The Chemical Constituents from Fruits of Catalpa bignonioides Walt. and Their α-Glucosidase Inhibitory Activity and Insulin Secretion Effect

Catalpa pod has been used in traditional medicine for the treatment of diabetes mellitus in South America. Studies on the constituents of Catalpa species have shown that it is rich in iridoids. In the present study, three previously undescribed compounds (2–4), including two secoiridoid derivatives along with twelve known compounds, were isolated from the fruits of Catalpa bignonioides Walt. In addition, fully assigned 13C-NMR of 5,6-dihydroxy-7,4’-dimethoxyflavone-6-O-sophoroside (1) is reported for the first time in the present study. The structures of compounds were determined on the basis of extensive spectroscopic methods, including UV, IR, 1D, and 2D NMR, mass spectroscopy, and CD spectroscopic data. All the isolated compounds were evaluated for α-glucosidase inhibitory activity. Among the tested compounds, compounds 2, 3, and 9 exhibited significant inhibitory activity against α-glucosidase enzyme assay. Meanwhile, the effect of compounds 2, 3, and 9 on glucose-stimulated insulin secretion (GSIS) was measured using pancreatic β-cells. Compounds 2, 3, and 9 exhibited non-cytotoxicity-stimulated insulin secretion in INS-1 cells. The expression levels of proteins associated with β-cell function and insulin secretion such as phosphorylation of total insulin receptor substrate-2 (IRS-2), phosphatidylinositol 3-kinase (PI3K), Akt, activated pancreatic duodenal homeobox-1 (PDX-1), and peroxisome proliferator-activated receptor-γ (PPAR-γ) were increased in INS-1 cells after treatment with compounds 2, 3, and 9. The findings of the present study could provide a scientific warrant for their application as a potential antidiabetic agent.


Introduction
Diabetes mellitus (DM) is a metabolic disease characterized by glucose intolerance and changes in lipid and protein metabolism [1]. DM can be divided into insulin-dependent diabetes mellitus (type 1) and non-insulin-dependent diabetes mellitus (type 2). Type 2 diabetes is a metabolic disorder resulting from the body's inability to produce, and properly utilize insulin, which causes hyperglycemia [2]. Many oral hypoglycemic agents for clinical use are available for the treatment of DM, but these synthetic agents produce severe side effects such as hypoglycemia, weight gain, gastrointestinal disturbances, and liver toxicity [3]. With the recommendations of the World Health Organization (WHO) expert committee on DM, many researchers have attempted to find more effective natural products with fewer side effects [4].
Catalpa bignonioides Walt. (Bignoniaceae), commonly known as a bean tree, is a traditional folk medicine in South America for the treatment of gastric diseases, helmintic infections, oncological diseases, bronchial diseases, carbuncles, scabs, and abscesses [5].
Compound 2 was isolated as a yellow amorphous powder. Its molecular formula was determined as C42H48O22 on the basis of HR-ESI-MS pseudo-ion at m/z 905.2669 [M + H] + (calcd. for C42H49O22, 905.2710). The 1 H-NMR spectrum of 2 showed the signals of two methoxy protons at δH 3.83 and 3.88 (each 3H, s), three anomeric protons at δH 4.71 (d, J = 7.7 Hz), 4.74 (d, J = 7.6 Hz), and 5.28 (d, J = 7.0 Hz), and two olefinic protons at δH 6.53 (s) and 6.43 (s). In addition, proton signals of A2B2-type aromatic protons at δH 7.08 (d, J = 8.4 Hz) and 7.90 (d, J = 8.4 Hz), and a benzoyl group at δH 7.13 (t, J = 7.4 Hz), 7.29 (t, J = 7.4 Hz), and 7.66 (2H, m) were shown. The 13 C-NMR and HSQC spectra of 2 showed the presence of 42 carbons, including eleven non-protonated carbons (two carbonyls), 26 methines (fifteen oxygenated), three oxygenated methylenes, and two methoxy carbons ( Table 1). The analysis NMR data of 2 indicated that the structure of 2 was similar to those of 5,6-       Table 1). The analysis NMR data of 2 indicated that the structure of 2 was similar to those of 5,6-dihydroxy-7,4 -dimethoxyflavone-6-O-sophoroside [9], except for the addition of a 6-O-benzoate glucopyranoside. The position of a sugar moiety at C-6 was suggested by HMBC correlations from glc H-1" (δ H 5.28) to C-6 (δ C 128.0). The sugar moiety was identified by comparing NMR data with those of sugar moiety in 6-hydroxyluteolin 7-O-[6"'-benzoyl-β-D-glucopyranosyl-(1→2)]-β-D-glucopyranoside (aphyllanthoside) [18]. The terminal sugar connectivity was confirmed by HMBC correlation between H-2"' (δ H 3.56) and C-1"" (δ C 104.4), which proved a 1→2 interglycosidic linkage by the downfield shift for C-2"' (δ C 83.1) of the inner glucose. The position of benzoyl group was verified by the 2D NMR and the MS/MS spectrum. The esterification site of the benzoic acid was found to be C-6"' of the terminal glucose, on the basis of the de-shielding of H-6"' (δ H 4.33) and C-6"' (δ C 64.0) as well as the cross-peak at C-7""' (δ C 166.3) in the HMBC spectrum. In addition, it was further confirmed by MS/MS spectrum of compound 2 to elucidate the location of the benzoyl group. Its two fragment ions, Y1 and Y2, generated by the cleavage of sugar moieties from compound 2 (Y1 [M + H] + and Y2 [M + Na -H 2 O] + ) using LC-QTOF-MS revealed that the benzoyl group is located at the second sugar. The positions of methoxy groups were verified by the HMBC correlation between protons of methoxy groups (δ H 3.83 and 3.88) and C-7 and C-4' (δ C 158.4 and 162.9), concluding that the methoxy groups are located at C-7 and C-4', respectively ( Figure 2). Thus, the structure of 2 was elucidated to be 5,6-dihydroxy-7,4'-dimethoxyflavone-6-O--β-D-glucopyranoside. The UV, IR, CD, NMR ( 1 H, 13  Compound 3 was isolated as a brown oil. Its molecular formula was determined as C 10 H 16 O 6 based on HR-ESI-MS pseudo-ion at m/z 233.0981 [M + H] + (calcd. for C 10 H 17 O 6 , 233.1020). The NMR spectra were similar to those of (7R)-hydroxyeucommic acid isolated from the fruits C. ovata [10], except for the replacement of a hydroxyl group into a methoxy group at C-3. The 1 H-NMR spectrum of 3 showed the signals of one methoxy proton at δ H 3.75 (3H, s). The methoxy group showed an HMBC correlation with ester carbonyl carbon (δ C 174.9), which was also correlated with methylene protons at H-4 (δ H 2.48) ( Figure 2). The stereochemistry of 3 was clarified based on the NOESY and the CD spectra. The NOE interaction was observed between H-6 (δ H 3.95) and H-7 (δ H 4.56), but no correlation was shown between H-5 (δ H 3.17) and H-6 (δ H 3.95) ( Figure 3). The absolute configuration was determined by the application of the inverse Octant rule for allylic oxygen substituent, which generally dominated the onset or appearance of the lower wavelength Cotton effect [19]. In addition, iridoids isolated from Catalpa genus such as (7R)-hydroxyeucommic acid n-butyl ester [∆ε −0.67 (204.4 nm)] [10] exhibited R configuration. As a biogenetic derivative, compound 3 is supposed to have an R configuration at C-7. It was further confirmed by CD spectrum with a negative Cotton effect at 204.0 nm (∆ε −12.65) ( Figure 4). The structure of 3 is, therefore, determined as (7R)-3-methoxyhydroxyeucommic acid. The IR, NMR ( 1 H, 13        Compound 4 was also isolated as a brown oil. Its molecular formula was confirmed as C 17 Table 2). Instead of a hydroxyl group at C-7 in compound 3, compound 4 consisted of a p-hydroxybenzoic acid. The HMBC spectrum gave a threebonded correlation between C-11 (δ C 168.1) and H-7 (δ H 5.87) ( Figure 2). The stereochemistry of 4 was also determined by NOESY and the CD spectra, as with compound 3. The NOE interaction was observed between H-6 (δ H 4.17) and H-7 (δ H 5.87), but not between H-5 (δ H 3.24) and H-6 (δ H 4.17) (Figure 3). The CD spectrum of 4 also showed a negative Cotton effect at 204.0 nm (∆ε −18.57), suggesting its R-configuration at C-7 ( Figure 4). Thus, the structure of 4 was elucidated to be (7R)-3-methoxy-(7-O-p-hydroxybenzoyl) eucommic acid. The UV, IR, NMR ( 1 H, 13

α-Glucosidase Inhibitory Activity
In spite of the introduction of various anti-DM medicines such as DPP-4 inhibitors, SGLT-2 inhibitors, or GLP-1 analogues, α-glucosidase inhibitors are still prevalently used to establish glycemic control over postprandial hyperglycemia. They can retard the liberation of glucose from carbohydrates and delay glucose absorptions from the gut. To find anti-DM phytochemicals, all of the isolated compounds were evaluated for α-glucosidase inhibitory activity using enzyme assay at the concentration of 20 µM (Table 3). The previous biological evaluation suggested that anthocyanidin, isoflavone, flavonol, and secoiridoid glucosides exhibit potent inhibitory effects on α-glucosidase activity [20,21]. In accordance with previous biological data, compounds 2 (flavone glycoside), 3, and 9 (both iridoids) demonstrated the most potent inhibitory activity, which is comparable to that of a well-known α-glucosidase inhibitor, acarbose. Several previous researchers have reported that flavones, such as isoquercitrin and isovitexin [22], as well as the iridoid, such as loniceranan B and swreoside [20], exerted significant α-glucosidase inhibitory activities. Thus, compounds 2, 3, and 9 could be potential natural resources as anti-DM phytochemicals. Table 3. α-glucosidase activity of compounds (20 µM) isolated from C. bignonioides.

Virtual Screening of α-Glucosidase Inhibitors
Virtual screening (VS) could be useful in searching for a novel lead compound that is appropriate for predictive new drug discovery studies. Compared with other de novo design methods, virtual screening suggests a proper understanding of the important structural and physicochemical features. To identify the putative binding conformation of bioactive compounds, the structure-based VS was performed against the α-glucosidase from the Protein Data Bank (PDB, http://www.rcsb.org/pdb) under code (PDB 3A4A). Compounds 3 and 9 were well-docked into the catalytic site of the α-glucosidase ( Figure  5) with the values of CDOCKER energy of −17.98 and −22.64 kcal·mol −1 , respectively. The key feature showed that 3 was docked into the active site via conventional hydrogen bonds with residues Arg 315, Arg 446, Gln 353, and Ile 440, and carbon-hydrogen bond with residues Glu 411 and Ser 441. Compound 9 formed conventional hydrogen bonding with residues Gln 353, Glu 411, and Ser 60.

Glucose-Stimulated Insulin Secretion (GSIS) Effect
Dysfunction and mass loss of pancreatic β cells are known to be risk factors for developing type 2 diabetes. Impairment of GSIS is mainly attributed to the initial dysfunction of pancreatic β cells. Therefore, amelioration of GSIS might be a strategy for the discovery of a potential antidiabetic agent. In the present study, compounds 2, 3, and 9, which exerted significant α-glucosidase inhibitory activities, were determined to increase GSIS in INS-1 cells. Since compounds 2, 3, and 9 were not toxic in less than 12.5 µM, their concentrations were used in the insulin secretion assay ( Figure 6A-C). As shown in Figure 6D-F, compounds 2, 3, and 9 led to an increase in GSI in a dose-dependent manner. The GSI levels were 1.96 ± 0.16, 2.71 ± 0.05, and 2.73 ± 0.51 for compound 2 at 2.5, 5, and 10 µM, respectively ( Figure 6D). The GSI levels were 1.37 ± 0.01, 3.56 ± 0.14, and 3.66 ± 0.03 for compound 3 at 2.5, 5, and 10 µM, respectively ( Figure 6E). The GSI levels were 1.43 ± 0.16, 3.02 ± 0.13, and 4.75 ± 0.21 for compound 9 at 2.5, 5, and 10 µM, respectively ( Figure 6F). Compounds 2, 3, and 9 stimulated insulin secretion in INS-1 cells without inducing cytotoxicity. In the amelioration of GSIS in INS-1 cells, compounds 3 and 9 (both iridoids) were more effective than compound 2 (flavone glycoside). The insulin secretion effect of compound 3 was the best and increased in a concentration-dependent manner. It has been shown that lyonofolin B, an iridoid isolated from lyonia ovalifolia, potentiated a GSIS from mice pancreatic islets in the male BALB/c mice [23]. Moreover, another study has indicated that rutin, a flavonoid glycoside, increases GSIS from pancreatic islets in the male Wistar rats [24]. However, its underlying mechanism has yet to be revealed. A further study revealed the underlying mechanism of compounds 2, 3, and 9 on amelioration of GSIS using the Western blot assay.
3.66 ± 0.03 for compound 3 at 2.5, 5, and 10 μM, respectively ( Figure 6E). The GSI levels were 1.43 ± 0.16, 3.02 ± 0.13, and 4.75 ± 0.21 for compound 9 at 2.5, 5, and 10 μM, respectively ( Figure 6F). Compounds 2, 3, and 9 stimulated insulin secretion in INS-1 cells without inducing cytotoxicity. In the amelioration of GSIS in INS-1 cells, compounds 3 and 9 (both iridoids) were more effective than compound 2 (flavone glycoside). The insulin secretion effect of compound 3 was the best and increased in a concentration-dependent manner. It has been shown that lyonofolin B, an iridoid isolated from lyonia ovalifolia, potentiated a GSIS from mice pancreatic islets in the male BALB/c mice [23]. Moreover, another study has indicated that rutin, a flavonoid glycoside, increases GSIS from pancreatic islets in the male Wistar rats [24]. However, its underlying mechanism has yet to be revealed. A further study revealed the underlying mechanism of compounds 2, 3, and 9 on amelioration of GSIS using the Western blot assay.  To evaluate the role of peroxisome proliferator-activated receptor γ (PPAR-γ), insulin receptor substrate-2 (IRS-2), phosphatidylinositol 3-kinase (PI3K), Akt, and pancreatic duodenal homeobox-1 (PDX-1) in the effect of compounds 2, 3, and 9 on GSIS, we measured these protein levels in pancreatic β-cells and demonstrated that the protein expression levels of PPAR-γ, P-IRS-2 (Ser731), P-PI3K, P-Akt (Ser473), and PDX-1 were increased by treatment with compounds 2, 3, and 9 at 10 µM compared to untreated controls. As reported, PPARγ has been shown to regulate the expression of genes involved in insulin secretion in pancreatic β cells [25]. However, in order to avoid the potential side effects of synthetic antidiabetic compounds such as thiazolidinediones and troglitazone, the PPAR-γ agonists, finding antidiabetic compounds from natural products is necessitated. Another role of PPAR-γ is known to regulate the PDX-1 gene promoter in pancreatic β cells [26,27]. Others have reported that troglitazone increases the expression of PDX-1 in INS-1 cells [26]. This prompted us to study the effect of compounds 2, 3, and 9 on the expression of PPAR-γ and PDX-1 in INS-1 cells. Thus, the expression of PPAR-γ and PDX-1 was evaluated with Western blot. As shown in Figure 7, treatment with compounds 2, 3, and 9 increased the expression of PPAR-γ and PDX-1. Subsequently, we assessed whether treatment with compounds 2, 3, and 9 increases the serine phosphorylation of IRS-2 (Ser 731), phosphorylation of PI3K, and serine phosphorylation of Akt (Ser473). It is well known that the IRS-2 signaling pathway is essential to the function of pancreatic β cells. In addition, the loss of IRS-2 expression in mice is linked to development of type 2 diabetes due to insuffi-ciency of pancreatic β cells [28]. In this signaling pathway, Akt and its downstream protein, PI3K, can be activated through phosphorylation of IRS-2 [29]. It has also been reported that upregulation of the PI3K/Akt signaling pathway promotes proliferation of pancreatic β cells [30]. As shown in Figure 7, treatment with compounds 2, 3, and 9 increased the expression of IRS-2, PI3K, and Akt. Taken together, these results suggested that compounds 2, 3, and 9 not only upregulate the expression of PPAR-γ and PDX-1, but also upregulate the phosphorylation of IRS-2, PI3K, and Akt in INS-1 cells. These results enhanced the understanding of the underlying mechanism of compounds 2, 3, and 9 on amelioration of GSIS. However, further study on the mode of entry of compounds 2, 3, and 9 into the pancreatic β cells and its effect in animal models for diabetes should be evaluated.

Plant Material
The fruits of C. bignonioides were collected from Arboretum of Seoul National University in Suwon, Korea

Structured-Based Virtual Screening
Structure-based virtual screening was performed using CDOCKER docking protocol in Discovery studio (BIOVIA/Accelrys Inc. San Diego, CA, USA). The 3A4A of α-glucosidase was downloaded from RCSB Protein Data Bank (www.rcsb.org). To perform docking in CDOCKER, the following docking parameters were used: grid center (X: 20.62, Y: −2.54. Z: 18.25), grid size (14.2 Å), grid spacing (0.5 Å). The most potent α-glucosidase inhibitory compounds, 2, 3, and 9, and positive control, acarbose, were used as a ligand. The hydrogen atoms were added to ligands. The 3D structure of ligands was generated through molecular dynamics, and the initial poses were generated by various rotations/translations of structures. Both ligand and receptor docking were performed using Chemistry at Harvard Macromolecular Mechanics (CHARMm) force field docking algorithm (Runs 10). For defined electrostatic or van der Waals interaction, a +1 charge probe was used for electrostatic interaction, and probe radii (0.65 through 2.55 Å) was used for van der Waals interaction. The known active site of α-glucosidase was selected as the binding site and −CDOCKER energy level was used for the result of interactions between the ligands and the receptor.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.