Cytotoxic Polyketides with an Oxygen-Bridged Cyclooctadiene Core Skeleton from the Mangrove Endophytic Fungus Phomosis sp. A818

Plant endophytic microorganisms represent a largely untapped resource for new bioactive natural products. Eight polyketide natural products were isolated from a mangrove endophytic fungus Phomosis sp. A818. The structural elucidation of these compounds revealed that they share a distinct feature in their chemical structures, an oxygen-bridged cyclooctadiene core skeleton. The study on their structure–activity relationship showed that the α,β-unsaturated δ-lactone moiety, as exemplified in compounds 1 and 2, was critical to the cytotoxic activity of these compounds. In addition, compound 4 might be a potential agonist of AMPK (5′-adenosine monophosphate-activated protein kinase).


Introduction
Mangrove endophytic fungi are a rich source of structurally novel and biologically diverse natural products that could be useful in the development of new pharmaceutical agents [1,2]. During the course of our exploration for chemical constituents from the endophytic microorganisms of mangrove, we isolated a series of new compounds with various bioactivities [3][4][5][6][7][8]. Natural products containing an oxygen-bridged cyclooctadiene core skeleton and a δ-lactone moiety, as represented by mycoepoxydiene (MED), are recognized as a new class of fungal metabolites [9]. The distinct structural features of this group of compounds have attracted several research groups for total chemical syntheses [10,11]. Previous bioactivity studies suggested that MED could be a promising novel candidate for development of low toxicity antitumor agents [12,13], nonsteroidal anti-inflammatory drugs [14], and anti-atherosclerotic therapeutics [15].
We have been studying a MED-producing strain, Phomosis sp. A123, isolated from foliage of Kandelia candel (L.) Druce [9][10][11]. Despite the attractive features in structure and activity, MED research suffered from the low yield in strain A123. Subsequently, we used the genome shuffling approach to generate high MED-producing strains, which were screened by the high-throughput screening method, 'Antimicrobial-TLC-HPLC' (ATH) [16,17]. These efforts led to several high yield strains for MED analogs, including strain A818, in which the yield increased over 200-fold. The high yield strains produced new analogs in addition to MED. Herein, we describe the isolation, structure elucidation, and bioactivity test of eight natural products of the MED class from strain A818, which was derived from the mangrove endophytic fungus Phomosis sp. A123.
Compound 5 was obtained as colorless needle-shaped crystals, and the molecular formula was determined as C 14 H 14 O 3 by ESI-MS data ([M + H] + = 231.1015, calc. 231.1016). The DEPT spectrum revealed 14 signals: one carbonyl carbon signal, eight olefinic carbon signals, two methines bearing oxygen signals, two methines, and one methyl group (Table S2). Compared to the NMR data reported for compound 1893A previously isolated from the mangrove fungal 1893 [21], HMBC and 1 H-1 H COSY correlations are consistent with those data for 1893A ( Figure 1).
Compound 6 was obtained as a white amorphous solid, whose molecular formula was determined as C 14 H 14 O 3 by ESI-MS data ([M + H] + = 231.1013, calc. 231.1016). A comparison of 13 C spectra between 5 and 6 showed that 6 also had one carbonyl carbon signal, eight olefinic carbon signals, two methines bearing oxygen signals, two methines, and one methyl group (Table S2). The clear difference between these two structures was at the chemical shift of C-3, which shifted from δ144.3 to δ140.1. 1 H spectrum showed that the coupling constant of J 2,3 = 5.5 Hz of compound 5 changed to J 2,3 = 6.1 Hz of compound 6. A NOESY signal between H3 and H6 was observed, while no coupling signal was found between H-5 and Me-14 ( Figure 2, Table S2). Based on these differences and taking the relative configuration of compound 5 (1893A) into consideration, the structure of 6 was identified as the E-isomer of 1893A (phomoxydiene C) [20].  1 H and 13 C spectra of 4 shared an overall similarity with those of compound 1, suggesting that the two molecules were closely related. The only difference between 4 and 1 was the H-2/H-3 double bond in 1 was reduced to methylene groups in 4, showing that 4 shared the same structure as 2,3-dihydromycoepoxydiene ( Figure 1, Table S1) [18].
Compound 5 was obtained as colorless needle-shaped crystals, and the molecular formula was determined as C14H14O3 by ESI-MS data ([M + H] + = 231.1015, calc. 231.1016). The DEPT spectrum revealed 14 signals: one carbonyl carbon signal, eight olefinic carbon signals, two methines bearing oxygen signals, two methines, and one methyl group (Table S2). Compared to the NMR data reported for compound 1893A previously isolated from the mangrove fungal 1893 [21], HMBC and 1 H-1 H COSY correlations are consistent with those data for 1893A ( Figure 1).
Compound 6 was obtained as a white amorphous solid, whose molecular formula was determined as C14H14O3 by ESI-MS data ([M + H] + = 231.1013, calc. 231.1016). A comparison of 13 C spectra between 5 and 6 showed that 6 also had one carbonyl carbon signal, eight olefinic carbon signals, two methines bearing oxygen signals, two methines, and one methyl group (Table S2). The clear difference between these two structures was at the chemical shift of C-3, which shifted from δ144.3 to δ140.1. 1 H spectrum showed that the coupling constant of J2,3 = 5.5 Hz of compound 5 changed to J2,3 = 6.1 Hz of compound 6. A NOESY signal between H3 and H6 was observed, while no coupling signal was found between H-5 and Me-14 ( Figure 2, Table S2). Based on these differences and taking the relative configuration of compound 5 (1893A) into consideration, the structure of 6 was identified as the E-isomer of 1893A (phomoxydiene C) [20]. Compound 7 was obtained as colorless needle-shaped crystals, and its molecular formula of C16H20O5 was determined by ESI-MS ([M + H] + = 293.1381, calc. 293.1383). 1 H-and 13 C-NMR spectra of compound 7 were similar to those of 1893A, and the differences between these two compounds were a series of correlated saturated resonances (δH 2.48, 2.28, 2.06, 4.56, and 5.13), and an acyl methyl signal was observed instead of three olefinic group resonances at δH 7.68, 6.28, and 5.57 ppm in 1893A (Figure 1, Table S2). These spectroscopic data were identical to that of a previously reported compound 1893B [21].
Compound 8 was obtained as a white amorphous solid, and the molecular formula was C14H18O4 as established by HR-ESI-MS ([M + H] + = 251.1275, calc. 251.1277). The 1 H-and 13 C-NMR spectra of 8 were similar to those of 7, except the signals for the acetyl group of 7 ( Figure 1, Table S2). From the DEPT spectrum, 14 carbon signals were observed, including 1 carbonyl signal, 4 olefinic carbon signals, 4 oxygen-connected methines, 2 methines, 2 methylenes, and 1 methyl group. Both the 1 H-1 H COSY correlations and HMBC data supported the structure of 8 to be the deacetylated form of 1893B (phomoxydiene B) [20,21].

Cytotoxicity Study
Compounds 1−8 (except compound 3 because only a small amount was obtained) were studied for cytotoxic activity against MDA-MB-435 (a human breast cancer cell line) by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay [22,23]. As shown in Table S3, these polyketide metabolites displayed cytotoxic activity. The results were in agreement with the previous reported Compound 7 was obtained as colorless needle-shaped crystals, and its molecular formula of C 16 H 20 O 5 was determined by ESI-MS ([M + H] + = 293.1381, calc. 293.1383). 1 H-and 13 C-NMR spectra of compound 7 were similar to those of 1893A, and the differences between these two compounds were a series of correlated saturated resonances (δ H 2.48, 2.28, 2.06, 4.56, and 5.13), and an acyl methyl signal was observed instead of three olefinic group resonances at δ H 7.68, 6.28, and 5.57 ppm in 1893A ( Figure 1, Table S2). These spectroscopic data were identical to that of a previously reported compound 1893B [21].  Table S2). From the DEPT spectrum, 14 carbon signals were observed, including 1 carbonyl signal, 4 olefinic carbon signals, 4 oxygen-connected methines, 2 methines, 2 methylenes, and 1 methyl group. Both the 1 H-1 H COSY correlations and HMBC data supported the structure of 8 to be the deacetylated form of 1893B (phomoxydiene B) [20,21].

Cytotoxicity Study
Compounds 1−8 (except compound 3 because only a small amount was obtained) were studied for cytotoxic activity against MDA-MB-435 (a human breast cancer cell line) by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay [22,23]. As shown in Table S3, these polyketide metabolites displayed cytotoxic activity. The results were in agreement with the previous reported data which revealed that compound 1 exhibited more potent activity than compound 2, while compound 4 was inactive [18]. The result also indicated that α,β-saturated moiety was related to the loss of cytotoxic activity. Compounds 1 and 2 gave the IC 50 values of 7.85 and 14.61 µM, respectively, against MDA-MB-435, whereas the other compounds showed no significant activity, even at a high concentration (Table S3). Taking the structural features into consideration, the activity results indicate that the δ-lactone moiety may play a more critical role in the cytotoxicity than the γ-lactone in these metabolites.
Our previous studies showed that MED could induce the arrest of cell cycle and apoptosis [13]. MED could directly activate AMPK in vitro by AMPK kinase assay. However, we did not observe detectable activation of AMPK at the cellular level, by monitoring the phosphorylation of AMPKα and its substrate acetyl-CoA carboxylase (ACC) (data not shown). Indeed, we also found that MED induced the production of the intracellular Reactive Oxygen Species (ROS) in MEDA-MB-435 cells in a dose-dependent manner ( Figure 3A). Due to the role of ROS in AMPK activity [24], we speculate that ROS production may impair AMPK activation induced by MED.
Compound 4, which is inactive in cytotoxicity even at a high dose (up to 800 µM) against NIH/3T3 (mouse embryo fibroblast cell line) (data not shown), offered a new opportunity to study the AMPK activation by new MED analogs. The structure of compound 4 is very close to that of MED. Both contain an oxygen-bridged cyclooctadiene core skeleton and a δ-lactone moiety carrying an acetyl group. The only difference is a double bond in the lactone (α,β-unsaturated δ-lactone in MED). AMPK activation of compound 4 in NIH/3T3 cells was evaluated by monitoring the phosphorylation of AMPKα at Thr172 and its substrate acetyl-CoA carboxylase (ACC) at Ser79, with acadesine (AICAR) used as a positive control. Western blot analyses showed that compound 4 dramatically increased the phosphorylation of AMPKα and ACC ( Figure 3B). The phosphorylation of AMPKα and ACC reached a maximal level at a concentration of 400 µM, whereas no detectable activation was observed in the cells treated with higher concentrations (600 and 800 µM). Furthermore, the activation of AMPK by compound 4 was observed as early as 30 min at 100 µM, and reached a peak value around 60 min without affecting the total content of AMPK ( Figure 3C). In addition, compound 4 induced a significant increase in AMPK activation in 3T3-L1 adipocytes ( Figure 3D). These results indicated that compound 4 might be a potential AMPK agonist. data which revealed that compound 1 exhibited more potent activity than compound 2, while compound 4 was inactive [18]. The result also indicated that α,β-saturated moiety was related to the loss of cytotoxic activity. Compounds 1 and 2 gave the IC50 values of 7.85 and 14.61 µM, respectively, against MDA-MB-435, whereas the other compounds showed no significant activity, even at a high concentration (Table S3). Taking the structural features into consideration, the activity results indicate that the δ-lactone moiety may play a more critical role in the cytotoxicity than the γ-lactone in these metabolites.
Our previous studies showed that MED could induce the arrest of cell cycle and apoptosis [13]. MED could directly activate AMPK in vitro by AMPK kinase assay. However, we did not observe detectable activation of AMPK at the cellular level, by monitoring the phosphorylation of AMPKα and its substrate acetyl-CoA carboxylase (ACC) (data not shown). Indeed, we also found that MED induced the production of the intracellular Reactive Oxygen Species (ROS) in MEDA-MB-435 cells in a dose-dependent manner ( Figure 3A). Due to the role of ROS in AMPK activity [24], we speculate that ROS production may impair AMPK activation induced by MED.
Compound 4, which is inactive in cytotoxicity even at a high dose (up to 800 µM) against NIH/3T3 (mouse embryo fibroblast cell line) (data not shown), offered a new opportunity to study the AMPK activation by new MED analogs. The structure of compound 4 is very close to that of MED. Both contain an oxygen-bridged cyclooctadiene core skeleton and a δ-lactone moiety carrying an acetyl group. The only difference is a double bond in the lactone (α,β-unsaturated δ-lactone in MED). AMPK activation of compound 4 in NIH/3T3 cells was evaluated by monitoring the phosphorylation of AMPKα at Thr172 and its substrate acetyl-CoA carboxylase (ACC) at Ser79, with acadesine (AICAR) used as a positive control. Western blot analyses showed that compound 4 dramatically increased the phosphorylation of AMPKα and ACC ( Figure 3B). The phosphorylation of AMPKα and ACC reached a maximal level at a concentration of 400 µM, whereas no detectable activation was observed in the cells treated with higher concentrations (600 and 800 µM). Furthermore, the activation of AMPK by compound 4 was observed as early as 30 min at 100 µM, and reached a peak value around 60 min without affecting the total content of AMPK ( Figure 3C). In addition, compound 4 induced a significant increase in AMPK activation in 3T3-L1 adipocytes ( Figure 3D). These results indicated that compound 4 might be a potential AMPK agonist.

Discussion
The constant emergence of drug resistant diseases and pathogens demands the continuous search for new therapeutic agents. Ideally, the new agents should possess novel features in both chemistry and mode of action that are distinct from the existing drugs. Plant endophytic fungi are a huge source of bioactive natural products with novel structural features and biological activities. Natural products of the mycoepoxydiene family contain an unprecedented structure, 9-oxabicyclo[4.2.1]nona-2,4-diene skeleton. So far, the exact biosynthetic mechanism for this structure has not been reported, although it was thought to be of polyketide origin [9]. The mechanism by which the oxygen bridge is formed within the cyclooctadiene ring is particularly intriguing. The post-polyketide tailoring enzymes involved in the biosynthesis of this rare structural feature could have interesting new features. The isolation and structural elucidation of the new MED analogs provide a new opportunity for the understanding of the biosynthetic mechanism of this family of natural products. Furthermore, these compounds offer the opportunity to investigate the structure-activity relationship. The results revealed that the α,β-unsaturated δ-lactone moiety of the compounds is important for the cytotoxic activity against MDA-MB-435. In addition, compound 4 exhibited activation of AMPKα and ACC in NIH/3T3 and 3T3-L1 adipocytes, which indicated that compound 4 might be a potential AMPK agonist. The results suggest that MED analogs could provide lead compounds for structure modifications in development of new agents for AMPK activation. As a key player in the regulation of energy metabolism, AMPK is of central importance in energy metabolism related diseases. AMPK activators hold a great potential in treating metabolic diseases such as type 2 diabetes and obesity.

General Experimental Procedures
The structures were elucidated based on ESI-MS, 1D, and 2D NMR. NMR spectra were recorded on a Bruker Avance III-600 NMR spectrometer (Bruker, Billerica, MA, USA) with TMS as an internal standard. Mass spectrometry analysis was performed using an XTerraMS (Waters, Milford, MA, USA) equipped with an electrospray ionization (ESI) source. HR-ESI-MS data were acquired in m/z by using a BioTOF TM -Q mass spectrometer (Bruker, Billerica, MA, USA) and a Dionex Ultimate 3000 coupled to a Bruker Maxis Q-TOF.

Microbial Strains
Strain A123 was isolated from the foliage of Kandelia candel (L.) Druce, a mangrove plant in the Mangrove Nature Conservation Area of Fugong, Fujian Province, China. It was identified as a non-sporulating fungus by traditional morphology. By sequencing the ITS rDNA and comparing it with sequences in GenBank, strain A123 was identified as a Phomopsis sp., showing a 98% similarity to Phomopsis liquidambari (Accession No. AY 601919) [19]. Strain A818 was screened from a genome shuffling mutagenesis library of strain A123 with enhanced yield of mycoexydiene [16,17].

Culture Conditions and Extraction
All strains derived from Phomopsis sp. A123 were maintained on potato dextrose agar (PDA) slants containing 20% (v/v) stored sea water. For regular cultures, the strains on stock slants were inoculated on PDA plates containing 20% artificial sea water and allowed to grow for 14 days at 28 • C.
Strain A818 was cultured at 28 • C, 180 rpm with 150 L of PDB (Potato Dextrose Broth) medium containing 20% (v/v) stored sea water. After 14 d cultivation, the mycelia were separated from liquid with ultracentrifugation. The supernatant was collected and extracted with ethyl acetate for three times, ethyl acetate phase was collected together and concentrated under vacuum to afford 60 g of residue.

Cell Culture and Cytotoxicity Assays
NIH/3T3 (mouse embryo fibroblast cell line) was maintained in Dulbecco's minimal essential medium (DMEM, Gibico, Waltham, MA, USA) supplemented with 10% inactivated fetal bovine serum (FBS, Hyclone, Waltham, MA, USA). 3T3-L1 preadipocytes were cultured in DMEM supplemented with 10% calf serum (Hyclone). Differentiation was induced by treating the cells with differentiation inducers (DMEM) containing 0.5 mM 3-isobutyl-1-methylxanthane (IBMX), 0.25 µM dexa-methasone, 10 µg/mL insulin, and 10% fetal bovine serum for 72 h. More than 90% of the cells expressed the adipocyte phenotype between 8 and 10 days after initiation of differentiation and were used for the experiments. The cells were refed with DMEM supplemented with 10 µg/mL insulin and 10% FBS for the following 48 h and changed every two days. The cell lines were grown in logarithmic growth at 37 • C in a humidified atmosphere consisting of 5% CO 2 and 95% air.
The cytotoxicity was measured by the MTT (microculture tetrazolium [3-(4,5-dimethylthiazol-2-yl)-2,5]-diphenyl-tetrazalium bromide, Sigma-Aldrich, St. Louis, MO, USA) assay [22,23]. Briefly, the cells plated in the wells of 96-well plates (BD Biosciences, San Jose, CA, USA), were treated in triplicate with various concentrations of compounds for 72 h at 37 • C. After change fresh medium, a 20 µL aliquot of MTT solution (5 mg/mL) was added and incubated for 4 h at 37 • C. 100 µL of triplex solution (10% SDS, 5% isobutanol, 12 mM HCl) was added to each well and incubated overnight at 37 • C. The optical density of each well was measured with a microplate reader (M-3350, Bio-Rad, Hercules, CA, USA) at 595 nm. Growth inhibition rates were calculated with the following equation: The IC 50 was defined as the concentration of compound that resulted in a 50% inhibition of growth rate. Data were obtained from five different experiments and present as mean ± SD.