New Alpha-Amylase Inhibitory Metabolites from Pericarps of Garcinia mangostana

Two new benzophenones: garcimangophenones A (6) and B (7) and five formerly reported metabolites were purified from the pericarps EtOAc fraction of Garcinia mangostana ((GM) Clusiaceae). Their structures were characterized by various spectral techniques and by comparing with the literature. The α-amylase inhibitory (AAI) potential of the isolated metabolites was assessed. Compounds 7 and 6 had significant AAI activity (IC50 9.3 and 12.2 µM, respectively) compared with acarbose (IC50 6.4 µM, reference α-amylase inhibitor). On the other hand, 5 had a moderate activity. Additionally, their activity towards the α-amylase was assessed utilizing docking studies and molecular dynamics (MD) simulations. The docking and predictive binding energy estimations were accomplished using reported crystal structure of the α-amylase (PDB ID: 5TD4). Compounds 7 and 6 possessed highly negative docking scores of −11.3 and −8.2 kcal/mol, when complexed with 5TD4, respectively while acarbose had a docking score of −16.1 kcal/mol, when complexed with 5TD4. By using molecular dynamics simulations, the compounds stability in the complexes with the α-amylase was analyzed, and it was found to be stable over the course of 50 ns. The results suggested that the benzophenone derivative 7 may be potential α-amylase inhibitors. However, further investigations to support these findings are required.


Introduction
Diabetes is a major severe health problem worldwide. Its treatment without any side effects is still a great challenge. It is characterized by chronic hyperglycemia with carbohydrate, protein, and lipid metabolic disturbances due to defects in insulin action and/or

Plant Material
The fruits were obtained from a Saudi local market in March 2019. Its verification was proven by Dr. Emad Al-Sharif (Faculty of Science and Arts, King Abdulaziz University) and a voucher (no. GM1424) specimen was maintained in the Faculty of Pharmacy`s herbarium, KAU.

Plant Material
The fruits were obtained from a Saudi local market in March 2019. Its verification was proven by Dr. Emad Al-Sharif (Faculty of Science and Arts, King Abdulaziz University) and a voucher (no. GM1424) specimen was maintained in the Faculty of Pharmacy's herbarium, KAU.

α-Amylase Inhibition
The new benzophenone derivatives (6 and 7) and the known xanthone (5) were evaluated for their AAI potential at concentrations of 5, 10, 20, and 40 µM [2][3][4]35,36]. The method is based on the assay of α-amylase using EnzChek ® Ultra-Amylase Assay Kit (E33651) (Thermo-Fisher Scientific Inc., Waltham, MA, USA). The provided stock solution of dye quenching (DQ TM ) starch and porcine pancreatic α-amylase enzyme (Sigma-Aldrich, Hamburg, Germany) were diluted with the reaction buffer (pH 6.9) according to the reported protocol [36]. To the microplate wells, the tested compound (10 µL) in DMSO, diluted enzyme (50 µL), and 40 µL of the reaction buffer were added and allowed to stand at room temperature for 5 min, then DQ TM starch (100 µL) was added. The fluorescence intensity of the digestion products from the DQ TM starch (with or without compounds) was measured using a Tecan Genios microplate reader at λ max 485 ± 10 nm starting from zero min to 60 min at 10 min intervals. All experiments were performed in triplicate. Acarbose was utilized as positive control. The IC 50 values were calculated by linear regression analysis [2][3][4]35,36].

Preparation of PDB Structures
The human pancreatic α-amylase (5TD4) was selected for docking studies. It was reported that the molecular models for the human and porcine pancreatic α-amylases are extremely similar [37][38][39]. The structure of the target, 5TD4, was downloaded from the Protein Data Bank and prepared and optimized using the "Protein preparation wizard" tool of the Schrödinger suite [40]. Some modifications were performed for optimization such as breaking the bonds to metals, adding zero-order bonds between metals and close atoms, and correcting charges. LigPrep was also used for preparing ligands in which a pH of 7.0 ± 2.0 was used for the generation of cofactors and metals het states. In addition, PROPKA was used for enhancing hydrogen bonds at pH 7, water molecules more than 3 Å from HET groups were extracted, and OSPL4 field was used for restrained minimization.

Receptor Grids Generation and Docking
Grid generation and the docking of ligands were both performed using Glide [41]. The grid was generated by using the crystal structure (PDB ID: 5TD4). For the PDB 5TD4, the binding region was established by selecting the native inhibitor T163. The nonpolar atoms were set for Van der Waals radii scaling factor to 1, and the partial charge cut-off was 0.25. The docking of ligands was performed using the "ligand docking" tool of Schrödinger suite. The protocol was extra precision (XP) and all the other settings were kept in their default form.

MD Simulation of Compound 7 and Acarbose in Complex with 5TD4
The Schrödinger suite [42,43] was used to run the molecular dynamic simulation. The TIP3P solvent was selected, and an orthorhombic shape box was chosen. A side distance box was set to 10 Å, and Na + ions were added for the system neutralization. The molecular dynamic calculations were continued for 50 ns per trajectory, and the number of atoms, pressure, and temperature were kept constant throughout the simulation.

Purification of Compounds
The MeOH extract was suspended in H 2 O and partitioned between n-hexane and EtOAc. The EtOAc fraction was successively treated utilizing Sephadex, SiO 2, and RP-18 columns to provide two new (6 and 7) and five known metabolites (1-5) (Figure 1). These metabolites were identified based on the spectral analyses and comparison with literature.

Structural Assignment of Compounds 6 and 7
Compound 6 was separated as a light-yellow powder and provided an FeCl 3 positive test, revealing its phenolic nature. Its HRESIMS demonstrated a pseudo-molecular peak at m/z 279.0515 [M+H] + (calcd for C 13 H 11 O 7 , 279.0505), consistent with a molecular formula C 13 H 10 O 7 . This formula required nine degrees of unsaturation. The UV revealed bands at 206, 296, and 318 nm, suggesting 6 to have a benzophenone skeleton [44]. Characteristic bands at 3317, 1623, and 1601 and 1589 cm −1 for chelated OH, carbonyl, and aromatic C=C functionalities, respectively were observed in the IR spectrum [45]. The 13 C and HSQC spectra showed 13 carbon resonances, comprising four aromatic methines and nine quaternary carbons, including one carbonyl at δ C 200.6 (C-7) and six for oxygen-linked carbons (Table 1). In the 1 H NMR of 6, the appearance of an aromatic signal at δ H 5.83 Life 2022, 12, 384 6 of 16 (2H, brs, H-3, 5), relating with the carbon at δ C 95.8 (C-3, 5) in the HSQC suggested the presence of a symmetrically substituted phloroglucinol ring (ring A) in 6 [45]. This was confirmed by the HMBC-cross-peaks of H-3/C-2, C-1, C-4, and C-5 and H-5/C-1, C-3, C-4, and C-6. Moreover, the 1 H NMR signals at δ H 6.52 (d, J = 2.4 Hz, H-2') and 6.36 (d, J = 2.4 Hz, H-4') were attributed to two m-coupled aromatic protons. They had HSQCcross-peaks to the carbons at δ C 107.4 (C-2') and 112.1 (C-4'). This was consistent with the existence of a hydroxyquinol moiety (1,2,3,5-tetrasubstituted phenyl, ring B). In the HMBC, the cross-peaks of H-2'/C-4' and C-6' and H-4'/C-2', C-5', and C-6' assured this moiety ( Figure 2). The link of rings A and B via the carbonyl group to provide the benzophenone core was established based on HMBC correlations from H-3 and H-5/C-7 and H-2'/C-7 and C-1. Based on the fore-mentioned evidences, the structure of 6 was elucidated and named garcimangophenone A.

α-Amylase Inhibitory Activity
Benzophenones are known to have diverse bioactivities such as antioxidant, antibacterial, cytotoxic, antitrypanosomal, and leishmanicidal activities [51][52][53][54]. However, there are limited reports regarding the AAI potential of Garcinia's benzophenones. Therefore, the AAI potential of the new benzophenone derivatives (6 and 7), along with compound 5 was assessed. The results showed that compounds 7 and 6 demonstrated noticeable AAI potential (IC 50 9.3 and 12.2 µM, respectively), compared with acarbose (IC 50 6.4 µM). These results are in good agreement with the previous study by Akoro et al. that reported the potent AAI effect of gakolanone, a benzophenone derivative reported from G. kola [55]. On the other hand, 5 displayed moderate activity with an AAI (IC 50 16.1 µM). It is noteworthy that previous studies revealed the AAI potential of 1-4 [2][3][4].

Molecular Docking and Dynamics Studies
The hydrolysis of the glycosidic linkage in starch is catalyzed by the α-amylase enzyme [56]. The α-amylase active site consists of five major subsites that are necessary for the binding of longer substrates [57]. Three residues in the active site, Asp 300, Asp 197, and Glu 233 are important for the catalysis. The catalytic nucleophile, Asp 197, forms a covalent bond with the glucosyl in the S-1 subsite. The Glu233 is the acid-base catalyst as it protonates the leaving group and deprotonates the water. The Glu300 is also essential for positioning the catalytic water for hydrolysis [58].

Preparations of Ligands and Proteins
Converting 2D structures to 3D structures, tautomerization, and ionization were all performed using LigPrep and resulted in the generation of 121 minimized structures. The 5TD4 was prepared and optimized by the Protein Preparation Wizard. Optimization included the H-bonding and minimization of the geometry, in addition to assigning the appropriate charges and force field treatment.

Molecular Docking Studies
The Receptor Grid Generation tool of Glide in Maestro was used for defining the grid box with the prepared α-amylase and then the 3D molecular structures (native inhibitor T163, acarbose, 7, 6, and 5) were docked with the 5TD4 binding site (Figure 3) and the docking scores were reported ( Table 2). The scores are an indication of the most strongly bound ligand and binding affinities. The T163 exhibited the highest docking score, followed by acarbose, 7, 6, and 5 which were −19.081, −16.078, −11.297, −8.185, and −6.726 complexed with 5TD4, respectively. docking scores were reported ( Table 2). The scores are an indication of the most strongly bound ligand and binding affinities. The T163 exhibited the highest docking score, followed by acarbose, 7, 6, and 5 which were −19.081, −16.078, −11.297, −8.185, and −6.726 complexed with 5TD4, respectively.   The docking analysis was performed to study the possible binding interactions between the different structures and the α-amylase (PDB ID:5TD4). The analysis of the docking between T163, the native inhibitor, and 5TD4 revealed various hydrogen bonds with the amino acid residues, most important are the hydrogen bonds with Asp197 and Glh233 which are the catalytic amino acids (Figure 4).  The docking analysis was performed to study the possible binding interactions between the different structures and the α-amylase (PDB ID:5TD4). The analysis of the docking between T163, the native inhibitor, and 5TD4 revealed various hydrogen bonds with the amino acid residues, most important are the hydrogen bonds with Asp197 and Glh233 which are the catalytic amino acids (Figure 4). Acarbose complexed with 5TD4 also showed hydrogen bonds with the catalytic amino acid residues ( Figure 5).
Compounds 7 ( Figure 6) and 6 ( Figure 7) also showed similar binding interactions; however, compound 5 ( Figure 8) lacked some of the major binding interactions which decreased its activity. Acarbose complexed with 5TD4 also showed hydrogen bonds with the catalytic amino acid residues ( Figure 5).
Compounds 7 ( Figure 6) and 6 ( Figure 7) also showed similar binding interactions; however, compound 5 ( Figure 8) lacked some of the major binding interactions which decreased its activity.

Molecular Dynamic Simulation
The movement of atoms with respect to time is computed through a molecular dynamics (MDs) simulation. The dynamics of the different atoms and the conformational stability both play important roles in the functioning of the essential biological macromolecules such as receptors and proteins [59]. For this reason, different properties of the systems such as root mean square deviation (RSMD), Cα-root mean square fluctuations (RMSF), the gyration radius (Rg), and location of inter-molecular H-bonds were studied

Molecular Dynamic Simulation
The movement of atoms with respect to time is computed through a molecular dynamics (MDs) simulation. The dynamics of the different atoms and the conformational stability both play important roles in the functioning of the essential biological macromolecules such as receptors and proteins [59]. For this reason, different properties of the systems such as root mean square deviation (RSMD), Cα-root mean square fluctuations (RMSF), the gyration radius (Rg), and location of inter-molecular H-bonds were studied

Molecular Dynamic Simulation
The movement of atoms with respect to time is computed through a molecular dynamics (MDs) simulation. The dynamics of the different atoms and the conformational stability both play important roles in the functioning of the essential biological macromolecules such as receptors and proteins [59]. For this reason, different properties of the systems such as root mean square deviation (RSMD), Cα-root mean square fluctuations (RMSF), the gyration radius (Rg), and location of inter-molecular H-bonds were studied in order to Life 2022, 12, 384 12 of 16 determine the dynamics of the system. The molecular docking predicts the binding mode of ligands to receptors which is a good starting point to study the stability of interactions. The Desmond software was used to study the frequency and stability of compound 7 and acarbose complexed with the α-amylase protein (PDB ID:5TD4). Two MD simulations were run for the complexes for 50 ns, and the complexes' structures were fixed at a pH of 7.0 ± 2.0. The stability of the complexes was tested by analyzing the interaction map and the root mean square deviation plot of the ligand and the protein. Figure 9a,b represent the plots of the RMSD for the α-amylase (PDB ID: 5TD4) complexed with compound 7 and acarbose, respectively. Looking at the RMSD values, it is clear that the acarbose was stabilized with the protein at about 2 Å after 35 ns and remained stable until 50 ns. However, the plot shows that the complex with compound 7 remained stable during the whole simulation time (50 ns) in relation to the reference time (0 ns). Although the plots showed some fluctuations at the time of simulation, all fluctuations are considered non-significant, as they were within an acceptable range between 1 and 3 Å.
Life 2022, 12, x FOR PEER REVIEW 12 of 16 in order to determine the dynamics of the system. The molecular docking predicts the binding mode of ligands to receptors which is a good starting point to study the stability of interactions. The Desmond software was used to study the frequency and stability of compound 7 and acarbose complexed with the α-amylase protein (PDB ID:5TD4). Two MD simulations were run for the complexes for 50 ns, and the complexes' structures were fixed at a pH of 7.0 ± 2.0. The stability of the complexes was tested by analyzing the interaction map and the root mean square deviation plot of the ligand and the protein. Figure 9a,b represent the plots of the RMSD for the α-amylase (PDB ID: 5TD4) complexed with compound 7 and acarbose, respectively. Looking at the RMSD values, it is clear that the acarbose was stabilized with the protein at about 2 Å after 35 ns and remained stable until 50 ns. However, the plot shows that the complex with compound 7 remained stable during the whole simulation time (50 ns) in relation to the reference time (0 ns). Although the plots showed some fluctuations at the time of simulation, all fluctuations are considered non-significant, as they were within an acceptable range between 1 and 3 Å. The binding interactions between compound 7 and the residues of the α-amylase (PDB ID:5TD4) are illustrated in Figure 10 which shows several interactions with residues The binding interactions between compound 7 and the residues of the α-amylase (PDB ID:5TD4) are illustrated in Figure 10 which shows several interactions with residues Asp197 (one of the catalytic amino acids), His305, Thr163, Gln63, Asn300, Asp356, and Trp59. The interactions are mostly hydrogen bonds and water bridges with a few hydrophobic interactions. The amino acid residue Asp197 interacts with the ligand through a hydrogen bond directly or a hydrogen bond mediated by a water molecule which were retained for more than 100% of the simulation time. The amino acid residue His305 interacts with the ligand through three types of interactions: H bond, H bond mediated by a water molecule, and a hydrophobic interaction during the simulation. Gln63 forms a hydrogen bond with the ligand which is retained for 90% of the simulation time in addition to a hydrogen bond through a water bridge. Hydrophobic interaction between Trp59 and an aromatic ring in the ligand was maintained for 60% of the 50 ns simulation time. Asp197 (one of the catalytic amino acids), His305, Thr163, Gln63, Asn300, Asp356, and Trp59. The interactions are mostly hydrogen bonds and water bridges with a few hydrophobic interactions. The amino acid residue Asp197 interacts with the ligand through a hydrogen bond directly or a hydrogen bond mediated by a water molecule which were retained for more than 100% of the simulation time. The amino acid residue His305 interacts with the ligand through three types of interactions: H bond, H bond mediated by a water molecule, and a hydrophobic interaction during the simulation. Gln63 forms a hydrogen bond with the ligand which is retained for 90% of the simulation time in addition to a hydrogen bond through a water bridge. Hydrophobic interaction between Trp59 and an aromatic ring in the ligand was maintained for 60% of the 50 ns simulation time. Figure 10. The α-amylase binding interactions with compound 7 throughout the simulation. The major interactions that occurred are Hydrogen bonds, water bridges, and hydrophobic interaction.
The detailed schematic diagram of binding interactions between acarbose and the residues of the α-amylase (PDB ID:5TD4) are illustrated in Figure 11. The docked poses were controlled through the simulation time of 50 ns, i.e., molecular interactions with residues Thr163, His305, Glu240, Trp59, Gln63, Asp197, Tyr151, Leu162, and Asn300 were observed. Figure 11 shows the ligand-protein interactions classified into different types: ionic, hydrophobic, hydrogen bonds, and water bridges. The interactions are mostly hydrogen bonds, which play a significant role in ligand binding. Most of the interactions remained for more than 100% of the simulation time which indicates multiple contacts of the same subtype with the ligand. Figure 11. The α-amylase binding interactions with acarbose throughout the simulation. The major interactions that occurred are Hydrogen bonds, water bridges, and hydrophobic interaction. The detailed schematic diagram of binding interactions between acarbose and the residues of the α-amylase (PDB ID:5TD4) are illustrated in Figure 11. The docked poses were controlled through the simulation time of 50 ns, i.e., molecular interactions with residues Thr163, His305, Glu240, Trp59, Gln63, Asp197, Tyr151, Leu162, and Asn300 were observed. Figure 11 shows the ligand-protein interactions classified into different types: ionic, hydrophobic, hydrogen bonds, and water bridges. The interactions are mostly hydrogen bonds, which play a significant role in ligand binding. Most of the interactions remained for more than 100% of the simulation time which indicates multiple contacts of the same subtype with the ligand. Asp197 (one of the catalytic amino acids), His305, Thr163, Gln63, Asn300, Asp356, and Trp59. The interactions are mostly hydrogen bonds and water bridges with a few hydrophobic interactions. The amino acid residue Asp197 interacts with the ligand through a hydrogen bond directly or a hydrogen bond mediated by a water molecule which were retained for more than 100% of the simulation time. The amino acid residue His305 interacts with the ligand through three types of interactions: H bond, H bond mediated by a water molecule, and a hydrophobic interaction during the simulation. Gln63 forms a hydrogen bond with the ligand which is retained for 90% of the simulation time in addition to a hydrogen bond through a water bridge. Hydrophobic interaction between Trp59 and an aromatic ring in the ligand was maintained for 60% of the 50 ns simulation time. Figure 10. The α-amylase binding interactions with compound 7 throughout the simulation. The major interactions that occurred are Hydrogen bonds, water bridges, and hydrophobic interaction.
The detailed schematic diagram of binding interactions between acarbose and the residues of the α-amylase (PDB ID:5TD4) are illustrated in Figure 11. The docked poses were controlled through the simulation time of 50 ns, i.e., molecular interactions with residues Thr163, His305, Glu240, Trp59, Gln63, Asp197, Tyr151, Leu162, and Asn300 were observed. Figure 11 shows the ligand-protein interactions classified into different types: ionic, hydrophobic, hydrogen bonds, and water bridges. The interactions are mostly hydrogen bonds, which play a significant role in ligand binding. Most of the interactions remained for more than 100% of the simulation time which indicates multiple contacts of the same subtype with the ligand. Figure 11. The α-amylase binding interactions with acarbose throughout the simulation. The major interactions that occurred are Hydrogen bonds, water bridges, and hydrophobic interaction. Figure 11. The α-amylase binding interactions with acarbose throughout the simulation. The major interactions that occurred are Hydrogen bonds, water bridges, and hydrophobic interaction.

Conclusions
Two new benzophenones and five known metabolites were purified from the EtOAcsoluble fraction of GM pericarps. They were characterized by various spectral techniques.