Fermentation and extraction Swietenia macrophylla seed

This method was modified from a prior study [11], 20 grams of mahogany seeds were placed into each of three containers, and 100 mL of sterile distilled water was added (moisture content 70%). The mixtures were sterilized for 45 minutes and then inoculated with Aspergillus niger ATCC 9142 was obtained from the Bogor Agricultural Institute (IPB). spore suspension was inoculated at 1% (v/v) of a spore suspension prepared in sterile water; spore concentration of the inoculum was approximately 1.5 × 10⁸ spores/mL. After a 48-hour incubation at 37° Celsius, the fermented mahogany seeds underwent defatting using n-hexane in a 1:1 ratio for 6 hours. The mixture was then filtered, and the residue was dried. Subsequently, the dried residue was extracted using 96% ethanol at a 1:10 ratio maceration for 24 hours. The resulting filtrate was filtered and concentrated using a rotary evaporator to obtain a thick extract as the final sample, Extract yield was 0.05% obtain and the sample called Ethanol Fraction of fermented Swietenia macrophylla (EFFSM).

Metabolite secondary

Total flavonoid content

The total flavonoid test follows the method carried out [15] with minor modifications. EFFSM added 1 mL with 3 mL methanol, 200 μL 2% of AlCl₃∙6H₂O, 200 μL CH₃COONa and the volume is set to 10 mL using aquades, then incubated for 30 minutes at a temperature of 25°C. Absorbance is measured at wavelength 432 nm. Using spectrophotometry UV-VIS, methanol is used as blanks and quercetin is used as standard. The blank consists of all reagents and solvents without sample solution. The quercetin calibration curve is used to determine its content. The results were given as mg Quercetin equivalent (QE)/g dry extract.

Total phenolic content

The total phenol test follows the method performed by [15] with minor modifications. 0.5 mL of EFFSM was added with 2.5 mL of distilled water, 2.5 mL 1N of FolinCiocalteu reagent. Incubated at 25°C for 15 minutes and added 2 mL 5% (w/v) sodium carbonate, incubated at 25°C for 30 minutes, and measured absorbance at 765 nm with UV-VIS spectrophotometer. Methanol is used as the control blank and gallic acid is used as standard. The blank consists of all reagents and solvents without a sample solution. A standard gallic acid calibration curve is used to determine the content. The result was given as mggallic acid equivalent (GAE)/g dry extract.

Total triterpenoid content

This method follows [16] with slight modification, 0.2 mL of EFFSM solution in a 10 mL measuring flask was heated to evaporate in a water bath, 0.5 mL of a new mixture of 5% vanillin-acetate solution (W/V) and 0.9 mL of sulfuric acid was added, mixed and incubated at 70°C for 30 min. Then the mixed solution is cooled and diluted to 5 mL with acetic acid. Absorbance was measured at 573 nm with a UV-VIS spectrophotometry. Methanol is used as a blank, and ursolic acid is used as standard. The blank consists of all reagents and solvents without a sample solution. A standard ursolic acid calibration curve is used to determine the content. The results were given as mg ursolic acid equivalent (UAE)/g dry extract.

in vitro amyloidosis assay

Thermal induced

BSA protein solution 10 mg/mL. Then 1 mL BSA solution added with an EFFSM 0.2 mL of 0.1 mg/mL aquadest as negative control is added and quercetin 100 ppm positive control. Then all samples are heated to 700°C for 45 minutes. After clumping, the sample was measured for absorbance at a wavelength of 600 nm [17].

Induction dithiolteriolriol (DTT)

BSA protein aggregation process induced by DTT. Aggregate formation follows the method [18]. Each mixture contains 20 mM phosphate buffer and 10 mM DTT. The procedure is to add quercetin then incubate at 37°C for 20 hours. The same procedure is applied to the bioactive compound. The turbidity of the BSA protein mixture aggregates with and without inhibitors was measured using a UV-Vis spectrophotometer at a wavelength of 600 nm.

Congo red binding assay

The anti-aggregation assay was performed with slight modifications. This method was use after thermal induced and DTT Induced. 7 mg/mL Congo Red solution was freshly prepared in phosphate buffer and filtered using a 0.2 μm syringe filter immediately before use. A volume of 5 μL of the Congo Red solution was added to phosphate buffer in a quartz cuvette, and the absorption spectrum was recorded in the range of 400–700 nm as a baseline. Subsequently, 5–10 μL of the protein aggregate solution was added to the cuvette, followed by incubation at room temperature for 30 minutes, during which a visible red precipitate indicated potential amyloid formation. After incubation, the absorption spectrum was recorded again in the same range (400–700 nm). To analyze amyloid fibril formation, the spectrum of Congo Red alone was mathematically subtracted from the spectrum of the protein– Congo Red mixture. A maximum spectral difference at approximately 540 nm was taken as an indication of amyloid fibril presence.
All measurements were conducted in triplicate (n = 3), and the results were expressed as mean ± standard deviation (SD). Statistical analysis was performed using one-way ANOVA followed by Tukey’s post hoc test, with P < 0.01 considered statistically highly significant, P < 0.05 significant, and while ns considered in significant difference.

Untargeted metabolomic fermented swietenia macrophylla using LCMS

The fermented mahogany sample was diluted with methanol. Soluble extract was then filtered using 0.45-micron Millipore filter. 5 μL filtrate of the sample was injected into the LC-ESI-QTOF system. LC-MS analysis was done by using UPLC-M Equipped with a binary pump. LC is connected to QTOF mass spectrometer coupled to ESI. The MS used was Xevo G2-S QTOF with positive ionization mode. ESI parameters used capillary temperature 120°C, gas atomizer 50 L/h, and source voltage + 2.9 kV. Full scan mode from m/z100–500 was done with source temperature 41°C. UPLC column used was Acquity UPLC HSS C181.8 μm, 2.1 × 150 mm. Solvent A was 0.1% formic acid in water, solvent B was 0.1% formic acid in acetonitrile. Solvents were set at a total flow rate of 0.4 mL/min. Isocratic elution system was run at 0–0.5 min with ratio 95:5; linear gradient of solvent A was from 95% to 5% in 15 min.

In silico analysis molecular docking and molecular dynamic

This method follow [7] with slightly modification. The three-dimensional structure of the target protein was obtained from the Protein Data Bank (PDB) in PDB format with code PDB ID 5OQV, representing Aβ42 fibril, which only one chain of 5OQV was used for docking, in this study we chose chain A beca. 5OQV had no attached or water molecules. Polar hydrogens and Kollman charges were added to the receptor and AD4 type were assigned. The receptor was then saved as PDBQT format.
Ligands identified by LC-MS (top 5 by %TIC: myricetin, gallocatechin, swietenolide, swietenine, pyrogallol) were retrieved from PubChem (SDF format), converted to 3D and protonated for pH 7.4 (OpenBabel). Gasteiger charges were assigned and ligands saved in PDBQT format.
Docking was performed using AutoDock Vina (PyRx 0.8 frontend). Blind docking has been used for this method. The grid box was centered at the coordinates (x = 38.6598, y = 60.0364, z = 53.5879) with dimensions (size x = 35.994 Å, size y = 46.7367 Å, size z = 15.5034 Å) covering all the surface of the protein. Interaction analysis was carried out using BIOVIA Discovery to identify hydrogen bonds and hydrophobic contacts, with ligands forming at least two hydrogen bonds with active site residues considered favorable. Visualization in BIOVIA Discovery Studio confirmed ligand orientation, and selection was based on energy threshold, hydrogen bonding, and site specificity to identify potential amyloidosis inhibitors.

Molecular dynamic simulation (coarse-grained)

CABS-Flex 2.0 was used with default reference settings to explore conformational variability of the docked complexes (100 cycles, generating 100 frames, roughly equiv-alent to 10 ns sampling). RMSF per residue was computed and compared across complexes [14]. CABS-Flex is a coarse-grained method that provides a fast approximation of backbone flexibility; it does not substitute for atomistic MD with explicit solvent. We therefore interpret RMSF results qualitatively and recommend follow-up atomistic MD (e.g., GROMACS: 100 ns) for the top compounds to quantify stability (RMSD, RMSF, Rg, H-bonds, SASA).

Fermentation of Swietenia macrophylla seeds by Aspergillus niger markedly increased the levels of secondary metabolites (Table 1). Total flavonoid content in F-28 (0.115 QE/g) increased by approximately 130% compared to NF (0.05 QE/g), while total phenolic content showed a smaller increase of 4% (4.8 vs. 4.6 GAE/g). The most pronounced change was observed in total triterpenoid levels, which rose by 179% from 69.7 to 194.5 UAE/g. This trend aligns with previous studies reporting that fungal fermentation enhances the metabolite such as flavonoid and phenolic content [11]. Among the three metabolite groups, triterpenoids were the most abundant in both extracts.

Based on Table 1, the observed increases after fermentation can be attributed to the enzymatic activity of Aspergillus niger. This fungus produces lignolytic enzymes capable of breaking down cell walls, releasing bound phenolics into free phenolic forms. It also produces β-glucosidase, which hydrolyzes glycosides into aglycones, facilitating the biotransformation of flavones into aglycones. A similar mechanism likely explains the increase in triterpenoids, where β-glucosidase converts glycosylated saponins into aglycones. These processes collectively enhance the availability and concentration of bioactive compounds in the fermented extract. Given these findings, the fermented extract (F-48) will be further analyzed to profile its bioactive compounds in detail and to evaluate its anti-amyloidosis in vitro and in silico.

in vitro anti amyloidosis

Anti-aggregation effect of the ethanol extract of Swietenia macrophylla seeds fermentation, two complementary assays were used: turbidimetric assay, which measures light scattering due to aggregate formation, and Congo red binding assay, which specifically detects amyloid fibril structures through dye interaction. The assays were performed under two stress conditions: thermal-induced denaturation and DTT-induced disulfide bond disruption. Quercetin (100 ppm) was used as a positive control. The in vitro analysis of amyloidosis shows at Figure 1.

Based on Figure 1, under thermal-induced stress, the ethanol fraction of fermented Swietenia mahagoni seeds (EFFSM) exhibited consistently stronger inhibitory activity than quercetin in both analytical assays. In the turbidimetric assay, EFFSM inhibited protein aggregation by 57.12 ± 5.37%, compared to 43.24 ± 6.84% for quercetin.
Similarly, in the Congo red assay, inhibition reached 79.09 ± 6.47% for the ethanol fraction, which was significantly higher (P < 0.01) than the 44.44 ± 12.23% observed for quercetin. Post hoc Tukey HSD analysis confirmed the significance of this difference, highlighting that the ethanol fraction outperformed the positive control specifically in the Congo red assay. Interestingly, inhibition values obtained from Congo red binding were consistently higher than those from turbidimetric analysis, suggesting that Congo red is more sensitive in detecting early fibrillar intermediates or β-sheet structured aggregates that remain invisible to turbidity measurements. This observation underscores the importance of employing complementary methods to fully capture the complexity of protein aggregation processes.
In the DTT-induced aggregation model, both quercetin and EFFSM displayed reduced inhibitory effects, reflecting the more challenging conditions imposed by reductive stress. The disruption of disulfide bonds by DTT facilitates protein unfolding and exposes hydrophobic residues, promoting aggregation that is more resistant to inhibition. Under these conditions, quercetin inhibited aggregation by only 11.33 ± 4.85% in the turbidimetric assay, while the ethanol fraction achieved a significantly higher inhibition of 27.15 ± 6.09% (P < 0.01). By contrast, Congo red assay results remained relatively high for both treatments, with 69.09 ± 7.43% inhibition for quercetin and 63.00 ± 4.39% for the ethanol fraction, with no significant difference between them. This indicates that while the ethanol fraction was more effective at reducing amorphous aggregate formation (turbidity), its capacity to prevent β-sheet fibrillation under reductive stress was comparable to quercetin.
These findings indicate that the ethanol fraction is more effective against aggregation driven by thermal denaturation than by reductive stress, likely due to differences in aggregation mechanisms. Whereas thermal denaturation mainly involves hydrophobic interactions, DTT-induced aggregation is driven by disulfide bond reduction and subsequent misfolding. Given these promising results, further investigations will be performed to identify the specific bioactive compounds using LC-MS and to elucidate their molecular interactions with target proteins through in silico analysis.

Profiling metabolite Swietenia macrophylla seed fer-mented

Based on Table 2, pyrogallol acid was detected at a retention time of 0.47 minutes with 6.1% area. This simple phenolic compound is well known for its antioxidant capacity, and together with gallocatechin (4.99 min, 12.09%), a flavonoid recognized for its anti-amyloidogenic activity, may interact with protein surfaces via hydrogen bonding and aromatic stacking, thereby preventing early protein misfolding events. At 8.28 minutes, myricetin (13.89%) was identified [19]. Previous study posses that myricetin have anti-amyloidogenic activity against superoxide dismutase 1 aggregation [20].

The dominant percentage area TIC was observed at 16.23 minutes, corresponding to swietenine (25.33%), another limonoid specific to mahogany along with swietenolide (17.68%), a characteristic limonoid of mahogany seeds known for its antibacterial and anti-inflammatory activities [8].
These LC-MS findings provide the basis for further in silico analysis to evaluate whether compounds with higher percentage areas also demonstrate stronger interactions with amyloidogenic targets. By correlating relative abundance with binding affinity and interaction profiles, the computational approach will help to clarify whether the predominant metabolites, particularly limonoids such as swietenine and swietenolide, play a central role in modulating protein aggregation, or whether less abundant phenolics and flavonoids contribute disproportionately through more favorable molecular interactions.

In silico: molecular docking analysis

Previous study has predicted the potential of triterpenoid compounds such as swietenine [14], swietenolide, and khayasin T to interact with Bovine Serum Albumin (BSA) through in silico docking, aiming to identify possible binding sites. Following LC-MS analysis, several new compounds were identified, namely myricetin, gallocatechin, and pyrogallol acid. However, BSA-based aggregation has limitations when used as a model of amyloid fibrillogenesis: in many experimental conditions BSA predominantly reports amorphous or non-fibrillar aggregation rather than the ordered β-sheet fibrils typical of pathogenic amyloids [14]. To strengthen interpretations of Congo red binding results and to elucidate the molecular mechanisms and candidate inhibitory compounds, we therefore complemented BSA experiments with in silico studies using β-amyloid fibril models thereby expanding the understanding of bioactive components in the sample and their possible interaction mechanisms with the target protein. Latrepiridine used for control positive through their anti-amyloidosis activity [21].
As shown as Figure 2, molecular docking was performed to evaluate the interaction between selected compounds from mahogany seed extract and β-amyloid. The binding affinities, hydrogen bonds, hydrophobic contacts, and van der Waals interactions of each ligand are summarized in Table 3 and Supplementary Table 1. Latrepirdine, used as a positive control, exhibited the strongest binding affinity (−7.2 kcal/mol), and thus served as the reference drug throughout the analysis. (−7.2 kcal/mol). Its stability was supported by a hydrogen bond with PHE19, hydrophobic interaction with PHE20, and van der Waals contacts with LEU17, VAL18, ASN27, and ILE31.

Compared to latrepirdine, gallocatechin (−6.1 kcal/mol) and myricetin (−6.2 kcal/mol) relatively high affinities, suggesting comparable binding potential though still weaker than the reference. Gallocatechin formed one hydrogen bond with GLU22 and several hydrophobic interactions (PHE19, ALA21, VAL24), while myricetin established two hydrogen bonds (PHE20, ALA30) and hydrophobic contacts with PHE19 and ILE31.
Swietenolide (−5.8 kcal/mol) and swietenine (−5.5 kcal/ mol) showed moderate binding affinities. Both compounds weaker than latrepirdine, interacted mainly through hydrogen bonding at PHE20 or GLU22, supported by van der Waals contacts with residues such as ALA21, VAL18, and ASP23. Pyrogallol acid exhibited the weakest binding affinity (−3.4 kcal/mol), forming only a single hydrogen bond with PHE20 and limited van der Waals interactions. Overall, gallocatechin and myricetin were identified as the most promising test compounds, as their affinities approached that of the reference drug latrepirdine. This benchmark comparison strengthens the interpretation of their potential role as inhibitors of amyloid aggregation. To further validate these findings, Root Mean Square Fluctuation (RMSF) analysis was performed to evaluate the dynamic stability of β-amyloid residues upon ligand binding.

In silico: molecular dynamic RMSF

The Root Mean Square Fluctuation (RMSF) analysis was performed to evaluate the residue-level flexibility of β-amyloid in complex with each ligand, using the interaction sites of latrepirdine (Leu17, Val18, Phe19, Phe20, Asn27, and Ile31) as reference position (Supplementary Table 2).
As visualized in Figure 3 and summarized in Table 4, Latrepirdine, the positive control and reference drug, showed consistently low RMSF values across all residues, particularly at Ile31 (0.654 Å), indicating a highly stable interaction and minimal structural fluctuation of the β-amyloid backbone. This profile serves as a stability benchmark for the tested compounds.

Compared with latrepirdine, gallocatechin displayed a relatively stable profile with low RMSF values at Leu17 (3.171 Å) and Val18 (3.598 Å), although higher fluctuations were observed at Asn27 (6.937 Å). Myricetin also showed moderate stability, with RMSF values ranging from 3.856–4.770 Å across residues and lower fluctua-tions at Ile31 (3.201 Å) compared to gallocatechin.
Swietenolide exhibited moderate fluctuations (3.518– 6.432 Å), while swietenine showed higher instability at Phe20 (7.667 Å) and Asn27 (7.408 Å), suggesting weaker stabilization of the amyloid residues. Pyrogallol acid demonstrated the highest RMSF across almost all residues (up to 7.545 Å at Asn27), consistent with its weak binding affinity in docking results.
Overall, RMSF analysis confirmed that gallocatechin and myricetin provide more stable interactions at the key binding residues compared to other test compounds. These results complement the docking findings, strengthening the evidence that both compounds may effectively stabilize β-amyloid structure and interfere with fibril formation.