Computational screen flags five natural compounds that bind a tuberculosis virulence enzyme

A computational drug-discovery study has identified five natural compounds that appear to bind strongly to a tuberculosis virulence protein called Rv1509, a DNA methyltransferase encoded by the Mycobacterium tuberculosis gene of the same name. The work, published in Applied Biological Chemistry, used a multi-stage virtual screening pipeline against a three-dimensional model of the protein, then filtered the top candidates through predicted absorption, distribution, metabolism and excretion properties, followed by 200-nanosecond molecular dynamics simulations and principal component analysis. The five lead compounds, carrying the identifiers ZINC00338392, ZINC01662782, ZINC04104877, ZINC85341079 and ZINC96316367, achieved docking scores of at least -8.0 kilocalories per mole, a threshold well beyond the -4.5 kilocalories per mole recorded by the natural cofactor S-adenosyl-L-methionine, which was docked as a positive reference control under identical parameters.

The target protein sits at an interesting intersection of epigenetic regulation and immune manipulation. Previous laboratory work summarized in the paper describes the Rv1509-encoded DNA methyltransferase as an S-adenosyl-L-methionine-dependent enzyme that dynamically modulates gene expression and affects M. tuberculosis virulence. The gene encoding it is specific to the M. tuberculosis complex and absent from other mycobacterial species, including the BCG vaccine strain. In vitro studies have shown that the protein activates the TLR-2 pathway, triggering the release of pro-inflammatory cytokines and the elicitation of Th1-type immunity. Mouse experiments found that immunization with the protein augmented CD4-positive and CD8-positive T cell and effector memory T cell responses, and increased IgG responses in mice and in tuberculosis patients, prompting the authors to note its potential as an immunodiagnostic biomarker.

Further evidence of the protein’s role in pathogenesis comes from overexpression experiments. When the Rv1509-encoded enzyme was overexpressed in the non-pathogenic bacterium Mycobacterium smegmatis, the modified bacteria survived better by suppressing phagolysosomal maturation and aiding phagolysosomal escape, the mechanism by which the pathogen avoids being destroyed inside host immune cells. Overexpression also induced disease-associated alterations in the pancreas of infected mice, including lymphocyte infiltration, enhanced regulatory T cells and the presence of multinucleated giant cells, findings that implicate the protein in granuloma formation. Because the protein plays critical functions in virulence and persistence through epigenetic regulation and immune modulation, and shows no homology with human proteins, the authors argue it is an attractive candidate for new therapeutic discovery. Blocking it, they hypothesize, could interfere with immune evasion and intracellular persistence.

The computational pipeline began with the protein structure itself. Because no experimentally derived three-dimensional structure was available, the team built homology models using three servers: Swiss-Model, Phyre2 and AlphaFold. Each best model was evaluated on the PSVS server, which generated Ramachandran plots dividing residues into most-favored, allowed and disallowed regions. The AlphaFold model showed the maximum allowed residues, 99 percent, and the fewest disallowed residues compared with the other models, so it was selected for molecular docking. The protein was then prepared with the Protein Preparation Wizard in the Schrödinger Suite, which stripped non-relevant water molecules and heteroatoms, added missing hydrogen atoms, assigned disulfide bonds and optimized the structure through energy minimization with the OPLS3 force field.

The screening library came from the ZINC database of natural compounds and contained approximately 2,000 compounds, which were converted into 6,700 potential ligand conformations using LigPrep, accounting for ring conformations, ionization states, stereoisomers and tautomeric states. The active site was predicted with the CastP server, and a cubical docking grid of 10 cubic angstroms was built around the active site residues. Screening proceeded in three hierarchical stages through Schrödinger’s Glide module: High-Throughput Virtual Screening rapidly filtered the library, Standard Precision docking re-evaluated the top 500 compounds with greater conformational sampling, and Extra Precision docking, the most computationally intensive stage, ranked the top 100 compounds to produce the final candidates. S-adenosyl-L-methionine and its byproduct S-adenosyl-L-homocysteine were docked through the same workflow as reference controls.

Compounds with docking scores greater than -8.0 kilocalories per mole were flagged as the most promising, and the five leads that survived this cut also showed favorable theoretical physicochemical properties when analyzed with the SwissADME web server. That analysis covered lipophilicity, water solubility, gastrointestinal absorption, blood-brain barrier penetration potential, cytochrome P450 interactions and compliance with Lipinski’s Rule of Five, along with a bioavailability radar summarizing six physicochemical properties. The authors report that the five selected candidates fell within favorable ranges for these predicted pharmacokinetic parameters. Protein-ligand interaction analysis with LigPlot+ and visualization in PyMOL revealed networks of hydrophobic interactions and hydrogen bonds, including steady hydrogen bond networks with residues such as Phe170 and Tyr172 within the CastP-predicted catalytic pocket, suggesting the phytochemicals tightly occupy the active pocket architecture near functionally relevant residues.

Molecular dynamics simulations then tested whether the docked complexes would hold together under more realistic, dynamic conditions. All-atom simulations were run for 200 nanoseconds in GROMACS 2021 with the CHARMM27 force field for the protein and SwissParam-generated topologies for the ligands. Systems were neutralized with sodium ions, solvated with TIP3P water molecules, energy-minimized, and equilibrated in two stages: a canonical ensemble simulation at 300 kelvin followed by an isothermal-isobaric ensemble simulation at 1 bar. The root-mean-square deviation of both the unbound protein and the drug-protein complexes was stable throughout the 200-nanosecond run, particularly from 150 to 200 nanoseconds, with minimal deviations of 0.1 to 0.15 nanometers from the docked starting conformations, and the systems settled around 50 to 60 nanoseconds at values between 0.25 and 0.45 nanometers. Notably, binding of the five compounds reduced the average RMSD of the protein relative to the ligand-free enzyme, which the authors interpret as a positive effect on overall structural stability.

Hydrogen bond analysis painted a picture of persistent anchoring. Across the simulation, the complexes maintained the following ranges of intermolecular hydrogen bonds: two to four for ZINC00338392, four to six for ZINC01662782, one to three for ZINC96316367, one to five for ZINC04104877, and two to four for ZINC85341079. The authors state that these interactions suggest the compounds are firmly anchored in the active site. Root-mean-square fluctuation analysis showed that the unbound protein had a prominent flexible peak in the 90 to 110 amino acid region, characteristic of a flexible loop or disordered region, and that this peak was significantly reduced upon drug binding. Average RMSF values for the complexes ranged from 0.24 to 0.30 nanometers, compared with 0.27 nanometers for the unbound protein, supporting the conclusion that binding reduces overall flexibility. Radius of gyration values changed only slightly, from 1.85 nanometers unbound to between 1.83 and 1.90 nanometers for the complexes, and solvent-accessible surface area shifted from 157 square nanometers unbound to between 155 and 168 square nanometers, indicating subtle structural modifications concentrated in the flexible loop regions rather than wholesale denaturation.

Principal component analysis added a final layer of evidence. Applied to the simulation trajectories, PCA showed that the conformational space occupied by the drug-bound complexes was more restricted than that of the unbound protein, a reduction in conformational flexibility that the authors read as drug binding stabilizing the protein and slowing its functional dynamics. Taken together, the docking scores, predicted ADME profiles, stable hydrogen bond networks and restricted conformational dynamics form a consistent computational case that these five natural compounds bind with high predicted affinity to the active site of the Rv1509-encoded DNA methyltransferase.

The authors are careful about what these results do and do not establish. Everything reported here is computational: the protein structure is a predicted model rather than an experimentally solved structure, the binding affinities are docking estimates, and the pharmacokinetic profiles are predictions from a web server rather than measurements in cells, animals or patients. No compound has yet been shown to inhibit the enzyme in a test tube, let alone to treat tuberculosis. The paper states explicitly that experimental testing and optimization are still needed to determine actual efficacy and safety in preclinical and clinical trials, and the authors call for systematic in vitro, ex vivo and in vivo studies to confirm inhibitory potential, along with exploration of possible synergistic effects with existing anti-TB drugs. The context for that caution is stark: the World Health Organization statistics cited in the paper record 10 million new tuberculosis cases and 1.25 million tuberculosis-related deaths in 2024, with India bearing 26 percent of the global case burden, and multidrug-resistant M. tuberculosis producing lengthy treatment regimens, high treatment failure rates and mortality. Against that backdrop, a virulence-specific enzyme with no human homolog remains an appealing target, and these five compounds now join the long queue of computational leads awaiting the laboratory experiments that will decide whether the promise survives contact with biological reality.

Subject of Research: In silico identification of natural compound inhibitors targeting the Mycobacterium tuberculosis Rv1509-encoded DNA methyltransferase

Article Title: In silico assessment of natural compounds against mycobacterium tuberculosis Rv1509 protein

Article References: Fayaz, H., Ali, W., Shrivastava, G., Hasnain, S. E., Ehtesham, N. Z., & Alam, A. (2026). In silico assessment of natural compounds against mycobacterium tuberculosis Rv1509 protein. Applied Biological Chemistry, 69(1), Article 53. https://doi.org/10.1186/s13765-026-01118-w

Image Credits: AI Generated

DOI: 10.1186/s13765-026-01118-w

Keywords: tuberculosis, Mycobacterium tuberculosis, Rv1509, DNA methyltransferase, virtual screening, molecular docking, molecular dynamics simulation, ADME prediction, natural compounds, drug discovery, principal component analysis, antimicrobial resistance