Screening of Himalayan herbs for potential antitubercular phytochemicals
Recent studies have highlighted the pharmacological potential of Himalayan herbs traditionally used in Hawan Samagri. The present study focuses specifically on Himalayan plant species traditionally incorporated into Hawan Samagri preparations. This regional emphasis was adopted because Himalayan medicinal plants are known to produce unique and chemically diverse secondary metabolites as adaptive responses to ecological stressors such as high altitude, intense UV radiation, and temperature fluctuations, thereby offering a biologically enriched resource for drug discovery. From these selected herbs, forty-nine phytochemicals were systematically curated based on their documented occurrence, structural diversification, chemical diversity, pharmacological relevance, and potential volatility or persistence during fumigation (Supplementary Table 1 and Supplementary Figure 1). This curated set represents a rationally defined subset rather than the complete phytochemical repertoire of Hawan Samagri, ensuring both novelty and biological plausibility in the present screening approach.

Building on this, the present study identifies novel and previously unreported phytochemicals from these Himalayan herbs that may act as multitarget inhibitors of three key FAS-II enzymes in M.tb namely HadAB, FabG1, and KasA. These enzymes have been selected because they are essential, structurally conserved, and absent in humans, making them promising and underexplored targets for inhibiting mycolic acid biosynthesis in M.tb.⁷

Each selected compound had a molecular weight under 400 Da and was reported from various plant parts, including aerial parts, bark, roots, fruits, and seeds⁴² (Supplementary Table 1). The focus on relatively smaller molecules was intended to enhance drug-likeness and oral bioavailability, consistent with established medicinal chemistry principles. The chemical diversity included tropane alkaloids, phenylpropanoids, lignans, and anthraquinones.^43,44 This diverse range of chemical scaffolds enhances the probability of targeting different binding environments within the enzyme active sites, thereby increasing the potential for multitarget inhibition.

Lipinski’s rule of five and ADMET profiling
The screening of selected phytochemicals began with evaluating the drug-likeness and pharmacokinetic properties to ensure their suitability for oral administration and clinical progression. Lipinski’s Rule of Five is a broadly accepted with criterion that helps predict oral bioavailability by assessing key physicochemical properties. Compounds that follow these rules are more likely to be absorbed and utilized effectively when administered orally.⁴⁵ Physicochemical profiling of the forty-nine selected phytochemicals was performed to evaluate their drug-likeness based on Lipinski’s rule of five parameters. All compounds exhibited molecular weights below 500 Da (range: 104-387 Da), indicating favorable molecular size for oral absorption and cellular permeability. Lipophilicity assessment (XlogP3) demonstrated that the majority of compounds (46 out of 49) fell within the optimal range (-0.4 to 4.2; predominantly 1-4), reflecting a balanced hydrophilic–lipophilic profile conducive to membrane permeability. Only one compound showed slightly negative lipophilicity (Carpasemine, -0.4), suggesting higher hydrophilicity, while none exceeded the critical threshold of XlogP3 > 5, indicating minimal risk of excessive hydrophobicity-related solubility or toxicity issues.

Hydrogen-bonding capacity remained within acceptable limits for all compounds, with hydrogen bond donors ≤5 and hydrogen bond acceptors ≤10, supporting favorable membrane diffusion without compromising target-binding potential. Molar refractivity values ranged from 36.7-105.6, with the vast majority falling within the optimal range of 40-130, indicating appropriate molecular volume and polarizability for stable protein-ligand interactions.

Collectively, these findings demonstrate that the curated phytochemical library largely conforms to established drug-likeness criteria, supporting its suitability for structure-based virtual screening against FAS-II targets (Table 1 and Supplementary Table 2).

Table 1. Lipinski rule of five and ADMET properties of selected compound

ADMET profiling is an essential part of initial drug screening that evaluates the pharmacokinetic and toxicological parameters of compounds in biological systems. This approach helps to screen drugs having high therapeutic potential and low toxicity risk, reducing costly failures in further experimental stages.⁴⁶ In silico ADMET analysis revealed that the majority of the selected phytochemicals exhibited high predicted human intestinal absorption (HIA > 90%), indicating favorable oral bioavailability (Table 1 and Supplementary Table 3). Caco-2 permeability values suggested moderate to good intestinal transport for most compounds, while MDCK permeability varied, reflecting structural diversity and differences in passive membrane diffusion. Skin permeability predictions were generally low to moderate, supporting suitability for systemic administration. Plasma protein binding (PPB) values ranged from moderate to high, with several compounds maintaining acceptable binding levels (<90%), ensuring sufficient free drug availability. Predicted BBB penetration values were mostly low to moderate, suggesting limited central nervous system exposure an advantageous feature for anti-tubercular agents. Most compounds were predicted to be non-inhibitors of CYP2D6, indicating a low risk of drug-drug interactions. Although some compounds showed predicted mutagenicity, carcinogenicity outcomes were variable and not consistently positive. Overall, the ADMET profile supports the pharmacokinetic suitability of a substantial proportion of the phytochemical library for further investigation.⁴⁷

Molecular docking and molecular dynamics simulation analysis
Molecular docking was performed to predict the binding affinities of forty-nine phytochemicals with the target enzymes. A blind docking approach was employed to explore potential binding pockets without introducing positional bias. This strategy was particularly important to ensure unbiased identification of both canonical active site interactions and possible alternative or allosteric binding regions, thereby minimizing the risk of overlooking novel interaction sites. Such an approach is especially relevant when screening structurally diverse natural compounds, where binding preferences may not be restricted to previously characterized pockets. Supplementary Table 4 shows a summary of the docking data, which includes binding energies and inhibition constants for all compounds. To facilitate comparison of overall binding preferences, the docking scores were shown as heatmaps and bar graphs (Figures 1A and B).⁷ The heatmap provides an intuitive overview of relative binding affinities, where strong binders (ΔG < -8.5 kcal/mol) are highlighted in blue, moderate binders in red, and weak binders (ΔG > -6.0 kcal/mol) in yellow (Figure 1A) allowing rapid identification of compounds with high affinity across multiple targets. The bar graph (Figure 1B) displays the average binding energies alongside their corresponding inhibition constants (Ki values) enabling a direct comparison of both binding strength and predicted potency for each compound.

Seven compounds were selected based on their binding affinities and inhibition constants: 16α-hydroxyprogesterone, guggulsterone Z, cinnamonol, norhyoscine, episesamin, apohyoscine and chrysophanol. Among them, both guggulsterone Z and 16α-Hydroxyprogesterone showed high docking affinities for all three targets in our screening (Supplementary Table 4). However, PreADMET toxicity predictions indicated potential mutagenicity and/or carcinogenicity, and their steroidal structures could interact with hormone-related pathways, limiting their suitability as lead candidates. Therefore, these compounds were excluded from further selection in the lead optimization pipeline.

The remaining five compounds were advanced for detailed analysis, as they demonstrated strong binding interactions across all three targets (Table 2). Each compound, however, showed its highest binding affinity toward a specific enzyme: Cinnamonol with HadAB (-9.25 kcal/mol), Episesamin with FabG1 (-7.70 kcal/mol), and both Norhyoscine (-9.95 kcal/mol) and Apohyoscine (-9.70 kcal/mol) with KasA. Chrysophanol, though relatively moderate, demonstrated extensive activity against all three proteins.⁴⁸

Table 2. Binding energy, inhibition constant, and interacting amino acid of selected five lead compounds

Cinnamonol was identified as the most promising candidate among these leads due to its strong binding affinity toward HadAB (-9.25 kcal/mol), FabG1 (-7.95 kcal/mol) and KasA (-7.59 kcal/mol). In HadAB, Cinnamonol established an extensive network of interactions (Figures 2A and D), where it was stabilized by several hydrogen bonds and hydrophobic interactions. Gln68, Glu81, and Thr79 formed strong hydrogen bonds with the ligand, while van der Waals and alkyl/π-alkyl interactions strengthened binding with hydrophobic residues Leu142, Val85, Ile84, Ala69, and Phe72. Residues Cys61 and Gly64 were positioned close to the ligand, helping define the catalytic cleft. In FabG1, Cinnamonol was located near the NADH-binding site, forming interactions with Arg25, Ser140, and Asn88, while Pro183 formed π-π interactions. Additional contacts with Ala89, Gly90, and Val141 further stabilized the complex (Figures 2B and E). In KasA, the ligand was anchored within the malonyl-ACP groove and stabilized by hydrogen bonds with His311 and Glu322 (Figures 2C and F).

MD simulations over 100 ns validated stable complexation among all three enzymes, exhibiting RMSD deviations below 2.8 Å (Figures 2G-I) and RMSF fluctuations below 3.5 Åֵ, predominantly localized in flexible loops regions (Figures 2J-L). Cinnamonol is predicted to block substrate entry into the catalytic site by occupying the hydrophobic groove of HadAB and forming stabilizing interactions, thereby inhibiting the dehydration step required for mycolic acid chain elongation.^49,50

Episesamin, a phenylpropanoid lignan, exhibited stable binding to all three FAS-II enzymes (Figure 3). In the case of HadAB, it formed hydrogen bonds with Gln68 and Gln86, while Ile60, Gly64 and Cys61 contributed additional van der Waals contacts. Hydrophobic residues such as Leu142, Ile84, and Val85 helped keep the ligand stable in the hydrophobic cleft (Figures 3A and D). In FabG1, Episesamin firmly was positioned within the substrate-binding region to the area where the substrate binds, which is next to the NADH cofactor pocket (Figures 3B and E). It formed hydrogen bonds with Asn88 and Ser140 which are important for maintaining the catalytic environment. It also formed van der Waals and α-alkyl interactions with Tyr153, Pro183, Val141, and Gly139. Residues like Gly22, Arg25, and Ile27 helped stabilize the structure even more by hydrophobic and shape-complementary packing interactions (Figures 3B and E). In KasA, with hydrogen bond interactions with Arg214, and Gly318, along with hydrophobic interactions with Ala215, Ile317, Pro316, and Gly403 contributed to the stability of binding (Figures 3C and F). In conclusion, Episesamin had the strongest binding affinity with FabG1 (-7.70 kcal/mol), while HadAB (-8.74 kcal/mol) and KasA (-8.32 kcal/mol) had weaker binding affinities. The docking protocol was further validated by docking experimentally known inhibitors for each target protein under the same computational conditions. The results are presented in Table 3. The identified phytochemicals showed docking scores and binding interactions comparable to the reference inhibitors. These results support the reliability of the docking methodology and further suggest the potential of the chosen phytochemicals as potential inhibitors of HadAB, FabG1 and KasA.

Table 3. Docking Results of Experimentally Validated Inhibitors Against Target Proteins Used for Method Validation

Molecular dynamics simulations demonstrated that the FabG1-Episesamin complex exhibited the most stable. RMSD values stabilized after 30 ns (Figure 3G-I), and RMSF fluctuations remained below 5 Å for the majority of residues, except for minor loop motions in HadAB (Figures 3J-L). These findings collectively identify FabG1 as the principal target of Episesamin, with its direct interactions with catalytically crucial residues (Asn88 and Ser140) and neighbouring hydrophobic contacts potentially hindering β-ketoacyl intermediate positioning or obstructing NADH binding, thereby disrupting fatty acid chain elongation and subsequent mycolic acid biosynthesis.^51,52

Norhyoscine (scopolamine) and Apohyoscine (aposcopolamine), both tropane alkaloids distinguished by their bicyclic tropane core and ester linkages, demonstrated significant affinities for all three FAS-II enzymes, with KasA identified as their principal target (-9.95 and -9.70 kcal/mol, respectively). In HadAB, Norhyoscine formed hydrogen bonds with Tyr65, Gln68 and Thr138 and it also formed hydrophobic interactions with Gly64, Leu91, Ile60 and Leu91. Additional van der Waals contacts with Cys61 and Thr140 helped keep the binding orientation stable in a shallow pocket next to the catalytic site (Figures 4A and D). In FabG1, Norhyoscine was bound near the cofactor-binding domain where it was stabilized by forming hydrogen bonds with Asn24, Arg47, and Arg25 and hydrophobic contacts with Val60, Gly22, Gly28, Gly26, and Val60 residues (Figures 4B and E). Additional van der Waals contacts with Ala89 and Val62 helped stabilize the binding orientation in a shallow pocket next to the catalytic site. In comparison with HadAB and FabG1, Norhyoscine had the strongest binding with KasA (-9.95 kcal/mol), where it was deeply embedded in the tunnel-like substrate channel (Figures  4C and F). A strong hydrogen bond with Gly318 and hydrophobic contacts with Ala279, Val278, Gly280, and Pro373 enhanced stability. Phe402 formed π-π stacking with the ligand, which is also observed for thiolactomycin. Additional stabilization came from Ala215, Met213, Thr313, and His311, confirming its occupancy of the catalytic groove and potential to block substrate access. MD simulations demonstrated stable complex formation, with RMSD values plateauing below 3 Å and RMSF fluctuations between 4-6 Å (Figures 4J-L), confirming the conformational stability of the Norhyoscine-KasA complex.

Similarly, apohyoscine showed a similar interaction profile. HadAB was stabilized by hydrogen bonding with Gln68 and formed van der Waals and π-π stacking interactions with Thr140, Gln89, Cys61, Leu142 and Ile84 (Figures 5A and D). Apohyoscine formed hydrogen bonds with Ile27 and Gly90 in FabG1 and hydrophobic interactions with Gly26, Gly25, Val62, and Gly22, whereas additional interactions with Ala111, Leu91, Arg47, Arg25 and Ala89 improved stability at the cofactor-binding region (Figures 5B and E).

However, its highest binding affinity was observed with KasA (-9.70 kcal/mol), where the ligand occupied the same substrate tunnel as Norhyoscine (Figures 5C and F). Strong π-π stacking with Phe402 and Phe404, and hydrophobic interactions involving Gly403, Gly318, Gly405, and Gly406 formed an extensive interaction network. The broader contact surface compared to Norhyoscine suggests improved stability and tighter binding. MD simulations corroborated this, demonstrating RMSD stabilization after 25 ns (Figure 5G-I) and uniform RMSF values across catalytic residues (Figure 5J-L).

In general, both alkaloids had stable but different affinities for all three enzymes. They bound moderately to HadAB and FabG1 and strongly to KasA. Their occupation of the substrate channel and π-π stacking with Phe402, a conserved residue in known inhibitors, supports their role as competitive blockers of the condensation step in mycolic acid chain elongation, thereby disrupting the FAS-II pathway.

Chrysophanol, an anthraquinone derivative previously documented for its antimycobacterial efficacy (MICs of 64 µg/mL against M.tb H37Ra and 64-128 µg/mL against M. bovis), demonstrated extensive yet moderate binding across HadAB, FabG1, and KasA (Figures 6A-C). Unlike the other ligands that exhibited target preference, Chrysophanol maintained consistent interaction patterns with all three enzymes.

In HadAB, Chrysophanol formed hydrogen bonds with Gln68 and Gln86, which were further strengthened by hydrophobic interactions with Cys61, Ile60, and Gln89 (Figures 6A and D). In FabG1, the ligand formed hydrogen bonds with Asn24 and Val60. It also formed van der Waals and π-π stacking interactions with Val62, Ala111, Ala89, and Arg25, which could disrupt cofactor stabilization near the NADH-binding pocket (Figure 6B and E). In KasA, Chrysophanol was found in the substrate tunnel, where it formed hydrogen-bond interactions with Asp273 and Val278 and strong π-π stacking interactions with Phe402, Phe404, His311. It was further stabilized by hydrophobic contacts with Gly403, Gly406, Cys171 and Ala279 (Figures 6C and F).

MD simulations confirmed stable binding in all three protein-ligand complexes, with RMSD convergence occurring between 5-35 (Figure 6G-I) ns and RMSF values typically below 6 Å, suggesting restricted flexibility in critical active-site regions (Figures 6J-L). Comprehensive analyses of temperature, velocity, potential energy, density gradient, box volume, and radius of gyration for all five selected compounds in relation to HadAB, FabG1, and KasA are illustrated in Supplementary Figures 2, 3, and 4, respectively. Chrysophanol’s ability to consistently block multiple targets shows that it could be a broad-spectrum inhibitor that can block substrate access and interfere with multiple steps.

All five selected phytocompounds Cinnamonol, Norhyoscine, Episesamin, Apohyoscine, and Chrysophanol, were found to bind directly within or next to the substrate-binding pockets of HadAB, FabG1, and KasA. These pockets are normally occupied by other intermediates or cofactors, like β-hydroxyacyl-ACP, NADH, and malonyl-ACP. It is predicted that these phytochemicals compete for the binding pockets and block substrate entry, thereby either slowing down or completely inhibiting with the FAS-II pathway. This inhibition process compromises mycolic acid synthesis, an essential component of the mycobacterial cell wall, ultimately weakening the cell wall structure and leading to the bacterial cell death. The ability of these compounds to bind to multiple targets makes them more likely to be tested further as broad-spectrum FAS-II inhibitors.

Binding site analysis and inhibition mechanism
Comparison with reference inhibitors reinforced the predicted mechanisms. Cinnamonol and Norhyoscine were found to bind near the KasA malonyl-ACP tunnel. In HadAB, Cinnamonol occupied a hydrophobic cleft similar to the binding region of thiolactomycin. In FabG1, Cinnamonol and Episesamin occupied residues surrounding the NADH cofactor pocket, resembling NADH-mimetic inhibitors reported in similar systems. These findings suggested that FAS-II enzymes are likely inhibited through a competitive mechanism, targeting multiple steps of mycolic acid biosynthesis and ultimately weakening the structural integrity of the M.tb cell wall.¹⁷