As of 2020, the global success rate for the treatment of MDR TB was 58% [7]. Thus, there is an urgent need to develop new chemical entities which would be highly effective in treating TB with a shorter duration coupled with a very good safety profile. Recently, Mycobacterium Membrane Protein Large (MmpL3) has come up as an attractive target for antitubercular drug discovery. This membrane protein transports trehalose monomycolate (TMM) from the cytoplasm to the cell wall of the bacteria for further conversion to trehalose dimycolate (TDM) and arabinan-linked mycolates [8]. MmpL3 is essential for mycobacterial growth and survival which has made it a promising drug target. Different classes of chemical moieties inhibit this protein and these findings have initiated extensive explorations to develop new chemical entities [8–11]. In our work, we have focused on indole compounds since previous research have established their antitubercular potency and along with their simple structures and synthetic feasibility, laid the path for further exploration of their structure-activity relationships [12–14].

In this study, we intended to design and synthesize novel indole molecules that would show antitubercular activities possibly by inhibiting MmpL3. Since the indole compounds reported to date as Mmpl3 inhibitors are not substituted at their third position, we wanted to explore that region by doing relevant substitutions and understand their importance in the context of interacting with MmpL3. Over the years, physics based computational models have gained importance in helping us understand complex biological systems. These models have not only helped to get a better picture of protein dynamics but also the intricacies of how well a molecule binds to its protein target [15, 16]. In this research, several computational techniques including homology modelling, molecular docking and binding free energy calculations were used to analyse the binding pocket of Mmpl3 and novel indole molecules were designed based on the computational data [15–17]. The compounds were synthesized and their antitubercular activities determined. Finally, molecular dynamics (MD) simulation was used to understand the intricacies of binding of the potent molecules (from previously reported and newly synthesized) with MmpL3 by correlating their antitubercular activities with the simulation data.

Experimental Section

Maestro 11.3 from Schrodinger Suite 2018-3 was used to perform all the computational modelling and calculations. A database of 64 compounds was created using molecules with antitubercular minimum inhibitory concentration (MIC) data from two research papers by the same group (Table S1) [14, 18].

Homology Model

A Cryogenic electron microscopy (Cryo-EM) structure of Mycobacterium tuberculosis Mmpl3 (PDB ID 7NVH) was reported recently by Adams et al. [19]. Although the resolution of the structure had an acceptable value of 3 Å, one of the major issue we faced with the structure was defining the binding site space for our molecules of interest. As the reported structure was not in a ligand bound state, the known binding residues were used to define the binding site. It was followed by docking with known inhibitors and we observed that the poses obtained were highly unrealistic, like the inversion of the molecules inside the binding site by 180° during docking even with very slight structural changes. This problem could not be solved even after considering induced-fit docking (IFD) and MD poses. Hence, we looked forward to resolve the problem by developing homology model from other structures instead of using the Cryo-EM structure. All the other crystal structures for MmpL3 reported till then were of Mycobacterium smegmatis which shares about 60%-65% sequence identity with MmpL3 from Mycobacterium tuberculosis [8, 20]. A homology model for Mycobacterium tuberculosis Mmpl3 was previously reported by Dupont et al. [21]. The template used for this model was the one with PDB ID 6AJJ. Though PDB ID 6AJJ had an indole compound (ICA38) in its binding pocket, 6AJI was chosen as the template structure for our model because of the larger receptor space its ligand occupied [20]. Molecules with substitution at the 3rd position of indole-2-carboxamides would resemble the shape of the ligand found in 6AJI. In addition to that, there was template gap of 51 residues in both 6AJI and 6AJJ. This resulted in chain breaks in the protein or formation of highly unstable structures when the chains were connected. With such a large gap, prediction of secondary structures and generation of random coil would not result in a realistic model. As a result, another PDB 6N40 was also used as a template, particularly for that region (had a gap of 26 residues) [8]. Finally, energy based Chimera modelling technique was used to select relevant residues from 6N40 and 6AJI to develop the homology model (Fig. 1). The model protein was minimized using the Macromodel tool from Maestro and the Ramachandran plot was generated (Fig. 2). An intrinsic membrane was setup on the residues 14-34, 186-206, 210-230, 236-256, 287-307, 315-335, 397-417, 563-583, 587-607, 617-637, 673-693, and 699-19 from the loop refinement panel. Based on the observations from the Ramachandran plot, loop refinement was done on the non-template loop, loop7 (res 772-792) [22, 23]. The developed homology model was then prepared at pH 7.0 by adding the necessary hydrogens, filling the missing loops and removing the unwanted water molecules. Finally, energy minimization was performed again using MacroModel tool (it utilises OPLS3E as the force field). The ligand from the database was prepared at pH 7.0 using the LigPrep tool and was minimized using OPLS3E force field. Thereafter, the receptor grid was generated from the protein which would be imported for the subsequent docking studies [3].

Figure 1.

The query (P9WJV5) and the template sequences (6N40 and 6AJI) are displayed where in (a) the region highlighted in pink and green represent the template regions used for building the homology model from 6N40 and 6AJI respectively. The first major template residue gap is observed from residue numbers 357 to 382. A random loop was created for this region. (b) The second major template residue gap is observed in this region from 783 to 797 where another random loop was created.

Figure 2.

(a) Ramachandran Plot before loop refinement (b) Ramachandran Plot after loop refinement. Red colour indicate favoured area, yellow colour indicates allowed area and white colour indicate disallowed region in the plot. Glycine and proline residues are represented by dark squares and triangles, respectively. Residues other than glycine and proline are represented by dark round blocks.

Molecular Docking and MMGBSA of reported indoles

The prepared model consisted of the molecule Rimonabant as the co-crystallised ligand. Though chemically dissimilar to our designed molecules, this molecule helped to determine the binding pocket for docking. This was majorly due to the shape of the ligand resembling the shape of the designed indole-2-carboxamides. The grid for docking was developed by using the Receptor Grid Generation tool from Glide in Maestro where the co-crystallised ligand was picked and a cubical box from the centroid of the ligand determined the binding space. The van der Waals radius scaling factor and partial charge cutoff were set at 1.0 and 0.25 respectively [24]. Flexible molecular docking was done with all the molecules in the database (using extra precision mode (XP) from Glide) on the prepared grid. The docked poses were analysed to understand the properties of the regions of the protein taking part in binding with the ligands. The regions of the ligand which had to be left conserved were also assessed by analysing the docking poses. Finally, Molecular Mechanics Generalized Born Surface Area (MMGBSA) method was used for all the docked complexes (implicit membrane was generated) to calculate their binding free energy (dG Bind) [25–27].

Design of novel molecules

The indole-2-carboxamides that had been reported so far showing antitubercular activity did not possess any substitution at the third position. The most prominent feature in these molecules is a bulky group attached to the amide nitrogen. Structure activity relationship (SAR) studies with this molecules have clearly shown the importance of the bulky lipophilic group attached to amide nitrogen [28, 29]. We thought of introducing new groups on the third position keeping the other parts conserved. This was based on the initial observations from the docking studies which showed residues LEU633, LEU637 and VAL681 forming hydrophobic enclosures and we thought introducing new parts on the molecules to fill the pocket would be conducive. Initially, a library of 1000 molecules was generated using the R-group Enumeration interface in Maestro. The enumeration was done on the third position of indole. The molecules were docked on the developed model using XP mode and their dG Bind scores were calculated using MMGBSA. Compounds with dG bind scores ≤ −40 kcal/mol were considered, and synthetic feasibility, cost and avalibility of chemicals were also taken into account to finalise a list of 100 molecules. Finally, 8 compounds with logP ≤ 4.5 was selected along with another two compounds (owing to their dG bind ≤ −50) with 4.5 < logP ≤ 5. Since the target is a membrane protein and the binding site forms a hydrophobic enclosure, it was evident that most of the molecules showing good binding affinity would be highly lipophilic in nature. But from the perspective of developing drug like molecules, it was important that logP be reduced to a certain extent without compromising the binding affinity. For this purpose, it may be noted that the polar groups have been introduced as linkers with lipophilic portions of the molecules covering the edges.

General Chemistry

The starting materials, reagents and catalysts were procured from Sigma-Aldrich (St. Louis, MO, USA) and Spectrochem Pvt. Ltd. (Bengaluru, India). The reactions were performed with LR grade solvent. Purified solvents were used for column chromatography (CC). Reactions were monitored using Silica gel 60 F254-precoated TLC (thin-layered chromatography) plates and by visualizing in UV chamber. Purification of compounds were done by CC or salt preparation. Column chromatography was performed on silica gel 60120 (100–200 mesh) purchased from Merck (Germany). 1H, 13C NMR and mass spectrometry were used to characterise the novel compounds. 1H and 13C-NMR spectra were recorded using Bruker Ascend 400 MHz; chemical shift values (δ) are expressed in ppm relative to internal standard, tetramethylsilane (TMS). LTQ21532 series LC/MS was used to record the mass spectra.

Synthesis of Ethyl 3-methyl-1H-indole-2-carboxylate (1)

To the solution of 2’-bromoacetophenone (15 mmol) in DMSO (15 ml), Copper (I) iodide (0.45 mmol) and cesium carbonate (30 mmol) were added. To the mixture, ethyl isocyanoacetate (58.5 mmol) was added slowly and it was stirred at 90° C until reaction was complete as determined by TLC. The reaction mixture was quenched by the addition of water (10 ml) and followed by extraction with ethyl acetate (3 X 20 ml). The organic layer was separated and dried over anhydrous sodium sulphate and distilled under vacuum at 50° C to give a crude reddish brown solid. Column chromatography was performed to purify the compound and white coloured solid was obtained [30].

Synthesis of 3-Methyl-1H-indole-2-carboxylic acid (2)

1 (6 mmol) was added to a solution of NaOH (18 mmol) in methanol (10 ml) and water (5 ml). The mixture was stirred for 2 hours at 80°C for completion of the reaction. The methanol was distilled out under vacuum to give a light brown solution. The pH was adjusted to 3-4 by the addition of dilute HCl to give white solid which was subsequently filtered and dried under vacuum [31].

Synthesis of 3-Methyl-N-(2-methylcyclohexyl)-1H-indole-2-carboxamide (3a)

EDC.HCl (5.52 mmol) and HoBt (5.52 mmol) were added to a solution of 2 (4.6 mmol) in 5 ml dichloromethane (DCM) and the reaction mixture stirred for 30 minutes at room temperature. A mixture of triethylamine (5.52 mmol) and 2-methylcyclohexylamine (5.52 mmol) was added to the reaction mixture and stirred for 16 hours. The reaction mixture was evaporated under vacuum and a solution of DMF in water (3 ml/10 ml) was added to it and stirred for 10 minutes. The obtained solid was then filtered and subsequently washed with methanol to give a pure white compound [13, 14].

Synthesis of 3-Methyl-N-(4-methylcyclohexyl)-1H-indole-2-carboxamide (3b) and N-Cyclohexyl-1H-indole-2-carboxamide (4)

The general procedure for 3a was followed with DCM (5ml) as solvent. The reagents used were EDC.HCl (5.52 mmol), HoBt (5.52 mmol), triethylamine (5.52 mmol), and 4-methylcyclohexylamine (5.52 mmol).

Synthesis of 3-(Benzylamino)-N-cyclohexyl-1H-indole-2-carboxamide (5a)

Benzylamine (1.64 mmol) and formaldehyde (1.07 mmol) were added to acetic acid (3 ml) and the mixture was stirred for 30 mins at room temperature. To the reaction mixture, and N-cyclohexyl-1H-indole-2-carboxamide (0.82 mmol) was added and stirred for 30 more minutes for the completion of the reaction. The reaction was quenched with water and sodium carbonate solution was added to it to basify the mixture. The resultant solid was filtered to give the crude product. The solid was purified by making oxalate salt in methylene chloride. The salt was then converted into free base by basifying with sodium carbonate solution and filtered. The white pure solid was dried under vacuum at 45 °C for 24 hours [32, 33].

Synthesis of N-Cyclohexyl-3-((4-methoxybenzyl)-amino)-1H-indole-2-carboxamide (5b), 3-(Allylamino)-N-cyclohexyl-1H-indole-2-carboxamide (5c), N-Cyclohexyl-3-(piperidin-1-yl)-1H-indole-2-carboxamide (5d), N-Cyclohexyl-3-(4-hydroxypiperidin-1-yl)-1H-indole-2-carboxamide (5e), N-Cyclohexyl-3-(diethylamino)-1H-indole-2-carboxamide (5f), N-Cyclohexyl-3-(isopropylamino)-1H-indole-2-carboxamide (5g) and 3-(Tert-butylamino)-N-cyclohexyl-1H-indole-2-carboxamide (5h)

The general procedure for 5a was followed with the respective benzylamines (1.64 mmol) , formaldehyde (1.07 mmol), and acetic acid (3 ml) as solvent.

3-Methyl-N-(2-methylcyclohexyl)-1H-indole-2-carboxamide (3a).

Yield 80% (White Powder). ¹H NMR (400 MHz, CDCl₃-d) δ 9.05 (s, 1H), 7.62 (d, J = 8 Hz, 1H), 7.38 (d, J = 8 Hz, 1H), 7.28 (t, J = 8 Hz, 1H), 7.14 (t, J = 8 Hz, 1H), 5.78 (d, J = 8 Hz, 1H), 3.81 - 3.72 (m, 1H), 2.61 - 2.57 (2s*, 3H), 2.15 – 2.11 (m, 1H), 1.85 – 1.71 (m, 2H), 1.47 – 1.13 (m, 6H), 1.04 – 0.96 (2d**, J = 8 Hz, 3H); ¹³C NMR (DMSO-d6) δ 161.87, 135.66, 128.52, 128.42, 124.13, 123.97, 120.04, 119.42, 114.58, 113.78, 112.20, 54.09, 49.63, 37.59, 34.59, 33.55, 30.11, 25.99, 25.78, 19.85; MS (ESI) m/z calcd for C₁₇H₂₂N₂O ([M + H]+) 271.17; found 271.30.

3-Methyl-N-(4-methylcyclohexyl)-1H-indole-2-carboxamide (3b).

Yield 83% (White Powder). ¹H NMR (400 MHz, CDCl₃-d) δ 9.06 – 9.03 (2 s, 1H), 7.62 (t, J = 8 Hz, 1H), 7.40 – 7.37 (dd, J = 8Hz, 1H), 7.29 (td, J = 8 Hz, 1H), 7.16 – 7.12 (m, J = 8 Hz, 1H), 6.21 – 5.81 (2 d, J = 8 Hz, 1H), 4.35 - 3.93 (2 m, 1H), 2.60 – 2.55 (2s*, 3H), 2.14 (d, J = 8 Hz, 1H), 1.86 – 1.66 (m, 5H), 1.32 – 1.09 (m, 4H), 0.98 – 0.93 (2d**, J = 4 Hz, 3H); ¹³C NMR (DMSO-d6) δ 161.80, 161.60, 135.69, 135.66, 128.52, 128.23, 124.06, 120.08, 119.44, 114.24, 114.16, 112.22, 48.50, 46.02, 34.15, 32.85, 32.04, 30.12, 29.02, 22.69, 21.13; MS (ESI) m/z calcd for C17H22N2O ([M + H]+) 271.17; found 271.20.

3-(Benzylamino)-N-cyclohexyl-1H-indole-2-carboxamide (5a).

Yield 80% (White Powder). ¹H NMR (400 Mhz, DMSO-d6) δ 11.50 (s, 1H), 7.54 (d, J = 8 Hz, 1H), 7.40 (d, J = 9.6 Hz, 1H), 7.33 – 7.32 (m, 4H), 7.29 – 7.23 (m, 1H), 7.16 (t, J = 7.4 Hz, 1H), 7.02 (t, J = 7.4 Hz, 1H), 4 (s, 2H), 3.812 – 3.71 (m, 3H), 1.84 (d, J = 11.6 Hz, 2H), 1.67 – 1.64 (m, 2H), 1.56 – 1.53 (m, 1H), 1.35 – 1.05 (m, 6H); ¹³C NMR (DMSO-d6) 161.11, 140.14, 135.27, 131.26, 128.59, 128.56, 128.31, 127.24, 123.51, 119.67, 119.64, 112.81, 112.62, 51.88, 48.07, 42.24, 40.61, 40.40, 40.19, 39.36, 33.04, 25.73, 24.93; MS (ESI) m/z calcd for C₂₃H₂₇N₃O ([M - H]-) 360.22 found 360.32.

N-Cyclohexyl-3-((4-methoxybenzyl)-amino)-1H-indole-2-carboxamide (5b).

Yield 82% (White Powder). ¹H NMR (400 Mhz, DMSO-d6) δ11.64 (s, 1H), 10.80 (d, J = 6 Hz, 1H), 7.56 (d, J = 8 Hz, 1H), 7.64 (d, J = 8, 1H), 7.39 (d, J = 8, 1H), 7.15 (t, J = 7.6, 1H), 7.03 (t, J =7.2, 1H), 5.96 – 5.86 (m, 1H), 5.20 – 5.10 (m, 2H), 3.961 (s, 2H), 3.8 – 3.76 (m, 1H), 1.92 – 1.89 (m, 2H), 1.74 – 1.70 (m, 2H), 1.61 – 1.58 (m, 1H), 1.41– 1.14 (m, 6H); ¹³C NMR (CDCl₃-d) 161.57, 135.60, 134.82, 131.18, 128.25, 124.09, 120.13, 119.28, 117.02, 112.50, 112.18, 51.23, 48.76, 42.74, 33.50, 25.98, 25.18; MS (ESI) m/z calcd for C₁₉H₂₅N₃O ([M - H]-) 390.22 found 390.28.

3-(Allylamino)-N-cyclohexyl-1H-indole-2-carboxamide (5c).

Yield 75% (White Powder). ¹H NMR (400 Mhz, DMSO-d6) δ11.5 (s, 1H), 7.64 (d, J = 8, 1H), 7.39 (d, J = 8, 1H), 7.15 (t, J = 7.6, 1H), 7.03 (t, J = 7.2, 1H), 5.96 – 5.86 (m, 1H), 5.20 – 5.10 (m, 2H), 3.96 (s, 2H), 3.80 – 3.76 (m, 1H), 1.92 – 1.89 (m, 2H), 1.74 – 1.70 (m, 2H), 1.61 – 1.58 (m, 1H), 1.41 – 1.14 (m, 6H); ¹³C NMR (CDCl₃-d) 161.57, 135.60, 134.82, 131.18, 128.25, 124.09, 120.13, 119.28, 117.02, 112.50, 112.18, 51.23, 48.76, 42.74, 33.50, 25.98, 25.18; MS (ESI) m/z calcd for C₁₉H₂₅N₃O ([M - H]-) 310.20 found 310.21.

N-Cyclohexyl-3-(piperidin-1-yl)-1H-indole-2-carboxamide (5d).

Yield 82% (White Powder). ¹H NMR (400 Mhz, DMSO-d6) δ 11.53 (s, 1H), 10.16 (d, J = 8 Hz, 1H), 7.64 (d, J = 8 Hz, 1H), 7.39 (d, J = 8 Hz, 1H), 7.15 (t, J = 8 Hz, 1H), 7.028 (t, J = 8 Hz, 1H), 3.84 – 3.75 (m, 1H), 3.67 (s, 2H), 2.40 – 2.31 (m, 3H), 2.07 (s, 1H), 2.0 – 1.63 (m, 6H), 1.50 – 1.14 (m, 11H); ¹³C NMR 161.68, 134.83, 130.95, 129.16, 124.08, 120.12, 119.47, 112.07, 111.87, 109.48, 53.80, 52.76, 48.86, 34.19, 25.99, 25.64, 24.55; MS (ESI) m/z calcd for C₂₁H₂₉N₃O ([M - H]-) 338.23 found 338.35

N-Cyclohexyl-3-(4-hydroxypiperidin-1-yl)-1H-indole-2-carboxamide (5e).

Yield 79% (White Powder). ¹H NMR (400 Mhz, DMSO-d6) δ 11.57 (s, 1H), 10.15 (d, J = 7.2 Hz, 1H), 7.65 (d, J = 8, 1H), 7.40 (d, J = 8, 1H), 7.16 (t, J = 7.4, 1H), 7.04 (t, J = 7.4, 1H), 4.69 (br s, 1H), 3.80 (d, J = 6.8 Hz, 1H), 3.70 (s, 2H), 3.53 (s, 1H), 2.16 (s, 2H), 2.32 (d, J = 6.8 Hz, 2H), 1.75-1.56 (m, 6H), 1.56-1.36 (m, 5H), 1.39 – 1.15 (m, 4H) ¹³C NMR 163.95, 161.58, 134.79, 129.02, 125.71, 124.26, 120.30, 117.16, 119.40, 112.09, 111.70, 92.47, 52.01, 48.89, 34.38, 34.22, 25.95, 25.56; MS (ESI) m/z calcd for C₂₀H₂₉N₃O ([M + H]+) 356.23 found 356.10.

N-Cyclohexyl-3-(diethylamino)-1H-indole-2-carboxamide (5f).

Yield 75% (White Powder). ¹H NMR (400 Mhz, DMSO-d6) δ 11.52 (s, 1H), 10.59 (d, J = 7.6, 1H), 7.66 (d, J = 8, 1H), 7.41 (d, J = 7.4, 1H), 7.16 (t, J = 7.9, 1H), 7.04 (t, J = 7.7, 1H), 3.83 (s, 2H), 3.80 – 3.76 (m, 1H), 2.60 – 2.50 (m, 5H), 1.95 (d, J = 10.4, 3H), 1.76 – 1.62 (m, 4H), 1.37 – 1.14 (m, 6H), 1.01 (t, J = 7, 6H); ¹³C NMR (CDCl₃-d) 161.67, 160.02, 134.80, 130.88, 129.14, 124.04, 120.05, 119.48, 112.48, 49.01, 47.74, 45.35, 33.93, 25.99, 25.60, 10.72; MS (ESI) m/z calcd for C₂₀H₂₉N₃O ([M + H]+) 328.23 found 328.

N-Cyclohexyl-3-(isopropylamino)-1H-indole-2-carboxamide (5g).

Yield 80% (White Powder). ¹H NMR (400 Mhz, DMSO-d6) δ 11.16 (s, 1H), 8.1 (d, J = 7.6, 1H), 7.34 (d, J = 8.4, 1H), 7.03 (t, J = 7.4, 1H), 6.82 (t, 1H), 4.89 (s, 1H), 4.07 (s, 2H), 3.97 (s, 2H), 3.81 (d, J = 4.4, 2H), 2.08 (s, 2H), 1.95 – 1.86 (m, 4H), 1.75-1.73 (br d, 3H), 1.61-1.60 (br d, 3H), 1.37-1.23 (m, 9H), 1.08 (d, J = 6.4, 6H); ¹³C NMR (DMSO-d6) δ 161.15, 135.26, 131.14, 128.14, 123.48, 119.69, 119.63, 113.14, 112.60, 48.39, 47.65, 33.45, 25.86, 25.17, 22.72, 19.91; MS (ESI) m/z calcd for C₁₉H₂₇N₃O ([M - H]-) 312.21; found 312.32.

3-(Tert-butylamino)-N-cyclohexyl-1H-indole-2-carboxamide (5h).

Yield 75% (White Powder). ¹H NMR (400 Mhz, DMSO-d6) δ 10.51 (d, J = 6.8, 1H), 9.22 (s, 1H), 7.63 (d, J = 8, 1H), 7.41 (d, J = 8, 1H), 7.26 (t, J = 7.4, 2H), 7.14 (t, J = 7.4, 1H), 4.05 (s, 2H), 4.02–3. 99 (m, 1H), 2.08 (d, J = 11.2, 2H), 1.81 (d, J = 13.2, 2H), 1.70 (d, J = 12.8, 1H), 1.45-1.35 (m, 2H), 1.32-1.30 (m, 11H), 1.21-1.17 (m, 1H); ¹³C NMR (DMSO-d6) δ 161.40, 135.64, 128.00, 127.79, 123.91, 120.11, 119.94, 119.45, 112.26, 51.32, 48.29, 33.35, 32.95, 25.74, 25.19, 24.08; MS (ESI) m/z calcd for C₂₀H₂₉N₃O ([M - H]-) 326.23; found 326.33.

*Three protons (-CH₃) at third positions of indoles split into two singlets because the compounds 3a and 3b are mixtures of their respective geometric isomers.

**Three protons (-CH₃) attached to cyclohexyl rings split into two doublets for compounds 3a and 3b as they are mixtures of their respective geometric isomers.

Mycobacterium tuberculosis inhibition assay

All the final synthesized indole-2-carboxamide derivatives were evaluated for in vitro antitubercular activity against Mycobacterium tuberculosis H37Rv strain. Minimum inhibitory concentrations (MIC) were determined in two different media: (1) 7H9/ADC/Tw consisting of Middlebrook 7H9 broth base supplemented with 0.2% glycerol, 0.2% glucose, 0.5% BSA fraction V, 0.08% NaCl and 0.05% Tween 80 and (2) 7H9/glucose/casitone/Tx consisting of Middlebrook 7H9 broth base supplemented with 0.4% glucose as carbon source, 0.03% Bacto casitone, 0.08% NaCl and 0.05% tyloxapol. The assay was performed using the round-bottom clear 96-well plates containing 50 μL of medium of choice in each well with or without serially diluted compound to which an equal volume (50 μL) of Mycobacterium tuberculosis diluted from an early log-phase culture in the same medium of choice was added. The final cell concentration in each well was 1x10⁴ cfu/well and the final compound concentrations ranged from 50 – 0.024 μM. Plates were incubated at 37 °C and growth monitored using an inverted enlarging mirror plate reader after 1- and 2-weeks of growth. The solvent used for preparing the stock solutions of the compounds was DMSO. The lowest concentration of compound that completely inhibited the growth of bacterium was reported as MIC [17, 34]. The positive control used for the study was isoniazid (Table 1).

Table 1.

Synthesized indole-2-carboxamides and their antitubercular activity.

Compound Name	dGBind (kcal/mol)	1-week MIC 7H9/ADC/Tw (μM)	2-week MIC 7H9/ADC/Tw (μM)	1-week MIC 7H9/glucose/casitone/Tx (μM)	2-week MIC 7H9/glucose/casitone/Tx (μM)	PlogP**
3a* (3a2C/3a2’C/3a2T/3a2’T)	−42.20/−44.25/−40.94/−39.17	12.5	12.5	3.13	6.25	4.04/4.04/4.03/4.03
3b* (3bC/3bT)	−39.72/−42.43	4.7	4.7	1.56	1.56	4.03/4.04
5a	−53.067	>50	>50	6.25	12.5	4.72
5b	−58.156	>50	>50	3.13	4.7	4.70
5c	−52.32	37	37	19	19	3.56
5d	−54.2	>50	>50	>50	>50	4.07
5e*	−32.75/−42.70	>50	>50	>50	>50	3.07/3.03
5f	−52.84	>50	>50	≥50	>50	4.03
5g	−40.44	50	≥50	25	37	3.66
5h	−59.66	>50	>50	25	37	4.03
Isoniazid	-	0.2 ± 0.1	0.2±0.1	0.1±0.05	0.15±0.05	-

Note:

^*3a2C and 3a2’C represent the mirror images of the cis form and 3a2T and 3a2’T represent the mirror images of the trans form of 3a, 3bC and 3bT represent the cis and trans forms of 3b whereas 5eC and 5eT represent the cis and trans form of 5e respectively;

^**PlogP = Predicted logP;

Molecular dynamics simulations

Molecular dynamics (MD) simulations were performed using Desmond from Schrodinger Suite with the previously docked complexes B38-Mmpl3, 3bC-Mmpl3, and 3bT-Mmpl3 complexes in order to analyse the intricacies of the molecules’ binding to Mmpl3 for a certain time period. The system building and the MD protocols were kept same for all the complexes. The initial step included building of a system through the System Setup panel in Maestro for each of the complexes. The systems were set up with single point charge (SPC) explicit solvent model inside an orthorhombic box and a membrane was setup with the Dipalmitoylphosphatidylcholine (DPPC) model for the transmembrane part of Mmpl3. Maestro’s default relaxation protocol, which included two stages of minimization (restrained and unrestrained) followed by four stages of MD runs with gradually diminishing restraints, was used. The simulations were run using the NPT ensemble which essentially meant that a constant number of atoms/molecules, and a constant temperature and pressure was maintained for the system during the simulation. Noose-Hover thermostat method and Martyna-Tobias-Klein barostat method were used to maintain constant pressure and temperature. OPLS3e was the force field used in the simulations with the neutralization of the systems were done by adding 19 Cl⁻ ions. The time step used for the simulations was 2 femtoseconds [35–37].

Results and Discussion

Refinement of the Homology Model

The Ramachandran plot after the MacroModel energy minimization showed that seven amino acid residues (ASP787, GLU814, THR789, ALA778, PHE611, HIS461, and LEU792) along with proline or glycine residues were present in the white region of the plot which suggested dihedral clashes (Fig 2a). The issue was resolved with loop refinement on the non-template regions and subsequent refinement through the Protein Preparation Wizard (Fig 2b). As a result, only one residue (ARG939) along with proline or glycine remained in the disallowed region, but it being nowhere near the binding site, we believed that this would cause an insignificant effect during simulation of the whole protein.

Analysis of docked complexes

From the molecular docking results, the essential regions in the binding pocket were identified. The type and number of interactions were also analysed for the potent molecules.

The dG bind scores from MMGBSA gave a quantitative idea about how strongly the ligands bound to the target (Table S1). In figure 3, docked poses of six MmpL3-ligand complexes are shown. Among them, three were potent molecules and the other three were inactive. These structures can be considered a representation of all the molecules in the list for understanding the interactions of the indole-2-carboxamides with MmpL3. In the complexes formed by both B38 and A12, a hydrophobic enclosure was observed, formed by the residues LEU637, TYR641, PHE644, PHE255, TYR252 and ILE248. Interestingly, not a single polar interaction was seen for these two molecules (Fig. 3a & S1a). It is clear from these poses that a bulky group would interact with the hydrophobic residues in the pocket through van der Waal’s forces, was the most essential factor. In B38, there is a bi-cyclohexyl ring attached to the amide nitrogen and similarly, whereas in the case of A12, there is a cyclooctyl ring attached that would fit inside the pocket. For the less potent molecules A16 and A27, the cycloalkyl rings were positioned opposite to the hydrophobic pockets (Fig. S1b and 3b). In case of A16, it is possible that the N-methyl group resulted in a stable position of the molecule where the indole part being placed in the hydrophobic pocket. On the other hand, in case of A27, the absence of the bulky cycloalkyl group resulted in the indole region to position itself inside the hydrophobic pocket. In both the cases, the hydrophobic pockets were relatively emptier than the complexes of MmpL3 with the potent molecules, A12 and B38. The importance of the indole ring is asserted in Fig. 3c & 3d. Since there were no polar interactions observed between the indole nitrogen and the amino acid residues, it is quite evident that the nitrogen atom interacted with ILE248 through Debye forces (dipole-induced dipole) for A12 (Fig. 3c). But in the case of A19, although there were possible Debye interactions between the benzoxazole oxygen and ILE248, there was a total inversion of the structure i.e. the indole end was positioned inside the hydrophobic pocket instead of the cyclohexyl end (Fig. 3d). Due to higher electronegativity of oxygen than nitrogen, the resonance effect is probably higher in indole than in benzoxazole. Higher resonance resulted in stronger affinity of indole nitrogen towards ILE248 through the Debye forces. Thus, the free rotation at that position would be more in case of A19. We believe, this resulted in van der Waal’s clashes at that end of the binding pocket and hence the molecule found a more stable pose by inverting itself to the opposite end leaving the hydrophobic pocket empty. This explains the 100-fold better activity of A11 over A19 (Table S1).

Figure 3.

(a) Protein-ligand complex of MmpL3 and B38. (b) Protein-ligand complex of MmpL3 and A27. (c) Orientation of A12 in the vicinity of ILE248 at the binding site of MmpL3. (d) Orientation of A19 in the vicinity of ILE248 at the binding site of MmpL3. In the figures, the green coloured ball and stick structures represent the ligand, the grey coloured ball and stick structures represent the residues at the binding site, and the grey coloured surface signify the hydrophobic pocket formed by the non-polar residues at the binding site. The magenta dotted lines represent the distances between atoms.

Molecular Docking and MMGBSA results of the designed molecules

The docked poses of the finally selected ten molecules are shown in Fig. 3. 3a, 3b and 5e are conformational isomeric mixtures of their respective isomers and each isomer for these compounds behaved differently at the binding site. The isomers are represented by the names 3a2T, 3a2’T, 3a2C and 3a2’C (Fig. S4). The cis isomers (dG bind 3a2C = −42.20 kcal/mol, dG bind 3a2’C = −44.25 kcal/mol) had relatively better binding scores than the trans isomers (dG bind 3a2T = −40.94 kcal/mol, dG bind 3a2’T = −39.17 kcal/mol) due to the methyl groups (at 3rd position of indole) occupying the extended hydrophobic pocket in the trans configurations (Fig S2a, S2b, S2c and 4a). The cis and the trans isomers of 3b (Fig. S4) were represented by the names 3bC (dg bind = −39 kcal/mol) and 3bT (dg bind = −42 kcal/mol) and their docking poses with Mmpl3 are shown in Fig. S2d and 4b, respectively.

Both the 4-methyl substituted cyclohexyl group and the methyl group at the third position of the indole moiety were ideally placed in the hydrophobic pockets in both the isomers which was reflected in their dG bind scores. On the other hand, the trans and cis isomers of 5e (Fig. S4) are represented by names 5eT (dG bind = −32.74 kcal/mol) and 5eC (dG bind = −42.70 kcal/mol) in Fig. S2e and 4c, respectively. The docked poses show that the hydrophobic pockets for both the isomers were empty at the location where the cyclohexyl groups of 3a and 3b had found a good fit. The docking poses of molecules 5a and 5b with MMPL3 are represented in Fig. S3a and 4d, respectively. Quite predictably, 5b (dG bind = −58.156 kcal/mol), which has a methoxy substitution at the fourth position of the benzyl ring, had covered a better part of the hydrophobic pocket than 5a (dG bind = −53.067 kcal/mol). The difference in their dG bind scores can be attributed to this. The docking poses of 5c (dG bind = −52.32 kcal/mol), 5d (dG bind = −54.20 kcal/mol), 5f (dG bind = −52.84 kcal/mol), 5g (dG bind = −40.44 kcal/mol) and 5h (dG = −59.66 kcal/mol) with MmpL3 are illustrated in Fig. S3b, S3c, S3d, S3e and S3f, respectively. It can be seen that except for 5g, all the other molecules have shown excellent dG bind scores. Although the hydrophobic pocket is not filled by these molecules in the manner it did for the already reported potent molecules as discussed in section 3.2, these molecules covered a large area of the inner non-polar surface of the protein (except for 5c).This maybe one of the major reasons for their high binding scores. It must be noted that, for the molecules 5c, 5d and 5h, it was the 3rd position substituted component that was present near the hydrophobic site. Also, the indole moiety was not in the plane of the binding pocket (except for 5g) for these molecules.

Precision of MMGSBA dG Bind

In order to estimate the precision of the MMGBSA dG bind score of the minimised (docked) structures, a sample molecule (B38) was and its complex with Mmpl3 was used. The method included running 5 MD simulations of 25 ns with the initial B38-Mmpl3 complex (previously mentioned MD protocol was followed), each time with a random seed and a random velocity. Based on the time taken for equilibration, relevant frames/snapshots were used and MMGBSA was performed for all the frames. Finally, the ensemble average of the dG binds from each of the MD runs were used to calculate the standard deviation and the standard error with respect to the previously calculated dG bind score (docked complex). The standard deviation and the standard error was calculated to be 0.239 kcal/mol and 0.107 kcal/mol respectively, which we assume was highly within acceptable limit considering the applicability domain of our model (Table S2) [38, 39].

Synthesis and spectral characterization of designed molecules

Scheme 1 shows the multi-step synthesis of indole-2-carboxamides from a 2-halo aryl ketone as the starting material. The first step involved the Cu-catalysed condensation/deformylation of 2’-bromo acetophenone with ethyl isocyanoacetate to give 1 which was then purified by column chromatography (overall 75 % yield) [30]. In the next step, 1 was hydrolysed in presence of NaOH and the reaction was carried out at 80°C. Pure solid was extracted after the mixture was turned slightly acidic by adding dil HCl to give 2 with a yield of 90 % [31]. The ¹H NMR spectra of 2 (Fig. S5) clearly shows the presence of four aromatic protons at 7.03 - 7.64 ppm, one aromatic N-H at 13.63 ppm, and an amide N-H at 12.89 ppm. In the final step, amide coupling reaction took place with EDC.HCl and HoBt was used as coupling agents and triethylamine as a proton abstractor from primary amines. 2-methylcyclohexylamine and 4-methylcyclohexylamine were the primary amines used in these reactions to give 3a and 3b, respectively [13, 14]. These compounds were purified by washing with methanol to give yields of 80% and 83% respectively. The characteristic peaks for the aliphatic protons were observed at 0.9 – 2.2 ppm for both the compounds. The individual aromatic N-H peaks and the four aromatic proton peaks were quite expectedly observed at 9.05 - 9.06 ppm, and at 7.12 - 7.65 ppm, respectively. Interestingly, three protons of the methyl group at the third position of indole appeared as two peaks instead of a single one. The same phenomenon was seen for the methyl groups attached to the cyclohexyl rings of the respective compounds. This was caused due to the compounds being present as a mixtures of their repective conformational isomers viz. trans and cis conformations, for 3a and 3b. The MS (ESI) showed ([M + H]⁺) peaks corresponding to their molecular formula weights (Fig. S27 and S28). The synthesis of compounds 5a – 5h are shown in Scheme 2. In the first step, 4 was synthesized from indole-2-carboxylic acid following the same amide coupling reaction as for 3a and 3b [13, 14]. Similarly, the compound was purified by washing with methanol giving a yield of 85%. The ¹H NMR spectra of 4 (Fig. S16) clearly show the presence of 10 aliphatic protons at 1.30 – 1.85 ppm, a characteristic multiplet for 1 proton at 3.78 ppm, 4 aromatic protons from 8.18 – 7.0 ppm, and the characteristic aromatic N-H proton peak at 11.50 ppm. In the next step, Mannich type reactions were performed with formaldehyde, respective amines, and 4 where the third position of the indole acted as the nucleophile [32, 33]. Quite interestingly, the reactions only progressed significantly in weakly acidic condition with acetic acid being used as the solvent. The yields varied from 75% to 83%. The ¹H NMR spectra of 5a – 5h (Fig. S7 – S14) illustrate similar results for aliphatic, aromatic, and aromatic N-H protons like the previously mentioned indole-2-carboxamides (compounds 3a, 3b, and 4). The characteristic peaks for the N-attached aliphatic protons were observed at 3.33 – 4 ppm. The MS (ESI) showed ([M + H]⁺/([M - H]⁺) peaks corresponding to the molecular formula weights of 5a – 5h (Fig. S29 – S36).

Scheme 1.

^aReagents and conditions: (i) CuI, Cs₂CO₃. ethyl isocyanoacetate, DMSO, 90° C, 16 hours; (ii) NaOH, MeOH, water, 80° C, 2 hours; (iii) EDC.HCl, HoBt, methylene chloride, triethylamine, corresponding cyclohexylamines, R.T., 16 hours.

Scheme 2.

^bReagents and conditions: (i) EDC.HCl, HoBt, methylene chloride, triethylamine, cyclohexylamine, R.T, 16 hours; (ii) corresponding amines, formaldehyde, acetic acid, R.T, 1.5 to 2 hours.

Antitubercular Evaluation Results

The MIC values for the synthesized compounds are shown in Table 1. Compounds 3a, 3b, 5a and 5b are the potent compounds in the list (MIC ≤ 12.5 μM). Compounds 3a and 3b showed high potency in both the media whereas 5a and 5b were significantly active in 7H9/glucose/casitone/Tx media which was devoid of BSA suggesting probable BSA binding of these analogues. Among all the molecules, 3b showed the lowest MIC (1.56 μM for 1-week and 2-weeks 7H9/glucose/casitone/Tx medium) and thus it was our molecule of interest for further investigation.

Molecular Dynamics Simulations

The MD simulation results for the complex of Mmpl3 and B38 is shown in Fig. 5. The protein root-mean-square deviation (RMSD) showed a steady rise till 50 ns followed by a steep spike and again stabilising till the end of the simulation. The RMSD after 50 ns was in the range of 10.5 Å to 12.0 Å. But quite interestingly, the ligand RMSD (calculated with reference to the protein backbone) was stable throughout and was in the range of 4.5 Å to 6.0 Å (Fig 5a). This signify that the protein-region complex region was highly stable and the high RMSD values of the protein resulted from deviations in the loop regions and it did not affect the binding site. It was understood that the simulation had converged after a certain amount of time. The protein root-mean-square fluctuation (RMSF) which was calculated for the whole simulation provides the justification for the overall high RMSD in the protein (Fig. 5b). The green strands intersecting the residues in the graph signify the ligand interaction sites whereas the light pink strands intersect the residues which make up the secondary structures of the protein. It is quite clear that the regions of the protein which contributed to binding were in the low RMSF regions. The protein region from residue 780 onwards showed high spikes in RMSF and thus this loop region contributed to the higher values of deviation. It can be noted that this region took no part in binding interactions and was not a part of the secondary structure of the protein. Fig. 5(c) represents the 2D interaction summary of B38 with Mmpl3 in terms of percentage for the entire simulation. The filling of the hydrophobic pocket by the bi-cyclohexyl ring was in tandem with the docking results. But interestingly, the amide N-H formed two water bridges with GLY636 and ASP640 for 76 % and 77 % of the time respectively and this observation was missing in XP docking pose. It is highly likely, that the polar interactions also played a deterministic role in the molecule’s binding to Mmpl3. The protein-ligand RMSD graph for the MmpL3-3bT is shown in Fig. 6a. The protein RMSD became stable post 20 ns and the values ranged from 9 Å to 10.5 Å till the end of the simulation. The ligand RMSD dipped post 4 ns, showed a jump after 40 ns and stabilised thereon. Majority of the RMSD curve of the ligand was positioned between 3 Å to 6 Å (Fig. 6a). As in the case of MmpL3-B38 complex, the high RMSD range of the protein could be attributed to the unstable nature of the loop region (residue 800 onward). Both the protein and that ligand RMSD indicate that the simulation had converged. Fig. 6b justify the reason for high protein RMSD where it can be seen that the ligand binding region of the protein was much stable and unaffected by the unstable loop region. The 2D interaction summary is represented in figure 6c. In addition to the 4-methyl substituted cyclohexyl ring filling the hydrophobic pocket, there was an H-bond between the amide N-H and ASP251 for 93 % of the simulation time. The protein-ligand pose which resulted from Glide docking did not show any H-bond interaction. Fig. 7a represents the protein-ligand RMSD graph for MmpL3-3bC. It is quite evident that the protein was significantly unstable throughout the simulation. The RMSD ranged from 6 Å to 12 Å for majority of the time period. Whereas, the ligand RMSD became stable from 30 ns and hovered over 6 A to 8 A thereon. Quite expectedly, the protein RMSF of this complex was similar to that of MmpL3-3bT and MmpL3-B38, where it was evident that the loop region was contributing majorly for the high values of deviation in the protein structure (Fig. 7b). Interestingly, the 2D interaction summary shows that the 4-methyl substituted cyclohexyl part was outside the hydrophobic pocket and the amide carbonyl formed water bridges with ASP640 and GLY636 for mere 27% and 23% of the simulation time respectively (Fig. 7c). These observations clearly indicated that the protein-ligand complex was highly unstable in comparison to the previous complexes and it was quite evident that the potency of 3b was majorly due to the trans-isomeric from (3bT) which showed appropriate stability and binding in the protein pocket.

Fig. 5.

(a) Graph displaying protein (blue) and ligand (red) RMSD vs simulation time for MmpL3-B38 complex. (b) Graph displaying protein RMSF (blue) vs residue numbers. Pink lines represent the alpha α-helix, light blue lines represent the β-pleated sheets and the green lines represent the ligand contact points of the protein for Mmp13-B38 complex, (c) 2D contact summary of MmpL3 and B38 where pink arrows represent polar hydrogen bonding, green colour ribbon indicates hydrophobic region, and the polar regions are represented by the red ribbon.

Fig. 6.

(a) Graph displaying protein (blue) and ligand (red) RMSD vs simulation time for MmpL3-3bT complex. (b) Graph displaying protein RMSF (blue) vs residue numbers. Pink lines represent the alpha α-helix, light blue lines represent the β-pleated sheets and the green lines represent the ligand contact points of the protein MmpL3-3bT complex, (c) 2D contact summary of MmpL3 and 3bT where pink arrows represent polar hydrogen bonding, green ribbons indicate hydrophobic regions, and the polar regions are represented by red ribbons. 3bT represents the trans form of 3b.

Fig. 7.

(a) Graph displaying protein (blue) and ligand (red) RMSD vs simulation time for MmpL3-3bC complex, (b) Graph displaying protein RMSF (blue) vs residue numbers. Pink lines represent the alpha α-helix, light blue lines represent the β-pleated sheets and the green lines represent the ligand contact points of the protein MmpL3-3bC complex, (c) 2D contact summary of MmpL3 and 3bC where pink arrows represent polar hydrogen bonding, green ribbons indicate hydrophobic regions, and the polar regions are represented by red ribbons. 3bC represents the cis form of 3b.

Molecular Insights

A few important aspects of the MmpL3 binding pocket were understood from this study. The hydrophobic pocket formed by the residues TYR641, PHE644, PHE255, TYR252 and ILE248 was identified and its importance was asserted. It was observed from the docked poses of the reported molecules that hydrophobicity of the molecules was the most important driving factor that determined affinity at the binding site. We also predicted that the indole nitrogen participating in ring resonance was interacting with ILE248 through Debye forces. Inability of molecules to interact through this force resulted in inversion of structures with the indole end of the molecule placed in the hydrophobic pocket instead of the cyclic hydrophobic groups. Keeping in mind synthetic feasibility, we designed multiple molecules with different types of substitutions at the third position. Docking poses of the isomers of our designed molecules 3a and 3b (both 3a and 3b are composed of their respective stereoisomeric mixtures) showed that another hydrophobic interaction was possible with ILE248 for molecules with non-polar groups at the third position of indole. The methyl substituted cyclohexyl group of 3b circumvented the hydrophobic pocket in a much ideal way than 3a and this observation was in tandem with the 6 times higher potency of 3b than 3a (in 2-week 7H9/glucose/ casitone/Tx medium) . The docked complexes of the Mannich products (5a to 5h) with MmpL3 revealed another hydrophobic surface with which interaction was possible. ILE244, LEU243, VAL240, LEU629 and LEU637 formed the surface. In case of 5a, it was observed that the non-polar portions of the molecule were in contact with this surface but the hydrophobic pocket was empty. Whereas in 5b, both the surface and the pocket were filled by the non-polar portions of the molecule. The lower dG bind score and MIC (dG Bind = −53.067 kcal/mol for 5a and −58.156 kcal/mol for 5b, MIC = 6.25 μM for 5a and 3.13 μM for 5b) of 5b than 5a supported this prediction. It was interesting to note that 5a and 5b showed potency in the 7H9/glucose/casitone/Tx medium and were totally inactive in 7H9/ADC/Tw suggesting high protein binding to the bovine serum albumin component of 7H9/ADC/Tw. The docked complexes of both the cis and trans isomers of 5e with MmpL3 clearly showed that the molecules were placed perpendicularly to the inner surface of the hydrophobic pocket thus resulting in decreased non-polar interactions. As expected, both the dG bind (dG bind = −32.75 kcal/mol for 5eT and −42.70 kcal/mol for 5eC) and MIC (>50 μM for all media) correlated with the observations made from the docked complexes. The docked pose of 5c-MmpL3 complex (dG bind = −52.32 kcal/mol) showed that the indole side of the molecule was bound to the hydrophobic pocket while the cyclohexyl group interacted with the other hydrophobic surface. In case of 5d (dG bind = −54.2 kcal/mol), the cyclohexyl group was bound to the hydrophobic pocket and the 3rd position substitution with methylene piperidinyl group interacted with hydrophobic surface. Even with very promising dG bind scores, these two compounds were inactive against M.Tb (MIC = 19 μM for 5c and >50 μM for 5d in 2-week 7H9/glucose/ casitone/Tx medium). Similarly for 5f (dG bind = −52.84 kcal/mol) and 5h (dG bind = −59.66 kcal/mol), the very strong binding scores did not correlate with the MIC values (> 50 μM for 5f and 37 μM for 5h). While in case of 5g, the 3rd position methylene N-alkyl part was present near the hydrophobic pocket and the cyclohexyl moiety interacted with the hydrophobic surface. Lower hydrophobicity of N-alkyl in comparison to the previously mentioned Mannich compounds might be reason of its relatively higher dG bind score of −40.44 kcal/mol. As observed in case of 5c, 5d, 5f and 5h, the dG bind score of 5g did not correlate with its high MIC value (37 μM in 2-week 7H9/glucose/casitone/Tx medium). The lack of whole cell activity may be driven by permeability barriers and resultant inability of these molecules to reach the target protein in sufficient concentrations.

3b, which had the lowest MIC value, was our molecule of interest for further investigation. Initially, a MD simulation was run for the homology model of MmpL3 and B38. The data generated from this model was used as standard since B38 was the most potent among the previously reported molecules. Two separate MD simulations were run with the isomers of 3b (3bT and 3bC). The high values of protein RMSD for 3bT and 3bC throughout the simulation was consistent with that of B38. In all the three cases, it was discovered that the protein RMSF was exceedingly high at the loop regions from residue 750 and above. Although this resulted in an overall higher RMSD graph for the protein, the binding site was unaffected by the loop instability at these regions of the protein. In case of B38, the highly stable ligand RMSD graph, presence of two consistent water bridges with GLY636 and ASP640, and non-polar interactions between the bi-cyclic ring and the hydrophobic pocket correlated with its very high potency. Whereas in case of 3bC, the 2D ligand-protein interaction for the entire simulation justified the ligand’s unstable RMSD plot. It is clear from these observations that the 4-methyl substituted cyclohexyl group clearly remained outside the hydrophobic pocket for most of the simulation time and there were no significant polar interactions. The ligand RMSD of 3bT was much more stable when compared to 3bC. The consistent polar and non-polar interactions (2D ligand-protein interaction summary) for this isomer supported this prediction. The high potency of 3b (1.56 μM in 2-week MIC 7H9/glucose/casitone/Tx) likely resulted from 3bT binding to the MmpL3 binding site as 3bC did not have play any role in binding and inhibiting the protein. We predict that the MIC value of the specific isomer 3bT will be superior to that of 3b. In summary, we identified the compound 3b as a potent antitubercular hit and predicted that its trans-isomer 3bT is the active configuration that inhibits MmpL3 holding promise for its further development in the drug discovery pipeline.

Conclusion

In this work, we developed a homology model of Mycobacterium tuberculosis MmpL3 from the template structure of Mycobacterium smegmatis MmpL3. The model was validated by analysing the internal parameters and molecular dynamics simulations. Docking with several reported indole-2-carboxamides helped us to identify significant regions of the protein pocket responsible for binding with ligands. Based on this data, we designed and synthesized ten novel indole-2-carboxamides with substitutions at the third position of the indole molecules, which we predicted would inhibit MmpL3. From the in vitro whole cell assay, we identified four molecules as antitubercular agents with promising potency (MIC ≤ 12.5 μM). Among them, the compound 3b, showed the highest potency (MIC = 1.56 μM). From the molecular dynamics simulation studies, we determined that it is the trans-isomer of 3b, i.e. 3bT, which must be predominantly responsible for the molecule’s potency. This creates opportunities for further studies with the molecule and especially its trans-isomer. We believe that 3bT holds potential for further assessment in the drug discovery pipeline. This work identified novel antitubercular hits and helped to illustrate the interaction patterns of the novel indole-2-carboximides with MmpL3.

Figure 4.

(a) Protein-ligand complex of MmpL3 and 3a2’C. (b) Protein-ligand complex of MmpL3 and 3bT. (c) Protein-ligand complex of MmpL3 and 5eC. (d) Protein-ligand complex of MmpL3 and 5b. In the figures, the green coloured ball and stick structures represent the ligand, the grey coloured ball and stick structures represent the residues at the binding site, and the grey coloured surface signify the hydrophobic pocket formed by the non-polar residues at the binding site. 3a2’C represents 2’ mirror image of the cis form of 3a whereas 3bT and 5eC represent the trans and cis form of 3b and 5e respectively.

Design, System, Application.

This research work delves into the molecular system of the Mmpl3 protein present in Mycobacterium tuberculosis. A computational model was developed based on a template structure thus initiating the step towards gaining deeper insights into the protein’s molecular structure. New molecules were designed based on the chemical information. The regions responsible for binding with the inhibitors were identified and the essential interactions along with binding free energies were used as parameters for screening the newly designed molecules. Novel molecules were thereby synthesized and their biological screening was done against Mycobacterium tuberculosis to validate our model. Finally, molecular dynamics simulations were performed to understand how one of the hits interacted with Mmpl3 leading to its antitubercular property. This work has not only lead to the development of new antitubercular molecules but also helped us to understand how each of the isomeric forms were possibly superior to the others in binding with Mmpl3 and thus opening new doors in the field of antitubercular drug discovery.