The initial structures for molecular modeling were obtained from the crystal structures of β-catenin (PDB ID: 2GL7, resolution 2.60 Å) and LRP6 (PDB ID: 3S2K, resolution 2.80 Å) (Ahn et al., 2011; Sampietro et al., 2006) (Table 1). Missing loop regions were reconstructed using the auto model and loop model functions of MODELER in Discovery Studio (DS) 2019 Client (Kemmish et al., 2017). Prior to molecular docking simulations, water molecules and ligands were removed, and hydrogen atoms were added to the protein structures. Further preparation of β-catenin and LRP6 was carried out using UCSF Chimera and AutoDockTools 4.2.6, with AD4 atom types and Gasteiger charges assigned to optimize docking accuracy (Forli et al., 2016; Pettersen et al., 2004).

β-Catenin possesses two primary binding sites: the canonical β-catenin union site (β-catenin US), which includes Asp16 from TCF4, Lys435, and His470, with the NZ atom of Lys435 defined as the grid center; and an allosteric site (β-catenin AS), characterized by Pro521, Arg528, and Asp583, with the ND1 atom of His524 designated as the center. For LRP6, the docking grid was centered at Gly227 from DKK1, consistent with previously reported binding interactions. Initial docking poses for subsequent molecular dynamics (MD) simulations were generated using AutoDockTools 4.2.6, with docking parameters optimized according to a previous study (Vélez-Vargas et al., 2023). The best binding poses were selected based on binding affinity and further validated by visual inspection using Discovery Studio (DS) 2019 Client, PLIP, and UCSF Chimera. For comparative analysis, hit derivatives were also evaluated using Schrödinger’s Glide software (version 2019–1, Schrödinger, LLC, New York, NY, United States, 2017) (Durrant and McCammon, 2011). These docking results provided the foundation for further MD simulations, enabling a detailed These docking results served as the basis for MD simulations, allowing detailed investigation of the stability and interactions of α-mangostin and its derivatives with β-catenin and LRP6.

All-atom molecular dynamics (MD) simulations of the top-ranked docking poses were performed using GROMACS 2016.3 with the PLUMED 2.4 plugin (Abraham et al., 2015). Protein and ligand parameters were generated using the CHARMM36 force field and CGenFF (Wurl and Ferreira, 2023), with systems solvated in a TIP3P water box (10 Å buffer) and neutralized with 0.15 M NaCl. Energy minimization was performed using the steepest descent method until a tolerance of 1000 kJ/mol was achieved, followed by NVT equilibration (25 ps, 303.15 K). NPT production runs for 500 ns at 303.15 K and 1 bar, controlled by the Nosé–Hoover thermostat and Parrinello–Rahman barostat (Shiga et al., 2023). Bond constraints involving hydrogen atoms were applied using the LINCS algorithm (Hess et al., 1997). The production runs employed a 2 fs timestep with trajectories saved every 1 ps; electrostatic interactions were treated with PME, and van der Waals interactions were truncated at 12 Å. An upper-wall restraint (200 kJ/mol·nm^-2) was applied when the ligand center of mass exceeded 12 Å from the binding site. Protein–ligand interactions were analyzed over the final 200 ns using GROMACS 2016.3, Discovery Studio 2019 (Kemmish et al., 2017), and VMD 1.9.4 (Humphrey et al., 1996).

MDA-MB-231 and MCF-7 cell lines were obtained from the European Collection of Authenticated Cell Cultures (ECACC; MDA-MB-231, catalog no. 92020424; MCF-7, catalog no. 86012803) and maintained at the Translational Pharmaceutical Research Laboratory, Faculty of Pharmacy, Universitas Padjadjaran, Sumedang, Indonesia.

MDA-MB-231 cells were maintained at 37 °C in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 15% fetal bovine serum (FBS), 2 mM L-glutamine, 100 IU/mL penicillin, and 10 μg/mL streptomycin in a humidified atmosphere containing 5% CO₂. The cells reached optimal growth at a density of 1–3 × 10^4 cells/cm². Once confluent, cultures were harvested by adding 1–2 mL of 0.25% trypsin. The detached cells were transferred to a conical tube, neutralized with fresh DMEM containing FBS to a final volume of 10 mL, and centrifuged at 2000 rpm for 5 min. The supernatant was discarded, and the resulting pellet was resuspended in 1 mL of medium. Viable cells were then counted using a hemocytometer. MCF-7 cells were cultured following the same procedure, except that Eagle’s Minimum Essential Medium (EMEM, EBSS) supplemented with 2 mM L-glutamine, 1% non-essential amino acids (NEAA), and 10% FBS was used.

Cell proliferation was assessed using the WST-8 assay. MDA-MB-231 and MCF-7 cells were seeded into 96-well plates and treated for 24 h with α-mangostin (60, 120, 240, and 480 µM), LiCl (2.5, 5, and 10 mM) or their combinations. After treatment, cells were washed with PBS and incubated with 100 µL of 0.5 mg/mL WST-8 solution at 37 °C for 2–4 h. Absorbance was measured at 450 nm using a Tecan Infinite spectrophotometer, and cell viability was calculated relative to untreated controls.

The MDA-MB-231 cell line was seeded into 6-well plates and incubated for 24 h. Cells were then treated with α-mangostin at concentrations of 5 μM and 10 µM or with cisplatin at 1.25 µM and 2.5 µM, either alone or in combination with LiCl (5 mM). RNA was isolated using the GeneZol™ Kit according to the manufacturer’s instructions, and the resulting RNA was resuspended in nuclease-free water (NFW). RT-qPCR was performed with the SensiFAST™ SYBR^® No-ROX One-Step Kit in a total reaction volume of 20 μL. The amplification program consisted of an initial denaturation at 95°C for 5 s, followed by 40 cycles of 95 °C for 30 s, 55 °C for 60 s, and 72 °C for 60 s. Primers specific to the target genes are listed in Table 2, with GAPDH used as the housekeeping control gene for normalization (Sun et al., 2015; Xiang et al., 2021). Relative gene expression was calculated using the 2^−ΔΔCT method, and all experiments were conducted in biological triplicates (n = 3).

Statistical analyses were performed using Microsoft Excel and GraphPad Prism 10.0.0 (GraphPad Software). All data are expressed as mean ± SD. Differences between groups were analyzed using one-way ANOVA followed by Tukey’s post hoc test, with significant p-values indicated in the figures. Significance levels are indicated as follows: P < 0.05; *P < 0.01; and **P < 0.001.

Previous findings suggest that α-mangostin exerts antiproliferative effects in MDA-MB-231 cells by modulating the transcription of key Wnt-regulated genes involved in proliferation. To further elucidate the molecular basis of this effect, in silico analyses were performed to investigate the interaction of α-mangostin with β-catenin and LRP6. Figure 1 shows the binding profiles of α-mangostin and XAV939 within the β-catenin union site (US), the allosteric site (AS), and the LRP6 receptor. The surface representations highlight the spatial orientation and electrostatic characteristics of the binding pockets, providing context for the stability of interactions. At the β-catenin US, α-mangostin exhibited a docking score of −5.03 kcal/mol with a relatively weak binding affinity (Kd = 206.27 μM). By comparison, XAV939 demonstrated slightly stronger binding at −5.62 kcal/mol with a Kd of 76.33 μM, suggesting better accommodation within the same pocket. The β-catenin AS displayed similar energetics, with α-mangostin docking at −5.03 kcal/mol (Kd = 204.77 μM) and XAV939 at −5.40 kcal/mol (Kd = 109.27 μM), consistent with the more solvent-exposed and structurally relaxed nature of this site. In contrast, α-mangostin exhibited substantially stronger binding to LRP6, with a docking score of −7.23 kcal/mol and a low dissociation constant of 5.02 μM, indicating high affinity. XAV939 also bound LRP6 with favorable energetics (ΔG = −6.63 kcal/mol, Kd = 13.70 μM), but with lower affinity compared to α-mangostin. The binding orientation of α-mangostin in LRP6 was more deeply encapsulated, suggesting enhanced shape complementarity. Electrostatic surface mapping further revealed that LRP6 provides superior charge compatibility, facilitating stronger non-covalent interactions, including hydrogen bonding and π-interactions. Taken together, these results indicate that LRP6 serves as a more selective and energetically favorable receptor for α-mangostin. The improved structural anchoring in LRP6 supports its potential for greater biological efficacy and retention, which may be critical for downstream pharmacological activity (Medina-Barandica et al., 2023).

Structural stability of the protein–ligand complexes was assessed by root mean square deviation (RMSD) analysis (Figure 2). In the β-catenin US complex, α-mangostin induced a gradual increase in protein RMSD, exceeding 0.35 nm toward the end of the simulation, whereas XAV939 maintained greater stability with values below 0.30 nm. Ligand RMSD plots showed that α-mangostin fluctuated around 0.20–0.22 nm, while XAV939 remained slightly lower at 0.15–0.20 nm. The β-catenin AS complex demonstrated improved stability, with both ligands maintaining consistent protein RMSD values between 0.25 and 0.30 nm. Ligand fluctuations in AS were also reduced compared to US, although α-mangostin still exhibited slightly higher deviation than the reference ligand. By contrast, LRP6 complexes exhibited the highest stability, with protein RMSD values consistently within the range of 0.20–0.30 nm throughout the simulation. Ligand RMSDs in LRP6 remained particularly low, with α-mangostin deviating by less than 0.15 nm. These findings indicate that LRP6 provides a more rigid and stable binding environment compared to β-catenin. The stable RMSD patterns observed in LRP6 suggest minimal structural rearrangements during ligand binding. Overall, these results highlight LRP6 as a conformationally stable and favorable target, supporting its potential for further drug development under physiological conditions (Aulifa et al., 2024).

To examine residue-level dynamics, root mean square fluctuation (RMSF) analyses were performed, as shown in Figure 3. In the β-catenin US complex, RMSF values for both ligands remained below 0.5 nm, with α-mangostin producing lower fluctuations than XAV939 across several regions. A minor peak was observed near residue index 600 for both ligands, indicating localized flexibility. In the AS configuration, the pattern changed markedly, as XAV939 displayed prominent peaks exceeding 2.0 nm, while α-mangostin exhibited restrained movements, with most residues fluctuating by less than 0.5 nm. These results suggest that α-mangostin stabilizes local dynamics more effectively in the AS. For LRP6, residue fluctuations were even more controlled, particularly in chain A, where RMSF values ranged between 0.2 and 0.4 nm. Chains B and C demonstrated similarly low fluctuations, with α-mangostin yielding slightly smoother profiles and values consistently below 0.3 nm. Across all chains, α-mangostin reduced residue mobility more effectively than the reference ligand. Reduced flexibility at the binding site contributes to enhanced ligand residency and complex stability. Overall, these RMSF trends support the conclusion that LRP6 provides a more rigid and stable interaction interface compared to β-catenin, a property advantageous for the design of ligands targeting dynamic protein systems.

Figure 4 presents the radius of gyration (Rg) and solvent-accessible surface area (SASA) analyses to evaluate protein compactness and surface exposure. In the β-catenin US complex, Rg values for α-mangostin fluctuated between 3.30 and 3.50 nm, slightly higher than those observed with XAV939. SASA plots indicated that α-mangostin increased solvent exposure, reaching approximately 240 nm². In the AS complex, both ligands exhibited higher Rg values, up to 4.35 nm, while SASA values remained comparable, suggesting moderate surface breathing and reduced folding compactness. By contrast, LRP6 demonstrated tighter Rg distributions (4.10–4.30 nm) and consistently lower SASA values compared with the β-catenin systems. α-Mangostin further reduced SASA in LRP6 relative to XAV939, reflecting deeper ligand embedding within the protein pocket. Compact protein structures are generally associated with enhanced energetic stability and resistance to denaturation or degradation (Buchberger et al., 2010). while reduced solvent exposure favors stronger hydrophobic interactions and improved binding efficiency. Collectively, the Rg and SASA results confirm that LRP6 maintains a more compact and tightly folded conformation during simulation, consistent with earlier findings on backbone and residue-level dynamics. Moreover, α-mangostin contributes to sustaining protein compactness in LRP6, reinforcing its potential as a stable binder.

Hydration behavior around the ligand-binding regions was evaluated using radial distribution function (RDF) analysis, as shown in Figure 5. RDF describes the spatial arrangement of solvent molecules in the vicinity of ligands. In the β-catenin US complex, both ligands exhibited broad and less intense RDF peaks, suggesting a disordered solvation shell. XAV939 reached a maximum g(r) value of approximately 13, whereas α-mangostin was slightly lower at around 12. The β-catenin AS complex displayed sharper peaks, although still less structured compared with LRP6. In this system, XAV939 peaked at g(r) = 14, while α-mangostin remained lower, indicating that α-mangostin generates a less disrupted hydration environment. By contrast, LRP6 showed the most structured hydration profile, with sharp peaks centered at 0.5–0.6 nm. Notably, α-mangostin reached a g(r) value close to 25, significantly higher than XAV939. These results suggest that LRP6 forms a compact and stable hydration shell around the ligand. A well-ordered water layer enhances molecular rigidity and contributes favorable enthalpic interactions, thereby influencing ligand stability and retention within the binding pocket. Collectively, the RDF patterns support the conclusion that LRP6 provides a more dynamically favorable environment, consistent with the RMSD and RMSF analyses, which demonstrate the stability of LRP6–ligand complexes.

Altogether, the simulation results provide a cohesive view of the structural behavior of β-catenin and LRP6 when bound to XAV939 and α-mangostin. β-Catenin, particularly in its AS binding configuration, exhibited greater flexibility and higher solvent exposure, which may limit its suitability as a stable target for inhibitors (Hankey et al., 2018). Although the US configuration showed somewhat improved stability, it still demonstrated more fluctuations compared with LRP6. Across all structural metrics evaluated (RMSD, RMSF, RDF, Rg, and SASA), LRP6 consistently outperformed β-catenin, indicating that LRP6 is a structurally rigid and solvent-protected receptor well suited for ligand interactions (Raisch et al., 2019). α-Mangostin performed favorably by stabilizing both protein and solvent dynamics, particularly within the LRP6 binding pocket, suggesting its potential role as a multi-site inhibitor of Wnt signaling. Given that β-catenin regulates downstream oncogenic pathways, especially in colorectal and breast cancers. At the same time, LRP6 functions as a co-receptor in the upstream Wnt complex; targeting both proteins could provide synergistic suppression of Wnt signaling. Previous studies have also shown that LRP6-targeted agents reduce cancer proliferation with fewer off-target effects. The consistent molecular stability observed in this study further supports α-mangostin’s candidacy for development as a Wnt pathway inhibitor (Ariyanto et al., 2023). Nevertheless, additional in vitro and in vivo studies are warranted to validate these simulation-based insights.

The MM/PBSA method was applied to estimate the binding free energies of α-mangostin and XAV939 with β-catenin (US and AS) and LRP6, providing insights into the stability of the complexes. As shown in Table 3, the calculated ΔG_bind values for α-mangostin were −76.437 kJ/mol at the β-catenin US site, −76.167 kJ/mol at the AS site, and −96.659 kJ/mol with LRP6. These results indicate that LRP6 forms the most energetically favorable complex with α-mangostin. The electrostatic contribution (ΔE_elec) was particularly significant in the LRP6–α-mangostin complex (−53.275 kJ/mol), whereas the AS site showed negligible electrostatics (0.042 kJ/mol). Van der Waals interactions (ΔE_vdW) followed a similar pattern, being strongest in the LRP6 complex (−187.046 kJ/mol). These findings suggest that LRP6 provides a deeply hydrophobic and structurally complementary binding pocket for α-mangostin. The gas-phase interaction energy (ΔG_gas), which combines ΔE_vdW and ΔE_elec, was lowest for the LRP6 complex (−19.664 kJ/mol), further supporting its thermodynamic advantage. Although LRP6 showed the highest desolvation penalty (ΔG_solv = 163.326 kJ/mol), this was offset by its favorable gas-phase and non-covalent interactions. In comparison, the β-catenin US and AS sites demonstrated similar ΔG_bind values, with weaker electrostatic contributions in the AS site balanced by stronger hydrophobic interactions. Collectively, the MM/PBSA results confirm that α-mangostin exhibits a stronger binding affinity for LRP6 than for β-catenin.

A similar trend was observed with the reference ligand XAV939, although its ΔG_bind values were consistently less negative than those of α-mangostin. XAV939 bound to the β-catenin US with a ΔG_bind of −38.929 kJ/mol and to the AS with −26.072 kJ/mol, indicating relatively weak complex formation. Its interaction with LRP6 was more favorable (ΔG_bind = −60.925 kJ/mol) but remained notably less negative than that of the α-mangostin–LRP6 complex. In all comparisons, α-mangostin demonstrated stronger binding, particularly with LRP6, thereby reinforcing its potential as a superior inhibitor scaffold for modulating the Wnt/β-catenin pathway. The strong electrostatic and hydrophobic contributions observed in the α-mangostin–LRP6 complex suggest robust anchoring and reduced dissociation, which may support prolonged biological activity. These findings are consistent with structural dynamics results, including the low RMSF and stable RDF profiles observed in LRP6 simulations. The pronounced ΔG_solv further underscores the role of hydrophobic stabilization within the binding pocket. The comparable ΔG_bind values at the β-catenin US and AS imply that both regions remain accessible targets, albeit thermodynamically less favorable than those at LRP6. Collectively, LRP6’s superior energetic profile highlights it as a more attractive therapeutic target. These binding energy trends provide a rationale for ligand optimization strategies aimed at strengthening van der Waals and electrostatic interactions. Experimental validation using biophysical approaches such as isothermal titration calorimetry (ITC) or surface plasmon resonance (SPR) will be essential to confirm these computational predictions.