Rv1258c is a significant mycobacterial efflux pump that contributes to multidrug resistance by conferring resistance to numerous antibiotics. As a result, finding a potent EPI against this pump is a crucial area of research. Therefore, the present study aims to look for potent EPIs from plant sources, as these have a number of benefits as compared to synthetic sources. For the study, we retrieved the amino acid sequence of Rv1258c from KEGG having accession number T30239:34247, and it was seen that the protein consists of 419 amino acids. Initially, a BLAST search was performed, which indicated that no close homolog of the protein of interest was available. Consequently, the distant homologs algorithm was employed for predicting the structure using the PSIPRED web server. The search resulted in six templates based on a significant p value (less than 0.001), out of which a glycerol 3-phosphate transporter (1pw4) from E.coli was ranked at the top with a score of 331 (p = 1*10–8). Templates were also searched using the MODELLER 9.24 web server using the profile.buildcommand, which yielded the same outcome, suggesting 1pw4 as the template. The results were found to correlate with the structure prediction results of Cloete et al. 2018. Further, the model was built up using using Align2D command of Modeller 9.24, which yielded 10 models, and the model with the lowest DOPE (Discrete Optimized Protein Energy) score was selected as our model. A sample of native protein structures was used to calculate the DOPE, or atomic distance-dependent statistical potential. Its sole foundation is the probability theory. The joint probability density function (pdf) of the atomic Cartesian coordinates' negative logarithm is a statistical potential. As a result, the DOPE method is a statistical method for evaluating built structures and determining the most accurate structure. From a group of models created by MODELLER 9.24, the best structure can be chosen using the DOPE model score. A model is better if its DOPE score is lower. The DOPE scores along with the model are shown in table (Additional file 1: Table S2). The 9th model has the lowest DOPE score and hence, selected as the best model for the molecular docking.

This model was further refined using the 3Drefine web server, and the refined model with the lowest energy is selected for further analysis. The refined model was analysed by studying the Ramachandran plot as shown (Additional file 1: Fig. S1a) which showed that 89.1% of the residues fall in the allowed region, with 2 residues falling in the disallowed region, and a ProSa Z score of − 4.61 was obtained (Additional file 1: Fig. S1b). The combination of the backbone dihedral angles is statistically represented in the Ramachandran plot. To further analyse the predicted structure, the templates were superimposed, which showed less deviation from the folds predicted (Additional file 1: Fig. S2).

After the model was created and validated, molecular docking was carried out. In the present study, blind and site-specific docking, both approaches were adopted. Blind docking involves attaching a ligand to the entire surface of a protein without any knowledge of the target pocket. Blind docking necessitates numerous trials/runs and energy calculations prior to identifying a favourable protein–ligand complex pose. In site-specific docking, locus for ligand to bind is predicted using the f pocket web server and based on the results the grid box is designed as per the x, y and z axis values given.

In the present study, for the site-specific docking, the active site for the ligand to bind was analysed by using f pocket server. The active site of the protein and grid box around the protein is shown in figures in the supplementary file (Additional file 1: Figs. S3 and S4).

For molecular docking,AutoDock Vina software was used for both blind and site-specific docking experiments for the 210 compounds from various medicinal plants selected based on the traditional and scientific reports that they work against tuberculosis. It was found that 10 out of 210 phytocompounds showed low binding energy. These 10 ligands with their docking scores, along with the standard error of the dataset, are depicted (Table 1). The ligands N-transferuroyl-4’-O-methyldopamine, ellagic acid, abyssinone II, mollic acid glucoside, glabridine, chrysoeriol, naringenin, luteolin, isoliquiritigenin, and baicalein had blind docking scores of −9.1, −9.6, −9.6, −9.2, −9, −8.6, −8.5, −8.4, −7.9 and −9 respectively. Similarly, for site-specific docking, these compounds exhibited docking scores of −8.7, −8.6, −8.6, −8.5, −8.5, −8.5, −8.4, −8.4, −8.4, −8.3 respectively. The docking experiments were performed in triplicate, and the standard error was found to be negligible in all the sets of experiments. In the study, it was found that in blind docking, the binding energy of piperine was −9.3. Blind docking scores of −9.2 and −8.5 for piperine and verapamil, respectively, site-specific docking of verapamil and piperine showed docking scores of −5.6 and −9.2, respectively, in the present study.

The ten docked complexes were subjected for interaction studies in the LigPlot + v. 2.2.5 software between the ligands and Rv1258c and these are depicted in figures (Additional file 1: Fig. S5ai–tii) in the supplementary file. From the figures, we see that the ligands and Rv1258c interact mainly by hydrogen bonding and hydrophobic interactions. Blind docking analysis of N-transferuroyl-4'-O-methyldopamine reveals that the ligand and macromolecule interact via two H bonds: one with Thr51 bond length 3.04 and one with Pro361, bond length 2.81. For site-specific docking with the same compound, we find 2 H bonds- 1 H bond with Ala301, bond length 2.70 Å, 1 H bond with Ser244 bond length 3.26 Å. In the case of ellagic acid, for blind docking, we find 3 H bonds, 1 H bond with Ala48 bond length 3.14 Å, 1 H bond with Leu364 bond length 2.70 Å, 1 H bond with Ser244 bond length 3.00 Å. For site-specific docking, we find 3 H bonds- 1 H bond with Tyr146 bond length 3.16 Å, 2 H bonds with Asn151 bond length 2.80 Å each. Blind docking yields two H bonds for Abyssinone II: one with Gly 117 (bond length 2.70) and one with Asp23 (bond length 2.93).In case of site-specific docking we not see any H bonds, but hydrophobic interactions are responsible for docking. The residues involved in hydrophobic interactions are Val245, he246, Phe330, Ala 306, Leu311, Lys249, Gly368, Ser45, Tyr250, Ser45, Ala367, Val245, Leu343, Phe247, Ser244. The ligand, mollic acid glucoside, has 5 H bonds involved in the blind docking interaction.These are two H bonds with Asn221 with bond lengths of 3.28 and 2.21, respectively, two H bonds with Val224 with bond lengths of 3.02 and 2.70, and one H bond with Leu222 with bond length of 2.64.In the case of site-specific docking, we see 3 H bonds- H bond with Ala319 bond length 3.10, 1 H bond with Leu317 bond length 3.05, 1 H bond with Ala266, bond length 3.01 Å. In the case of glabridine for blind docking, we have found 2 H bonds with Arg124 with bond length 2.82 Å and 3.35 Å respectively. For site-specific docking, we find only 1 H bond at Gly368 bond length 2.93 Å. In the case of chrysoeriol, for blind docking, we see 4 H bonds- 1 H bond with Arg134, bond length 3.24 Å, 3 H bonds with Arg124 with length 3.10 Å, 2.82 Å, 2.99 Å respectively. For site-specific docking with the same compound, we find 2 H bonds- 1 H bond with Thr290, bond length 2.91 Å, 1 H bond with Asn151, bond length 2.92 Å. In the case of naringenin, for blind docking, we see 4 H bonds Bond lengths for three H bonds with Asp 79 are 2.72, 2.91, and 3.31, respectively, and one H bond with Ser 26 is 2.75.In the case of luteolin, for blind docking, we find 2 H bonds- 1 H bond with Thr290, bond length 2.96 Å, 1 H bond with Asn151 bond length 3.01. For site-specific docking, we find 3 H bonds- 1 H bond with Thr290 bond length 2.92 Å, 2 H bonds with Asn151 bond length 2.92 Å and 3.00 Å respectively. For isoliquiritigenin, in blind docking we find 2 H bonds- 2 H bonds with Gln329 with bond length 3.02 Å and 3.22 Å respectively. For site-specific docking, we do not find any H bonds. Only hydrophobic interactions are present and the residues involved are Val245, Phe330, Leu383, Lys249, Thr379, Thr371, Tyr250, Ala306, Phe245, Val245, Phe380. In the case of baicalein, for blind docking, we find 4 H bonds- 1 H bond with Leu246 bond length 2.71 Å, 2 H bonds with Ser244 bond length 2.78 Å, 3.06 Å, 1 H bond with Leu364 bond length 2.71 Å respectively. For site-specific docking, we find 3 H bonds- 1 H bond with Tyr146 bond length 2.88 Å, 2 H bonds with Asn151 with bond length 2.98 Å, 2.94 Å respectively.

Additionally, 10 compounds selected based on the docking results were subjected to drug likeliness studies using different web servers, namely Molinspiration, pkCSM, the Lipinski filters server, DruLiTo, and SwissADME and the results are depicted in table (Table 2). The ADMET properties of these compounds were also assessed using various web servers, namely, pkCSM, ProTox-II, admetSAR 2.0, and admetLab 2.0. and the results are compiled (Table 3). From the drug-likeness studies, we find that except for mollic acid glucoside, all the other phytocompounds follow all the Lipinski Rule of 5. Mollic acid glucoside has two violations: a higher molecular weight and a larger number of H bond donors. ADMET analysis showed that compounds showed acceptable ADMET properties, and the results were consistent across four different web servers. However, one of the compounds, glabridine, exhibits AMES toxicity using the admetSAR 2.0 tool and is carcinogenic using the ADMETlab 2.0 tool. Naringenin, isoliquiritigenin were also carcinogenic when using the same tool. Luteolin and isoliquiritigenin are also seen to be positive for AMES toxicity using ADMETlab 2.0. Luteolin is also seen to be positive for mutagenicity and carcinogenicity using ProTox-II software. Using ADMETlab 2.0, we can see that our control piperine is toxic and carcinogenic to AMES.

Using two programmes, admetSAR 2.0 and SwissADME, the bioavailability scores of the 10 compounds were assessed (Table 4). In order to evaluate acute oral toxicity, we used the LD50 test. The LD50 is the dosage of a medication required to render 50% of the test animals dead. The lower a drug's LD50 value, the more lethal it is. The substances are classified according to their toxicity, with class 1 being the most dangerous, classes 2 and 3 being moderately harmful, and classes 4 and 5 being the least toxic. None of the ten phytocompounds were found to be harmful. The ProTox-II software was used to analyse the LD50 (rat) of the 10 compounds, and the same server was also used to determine their toxicity class (Table 5). The pkCSM software showed that these substances are efficiently absorbed by the human digestive system. The oral route of drug administration is the most widely used and preferred method because it is cost-effective, non-invasive, and patient-compliant, making it a crucial factor in determining the fate of novel medications and EPIs. Using SwissADME and admetSAR 2.0 software, it is observed that the compounds ellagic acid and baicalein have high bioavailability scores. They have a bioavailability score of 0.55 in SwissADME and 67.14 and 67.1 in admetSAR 2.0 for ellagic acid and baicalein, respectively. None of the other compounds have a bioavailability score of more than 50, which is the threshold for good bioavailability according to admetSAR 2.0. Ellagic acid and baicalein both have excellent LD50 values of 2991 mg/kg and 3919 mg/kg, respectively, and are classified as having toxicity classes 4 and 5, according to the ProTox-II web server. This is much better than piperine, which has a bioavailability score of 42.86 using admetSAR 2.0 and 0.55 using SwissADME software. Additionally, it belongs to toxicity class 4 using the ProTox-II web server and has a very low LD50 of 330 mg/kg. The 3D chemical structures of ellagic acid and baicalein are depicted in Figs. 1 and 2 (Table 6).