Cytotoxic Rocaglate Derivatives from Leaves of Aglaia perviridis

Rocaglates are a series of structurally complex secondary metabolites with considerable cytotoxicity that have been isolated from plants of the Aglaia genus (Meliaceae). A new rocaglate (aglapervirisin A, 1) and its eight new biosynthetic precursors of rocaglate (aglapervirisins B-J, 2–9) together with five known compounds, were isolated from the leaves of Aglaia perviridis. Their structures were elucidated based on a joint effort of spectroscopic methods [IR, UV, MS, ECD, 1D- and 2D-NMR, HRESIMS], chemical conversion and single-crystal X-ray diffraction. Among these isolates, three (1, 10–11) were silvestrols, a rare subtype rocaglates, exhibiting notable cytotoxicity against four human tumor cell lines, with IC50 values between 8.0 and 15.0 nM. Aglapervirisin A (1) induces cell cycle arrest at the G2/M-phase boundary at concentration 10 nM accompanied by reductions in the expression levels of Cdc2 and Cdc25C in HepG2 cells after 72h co-incubation, and further induces the apoptosis of HepG2 cells at concentrations over 160 nM.

aromatic protons of a monosubstituted benzene, two aromatic protons of a 1,2,3,5-tetrasubstituted benzene, and four methoxy groups. Its 1D-NMR (Table 1) data, particularly the three characteristic proton signals at δ H 5.11(1H, d, J = 7.0 Hz), δ H 3.91 (1H, dd, J = 14.5, 7.0 Hz), and δ H 4.32(1H, d, J = 14 Hz), featured a cyclopenta [b] benzofuran derivative nature of 1 23 . A series of proton signals at δ H 3.5 to 5.5, carbon signals at δ C 60.0 to 70.0, and two characteristic acetal carbons at δ C 93.7 and 95.3, suggested the presence of a dioxanyloxy unit in the structure of 1. The above-described analysis indicated that compound 1 was a silvestrol (10) analogue. Compared with 10, the presence of an acetyl group in 1 was evidenced by characteristic NMR signals (δ H 1.89, δ C 51.9, 171.7) and 42   13 C NMR spectra measured at 125 MHz; a was obtained in CDCl 3 . The assignments are based on the 2D-NMR spectra.
mass unit more than 10. The key HMBC correlation (Fig. 2) from H-6′′′ (δ H 4.08 and 3.99) to the acetyl group (δ C 20.5) indicated that the acetyl group was located at 6′′′-OH. The key ROESY correlations (Fig. 2) of H-1/H-2, H-2′ , 6′ and H-2/H-2′ , 6′ , H-2″ , 6″ , also observed for 10 23 , indicate that H-1, H-2, 1,4-disubstituted benzene and monosubstituted benzene were co-facial. Thus, the planar structure and relative configuration of 1 was determined. The similar ECD spectra of 1 and silvestrol (10) indicated that the absolute stereochemistry of the basic skeleton of 1 was the same as that of silvestrol. However, the absolute configuration of C-5′′′ in 1 was difficult to determine based on the ECD comparison. To completely determine the absolute configuration of 1, the acetyl derivative of 1 (1a), silvestrol (10a) and episilvestrol (11a) (see Figure S16) were prepared by acylation. The identical HRESIMS (1a, m/z 756.2857, [M + NH 4 ] + , 10a, m/z, 756.2859, [M + NH 4 ] + ), retention times in HPLC (1a, t R = 8.1 min, 10a, t R = 8.1 min, 70% MeOH-H 2 O), and the NMR data (see in SI), as well as the similar optical values obtained for 1a and 10a and the different retention times in HPLC (1a, t R = 8.1 min, 11a, t R = 9.3 min, 70% MeOH-H 2 O) and the NMR data (see in SI) obtained for 1a and 11a, indicate that the absolute configuration of 1 was consistent with that of silvestrol (10) 23 . The C-5′′′ in 1 was adopted as R based on the R configuration of C-5′′′ in compound 10 23 . Thus, the structure of 1 was determined, and named as aglapervirisin A. Aglapervirisin B (2) was obtained as colourless crystal ( α [ ] D 23 − 28.3) and gave a sodiated molecular ion [M + Na] + at m/z 675.2672 (calcd 675.2677) in the HRESIMS corresponding to a molecular formula of C 38 H 40 N 2 O 8 , which requires 20 indices of hydrogen deficiency. In the 1 H NMR spectrum, the 16 aromatic hydrogen signals in the low-field region (δ H 6.09-7.85) presented four benzene rings, including two monosubstituted benzene rings, a 1,4-disubstituted benzene ring, and a 1,2,3,5-tetrasubstituted benzene ring. Two amide protons at δ H 6.47 (NH-17) and 5.26 , and the remaining methylenes suggested the presence of a 1,4-butanediamide chain. These characteristic proton signals revealed that 2 was a cyclopenta[bc]benzopyran derivative 24-27 . In the HMBC spectrum (Fig. 3), a cross peak between signals at δ H 7.76 (H-20, 24) and δ C 167.6 (C-18) established the connection of one of the monosubstituted benzene rings to the butanediamide chain. The correlated peaks from H-3 to C-11 and from H-4 to C-1″ and C-2″ , 6″ allowed the placement of the 1, 4-butanediamide moiety and the second monosubstituted benzene ring at C-3 and C-4, respectively. Thus, the planar structure of 2 was depicted in Fig. 3. The relative configuration of 2 was determined according to the coupling constants between H-3 and H-4, and the cross peaks in the ROESY spectrum (Fig. 3). The vicinal coupling constant value between H-3 and H-4 was 6.5 Hz, allowed the assignment of H-3β and H-4α configuration [24][25][26][27] . The ROESY correlations between H-10 and H-2″ , 6″ and the lack of any correlation between H-3 and H-10 in compound 2 further confirmed the H-3β and H-4α. A single-crystal X-ray diffraction experiment (Fig. 4) using Cu Kα radiation was performed, and the absolute configuration of five asymmetric carbons in 2 was unambiguously established as 2R, 3S, 4R, 5R, and 6S. The absolute configuration of such a flavonol and diamide [3 + 2] adduct containing a benzoyl-1,4-butanediamide moiety was determined for the first time via the single-crystal X-ray diffraction method.    Table 2) of 3 were similar to those of 2, which indicated that 3 was also a cyclopenta[bc]benzopyran derivative with a benzoyl-1,4-butanediamide moiety 28 . However, the methine signals of H-10 at δ H 4.67 and C-10 (δ C 76.1) in 3 showed notable differences from those of 2 (δ H 4.19, δ C 82.6) ( Table 1), which combined with the observed obvious ROESY correlation between H-3 (δ H 3.51) and H-10 (δ H 4.68) in 3, absent in 2, revealed that 2 and 3 were epimers at C-10.
Aglapervirisin D (4), α [ ] D 23 − 0.3 (c 0.12, MeOH), was obtained as a white amorphous powder, and its molecular formula was elucidated to be C 38 H 38 N 2 O 8 (m/z 673.2516, [M + Na] + ) based on its 13 C NMR data and HRESIMS, with one more degree of unsaturation than 2. The similarity between the NMR data ( Table 2) of 4 and 2 and the key HMBC correlations between H-3/ C-2″ , 6″ and H-4/C-11 suggest that 4 was also a cyclopenta[bc] benzopyran 4,27,28 derivative, similar to 2. Compared with 2, the presence of an additional carbonyl group (δ C 209.5) and the absence of the characteristic signal of C-10 (δ H 4.19, δ C 82.6) indicated that 4 was an oxidation derivate at C-10 of 2. This deduction was confirmed by the HMBC correlations between H-4 (δ H 4.66) and C-10. Four sets of signals for benzene-ring, including one monosubstituted, two p-disubstituted, and one 1,2,3,5-tetrasubstituted benzene rings, three methoxyl groups in the 1 H NMR spectrum, and its similar HMBC correlations to those of 2 suggested that 5 was also a typical cyclopenta[bc]benzopyran 4,27 derivative. The 1D-NMR data ( Table 2) for 5 and 2 showed one more 4-hydroxybenzyl group [δ H 7.06 (2H, d, J = 8.5 Hz, δ C 129.4), δ H 6.84 (2H, d, J = 8.5 Hz, δ C 113.9)], and (δ H 3.45, δ C 43.1), in 5 than in 2, which suggested that a p-hydroxybenzyl group moiety presented in 5 other than a benzoyl group in 2. The obvious HMBC correlations from H-16 (δ H 2.95) and H-19 (δ H 3.45) to the amide carbonyl carbon (172.0, C-18) placed the p-hydroxybenzyl group at N-17. Thus, the planar structure of 5 was determined as depicted. In the 1 H NMR spectrum of 5 (Table 3), an extremely upfield shifted methoxyl signal was observed at δ H 3.09. The coupling constant value (8.5 Hz) of J (H-3,H-4) was also differ greatly from that of 2 (6.5 Hz). These data characterized H-3α and H-4β in 5 3,24,28 , respectively, in opposite to those in 2. A key ROESY correlation between H-10 (δ H 4.89) and H-3 (δ H 3.86) confirmed that 5 had opposite relative configurations at position C-3 and C-4 compared with 2, which also established an endo relationship between H-3 and H-10 28 . Thus, the structure of 5 was established as shown.
The molecular formula of aglapervirisin − 5.4 (c 0.28, MeOH)} was determined to be C 38 H 40 N 2 O 9 by HRESIMS (m/z 691.2625, [M + Na] + , calcd 691.2626), with one CH 2 unit less than 5. Its 1 H and 13 C NMR data, particularly the characteristic methoxyl signal at δ H 3.10 and the vicinal coupling constant value (8.5 Hz) of H-4/H-3, resemble greatly with those of 5, which indicated that the basic skeleton of 6 was the same as that of 5. The absence of a characteristic methine (δ H 3.45, δ C 43.1, C-19) indicated that the p-hydroxyphenylacetyl group at N-17 in 5 was replaced by a 4-hydroxybenzoyl group moiety in 6, as supported also by the observed HMBC correlations from H-16 (δ H 3.17) and H-20/24 (δ H 7.67) to C-18 (δ C 170.0). The ROESY correlations of H-10/H-3 indicated that the relative configuration of 6 was also the same as that of 5. Finally, the structure of 6 was assigned as shown.
Aglapervirisin G (7)   The key cross peak between H-10 and H-4 indicated an endo relationship between H-10 and H-4 28 . Thus, the structure of 7 was proposed as depicted.
Aglapervirisin H (8) was obtained as a colourless powder, α [ ] D 23 + 96.9 (c 0.10, MeOH), exhibited a sodicated molecular ion at m/z 600.2202 [M + Na] + (calcd for C 32 H 35 NO 9 Na, 600.2204) in the HRESIMS. The eleven characteristic aromatic protons at δ H (7.71−5.78) and 18 carbon signals observed in the 1 H and 13 C NMR data of 8, indicated that this compound had four aromatic protons and six aromatic carbons less than 5. The presence of a characteristic methoxyl at δ H 3.62 and obvious HMBC correlations from H-15 (δ H 1.94) and the methoxyl signal (δ H 3.62) to a carbonyl carbon at δ C 173.4 indicated that the benzoyl-1,4-butanediamide moiety in 2 was replaced by a 4-aminobutanoate methyl ester moiety in 8 3 . The key HMBC correlations (Fig. 5) of the resonances of H-3 (δ H 3.88) with C-11 (δ C 173.8) and H-4 (δ H 4.10) with C-2″ , 6″ (δ C 128.7) indicated that the basic connection of the core planar structure in 8 was the same as that of 5. The characteristic deshielded shift of the 6-OMe (δ H 3.08) signal and the coupling constant J (H-3,H-4) (8.5 Hz) indicated that the relative configurations of 8 at C-3 and C-4 were the same as those of 5 3,24,28 . The absence of a cross peak between H-10 (δ H 4.89) and H-3 (δ H 3.88) indicated   Table 3. 1H NMR and 13 C NMR Spectroscopic Data for Compounds 5-6 and 8-9. a,b 1 H NMR spectra measured at 500 MHz; 13 C NMR spectra measured at 125 MHz; a was obtained in CDCl 3; b was obtained in methanol-d 4 . The assignments are based on the 2D-NMR spectra. an exo relationship between H-10 and H-3 28 . The absolute configuration of 8 was assigned as 2R, 3R, 4S, 5R, and 6S based on the calculated ECD (Fig. 6) 13 C NMR data of 9 were similar to those of 8. The obvious differences between them were the presence of two multiplet protons at δ H 3.40 and the absence of a carbonyl carbon and a methoxy group in 9. The above information indicated that the methyl side chain was reduced to a hydroxymethyl group, as confirmed by the key HMBC correlations between H-15 (δ H 1.10) and C-16 (δ C 62.5). The coupling constant of J (H-3,H-4) (8.5 Hz) and characteristic deshielded shift of the 6-OMe (δ H 3.07) signal indicated that the planar structure and relative configuration of 9 were consistent with those of 8. Furthermore, the absolute configuration of 9 was elucidated as 2R, 3R, 4S, 5R, and 6S based on its ECD data, which was similar to that of 8. Thus, 9 was elucidated to be a new cyclopenta[bc] benzopyran derivative, and was named as aglapervirisin I.
Five known compounds were identified as silvestrol (10)
Cell cycle arrest is one of the important factors that influence the proliferation of tumour cells 29 . Thus, the regulation of the cell cycle of HepG2 cells by aglapervirisin A (1), a new silvestrol analogue possessing significant   cytotoxicity, was investigated through flow cytometry analysis. Treatment of HepG2 cells with 1 at concentrations of 40 or 160 nM for 48 h increased percentage of HepG2 cells arrested at the G2/M boundary from 10% to 25% or 35%, respectively. These results indicated that 1 can induce the arrest of HepG2 cells at the G2/M boundary in a dose-dependent manner, similarly to hydroxycamptothecin 14,15 , as shown in Fig. 7. The western blotting results indicated that 1 down-regulates the expression of Cdc2 and Cdc25C, which may be responsible for the induction of G2/M arrest.

Aglapervirisin A (1) causes HepG2 cell death by apoptosis. The induction of apoptosis is also a mech-
anism through which antitumour agents exert their therapeutic effects 30,31 . Thus, the induction of apoptosis by 1 was also examined in this study (Fig. 7). The assay showed that incubation with 1 for 72 h induced the apoptosis of HepG2 cells in a dose-dependent manner. The percentages of apoptotic cells were 6.2% and 14.5% after treatment with 1 at concentrations of 160 and 2560 nM, respectively, as compared with 0.6% in the negative control group. Thus, these results demonstrated that 1 induced HepG2 cell death by apoptosis (Fig. 8).

Conclusion
In summary, this paper describes the isolation and structural elucidation of a rare silvestrol analogue (aglapervirisin A, 1), eight new biosynthetic precursors of rocaglate (cyclopenta[bc]benzopyrans, aglapervirisins A-H, 2-9), and five known analogues isolated from the leaves of A. perviridis. The absolute configurations of 1, 2, and 8 were confirmed through chemical conversion, single-crystal X-ray diffraction, and quantum chemical calculations of ECD, respectively. The new silvestrol analogue 1 showed significant cytotoxicity at the nanomolar level against several cancer cell lines and a further mechanistic study indicated that this cytotoxicity was associated with the induction of G2/M phase arrest through reductions in the expression levels of Cdc2 and Cdc25C and the induction of apoptosis. Taken together, these findings indicate that the isolated compounds are potential natural prodrugs with anticancer activity.

Methods
General Experimental Procedures. The optical rotations were measured on a JASCO P-1020 polarimeter at room temperature. The melting points were measured using an X-4 digital display micromelting apparatus and are uncorrected. The IR spectra were recorded on a Bruker Tensor 27 spectrometer using KBr pellets. The  tested with a diffractometer using Olex2 32 , and the structure was solved through direct methods with the ShelXS 33 structure solution program and refined with the ShelXL 33 refinement package using least squares minimization. The crystallographic data for 2 were deposited in the Cambridge Crystallographic Data Centre (deposition number CCDC 1042328 Theoretical calculated and experimental observed ECD of 8. The phenomena of ECD have been extensive applied in the determination of the absolute configurations of natural chiral molecules 34 . The conformations were generated and optimized by Gaussian 09 package 35 . For the TD calculations Gaussian 09 was used. TDDFT calculations employed the B3LYP functional and the 6-311+ g (d, 2p) basis set (Nstates = 40, root = 3). An overall ECD spectrum was generated on the basis of Boltzmann weighting of 5 individual conformers (SI- Table 1) applying a shift based upon the difference between observed and calculated UV spectra. Comparisons of the experimental and calculated spectra were done using SpecDis with UV shift (36 nm) and a half-bandwidth of 0.21 eV. Finally, the calculated ECD spectrum of conformer 4 was adjacent to experimental ECD data. Through comparison with the experimental ECD of 8, the absolute configurations of 8 were assigned as 2R, 3R, 4S, 5R, and 6S, respectively. Determination of Cytotoxic Activities. As reported recently in the literatures, rocaglates exert significant cytotoxic activities [24][25][26][27][28] . In this investigation, the IC 50 values in human leukemic (HL-60), colon cancer (HT-29), human breast cancer (MCF-7), and human liver hepatocellular carcinoma (HepG2) cell lines were determined to evaluate the cytotoxicity of the isolated compounds. The isolated compounds were evaluated based on their cytotoxic activities against the above-mentioned tumor cell lines. The silvestrol analogues exhibited more potent cytotoxic activity than the cyclopenta[bc]benzopyrans, as shown in Table 4. The cytotoxicity assay used in this study is based on the MTT method and was performed in 96-well microplates 36 . The cells were cultured in DMEM medium (Hyclone, Logan, UT, USA) with 10% foetal bovine serum in an atmosphere with 5% CO 2 at 37 °C prior to the assay. Then, 150 μL of the cell suspension was seeded into each well of 96-well the cell culture plates, and the cells were allowed to adhere for 12 h before testing. The initial density of the cells was 10 5 /mL. Each tumor cell line was exposed to each test compound at concentrations of 0.001, 0.01, 0.1, 1, and 10 μM in triplicate for 48 h, and paclitaxel and cis-platinum were used as positive controls. Using the Reed-Muench method, the IC 50 values were calculated based on the obatained cell viability 37 .
Statistical analysis. Statistical analysis of the data was processed with GraphPad Prism 4.0 software. Statistical analysis of the data was expressed as mean ± SD. Values were analyzed by one-way analysis of variance (ANOVA) using SPSS version 12.0 software. p < 0.05 were considered statistically significant.