Antioxidant, Anti-Inflammatory, and Antidiabetic Activities of Bioactive Compounds from the Fruits of Livistona chinensis Based on Network Pharmacology Prediction

In this study, a chemical investigation on the fruits of Livistona chinensis (FLC) led to the isolation and identification of 45 polyphenols and 5 alkaloids, including two new compounds (Livischinol (1) and Livischinine A (46)), an undescribed compound (47) and 47 known compounds. FLC was predicted with novel potential antidiabetic function by collecting and analyzing the potential targets of the ingredients. Compound 32 exhibited significant α-glucosidase inhibitory activity (IC50 = 5.71 μM) and 1, 6, and 44 showed the PTP1B inhibitory activity with IC50 values of 9.41-22.19 μM, while that of oleanolic acid was 28.58 μM. The competitive inhibitors of PTP1B (compounds 1 and 44) formed strong binding affinity, with catalytic active sites, proved by kinetic analysis, fluorescence spectra measurements, and computational simulations, and stimulated glucose uptake in the insulin-resistant HepG2 cells at the dose of 50 μM. In addition, FLC was rich in antioxidant and anti-inflammatory bioactive compounds so that they could be developed as nutraceuticals against diabetes.

Type 2 diabetes is a complex metabolic disorder associated with developing insulin resistance, impaired increased oxidative stress, inflammation, insulin signaling, abnormal glucose metabolism, and so on. The disorder leads to a consequent decrease in quality of life and an increase in the rate of mortality [18]. At present, α-glucosidase inhibitors have been used against type 2 diabetes in clinic, such as acarbose and miglitol [19][20][21][22][23]. However, the above drugs could cause some side effects such as diarrhea, flatulence, and acute hepatitis. Protein tyrosine phosphatase 1B (PTP1B), another key enzyme related to type 2 diabetes, works as a negative governor for the insulin signaling pathway by dephosphorylating both the insulin receptor and the downstream insulin receptor substrate proteins [24,25]. Consequently, new α-glucosidase inhibitors and PTP1B inhibitors are promising therapeutic agents to treat type 2 diabetes. Moreover, a large number of investigations have suggested that a diet rich in foods with antioxidant polyphenols is related to a lower risk of diabetes and predisposing factors [26].
To develop the potential edibleness of FLC, systematic phytochemical investigations were undertaken in this study. Then, FLC was predicted with novel antidiabetic function utilizing chemical profile and network pharmacology. The network pharmacology analysis also revealed that the biological process of FLC against diabetes might be involved in the regulation of inflammatory response, oxidationreduction process, glucose metabolic process, insulin signaling pathway, and so on. Therefore, the extract, the fractions, and all isolated compounds were assayed for their antioxidant, anti-inflammatory, and inhibitory effect on α-glucosidase and PTP1B. The PTP1B inhibitory mechanism and effect on glucose uptake in insulin-resistant HepG2 cells of FLC-derived polyphenolic PTP1B inhibitors were further explored.

Extraction and Isolation.
The whole dry fruits (40 kg) of Livistona chinensis R. Brown were extracted with 70% ethanol by heat reflux extraction (3 × 320 L, 2 h each time). The ethanol extract was suspended in water and successively partitioned with petroleum ether, dichloromethane, ethyl acetate, and butyl alcohol, respectively, to obtain petroleum ether extract (8 g), dichloromethane extract (26 g), ethyl acetate extract (55 g), and butyl alcohol extract (420 g). The petroleum ether extract was not further fractioned, since fatty acids are primary constituents in it [8]. The dichloromethane layer (

Network Construction and Analysis.
The related diseases were predicted by the targets of the ingredients. And the disease enrichment analysis was obtained by DAVID database (http://david.ncifcrf.gov/). The overlap targets were the candidate targets of FLC against diabetes. To cluster the biological functions and clarify the pathways that were involved in putative drug targets, the analysis of GO function and KEGG (http://www.genome.jp/kegg/mapper .htmL) signaling pathways was performed using the cluster-Profiler and ggplot2 packages in R software (version.3.3.0). GO enrichment analysis mainly consisted of biological processes. Besides, the enrichment analysis was also obtained by DAVID database. P ≤ 0:05, as the cutoff value, was calculated by the two-sided hypergeometric test method to identify enriched GO terms and the localization of the biological and molecular functions of the proteins, which indicated the relative importance of enriched GO terms and pathways. The ingredient-target network was established by Cytoscape.3.4.0 software [27].
2.5. Antioxidant Activities 2.5.1. DPPH Assay. The DPPH free radical scavenging assay was performed according to a previously reported method [28]. A series of different concentrations of the tested samples and the positive control ascorbic acid were mixed with freshly prepared DPPH (0.2 mM) in a 96-well microplate with a total volume of 150 μL. After leaving the mixture to react in the dark for 30 min, its absorbance was determined at 517 nm. All of the wells with only samples were set as blanks to obtain an absorbance value, which was subtracted from the test sample readings. The percentage inhibition (%) for each sample was calculated by the following formula: where Abs blank and Abs samples correspond to the absorbance units (at 517 nm) of DPPH solutions incubated in the absence and presence of the tested samples.

ABTS Assay.
The ABTS radical cation (ABTS·+) was prepared by reacting ABTS with potassium persulfate following the method previously described [28]. The ABTS radical cation (ABTS·+) was prepared by mixing with 3 Oxidative Medicine and Cellular Longevity 2.45 mM potassium persulfate and equal volumes of an ABTS solution (7 mM), and then, they were permitted to react in the darkness at room temperature for 12-16 h. Thereafter, the ABTS reagent was diluted with methyl alcohol until reaching an absorbance of 0:70 ± 0:02 units at 734 nm. In a 96-well microplate, ABTS reagent (190 μL in each well) and samples (10 μL in each well) were incubated at room temperature for 20 min. To determine the antioxidant capacity by the ABTS method, the experimental conditions and protocol were identical to those of the DPPH assay. The only difference was that the absorbance was measured at 734 nm.
2.6.1. MTT Assay for Cell Viability. RAW 264.7 cells were plated into a 96-well plate (2 × 10 4 cells/well) and cultured in RPMI-1640 with 10% FBS at 37°C under 5% CO 2 . After incubation for 12 h, the cells were pretreated with the samples for 4 h, which were then stimulated with LPS (1 μg/mL) for 20 h. And the control group was without the samples. Besides, 0.5 mg/mL MTT was then added to plates, which were incubated for further 4 h. Subsequently, 100 μL dimethyl sulfoxide (DMSO) was then added to each well to dissolve the crystals. The absorbance was recorded at a wavelength of 490 nm by a BioTek microplate reader.
2.6.2. Nitrite Assay. The level of NO production was tested by measuring the nitrite level in the culture medium with the Griess method [29]. Cells were seeded at a density of 2 × 10 4 cells per well in 96-well plates. After the cells were pretreated with indomethacin and the test samples for 4 h, LPS (1 μg/mL) was added to the medium for further incubation. 20 h later, nitrite production was measured by mixing 100 μL of supernatant and 100 μL of Griess reagent in a 96-well plate for 10 mins. A microplate reader was employed for the optical density measurement at 570 nm. The percentage inhibition of NO was determined as follows: where A c is the absorbance of the control and A s is the absorbance of the test sample.
2.7. α-Glucosidase Inhibitory Assay. According to the previously published method [28], the α-glucosidase inhibition was assayed using p-nitrophenyl-α-D-glucopyranoside (p-NPG) as the substrate, the release of p-nitrophenol was measured with a microplate reader at 405 nm. Acarbose was used as the positive control. The extract, fractions, and compounds were dissolved in 5% DMSO and then diluted with phosphate buffer. Briefly, 80 μL PBS, 20 μL of varying concentrations of sample solutions, and 20 μL α-glucosidase solution (1.3 U/mL) were mixed and incubated in 96-well where A a represents the absorbance of the sample group with the enzyme, A b represents the absorbance of the sample control group without enzyme, A c represents the absorbance of the control group without samples, and A d represents the absorbance of the blank control group without samples and enzyme. . The optical density of each well was measured at 405 nm using an MTP-500 microplate reader. The PTP1B inhibitory activity was calculated by formula (2).

Enzyme Inhibition Kinetic
Study. Lineweaver−Burk double reciprocal plot and Dixon plot were used to study the kinetic behavior of active compounds against PTP1B and the corresponding inhibition constants (K i values). Enzyme reactions were conducted at various concentrations of p-NPP substrate (0.3, 0.6, and 1.2 mM) with active compounds at different concentrations (0, 3.125, 6.25, and 12.5 μM for compound 1, 0, 6.25, 12.5, and 25 μM for compounds 44 and 6) and 100 nM PTP1B in 96-well plates. The plate reader recorded the absorbance of the reaction mixture every 3 mins. The enzymatic velocity of the enzyme reaction was calculated based on the time − Δabs plot [31]. The Lineweaver−Burk plot and Dixon plot were generated by GraphPad Prism 8 in order to determine the type of PTP1B inhibition and K i .

Fluorescence Measurement.
180 μL of the buffer with 10 μL of the PTP1B at the concentration used in the assays were added into the 96-well black plates followed by 10 μL different concentrations (0, 6.25, 9.37, 12.5, 18.75, 25, and 50 μM) of inhibitor. The excitation wavelength was set at 250 nm, and the fluorescence emission spectra were collected from 300-400 nm with an emission bandwidth of 2 nm [32].
The quenching parameters such as the Stern-Volmer constant (K SV ), binding constant (K A ), and the number of binding sites (n) were determined by the formula [33].
where F 0 and F are the fluorescence intensities in the absence and presence of a quencher and ½Q f is a concentration of compounds. 2.8.5. Molecular Dynamics Simulations. Molecular dynamics simulations were carried out of PTP1B apoprotein-and PTP1B-compound complexes, respectively, following the method previously described with modifications [35]. In order to obtain a stable and low-energy protein conformation, the mode was optimized by using the Desmond (v3.8) module in the Schrödinger software for molecular dynamics simulation with optimized potentials for liquid simulations (OPLS) all-atom force field 2005. 50 ns molecular dynamics simulations were carried out of PTP1B apoprotein-and PTP1B-compound complexes, respectively. The system was solvated with simple point charge (SPC) water and neutralized by adding an appropriate amount of counterions in a 10 Å×10 Å×10 Å orthorhombic box, so as to form a buffer region between protein atoms and box sides. Additionally, the OPLS_2005 force field was used to minimize the energy of the complex system, and the maximum iterations during minimization were set to 5000, and the convergence threshold was kept at 1.0 kcal/mol/Å. Hybrid methods of the steepest descent and the limited memory Broyden Fletcher Goldfarb Shanno (LBFGS) algorithms with a maximum of 5000 steps were performed to minimize the system energy until a gradient threshold of 25 kcal/mol/Å was reached. Before molecular dynamics simulations, a 10 ns simulation 5 Oxidative Medicine and Cellular Longevity was performed to relax the whole system, applying a normal pressure temperature (NPT) ensemble with a Nose-Hoover thermostat at 300 K and Martyna-Tobias-Klein barostat at 1.01325 bar pressure. At last, the molecular dynamics simulations were carried out Accepted Manuscript for 50 ns under the pressure at 1.01325 bar and the temperature at 300 K. The energy and trajectory atomic coordinate data were recorded at every 1.2 ps and 100 ps, respectively, and the obtained data were used for statistical analysis. Besides, root means square deviation (RMSD) of the protein-ligand complex was monitored during the whole simulation, and the obtained protein-ligand interaction histogram showed which amino acid residues interacted with the ligand and the proportion of each interaction.

Results and Discussion
3.1. Structure Elucidation. The chemical structures of the compounds isolated from FLC were shown in Figure 1. Compound 1 was obtained as an orange amorphous powder. Its molecular formula was determined to be C 19 H 20 O 7 on the basis of its 13 (Table 1) and HSQC spectrum displayed 19 carbon resonances assignable to two methyls (one methoxy), one methylene, eight methines (five olefinic and two oxygenated), and eight quaternary carbons (seven olefinic). As shown in Figure 2, 1 H-1 H COSY data showed correlations between H-3/H-2, and H-2/H-4, indicating that the three methine protons were connected. The above data suggested that compound 1 was a flavanol derivative [37][38][39][40]. The 1 H-1 H COSY spectrum also displayed a correlation between H-4 and H-11, and the heteronuclear multiple-bond connectivity (HMBC) spectrum displayed correlations of H-4 (δ H 2.54) with C-11 (δ C 34.1), C-12 (δ C 98.8) and H-11 (δ H 1.29 and 2.13) with C-3 (δ C 69.2), C-4 (δ C 25.6). Therefore, a C-4-C-11 connection was confirmed. There were 10 degrees of unsaturation evident in the molecule of 1, of which nine were represented by flavanol and one unsaturation was left. The chemical shift of C-5 was shifted downfield by 5.4 ppm, compared with carbon atom in (-)-epicatechin [40], indicating that the ether was incorporated between C-5 and C-12 [37][38][39]. Moreover, the HMBC correlations of C-12 with H-11, H-13, and H-14, indicated that the methyl group and methoxy group were attached to C-12. Thus, the planar structure of compound 1 was showed as Figure 2.
The relative configuration of C-2 and C-3 in 1 was concluded to be of the epicatechin type from the characteristic feature of the H-2 resonance in the 1 H-NMR spectrum: δ H 5.12 (d, J = 5:1 Hz) [39,41]. And the C-2 aryl substituent was suggested to be the equatorial orientation according to the thermodynamically favoured conformation existing in natural flavanones [42,43]. The J 2, 3 coupling constant of 5.1 Hz confirmed a 2, 3-cis configuration of 1, whereas the  Oxidative Medicine and Cellular Longevity coupling constant (J = 11:1) between H-3 and H-4 indicated a 3, 4-trans-configuration. Due to the NOE correlation between H-4 and H-14, the relative configuration was determined and they were oriented to the same direction on the ring of inner ether. The CD spectrum showed a negative cotton effect at 240 nm, indicating the existence of 3R configuration [44,45]. It was consistent with the CD data of 2,3cis-3,4-trans (2R,3R,4S) flavanol [46]. On the basis of the above discussion and published same skeleton derivatives, the structure of 1 (shown in Figure 1) was finally assigned and named Livischinol.
Compound 46 was obtained as a light yellow amorphous powder, whose molecular formula was determined to be C 10 H 13 N 5 O 2 based on its 13 90. Besides, the oxygen atom was attached to C-8 rather than C-2, which was finally deduced from the analysis of the HMBC spectroscopic data. Both methylene protons and the N 10 -H (δ H 6.55) showed HMBC correlations with C-6 (δ C 145.9) and a correlation a correlation between C-6 and δ H 8.02; thus, the oxygen atom was present at C-8. Moreover, the NOE spectroscopy experiment showed a correlation between N 9 -H (δ H 11.22) and O-H (δ H 10.19), which strongly suggested the existence of this compound as an enol rather than an 8-keto tautomer. The 1 H-1 H COSY spectrum also displayed correlations of H-10 (δ H 6.55) with H-11 (δ H 3.67 and 3.31), H-11 with H-12 (δ H 4.07) and the HMBC spectrum displayed correlations of H-11 with C-6 (δ C 145.9), C-12 (δ C 72.9). In addition, two methylene protons were at different chemical shifts due to the chiral carbon C-12. Thus, the connection of the branch was confirmed as shown in Figure 2, because ½α 20 D was -33 (c 0.23, MeOH), which was the same specific rotation with the known similar compound (S)-5-amino-pent-1-en-3-ol hydrochloride [48]. As reported, the specific rotation direction of the sidechain was the same with the side-chain attached to the purine ring [49,50]. Thus, the absolute configuration of the chiral carbon C-12 was identified as S. The molecule was named livischinine A.
Compound 47 was obtained as a light yellow amorphous powder. Its molecular formula was determined to be C 10 H 13 N 5 O 2 based on its 13   confirmed an E configuration. The spectroscopic data suggested the structure of this compound was the same with 8-hydroxy-zeatin [51]. The spectroscopic data was first described except for the mass spectra.

Potential Antidiabetic Effect Predicted by Network
Pharmacology Analysis. In our previous study, we summarized 65 compounds that had been isolated from FLC through literature research [8]. In order to build the chemical ingredient database of FLC, we collected 105 compounds in all, consisting of the reported ingredients and those which were isolated in our laboratory. Additionally, the database mainly included flavonoids, ligands, phenols, fatty acids, and alkaloids. And then, a total of 811 candidate targets of ingredients were searched from PubChem, SwissTargetPrediction, and SuperPred databases. The relationship between the number of targets and the types of compounds was displayed in Figure 3. To investigate the diseases related to the candidate targets, the targets were mapped to DAVID to systematically analyze potential related diseases. The result shown in Figure 4 illustrated that FLC mainly had a potential effect against type 2 diabetes, cancer (such as prostate cancer, lung cancer, esophageal adenocarcinoma, breast cancer, and bladder cancer), and so on. Type 2 diabetes ranked top one with the smallest P value (1:19 × 10 −84 ), the largest count (324) and the highest gene ratio (44.57%). It had been reported that the extracts of FLC showed antiproliferative activities in vitro (such as lung cancer A549, human liver cancer HepG2, and breast cancer MCF-7) and in vivo [8,10,12]. However, the anti-diabetic function of 9 Oxidative Medicine and Cellular Longevity FLC has not been reported to date. Typical habits of modern societies push individuals toward a nutritional overload lifestyle. This disturbing reality is evidenced by the exponential rise in the prevalence of type 2 diabetes. And research showed that diabetes patients had an increased risk of developing tumors, and diabetic patients with tumors had a poor prognosis and a high risk of death, because chemotherapy drugs could reduce insulin secretion, aggravate glucose metabolism disorders, and increase the incidence of acute complications such as hypoglycemia, ketoacidosis, and even death [92]. Thus, FLC possessed the potential to be developed as functional food for diabetic patients with tumors.
The therapeutic targets for type 2 diabetes were collected from Comparative Toxicogenomic Database, Therapeutic Target Database and DrugBank database. Six hundred and eighty-one overlapped targets were both related to FLC and type 2 diabetes. The overlapped targets were mapped to DAVID to systematically analyze their biological processes [27]. According to the result of GO enrichment analysis ( Figure S121), the biological process of FLC against diabetes might involve in the regulation of inflammatory response, oxidation-reduction process, protein phosphorylation, insulin secretion, glucose metabolic process, and so on. To understand the potential mechanism of FLC against diabetes, KEGG pathway enrichment analysis was applied. The results ( Figure S122) of the analysis demonstrated that the pathways of FLC against diabetes including PI3K-Akt, MAPK, RAS, insulin, insulin resistance, TNF, FoxO, and cancer-related signaling pathways. According to the analysis of the ingredient-target network, 15 key targets of FLC against diabetes identified with network analyzer were listed in Table S1. Among the key targets, PTGS1, PTGS2, and ALOX5 were closely related to inflammation. GAA was the gene encoded α-glucosidase in lysosome, and PTPN1 was the gene encoded PTP1B protein.

10
Oxidative Medicine and Cellular Longevity  3.5. α-Glucosidase Inhibitory Activity. Inhibition of α-glucosidase or PTP1B enzymes was an effective strategy to treat metabolic diseases such as diabetes. All isolated compounds and fractions were assayed in vitro for their α-glucosidase and PTP1B inhibitory activities. The results were presented in Tables 2 and 3. The results showed that the butyl alcohol fraction had the highest inhibitory effect of α-glucosidase with an IC 50  3.6. PTP1B Inhibitory Activity. The results showed that the butyl alcohol fraction had the highest inhibitory effect of PTP1B with an IC 50 value of 0.017 μg/mL. The sesquineolignans in FLC performed stronger inhibitory activity of α-glucosidase than that of PTP1B. Besides, the catechol derivatives (compounds 1, 4, 42) displayed appropriate inhibition activities against PTP1B and antioxidant. It had been reported that some catechol derivatives were successfully synthesized as PTP1B inhibitors, implying that the catechol moiety played a key role in the PTP1B inhibitory activity [94]. The compounds with catechol moiety were main flavonoids in FLC. Evidence from studies in experimental animals suggested that dietary (-)-epicatechin (compound 4) could mitigate insulin resistance through the modulation of redox-regulated mechanisms in a rat model of metabolic syndrome [95]. Compounds 1, 6, and 44 exhibited the most remarkable activity with IC 50 values ranging from 9.41 to 22.19 μM, while that of the positive control oleanolic acid was 28.58 μM. They were potential PTP1B inhibitors; therefore, they were used to explore the inhibitory mechanism.
3.6.1. Inhibitory Kinetics and Binding Affinities to PTP1B. To study the kinetic behavior of the most active compounds (compounds 1, 6, and 44) against PTP1B, Lineweaver−Burk double reciprocal plot and Dixon plot were further applied. As displayed in Figures 5(a) and 5(c), compounds 1 and 44 were identified as competitive inhibitors, the regression lines for each compound intersected at the y-axis. And compound 6 was educed as a mixed-type inhibitor for the reason that increasing concentration of inhibitor results in the family of lines shared a common intercept on the left of the y-axis and above the x-axis. Based on the results of Dixon plots ( Figures 5(b), 5(d), and 5(f)), the inhibition constant (K i ) values (Table 4) of compounds 1, 44, and 6 were 1.49, 2.80, and 4.18 μM, respectively.
The PTP1B enzyme has a large number of intrinsic fluorescence residues consisting of 19 phenylalanines, 13 tyrosines, and 18 tryptophans. Therefore, the intrinsic   13 Oxidative Medicine and Cellular Longevity fluorescence might be changed by the function of inhibitor affinity [96]. Then, we investigated the interactions between PTP1B and the potential inhibitors (compounds 1, 6, and 44) that had an obvious difference in inhibitory potencies at different concentrations. There was no significant emission from any assay mixture (including the inhibitors) under our measurement conditions. Figure 6 showed the fluorescence emission spectra of PTP1B at different concentrations (0, 6.25, 9.37, 12.5, 18.75, 25, and 50 μM) of the potential inhibitors. The dose-dependent lowering of the fluorescence intensity was observed on the increase of the concentrations of inhibitors. Importantly, the decreasing tendency of fluorescence quenching was highly correlated with inhibitory potencies (IC 50 ). The binding affinity constant (K SV ) was calculated using Stern-Volmer equation (Table 4). The K SV values were ranked in order of inhibitory potencies as follows: compound 1 (IC 50  14 Oxidative Medicine and Cellular Longevity binding energy of PTP1B (PDB ID: 1AAX) and the optimized binding pose were displayed in Figure 7 through molecular docking study, respectively. The 3D and 2D binding mode as shown in Figures 7(a) and 7(d) of compound 1 in the best site of PTP1B with the smallest binding energy (-4.7 kcal/mol). Compound 1 formed two hydrogen bonds, including the 3-OH with residues Asp48 and the phenolic hydroxyl group on the B ring with residues Met258. And the π-π and π-alkyl hydrophobic interactions were connected with residues Tyr46, Ala217, and Val49. The analysis also indicated that compound 1 was surrounded by Arg47, Phe182, Ile219, Gly259, and Gln262. These suggested that compound 1 was located in the secondary phosphatebinding active site, and it could affect the recognition and binding affinity of the substrate [97]. Figures 7(b) and 7(e) showed the pose of compound 44 in PTP1B with the binding energy (-6.1 kcal/mol), surrounded by Lys120, Ile219, and Gln62. Compound 44 formed hydrogen bonds with residues Ser216, Ala217, Arg221, and Gln266. It also formed π-σ interaction with residues Ala217, π-π interactions with Tyr46 and Phe182, and π-alkyl interaction with residues Val49. The result directed that compound 44 bonded to the catalytic site (Arg221) and then blocked the dephosphorylation of the substrate [98]. The pose (Figure 7(c) and 7(f)) of compound 6 molecule in the PTP1B crystal structure with the smallest binding energy (-6.2 kcal/mol), which was surrounded by Asp48, Ser216, and Gly218. It formed π-σ inter-action with residues Ala217, π-π interactions with Tyr46 and Phe182, and π-alkyl interaction with residues Val49 and Ile219. Based on the molecular docking analysis, the potential PTP1B inhibitors were all located in the active   To predict the stability and flexibility of the binding modes, the protein of PTP1B-and PTP1Bcompound complexes were submitted to molecular dynamics simulations, respectively. Additionally, RMSD could reflect the changes of protein structural conformation and suggest whether the molecular dynamics simulations had reached the equilibrium. As shown in Figure 8, during the whole  TYR_20  ARG_24  ASP_29  LYS_36  TYR_46  ARG_47  ASP_48  VAL_49  SER_50  GLU_115  LYS_116  LYS_120  ASP_181  PHE_182  SER_215  SER_216  ALA_217  GLY_218  ILE_219  ARG_221  ARG_257  MET_258  GLY_259 TYR_46  ARG_47  ASP_48  VAL_49  GLN_85  GLU_115  LYS_116  LYS_120  ASP_181  PHE_182  SER_215  SER_216  ALA_217  GLY_218  ILE_219  GLY_220  ARG_221  SER_222  MET_258  GLN_262  THR_263  GLN_266 H-bonds Hydrophobic (c) The above results indicated that the PTP1B-compound 1 complex and PTP1B-compound 44 complex were more stable than PTP1B-compound 6 complex after the systems reached equilibrium, which indicated that compound 1 and compound 44 had a better binding affinity. And the RMSD value also suggested that compound 1 might deviate molecular docking position. The histograms (Figure 9) showed the type and proportion of interactions between receptor and ligand. In Figure 9(a), compound 1 formed strong hydrogen bonds with catalytic active site (Asp181), which accounted for 152% of the simulation time. It also formed hydrogen bonds and water bridges with residues Asp48, Lys120, and catalytic active residue Gln262. Meanwhile, residues of Tyr46, Ala217, and Ile219 formed hydrophobic contacts with compound 1, accounting for 73%, 38%, and 35% of the simulation time, respectively. In conclusion, hydrogen bonds and hydrophobic contacts were the main interactions between PTP1B and compound 1. In Figure 9(b), compound 44 formed hydrogen bonds with catalytic active sites Arg221 and Gln262, which accounted for 280% and 220% of the simulation time, respectively. In addition, compound 44 formed hydrophobic contacts with residues of Phe182, Val49, Ala217, and Tyr46, accounting for 90%, 60%, 54%, and 51% of the simulation time, respectively. As shown in Figure 9(c), compound 6 formed hydrophobic contacts with residues Tyr46 and Val49, accounting for 90% and 19% of the simulation time, respectively. At the same time, it also bounded to residues Lys120 and Arg221 with hydrogen bonds.
The results obtained from the molecular dynamics simulations indicated that compounds 1 and 44 formed better bind affinity, especially with catalytic active sites (Asp181, Arg221, and Gln262), compared with compound 6. It also explained that compounds 1 and 44 were potential competitive inhibitors of PTP1B with better biological activity, which was consistent with kinetic mode and made up the shortage of molecular docking analysis [94]. The results above all further deepen the understanding of the inhibitory mechanism of potential competitive PTP1B inhibitors.
3.7. Glucose Uptake Stimulations. Encouraged by the result of the PTP1B inhibition assay, compounds 1, 44, and 6 were tested for their effects on glucose uptake in insulin-resistant HepG2 cells. All tested compounds showed no cytotoxicity against HepG2 cells in the concentration range from 12.5 to 50 μM. At the dose of 50 μM, compounds 1 and 44 increased glucose consumption by 48.9% and 42.8%, respectively, which were similar to that of the positive control of rosiglitazone (46.7%) in an insulin resistance model of HepG2 cells as shown in Figure 10. Thus, compounds 1 and 44 were also potential insulin sensitizers. However, compound 6 displayed little effect on glucose uptake.

Conclusions
In conclusion, we studied the novel potential antidiabetic function predicted by chemical profile and network pharmacology. This provided a new strategy to identify the potential pharmacological activity. The bioactive compounds responsible for the antioxidant, anti-inflammatory, and antidiabetic activities were identified. In addition, the competitive inhibitors of PTP1B (compounds 1 and 44) formed strong binding affinity with catalytic active sites and they were also potential insulin sensitizers. With the attained results, it could be speculated that FLC could be developed as nutraceuticals for the prevention and treatment of diabetes.

Data Availability
The data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
The authors declare no conflict of interest, financial or otherwise.