The motivation behind the work lies in a persistent gap between what is known about Bacopa monnieri and what can be done with it. The plant has a long history of use for enhancing memory and cognitive function, and systematic reviews and meta-analyses of randomized trials have reported potential advantages for selected memory and attention metrics, although heterogeneity of studies and variation in extracts, dosages, and durations limit broad conclusions. Bacoside A, the dammarane-type triterpenoid saponin mixture responsible for much of the plant’s pharmacological reputation, yields jujubogenin and pseudojujubogenin-related aglycones upon deglycosylation, whereas acid hydrolysis generates ebelin lactone and bacogenin A1. Compared with the parent glycosides, ebelin lactone carries fewer hydrogen-bonding groups and a smaller polar surface, properties that may improve membrane partitioning while elevating hydrophobicity and reducing aqueous solubility. A foundational computational and receptor-binding study had previously reported micromolar binding affinities of ebelin lactone for the M1 muscarinic and 5-HT2A receptors, but its structural compatibility with proteins involved in synaptic function, proteostasis, and cellular resilience had never been evaluated in a unified workflow.

The six targets were selected through a literature-guided review because they collectively represent complementary cognition-associated processes rather than a single canonical pathway. Proteome-wide studies have linked alpha-actinin-4, or ACTN4, complexin-1, or CPLX1, and neuritin to the preservation of cognitive function in elderly individuals. BDNF signaling plays an important role in dendritic-spine development and structural synaptic plasticity, while complexin-1 is part of the presynaptic release machinery, and changes in its levels have been observed in the hippocampal tissue of individuals with Alzheimer’s disease. HSP70-1, the stress-responsive molecular chaperone also known as heat shock protein 72, participates in ATP-dependent protein folding and proteostasis, pathways considered important for limiting protein misfolding associated with neurodegenerative diseases. HSPA9, or mortalin, is a mitochondrial chaperone involved in mitochondrial protein import and cellular stress responses, and differential expression of mortalin isoforms has been observed in hippocampal tissue from individuals with Alzheimer’s disease in a pattern influenced by APOE genotype. Human proteomic studies have associated NRN1 with cognitive resilience in Alzheimer’s disease, with emerging evidence suggesting contributions to neuronal plasticity and structural remodeling.

Methodologically, the study integrated ADMET-AI predictions with blind docking using GNINA, in which the search space was generated automatically and poses were ranked with the Vina scoring function. Because experimentally validated binding sites were not uniformly available across the targets, most of which were represented by AlphaFold models, blind docking was performed for all six proteins. Only ACTN4 was represented by an experimentally determined structure, the 237-residue K255E actin-binding domain fragment in PDB 2R0O resolved at 2.20 angstroms, while BDNF was represented by its precursor form and the remaining targets by AlphaFold-predicted models covering specific portions of their reference sequences. Docking was carried out with a flexibility setting of 8, yielding nine binding modes per ligand. The five most favorable complexes, ACTN4-EL, BDNF-EL, HSP70-1-EL, HSPA9-EL, and NRN1-EL, were then advanced to 100-ns molecular dynamics simulations in GROMACS 2024.3 using the CHARMM36 force field for protein parameterization and SwissParam for ligand topology. Complexin-1 was not advanced because it had the least favorable docking score and occupied a comparatively small, predicted binding pocket, a decision the authors note should not be interpreted as evidence against a potential ebelin lactone-CPLX1 interaction.

The docking scores ranged from minus 9.13 to minus 7.00 kcal per mole. HSPA9 led the field at minus 9.13 kcal per mole, with one conventional hydrogen bond to ALA225 and an interaction profile dominated by van der Waals and hydrophobic contacts involving ILE51, VAL55, ALA75, VAL190, LEU217, and LEU228. ACTN4 followed at minus 8.65, BDNF at minus 8.57, HSP70-1 at minus 8.19, NRN1 at minus 7.87, and CPLX1 at minus 7.00. Across the six complexes, van der Waals and hydrophobic or pi-associated contacts were far more common than conventional hydrogen bonds, which appeared only in the CPLX1 and HSPA9 poses. The authors attribute this contact pattern to the rigid triterpenoid framework of ebelin lactone, its high predicted lipophilicity, and its limited hydrogen-bonding capacity of one donor and three acceptors, together with the prevalence of nonpolar residues in the predicted binding pockets.

The molecular dynamics trajectories revealed a target-dependent picture rather than a uniform ranking. The ACTN4 fragment showed the lowest mean protein RMSD of any complex, 0.163 plus or minus 0.021 nanometers, indicating remarkably low structural deviation from its starting conformation, while the full precursor model of BDNF showed the highest mean RMSD at 2.125 plus or minus 0.388 nanometers. Comparison with the corresponding apo trajectories showed that the ebelin lactone-bound forms of HSP70-1, HSPA9, and NRN1 displayed lower mean RMSD than their unbound counterparts, whereas ACTN4 and BDNF showed lower RMSD values in the apo forms. Residue-level fluctuation analysis told a similar story, with ACTN4-EL, HSP70-1-EL, and NRN1-EL showing lower mean RMSF than their apo forms, while BDNF and HSPA9 fluctuated more when bound. Across all five targets, the bound forms showed higher mean radius of gyration and higher solvent-accessible surface area than the apo forms, consistent with lower average compactness and greater surface exposure during the ligand-bound trajectories. Because the protein inputs differed in size and structural coverage, the authors interpreted absolute values within their respective target comparisons rather than using them alone to rank stability.

Protein-ligand hydrogen bonds occurred intermittently across all five bound complexes, and here NRN1 stood out clearly. NRN1-EL recorded the highest mean hydrogen-bond count at 0.575 plus or minus 0.501, followed by BDNF-EL at 0.352, HSP70-1-EL at 0.276, HSPA9-EL at 0.141, and ACTN4-EL at 0.094. HSP70-1-EL formed multiple simultaneous hydrogen bonds during short intervals, although these were not maintained throughout the trajectory. Principal component analysis and free-energy landscape mapping added a conformational dimension to the comparison. NRN1-EL displayed a prominent low-energy region in its landscape, indicating preferential sampling of a dominant conformational state during the simulation, while BDNF-EL and HSPA9-EL also showed identifiable low-energy regions with broader distributions. These findings were considered together with the trajectory statistics and binding-energy estimates when interpreting the conformational behavior of each system.

The MM/GBSA binding-energy calculations, performed with gmx_MMPBSA over the final 1,001 frames of each trajectory using the Generalized Born OBC2 model, produced the study’s most consequential numbers. NRN1-EL showed the most negative mean total binding-energy estimate at minus 39.21 plus or minus 3.24 kcal per mole, followed closely by HSPA9-EL at minus 37.93 plus or minus 2.60, BDNF-EL at minus 33.83 plus or minus 3.49, HSP70-1-EL at minus 23.20 plus or minus 5.24, and ACTN4-EL at minus 8.03 plus or minus 2.15. Per-residue energy decomposition identified pronounced favorable contributions from LEU70, LEU109, and TYR113 in BDNF-EL; VAL19, TYR21, ILE33, and MET36 in HSP70-1-EL; MET70, VAL190, ALA225, and LEU228 in HSPA9-EL; and PHE38, LEU94, LEU107, and PHE108 in NRN1-EL. Because entropic contributions were excluded and the calculations derived from single trajectories, the authors treated these values as comparative computational estimates rather than absolute experimental binding free energies, and the close NRN1 and HSPA9 estimates were not used to establish a rigorous quantitative ranking between the two complexes.

The ADMET predictions added a sobering developability dimension to the enthusiasm. Ebelin lactone has a molecular weight of 454.695 daltons, a calculated logP of 7.018, a topological polar surface area of 46.53 square angstroms, and a predicted solubility output of minus 6.79951, indicating high predicted lipophilicity and limited predicted aqueous solubility, properties that may create challenges for formulation and bioavailability. The blood-brain barrier Martins model produced an output of 0.206425, and because the ADMET-AI output did not provide endpoint-specific decision thresholds or confidence information, all values were retained as model-generated predictions rather than definitive claims regarding intestinal absorption, oral bioavailability, or central nervous system exposure. The authors also acknowledge several limitations that define the scope of the findings: the protein inputs differed in structural source and sequence coverage, docking poses were generated without experimental binding-site validation or redocking controls, each system was represented by a single 100-ns trajectory without independent replicates, and the dynamic behavior of CPLX1 remains unknown. Considering the docking, dynamics, free-energy landscape, and MM/GBSA results together, NRN1 was prioritized as the overall most promising computational candidate for experimental follow-up, with HSPA9 close behind. The authors conclude that experimental binding studies and functional cellular assays are required to validate the predicted ebelin lactone-NRN1 and ebelin lactone-HSPA9 interactions and determine their biological relevance, and that the predictions should be interpreted as hypothesis-generating until experimentally confirmed.

Subject of Research: Computational profiling of the Bacopa-derived triterpenoid ebelin lactone against cognition-associated proteins

Article Title: Comparative multi-target computational profiling of ebelin lactone across cognition-associated proteins

Article References: Lokanadham, M., & Chidambaram, R. (2026). Comparative multi-target computational profiling of ebelin lactone across cognition-associated proteins. Frontiers in Bioinformatics, 6, Article 1952759. https://doi.org/10.3389/fbinf.2026.1952759

Image Credits: AI Generated

DOI: 10.3389/fbinf.2026.1952759

Keywords: ebelin lactone, Bacopa monnieri, neuritin, mortalin, molecular docking, molecular dynamics, MM/GBSA, cognition, Comparative, multi-target, computational, profiling