Structurally Modified Cyclopenta[b]benzofuran Analogues Isolated from Aglaia perviridis

Abstract

Four new cyclopenta[b]benzofuran derivatives based on an unprecedented carbon skeleton (1–4), with a dihydrofuran ring fused to dioxanyl and aryl rings, along with a new structural analogue (5) of 5‴-episilvestrol (episilvestrol, 7), were isolated from an aqueous extract of a large-scale recollection of the roots of Aglaia perviridis collected in Vietnam. Compound 5 demonstrated mutarotation in solution due to the presence of a hydroxy group at C-2‴, leading to the isolation of a racemic mixture, despite being purified on a chiral-phase HPLC column. Silvestrol (6) and episilvestrol (7) were isolated from the most potently cytotoxic chloroform sub-fraction of the roots. All new structures were elucidated using 1D- and 2D-NMR, HRESIMS, IR, UV, and ECD spectroscopic data. Of the five newly isolated compounds, only compound 5 exhibited cytotoxic activity against a human colon cancer (HT-29) and human prostate cancer cell line (PC-3), with IC₅₀ values of 2.3 μM in both cases. The isolated compounds (1-5) double the number of dioxanyl ring-containing rocaglate analogues reported to date from Aglaia species and present additional information on the structural requirements for cancer cell line cytotoxicity within this compound class.

Graphical Abstract

Over the last four decades, several species of the genus Aglaia, belonging to the plant family Meliaceae, have been investigated for their phytochemical composition, resulting in the isolation of cyclopenta[b]benzofuran compounds (flavaglines) such as rocaglamide,¹ didesmethylrocaglamide,² silvestrol (6),^3–6 and various thapkapsins and thapoxepines.^7,8 These phytochemicals are of considerable interest due to their potential biological activities such as antibacterial, antitumor, antiviral, insecticidal, and neuroprotective effects.^9–15 In particular, rocaglamide, silvestrol (6), and its structural analogue 5‴-episilvestrol (episilvestrol, 7) have been found to act as potent apoptotic and cytostatic agents^9,16,17 leading to a growing interest in their biological activity and total synthesis. Several groups have worked successfully on the complete enantioselective synthesis of silvestrol and episilvestrol, which contain a dioxanyl ring in their structures,^18–22 enabling a better understanding of the structure-activity relationship of flavaglines and their possible development as therapeutic leads, such as against B-cell malignancies.^7,9 Mechanistic investigations of the flavaglines have demonstrated them to arrest the G2/M phase of the cell cycle,^10,23 with rocaglamide acting as a prohibitin (PHB) 1 and 2 inhibitor,^24,25 and silvestrol (6) and episilvestrol (7) determined as eIF4A protein translation inhibitors.^16,22,26 A recent study further evaluated the structural mechanism of rocaglamide as a translation initiation inhibitor where the ATP-independent high structural selectivity of these compounds for the molecular interface formed between the eIF4A1 protein and the polypurine bases on mRNA was shown.²⁵ On the other hand, it was demonstrated that silvestrol acts as an FMS-like tyrosine kinase receptor-3 (FLT3) translation and protein expression inhibitor, a protein found in increased amounts in acute myeloid leukemia (AML) patients.²⁷ Further studies have been done to evaluate the antineoplastic efficacy of this rocaglate analogue, in several in vitro cancer cell assays^3–8 and in vivo activity against various cancer cells, inclusive of acute lymphocytic leukemia (ALL), AML, Epstein-Barr virus (EBV)-driven lymphoma, and mantle cell lymphoma.^13,28Additionally, analogues of silvestrol antibody-drug conjugates (ADCs) were developed recently, of which an aminosilvestrol ADC displayed specificity and in vivo efficacy against CD-22-expressing B-cell malignant tumor cells.²⁹

Phytochemical analysis of more than 60 Aglaia spp. thus far has led to the isolation of five dioxanyl ring-bearing bioactive cyclopenta[b]benzofuran analogues from only three Aglaia species. All these biosynthesized natural dioxanyl ring-containing flavaglines have shown micromolar to nanomolar activity against various cancer cell lines.^3,5,6 Previous work on the leaves, twigs, fruits, bark, and roots of Aglaia perviridis Hiern, a plant native to tropical forests of South and South-eastern Asia, has been found to produce secondary metabolites belonging to the bisamide, cyclopenta[b]benzofuran, cyclopenta[b]benzopyran, lignan, sesquiterpenoid, sterol, and triterpenoid structural classes.^6,30–34 Some of these compounds have exhibited in vitro cytotoxic activity against liver⁶ and colon cancer³³ cell lines with IC₅₀ values in the high nanomolar range.

In our continued efforts to discover antineoplastic compounds from higher plants,^13,28 the chloroform and ethyl acetate partitions of the roots of A. perviridis were studied for their phytochemical composition and evaluated against the HT-29 colon cancer cell line for activity. Bioassay-guided fractionation of the chloroform partition led to the isolation of silvestrol (6),³ and 5‴-episilvestrol (episilvestrol, 7)³ from the most active sub-fractions. Furthermore, ¹H NMR guided prioritization of fractions obtained from the ethyl acetate partition of the aqueous root extract led to the isolation of four modified cyclopenta[b]benzofuran structures (1-4) and a new silvestrol analogue (5). Compounds 1-4 are representative of a new carbon skeleton with a fused 2,3-dihydrobenzofuran unit attached to the dioxanyl ring, and compound 5 was observed to be a hybrid derivative of silvestrol and didesmethylrocaglamide, another flavagline isolated previously from A. argentea and A. perviridis.^2,33 All isolates were tested for their cytotoxic activity against HT-29 and PC-3 cells, with only compound 5 showing activity (IC₅₀ of 2.3 μM for both cell lines).

RESULTS AND DISCUSSION

Compound 1, isolated as a colorless solid, showed a sodiated molecular ion peak at m/z 630.1946 [M+Na]⁺, corresponding to a molecular formula of C₃₂H₃₃NO₁₁Na, as displayed in the high-resolution electrospray ionization mass spectrum (HRESIMS). The ¹H NMR spectrum of 1 showed the characteristic signals of a para-disubstituted benzene ring at δ_H 7.15 (2H, d, J = 8.9 Hz, H-2′ and H-6′) and 6.61 (2H, d, J = 8.9 Hz, H-3′ and H-5′), and a multiplet at δ_H 7.03 (5H, m, H-2″, H-3″, H-4″, H-5″, and H-6″) for a monosubstituted benzene ring. These observations combined with a set of doublets at δ_H 4.73 (1H, d, J = 5.3 Hz, H-1) and 4.39 (1H, d, J = 14.2 Hz, H-3), along with a doublet of doublets at δ_H 3.95 (1H, dd, J = 14.2, 5.4 Hz, H-2) for the (-OCH-CH-CH-) spin system of ring D (Figure 1), were consistent with the presence of a cyclopenta[b]benzofuran skeleton, characteristic of such compounds in Aglaia species.^3,5

Figure 1.

Structures of compounds 1–7 from Aglaia perviridis

The ¹H NMR chemical shifts and splitting patterns for two methylenes [δ_H 3.88 (1H, dd, J = 11.7, 9.4 Hz, H-3‴a), 3.82 (1H, dd, J = 11.7, 3.0 Hz, H-3‴b), 3.73 (1H, dd, J = 10.4, 5.2 Hz, H-6‴a), and 3.64 (1H, dd, J = 10.6, 6.1 Hz, H-6‴b), and two methine protons [δ_H 4.13 (1H, dt, J = 9.3, 3.3 Hz, H-4‴), 3.68 (1H, td, J = 5.7, 3.5 Hz, H-5‴)], with their corresponding oxygenated carbon signals [δ_C 62.9 (C-3‴), 69.9 (C-4‴), 71.4 (C-5‴), and 62.1 (C-6‴)] in the ¹³C NMR spectroscopic data (Tables 1 and 2) were similar to those of a 1,4 dioxanyl ring system, as seen in the structure of silvestrol (6).^3,5 The ¹H and ¹³C NMR data of 1 (Tables 1 and 2) displayed signals for only two methoxy groups [δ_H 3.85 (3H, OCH₃-8) and 3.66 (3H, OCH₃-4′)] and four quaternary carbons, respectively, instead of four methoxy and three quaternary carbon signals for silvestrol (6),⁵ showing the structural differences of 1 and 6. The presence of a primary amide was ascertained based on the MS data and the ¹³C NMR signal, at δ_C 174.2 ppm, in compound 1. ¹H-¹³C HMBC correlations observed between δ_C 174.2 and δ_H at 3.95 (1H, dd, J = 14.2, 5.4 Hz, H-2) led to the assignment of the amide bonded to C-2, rather than a methoxycarbonyl group as in silvestrol (6). Further analysis of the HMBC spectrum showed correlations between δ_H 6.19 (1H, s, H-5) and the carbons at δ_C 160.4 (C-4a), 162.6 (C-6), 102.6 (C-7), and 108.4 (C-8a).

Table 1.

¹H NMR Spectroscopic Data of Compounds 1–5 (700 MHz, methanol-d₄)^a

proton	1	2	3	4	5 (major)	5 (minor)
1	4.73, d (5.3)	4.74, d (5.3)	4.75, d (5.4)	4.75, d (5.4)	4.81, d (5.8)	4.81, d (5.8)
2	3.95, dd (14.2, 5.4)	3.95, dd (14.1, 5.4)	3.91, dd (14.2, 5.4)	3.91, dd (14.2, 5.4)	3.91, dd (14.2, 5.8)	3.91, dd (14.2, 5.8)
3	4.39, d (14.2)	4.39, d (14.2)	4.37, d (14.2)	4.37, d (14.2)	4.31, d (14.2)	4.31, d (14.2)
5	6.19, s	6.19, s	6.23, s	6.23, s	6.44, d (1.9)	6.51, d (1.9)
7	-	-	-	-	6.37, d (1.9)	6.45, d (1.9)
2′, 6′	7.15, d (8.9)	7.15, d (9.0)	7.14, d (9.0)	7.14, d (9.0)	7.16, d (8.9)	7.16, d (8.9)
3′,5′	6.61, d (8.9)	6.61, d (9.0)	6.62, d (9.0)	6.62, d (9.0)	6.64, d (8.9)	6.64, d (8.9)
2″,6″	7.03, m	7.04, m	7.03, m	7.03, m	7.03, m	7.03, m
3″,5″
4″
1‴	5.77, d (4.0)	5.76, d (4.1)	5.66, d (3.8)	5.66, d (3.9)	5.26, s	5.29, br s
2‴	5.03, d (4.0)	5.05, d (4.1)	5.05, d (3.8)	5.05, d (3.9)	4.97, s	4.89, d (1.6)
3‴Ha	3.88, dd (11.7, 9.4)	3.78, dd (11.6, 8.6)	3.84, dd (11.6, 9.1)	3.95, dd (11.7, 3.0)	4.19, t (11.1)	4.16, dd (11.1, 2.7)
3‴Hb	3.82, dd (11.7, 3.0)	3.76, dd (11.3, 3.5)	3.82, dd (11.6, 3.3)	3.72, dd (11.7, 9.1)	3.74, dd (11.6, 2.5)	3.84, br d (11.0)
4‴	4.13, dt (9.3, 3.3)	3.99, ddd (9.3, 6.8, 2.9)	4.15, dt (9.1, 3.5)	4.01, ddd (9.2, 6.4, 2.7)	4.08, ddd (11.0, 7.3, 2.7)	4.01, ddd (10.5, 6.7, 2.7)
5‴	3.68, td (5.7, 3.5)	3.68, td (6.4, 3.7)	3.68, ddd (6.1, 5.8, 2.9)	3.65, td (6.9, 2.9)	3.6, m	3.6, m
6‴Ha	3.73, dd (10.4, 5.2)	3.94, m	3.73, dd (10.6, 5.2)	3.73, dd (10.2, 2.6)	3.71, dd (11.4, 3.6)	3.69, dd (11.4, 3.7)
6‴Hb	3.64, dd (10.6, 6.1)	3.64, dd (11.2, 6.0)	3.63, dd (10.5, 6.1)	3.63, dd (10.3, 3.8)	3.43, dd (11.4, 6.7)	3.47, dd (6.7, 11.4)
OMe-4′	3.66, s	3.66, s	3.67, s	3.67, s	3.68, s	3.68, s
OMe-8	3.85, s	3.85, s	4.13, s	4.13, s	3.87, s	3.87, s

^aChemical shifts are presented in ppm. J values are given in Hz in parentheses. Assignments are based on ¹H-¹H COSY and HMBC spectra.

Table 2.

¹³C NMR Spectroscopic Data of Compounds 1–5 (175 MHz, methanol-d₄)^a

position	1	2	3	4	5 (major)	5 (minor)
1	78.9, CH	78.9, CH	79.1, CH	79.7, CH	79.2, CH	79.2, CH
2	51.2, CH	51.2, CH	51.3, CH	51.5, CH	50.9, CH	50.9, CH
3	55.6, CH	55.5, CH	55.6, CH	55.5, CH	55.5, CH	55.5, CH
3a	101, C	102.8, C	101.8, C	102.2, C	101.3, C	101.3, C
4a	160.4, C	162.6, C	162.5, C	163, C	160.1, C	160.8, C
5	86.9, CH	86.9, CH	87.1, CH	86.7, CH	91.6, CH	92.2, CH
6	162.6, C	162.6, C	162.5, C	163, C	160.8, C	160.8, C
7	102.6, C	100.8, C	105.7, C	105.7, C	93.5, CH	93.9, CH
8	157.3, C	160.4, C	155.7, C	155.7, C	159, C	158.3, C
8a	108.4, C	108.4, C	109.8, C	109.1, C	109.5, C	109.6, C
8b	93.7, C	93.7, C	93.6, C	93.5, C	93.5, C	93.7, C
1′	128.1, C	128.1, C	128.1, C	127.9, C	128, C	128, C
2′, 6′	128.8, CH	128.8, CH	128.7, CH	128.7, CH	128.9, CH	128.9, CH
3′,5′	111.7, CH	111.7, CH	111.7, CH	110.9, CH	111.7, CH	111.7, CH
4′	158.4, C	158.4, C	158.4, C	157.6, C	158.7, C	158.7, C
1″	137.8, C	137.8, C	137.7, C	137.9, C	137.7, C	137.7, C
2″,6″	128, CH	128, CH	128, CH	128, CH	128, CH	128.1, CH
3″,5″	127.1, CH	127.1, CH	127.1, CH	127.2, CH	127.1, CH	127.1, CH
4″	125.8, C	125.8, C	125.8, C	125.9, C	125.8, C	125.8, C
1‴	102.9, CH	102.7, CH	102.1, CH	101.9, CH	95.1, CH	94.9, CH
2‴	68.7, CH	68.9, CH	70.1, CH	70.2, CH	88.9, CH	91.7, CH
3‴	62.9, CH2	62.7, CH2	63.2, CH2	63.2, CH2	59.4, CH2	67.3, CH2
4‴	69.9, CH	69.6, CH	69.7, CH	69.5, CH	67.6, CH	67, CH
5‴	71.4, CH	71.8, CH	71.1, CH	71.6, CH	72, CH	71.7, CH
6‴	62.1, CH2	62.5, CH2	62.5, CH2	62.5, CH2	62.9, CH2	63.1, CH2
OMe-4′	54, CH3	54, CH3	54, CH3	53.8, CH3	53.8, CH3	53.9, CH3
OMe-8	54.8, CH3	54.8, CH3	57.7, CH3	57.5, CH3	54.7, CH3	54.7, CH3
CONH2−2	174.2, C	174.3, C	174.2, C	174.2, C	174.4, C	174.4, C

^aAssignments are based on HSQC and HMBC experiments.

Moreover, HMBC correlations between δ_H 5.03 (1H, d, J = 4.0 Hz, H-2‴) and δ_C 102.6 (C-7), 157.3 (C-8), and 162.6 (C-6), as shown in Figure 2, in addition to the 2D ¹H-¹H COSY, HSQC, and other HMBC correlations, suggested the presence of an unusual 2,3-dihydrofuran ring connection between the flavagline core and the dioxanyl moiety (Figure 1). Additionally, the coupling constant of 3.8 Hz for the two doublets at δ_H 5.77 (1H, d, H-1‴) and 5.03 (1H, d, H-2‴) indicated H-1‴ and H-2‴ to be co-facial and rings A and B to be cis-fused,³⁵ and was confirmed by the NOESY correlations (Figure 2) observed for H-1‴, H-2‴ and OMe-8.

Figure 2.

Select (A) HMBC and (B) NOESY correlations of 1

Further perusal of the ¹H NMR spectrum of 1 indicated similar chemical shifts and splitting patterns for the other protons of the dioxanyl ring [H-3a‴ (δ_H 3.88, 1H, dd, J = 11.7, 9.4 Hz), H-3b‴ (δ_H 3.82, 1H, dd, J = 11.7, 3.0 Hz), H-4‴ (δ_H 4.13, 1H, dt, J = 9.3, 3.3 Hz) and H-5‴ (δ_H 3.68, 1H, td, J = 5.7, 3.5 Hz)] to those found in silvestrol (6).^3,5 This suggested the configurations of C-4‴ and C-5‴ to be the same as in 6, and was consistent with the observed NOESY correlations (Figure 2).

The absolute configuration for most of the flavagline compound class has been predominantly determined based on the comparison of ECD and NMR spectroscopic data with the known compounds rocaglamide and silvestrol, for which the structures were determined unambiguously using X-ray crystallography. Owing to the small amounts of the isolated compounds, ECD calculation was used for the assignment of the absolute configurations of compound 1. Computational ECD calculations were performed on two different isomers 1‴S, 2‴R, 5‴R and 1‴R, 2‴S, 5‴S. The ECD spectra for both the isomers along with the experimental ECD spectrum of compound 1 are shown in Figure 3. After a blue shift of the computed spectra by ~10 nm, both peak sign and peak positions of the calculated spectra for 1‴S, 2‴R, 5‴R were observed to match those of the experimental data for compound 1. A wavelength shift is commonly applied during the computation/experiment comparison to account for any systematic over/underestimation of transition energies due to functional(s), basis set or solvation model used in calculation.³⁶ Particularly, the two negative Cotton effects at ~215 and ~235 nm (blue line in Figure 3A) that match the experimental data for compound 1 (black line in Figure 3A) are absent from the spectrum of (1‴R, 2‴S, 5‴S) isomer (red line in Figure 3A). The ~215 nm Cotton effect seems further blue-shifted by 5 nm while the ~235 nm Cotton effect is of positive sign in the calculated spectrum of the (1‴R, 2‴S, 5‴S) isomer. Notably, an additional broad and low-amplitude Cotton effect at 275 nm was observed in the experimental spectrum, shown to be characteristic of the rocaglate core,^3,37 which was not reflected in the computed data, possibly due to the low amplitude. Taken together, based on the comparison of the computed and experimental ECDs, a good match was obtained between the (1‴S, 2‴R, 5‴R) diastereomer and compound 1. Thus, compound 1 was defined as (1R,2R,3S,3aR,8bS,1‴S,2‴R,4‴R)-4‴-[(R)-1,2-dihydroxyethyl]-1,8b-dihydroxy-8-methoxy-3a-(4-methoxyphenyl)-3-phenyl-1,2,3a,8b,1‴,2‴,3‴,4‴-octahydro-8H-cyclopenta[4,5]furo[3,2-f][1,4]dioxino[2,3-b]benzofuran-2-carboxamide.

Figure 3.

Overlay of the experimental spectrum of compound 1 (black line) with the (A) calculated spectra of the (1‴S, 2‴R, 5‴R) (blue) and (1‴R, 2‴S, 5‴S) (red) diastereomers, TD-DFT calculated at the wB97XD/Def2TZVP level. (B) TD-DFT calculated ECD spectra; functional sets are noted in the legend and all calculations were ran with the Def2TZVP basis set

Compounds 2-4 had the same molecular formula (C₃₂H₃₃NO₁₁Na), as compound 1, as determined by the sodiated molecular ion peaks at m/z 630.1947 (2), 630.1956 (3), and 630.1958 (4) ([M+Na]⁺, calcd for C₃₂H₃₃NO₁₁Na, 630.1951). The ¹H NMR (Table 1) and ¹³C NMR (Table 2) data of these molecules suggested the presence of a 2,3-dihydrofuran[1,4]dioxane unit fused to the cyclopenta[b]benzofuran core, as deduced for compound 1. A primary amide chemical shift at δ_C 174.2 attached to C-2 was observed for all three compounds as well. Comparable chemical shifts for H-1‴ and H-2‴ as doublets with a coupling constant of ca. 3.9 Hz in the ¹H NMR spectra of 2–4, suggested these two protons to be co-facial in all three compounds. This was substantiated by the observed NOE correlations between H-1‴/H-2‴ and OCH₃-8/H-2‴, as seen in compound 1. However, the ¹H and ¹³C NMR chemical shifts and splitting patterns for the dioxanyl ring protons (C/H-1‴ to C/H-6‴) were different from 1 (Tables 1 and 2), indicating compounds 2-4 to be structural isomers of compound 1. For compound 2, the splitting pattern and coupling constants of H-1‴ (δ_H 5.76, d, J = 4.1 Hz), H-2‴ (δ_H 5.05, d, J = 4.1 Hz), H-3a‴ (δ_H 3.78, dd, J = 11.6, 8.6 Hz), and H-3b‴ (δ_H 3.76, dd, J = 11.3, 3.5 Hz) were similar to those of compound 1, suggesting the same configurations for C-1‴ (S), C-2‴ (R), and C-4‴ (R). Thus, the only difference between compounds 1 and 2 was the different configuration at C-5‴, as supported by the different chemical shifts, splitting pattern, and coupling constants of H-4‴ (δ_H 3.99, ddd, J = 9.3, 6.8, 2.9 Hz), H-5‴ (δ_H 3.68, td, J = 6.4, 3.7 Hz), H-6‴a (δ_H 3.94, m), and H-6‴b (δ_H 3.64, dd, J = 11.2, 6.0 Hz), resembling the observed shifts for 5‴-episilvestrol (7).³ The relative configurations of 3 and 4 were also assigned in a similar fashion, based on the observed NOESY correlations for H-1‴/ H-2‴/ H-3‴a/ H-3‴b/ H-4‴, and H-5‴ and comparison with the published values of silvestrol and its natural analogues, 2‴-, 5‴-, and 2‴,5‴-diepisilvestrol.^3,5,6,21 Both compounds 3 and 4 had different chemical shifts and NOESY correlations from 1 for H-1‴ (3 δ_H 5.66, d, J = 3.8 Hz; 4 δ_H 5.66, d, J = 3.9 Hz), and H-2‴ (3 δ_H 5.05, d, J = 3.8 Hz; 4 δ_H 5.05,d, J = 3.9 Hz), suggesting the configurations at C-1‴ and C-2‴ to be R and S, respectively. The signals for H-3‴ of both these molecules were similar to 1 and 2, indicating a 4‴R absolute configuration for all four compounds. However, the H-4‴ signals for 3 (δ_H 4.15, dt, J = 9.1, 3.5 Hz) were similar to 1 and silvestrol, while that of 4 (δ_H 4.01, ddd, J = 9.2, 6.4, 2.7 Hz) was similar to 2 and 5‴-episilvestrol (7), thereby rendering these 5‴-epimers (3: R and 4: S). The ECD spectroscopic data for molecules 2-4 were also consistent with the above deductions and comparable to each other. Thus, the structures of compounds 2, 3, and 4 were assigned, in turn, as the C-5‴-epimer, the C-1‴, C-2‴ di-epimer, and the C-1‴, C-2‴, C-5‴ tri-epimer of 1, respectively.

For compound 5, isolated as a colorless amorphous powder, a sodiated molecular ion peak m/z 648.2052 [M+Na]⁺ (calcd for C₃₂H₃₅NO₁₂Na) was observed in the HRESIMS data. Analysis of the ¹H NMR spectrum of 5 showed the presence of another set of protons attributed to a minor compound with comparable shifts to those of the major compound (2:1 ratio in solution). Further perusal of the ¹H and ¹³C NMR spectra (Tables 1 and 2) showed the structure of 5 (both major and minor) to be related to known cyclopenta[b]benzofurans, particularly silvestrol (6).^3,5 For both the compounds, the ¹H NMR spectrum showed similar chemical shifts and splitting patterns for H-1, H-2, and H-3, H-1′ to H-6′ and H-1″ to H-6″, as observed for known rocaglates.^3,7,38 Thus, these positions were assigned the same configurations (1R, 2R, 3S, 3aR, 8bS) as silvestrol based on the 1D and 2D NMR data (particularly the NOESY NMR spectrum, Figure 4). Similar to the previous isolates, the chemical shift at δ_C 174.4 showed a HMBC correlation with H-2 (δ_H 1H, dd, 3.91, J = 14.2, 5.8 Hz), and, in combination with the HRESIMS data, could be attributed to the presence of an amide group as opposed to an ester group at C-2 occurring in known dioxanyl ring-containing rocaglates.^3,5,6 Signals for multiple oxygenated methine and methylene signals were observed for a 1,4-dioxanyl ring, consistent with this compound being a silvestrol analogue. A key difference in compound 5, as observed for compounds 1-4, was the presence of only two methoxy group signals at δ_H 3.87 (OCH₃-8) and 3.68 (OCH₃-4′) instead of four methoxy group signals observed for 6.^3,5 Furthermore, the upfield shift of ca. 10 ppm for C-2‴ at δ_C 88.9 (H-2‴ 1H, s, 4.97) suggested a hydroxy group at C-2‴ in 5, instead of a methoxy group as observed for silvestrol (6).³

Figure 4.

Selected COSY and HMBC correlations of 5

The key difference between the two compounds was the observed splitting patterns and coupling constants for H-1‴ (major: δ_H 5.26, 1H, s; minor: δ_H 5.29, br s, 1H) and H-2‴ (major: δ_H 4.97, 1H, s; minor: δ_H 4.89, dd, 1H, J = 1.7 Hz). In association with the ¹³C NMR data and by comparison with the known 2‴, 5‴-di-epimer of silvestrol from A. foveolata,⁵ the two compounds were suggested to be C-2‴ epimers. Accordingly, the major compound was determined to have a β-orientation of H-2‴, with an α-orientation for the minor compound, as substantiated by the small coupling constant of 1.6 Hz for H-1‴ and H-2‴.⁵ The ¹H NMR spectrum along with HPLC analysis suggested compound 5 to show mutarotation at C-2‴. The comparable splitting patterns and coupling constants with those of 5‴- and 2‴, 5‴-episilvestrol led to the conformational assignment for the major (β) and minor (α) compounds as shown, consistent with the 2D NOESY spectrum observed. The ECD spectrum of compound 5 was similar to that of silvestrol (Figure 5), for which a characteristic negative Cotton effect between 215 and 220 nm was observed.

Figure 5.

Comparison of the ECD spectra of compound 5 and silvestrol (6)

Thus, the structure of compound 5 (major isomer) was determined as an analogue of silvestrol (6) and assigned as (1R,2R,3S,3aR,8bS)-4‴-{[(2‴R,4‴R)-4‴-[(S)-1,2-dihydroxyethyl]-3-hydroxy-1,4-dioxan-2-yl]oxy}−1,8b-dihydroxy-8-methoxy-3a-(4-methoxyphenyl)-3-phenyl-2,3,3a,8b-tetrahydro-1H-cyclopenta[b]benzofuran-2-carboxamide. The present report of the isolation of compounds 1-5 has led to a two-fold increase in the number of known dioxanyl ring-containing cyclopenta[b]benzofurans in the genus Aglaia.^3,5,6

HPLC analysis of the fractions obtained from the chloroform partition, on an ODS column showed the most active sub-fractions (against HT-29 human colon cancer cell lines), D3F8, F81, F82, F83, F9, and F10 to contain silvestrol (6) and 5‴-episilvestrol (7). Furthermore, HPLC analysis of these fractions using standard stock solutions of 6 and 7, showed episilvestrol (7) to be present at a higher concentration as compared to silvestrol (6), making A. perviridis the first species in this genus to have 5‴-episilvestrol (7) as the major bioactive rocaglate constituent.

All isolated compounds were evaluated for their cytotoxic activity against the HT-29 human colon cancer and PC-3 human prostate cancer cell lines. Compounds 1-4 were observed to be inactive (IC₅₀ >10 μM), which may be attributed to the rigidity of the 1,4 dioxanyl ring due to the presence of the fused 2,3-dihydrobenzofuran system in these compounds. Since these compounds are structurally similar to silvestrol (7), this demonstrates the detrimental effect of rigidifying the dioxanyl ring, known to be required for both enhanced cytotoxic activity and the increased sensitivity of silvestrol (6) to multi-drug resistance.³⁹ Furthermore, compound 5 showed moderate activity in these cell lines (IC₅₀ = 2.3 μM for both), which could be ascribed to either or both the hydroxy group at C-2‴ or the amide group at C-2. Previous investigations have demonstrated the occurrence of an α-configuration of C-2‴ to be detrimental to the biological activity of silvestrol.⁵ Accordingly, the reduced activity of this compound 5 is most likely due to the presence of the C-2‴ hydroxy group. The biological test data obtained in this study contribute in further comprehending the structure-activity relationships of flavaglines against human cancer cells.^{7,23,28,40,41}

EXPERIMENTAL SECTION

General Experimental Procedures.

Optical rotations of all compounds were measured on an Anton Paar MCP 15 polarimeter (Anton-Paar, Ashland, VA, USA). The UV spectra were obtained using a Hitachi U-2910 spectrometer (Hitachi, Tokyo, Japan). The ECD spectra were recorded on a JASCO J-810 spectropolarimeter (JASCO, Easton, MD, USA). A Thermo-Nicolet 6700 FT-IR spectrophotometer (Thermo Scientific, Waltham, MA, USA) was used to record the infrared spectra. The NMR spectroscopic data were recorded on a Bruker Ascend 700 MHz NMR spectrometer and the data were processed using TopSpin 3.2 software (Bruker, Billerica, MA, USA). A Thermo LTQ Orbitrap mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) was used to obtain HRESIMS data. Column chromatography was conducted with 230–400 mesh or 40–63 μm mesh silica gel (Sorbent Technologies, Norcross, GA, USA). Analytical HPLC purification was performed on a 150 × 4.6 mm i.d., 5 μm Sunfire C₁₈ (Waters, Milford, MA, USA) and 250 × 4.6 mm i.d., 5 μm Luna C₁₈ column (Phenomenex, Torrance, CA, USA) equipped with a Hitachi 2130 controller, autosampler and a 2440 diode array detector. Chiral-phase separation of the compounds was carried out on an analytical 250 × 4.6 mm i.d., 3 μm Chiralpak IB-3 column (Diacel, West Chester, PA, USA). Preparative HPLC was conducted on a 150 × 19 mm i.d., 5 μm Sunfire C₁₈ (Waters, Milford, MA, USA) and a 250 × 19 mm i.d., 5 μm Luna C₁₈ column (Phenomenex, Torrance, CA, USA) connected to a Hitachi system with a Prep24 HPG pump system, autosampler and 2400 single wavelength UV detector.

Plant Material.

A re-collection of A. perviridis roots was conducted in July, 2011, in Nui Chua National Park, Ninh Thuan Province, Vietnam, by D.D.S. and T.N.N, who also identified this species. A voucher specimen (original collection Soejarto et al. 14863) of this species has been deposited in the John G. Searle Herbarium of the Field Museum of Natural History (under the accession number F-2300817), Chicago, IL, USA.

Extraction and Isolation.

The air-dried and ground root material of A. perviridis (2.5 kg) was extracted exhaustively with MeOH (3 × 5 L) for three days each and dried in vacuo to afford 232 g of a crude extract. This extract was partitioned sequentially with hexanes (3 X 1 L) and CHCl₃ (3 X 1 L, 126 g), followed by detannification of the CHCl₃ partition with 1% NaCl. The CHCl₃ and aqueous partitions were found to be most active with IC₅₀ values of 1.1 and 1.7 μg/mL, respectively, against the HT-29 human colon cancer cell line. The CHCl₃ partition was then fractionated on a silica gel column (CH₂Cl₂-acetone, 1% to 100% acetone), to afford 11 fractions, F1-F11. Of these, fraction F8 (13.37 g, IC₅₀ = 0.6 μg/mL, HT-29) was chromatographed on a Sephadex LH-20 column (H₂O/MeOH, 70–100% MeOH) to yield eight sub-fractions (F81-F88), of which F81 (1.55 g, IC₅₀ = 0.6 μg/mL, HT-29), F82 (1.12 g, IC₅₀ = 0.1 μg/mL, HT-29), and F83 (1.12 g, IC₅₀ = 0.3 μg/mL, HT-29) proved to be silvestrol (6)-rich fractions by HPLC. Sub-fraction F83 was separated on an open ODS column (MeOH/H₂O, 50–100% MeOH) to yield 5‴-episilvestrol (7, 4 mg). Fraction F82 (1.12 g) was subjected to purification over a semi-preparative RP-C₁₈ column (150 × 10 mm) using MeOH-H₂O (58:42) as solvent to afford silvestrol (6, 1.5 mg, t_R = 14.45 min) and 5‴-episilvestrol (7, 4.5 mg, t_R = 16.45 min).

The aqueous extract was partitioned with EtOAc (3 X 1 L) to afford 5.5 g of a dried EtOAc partition (D4). The ¹H NMR spectrum of the EtOAc partition (D4) showed characteristic peaks of cyclopenta[b]benzofurans (discussed in the “Results and Discussion” section) and was subjected to a RP-C₁₈ silica gel column chromatography, using a MeOH and H₂O gradient (30–100% MeOH) as solvent systems, to afford eight fractions (F1-F8). Fraction F5 (930.1 mg), the most active fraction, with 30% survival at 20 μg/mL against the HT-29 cell line, was then eluted on an open silica gel column to yield eight sub-fractions, F51-F58. One of these fractions, F55 (256.4 mg), showed inhibitory activity against various cancer cell lines (inclusive of 41% survival at 2 μg/mL against the HT-29 cell line) and was purified using a preparative ODS column (250 mm X 19 mm), with CH₃CN-H₂O (23:77, 17 mL/min), to give an anomeric mixture, compound 5 (12 mg). Compound 5 was purified on a chiral-phase column (Chiralpak IB-3, 250 X 4.6 mm) using a hexanes-EtOH (75:25) isocratic solvent system to afford two peaks, P1 and P2, found to be interchangeable epimers (see the “Results and Discussion” section). Fraction F53 (79.4 mg) was subjected to separation over a semi-preparative RP-C₁₈ column (250 mm x 10 mm), using isocratic conditions (CH₃CN-H₂O, 25:75), resulting in the purification of compounds 1 (1.4 mg) and 2 (1.4 mg). Similarly, compounds 3 (1.1 mg) and 4 (1.6 mg) were obtained from subfraction F54 (52.3 mg) using a semi-preparative ODS column (250 mm x 10 mm; MeOH-H₂O, 50:50).

Compound 1:

colorless gum; [α]²⁰_D +37 (c 0.1, MeOH); UV (MeOH) λ_max (log ε) 212 (4.59), 274 (3.47) nm; ECD (c 1.81 × 10⁻⁴ M, MeOH) λ_max (Δε) 217 (−19.11), 250 (−7.80), 294 (−0.45); IR (film) ν_max 3727, 3350, 2922, 2360, 2341, 1647, 1612, 1513, 1450, 1250, 1182, 1145, 1110, 949, 801, 754, 668 cm⁻¹; ¹H and ¹³C NMR (700 MHz, methanol-d₄) data, see Tables 1 and 2; HRESIMS m/z 630.1957 [M+Na]⁺ (calcd for C₃₂H₃₃NO₁₁Na, 630.1951).

Compound 2:

colorless gum; [α]²⁰_D +20 (c 0.1, MeOH); UV (MeOH) λ_max (log ε) 214 (4.78), 274 (3.55) nm; ECD (c 1.65 × 10⁻⁴ M, MeOH) λ_max (Δε) 217 (+15.37), 250 (−6.18), 294 (−0.31); IR (film) ν_max 3354, 2923, 1679, 1641, 1611, 1514, 1486, 1450, 1298, 1257, 1184, 1144, 1108, 1071, 1071, 1014, 949, 804 cm⁻¹; ¹H and ¹³C NMR (700 MHz, methanol-d₄) data, see Tables 1 and 2; HRESIMS m/z 630.1956 [M+Na]⁺ (calcd for C₃₂H₃₃NO₁₁Na, 630.1951).

Compound 3:

colorless gum; [α]²⁰_D +23 (c 0.1, MeOH); UV (MeOH) λ_max (log ε) 280 (3.39) nm; ECD (c 8.24 × 10⁻⁵ M, MeOH) λ_max (Δε) 217 (−3.36), 250 (+2.26), 294 (−1.18); IR (film) ν_max 3725, 3355, 2923, 2360, 2341, 1615,1455, 1134, 668 cm⁻¹; ¹H and ¹³C NMR (700 MHz, methanol-d₄) data, see Tables 1 and 2; HRESIMS m/z 630.1956 [M+Na]⁺ (calcd for C₃₂H₃₃NO₁₁Na, 630.1951).

Compound 4:

colorless gum; [α]²⁰_D +40 (c 0.1, MeOH); UV (MeOH) λ_max (log ε) 213 (4.60), 280 (3.53) nm; ECD (c 8.24 × 10⁻⁵ M, MeOH) λ_max (Δε) 217 (+12.47), 250 (+1.75), 294 (−1.33); IR (film) ν_max 3727, 3350, 2922, 2360, 2341, 1647, 1612, 1513, 1450, 1250, 1182, 1145, 1110, 949, 880, 801, 754, 668 cm⁻¹; ¹H and ¹³C NMR (700 MHz, methanol-d₄) data, see Tables 1 and 2; HRESIMS m/z 630.1958 [M+Na]⁺ (calcd for C₃₂H₃₃NO₁₁Na, 630.1951).

Compound 5:

colorless amorphous powder; UV (MeOH) λ_max (log ε) 208 (4.92), 273 (3.62) nm; ECD (c 0.89 × 10⁻⁴ M, MeOH) λ_max (Δε) 217 (−29.44), 250 (+0.74), 294 (+1.63); IR (film) ν_max 3408, 3005, 2937, 2839, 2360, 2341, 1668, 1609, 1513, 1498, 1453, 1428, 1342, 1300, 1250, 1218, 1129, 1045, 1030, 826, 700,614 cm⁻¹; ¹H and ¹³C NMR (700 MHz, methanol-d₄) data, see Tables 1 and 2; HRESIMS m/z 648.2052 [M+Na]⁺ (calcd for C₃₂H₃₃NO₁₁Na, 648.2057).

Silvestrol 6:

ECD (c 0.92× 10⁻⁴ M, MeOH) λ_max (Δε) 217 (−54.94), 250 (−10.27), 294 (−11.41).

5‴-Episilvestrol 7:

ECD (c 0.15× 10⁻⁵ M, MeOH) λ_max (Δε) 217 (−66.79), 250 (−1.19), 294 (+0.19).

ECD Calculations.

The 2D structures of the compounds were converted to 3D coordinates using OpenBabel 2.3.1^42,43 and the MMFF94^44-48 force field. Openbabel’s confab function was used to systematically generate local minima conformers with at least 0.5 Å RMSD cut-off and sampling at most one million conformers. The conformers were visually inspected, and 4–5 conformers were selected based on variance around the rotatable bonds. The selected conformers were optimized using Gaussian 16⁴⁹ at the wB97XD/6–31G(d)^50–60 level followed by a frequency calculation. The conformer with the lowest thermal free energy value was selected and three separate TD-DFT calculations on 10 N-states were done with the Def2TZVP^61,62 basis set using the wB97XD, CAM-B3LYP,⁶³ and M062X⁶⁴ functionals, respectively. All Gaussian 16 calculations were done on OSC⁶⁵ using the self-consistent reactant field (SCRF) implicit solvent model of MeOH. Functionals and basis sets used were chosen based on previously published work.³⁶ The calculated spectra were translated by −10 nm to match the experimental spectra.

Cytotoxicity Assays.

All extracts and fractions were prepared in 96-well plates at sample concentrations of 20 and 2 μg/mL, and were tested for cytotoxic activity against a panel of cancer cell lines (MDA-MB-231, MDA-MB-435, OVCAR3 and HT-29 cell lines), according to a previously published protocol.⁶⁶ The cytotoxicity of the purified test compounds (1–5) was screened against HT-29 and PC-3 cells by a previously published protocol.⁵ Paclitaxel was used as a positive control for both these cancer cell line types.