Potential of Icariin Metabolites from Epimedium koreanum Nakai as Antidiabetic Therapeutic Agents

The therapeutic properties of Epimedium koreanum are presumed to be due to the flavonoid component icariin, which has been reported to have broad pharmacological potential and has demonstrated anti-diabetic, anti-Alzheimer’s disease, anti-tumor, and hepatoprotective activities. Considering these therapeutic properties of icariin, its deglycosylated icaritin and glycosylated flavonoids (icaeriside II, epimedin A, epimedin B, and epimedin C) were evaluated for their ability to inhibit protein tyrosine phosphatase 1B (PTP1B) and α-glucosidase. The results show that icaritin and icariside II exhibit potent inhibitory activities, with 50% inhibition concentration (IC50) values of 11.59 ± 1.39 μM and 9.94 ± 0.15 μM against PTP1B and 74.42 ± 0.01 and 106.59 ± 0.44 μM against α-glucosidase, respectively. With the exceptions of icaritin and icariside II, glycosylated flavonoids did not exhibit any inhibitory effects in the two assays. Enzyme kinetics analyses revealed that icaritin and icariside II demonstrated noncompetitive-type inhibition against PTP1B, with inhibition constant (Ki) values of 11.41 and 11.66 μM, respectively. Moreover, molecular docking analysis confirmed that icaritin and icariside II both occupy the same site as allosteric ligand. Thus, the molecular docking simulation results were in close agreement with the experimental data with respect to inhibition activity. In conclusion, deglycosylated metabolites of icariin from E. koreanum might offer therapeutic potential for the treatment of type 2 diabetes mellitus.


Introduction
Diabetes mellitus (DM) is considered to be one of the foremost challenges of the 21st century. This disease is characterized by a group of metabolic disorders of heterogeneous etiology and is augmented by hyperglycemia resulting from defects in insulin action and secretion. Chronic hyperglycemia in DM is involved in long-term damage, dysfunction, and failure of various organs, especially the eyes, kidneys, nerves, heart, and blood vessels. Diabetes mellitus is a result of insulin deficiency, which is often combined with insulin resistance [1]. DM is prevalent throughout the world, and its incidence is rapidly increasing in most countries [2]. Insulin resistance, a hallmark of type 2 DM (T2DM), results from disequilibrium in auto-phosphorylation of the insulin receptor (IR) and tyrosine kinase activity. Protein tyrosine phosphatase 1B (PTP1B) interacts with the activated IR and with insulin receptor substrate (IRS) proteins, after which it dephosphorylates tyrosine residues on the IR and IRS proteins [3]. PTP1B has been suggested to be a negative regulator of insulin metabolism

Abilities of the MeOH Extract of the Aerial Parts of E. koreanum and Its Solvent-Soluble Fractions to Inhibit PTP1B and α-Glucosidase
To determine the anti-diabetic potential of the MeOH extract of the aerial parts of E. koreanum and its soluble fractions, their abilities to inhibit PTP1B and α-glucosidase were evaluated. As shown in Table 1, the CH 2 Cl 2 fraction strongly inhibited PTP1B at the tested concentrations, with an 50% inhibition concentration (IC 50 ) value of 3.51 ± 0.05 µg/mL. For comparison, the positive control ursolic acid had an IC 50 value of 5.54 ± 0.13 µg/mL. The MeOH extract of the aerial parts of E. koreanum significantly inhibited PTP1B activity in a dose-dependent manner, with an IC 50 value of 18.83 ± 0.14 µg/mL ( Figure 1). Moreover, the EtOAc, n-BuOH, and H 2 O fractions moderately inhibited PTP1B activity, with IC 50 values of 27.09 ± 0.30, 14.44 ± 0.68, and 38.36 ± 1.29 µg/mL, respectively. The abilities of the MeOH extract and its fractions to inhibit α-glucosidase are shown in Table 1. Among the four solvent-soluble fractions, the CH 2 Cl 2 and EtOAc fractions were the most potent inhibitors of α-glucosidase, with IC 50 values of 47.69 ± 0.32 and 75.87 ± 0.38 µg/mL, respectively. For comparison, the positive control acarbose had an IC 50 value of 87.83 ± 1.08 µg/mL. In contrast, the MeOH extract and two fractions thereof (n-BuOH and H 2 O) moderately inhibited α-glucosidase activity, with IC 50 values of 120.75 ± 3.06, 138.90 ± 0.02, and 310.01 ± 1.29 µg/mL, respectively. Table 1. Abilities of the MeOH extract of Epimedium koreanum and its fractions to inhibit protein tyrosine phosphatase1B (PTP1B) and α-glucosidase.

Abilities of the Individual Compounds to Inhibit PTP1B and α-Glucosidase Activity
To identify the active compounds of E. koreanum responsible for its potent anti-diabetic activity, the abilities of the compounds ( Figure 2) to inhibit PTP1B and α-glucosidase activity were evaluated. As shown in Table 2, icaritin and icariside II were the most potent inhibitors, with IC50 values of 11.59 ± 1.39 and 9.94 ± 0.15 μM, respectively. For comparison, ursolic acid had an IC50 of 8.24 ± 0.03 μM. However, none of the other compounds inhibited PTP1B at any of the tested concentrations. Among the tested compounds, icaritin and icariside II were the strongest inhibitors of α-glucosidase, with IC50 values of 74.42 ± 0.01 and 106.59 ± 0.44 μM, respectively. For comparison, acarbose had an IC50 value of 101.16 ± 3.69 μM. In contrast, epimedin A, epimedin B, and epimedin C only moderately inhibited α-glucosidase activity, with IC50 values of 444.09 ± 3.80, 451.78 ± 5.36 and 403.87 ± 1.53 μM, respectively. However, icariin did not inhibit α-glucosidase activity at any of the tested concentrations.

Abilities of the Individual Compounds to Inhibit PTP1B and α-Glucosidase Activity
To identify the active compounds of E. koreanum responsible for its potent anti-diabetic activity, the abilities of the compounds ( Figure 2) to inhibit PTP1B and α-glucosidase activity were evaluated. As shown in Table 2

Enzyme Kinetics Analysis of PTP1B Inhibition by Selected Compounds
Since the PTP1B activity assay revealed that two of the tested compounds (icaritin and icariside II) were active, these compounds were further analyzed to establish the compound-substrate, p-nitrophenyl phosphate (pNPP) relationships for PTP1B. The inhibition type and inhibition constant (Ki) of the two compounds were determined using Lineweaver-Burk and Dixon plots. Each line of the inhibitors penetrate the same point at the x-intercept, indicating noncompetitive-type inhibition, and the x-axis represent-Ki when 1/V = 0. Therefore, icaritin and icariside II are noncompetitive-type inhibitors of PTP1B ( Figure 3). The Dixon plot is an established method for determining the enzyme inhibition type and the Ki value for an enzyme-inhibitor complex, where the value of the x-axis indicates Ki. As shown in Table 2, the Ki values of icaritin and icariside II were 11.41 and 11.66 μM, respectively.

Enzyme Kinetics Analysis of PTP1B Inhibition by Selected Compounds
Since the PTP1B activity assay revealed that two of the tested compounds (icaritin and icariside II) were active, these compounds were further analyzed to establish the compound-substrate, p-nitrophenyl phosphate (pNPP) relationships for PTP1B. The inhibition type and inhibition constant (K i ) of the two compounds were determined using Lineweaver-Burk and Dixon plots. Each line of the inhibitors penetrate the same point at the x-intercept, indicating noncompetitive-type inhibition, and the x-axis represent-K i when 1/V = 0. Therefore, icaritin and icariside II are noncompetitive-type inhibitors of PTP1B ( Figure 3). The Dixon plot is an established method for determining the enzyme inhibition type and the K i value for an enzyme-inhibitor complex, where the value of the x-axis indicates K i . As shown in Table 2, the K i values of icaritin and icariside II were 11.41 and 11.66 µM, respectively.

Molecular Docking Simulation of PTP1B Inhibition
Next, a molecular docking simulation study was performed to validate the mechanism of interaction between the two compounds and the active site of PTP1B. Specifically, 3D docking of the active compounds (icaritin and icariside II) was simulated. For this simulation, 3-({5-[(N-acetyl-3-{4-[(carboxycarbonyl)(2-carboxyphenyl)amino]-1-naphthyl}-L-alanyl)amino]pentyl}oxy)-2-naphthoic acid (positive ligand of the PTP1B catalytic model, compound 23) and 3-(3,5-dibromo-4-hydroxybenzoyl)-2-ethyl-benzofuran-6-sulfonic acid (4-sulfamoyl-phenyl)-amide (positive ligand of the allosteric model, compound 2) were considered as the standard ligands to which the orientation of the tested compounds were compared. Docking analyses were conducted to determine the extent of agreement of the structure-activity compound relationships with the obtained experimental data, as well as to evaluate binding site-directed inhibition of PTP1B. The AutoDock 4.2 program [36] uses a semi-empirical free energy force field to predict the protein-ligand binding complexes of a known structure and the binding energies for the bound state. The results of the docking simulations of the four compounds are shown in Table 3 and Figure 4. The PTP1B-icaritin complex had a binding energy of −6.24 kcal/mol and one hydrogen bond interaction with residue Asn193. The nitrogen group of Asn193 was involved in a strong hydrogen bond with the hydroxyl group of icaritin; this bond had a distance of 2.69 Å. Hydrophobic interactions were also predicted between Ser187, Ala189, Ser190, Leu192, Phe196, Glu276, Gly277, and Phe280 ( Figure 5). These hydrophobic interactions are important for strengthening the protein-ligand interaction and positioning icaritin in the active pocket to inhibit PTP1B activity. The binding energy of the PTP1B-icariside II complex was −8.77 kcal/mol, and four

Molecular Docking Simulation of PTP1B Inhibition
Next, a molecular docking simulation study was performed to validate the mechanism of interaction between the two compounds and the active site of PTP1B. Specifically, 3D docking of the active compounds (icaritin and icariside II) was simulated. For this simulation, acid (positive ligand of the PTP1B catalytic model, compound 23) and 3-(3,5-dibromo-4-hydroxybenzoyl)-2-ethyl-benzofuran-6-sulfonic acid (4-sulfamoyl-phenyl)-amide (positive ligand of the allosteric model, compound 2) were considered as the standard ligands to which the orientation of the tested compounds were compared. Docking analyses were conducted to determine the extent of agreement of the structure-activity compound relationships with the obtained experimental data, as well as to evaluate binding site-directed inhibition of PTP1B. The AutoDock 4.2 program [36] uses a semi-empirical free energy force field to predict the protein-ligand binding complexes of a known structure and the binding energies for the bound state. The results of the docking simulations of the four compounds are shown in Table 3 and Figure 4. The PTP1B-icaritin complex had a binding energy of −6.24 kcal/mol and one hydrogen bond interaction with residue Asn193. The nitrogen group of Asn193 was involved in a strong hydrogen bond with the hydroxyl group of icaritin; this bond had a distance of 2.69 Å. Hydrophobic interactions were also predicted between Ser187, Ala189, Ser190, Leu192, Phe196, Glu276, Gly277, and Phe280 ( Figure 5). These hydrophobic interactions are important for strengthening the protein-ligand interaction and positioning icaritin in the active pocket to inhibit PTP1B activity. The binding energy of the PTP1B-icariside II complex was −8.77 kcal/mol, and four hydrogen bonds interacted with Glu276. The oxygen groups of the Glu276 residue were involved in the hydrogen bond interaction with the hydroxyl groups of icariside II; the bond distances ranged from 2.59 to 3.13 Å. In addition, hydrophobic interactions were observed between Ser187, Ala189, Leu192, Asn193, Phe196, Lly197, Glu200, Gly277, Phe280, and Ile281 (Figure 5c). Therefore, hydrogen bonds and hydrophobic interactions with PTP1B might be associated with substrate hydrolysis. The molecular docking simulation results corresponded well with the experimental data with respect to the abilities of the different compounds to inhibit PTP1B.  hydrogen bonds interacted with Glu276. The oxygen groups of the Glu276 residue were involved in the hydrogen bond interaction with the hydroxyl groups of icariside II; the bond distances ranged from 2.59 to 3.13 Å. In addition, hydrophobic interactions were observed between Ser187, Ala189, Leu192, Asn193, Phe196, Lly197, Glu200, Gly277, Phe280, and Ile281 (Figure 5c). Therefore, hydrogen bonds and hydrophobic interactions with PTP1B might be associated with substrate hydrolysis. The molecular docking simulation results corresponded well with the experimental data with respect to the abilities of the different compounds to inhibit PTP1B.

Discussion
DM is a major community-level concern. Epidemiological studies have revealed that the prevalence of DM continues to increase, having almost attained epidemic proportions worldwide. As of 2015, more than 415 million adults had DM. This number is expected to grow to 642 million by 2040 [37]. Over the past decade, the growing epidemic of DM and its associated complications have presented both challenges and opportunities. Significant therapeutic developments have been made, and many novel chemical compounds have been tested for their ability to treat diabetes; however, the long-term safety and efficacy of these chemicals are not yet established. Thus, finding new natural drug candidates for DM, with no or reduced side effects, is an important goal for improving treatment and prevention of diabetes. PTP1B is a negative regulator of the insulin metabolic pathway, and α-glucosidase promotes the absorption of carbohydrates into blood vessels by hydrolyzing oligosaccharides to glucose. Despite the identification of many potent compounds [38][39][40][41], it is still essential to search for new inhibitors with better safety and high efficacy. E. koreanum is used in TCM to treat impotence, female infertility, dysuria, rheumatic arthritis, geriatric depression, and angina pectoris [11,12]. Despite the broad pharmacological effects of E. koreanum, no systematic investigation prior to the present study investigated the possibility of identifying anti-DM drug candidates from E. koreanum.
As summarized in Table 1, the anti-DM activity of the MeOH extract was examined via in vitro PTP1B and α-glucosidase activity assays. Similarly, the effects of the CH2Cl2, EtOAc, n-BuOH, and H2O fractions were assessed in the same assays. The results showed that the MeOH extract of the

Discussion
DM is a major community-level concern. Epidemiological studies have revealed that the prevalence of DM continues to increase, having almost attained epidemic proportions worldwide. As of 2015, more than 415 million adults had DM. This number is expected to grow to 642 million by 2040 [37]. Over the past decade, the growing epidemic of DM and its associated complications have presented both challenges and opportunities. Significant therapeutic developments have been made, and many novel chemical compounds have been tested for their ability to treat diabetes; however, the long-term safety and efficacy of these chemicals are not yet established. Thus, finding new natural drug candidates for DM, with no or reduced side effects, is an important goal for improving treatment and prevention of diabetes. PTP1B is a negative regulator of the insulin metabolic pathway, and α-glucosidase promotes the absorption of carbohydrates into blood vessels by hydrolyzing oligosaccharides to glucose. Despite the identification of many potent compounds [38][39][40][41], it is still essential to search for new inhibitors with better safety and high efficacy. E. koreanum is used in TCM to treat impotence, female infertility, dysuria, rheumatic arthritis, geriatric depression, and angina pectoris [11,12]. Despite the broad pharmacological effects of E. koreanum, no systematic investigation prior to the present study investigated the possibility of identifying anti-DM drug candidates from E. koreanum.
As summarized in Table 1, the anti-DM activity of the MeOH extract was examined via in vitro PTP1B and α-glucosidase activity assays. Similarly, the effects of the CH 2 Cl 2 , EtOAc, n-BuOH, and H 2 O fractions were assessed in the same assays. The results showed that the MeOH extract of the aerial parts of E. koreanum inhibited both PTP1B and α-glucosidase in a dose-dependent manner (Figure 1). Of the four solvent-soluble fractions, the CH 2 Cl 2 fraction showed the most potent inhibition of PTP1B and α-glucosidase. In addition, icaritin and icariside II showed inhibitory potential, with IC 50 values of 11.59 ± 1.39 and 9.94 ± 0.15 µM against PTP1B and 74.42 ± 0.01 and 106.59 ± 0.44 µM against α-glucosidase, respectively. In conclusion, icaritin and icariside II, two deglycosylated metabolites of icariin, could potentially play therapeutically important roles as PTP1B and α-glucosidase inhibitors. In previous studies, E. koreanum extract was pharmacologically and biologically evaluated. These studies revealed that the extract exhibits osteoblastic proliferation stimulating activity [42], affects kidney-yang deficiency syndrome [43], and possesses estrogenic activity [44]. In addition, Oh et al. [15] reported that EH extract has significant anti-diabetic activity in rats with streptozotocin-induced diabetes. Furthermore, epimedin A-C did not inhibit α-glucosidase, similar to the findings of a previous study [45]. Moreover, icaritin and icariside II from E. koreanum also showed potent anti-diabetic activities. This study is the first to report the potencies of these compounds.
To establish the modes of inhibition of the two compounds with high inhibitory activity (icaritin and icariside II), enzyme kinetics and molecular docking studies were performed. The Lineweaver-Burk and Dixon plots revealed that both of the active compounds were non-competitive inhibitors of PTP1B because they decreased the V max values without changing the K m values. To predict binding sites and affinities, molecular docking simulations were performed. The best docked conformation was selected based on binding affinity, hydrogen bonds, and hydrophobic interactions. Amino acid residues of enzymes interact with compounds via H-bonds and hydrophobic interactions [46]. Computational docking simulations are one method for comprehensively understanding the mechanism behind active site binding interactions [47]. To avoid potential bottlenecks in the development of PTP1B inhibitors, much attention has focused on allosteric PTP1B inhibitors because allosteric inhibitors have higher specificity, fewer side effects, and lower toxicity [48,49]. The allosteric site is located on the C-terminal domain of PTP1B and is bordered by helices α3, α6, and α7, which constitute a hydrophobic pocket for allosteric inhibitors [50]. The WPD loop, which contributes to the recognition of acids/bases via Asp181, is the primary determinant for the binding of catalytic substrates in PTP1B. Moreover, allosteric inhibition of PTP1B was shown to be associated with closure of the catalytic WPD loop [51,52]. In the presence of allosteric inhibitors, the WPD loop adopts a different conformation compared to that in the Apo form of PTP1B [50]. Compound 2, which is a positive inhibitor of the allosteric PTP1B model, forms two hydrogen bonds with allosteric sites of PTP1B (e.g., Asp193 and Glu276) ( Table 3 and Figure 5a). The hydroxyl group of icaritin interacted with PTP1B, forming a single hydrogen bond with Asn193 with a binding energy of −6.24 kcal/mol (Figure 5b). Icariside II showed a lower binding score to PTP1B than icaritin, indicating high affinity. The Glu276 residue interacted with the hydroxyl groups of the α-L-rhamnose moiety of icariside II by forming four hydrogen bonds (Figure 5c). These hydrophobic interactions contributed to the higher binding affinity between PTP1B and the compounds. The molecular docking simulation results agreed well with the in vitro PTP1B activity assay results, where icariside II and icaritin were potent PTP1B inhibitors (Table 2).
Recently, Wu et al. [53] demonstrated that icariin is hydrolyzed to icariside II and icaritin by human intestinal microflora. Other studies have reported that icariin ameliorates streptozotocin-induced diabetic retinopathy [27], reduces mitochondrial oxidative stress injury in the hearts of diabetic rats [54], and protects mice from cisplatin-induced acute renal injury [55]. Furthermore, icariside II has been shown to ameliorate diabetic nephropathy in rats with streptozotocin-induced diabetes [56], and icaritin has been reported to have anti-inflammatory and anti-oxidative effects in rats under stress [57]. The findings of the present study indicate that the in vivo effects of icariin can be attributed to the activity of two of its metabolites, icariside II and icaritin.

Plant Materials
The aerial parts of Epimedium koreanum were collected from the Medicinal Herb Garden, Seoul, Korea, in November 2008 and confirmed by Professor Jinwoong Kim at the College of Pharmacy, Seoul National University. A voucher specimen (no. 20081105) was deposited in the authorized laboratory.

Extraction and Fractionation
The dried aerial parts of E. koreanum (2.4 kg) were extracted with MeOH (5 L × 3 times) for 3 h under reflux. After filtration, the entire filtrate was concentrated to dryness in vacuo at 40 • C, yielding the MeOH extract (655.2 g).

PTP1B Activity Assay
The abilities of the aerial parts of E. koreanum and their constituents to inhibit human recombinant PTP1B activity were evaluated using pNPP as a substrate [62]. Briefly, 40 µL of recombinant PTP1B enzyme [0.5 units diluted with a PTP1B reaction buffer containing 50 mM citrate (pH 6.0), 0.1 M NaCl,1 mM EDTA, and 1 mM DTT] was added to each well of a 96-well plate (final volume 100 µL), with or without a test sample. Then, 50 µL of 2 mM pNPP dissolved in PTP1B reaction buffer was added. After incubation at 37 • C for 20 min in the dark, the reaction was terminated by addition of 10 M NaOH. The amount of p-nitrophenylate produced by enzymatic dephosphorylation of pNPP was determined by measuring the absorbance at 405 nm using a microplate spectrophotometer (Molecular Devices). Non-enzymatic dephosphorylation of pNPP was corrected by measuring the increase in absorbance at 405 nm obtained in the absence of PTP1B. Ursolic acid was used as a positive control. The percent inhibition was calculated as {(A C − A S )/A C } × 100, where A C is the absorbance of the control, and A S is the absorbance of the sample. The IC 50 is expressed as the mean ± SEM of triplicate experiments.

α-Glucosidase Activity Assay
α-Glucosidase enzymatic activity was assayed spectrophotometrically as described previously [63]. Briefly, 60 µL of reaction mixture containing 20 µL of 100 mM phosphate buffer (pH 6.8), 20 µL of 2.5 mM pNPG, and 20 µL of sample (final concentration ranging from 20 to 300 µM, dissolved in 10% DMSO) was added to each well. After incubation at 37 • C for 5 min in the dark, 20 µL of yeast α-glucosidase (0.2 U/mL) in 10 mM phosphate buffer (pH 6.8) was added. The plate was then incubated at 37 • C for 15 min in the dark, after which the reaction was stopped by the addition of 80 µL of 0.2 M sodium carbonate solution. The absorbance was immediately recorded at 405 nm using a microplate spectrophotometer (Molecular Devices). The control consisted of the same reaction mixture but with the sample solution replaced by an equivalent volume of phosphate buffer. Acarbose dissolved in 10% DMSO was used as a positive control. The inhibition percentage (%) was obtained using the same equation as in the PTP1B enzymatic assay.

Enzyme Kinetics Analysis of PTP1B
In order to determine the inhibition mechanism, two complementary kinetics methods were used: Lineweaver-Burk and Dixon plots [64][65][66]. The level of enzymatic inhibition at various concentrations of test sample was evaluated by monitoring the effect of different substrate concentrations in Dixon plots (single reciprocal plots). Dixon plots for PTP1B inhibition were obtained in the presence of different concentrations of pNPP. Using Lineweaver-Burk plots (double reciprocal plots), the inhibition modes of the compounds were determined at various concentrations of pNPP with different concentrations of compounds. The enzymatic procedures were the same as for the PTP1B assay methods. The types of inhibition were investigated by interpreting the Dixon plots, in which the value of the x-axis indicates K i .

Molecular Docking Simulation of PTP1B Inhibition
The X-ray crystallographic structure of PTP1B complexed with its selective inhibitor 3-(3,5-dibromo-4-hydroxy-benzoyl)-2-ethyl-benzofuran-6-sulfonic acid (4-sulfamoyl-phenyl)-amide (compound 2) (PDB ID: 1T49) and the 3D structure of its catalytic inhibitor 3-({5-[(N-acetyl-3-{4-[(carboxycarbonyl)(2-carboxyphenyl)amino]-1-naphthyl}-L-alanyl)amino]pentyl}oxy)-2-naphthoic acid (compound 23) were obtained from the RCSB Protein Data Bank [67,68] website. The resolution of the crystal structure was 1.9 Å [69], and the PubChem Compound (NCBI) CID was 447410. The bound PTP1B inhibitor and water molecules were removed from the structure for the docking simulation using Accelrys Discovery Studio 4.1 software (Accelrys, Inc. San Diego, CA, USA). The 3D structures of icaritin and icariside II were obtained from PubChem Compound (NCBI), corresponding to compound CIDs of 5318980 and 44587252, respectively, and protonated (pH 7.0) using the MarvinSketch program (ChemAxon, Budapest, Hungary). Automated docking simulation was performed using AutoDock Tool (ADT) to assess the appropriate binding orientations and conformations of the PTP1B with the different compounds. A Lamarkian genetic algorithm method implemented in AutoDock 4.2 was employed. For docking calculations, Gasteiger charges were added by default, the rotatable bonds were set by the ADTs, and all torsions were allowed to rotate. Grid maps were generated by the Autogrid program; the grid box size was 80 × 80 × 80 and had a default spacing of 0.375 Å. The X, Y, and Z centers were 51, 19.034, and 14.048, respectively. The docking protocol for rigid and flexible ligand docking consisted of 10 independent genetic algorithms, whereas ADT default settings were used for the other parameters. Binding PTP1B residues and their corresponding binding affinity scores were regarded as the best molecular interactions. The results were analyzed using UCSF Chimera [70], while the hydrogen bonds and van der Waals interaction residues were visualized by Ligplot 1.4.5.

Statistics
All results are expressed as the mean ± SEM of triplicate samples. Statistical significance was analyzed using one-way ANOVA and Student's t-test (Systat Inc., Evanston, IL, USA).

Conclusions
The results of the current study indicate that the aerial parts of E. koreanum and their constituents possess anti-DM properties. Comparative evaluations of these activities of E. koreanum might be useful in the development of therapeutic agents for treating patients with DM. Among the tested compounds, icaritin and icariside II (two deglycosylated metabolites of icariin) significantly inhibited PTP1B. Furthermore, enzyme kinetics analysis and molecular docking simulations supported these results. Thus, approaches combining computational work with experimental validation are useful for developing new natural drug candidates against diabetes. These active candidates have significant potential as drug candidates for treating DM. Further investigations of the bioactivity of these natural products are needed to determine the precise therapeutic and/or preventive potential of E. koreanum and its constituents in vivo.