α-Glucosidase Inhibitors From the Coral-Associated Fungus Aspergillus terreus

Nine novel butenolide derivatives, including four pairs of enantiomers, named (±)-asperteretones A–D (1a/1b–4a/4b), and a racemate, named asperteretone E (5), were isolated and identified from the coral-associated fungus Aspergillus terreus. All the structures were established based on extensive spectroscopic analyses, including HRESIMS and NMR data. The chiral chromatography analyses allowed the separation of (±)-asperteretones A–D, whose absolute configurations were further confirmed by experimental and calculated electronic circular dichroism (ECD) analysis. Structurally, compounds 2–5 represented the first examples of prenylated γ-butenolides bearing 2-phenyl-3-benzyl-4H-furan-1-one motifs, and their crucial biogenetically related metabolite, compound 1, was uniquely defined by an unexpected cleavage of oxygen bridge between C-1 and C-4. Importantly, (±)-asperteretal D and (4S)-4-decarboxylflavipesolide C were revised to (±)-asperteretones B (2a/2b) and D (4), respectively. Additionally, compounds 1a/1b–4a/4b and 5 were evaluated for the α-glucosidase inhibitory activity, and all these compounds exhibited potent inhibitory potency against α-glucosidase, with IC50 values ranging from 15.7 ± 1.1 to 53.1 ± 1.4 μM, which was much lower than that of the positive control acarbose (IC50 = 154.7 ± 8.1 μM), endowing them as promising leading molecules for the discovery of new α-glucosidase inhibitors for type-2 diabetes mellitus treatment.


INTRODUCTION
Diabetes mellitus (DM) is one of the most serious chronic diseases with the ever-increasing incidence rates of obesity and aging of the general population throughout the world (Kopelman, 2000) In 2013, it was estimated that over 382 million people all over the world have DM and this number is predicted to increase up to 500 million in 2030, when this disease will be excepted to be the 7th leading cause of death (Lauritano and Ianora, 2016). Globally, type-2 diabetes (non-insulin-dependent DM) covered 90-95% of all the diabetes cases (Lauritano and Ianora, 2016). Postprandial hyperglycemia is an important factor for the induction of type-2 diabetes and complications related to the diseases, such as micro-and macro-vascular diseases (Baron, 1998). A good strategy to maintain the normal level of postprandial plasma glucose is to medicate in combination with dietary restriction and an exercise plan (Kim et al., 2008). In type-2 diabetes, delaying glucose absorption after meals by inhibition of α-glucosidase is known to help the therapy (Kim et al., 2008). For diabetic patients, α-glucosidase inhibitors (AGIs) are widely applied either as monotherapy or in combination with other oral hypoglycemic agents or insulin (Hung et al., 2012). However, AGIs-induced serious liver injuries and gastrointestinal side effects restricted the clinical practice (Yin et al., 2014;Kao et al., 2016). In view of the limited number of safe anti-diabetic drugs with low toxicity and everincreasing number of diabetic patients, the exploration for new α-glucosidase inhibitors, attracted, and still attract great interests from scientific community.

Fungus Material
The fungus Aspergillus terreus was isolated from the soft coral Sarcophyton subviride, which was collected from the Xisha Island in the South China Sea. This strain was cultivated on potato dextrose agar (PDA) medium and identified by one of the authors (JW), based on its morphological properties and ITS sequence analysis (GenBank access no. MF972904). The fungal strain was reserved in the culture collection of Tongji Medical College, Huazhong University of Science and Technology.

Cultivation, Extraction, and Isolation
The strain Aspergillus terreus was cultivated on PDA (Potato Dextrose Agar) medium at 28 • C for 1 week to prepare the seed cultures. Agar plugs were cut into small pieces (approximately 0.5 × 0.5 × 0.5 cm 3 ) and then was inoculated in 300 × 500 mL Erlenmeyer flasks which were previously sterilized by autoclaving, each containing 200 g rice and 200 mL distilled water. All flasks were incubated at 28 • C for 28 days. Then, the whole rice solid medium was extracted seven times in 95% aqueous EtOH at room temperature, and the solvent was removed under reduced pressure to afford a crude extract, which was partitioned with ethyl acetate against water to obtain the ethyl acetate soluble part (1.5 kg). The organic extract was separated by silica gel CC (100-200 mesh) with a stepwise gradient elution of petroleum ether-ethyl acetate-MeOH (10:1:0, 7:1:0, 5:1:0, 3:1:0, 1:1:0, 2:2:1, 1:1:1) to afford seven fractions (A-G).

ECD Calculations
The theoretical calculations of compounds 1a/1b and 2a/2b were performed using Gaussian 09 and figured using GaussView 5.0 (He et al., 2017a,b,c;Hu et al., 2017). Conformation search using molecular mechanics calculations was performed in the Discovery Studio 3.5 Client with MMFF force field with 20 kcal mol −1 upper energy limit (Smith and Goodman, 2010). The optimized conformation geometries and thermodynamic parameters of all selected conformations were provided. The predominant conformers were optimized at B3LYP/6-31G(d,p) level. The theoretical calculation of ECD was performed using time dependent Density Functional Theory (TDDFT) at the B3LYP/6-31G(d,p) level in MeOH with PCM model (Miertus et al., 1981). The ECD spectra of compounds 1a/1b and 2a/2b were obtained by weighing the Boltzmann distribution rate of each geometric conformation (Tähtinen et al., 2003).
The ECD spectra were simulated by overlapping Gaussian functions for each transition according to: The σ represented the width of the band at 1/e height, and E i and R i were the excitation energies and rotational strengths for transition i, respectively. R vel had been used in this work.

α-Glucosidase Inhibitory Assay
The α-glucosidase enzyme from Saccharomyces cerevisiae (Sigma Aldrich, USA) solution (1.5 U/mL) was prepared by dissolving the α-glucosidase in 200 M phosphate buffer (pH 6.8). The α-glucosidase enzyme solution (20 µL), test compounds (10 µL) and buffer (40 µL) were pipetted and mixed in a 96 well microtiter plate. The mixture was incubated at 37 • C for 10 min. After incubation, p-nitrophenyl-α-D-glucopyranoside (PNP-G) substrate solution (10 µL, in 20 mM phosphate buffer) was added. The increment of absorbance due to the hydrolysis of PNP-G by α-glucosidase was measured at the wavelength of 410 nm with a microplate reader (Thermo Scientific, Waltham, MA). Acarbose was used as a positive control and averages of three replicates were calculated. The α-glucosidase inhibitory activity was expressed as percentage inhibition and was calculated using the following formula: inhibition (%) = [1-(OD sample /OD blank )] × 100. The halfmaximal inhibitory concentration (IC 50 ) was calculated as the compound concentration that is required for 50% inhibition, and the IC 50 value of the acarbose was 154.7 ± 8.1 µM.

Molecular Docking Simulation
The virtual docking was implemented in the Surflex-Dock module of the FlexX/Sybyl software, which is a fast docking method that allows sufficient flexibility of ligands and keeps the target protein rigid. Molecules were built with Chemdraw and optimized at molecular mechanical and semiempirical level by using Open Babel GUI. The crystallographic ligands were extracted from the active site and the designed ligands were modeled. All the hydrogen atoms were added to define the correct ionization and tautomeric states, and the carboxylate, phosphonate and sulphonate groups were considered in their charged form. In the docking calculation, the default FlexX scoring function was used for exhaustive searching, solid body optimizing and interaction scoring. Finally, the ligands with the lowest-energy and the most favorable orientation were selected.  , and one methoxyl. The diagnostic data above indicated that compound 1 was a butenolide derivative.
Frontiers in Chemistry | www.frontiersin.org H 2 -5, and OMe-4 to C-4 (δ C 177.4) and H-2 to C-1 (δ C 178.3) indicated that a methyl ester and a carboxyl group were attached at C-3 and C-2, respectively. Thus, the planar structure of 1 was determined.
(±)-Asperteretone B (2a/2b) were also obtained as white, amorphous powders and assigned the molecular formula C 23 H 24 O 5 , as determined from the HRESIMS analysis at m/z 403.1526 [M + Na] + (calcd for C 23 H 24 O 5 Na, 403.1521) and 13 C NMR data. The 1D ( Table 2) and 2D NMR spectra of 2 were completely identical to that of the reported (±)-asperteretal D , which drove us to believe that they shared the same structures. The 1,3,4-trisubstituted phenyl group with a C-4" hydroxyl and a C-3" prenyl motif and para-disubstituted phenyl group with a C-4' hydroxyl motif were explicitly confirmed by detailed analysis of the 2D NMR data (Figure 2) of 2. However, a strong four-bond HMBC correlation from H 2 -5 to C-4 made us confused about the correctness of (±)asperteretal D. After careful examination of the HMBC spectrum (Figure 2) of (±)-asperteretal D, key correlations from H-2' to C-3 and from H-4 to C-1' were not observed in the HMBC spectrum; on the contrary, two four-bond HMBC correlations from H-2' to C-2 and from H 2 -5 to C-4 were observed, which were also found in the HMBC spectrum of 2. These data above suggested that (±)-asperteretal D should be structurally revised from 2-benzyl-3-phenyl-type to 2-phenyl-3-benzyl-type.
(±)-Asperteretone C (3a/3b), obtained as white, amorphous powders, were determined to have the molecular formula  Table 1) with those of 2 suggested that 3 was also a butenolide derivative, with the differences that the paradisubstituted phenyl group was attached with a C-4' methoxy motif in 3 rather than a C-4' hydroxyl motif in 2, and the 1,3,4-trisubstituted phenyl group attached with a C-4" hydroxyl and a C-3" prenyl motif in 2 was replaced by a 2-(2,3dihydrobenzofuran-2-yl)propan-2-ol motif in 3, as supported by the 2D NMR spectra (Figure 2), including HMBC and 1 H-1 H COZY correlations. Thus, the structure of 3 was determined.
Considering the similar structural features of 2 and 3, we deduced that compound 3 was likely a racemic mixture. As expected, by chiral HPLC resolution (Figure 3), two isolates were obtained. Since no apparent Cotton effects were decisive for the absolute stereochemistry of C-8", the experimental ECD spectra (Figure 5) of 3a and 3b were closely similar to those of (−)asperteretone B (2a) and (+)-asperteretone B (2b), respectively, indicating that compounds 3a and 3b possessed the 4S-and 4Rconfiguration, respectively. Regrettably, the configuration of C-8" was difficult to be determined (Liu et al., 2018b (Tables 1, 2) of 4 and 2 suggested that they shared the similar structural features, differing in that the 1,3,4-trisubstituted phenyl group attached with a C-4" hydroxyl and a C-3" prenyl motif in 2 was replaced by the 1,3,4-trisubstituted phenyl with the fusion of gem-dimethyl substituted tetrahydropyrane ring, as supported by the HMBC correlations from H 2 -8" to C-9", C-10", and C-11" and from H 2 -7" to C-2" and C-4", as well as the 1 H-1 H COZY correlation of H 2 -7"/H 2 -8". Thus, the planar structure of 4 was determined.
The optical rotation of zero in MeOH and inapparent Cotton effects in the ECD curve highlighted that 5 was racemic. Unluckily, despite for many attempts for several chiral columns using various mobile phase systems, we still failed to obtain the enantiomers of 5, which might own to that the rapid interconversion of these two enantiomers in the solvents prevented the separation on chiral columns. Compounds 1-5 represented two special classes of 7,8dimeric phenylpropanoids with unexpected architectures, and their plausible biogenetic pathways were proposed as follows (Scheme 1): two molecules, p-hydroxyphenyl pyruvic acid, underwent prenylation and decarboxylation reactions, respectively, followed by aldol condensation and dehydration reactions to create intermediate b. Alternatively, a further dehydration reaction of b could generate an acid anhydridecontaining intermediate c, which could furnish 2-5 through a series of reduction, cyclization, esterification, and so on. Meanwhile, the esterification at C-4 and reduction of 2,3 double bond could from 1, which was identified as a crucial biogenetically related metabolite, was the first report of 2,3disubstituted butenolide derivatives with an unexpected cleavage of oxygen bridge between C-1 and C-4. This finding would greatly expand the chemical space and biosynthesis study for butenolide derivatives.
Biological Evaluation of Compounds 1a/1b-4a/4b and 5 Compounds 1a/1b-4a/4b and 5 were evaluated for the αglucosidase inhibitory activity. As shown in Table 3, all the compounds exhibited potent inhibitory potency against αglucosidase, with IC 50 values ranging from 15.7 ± 1.1 to 53.1 ± 1.4 µM, which was much lower than that of the positive control acarbose (IC 50 = 154.7 ± 8.1 µM). All enantiomers displayed nearly horizontal IC 50 values against α-glucosidase inhibitory activity, indicating that the difference of chirality might have a negligible impact on the activity. Most importantly, compounds 1a/1b-4a/4b and 5 may provide novel chemical scaffolds for the discovery of new α-glucosidase inhibitors.
To investigate the binding mode of these compounds with α-glucosidase, molecular docking study was carried out by using the SYBYL 2.0 software. Due to the unavailable of crystal structure of α-glucosidase from Saccharomyces cerevisiae, the crystal structure of isomaltase (PDB ID: 3A4A) from S. cerevisiae, which is 84% similar to that of S. cerevisiae α-glucosidase, was conducted as docking model (Shen et al., 2015). The theoretical binding mode between 4a and the enzyme was shown in Figure 7. Compound 4a adopted a "V-shaped" conformation in the pocket. Detailed analysis showed that the phenolic group and benzopyran group of 4a formed π-π stacking interaction with the residue Phe303 and Phe173, respectively. It was also shown that the residue Asp307, Asp352, and Glu411 formed key hydrogen bonds with 4a, which were the main interactions between 4a and the enzyme. All these interactions helped 4a to anchor in the binding site of the enzyme.

CONCLUSIONS
In conclusion, nine novel butenolide derivatives belonging to two undescribed structural types, including four pairs of enantiomers (1a/1b-4a/4b) and a racemate (5), were isolated from the coral-associated fungus Aspergillus terreus. More importantly, (±)-asperteretal D and (4S)-4-decarboxylflavipesolide C were structurally revised to 2a/2b and 4, respectively. This study further enriched secondary metabolites in the Aspergillus species and was also a strong structural supplement to the new class of γ -butenolides. In addition, bioactivity evaluation results showed that all the isolates exhibited potent α-glucosidase inhibitory activity with IC 50 values ranging from 15.7 ± 1.1 to 53.1 ± 1.4 µM. On the background that DM is becoming a global public health problem and more new effective therapeutic agents are in the urgent need, our findings provide a basis for further development and utilization of butenolide derivatives as source of potential α-glucosidase inhibitors as therapeutic agents for type-2 diabetes mellitus.