Potential anti‐obesity effect of saponin metabolites from adzuki beans: A computational approach

Abstract In contrast to its widespread traditional and popular culinary use to reduce weight, Vigna angularis (adzuki beans) was not subjected to sufficient scientific scrutiny. Particularly, its saponins whose role was never investigated before to unveil the beans’ antidiabetic and anti‐obesity effects. Four vital pancreatic and intestinal carbohydrate enzymes were selected to assess the potency of the triterpenoidal saponins of V. angularis to bind and activate these proteins through high‐precision molecular modeling and dynamics mechanisms with accurate molecular mechanics Generalized Born Surface Area (MMGBSA) energy calculations; thus, recognizing their anti‐obesity potential. Our results showed that adzukisaponin VI and adzukisaponin IV were the best compounds in the α‐amylase and α‐glucosidase enzymatic grooves, respectively. Adzukisaponin VI and angulasaponin C were the best fitting in the N‐termini of sucrase‐isomaltose (SI) enzyme, and angulasaponin C was the best scoring compound in maltase‐glucoamylase C‐termini. All of them outperformed the standard drug acarbose. These compounds in their protein complexes were selected to undergo molecular simulations of the drug‐bound protein compared to the apo‐protein through 100 ns, which confirmed the consistency of binding to the key amino acid residues in the four enzyme pockets with the least propensity of unfolding. Detailed analysis is given of the different polar and hydrophobic binding interactions of docked compounds. While maltase–adzukisaponin VI complex scored the lowest MMGBSA free energy of −67.77 Kcal/mol, α‐amylase complex with angulasaponin B revealed the free binding energy of −74.18 Kcal/mol with a dominance of van der Waals energy (ΔEVDW) and the least change from the start to the end of the simulation time. This study will direct researchers to the significance of isolating the pure adzuki saponin components to conduct future in vitro and in vivo experimental works and even clinical trials.


| INTRODUC TI ON
Natural products have been the source of bioactive molecules since antiquity with marked effectiveness against various ailments, such as inflammation, cancer, infection, bone diseases, and more (Ashraf et al., 2020(Ashraf et al., , 2022;;Moussa et al., 2020;Fayez et al., 2022;Moussa, 2013).In Japan, V. angularis, known as adzuki beans, were considered a part of the traditional Kampo medicine, which prescribed its juice and decoction to reduce aging and cardiovascular diseases (CVDs) (Maruyama et al., 2008).Eating beans to lose weight is a common folk practice in many countries, such as China, Egypt, Middle Eastern region, Japan, and South Korea.Vigna angularis, or Chi Xiao dou in Chinese, were used in the treatment of dropsy and beriberi diseases in China and even mentioned in the first medical herbal book in the sixteenth century, the Ben Cao Gang Mu (Compendium of Materia Medica) (Jiang et al., 2014;Wang et al., 2022).Obesity-related diseases can dramatically retard healthcare efforts and incur extra costs on governments, which renders safe and effective anti-obesity solutions an urgent need.
Saponins were never thought before to be of a significant role in adzuki beans' favorable effect in diabetes and glucose intolerance (Shi et al., 2020), and adzuki studies about obesity and weight loss were very few despite the popular use, which necessitated scientific work to rationalize these notable health benefits (Liu et al., 2017).
Reports included reduction of total cholesterol levels, triglycerides, hepatic cholesterol, and low-density lipoprotein (LDL) cholesterol by the virtue of pancreatic lipase and α-glucosidase inhibitory effects of adzuki beans flavonoids.Our previous work (Liu et al., 2017) identified through high-performance liquid chromatography-diode array detection-electrospray ionization-mass spectrometry analyses (HPLC-DAD-ESI-MS) four flavonoids and six oleanane-type triterpene oligoglycosides named adzukisaponins in the ethanolic extracts of the red beans (Liu et al., 2017).Most of the isolated saponins reported an inhibition of nitric oxide (NO) production (Sansbury & Hill, 2014), which was proved to be closely correlated with the anti-obesity mechanism (Qian et al., 2019) (Table 1).Yet, activities on pure compounds remained to be investigated.
Pure single saponin compounds from adzuki beans are not available until now, and all the data reported the whole activity of the crude saponin extract, which have its known disadvantages (Newman, 2021).Crude extracts are abundant as promiscuous players with multiple bioactivities, and with the advent of Artificial Intelligence (AI) techniques, the use of these claimed bioactive extracts with their main components referred to as the bioactive materials is no more accepted since it misleads large compound databases and other libraries for high-throughput screening (Rodrigues, 2019).
These databases captured a significant number of compounds that will hit screens for several reasons just because they were reported invalidly as biologically active.In silico computational approaches served to overcome this problem and are invaluable nowadays in order to investigate bioactive molecules in their binding sites, their types, and stability of interactions, and in many cases to identify unknown targets (Torky et al., 2021).The cost-effectiveness and time saving conferred by such techniques encouraged researchers to revive drug discovery from natural products.This grew in line with the limiting ethical laws of animal use, which rationalized the strategy of using in silico investigations for lead identification.Furthermore, false-negative molecules can be optimized for better binding in receptor sites by medicinal chemists.Diabetes and obesity are correlated strongly to starch metabolic disorders, and starch digestive enzymes were targeted to reduce postprandial sugar and reduce complications that are of major global concern (Al-Goblan et al., 2014).The mechanism of starch digestion encompasses four major enzymes, pancreatic and salivary-amylases, maltaseglucoamylase, and sucrase-isomaltase that conduct the hydrolysis process from starch to glucose.In this study, four α-glucosidases were selected, which include human pancreatic α-amylase, human lysosomal acidα-glucosidase, GAA, the N-terminal sucrase-isomaltase, and the C-terminal subunit of human maltase-glucoamylase to demarcate the activity of more than 10 adzukisaponin compounds as potential antidiabetic and anti-obesity agents (Figure 1).While continuing our work on adzuki beans (Liu & Xu, 2016), we embarked on this study aiming to virtually screen the triterpenoidal saponins isolated from adzuki beans against four vital carbohydrate-digesting enzymes through high-precision molecular modeling and dynamics mechanisms with Molecular Mechanics energies combined with Gerneralized Born and Surface Area (MMGBSA) Continuum solvation area energy calculations to recognize and predict the molecular basis of the proposed anti-obesity potential.The proposed hypothesis is that saponins are responsible for the anti-obesity effects of adzuki beans.All saponins isolated from V. angularis, namely, adzukisaponin I, II, III, IV, V, and VI, angulasaponin A, B, C, and D (Iida et al., 1997(Iida et al., , 1999) ) were included.The selection of the study was due to the wide folk use of adzuki beans for weight loss and the structural similarity between acarbose, the standard antidiabetic drug, and many of the adzukisaponins.

| Protein and ligand preparation
The four selected carbohydrate-digesting proteins for this study with their co-crystallized ligands were the human pancreatic α-amylase in complex with montbretin A (Pdb ID:4w93), human lysosomal acidα-glucosidase, GAA, in complex with acarbose (Pdb ID:5nn8), the N-terminal sucrase-isomaltase with kotalanol (Pdb ID:3lpp), and the C-terminal subunit of human maltase-glucoamylase in complex with acarbose (Pdb ID:3top) whose crystallographic data were obtained from the Protein Data Bank (PDB) website (https:// www.rcsb.org/ ).
The Schrödinger platform was employed to prepare the chosen proteins via Protein Preparation tool in order to add hydrogen and correct problems like incomplete loops or side chains, flipped residues or unclear protonation states.Forcefield of OPLS2005 was applied for energy minimization after optimization of the preprocessed protein; moreover, ionization states were selected as 7.4 pH and water molecules were kept based on their important role in interactions (AbdelRazek et al., 2023;Torky et al., 2021).

| Receptor grid generation and ligand preparation
The native ligands like montbretin (4w93), kotalanol (3lpp), and acarbose (5nn8 and 3top) were picked and defined to locate the docking position for the wizard using the Receptor Grid Preparation in Maestro 11.6.Furthermore, the native ligands' structures were copied to separate files.Their structures were subjected to ligprep tool for energy minimization in three-dimensional (3D) low-energy formats up to 2500 iterations, and redocked using the same ligand docking protocol of the chosen compounds library in each protein to calculate the root-mean-square deviation (RMSD) and validate the procedures (Figure 2).

| Molecular docking
The Glide ligand docking protocol in Maestro was utilized to evaluate the binding affinity and stability of each of the selected compounds in the enzymatic groove of the four prepared proteins through high extra precision standards (XP).The filtering results are used to generate the complex Glide XP scores, which are created by ranking single ligand poses of separate chemical entities to separate those that bind strongly from compounds that bind weakly or are inactive through Glide score and using the "XP Pose rank" in case of generating conformers of the same ligand and considering the following equation that involves both E-model and Glidescore values.
The E-model score expresses more the protein-ligand coulomb-VdW (van der Waals) energy considering also the Glide score.The physical aspects of the binding affinities, lipophilicity, hydrogen bonds, and protein-ligand coulomb-VdW energies are represented by the Glidescore term.The grid box dimensions were specified in the grid generation tool as 14 Å × 14 Å × 14 Å, which allows the selected ligands' fitting; moreover, van der Waals factor was set to 1.00 within an OPLS2005 forcefield and workload control factor of 0.25.Finally, the poses were selected based on the lowest GlideXP score and E-model provided by poses that bind vital core amino acids in the enzymatic grooves.All protein native ligands were redocked in the primary X-ray structures of the proteins to measure their RMSD.

| MD simulation
The docking results of targeted proteins were analyzed and one compound from each study was selected for protein-ligand stability analysis at 100 ns molecular dynamics (MD) simulation using Visual Molecular Dynamics (VMD) (William, 1996) and Nanoscale Molecular Dynamics (NAMD) (Phillips et al., 2020) tools.The input files required for simulation were generated in AmberTools 21 (Case et al., 2021), therein the Antechamber was used to generate the ligand topology files while the missing hydrogens were added by using LeaP program (Case et al., 2005).Each ligand-protein complex was then solvated in a solvation box of 10 Å containing TIP3P water model (Duan et al., 2003;Phillips et al., 2020).
To further neutralize the systems, counter Na + and Cl − ions were added to the complexes.To remove clashes, the complexes were minimized by using FF14SB and General Amber Force Fields (GAFFs) for protein and ligand, respectively (Duan et al., 2003).
Finally, the solvation systems were subjected to three additional equilibrations at 200, 250, and 300 K.The processed ligand-protein complexes were then subjected to 50 ns simulation, and the MD trajectories were saved at every 2 ps interval.The trajectories were analyzed by CPPTRAJ (Weinzierl, 2021) and R package (Grant et al., 2021).

| MD simulation
The MD trajectories were analyzed to calculate the RMSD, root mean square fluctuation (RMSF), and radius of gyration (Rg) of all selected ligand-protein complexes, separately.The RMSD plot of sucrase-angulasaponin C complex showed that the systems attained equilibrium at 2 ns with an RMSD value of 1 Å (Figure S1a).
After equilibration, the complex did not show any major deviation in the RMSD value till the end of simulation, which indicated that the protein did not undergo major conformational changes during simulation.The maximum value of RMSD attained by the protein was 1.5 Å at 80 ns.The Rg described the compactness of the protein structure where higher values inferred that the protein faced some unfolding events during simulation.The Rg plot of sucrase-angulasaponin C complex showed that the system had a stable Rg value starting at 28.2 Å and reached 28.4 Å at equilibration.It gradually increased to 28.6 Å at 70 ns before attaining the previous values toward the end of simulation.The overall results showed that the complex did not face any unfolding event during simulation and remained compact (Figure S1b).Similarly, the RMSF analysis revealed that no major fluctuations were observed in protein residues.The higher RMSF value showed the loop regions while the lower RMSF indicated the rigid secondary structures.The protein residues did The screened scaffolds of adzuki beans saponin compounds and a positive drug acarbose.
not show major fluctuations, except for the N-terminals.Meanwhile, other residues remained rigid indicating the stable secondary structures (Figures S1c and S2).
The complex of maltase with angulasaponin VI was selected for the MD simulation.The RMSD plot of maltase-angulasaponin VI complex backbone atoms showed that the RMSD values of the complex attained equilibrium at 5 ns where the RMSD value reached 2 Å before it was gradually increased to 2.5 Å at 40 ns.After 40 ns, it gradually decreased to 2 Å at 60 ns and deviated in the range of 2-2.25 Å till the end of simulation (Figure S3a).The Rg plots showed that the protein bound to the ligand remained compact, as the Rg values did not show major variations and remained between 28.7 and 28.9 Å during whole simulation time, after being equilibrated at 5 ns (Figure S3b).This indicated that the ligand did not induce any unfolding event in the protein structure.Similarly, the RMSF plots show that the major fluctuations were observed in the residues from 400 to 410 (Figure S4) and C-terminal restudies (Figure S3c).
In case of the glucosidase enzyme, the complex with adzukisaponin IV was selected for MD simulation analysis.The RMSD of the backbone atoms of the glucosidase complex did not show major deviations as the plot showed that the RMSD gradually increased to 1.5 Å at 15 ns and then attained stability in the range of 1.5-2 Å till the end of simulation (Figure S5a).The Rg plot showed deviations of 0.6 Å till 20 ns and then attained stability in the range of 28.3-28.5 Å till the end of simulation (Figure S5b).Similarly, the RMSF analysis showed no major fluctuation in the glucosidase residues, except for the loop regions.Additionally, some minor fluctuations were observed, and the overall structure did not show flexibility (Figure S5c).The stable amino acid contacts are shown in Figure S6.
For the α-amylase protein, angulasaponin B and adzukisaponin VI-complexes were selected for stability analysis.The RMSD of α-amylase-adzukisaponin VI complex attained stability in the range of 1.25-1.5Å after 10 ns and then remained in the same range till the end of simulation.While, α-amylase-angulasaponin B complex remained less than 2 Å till 20 ns and then showed deviations for some time and then attained stability in the range of 2-2.5 Å till the end of simulation with some minor deviations at 60 and 75 ns (Figure S7a).S7b).The RMSF analysis showed that the loop regions in the angulasaponin B complex exhibited more fluctuations than the adzukisaponin VI complex as the RMSF values increased to 6 Å at the residues ranging from 140 to 160 in the angulasaponin B complex (Figure S7c).The remaining residues remained rigid with minor fluctuations (Figure S8).

| MMGBSA
Molecular mechanics Generalized Born Surface Area (MM/ GBSA) method was used to calculate the total binding free energy (ΔGtotal) for both complexes.The ΔGtotal value is usually used to estimate the stability of protein-ligand complex.The lower values of ΔGtotal indicate that the complex is more stable and vice versa.
It was computed as a sum of the protein-ligand complex and the difference of protein and its ligands' free energies.The total binding free energy estimated using MM/GBSA model is the outcome of the contribution of various protein-ligand interactions, such as van der Waals energy (ΔEvdW), electrostatic energy (ΔEele), and ΔGGB (electrostatic contribution to solvation free energy by The validation RMSD values in the four selected enzymes, (a) 4w93, (b) 3top, (c) 3lpp, and (d) 5nn8.
Generalized Born (GB)).The total binding free energies are given in Table 6.The binding energy contributions of the α-amylase complexes are compared and shown in Figure S9.The ΔEvdW contribution of α-amylase-angulasaponin B complex was more than that of α-amylase-adzukisaponin VI complex, but the electrostatic contribution of amylase-adzukisaponin VI complex was even more.The GB contribution showed that α-amylase-adzukisaponin VI complex has a higher GB value than α-amylase-angulasaponin B complex.
The total binding free energies of both complexes were 46.77 for α-amylase-adzukisaponin VI complex and −74.18 for α-amylase-angulasaponin B complex (Figure S9).

| DISCUSS ION
Carbohydrate intake was highly correlated to the metabolic syndrome and its bad prognosis (Hyde et al., 2019).Dietary intake with low carbohydrate and high fat proved to be beneficial to metabolic syndrome patients than high carbohydrate and low-fat one.
Saponins were reported to improve insulin resistance and suppress visceral fat accumulation; for example, in the staple food of Chenopodium quinoa Willd.(Li et al., 2022); moreover, sea cucumber extracts revealed favorable effect in obese high-fat fed rats through its effect on pancreatic lipase, reducing total cholesterol, LDL, and total triglycerides (Guo et al., 2016).
They are legumes with significant economic importance and are con-  subject's study reported the marked increase of high-density lipoprotein (HDL) in volunteers who received the adzuki bean extract with no notable side effects during a period of 8 weeks (Kitano-Okada et al., 2012;Wang et al., 2022).For diabetic noninsulin-dependent patients, adzuki beans controlled the postprandial glucose level and regulated the oxidative damage induced by inflammation in ailments like cancer, CVD, and atherosclerosis via their phenolic components (Luo et al., 2016).Upon studying the black adzuki bean extract and its effect on pancreatic β-cells in high glucose-induced glucotoxicity condition, cells' viability was ameliorated and restored to the normal glucose level situation; furthermore, insulin secretion was controlled to reduce the insulin resistance index in high-fat diet-induced glucose-intolerant obese C57BL/6J mice (Kim et al., 2016).
Additionally, black adzuki beans affected the adipocytes in a positive anti-adipogenic way where it raised gene expression of lipolytic genes as adipose triglyceride lipase (ATGL) and hormone-sensitive TA B L E 3 Different interactions of the selected compounds in α-glucosidase protein.lipase (HSL) and lowered the messenger RNA (mRNA) expression of transcription factors like peroxisome proliferator-activated receptor gamma (PPARγ) and CCAAT/enhancer-binding protein α (C/EBPα) as well as cell proliferation, related to adipogenesis (Kim et al., 2015).
A detailed description of the reported obesity-related research of adzuki beans is shown in Table 1.

| Sucrase-isomaltase
Sucrase-isomaltase (SI) is another carbohydrate-digesting enzyme belonging to the GH31 subfamily for hydrolyzing the terminal starch products, in the form of 1,4-and/or 1,6-linked oligosaccharides with overlapping specifications.This binding site was extensively analyzed for the binding of 10 adzukisaponins together with acarbose and kotanalol as control active molecules.An overview of the SI active site revealed three characteristic features, the buried hydrophilic subsite −1 interacting with the nonreducing part of the substrate, the shallow hydrophilic surface subsite +1 interacting with the reducing terminals, and another hydrophobic +1 subsite also involved in nonspecific contacts with the reducing groups of substrates (Sim et al., 2010).The catalytic reaction takes place between these two hydrophilic subsites to cleave the glycosidic link via Koshland mechanism without altering the anomeric carbon configuration.The conserved Trp327 in most of the GH31 family was important for the 1,6-linkage specificity.Acarbose-binding mode was largely influenced by its hydrophilic nature allowing it to score −16.048Kcal/mol through more than eight ionic hydrogen bond TA B L E 4 Different interactions of the selected compounds in the maltase protein.C, were able to exceed other ligands by a high margin score clearly based on their auspicious three-dimensional (3D) orientation in the binding pocket exposing them to a larger surface area of hydrophobic contacts in the shallow +1 subsite (Figure S10).This was in accordance with Sim et al. who attributed substrate discrimination to the +1 subsite.By inspecting the two-dimensional (2D) interactions, Asp571 added more stability and better fitting to the highest two scoring compounds as well as angulasaponin B and adzukisaponin IV.
A careful analysis of compounds scoring below −7 Kcal/mol demonstrated that they were incapable of interacting with the vital amino acid residues, which also showed their lack of diglucoside moiety in the triterpenoidal E ring.
Kotanalol-binding Glide XP score was −12.559Kcal/mol which was rationalized mainly through the interaction of its hydroxyl groups with the −1 subsite amino acids such as Asp 472, Asp355, Asp571, and His629 as well as salt bridges to Asp571 and Asp472.
Kotanalol showed much tighter binding to the SI than acarbose and able to attain the best orientation in the active site (Sim et al., 2010).
The whole simulation analysis was examined and revealed that angulasaponin C interacted stably with Leu233 and Asn237 during the 100 ns simulation time, which indicated the reliability of the previous docking work; moreover, the impact of hydrophobic contacts was evident, since many of them were constant in the simulation run, such as Phe604, Leu233, Trp327, Met473, and Trp435 (Figure S1).

| Human maltase-glucoamylase
Human maltase-glucoamylase (MGAM-C) C-termini is one of the GH31 enzyme subfamily whose members are promising targets for obesity and type 2 diabetes.Previous investigations of the enzyme revealed the existence of additional 21 amino acids conferring two TA B L E 6 The binding free energy contributions of the complexes.unique subsites, not identified in other GH31 members before, the +2 and + 3 subsites in the MGAM-C catalytic site turned out to prefer longer substrates (Ren et al., 2011).MGAM-C binding pocket was illustrated as four subsites as manifested from the standard acarbose binding.Acarbose showed favorable interactions through nine ionic interactions with amino acids Arg1510, Arg1156, Asp1157, Lys1460, Asp1526, His1584, and Asp1279 within a distance between (1.77 and 2.46).Similarly, it shared the same polar residue contacts as with the highest scoring compound angulasaponin C, which are Lys1460, Asp1526.Asp1526, and His1584 (Figure S11; Table 6).

Maltase
Acarbose extended in the MGAM-C groove in such a way as to interact with vital residues through polar and hydrophobic contacts in each of the four critical subsites from −1 to +3.The imino nitrogen located between −1 and +1 subsites in the catalytic region.His1584 and Asp1279 formed hydrogen bonds with C3-OH and C4-OH of the acarbose cyclitol ring, and several hydrophobic contacts assisted its stabilization as Trp1523, Tyr1251, and Trp1418.In the +1 subsite, the dideoxy amino sugar ring interacted with Asp1157 and with Arg1510 nitrogen atom to form three hydrogen bonds.While Asp1526 interacted with the imino nitrogen of acarbose, π-π stacking was detected between Phe 1559and Trp1355 and acarviosine rings.In the +2 subsite, Trp1369 stacked with the third ring in acarbose and in the +3 subsite, nonpolar interactions rendered the fourth ring of acarbose more stable through Pro1159 and Phe1560 (Figure S12; Table 5).The −1-subsite groove is constrained by the hydrophobic patch Trp327, Tyr1251, Phe1559, and Phe1560, which render the 1,6-glucoside substrates more auspicious as in angulasaponin C, adzukisaponin IV and VI, the highest scoring compounds in our selected molecular skeleton.
In the angulasaponin C, the triterpenoidal moiety acted as an effective linker with adequate length to ensure binding of the 1,6-diglucoside hydroxyl groups to Lys1164, Tyr1162, Glu1138, Gln1533, and Asp1532 and the 1,2-linked di-maltose hydroxy-and methyl hydroxy moieties to Asp1526, His1584, Asp1279, and Trp1369 (Figure S12).It is worth mentioning that Trp1369 served to constrain the flexibility of the 1,6-linked sugars to guide the catalytic process.MGAM-C favors up to four glucose units for its catalytic activity, and this isoform is the most active of all the maltase enzymes; thus, its inhibition is a promising target for significant improvements in diabetic patients.−11.173, −10.285, and −11.335Kcal/mol, respectively.The negative binding scores denoted favorable interactions with the key amino acids.GAA is a hydrolase assuring the breakage of both α-1,4-and α-1,6-glycosidic linkages within the lysosomes (Table 4).5nn8 is a recombinant GAA protein with particular specificity directed to the hydrolysis of α-1,4-glucosidic linkage 32-fold than the α-1,6-link (Roig-Zamboni et al., 2017).The structure of the enzyme was previously illustrated as a typical GH311 hydrolase protein with major similarity to the maltase-glucoamylase (MGAM) and the sucrase-isomaltase (SI), which are the core starch-digesting enzymes within the intestinal border.While the maltose is formed of two glucose units joined by an α 1-4 bond, isomaltose is two glucose units linked through an α 1-6 bond.Catalytic reactions held in family GH311 enzymes retain the configuration of the anomeric proton and are of the Koshland double displacement reaction type, with Asp518and Asp616 playing the nucleophilic role and acid/base catalysis.
The binding subsites +2 and +3 were envisaged for their stabilizing role in binding of the tetrasaccharide ligands compared to the disaccharide counterparts, although their exact amino acid pattern was not clear.The reversed orientation of adzukisaponin IV, such that the single glucose unit fits into the +1 subsite, largely impacted its binding stability and interactions, which further confirmed that the α-1,6-attached diglucoside in the +1 and −1 subsites was essential to manifest the lowest fitting score (Figure S13).
Adzukisaponins bound with a terminal nonreducing end.While acarbose showed seven hydrophilic interactions in the α-rhGA A groove with Asp282 (1.93, 2.33, 2.11), Asp616 (2.23,1.71),and Asp404 (1.95), it was close enough to Phe525 to extend a hydrophobic contact at 2.11A°.The hydrophilic nature of the ligands appears to contribute to a better binding score, which is manifested form inspecting angulasaponin A (−7 Kcal/mol) and adzukisaponin I (−7.05Kcal/mol) whose skeletons encompass only one disaccharide unit, instead of two, in the better binding compounds.Moreover, the larger size of adzukisaponins, due to the triterpenoidal link, helped to provide a better orientation and interaction distances with amino acids than acarbose.Adzukisaponin IV and VI interacted with the same polar residues as acarbose, namely, Asp404, asp616, and Asp282; additionally, their better binding scores could be attributed to their larger number of ionic interactions with Arg600 and the salt bridge formation with Arg411 and Lys470, respectively.The hydrophobic contacts with Trp481 and Val480 might have contributed to a stable binding mode than acarbose (Figure S13).On the other hand, angulasaponins A and D recorded lower number of hydrogen bond interactions and lacked binding to Asp404 and Asp616, respectively, which largely compromised their Glide XP scores to be less than acarbose.Angulasaponins B and C were superior to acarbose, despite lacking polar interactions with key amino acids as Asp404 and Asp282, but this was compensated by their hydrogen bonds to Asp518 and Arg600 as well as the salt bridge formed between Lys479 and angulasaponin B or between Pro482 and angulasaponin C.
Adzukisaponin IV complex with α-glucosidase revealed constant interactions with Trp376, Arg411, and Trp481 during the simulation study.The trajectory analysis of both the apoα-glucosidase enzyme and its bounded form manifested homogeneity in their behavior through the 100 ns simulation time, as seen in the plateau formed upon investigating the RMSD within a range between 1.5 and 2A.Similarly, the RMSF values of the sampled amino acids showed little fluctuations with or without a bounded ligand, which further confirmed the stability of the α -glucosidase-adzukisaponin IV complex and its promising potential as anti-obesity molecule (Figures S5 and S6).

| α-Amylases
In order to assess the binding mode, energy, and interaction of the selected ligands inside the pancreaticα-amylase enzyme (4 W93), molecular docking was performed and validated with MD simulations and free energy calculations.Three core amino acids featured the conserved catalytic residues in α-amylase, Glu 233, Asp 197, and Asp 300 (Williams et al., 2015).Upon inspecting the 2D and 3D interactions of adzukisaponins inside the binding groove, these residues were detected forming clear ionic hydrogen bond interactions both directly and mediated by water molecules intrinsic to the protein pocket (Guan et al., 2022).
The most stable compound with the best fitting energy score −15.599Kcal/mol was adzukisaponin VI, assuming five hydrogen bond interactions with Glu240 (2.03, 1.92), Asp353 (2.69), Asp356 (1.70), and Asp300 (1.70).The highest Glide XP scores were achieved by adzukisaponin VI, angulasaponin B, and angulasaponin C whose structures possessed four sugar units attached as two maltose units in positions 3 and 29 of the tritepenoidal skeleton (Table 3).While the hydrophobic bed was effectively filled with the triterpenoidal skeleton, which revealed close contact to Trp357,Val354,Tyr151,Leu162,Ile235,Gly306,Thr163,Leu237,Trp58,and Trp59, the polar side pockets were clearly interacting with the disaccharide units of adzukisaponin VI, with an emphasis on the importance of binding distances, and the number of hydrogen bonds, polar and nonpolar contacts (Figure S14).Angulasaponin B displayed obvious hydrogen bonds with the conserved pair of amino acids, Asp197 (1.64) and Glu233 (2.62), impeded within hydrophobic patches formed of Trp269 and Ala307, Glu272, which strengthen their rewarding energy; moreover, a salt bridge was formed between the carboxylic group in ring G and Lys261 (3.02) Table 2.The native ligand montbretin conferred a better fitting to the enzyme core through its π-π-stacking interactions with Tyr62 and His290.A salt bridge was noted between Lys200 and the carboxylate group of the galacturonic acid unit, which contributed to the better binding conferred by adzukisaponin VI, even better than the standard drug acarbose (−13.082Kcal/mol) (Figures S15   and S16).The pancreatic α-amylase-binding site is characterized by key aspartate amino acids Asp300, Asp353, and Asp356 and the ππ-stacking interaction with residues Tyr62 and His290.The docking procedures were validated by redocking the montbretin A into the active site with an RMSD value of 0.0221 A° (Figure 2).The selected saponin compounds were bound in the active site of α-amylase assuming the same direction and orientation of the co-crystallized ligand montbretin A.
Upon conducting extensive analysis using molecular dynamics (MD) simulation for 100 ns, the adzukisaponin VI-amylase complex manifested stability of its binding interactions throughout most of the simulation time (Figure S7).This was evident by investigating the Rg analysis for both the free α-amylase and its bounded state trajectories.The RMSF analysis, which measured the displacement degree of atoms compared to reference structures, again confirmed the stability of adzukisaponin VI complex in the range 23-23.2Å than the angulasaponin B complex whose loop region fluctuated till 23.7 A at 60 ns.The RMSD measured the conformational fluctuation of a protein during the simulation from its initial position to its final conformation.The trajectory A similar trend was observed in Rg analysis as the Rg of adzukisaponin VI remained in the range of 23-23.2Å throughout the simulation, while the Rg of angulasaponin B showed stability till 20 ns at 23-23.3 Å and then gradually increased to 23.75 Å at 60 ns.It again attained stability toward the end of simulation in the range of 23.3 Å (Figure sidered dietary staple in many cultures.A growing number of studies revealed their role as anticancer, anti-obesity, and antidiabetic agents(Kris-Etherton et al., 2002;Sreerama et al., 2012).A human TA B L E 2 Different interactions of the selected compounds in α-amylase protein.
this was largely due to more hydrogen bonds formed in the −1 subsite, interactions with the hydrophobic patches as Phe479,Leu233,   Tyr242, Trp327, Trp568, Phe604, Ile356, Trp470, Ile392, Met473,   and Trp435.Kotanalol results and interactions were slightly better than angulasaponin C, probably due to the formation of about nine ionic hydrogen bonds encompassing two salt bridges with Asp472 and Asp571 and π-cation interaction with Trp327.In contrast to the human MGAM (nt MGAM) with its directed preference to the short-chain 1,4-oligosaccharides, the N-terminal of SI (Pdb ID: 3lpp) favored 1,6-linkage diglucosides, despite their 3-bond attachment and notable flexibility.Adzukisaponin VI bound to the hydrophobic +1 subsite in such a way as to constrain its 1,6-linked oligosaccharides through the −1 narrow groove and the hydrophobic adjacent interactions.1,6-linked malto-oligosaccharides fitted in the −1 subsite and 1,2-linked and /or 1,4-linked oligosaccharides fitted in the +1 subsite.Angulasaponin C 1, 2 linkages fitted in the +1 subsite and the 1,6-linkage fitted in the −1 subsite.Low scoring compounds lacked the 1,6-linked nonreducing dimaltose.Failure to meet these protein features dramatically declined the GlideXscore and reduced fitting, thus the prospected bioactivity as realized in adzukisaponin I, II, and III and angulasaponin A. Only the 1,6-linked diglucoside was Upon inspection of the −1-subsite binding with angulasaponin C, His1584 and Asp1529 formed two H bonds with the OH groups of ring I.While the +2-subsite encompassed polar interaction between Trp1369 and the OH group in ring H, the +3-subsite showed two hydrogen bonds between Phe1560 and both the pyranosyl oxygen and OH group.Angulasaponin C superseded acarbose through Lys1164 interaction with the methyl hydroxy group of the 1,2-linked diglucosides and Asp1532 with double hydroxy groups of the terminal ring.Clearly, these ionic interactions stabilized angulasaponin C to have a score of −13.103Kcal/mol.Angulasaponin B encountered the same orientation in the binding subsites −1, +1, and +3 and was able to extend hydrogen bonds in the distal subsite to Lys1164 and Asp1532, yet it lacked the hydrogen bond formed with Trp1369 in the +2-subsite, justifying its lower binding score.Angulasaponin VI and IV marginally outperformed acarbose in their binding stability.Adzukisaponin IV resulted in a lower binding score due to the single glucoside unit attachment in position 3.Interactions were summarized as follows: in the distal subsite, hydrogen bonds were formed between Lys1164 and COOH group in ring A, between Tyr1162 and the methyl hydroxy group in ring F, and between Asp1532 and the two hydroxy groups in ring F. Whereas the −1 subsite was marked by His1584 ionic interaction with OH group of ring H, the +1 subsite showed Asp1526 interacting with OH group of ring F, and the +3 subsite encompassing aromatic hydrogen bond formed between Phe1560 and the ester linking group.Adzukisaponin VI ring G methyl OH group extended hydrogen bond to Asp1526 in the −1 subsite.In the same vein, Arg1510 and Asp1157 interacted with the OH group of ring G.In the front edge of the site, Ser1452 and Arg1156 interacted with the COOH and OH groups of ring H, respectively, and Phe1247 formed ionic interaction with the methyl OH group of ring I.These interactions permitted enough fitting for adzukisaponin VI in the enzyme pocket with a score of −12.551Kcal/mol since the formation of hydrogen bonds within the region of hydrophobic contacts is of significant impact on good binding and fitting scores.Upon conducting the molecular dynamics (MD) study of the angulasaponin C complex with maltase enzyme, the consistency of all the ionic hydrogen bond interactions seen was evident as in the docking protocol throughout the whole simulation time, namely with Arg1156, Tyr1251, Trp1355, Trp1369, Phe1427, Arg1510, and Thr1586.The stable RMSD values revolved between 2 and 2.25 A° upon binding of the ligand, indicating the conformational uniformity of molecular interactions and their contributions to the favorable binding free energy (FiguresS3 and S4).

4. 3
| α-Glucosidases α-glucosidases are of notable importance in obesity and type 2 diabetes mellitus where they act to release α-glucose subunits from di-, oligo-, or aryl glucosides.The native ligand acarbose was used to validate the docking protocol in Maestro 11.0 where the co-crystallized ligand was removed and redocked with an RMSD value of 1.3756 A°.Many of the screened saponins outperformed acarbose, such as adzukisaponin IV, adzukisaponin VI, angulasaponin B, and angulasaponin C, whose Glide XP scores were −11.585, Reported biological assays on adzuki extracts.
TA B L E 1 Different interactions of the selected compounds in the sucrase protein.
furthermore, the quaternary nitrogen linker extended a salt bridge to Asp571 seemingly imparting more stability to acarbose fitting in the binding site (Table 5).While most of the tested saponin compounds interacted with Asp231 and Lys509 except adzukisaponin II, only the highest scoring ones, adzukisaponin VI and angulasaponin TA B L E 5