A high-resolution α-glucosidase inhibition profiling for targeted identification of natural antidiabetic products from Lycopodiella cernua (L.) Pic. Serm and their inhibitory mechanism study

Abstract The targeted identification of α-glucosidase inhibitors from the crude ethyl acetate of Lycopodiella cernua (L.) Pic. Serm (L.cernua) was guided by high-resolution inhibition profiling. The α-glucosidase inhibition profiling and HPLC-QTOF-MS showed tannins and serratenes were the corresponding antidiabetic constituents. Two new serratenes named 3β, 21β-dihydroxyserra-14-en-24-oic acid-3β-(4'-methoxy-5'-hydroxybenzoate) (4), 3β, 21α-dihydroxyserra-14-en-24-oic acid-3β-(4'-methoxy-5'-hydroxybenzoate) (7), together with two known compounds (5 and 6) were isolated. Their structures were elucidated by HR-ESI-MS and NMR. Compounds 5–7 inhibited the α-glucosidase activity in a non-competitive manner with Ki values ranging from 1.29 to 12.9 µM. The molecular docking result unveiled that 4–7 bound to the residues at the channel site, which enabled to block the substrate access. In addition, the molecular dynamics (MD) simulation of the most active compound 7 and α-glucosidase indicated the 4′-methoxy-5′-hydroxybenzoate group formed the stable hydrogen bonds and pi-pi T-shaped interactions with Arg312, Gln350 and Phe300 residues, while the rings D and E were stabilized by hydrophobic interaction.


Introduction
a-Glucosidase is a membrane-bound enzyme found in the brush border of the small intestines responsible for the digestion of dietary carbohydrates (Van de Laar et al. 2005).Inhibition of a-glucosidase is one of the existing therapeutic strategies for lowering postprandial plasma glucose level (Chehade and Mooradian 2000).The clinically used a-glucosidase inhibitors, including acarbose, voglibose, and miglitol, present some side effects including flatulence, abdominal discomfort, and bloating owing to the unexpected accumulation of maltose in the intestinal system (Joshi et al. 2015;Patel 2016).New a-glucosidase inhibitors with fewer side effects are, therefore, of urgent need.Medicinal plants are well known as a rich source of new chemical entities in drug discovery for novel a-glucosidase inhibitors (Zhao et al. 2019;Atanasov et al. 2021).Natural products based drugs constituted approximately 40% of all small-molecule drugs approved over the past three decades (Newman and Cragg 2016).Indeed, some bioactive constituents derived from traditionally used medicinal plants, like terpenoids, flavonoids, coumarins, alkaloids and phenolic acid have the potential to act as an important source of a-glucosidase inhibitors (Liu et al. 2016;Proenc¸a et al. 2017;Chu et al. 2019;Suthiphasilp et al. 2020).In our recent phytochemical research on L.cernua (Liu et al. 2022), some new and previous identified serratenes from methanol extract have showed potential inhibitory effects towards a-glucosidase in vitro, which encouraged us to continuously investigate the pharmacological potential of L. cernua as an anti-diabetic medicinal plant.
In the preliminary screening of a-glucosidase inhibitory effect of crude ethyl acetate extract, it showed more promising activity than that of crude methanol extract with the IC 50 value of 4.87 ± 0.69 lg/mL (Figure S1).In order to accelerate the identification of antidiabetic principles directly from crude plant extract, the off-line hyphenation of high-performance liquid chromatography (HPLC) with bioassays was implemented to correlate the individual secondary metabolites with their a-glucosidase inhibitory abilities.This microplate-based approach enables pinpoint bioactive constituents directly from complex mixtures, thereby avoiding repetitive evaluation of biological fractions.In addition, on the basis of the high-resolution pharmacological profile, subsequent separations were guided to specifically target those constituents, which was in favour to decrease the workload of isolation.This technology has been successfully employed for accelerated identification of antidiabetic drug candidates, i.e. a-glucosidase inhibitors (Liang et al. 2021;Petersen et al. 2022), a-amylase inhibitors (Okutan et al. 2014), aldose reductase inhibitors (Tahtah et al. 2015), PTP1B inhibitors as well as free radical scavengers (Zhao et al. 2019;Pedersen et al. 2020).We report herein, with the aid of high-resolution a-glucosidase inhibitory biochromatogram, the pharmacological profiling and chemical identification of bioactive serratene triterpenoids in crude ethyl acetate extract of L. cernua for the first time.Enzyme kinetics, molecular docking and molecular dynamics (MD) of isolates were discussed to reveal the possible inhibitory mechanism.

High-resolution inhibition profiling
As our continuous research for natural oriented a-glucosidase inhibitors from L. cernua, the crude ethyl acetate extract was, therefore, initially tested for a-glucosidase at a concentration of 100 lg/mL (Liu et al. 2022), and surprisingly displayed approximately 100% inhibition.The IC 50 values were further determined as 4.87 ± 0.69 lg/mL using a concentration-dependent inhibition assays (Figure S1).This showed that crude ethyl acetate extract of L. cernua.might contain the potential a-glucosidase inhibitors.With the aim of exploring bioactive individuals responsible for a-glucosidase inhibitory effect in a fast and efficient way, 20 lL crude ethyl acetate extract of L. cernua (45 mg/ mL) was subjected to generate high-resolution a-glucosidase profile.Thus, 900 lg of raw ethyl acetate extract were separated in 47 min, and the eluent was collected into two 96-well microplates from 12 to 47 min (resolution ¼ 5.0 wells/min).A high-resolution inhibition profiling (absorbance values vs. the corresponding retention times) was plotted beneath the HPLC chromatogram at 254 nm (Figure S2).Obviously, seven chromatographic peaks (1-7) were correlated with responses in a-glucosidase inhibitory profile.It was also noticeable that eluent between 10-30 min led to approximately 100% inhibition.This broad hump is typically attributed to tannin-rich fractions, which is concerned as false positive result due to their non-specific ability to precipitate proteins (Kongstad et al. 2014).
Compound 4 was obtained as a white powder.-, 469.3318) corresponded to the loss of 4 0methoxy-5 0 -hydroxybenzoate unit.In order to unambiguous identification of compound 4, its isolation from peak 4 for enough material to perform the 1 D and 2 D NMR was attempted by preoperative-scale and analytical-scale HPLC.This was unsuccessful due to the low content and only 1 H and 13 C NMR spectra were acquired.Detailed analysis of the 1 H and 13 C NMR spectra indicated that its structure was very similar to that of 7, but only differed in the relative configuration of the hydroxyl group at C-21.Compare to that of 7, the 21-OH at compound 4 was deduced to be b-orientation by the large coupling constant and upfield-shifted proton resonance at  ).The HMBC correlation between a methoxy resonance (d 3.92, s, 3-OMe) and C-4 0 confirmed the location of the methoxy group at C-4 0 (Figure S5.6).The additional hydroxyl unit was attached to C-5 0 on the basis of HMBC correlations of H-3 0 to C-1 0 (d C 166.2, C), together with H-7 0 to C-1 0 .HMBC correlations of H-3 and C-1 0 indicated the attachment of 4 0 -methoxy-5 0hydroxy benzene group as an ester substituent to C-3.The large coupling constant of H-3 (d H 4.75, dd, J ¼ 12.2, 4.2 Hz) was contributed to an a-oriented configuration, which was additionally confirmed by the distinctive NOESY relationships from H-3 to H-5 and 23-Me.As for the relative configuration of the carbonyl group at C-4, it was determined by the presence of NOESY correlations from H-5 to H-23.Thus, the chemical structure of 7 was very similar to that of 3b, 21b, 29-trihydroxyserrat-14-en-24-oic acid 3b-(4 0 -methoxy-5 0 -hydroxybenzoate), only differing in the groups at C-21 and C-29 (Liu et al. 2022) S6.Therefore, the chemical structure of 7 was established to be a new compound, 3b, 21a-dihydroxyserra-14-en-24-oic acid-3b-(4 0 -methoxy-5 0 -hydroxybenzoate).

a-Glucosidase inhibitory activity
Based on the results obtained from the biochromatogram, the IC 50 values of compound 5-7 were determined to verify their a-glucosidase inhibitory activity.The IC 50 value of 4 was unsuccessful acquired due to the low amount as mentioned in section 2.2.Compound 7 was the most active inhibitor towards a-glucosidase with the IC 50 value of 9.74 ± 0.99 lM, more promising than 5 (23.15 ± 1.37 lM) and 6 (22.70 ± 1.36 lM).Moreover, all of these compounds showed the more considerable inhibitory effect than that of positive control acrbose (IC 50 ¼ 610.30 ± 2.60 lM).Doseresponse curve was given in Figure S1.Their significant activity confirmed the important role of the carboxylic acid moiety at C-24 for enhancing the inhibitory effect of serratenes, which was in accordance with our previous report (Liu et al. 2022).Since IC 50 value of 4 was not determined, another serratene identified from L. cernua in our previous study, 3b, 21b, 29-trihydroxyserrat-14-en-24-oic acid 3b-(4 0 -methoxy-5 0hydroxybenzoate) with IC 50 value of 38.30 ± 1.00 lM, was selected whose chemical structures mainly differ in groups at C-21 and C-22.The comparison showed that the introduction of methyl group at C-29 and a-oriented hydroxyl group at C-21 was favourable for suppressing a-glucosidase activity (Liu et al. 2022).

Kinetics of inhibition
In order to investigate the inhibition type, Lineweaver-Burk plots of 1/rate versus 1/[substrate] were constructed.As displayed in Figure S7, all the data lines intersected in one point on the x-coordinate.Thus, with the increased concentrations of 5-7, the value of Km for the substrate remained unchanged, while the value of Vmax decreased, indicating that compounds 5-7 induced a non-competitive type of inhibition (Zheng et al. 2020).It means they may prefer to noncompetitively form the complex of enzyme-substrate-inhibitor to interrupt enzyme-substrate intermediate (Wang et al. 2019).According to Eqs. ( 2) and ( 3), the values of Ki were calculated to be 12.9 lM (5), 5.19 lM (6) and 1.29 lM (7), respectively.The varied Ki values may attribute to the interaction with different non-competitive sites and/or the different affinity.To the best of our knowledge, this is the first time to investigate kinetics mode of serratenes towards a-glucosidase.While, some structural similar pentacyclic triterpenes, such as oleanolic acid, ursolic acid, were reported as non-competitive inhibitors, which was supportive of our results partially (Ding et al. 2018).The secondary plots of slope vs.
[I] were linearly fitted, indicating they have a single inhibition site or a single class of inhibition sites on a-glucosidase (Dong et al. 2021).

Molecular docking
The top 50 low-energy conformers of compounds 4-7 were selected for further interactions analysis with protein as shown in Figure S8, Tables S1 and S2.The conformers analysis of the complex showed that they mainly bound to amino acids Phe157, His239, Asn241, Ala278, His279, Glu304, ValL305, Pro309, Arg312 and Asp408 in the entrance to the active site pocket of a-glucosidase through hydrogen bond and hydrophobic interaction(Figure S9) (Yamamoto et al. 2010).Apparently, it was difficult for the oligosaccharides to pass through the narrow entry channel to access to the catalytic site, thereby decreasing the glucose level.The amino acid residues involved in the interactions between 4, 6 and 7 and a-glucosidase, including amino acids His279, Pro309, Asn241, Arg312 and His239, were almost the same, suggesting that they probably bound to the enzyme in the identical site.Moreover, there was no significance, more evenly distributed, to form hydrogen bonds between the amino acids and 4, 6 and 7. Compare to that, the hydrogen bond between 5 and Glu304 displayed a high probability, which may be attributed to the absence of benzoate ring substituted at the C3 position of 5.

Molecular dynamics (MD)
The docking pose of the a-glucosidase-7 complex with lowest binding energy in molecular docking was selected for molecular dynamics.The overall root mean square deviation (RMSD) was used to investigate the stability of the system of a-glucosidase and 7.As shown in Figure S8, the RMSD value stayed well below 2.5 Ð in the MD simulation, and the system basically reached an equilibrium state.Dynamic root mean square fluctuation (RMSF) of protein showed that all amino acids fluctuated less 2.5 Ð during the simulation process.
The trajectories in the simulation process were extracted at intervals of 5 ns, and a total of 13 frames were obtained.After superimposing them, it was found that the binding position of 7 and the protein changed slightly with time.During the simulation, the 4 0 -methoxy-5 0 -hydroxybenzoate unit gradually approached Phe300, Arg312 and Gln350 residues, while the terminal part kept the binding position stable around Phe231, His239 and Lys233 (Figure S10).The 4 0 -methoxy group, 5 0 -hydroxyl group and 21-hydroxyl group formed stable hydrogen bonds with the residues Arg312, Gln350 and Lys233, the hydrogen bond distance map between above residues and 7 also proved the above changes (Figure S11).In addition, there was a pi-pi T-shaped interaction between the benzene ring and Phe300.The rings D and E, due to the lack of polar groups, were mainly stabilized by hydrophobic interaction with Phe231 and His239.According to the free energy calculation, the binding energy of the system was about À120 kJ/mol, indicating a strong binding interaction between 7 and a-glucosidase.As shown in Figure S12, vdW interaction in vacuum played a dominant role in the stability of complex, while the contribution of electrostatic interaction in vacuum and non-polar solvation energy in the system were weak, indicating that the binding of 7 and receptor were driven and facilitated primarily by vdW interaction (Bronowska 2011).The energy decomposition showed that residues Phe231 (-5.3819 kJ/mol), Glu276 (-2.2000 kJ/mol), His279 (-3.3688 kJ/mol), Gly280 (-2.6077 kJ/ mol), Phe300 (-2.4804 kJ/mol), Glu304 (-4.1502 kJ/mol), Pro309 (-2.7360 kJ/mol), Phe310 (-3.9634 kJ/mol), ASP408 (-3.6136 kJ/mol), mainly provided the binding region with large energy in a-glucosidase (Figure S13).Overall, the molecular simulation analysis showed that the 7 inhibited the a-glucosidase by binding to the residues of the channel and blocking the substrate to enter the catalytic centre, which was consistent with the experimental result.

Plant material and extraction
Whole plants of L.cernua were collected in December 2019 from Dehong, Yunnan, China (97 90 0 04 00 E, 24 27 0 10 00 N).Botanical identification was performed by Prof. Chu chu.A reference specimen (No.LC201912) was deposited at the College of Pharmacy, Zhejiang University of Technology.
Forty-five gram of dried ground whole plant of L.cernua was extracted with 900 mL of ethyl acetate under ultrosonication for 2 h and yielded 1.46 g of crude extract.The dark viscous crude extract was resuspended in methanol and H 2 O (v/v, 9:1) and extracted with petroleum ether (v/v, 4:3) to defeat the crude extract.The methanol fraction was dried at 40 C using a Buchi evaporator (Buchi, Switzerland).andstored in the freezer until use.Crude extract sample for bioassays as well as for HPLC-HRMS analysis was prepared in methanol at concentrations of 20 mg/mL and centrifuged for 3 min at 10,000 rpm before use.

a-Glucosidase inhibition assay
The a-glucosidase inhibitory activities were determined according to a previously reported procedure (Wubshet et al. 2019).0.1 M phosphate buffer consisting of 0.066 M Na 2 HPO 4 and 0.034 M NaH 2 PO 4 ・2H 2 O added 0.02% NaN 3 and adjusted to pH 7.5 with 1.0 M NaOH solution.Ten lL of DMSO stock solution of analytes, 90 mL of phosphate buffer, and 80 mL of a-glucosidase solution were added to each well.The samples were shaken for 2 min and then incubated at 28 C for 10 min and 20 lL substrate solution (10 mM p-NPG) were added to initiate the reaction.The total volume in each well was 200 lL, resulting in a final concentration of 5% DMSO, 0.01 U of a-glucosidase enzyme and 1 mM p-NPG per well.The absorbance of p-nitrophenol at 405 nm was measured every 30 s for 35 min using a Multiskan FC Microplate Photometer (Thermo Scientific, Waltham, MA, USA), controlled by SkanIt version 2.5.1 software.The maximum enzyme activity measured without the addition of the inhibitor was used as a blank control.The percentage enzyme inhibition was calculated using the following equation: All measurements were performed in triplicate, and the results are presented as means ± SD.IC 50 values were determined by non-linear regression using GraphPad Prism software (GraphPad Prism 7.0, GraphPad Software, Inc.).
The elutes from 12-47 min were fractionated into two 96-well microplates, resulting in a resolution of 5.0 points per min.The microplates were evaporated to dryness using an SPD121P Savant SpeedVac concentrator (Thermo Scientific) equipped with an OFP400 oil-free pump and a RVT400 refrigerated vapour trap.The a-glucosidase inhibitory activities of the fractions were determined according to the previously reported procedures.Cleavage rates were plotted at their respective retention times as a-glucosidase inhibition profiling underneath the HPLC.

Isolation and structure determination of active compounds
The separation of L.cernua extracts was performed by preparative-scale HPLC with an Agilent 1100 series instrument (Santa Clara, CA, USA) consisting of a quaternary pump, a degasser, a thermostatted column compartment, a photodiode-array detector, and a highperformance autosampler, all controlled by Agilent ChemStation version B.01.01 software and equipped with a reversed-phase Luna C18(2) column (Phenomenex, 250 Â 21.2 mm, 5 lm, 100 A) maintained at room temperature.The flow rate (with solvents as described above) was maintained at 20 mL/min.The gradient elution profile was according to section 2.4 described above.900 lL of a 40mg/mL solution of crude extract was injected eight times and peaks were collected manually.The fractions were evaporated to dryness under reduced pressure at 40 C on a Buchi evaporator.The isolated compounds were analysed by the above-mentioned 1200 Agilent chromatograph hyphenated with a Bruker micrOTOFQ II mass spectrometer equipped with an electrospray ionization (ESI) interface and controlled by Bruker Hystar software version 3.2 (Bruker Daltonik, Bremen, Germany).NMR experiments were recorded with a Bruker Avance III instrument (1H resonance frequency of 600.13 MHz) at 300 K, using CDCl 3 as solvent. 1 H and 13 C chemical shifts were referenced to the residual solvent signal at d7.26 and 77.16 ppm.NMR data processing was performed using Bruker Topspin version 3.2 software.

Enzyme kinetics
The inhibition type was determined graphically according to the Lineweaver-Burk plots, with different concentrations of substrate and inhibitors.Enzyme kinetics were evaluated in triplicate, using the standard assay conditions described in section 2.3.The non-competitive inhibition mechanism was presented by the following equations: Secondary plots were constructed from: where v is the enzyme reaction rate in the absence and presence of inhibitors.Ki is the inhibition constant for binding with enzyme-substrate complex.Km is Michaelis À Menten constant.a is the apparent coefficient.
[I] is the concentration of inhibitor and [S] for substrate (Ding et al. 2018).

Molecular docking
Since the crystalline structure of a-glucosidase from Saccharomyces cerevisiae (maltase, EC 3.2.1.20)was not available in the Protein Data Bank (PDB).Homology modelling of a-glucosidase was applied to predict its possible 3 D-structure.The amino acid sequence of a-glucosidase (P53341) was obtained from UniProt protein resource data bank (http://www.uniprot.org/uniprot/P53341.html).The crystal structure of isomaltase from S. cerevisiae (PDB ID: 3AJ7) (73% sequence identity) was selected as the template for modelling by SWISS MODEL (Yamamoto et al. 2010).(https://swissmodel.expasy.org/).AutoDock 4.2 program was used to perform the docking studies for compounds 4-7 (Morris et al. 2009).

Molecular dynamics
The docking pose of the a-glucosidase/7 complex with the lowest energy of binding in molecular docking was selected for molecular dynamics.MD simulations of proteinligand were performed using the GROMACS 2018 program (Abraham et al. 2015).The topology and coordinate files of 7 were generated by calculating the BCC charge using AmberTools.The topology information for the protein was obtained in the Amber99SB force field using GROMACS.A rectangular box was constructed into which the complex was placed with PIP3P water and Na þ cations and the system was subjected to energy minimization to relax the structure of the complex.The equilibration of NVT and NPT was carried out sequentially, and each equilibration time was 100 ps.
After the two equilibration phases, the system was well-equilibrated at the desired temperature and pressure.The constraints were released and the production simulation was performed for 60 ns for data collection.The RMSD of the system and RMSF of the protein were calculated.Furthermore, the intermolecular hydrogen bonds of protein-ligand were analysed.The last 15 ns of the simulated trajectory was selected for the free energy calculation using molecular mechanics Poisson-Boltzmann surface area (MM-PBSA) method (Kumari et al. 2014).
(Yan et al. 2009ion of the hydroxyl group at C-21 was established according to the featured coupling constant (J ¼ 2.2 Hz) between H-20 and H-21, as well as the key ROESY correlation from H-21 to 29-Me (FigureS5.7)(Yanetal. 2009).