Plant secondary metabolites as potential myostatin inhibitors: Computational insights for skeletal muscle atrophy

1. Introduction

Skeletal muscle (SM) makes up approximately 40% of total body mass and is essential for locomotion, maintaining posture, regulating body temperature, and supporting metabolic homeostasis. Muscle health relies heavily on balancing the synthesis and breakdown of muscle protein (Tipton et al., 2018; Mukund and Subramaniam, 2020). Disruption of this balance leads to SM atrophy, characterized by reduced muscle mass, strength, and function. Muscle atrophy is frequently associated with various pathological conditions, including aging, cancer cachexia, muscular dystrophies, chronic diseases, and metabolic disorders (Wilburn et al., 2021).

Myostatin (MSTN) is a well-researched negative regulator of muscle growth and differentiation. MSTN impedes myogenesis by interacting with its receptor, the activin type IIB receptor (ActRIIB) (Rodgers and Ward 2022). It initiates downstream signaling cascades that involve the phosphorylation of SMAD2/3, subsequently inhibiting anabolic signaling pathways such as the Akt/mTOR pathway (Sartori et al., 2021, Baig et al., 2022). MSTN expression is elevated in various muscle-wasting diseases and conditions, indicating its potential as a therapeutic target for SM atrophy. Inhibition of MSTN presents a promising approach for managing muscle-wasting diseases, given the well-documented association between MSTN and such disorders (Ahmad et al., 2024). Fig. 1 shows the molecular mechanisms by which MSTN controls muscle mass and how natural compounds might oppose MSTN-driven muscle atrophy to promote muscle growth.

Close

Fig. 1.

Schematic representation of the molecular mechanisms by which MSTN regulates SM mass. Active MSTN signals through the ActRIIB–ALK4/5 receptor complex to activate SMAD2 and suppress MyoD and IGF-1 expression, leading to muscle atrophy. Inhibition of MSTN signaling by natural compounds may restore Akt/mTOR activity, enhance protein synthesis, and promote muscle growth.

Export to PPT

Several strategies have been developed to inhibit MSTN activity, including monoclonal antibodies, peptides, and small-molecule inhibitors (Abati et al., 2022; Mitra et al., 2023; Wetzlich et al., 2025). However, many of these therapeutic agents have encountered limitations, including high costs, adverse side effects, or insufficient efficacy in clinical trials. Natural products have historically served as rich sources for drug discovery, offering structural diversity and lower toxicity profiles. Recent studies targeting MSTN using natural compound screening have yielded several papers providing in-depth insights into this approach (Ahmad et al., 2024, Ahmad et al., 2024, Wetzlich et al., 2025). Numerous natural compounds, such as epicatechin, curcumin, gingerol, and dithymoquinone, have demonstrated potential for inhibiting MSTN. Our previous research identified dithymoquinone from Nigella sativa, and compounds from Glycyrrhiza uralensis (licochalcone A and B) and Traditional Chinese Medicines, as promising MSTN inhibitors using computational and experimental approaches (Ahmad et al., 2021; Ali et al., 2022; Baig et al., 2022; Ahmad et al., 2024).

Given these factors, there is a strong interest in discovering novel, effective, and safe MSTN inhibitors with few or no side effects (Wetzlich et al., 2025). Plant secondary metabolites (PSMs) possess a broad spectrum of therapeutic properties, attributed to their structural diversity and advantageous safety profile (Mushtaq et al., 2018). PSMs generally exhibit fewer adverse effects, diverse mechanisms of action, and advantageous pharmacokinetic properties compared to many synthetic drugs, making them attractive candidates for pharmaceutical development. These compounds belong to various chemical classes, such as flavonoids, coumarins, terpenoids, polyphenolic acids, alkaloids, glycosides, and organosulfur compounds. They exhibit diverse pharmacological activities, including anticancer, antioxidant, antimicrobial, and immunomodulatory properties, underscoring their potential as therapeutic agents (Rao et al., 2023).

In this context, computational drug discovery offers a rapid, cost-effective, and efficient strategy for identifying novel therapeutic candidates. Although MSTN is biologically active as a covalently bound homodimer, most computational studies, including our previous work and those of others (Ahmad et al., 2021; Ahmad et al., 2024), have employed the monomeric unit as the docking target. Integrating virtual screening, ADMET (absorption, distribution, metabolism, excretion, toxicity) profiling, and molecular dynamics simulations (MDS) enables the systematic discovery of secondary metabolite-based MSTN inhibitors. Accordingly, the present study is designed to identify potent PSMs targeting MSTN using computational methodologies, including virtual screening, MDS, energy calculations, and ADMET analysis, to develop promising therapeutic avenues for managing SM atrophy in various pathological conditions.

2. Materials and Methods

3. Results and Discussion

This study involved a systematic virtual screening of a wide range of PSMs, building on our previous research where we successfully employed computational methods to identify natural inhibitors of MSTN and its signaling regulators (Ahmad et al., 2024). Our goal was to identify novel, effective inhibitory compounds with potential therapeutic applications for SM atrophy by targeting the MSTN protein.

The methodology included structure-based virtual screening, ADMET profiling, MDS, and M/PBSA, yielding a comprehensive in silico evaluation of potential inhibitors. The choice of curcumin was based on its established role as an MSTN modulator in our earlier computational studies (Baig et al., 2017; Ahmad et al., 2024), thereby serving as a reliable benchmark to validate the docking and simulation pipeline used in this work.

Several PSMs exhibited higher binding affinities for MSTN compared to the control, curcumin (−6.8 kcal·mol⁻¹). The top three hits, namely PSCdb02834 (−7.5 kcal·mol⁻¹), PSCdb02056 (−7.3 kcal·mol⁻¹), and PSCdb02107 (−7.3 kcal·mol⁻¹), had the strongest predicted interactions. All top ten candidates outperformed curcumin, with binding energies ranging from −6.9 to −7.5 kcal·mol⁻¹ (Table 1).

Following visual inspection of the superimposed binding poses and interaction patterns of the top ten PSMs, PSCdb02834, PSCdb02107, and PSCdb02056 emerged as the most promising candidates, exhibiting stable orientations and deep accommodation within the MSTN binding pocket. These PSMs consistently occupied the same region as curcumin (used as a control), but with enhanced interaction networks and tighter spatial fit (Fig. 2).

Close

Fig. 2.

Superimposed binding poses of the top three screened compounds, PSCdb02834 (yellow), PSCdb02107 (blue), PSCdb02056 (green), and curcumin (control, black), within the MSTN binding pocket. The left panel displays all compounds docked onto the MSTN protein surface. The right panel provides a close-up view of the active site, highlighting key interacting residues and the spatial orientation of each ligand.

Export to PPT

The 2D interaction diagrams further illustrate that these three PSMs establish multiple key contacts with crucial MSTN residues, including aromatic and polar side chains such as PHE87, TYR86, TRP31, ASN88, GLY89, GLU91, AND ILE98. Most of these residues have previously been associated with ligand recognition and binding stabilization in MSTN (Ahmad et al., 2024). Notably, these PSMs exhibit stronger hydrophobic and π-π stacking interactions, along with hydrogen bonds, than curcumin, indicating a stronger and more favorable binding profile. The interaction distances are mostly under 5 Å for key contacts, indicating a strong fit between the ligands and the binding pocket. This is supported by the higher binding energies observed for PSCdb02834, PSCdb02107, and PSCdb02056, which exceed those of curcumin (Fig. 3).

Close

Fig. 3.

2D interaction diagrams of the top three screened compounds, PSCdb02056 (green), PSCdb02107 (blue), PSCdb02834 (yellow), and the control, curcumin (black), within the MSTN active site. Key interacting residues and atomic distances are shown for each ligand.

Export to PPT

MDS analyses revealed the stability and conformational dynamics of MSTN complexes with the top three screened PSMs (PSCdb02056, PSCdb02107, PSCdb02834) and the control. The RMSD plots (Fig. 4a) show that all ligand–protein complexes reached equilibrium, with PSCdb02107 (blue) and PSCdb02834 (yellow) exhibiting lower fluctuation profiles than the control, indicating improved complex stability. The Rg values (Fig. 4b) remained consistent throughout the simulation, further confirming the structural stability of the MSTN–ligand complexes. RMSF analysis (Fig. 4c) demonstrated minimal residue-level fluctuations, especially for PSCdb02107 and PSCdb02834, suggesting increased rigidity in the protein backbone after ligand binding. The hydrogen bond analysis (Fig. 4d) revealed persistent hydrogen bonds across all top PSMs, with PSCdb02834 (yellow) showing the highest number of stable H-bonds, reflecting strong and consistent interactions within the MSTN binding site (Fig. 4).

Close

Fig. 4.

MDS analyses of MSTN complexes with the top three screened compounds, PSCdb02056 (green), PSCdb02107 (blue), PSCdb02834 (yellow), and the control (black), over 200 ns. (a) RMSD values of protein backbone Cα atoms relative to the minimized starting structure. (b) Radius of gyration (Rg) profiles showing global compactness. (c) RMSF plots highlighting residue-level flexibility (red line indicates apo-MSTN). (d) Hydrogen-bond analysis indicating the number and persistence of H-bonds formed with key binding site residues.

Export to PPT

To provide a more quantitative assessment, the average RMSD values of the backbone Cα atoms were 0.38 ± 0.05 nm for the control, 0.34 ± 0.04 nm for PSCdb02056, 0.36 ± 0.06 nm for PSCdb02107, and 0.33 ± 0.05 nm for PSCdb02834, confirming the lower fluctuation trends observed. The corresponding average Rg values were approximately 1.92 nm (control), 1.86 nm (PSCdb02056), 1.84 nm (PSCdb02107), and 1.85 nm (PSCdb02834), indicating ligand binding helped maintain the compactness of MSTN. RMSF analysis further revealed that key binding-site residues were less flexible in the PSCdb02107 and PSCdb02834 complexes than in the control, indicating localized stabilization of the binding pocket. In the hydrogen-bond analysis, PSCdb02834 maintained approximately 2-3 persistent hydrogen bonds, primarily with GLU91, ASN88, and GLY89, whereas PSCdb02107 exhibited around two consistent bonds with TYR86 and TRP31, indicating stable and specific protein-ligand interactions.

Following the RMSD, Rg, RMSF, and H-bond analyses (Fig. 4), Fig. 5 shows consistent global compactness (SASA) and a clear first hydration shell around the ligands (RDF). Persistent water-mediated contacts (water-bridge occupancy) further support a stable solvent-assisted recognition mode prior to MM/PBSA analysis.

Close

Fig. 5.

Solvent–solvation behavior of MSTN complexes from MD trajectories. (a) SASA of the MSTN protein over 0–200 ns shows consistent global compactness across apo/control (black), MSTN (red), PSCdb02056 (green), PSCdb02107 (blue), and PSCdb02834 (yellow). (b) Water–ligand RDF g(r) between ligand heavy atoms and water oxygen (OW), calculated over 100–200 ns, reveals a structured first hydration shell. (c) Water-bridge occupancy (the fraction of time a protein–water–ligand H-bond triad is present; distance < 0.35 nm, angle > 150°).

Export to PPT

All systems maintained compact conformations without upward drift (Fig. 5a). SASA fluctuated within approximately 70–86 nm² across trajectories, indicating no global unfolding; ligand-bound complexes showed variability similar to the apo/control system, consistent with local breathing of the binding region rather than large-scale rearrangement.

RDF curves (Fig. 5b) show a clear first hydration shell around the bound ligands; g(r) increases sharply between approximately 0.3–0.6 nm and nears unity beyond about 2.0–2.5 nm. The three lead complexes have slightly higher first-shell density than the control, indicating structured hydration at the pocket entrance during stable binding.

Time-resolved occupancies (Fig. 5c) show recurrent protein–water–ligand triads. Events frequently involve the pocket residues GLU91, TYR86, PHE87, and TRP31, supporting solvent-mediated stabilization that complements the direct H-bond and hydrophobic contacts observed in docking/MD contact maps.

Then, principal component analysis (PCA) was performed to explore the essential dynamics and conformational space sampled by the MSTN complexes during MDS. The 2D projection of the trajectories onto the first two PCs (Panel A) showed that PSCdb02834 (yellow) had the broadest distribution, indicating greater conformational flexibility in the MSTN-binding pocket. In contrast, PSCdb02107 (blue) and PSCdb02056 (green) formed more compact clusters, suggesting restricted motion and stable binding conformations. The eigenvalue spectrum of the covariance matrix (Panel B) indicated that the first few eigenvectors accounted for most of the atomic motion across all complexes, with PSCdb02834 again showing slightly higher values, consistent with its dynamic behavior. These results emphasize the distinct dynamic profiles of the MSTN–ligand complexes, with PSCdb02107 and PSCdb02056 demonstrating greater stability compared to the control and PSCdb02834 (Fig. 6).

Close

Fig. 6.

PCA of MSTN complexes with the top three screened compounds, PSCdb02056 (green), PSCdb02107 (blue), PSCdb02834 (yellow), and the control, curcumin (black). (a) 2D projection of the trajectories onto the first two principal components (eigenvectors 1 and 2). (b) Eigenvalue distribution of the covariance matrix representing atomic motion.

Export to PPT

Free energy landscape (FEL) analyses were conducted to visualize the conformational stability and transitions of the MSTN complexes during MDS. The FEL plots display distinct energy basins corresponding to stable conformational states for each MSTN–ligand complex. Complexes with PSCdb02056, PSCdb02107, and PSCdb02834 displayed deeper and broader energy minima compared to the control, indicating increased structural stability and a lower tendency to adopt high-energy, less favorable conformations. The presence of well-defined, low-energy basins for these top candidates demonstrates their ability to maintain stable binding within the MSTN pocket throughout the simulation, which aligns with their superior binding energies and dynamic profiles observed in previous analyses (Fig. 7).

Close

Fig. 7.

FEL plots of MSTN complexes projected along the first two principal components (PC1 and PC2) from MD simulations: (a) Control (curcumin), (b) PSCdb02056, (c) PSCdb02107, and (d) PSCdb02834. Color bars show free energy (kJ mol^-1), with deeper blue areas indicating more stable conformations.

Export to PPT

To assess the binding affinities of the selected PSMs, MM/PBSA calculations were performed to determine the binding free energies of the selected PSMs and the control with MSTN across the MDS trajectories. A total of 20,000 frames were extracted from the 200 ns MDS for each protein–ligand complex, enabling high-resolution analysis of binding free energy fluctuations. The energy profiles, as shown in Fig. 7, remained stable for most of the trajectory, indicating strong protein–ligand interactions and system convergence.

The energetic component plots reveal that all complexes maintain consistently negative total binding energies throughout the simulation, indicating favorable binding interactions. Notably, PSCdb02107 and PSCdb02834 showed more stable, lower average binding energies and fewer fluctuations than the control and PSCdb02056, supporting their superior, stable binding affinities to MSTN. These findings align with the dynamic and structural analyses, further confirming the potential of these hits as effective MSTN inhibitors (Fig. 8).

Close

Fig. 8.

MM/PBSA binding free energy component profiles for MSTN complexes throughout the MD simulation trajectory: (a) Control (curcumin), (b) PSCdb02056, (c) PSCdb02107, and (d) PSCdb02834. Total binding energies (black) and moving averages (red) are displayed as functions of simulation frames. Lower, more stable energy values suggest stronger and more stable ligand binding.

Export to PPT

The MM/PBSA analysis showed that all three screened PSMs had favorable binding energetics to MSTN, with ΔTOTAL values of –13.1 kcal·mol⁻¹ (PSCdb02056), –14.2 kcal·mol⁻¹ (PSCdb02107), and –17.6 kcal·mol⁻¹ (PSCdb02834), compared to –15.5 kcal·mol⁻¹ for the control (curcumin). Notably, PSCdb02834 exhibited the lowest total binding free energy, indicating the strongest interaction with MSTN among the tested ligands. The van der Waals (ΔVDWAALS) and electrostatic (ΔEEL) contributions were consistently negative, supporting potent ligand–protein interactions. These results demonstrate that all three PSMs bind favorably to MSTN, with PSCdb02834 exhibiting the highest binding affinity, as confirmed by docking and MDS analyses (Table 2).

Curcumin showed the most favorable net interaction energy in vacuum (ΔGGAS = −36.4 kcal·mol⁻¹), primarily due to strong van der Waals and electrostatic contributions. However, it also experienced the largest polar solvation penalty (ΔEPB = +20.9 kcal·mol⁻¹), resulting in a less favorable ΔTOTAL than PSCdb02834 (−17.6 kcal·mol⁻¹) and PSCdb02107 (−14.2 kcal·mol⁻¹). This suggests that the top screened PSMs better balance interaction energy with desolvation costs in the MSTN pocket.

Finally, the ADMET analysis of the selected PSMs was performed using the pkCSM platform. These hits showed good predicted intestinal absorption (>83%) and moderate to high Caco-2 permeability, indicating favorable oral bioavailability. The PSMs also demonstrated acceptable water solubility and skin permeability. Notably, only PSCdb02834 was not predicted to be a substrate of P-glycoprotein, which may lower the risk of efflux-mediated drug resistance. Distribution profiles indicated a moderate volume of distribution and limited blood–brain barrier and CNS permeability for all candidates, thus reducing potential central nervous system side effects. Regarding metabolism, the PSMs showed limited liability as CYP450 substrates or inhibitors, with some predicted inhibitory effects on CYP1A2, CYP2C19, and CYP3A4 isoforms. Total clearance values were within the acceptable range, and none of the PSMs were identified as renal OCT2 substrates. Toxicity predictions indicated that all PSMs, except PSCdb02107 (which showed positive AMES results), were non-mutagenic and posed no significant risk of hERG inhibition (except for some hERG II liability in PSCdb02834 and PSCdb02107). Additionally, hepatotoxicity was only predicted for PSCdb02056. The ADMET profiles demonstrate the drug-likeness and safety of PSCdb02834, PSCdb02056, and PSCdb02107, emphasizing their potential for further development as MSTN inhibitors (Table 3).

4. Conclusions

This study employed a comprehensive computational pipeline that included virtual screening, MDS, MM/PBSA energy calculations, and ADMET profiling to identify and evaluate potential MSTN inhibitors from a diverse library of PSMs. Three PSMs, namely PSCdb02834, PSCdb02056, and PSCdb02107, demonstrated strong binding affinity, favorable interaction profiles, structural stability, and promising pharmacokinetic and toxicological properties. These results provide a solid foundation for further experimental studies of these candidates as novel MSTN inhibitors for treating SM atrophy. Ultimately, these findings highlight the importance of in silico methods and natural product libraries in the early stages of drug discovery for muscle-wasting conditions.