Structure related α-glucosidase inhibitory activity and molecular docking analyses of phenolic compounds from Paeonia suffruticosa

High postprandial hyperglycaemia is an important determinant of the development and progression of type 2 diabetes. Thus, inhibition of key digestive enzymes such as α-amylase and α-glucosidase is considered an efficient approach to control glycaemic levels in diabetics. In search of α-amylase and α-glucosidase inhibitors, the root bark of Paeonia suffruticosa was screened for inhibitors, resulting in the isolation of eleven phenolic compounds (1–11). Their enzymes inhibitory activities and inhibition mechanism were investigated using an in vitro inhibition assay and molecular docking studies. Compounds 2, 5, 6, and 8–11 (IC50 between 290 and 431 µM) inhibited α-glucosidase more effectively than the reference compound acarbose (IC50 = 1463.0 ± 29.5 µM). However, the compounds (IC50 > 800 µM) were less effective against α-amylase than acarbose (IC50 = 16.6 ± 0.9 µM). Among them, compound 10 exhibited the highest α-glucosidase inhibitory effect with an IC50 value of 290.4 ± 9.6 µM. Compounds 2, 5, 9 10 and 11 were found to be competitive inhibitors, while compounds 6 and 8 were noncompetitive inhibitors of α-glucosidase. Computational analyses showed that the main binding forces between the compounds and the main residues were hydrogen bonds. The results suggest that these compounds have the potential to be developed as α-glucosidase inhibitors.


Introduction
Diabetes mellitus (DM) is a metabolic disorder characterised by excessive increases in plasma glucose levels and abnormalities in lipid and protein metabolism caused by deficient insulin secretion, insulin resistance, or both in combination over time [1]. Changes in human behaviour and lifestyle have led to a substantial increase in the prevalence of diabetes worldwide over the past century. In 2014, approximately 422 million individuals were reported by the World Health Organization to have diabetes worldwide, with this figure projected to increase to over 650 million by 2040. [2]. One of the most effective approaches to treat DM is to suppress postprandial hyperglycaemia by inhibiting key digestive enzymes such as α-amylase and αglucosidase. Therefore, commercial inhibitors such as voglibose, miglitol, and acarbose were developed for the treatment of hyperglycaemia and DM [3]. These oral drugs were efficient in inhibiting both α-amylase and α-glucosidase, which are responsible for the hydrolysis of polysaccharides to oligosaccharides and disaccharides and then to monosaccharides, subsequently lowering plasma glucose levels [4]. However, the use of these drugs (especially acarbose) is not without drawbacks, the most notable of which are diarrhoea and flatulence as a result of the prolonged suppression of starch hydrolysis [5]. The adverse effects (gastrointestinal complications) of the drugs are associated with excessive α-amylase inhibition leading to accumulation of undigested polysaccharides in the colon [6]. These limitations have highlighted the need for novel drug discovery techniques aimed at improving affinity and specificity while reducing existing adverse effects. Therefore, the focus in treating diabetes and managing its associated problems is shifting to widely available drugs with few side effects [7]. Medicinal plant extracts and their chemical constituents are gaining importance as potential therapies for diabetes and its sequelae because of their different modes of action and safety. Secondary metabolites of medicinal plants with pharmacological activity, including phenolic chemicals and flavonoids, are considered as potential sources of efficient and safe hypoglycaemic agents [8].
Paeonia suffruticosa (Paeoniaceae) is a medicinal plant indigenous to China with a long history of use in Traditional Chinese Medicine and has become an important ornamental plant worldwide [9]. Traditionally, the root of P. suffruticosa has been utilised as a crude medicine for the treatment of extravagant blood, elimination of stagnant blood, and cardiovascular complications [10]. The biological activities of the plant are mainly attributed to monoterpene glycosides, such as paeoniflorin, benzoylpaeoniflorin, albiflorin, and paeoniflorigenone, and the plant is also rich in galloylglucoses, gallic acid derivatives, flavonoids, triterpenoids, and acetophenones [11]. Although the crude extract of the plant is frequently employed in antidiabetic Chinese herbal formulations, scientific studies on its antidiabetic effects are limited [12]. A comprehensive study of its bioactive constituents against key digestive enzymes responsible for the hydrolysis of carbohydrates is still lacking. Hence, the root bark of P. suffruticosa was investigated for α-glucosidase inhibitors, resulting in the isolation of eleven compounds (1-11) (Fig. 1). The structurally related compounds were screened for their structure-activity association and interaction with α-glucosidase. Although the phenolic compounds have been studied for their α-amylase and α-glucosidase activities [13][14][15], their mode of interaction with α-glucosidase has rarely been fully explored, with the exception of gallic acid (1) [16]. Therefore, in the present study, kinetic analysis and molecular docking were conducted to understand the mechanism of interactions between phenolic compounds and α-glucosidase. This article presents the separation, characterisation, enzyme inhibitory effect and inhibition mechanisms of phenolic compounds from the root bark of P. suffruticosa.

Plant material
The roots bark of P. suffruticosa was collected from China's Anhui province and was purchased from traders. Samples were authenticated by Prof. Sheng-Zehn Yang, Herbarium Curator, Department of Forestry, National Pingtung University of Science and Technology. A voucher specimen (No. BT360) was deposited at the herbarium of the Department of Biological Science and Technology.

α-Amylase inhibition assay
The inhibition of α-amylase by the tested phenolic compounds was carried out using the method of Okutan et al. [28] with minor changes. In a 96-well plate, 5 µL of the compounds solution (0-1000 µM, dissolved in DMSO), 160 µL of phosphate buffer (0.1 M, pH 6.0) with 0.02% sodium azide, and 10 µL of α-amylase (1 U/mL, final concentration) were mixed and incubated for 5 min at 37°C. Then 10 µL of 2-chloro-4-nitrophenyl-α-D-maltotrioside (CNP-G3) solution (25 mM, in phosphate buffer) was introduced into the reaction and incubated at 37°C for 30 min. Thereafter, 5 µL NaOH (5 M in distilled water) was added to the reaction mixture and the absorbance was recorded at 405 nm using a microplate reader. Acarbose (0-50 µM in DMSO) was used as reference compound. The amount of DMSO (2.5%) did not affect the enzyme activity [28]. The inhibition effect of the compounds was calculated using the following formula.
where A sample and A control represent the absorbance of enzyme activity with and without the samples or standard, respectively.

α-Glucosidase inhibition assay
The α-glucosidase inhibitory activity of the tested compounds was measured following the methods of [29] and [30] with slight modifications. Briefly, 10 µL of α-glucosidase, 5 µL of the compounds solution and 170 µL of phosphate buffer (0.2 M, pH 6.8) were mixed and incubated at 37°C for 5 min. After incubation, the reaction was initiated with the addition of 10 µL of pNPG solution into the reaction mixture and incubated for 60 min at 37°C. After incubation, the reaction was stopped by adding 5 µL of NaOH and the absorbance was measured at 405 nm using a microplate reader. The enzyme (1 U/mL) and substrate (25 mM) stock solutions were prepared in phosphate buffer and NaOH was dissolved in distilled water, while the compounds (0-500 µM) and acarbose (0-1500 µM) were dissolved in DMSO. The amount of DMSO (2.5%) did not interfere with the experiment [26]. The percentage of inhibition was calculated using equation 1.

Mode of inhibition against α-glucosidase
The same procedure as the enzyme inhibition assay was used to analyse the inhibition mechanisms and the reaction mixture Kinetic parameters were determined using Lineweaver-Burk plots and described as follows [31,32].
Competitive type: Non-competitive and mixed type: Secondary plots were determined as follows Here K i and K is indicate the equilibrium constant of the inhibitor to the enzyme and the enzyme-substrate composite, respectively. K m represent the Michaelis-Menten constant, v denotes enzyme velocity, V max represents maximal velocity, [I] and [S] represent the concentration of the compounds and pNPG, respectively. K m was the negative reciprocal of the x-axis intercept while V max was the reciprocal of the y-axis intercept obtained from the double reciprocal plots of enzyme velocity versus pNPG with increasing concentrations of the compounds. K i was the additive inverse of the x-axis intercept derived from the secondary plots of the slope of the double reciprocal plots against the concentration of the compounds, whereas K is was the additive inverse of the x-axis intercept obtained from the y-axis intercept of the double reciprocal plots versus the concentrations of the compounds.

Molecular docking
The interactions between α-glucosidase and phenolic compounds were studied by computer simulations. A SWISS-MODEL was used to identify the protein for homology modelling from the Protein Data Bank (PDB) (http://www. rcsb.org/pdb) with high sequence identity to the target (αglucosidase from Saccharomyces cerevisiae used in the experimental procedures). The crystallographic structure of the S. cerevisiae glucosidase enzyme has not yet been published. Therefore, S. cerevisiae isomaltase (PDB code: 3A4A; resolution: 1.60 Å) was selected as the receptor model for simulation, which was 72% identical and had 85% similarity to the target [33]. Ligands and water were removed from the enzyme to create a stable receptor for the phenolic compounds and docking calculations (binding energies) were carried out using the default setting for Discovery Studio 3.0 and the docking protocol was validated through redocking of co-crystallised ligand in protein.
The angle of the grid box was 90 points (x, y, and z) with a spacing of 0.5 Å, and the grid box location was set at 11.9, −16.3, and 15.5 Å (x, y, and z), with 10 runs per molecule and up to 10 conformers per ligand. Binding events were visually analysed using Discovery Studio 3.0 software and geometry minimisation was performed using a CDOCKER (CHARMm-based DOCKER). ChemDraw Pro 5.0 software was used to create the three-dimensional structures of the compounds and standard acarbose. Blind docking was conducted on all the compounds and the conformation with the highest docking number and lowest binding energy was selected as the best results for analysis. The hydrogen bonding, pi-pi stacking and hydrophobic interactions generated between the phenolic compounds and the major residues in the active site of α-glucosidase were obtained from the docking results.

Statistical analysis
SPSS version 25 was used for all statistical analyses. Compounds were statistically compared using one-way analysis of variance (ANOVA), and differences between means were assessed using Tukey's HSD test, with p values < 0.05 considered significant. Each figure reflects three separate experiments, and results are reported as mean ± standard error of the mean (SEM). The variable slope nonlinear regression method was used to determine IC 50 values (GraphPad Prism 5.0.1, GraphPad Software, San Diego, California USA).

Alpha-Glucosidase and alpha-amylase inhibitory activity
All the isolated compounds were investigated for their in vitro α-glucosidase and α-amylase inhibition activities. As shown in Table 1, all tested phenolic compounds (IC 50 values between 290 and 431 µM) were more efficient in inhibiting α-glucosidase compared to the reference compound acarbose (IC 50 value 1463.0 ± 29.5 µM), except for benzoic acid (1) and 4-methoxybenzoic acid (3), which were not active at a concentration of 500 µM. Among the tested compounds, gallic acid (10) (IC 50 value 290.4 ± 9.6 µM) was the most potent α-glucosidase inhibitor, while isoacetovanillon (6) (IC 50 value 431.3 ± 11.7 µM) showed the least inhibitory effect. Gallic acid (10) contained three hydroxyl groups positioned at C-3, C-4 and C-5 and one carboxylic acid group connected to C-1. Substitution of the carboxylic acid group with an ester group resulted in a decline in the inhibitory effect of methyl gallate (11) compared to that of gallic acid (10) (Fig. 2). When the data of gallic acid (10) was compared with that of 4-hydroxybenzoic acid (2), it was observed that hydroxylation of the compound contributes to the effectiveness of the compound in suppressing α-glucosidase. Moreover, a comparison of the inhibition data of paeonol (5) and isoacetovanillon (6) with those of 1-(2,5-dihydroxy-4-methoxyphenyl)ethanone (8) and 1-(2,3-dihydroxy-4methoxyphenyl)ethanone (9) substantiated that the greater number of hydroxyl groups on the aromatic ring was favourable for their inhibitory activity. 1-(2,5-Dihydroxy-4methoxyphenyl)ethanone (8) and 1-(2,3-dihydroxy-4-methoxyphenyl)ethanone (9) had an additional hydroxyl group and showed lower IC 50 values than paeonol (5) and isoacetovanillon (6). A similar phenomenon has been observed from the literature [40], suggesting that hydroxylation may increase the inhibitory activity of flavonoid compounds. Moreover, methoxylation at C-4 and substitution of the carboxylic acid group with an acetyl group further decreased the IC 50 value of the compounds. The position of the hydroxyl group on the benzene ring had a minor effect on the potency of the compounds, which was observed when 1-(2,5-dihydroxy-4-methoxyphenyl)ethanone (8) was compared with 1-(2,3-dihydroxy-4-methoxyphenyl)ethanone (9), as well as paeonol (5) with isoacetovanillon (6).
Moreover, all tested compounds (1-11) were evaluated for their α-amylase inhibitory activity and the results showed that the compounds (IC 50 values >800 µM) were significantly less active compared to acarbose (IC 50 = 16.6 ± 0.85 µM) ( Table 1). These compounds showed 50fold lower α-amylase inhibition activities than acarbose while their α-glucosidase inhibition activities were threefold greater than acarbose. Therefore, α-glucosidase was selected for further kinetic analysis of tested compounds. 4-Hydroxybenzoic acid (2) and gallic acid (10) were previously reported for their α-amylase inhibitory activity with  [41], which were similar to our result. The compounds showed selectivity in inhibition of αamylase and α-glucosidase. Compared with α-amylase, these compounds have stronger selective inhibition of αglucosidase (Table 1). In contrast, the standard compound acarbose showed greater ability to selectively inhibit αamylase over α-glucosidase. This results demonstrated that the structural composition of a compound can lead to large differences in its ability to selectively inhibit one enzyme over others. The reason for the selectivity of phenolic compounds in inhibiting α-glucosidase over α-amylase may be that α-amylase must incorporate larger oligosaccharide units into its active site than the small molecule inhibitors [42]. Therefore, acarbose inhibited α-amylase more effectively than α-glucosidase compared to the other compounds, which may be due to the trisaccharide moiety of acarbose that favours additional interactions. Acarbose was found to have a stronger affinity for the C-terminal than for the N-terminal subunits of the enzymes, which is due to its extended substrate binding site that enhances the interaction of C-terminal enzymes with longer oligomers [43,44].

Inhibition mechanisms of α-glucosidase
The nature of inhibition on α-glucosidase of phenolic compounds was investigated using Lineweaver-Burk plots [45]. Figure 3 shows the reciprocal plots of the maximal velocity (V max ) of α-glucosidase reaction against pNPG concentration with increasing concentrations of the compounds. The apparent values of K m (Michaelis-Menten constant) were calculated from linear curve fitting equations, while the K i (equilibrium constant for enzymeinhibitor association) and K is (equilibrium constant for enzyme-substrate-inhibitor association) were derived from secondary plots of the slopes of the primary plots (Lineweaver-Burk plots) versus the concentration of the compounds. All the double reciprocal plots of the compounds (2,5,9,10, and 11) intersected on the y-axis, indicating competitive inhibition of α-glucosidase. The values of K m increased and the values of V max remained constant as shown by the increasing slope and constant y-intercept of the curves as the concentration of the compounds increased, indicating that the compounds formed the α-glucosidase-compound composite to slow down the catalytic efficiency of the enzyme, which showed that these compounds induce competitive inhibition.
In Fig. 3, the data lines of isoacetovanillon (6) and 1-(2,5-dihydroxy-4-methoxyphenyl)ethanone (8) crossed on the horizontal axis with a constant x-intercept. Additionally, both the y-intercept and gradient of the graphs increased with the increase in the concentration of the compounds, indicating that the V max values decreased and the K m values were constant. As shown in Table 2, the equilibrium constants (K i and K is ) were the same, suggesting that the affinity of the αglucosidase-compound complex is similar to the affinity of the α-glucosidase-pNPG-compound complex. These results indicated that isoacetovanillon (6) and 1-(2,5-dihydroxy-4-   (11) and (H), Acarbose methoxyphenyl)ethanone (8) were noncompetitive inhibitors of α-glucosidase. Moreover, the secondary plots (insert of Fig. 3) of the slope against the concentration of the compounds fitted linearly, suggesting that the compounds bind to a single inhibition site on the enzyme [46].

Molecular docking analysis
Computer-assisted docking was conducted to analyse the interaction mechanisms of the compounds with α-glucosidase by visualising binding in the receptor-ligand composite    [47]. As shown in Fig. 4A and B, the compounds were located at the active binding site of α-glucosidase. The major amino acid residues involved in the interaction of the compounds and α-glucosidase were Asp233, Asn235, Ser311, Leu313, Asn317, Val319, and Lys432, and these residues were found to be crucial for the catalytic mechanism (Table  3) [46]. All compounds formed π-interactions with the amino acid residue Leu313, and paeonol (5) and methyl gallate (11) also formed π-interactions with Val319. In Fig. 4C (6) is the compound with lower affinity. The change in the position of the hydroxyl group from C2 to C3 has significant implications for the interactions of the compounds with the enzyme, as we observed that isoacetovanillon (6) forms fewer hydrogen bonds with the amino residues compared to paeonol (5), which weakens the interaction of the compound with the residues in the deep pocket of the active site [48]. This is consistent with the noncompetitive inhibition mode of isoacetovanillon (6) (Fig. 3). In addition, the C2 and C3 hydroxyl group substitution in the structure of the compound appears to have a significant impact explaining the observed difference in α-glucosidase inhibition between compounds 5 and 6 ( Table 1). Similar to compounds 5 and 6, we observed that the C3 and C5 substitution of the hydroxyl group affected the manner in which compounds 8 and 9 inhibited the enzyme ( Table 2). The change in position of the hydroxyl group altered the selectivity of the compounds on the amino acid residues with which they interact (Fig. 4).
Druglikeness/ADME/toxicity parameters Many potential therapeutics do not make it into clinical trials due to poor ADME properties and toxicity and untested tolerability. The failure of compounds in vivo or in clinical trials results in monetary and time losses. Therefore, screening compounds for drug-like, ADME and toxicity properties before further studies in vivo or clinical trials can minimise losses. In this work, the phenolic compounds were analysed for drug likeness, pharmacokinetic and toxicity properties ( Table 4). The results showed that all the evaluated compounds complied with drug likeness rules except for compounds (2, 10 and 11) while the reference compound acarbose only complied with the MDDR-like rule. These results suggest that compounds (5, 6, 8 and 9) are less likely to cause problems related to oral bioavailability, showing the potential value of the compounds in developing a compound with good drug-like properties. The projections for ADME showed that the compounds had high blood-brain barrier (BBB) penetration and moderate Caco2 cell permeability, while acarbose showed moderate permeability for both. Compounds with high penetration of the BBB may be a dynamic regulatory interface for the brain and have the potential to develop effective central nervous system drugs [49]. Moreover, the projections for the phenolic compounds showed effective intestinal absorption and low plasma protein affinity along with optimal skin permeability. Overall, the projections showed an optimal pharmacokinetic profile of the compounds, as they met all the requirements for intestinal absorption as well as BBB penetration. Toxicity analysis predicted that the compounds are mutagenic, while acarbose is not. In addition, the compounds were predicted to be carcinogenic in mice, whereas acarbose is carcinogenic in rats. In addition, the predictions showed that the compounds have a low risk of cardiotoxicity, while the cardiotoxicity risk of acarbose was not clear.

Conclusions
The results of enzymatic activities and molecular docking suggested that hydroxylation of the aromatic ring was favourable for the inhibitory effect of phenolic compounds compared to methoxylation or hydrogenation. In addition, the position of the hydroxyl group and the substitution of the carboxyl group were important in improving the inhibitory activity of the compounds. The most effective phenolic compound found was gallic acid (10) with the most hydroxyl groups and carboxyl groups. The predictions for drug similarity, ADME and toxicity showed that the compounds (5, 6, 8 and 9) were suitable for further in vivo studies as they complied with most of the rules for druglikeness, ADME and toxicity properties. This research contributes significantly to the study of direct α-glucosidase inhibitory activity by phenolic compounds and provides detailed information on their interactions with α-glucosidase and pharmacokinetics properties.

Data availability
The NMR and MS spectra of the compounds are available as supplementary material.