Phenolics Profiling of Carpobrotus edulis (L.) N.E.Br. and Insights into Molecular Dynamics of Their Significance in Type 2 Diabetes Therapy and Its Retinopathy Complication

Adverse effects associated with synthetic drugs in diabetes therapy has prompted the search for novel natural lead compounds with little or no side effects. Effects of phenolic compounds from Carpobrotus edulis on carbohydrate-metabolizing enzymes through in vitro and in silico methods were assessed. Based on the half-maximal inhibitory concentrations (IC50), the phenolic extract of the plant had significant (p < 0.05) in vitro inhibitory effect on the specific activity of alpha-amylase (0.51 mg/mL), alpha-glucosidase (0.062 mg/mL) and aldose reductase (0.75 mg/mL), compared with the reference standards (0.55, 0.72 and 7.05 mg/mL, respectively). Molecular interactions established between the 11 phenolic compounds identifiable from the HPLC chromatogram of the extract and active site residues of the enzymes revealed higher binding affinity and more structural compactness with procyanidin (−69.834 ± 6.574 kcal/mol) and 1,3-dicaffeoxyl quinic acid (−42.630 ± 4.076 kcal/mol) as potential inhibitors of alpha-amylase and alpha-glucosidase, respectively, while isorhamnetin-3-O-rutinoside (−45.398 ± 4.568 kcal/mol) and luteolin-7-O-beta-d-glucoside (−45.102 ± 4.024 kcal/mol) for aldose reductase relative to respective reference standards. Put together, the findings are suggestive of the compounds as potential constituents of C. edulis phenolic extract responsible for the significant hypoglycemic effect in vitro; hence, they could be exploited in the development of novel therapeutic agents for type-2 diabetes and its retinopathy complication.


Introduction
Diabetes, one of the leading causes of death globally, is a chronic metabolic derangement leading to high levels of glucose in systemic circulation (hyperglycaemia) due to the inability of the body to manage available glucose levels, arising from ineffective or insensitive insulin secretion by islets of Langerhans beta cells of the pancreas [1]. The management of the disease is time-consuming, during which diverse secondary complications (nephropathy, neuropathy, cardiopathy, retinopathy, etc.) culminating in death may set in, if no viable treatment/management therapy is embraced. The prevalence of diabetes continues to rise with population growth, and alongside other non-communicable diseases accounted for 74% of the world's total deaths in 2019 [2]. Uncontrolled hyperglycaemia stimulates excessive free radical generation, culminating in oxidative stress associated with most diabetic complications such as neuropathy, nephropathy and retinopathy [3].

Phenolic Content
The total phenolic content of the extract was 96.05 mg/g GAE. Based on the standards used for the HPLC analysis, 11 major phenolic compounds including sinapic acid, cacticin, hyperoside, 1,3-dicaffeoxyl quinnic acid, procyanidin, luteolin-7-O-beta-D-glucoside, rutin, epicatechin, isorhamnetin-3-O-rutinoside, chlorogenic acid and myricetin were identified from the chromatogram of the extract ( Figure 1, Table 1). It is noteworthy that the prominent peaks as observed from the chromatogram with their highest relative abundances were for chlorogenic acid followed by rutin, luteolin-7-O-beta-D-glucoside and epicatechin, suggesting they are the major identifiable phenolic constituents of C. edulis (Figure 1).

Antidiabetic Activity
The antidiabetic effect of phenolic extract of C. edulis based on its half-maximal inhibitory concentrations (IC 50 ) of 0.06 and 0.75 mg/mL revealed significantly (p < 0.05) higher inhibitory activity against alpha-glucosidase and aldose reductase, respectively, than the standards [acarbose (0.72 mg/mL) and ranirestat (7.05 mg/mL)], except for alpha-amylase, where the extract (0.51 mg/mL) favourably competed with acarbose (0.55 mg/mL) ( Table 2). Alpha-amylase and alpha-glucosidase found in the pancreas and intestine, respectively, are prominent enzymes involved in the uncontrolled or unending hydrolysis of carbohydrate to glucose. In a diabetic state, the control (with inhibitors) of these enzymes is necessary for the management of hyperglycaemia. Due to the available synthetic inhibitors such as acarbose and miglitol triggering gastro-intestinal dysfunctional events including bloating, flatulence and diarrhoea, the use of plant-based alternative therapeutic agents has been advocated [15], provided they are able to strongly (low IC 50 score) and mildly (marginally high IC 50 value) inhibit the specific activity of alpha-glucosidase and alpha-amylase [16,17], as observed in this study with C. edulis phenolic extract. Although in a related study [14], among the studied extracts (50% methanol, 70% acetone and aqueous) of the plant, only aqueous extract was tested against alpha-glucosidase, the result obtained in the current study fared comparably in terms of activity with the previously determined aqueous extract, which not only corroborates the antihyperglycemic effect of the plant but the use of polar solvent as medium of extraction or herbal preparation in indigenous medicine. Overall, the exhibited antihyperglycemic effect by the extract could be attributed to its phenolic compounds, as studies have reported phenolic compounds such as cacticin, epicatechin, luteolin glucosides, cyanidin (identified in this study) in plants such as Bergenia ciliata, Diospyros kaki and Eleusine coracana as inhibitors of alpha-amylase and alpha-glucosidase [7,18]. Aldose reductase, an enzyme implicated in the complications of diabetes, particularly retinopathy, facilitates sorbitol production from glucose through polyol pathway during the glucose catabolism [19]. Drugs such as ranirestat, sorbinil and alrestatin act by improving the reduced glucose concentration, thus facilitating eventual absorption in tissues (such as neural, lens and glomeruli) [20]. However, like many other synthetic drugs, they exhibit side effects including nausea, fever and diarrhoea, hence, the need for alternative therapeutic molecules (inhibitors) from natural products in drug design. Based on its IC 50 value with respect to modulatory effect of inhibitors on aldose reductase, the elicited activity by C. edulis phenolic extract in this study could be indicative of its prospect in the management of long-term diabetic retinopathy complication [21][22][23][24][25]. Patel et al. [26] also advocated phenolic compounds such as luteolin-7-O-β-D-glucopyranoside and 4,5-di-O-caffeoylquinic acid which are similar compounds identified in this study, as inhibitors of aldose reductase with prospect for diabetes retinopathy.

Molecular Docking and Dynamics
To gain insight into the probable interactions between the identified phenolic compounds (as revealed by the HPLC analysis) and the study enzymes in this study, computational evaluation was performed through molecular docking and MDS. Molecular docking, a measure of fitness and pose of a compound at the active site of an enzyme, normally gives high negative scores as a reflection of better pose of the compound [27]. In this study, phenolic compounds such as 1,3-dicaffeoxyl quanic acid, chlorogenic acid, epicatechin, luteolin-7-O-beta-D-glucoside, isorhamnetin-3-O-rutinoside, myricetin and rutin, had significant and better poses based on their scores than the reference standard, ranirestat, when docked with aldose reductase (Table 3). Additionally, better poses were observed for epicatechin, luteolin-7-O-beta-D-glucoside, isorhamnetin-3-O-rutinoside, rutin, hyperoside and procyanidin with alpha-amylase compared to the resulting complexes with acarbose (Table 3). While most of the identified compounds including 1,3-dicaffeoxyl quanic acid, chlorogenic acid, epicatechin, isorhamnetin-3-O-rutinoside, luteolin-7-O-beta-D-glucoside, myrcetin, rutin, cacticin, hyperoside and procyanidin showed good docking with alpha-glucosidase as depicted by the higher negative values than acarbose, other compounds (epicatechin, isorhamnetin-3-O-rutinoside, chlorogenic acid and rutin) had commendable binding at the active sites of the three enzymes (Table 3), which is indicative of their prospective interaction with the enzymes [28]. However, since docking is only a preliminary reflection of the ligand's fitness within the binding pocket of a receptor, the binding orientations of the studied phenolics were subjected to further binding energy calculations and MDS. Typically looking at the thermodynamic calculations against alpha-amylase, procyanidin among other compounds had the highest (−69.834 kcal/mol) binding energy, which was better than the value for acarbose (−54.679 kcal/mol) and rutin (−46.826 kcal/mol) (Table 4). Similarly, against alpha-glucosidase, 1,3-dicaffeoxyl quinic acid and hyperoside had higher binding energies than acarbose, while isorhamnetin-3-O-rutinoside by luteolin-7-O-beta-D-glucoside and rutin had higher binding energies than ranirestat against aldose reductase (Table 4). Higher negative values are indicative of stronger affinity of these compounds with the respective enzymes and hence possible better stability of the resulting complex [29]. While potential stronger affinities of phenolic compounds (over synthetic inhibitors) in complex with antidiabetic enzymes have been reported [30,31], Rasouli et al. [18] observed higher binding free energy of some of the phenolic compounds (rutin, cacticin and epicatechin) reported in this study against alpha-amylase as well as alpha-glucosidase. Table 3. Molecular docking scores (kcal/mol) of the phenolic compounds with carbohydrate hydrolyzing enzymes and aldose reductase.

Compounds
Aldose Reductase α-Amylase α-Glucosidase The enzyme-ligand complex is prone to conformational changes inducible by the binding ligand, and this could result in possible alteration of the biological activity of the enzyme [31]. Based on the affinity of the enzymes for the phenolics through 100 ns MDS in this study, a further probe into understanding the degree of stability and compactness of the resulting complexes was undertaken. To examine the stability while ensuring the equilibration of the enzyme-ligand, parameters including RMSD, RMSF and RoG evaluated over an extended period of 100 ns for unbound (apo-enzyme, i.e., alpha-amylase, alphaglucosidase and aldose reductase) and bound complexes (enzyme + inhibitor, i.e., standards and phenolic compounds) are presented in Figures 2-4.

Phenolic Extract Preparation and Quantification
The method of Mulaudzi et al. [14] with slight modification was adopted for the extraction of the powdered materials from the leaves of the plant (10 g) in 50% methanol (250 mL) under sonication in a cold water-containing water bath for 120 min. The extract was filtered through a Whatman No. 1 filter paper, centrifuged (3000 rpm, 15 min), and the resulting supernatant concentrated in a water bath at 40 • C. The dried extract obtained was kept airtight and refrigerated (10 • C) in a glass vial prior to use for total phenolic quantification, HPLC analysis and in vitro antidiabetic assays.
For total phenolic quantification, 50 µL (1 mg/mL) of the phenolic extract was added to 6.950 mL distilled water in a test tube, and gently shaken prior to the addition of Folin-Ciocalteau phenol reagent (0.5 mL) and sodium carbonate (1.5 mL; 20%). Subsequently, distilled water (1 mL) was added to the mixture, shaken vigorously and allowed to stand for 45 min prior to absorbance measurement at 760 nm. The standard (gallic acid) was prepared (0.8 mg/mL in 40% methanol) and its varying concentrations were treated in a similar manner as the phenolic extract [36]. The phenolic content of the extract was estimated from gallic acid standard curve and expressed as milligram per gram gallic acid equivalent (mg/g GAE).

HPLC Analysis
The HPLC analysis was achieved based on the method of Peng et al. [37] with modifications. This was conducted using HPLC (Shimadzu Prominence-i LC-2030C 3D plus, Kyoto, Japan) coupled to a diode array UV detector (HPLC-DAD) and a high-resolution mass spectrometry (HPLC-HRMS) on an Ultimate 3000 RSL Cnano system (Thermo Scientific, Waltham, Massachusetts, United States of America). The mobile phase consisted of A (0.1% formic acid) and B (acetonitrile), and the flow rate was 0.25 mL/min with the temperature of the column (Sunfire C18, 5 µm, 4.6 mm × 150 mm, Waters Corporation, Milford, Massachusetts, United States of America) set at 35 • C and the sample volume maintained at 20 µL. The elution gradient varied from 1% A to 2% B linearly for 2 min and from 2-100% B in 50 min and thereafter, from 10% to 2% for 1 min and from 2% to 0% for 9 min. The chromatogram was based on photo diode array UV detector (DAD) with wavelengths spanning 190-800 nm based on the peak absorption of the analysed compounds. The identification of the compounds was achieved based on their individual retention times and MS fragment patterns compared with those of the standard phenolics (sinapic acid, cacticin, hyperoside, 1,3-dicaffeoxyl quinic acid, procyanidin, rutin, epicatechin, isorhamnetin-3-Orutinoside, chlorogenic acid, myricetin and luteolin-7-O-beta-D-glucoside) used in tandem with published data.
2.6. In vitro Assays 2.6.1. Alpha-Amylase Inhibitory Assay Using a previously reported protocol [38], the activity of the extract against α-amylase was evaluated. Aliquots (0.1 mL) of either acarbose (reference standard) or the phenolic extract at varying concentrations (0.065-1.000 mg/mL) were added to α-amylase solution (0.1 mL of 0.5 mg/mL). Following a 10-min pre-incubation (25 • C) of the resulting solution, 1% starch solution in 0.02 M sodium phosphate buffer, was added and further incubated (25 • C, 10 min), before the reaction was halted by DNS (0.5 mL). The resulting mixtures were boiled (100 • C, 5 min) and subsequently cooled (25 • C) before final dilution (distilled water, 7.5 mL) and spectrophotometric absorbance reading (540 nm) (OPTIZEN POP, Apex Scientific, Yuseong-gu, Daejeon, Republic of Korea). The results presented as IC 50 (halfmaximal inhibitory concentration) value in each case was non-linearly extrapolated from maltose standard calibration curve.

Alpha-Glucosidase Inhibitory Assay
For this assay, 50 µL of varying concentrations (0.065-1.000 mg/mL) of either acarbose or phenolic extract were added to 0.1 mL αglucosidase (1 M) before incubation (25 • C, 10 min). Thereafter, 0.05 mL of 5 mM p-NPG solution was added followed by a further 5 min incubation (25 • C) period before 0.05 mL Na 2 CO 3 (0.1 M). Following this treatment, a microplate reader (MULTISKAN GO 1519 Thermo Scientific, Vantaa, Finland) was used to take the absorbance readings (405 nm) and the IC 50 values were similarly non-linearly determined from a p-nitrophenol standard curve [38].

Aldose Reductase Determination
In this assay, glyceraldehyde and NADPH were used as substrate and cofactor, respectively [39]. The final concentration of the dissolving solvent was kept equivalent among reaction mixtures in the presence of varying concentrations of either the phenolic extract or ranirestat (reference standard). The rates of reaction were monitored spectrophotometrically (340 nm) at 25 • C and compared with the control not containing the phenolic extract, and the IC 50 value was non-linearly determined from a calibration curve. Ranirestat was used as reference standard and the experiments were performed in triplicate.
The docking of the prepared phenolic compounds and standards into binding pockets of the enzymes (α-amylase, α-glucosidase, and aldose reductase) was by Autodock Vina Plugin on Chimera V1.14. Judging by the docking scores, complexes identified to have the best pose for each compound were ranked, selected and further analyzed through 100 ns molecular dynamics simulation (MDS).
The MDS was achieved as recently reported [28], using the GPU (force fields) version obtainable in AMBER package, where the description of the system by FF18SB variant of the AMBER force field was carried out [42]. With the aid of Restrained Electrostatic Potential (RESP) and the General Amber Force Field (GAFF) methods of the ANTECHAMBER assisted with information on atomic partial charges for the compounds. Hydrogen atoms and Na+ and Cl-counter ions (to neutralize the system) were made possible with Leap module of AMBER 18. The residues were numbered 1-336, 913, and 496, respectively, for aldose reductase, α-glucosidase and α-amylase. The system in each case was then lowered implicitly within an orthorhombic box of TIP3P water molecules such that all atoms were within 8Å of any box edge. MDS total time carried-out were 100 ns. For each simulation, hydrogens atoms were constricted using the SHAKE algorithm. The step size of each simulation was 2 fs, and an SPFP precision model was used. The simulations align with the isobaric-isothermal ensemble (NPT), having randomized seeding, Berendsen barostat maintains 1 bar constant pressure, 2 ps pressure-coupling constant, 300 K temperature and Langevin thermostat with a collision frequency of 1.0 ps [43].
Using PTRAJ, the systems were subsequently saved, and each trajectory analyzed every 1 ps, and the RoG, RMSF, and RMSD were analyzed with CPPTRAJ module (AMBER 18 suit).
Molecular Mechanics/GB Surface Area method (MM/GBSA) was adopted to assess the free binding energy while comparison of the systems binding affinity followed afterwards [44]. Binding free energy was averaged over 100,000 snapshots extracted from the 100 ns trajectory. The ∆G for each system (enzyme, complex and phenolics) was estimated as earlier reported [45].

Statistical Analysis
For the in vitro experiments, data analyses were carried out by Graph pad Prism version 3.0 using t-test (and nonparametric tests), supplemented with Mann-Whitney test. Results are expressed as mean ± standard error of the mean (SEM). Except otherwise stated, the raw data plots for the in silico evaluations were generated using the Origin data analysis software V18 (OriginLab, Northampton, MA, USA) (Seifert, 2014).

Conclusions
While the in vitro studies result gave an insight into possible antidiabetic potential of C. edulis, the HPLC analysis suggested and identified 11 phenolic compounds, which were further analysed as probable hypoglycaemic candidates through in silico studies. Based on the findings from the binding free energy, structural stability and compactness in this study, procyanidin was a better inhibitor of alpha-amylase, 1,3-dicaffeoxyl quinic acid against alpha-glucosidase while luteolin-7-O-beta-D-glucoside showed good inhibitory potentials of aldose reductase among other phenolic compounds. Thus, these molecules could be exploited in developing novel therapeutic candidates against postprandial hyperglycaemia and diabetic retinopathy.

Data Availability Statement:
The data presented in this study are available in the article.