Cytotoxic Lignan and Flavonoid Derivatives from the Branches of Beilschmiedia yunnanensis

Abstract

An investigation of a cytotoxic MeOH extract of the branches of Beilschmiedia yunnanensis, collected in Vietnam, led to the isolation of four new compounds (1−4). Two of these, isolated from a CHCl₃-soluble partition, were characterized as the furofuran-type neolignans, beilschmiedianins A (1) [(7R,7′R,8S,8′S,8″R)-4′,4″,9″-trihydroxy-3,5,3′,3″-tetramethoxy-4,8″-oxy-7,9′:7′,9-diepoxy-8,8′-sesquilignan-7″-one] and B (2) [(7R,7′R,7″R,8S,8′S,8″R)-9″-feruloyl-4′,4″-dihydroxy-3,5,3′,3″-tetramethoxy-4,8′′-oxy-7,9′:7′,9-diepoxy-8,8′-dilignan-7″-ol]. In turn, the flavonoid glycosides 3 and 4 were obtained from an EtOAc-soluble partition and were assigned as (2R,3R)-dihydrokaempferol-5-O-β-l-arabinosyl-(2→1)-α-l-rhamnopyranoside and (2R,3R)-dihydrokaempferol-5-O-β-l-arabinopyranoside, respectively. The structures of these new compounds were determined using a combination of spectroscopic and spectrometric methods. Additionally, the known dilignan, (−)-9,9′-O-diferuloyl-secoisolariciresinol (5), showed selective cytotoxicity against the OVCAR3 ovarian cancer cell line, with an IC₅₀ value of 0.51 μM. Mechanistic studies showed that compound 5 increased the cPARP levels and decreased the expression of BCL-2 in OVCAR3 cells.

Graphical Abstarct

The Lauraceae is a plant family comprising 50 genera and approximately 3,000 species, mainly arboreal plants, found growing in tropical areas of Southeast Asia, South America, Australasia and Africa.^1,2 Beilschmiedia, a genus belonging to this group, is comprised of trees and shrubs that have been used in traditional medicine to treat conditions such as uterine tumors, rheumatism, and bacterial, fungal, and parasitic infections.^3,4 Laboratory studies have shown that Beilschmiedia erythrophloia (B. erythrophloia),⁵ B. mannii,⁶ B. acuta,⁷ B. obscura,⁸ and B. cinnamomea,⁹ possess antioxidant, anti-inflammatory, cytotoxic, and antibacterial and antifungal activities, respectively. Cytotoxic compounds have been reported from this genus, as exemplified by N-trans-feruloyloctopamine from B. erythrophloia¹⁰ and tsangin B from B. tsangii,¹¹ having IC₅₀ values of 10.3 μg/mL against a human lymphoblastic leukemia and 0.42 μg/mL against a mouse lymphocytic leukemia cell line, respectively. Williams and co-workers isolated eight new aliphatic tetracyclic beilshmiedic acids from a Gabonese Beilschmiedia species. Four of these compounds exhibited IC₅₀ values of <10 μM against the NCI-H460 large-cell lung carcinoma cell line.¹² In the present study, the branches of B. yunnanensis Hu have been investigated for their cytotoxic constituents. No previous studies have been reported on the cancer cell line cytotoxic constituents of this species. The plant was collected in Vietnam and its branches were extracted with MeOH, followed by several solvent partition and chromatographic purification steps. Four new (1–4) and 11 known compounds (5–12) (Chart 1), and (13-15, Figure S1, Supporting Information), belonging to diverse chemical architectures, including the phenylpropanoid, lignan, and flavonoid classes, were isolated and characterized using spectroscopic and spectrometric methods, and subsequently evaluated for their cytotoxic potential.

RESULTS AND DISCUSSION

To identify potential cytotoxic constituents of Beilschmiedia species, the dried and pulverized branches of B. yunnanensis were extracted with MeOH. The dried extract showed a moderate IC₅₀ value of 8.4 μg/mL against the human HT-29 colorectal cancer cell line and hence was successively partitioned with hexanes, CHCl₃, EtOAc, and H₂O. All resulting fractions were tested against four human cancer cell lines: MDA-MB-435 (melanoma), OVCAR3 (ovarian), HT-29 (colorectal), and MDA-MB-231 (breast). Of these four partitions, the CHCl₃-partition showed a promising activity, exhibiting a 21% cell survival at 20 μg/mL against the HT-29 cancer cell line. Using chromatographic methods, the new lignans 1 and 2, along with eight known compounds (5–9) (Chart 1) and 13–15 (Figure S1, Supporting Information) were obtained from a cytotoxic fraction from the CHCl₃-soluble sample. In addition, the EtOAc-partition showed a slight cytotoxicity against MDA-MB-231, and further purification resulted in the isolation of two new dihydroflavonol glycosides, 3 and 4, along with three known flavonoid derivatives, catechin (10),¹³ epicatechin (11),¹⁴ and (+)-taxifolin (12)¹⁵. The structures of the aforementioned compounds were characterized by analysis of their 1D- and 2D-NMR spectroscopic and HRESIMS data. Furthermore, chiroptical measurements [optical rotation (OR), electronic circular dichroism (ECD), and quantum chemical calculations of ECD, OR and NMR] were conducted to assign the stereostructures of the new lignans. Identification of the sugar components of the glycosides was performed by acid hydrolysis and derivatization processes as described below.

Compound 1 was obtained as a light-brown gum. Its molecular formula, C₃₁H₃₄O₁₁, was generated from the protonated molecular ion at m/z 583.21715 [M + H]⁺ (calcd for C₃₁H₃₅O₁₁⁺, 583.21739) observed in the HRESIMS data. From its IR spectrum, the broad absorption bands at ν_max 3380 cm⁻¹ and the sharp bands at 1591, 1458, and 1423 cm⁻¹ indicated the presence of hydroxy and aromatic functionalities, respectively. The ¹H, ¹³C and HSQC NMR data of 1 (Table 1) revealed the occurrence of eight aromatic methines, three oxymethylenes, three oxymethines, two methines, seven oxygenated aromatic carbons, three quaternary aromatic carbons, a carbonyl carbon, and four methoxy groups. In the ¹H NMR spectrum, three different sets of aromatic protons were discernible. One of these at δ_H 6.68 (1H, s, H-2, 6) integrated for two protons, alluding to symmetry, and thus the presence of a 1,3,4,5-tetrasubstituted aromatic ring. The remaining two sets of aromatic resonances showed the typical splitting patterns and coupling constants for 1,3,4-tri-substituted aromatic rings: δ_H 6.94 (1H, d, J = 1.6 Hz, H-2′), δ_H 6.76 (1H, d, J = 8.1 Hz, H-5′), δ_H 6.81 (1H, dd, J = 8.2, 1.7 Hz, H-6′), and δ_H 7.54 (1H, br. d, J = 1.9, H-2″), δ_H 6.76 (1H, d, J = 8.1 Hz, H-5″) and δ_H 7.56 (1H, dd, J = 8.2, 1.9 Hz, H-6″).

Table 1.

¹H and ¹³C NMR Data of Compounds 1 and 2 (Methanol-d₄).

	1			2
position	δH, m (J in Hz)a	δH, m (J in Hz)b	δC, type	δH, m (J in Hz)c	δC, type
1			139.1, C		139.1, C
2	6.68, s	6.68, s	104.1, CH	6.66, s	104.0, CH
3			154.0, C		154.4, C
4			137.0, C		136.8, C
5			154.0, C		154.4, C
6	6.68, s	6.68, s	104.1, CH	6.66, s	104.0, CH
7	4.76, d (3.9)	4.76, d (4.3)	87.3, CH	4.71, d (4.4)	87.3, CH
8	3.12, m	3.12, m	55.3, CH	3.06, m	55.7, CH
9a	4.26d, m	4.26d, m	72.9, CH2	4.21, m	72.80h, CH2
9b	3.88e,j	3.88e,j		3.85, sj
OMe-3,5	3.73, s	3.73, s	56.6, CH3	3.83, sj	56.6, CH3
1′			133.7, C		133.7, C
2′	6.94, d (1.6)	6.94, d (1.9)	110.9, CH	6.93, d (1.4)	111.0, CH
3′			149.1, C		149.2, C
4′			147.4, C		147.5, C
5′	6.76f, d (8.1)	6.76, d (8.1)	116.5, CH	6.77, d (8.1)k	116.1, CH
6′	6.81, dd (8.2, 1.7)	6.81, dd (8.2, 1.9)	120.0, CH	6.79, dd (8.1, 1.4)k	120.1, CH
7′	4.71, d (4.3)	4.71, d (4.5)	87.5, CH	4.68, d (4.4)	87.5, CH
8′	3.12, m	3.12, m	55.8, CH	3.05, m	55.3, CH
9′a	4.26d, m	4.26d, m	72.7, CH2	4.21, m	72.81h, CH2
9′b	3.88e,j	3.88e,j		3.85j
OMe-3′	3.86g, s	3.86, s	56.2, CH3	3.86, s	56.4, CH3
OH-4′
1″			133.7, C		132.9, C
2″	7.54, br. d (1.9)	7.51, d (2.1)	112.2, CH	7.03, d (1.5)	111.8, CH
3″			150.0, C		148.9, C
4″			157.3i, C		147.5, C
5″	6.76f, d (8.1)	6.71, d (8.3)	116.5, CH	6.76, d (8.2)	115.9, CH
6″	7.56, dd (8.2, 1.9)	7.53, dd (8.3, 2.1)	125.8, CH	6.89, dd (8.2, 1.5)	120.9, CH
7″			196.9, C	4.97, d (6.6)	75.3, CH
8″	5.31, t (5.3)	5.33, t (5.4)	85.8, CH	4.44, ddd (6.6, 5.1, 3.3)	85.8, CH
9″a	3.91, br. d (5.7)	3.90, br. d (5.4)	64.3, CH2	4.30, dd (11.9, 3.2)	65.5, CH2
9″b	3.91, br. d (5.7)	3.90, br. d (5.4)		4.09, dd (11.9, 5.1)
OMe-3″	3.86g, s	3.84, s	56.4, CH3	3.82, s	56.5, CH3
OH-4″
1‴					126.9, C
2‴				7.13, d (1.5)	111.6, CH
3‴					149.8, C
4‴					152.1i, C
5‴				6.78, d (8.1)k	116.8, CH
6‴				7.01, dd (8.1, 1.5)	124.5, CH
7‴				7.37, d (15.9)	147.1, CH
8‴				6.18, d (15.9)	114.6, CH
9‴					168.9, C
OMe-3‴				3.88, s	56.5, CH3
OH-4‴

^a1H, 400 MHz, ¹³C, 100 MHz.

^b1H, 600 MHz.

^c1H, 700 MHz, ¹³C, 175 MHz.

^d-gProtons in the same column may be exchanged.

^hCarbons in the same column may be exchanged.

ⁱAssigned by HMBC.

^jOverlapped.

^kMultiplicity extracted from the 1D-selective TOCSY spectrum.

In addition, a carbonyl group in 1 was assigned to the resonance at δ_C 196.9 (C-7″) in the ¹³C NMR spectrum, with the HMBC spectrum being used to determine its position. In the HMBC spectrum, this carbon was correlated to the protons at δ_H 7.54 and 7.56 from one of the 1,3,4-trisubstituted aromatic rings. Additionally, correlations of the oxymethine proton at δ_H 5.31 (1H, t, J = 5.3 Hz, H-8″) and the oxymethylene protons δ_H 3.91 (2H, br. d, J = 5.7 Hz, H-9″) with the carbonyl carbon indicated the presence of a 2-O-substitued 2,3-dihydroxy-1-(4-hydroxy-3-methoxyphenyl)propan-1-one unit in compound 1 (Figure S5, Supporting Information), also known as veratroylglycol.^16,17 The position of the O-methyl group (3H, δ_H 3.86, s) of the veratroylglycol unit at C-3″ was assigned by 2D-NOESY data acquired using a 600 MHz ¹H NMR spectrum, which gave better resolution in the aromatic region. A NOE cross peak between MeO-3″ and H-2″ indicated their ortho relationship (Figure S7, Supporting Information). Further analysis of the remaining unassigned resonances of 1 suggested the presence of a furofuran unit in this compound. In particular, the two pairs of doublets at δ_H 4.76 (1H, d, J = 3.9 Hz, H-7; δ_C 87.3) and δ_H 4.71 (1H, d, J = 4.3 Hz, H-7′, δ_C 87.5), each with a small J value of 3.9 and 4.3 Hz, respectively, along with the unresolved protons at δ_H 3.12 (2H, overlapped, H-8 and H-8′; δ_C 55.3, 55.8), were similar to values found in the ¹H NMR spectra of other furofuran lignans [e.g., (+)- and (−)-idaeusinols B and C].¹⁸ This was supported further by the signals for the unresolved protons at δ_H 3.88 (overlapped) and δ_H 4.26 (overlapped), corresponding to the oxymethylene protons H₂-9 and H₂-9′. In the HMBC spectrum of 1, while the signal at δ_H 4.76 (H-7) correlated with the equivalent C-2/6 (δ_C 104.1) of the 1,3,4,5-tetrasubstituted aromatic ring, H-7′ correlated with C-1′ (δ_C 133.7), C-2′ (δ_C 110.9) and C-6′ (δ_C 120.0) of the remaining 1,3,4-trisubstituted aromatic ring (Figure S5, Supporting Information). These observations indicated that the furofuran portion of compound 1 comprised the lignan medioresinol, for which a literature data comparison showed the similarity of their NMR spectra.¹⁹ The location of the methoxy groups was assigned by analysis of the 2D-NOESY spectrum (Figure S7, Supporting Information). Finally, the veratroylglycol and medioresinol units of compound 1 were connected by the correlations of H-8″ to the oxygenated aromatic C-4 (δ_C 137.0) of the former unit. The structure proposed for compound 1 is similar to that of ficusesquilignan B, which was isolated from the roots of Ficus hirta,²⁰ but with all of its aromatic rings being 1,3,4,5-tetrasubstituted. These differences are reflected in the changes in chemical shifts of corresponding and neighboring positions found in the ¹H NMR spectrum of 1. Additional comparison with sesqui-illisimonan A, a sesquineolignan isolated from Illicium simonsii, confirmed the 1,3,4-trisubstituted aromatic rings in compound 1.²¹ This compound (1) recently was proposed tentatively from Tarenaya aculeata stems via MS data and biosynthetic considerations after detection by LC-MS, although it was not purified and the stereostructure not resolved.²² Thus, according to the data obtained, the structure of compound 1 is proposed to be (7R*,7′R*,7″R*,8S*,8′S*,8″R*)-4′,4″,9″-trihydroxy-3,5,3′,3″-tetramethoxy-4,8″-oxy-7,9′:7′,9-diepoxy-8,8′-sesquilignan-7″-one, and this has been named beilschmiedianin A.

Compound 2 was obtained as an off-white solid. Its molecular formula, C₄₁H₄₄O₁₄, was supported by the observation of a deprotonated molecular ion at m/z 759.26626 [M − H]⁻ (calcd for C₄₁H₄₃O₁₄⁻, 759.26583) present in the HRESIMS data. Its IR spectrum exhibited similar bands to compound 1 at ν_max 3384 cm⁻¹ and 1592, 1515, 1271 cm⁻¹, corresponding to hydroxy groups and an aromatic skeleton, respectively. The ¹H, ¹³C, and HSQC NMR spectra of 2 showed close similarities to those of compound 1, such as the medioresinol core and the phenylpropanoid unit at C-4 of 1. However, additional signals were present (Table 1). Differences were observed for the chemical shifts of C-3″, C-4″, C-6″, and C-7″. The absence of any signal above 190 ppm suggested that the carbonyl group (C-7″, δ_C 196.9) in compound 1 was absent in compound 2. However, the observed shift corresponding to C-7″ at δ_C 75.3, indicated that a hydroxy group is attached. NMR comparison with those of sesqui-illisimonan A²¹ and hedyotol C 7″-O-β-d-glucopyranoside²³ confirmed the sesquilignan structure in 2, with a hydroxy group at C-7″ instead of a carbonyl group. In addition, small differences such as the chemical shifts of C-9″ of compound 1 (δ_C 64.3) and 2 (δ_C 65.5), and the adjacent proton signals, suggested that extra groups are attached at this position.

The analysis of the additional aromatic proton resonances led to the proposal of a further 1,3,4-trisubstituted ring, with one of the substituents corresponding to a methoxy group, a second to a hydroxy group, and a quaternary carbon. The attachment of the methoxy group at C-3‴ (δ_C 149.8) was confirmed by a correlation with the protons of the methyl group (3H, s, δ_H 3.88) and this carbon, in the HMBC spectrum (Figure S14, Supporting Information). The hydroxy group was attached to C-4‴ (δ_C 152.1) due to its higher chemical shift.

The remaining signals (δ_C 114.6, 147.1, 168.9) were attributed to olefinic and ester functionalities. Correlations between C-9‴ (δ_C 168.9) and the olefinic protons at δ_C 7.37 and 6.18 (H-7‴ and H-8‴) in the HMBC spectrum suggested the connection of these two groups. Furthermore, the H-8‴ (1H, d, J = 15.9 Hz, δ_H 6.18) resonance showed a HMBC correlation with C-1‴ (δ_C 126.9 ppm), connecting this fragment to the 1,3,4-trisubstituted ring. All these signals enabled a feruloyl substructure to be proposed. Finally, a HMBC resonance occurring between C-9‴ (δ_C 168.8) and H-9″b (δ_H 4.09) linked the feruloyl to the sesquilignan moiety, giving a dilignan core. A literature search and consideration of the absolute configuration (see below) indicated that compound 2 is an enantiomer of the reported but misnamed dilignan (7S,7′S,7″S,8R,8′R,8″S)-9″-feruloyl-4″,4″-dihydroxy-3,3′,3″,5-tetramethoxy-7,9′:7′,9-diepoxy-8′-oxy-8,8′-dilignan-7″-ol (the authors’ compound 17).²⁴ However, the chemical shift differences between the two enantiomers could be due to the use of different NMR solvents. According to this analysis, the structure of compound 2 is proposed as [(7R*,7′R*,7″R*,8S*,8′S*,8″R*)-9″-feruloyl-4′,4″-dihydroxy-3,5,3′,3″-tetramethoxy-4,8″-oxy-7,9′:7′,9-diepoxy-8,8′-dilignan-7″-ol.

Establishing the absolute configurations of compounds 1 and 2 was challenging due to the limited sample quantities, which prevented the utilization of traditional methods like chiral derivatization and/or chemical degradation. Additionally, the high degree of conformational flexibility of multiple rotatable bonds in the pendant C₆−C₃ side chain tethered to the furofuran-type medioresinol core structure complicated the reliable computation of chiroptical and spectroscopic properties. An extensive conformational search with the Mixed Torsional/Low-Mode sampling method in MacroModel, followed by DFT optimization and Boltzmann-weighted ECD spectrum simulations eventually permitted definition of the absolute stereostructure of compound 1.

To establish the absolute configuration of the substituted furofuran-type lignan scaffold in 1 and 2, electronic circular dichroism (ECD) calculations were conducted for the reference compounds (−)- and (+)-medioresinol (Figure 1), which are related structurally to the lignan core present in both natural products. A plausible biogenetic pathway from (−)-medioresinol to compounds 1 and 2 is provided in Figure S19, Supporting Information. These reference compounds have been previously characterized.^19,25 The computed average ECD spectra for (−)-medioresinol displayed a diagnostic negative exciton couplet around 240 nm and a pronounced negative Cotton Effect (CE) in the 207–211 nm region (Figure 2). The negative couplet is reminiscent of exciton coupling between the aromatic rings located at C-7 and C-7′. Taken in conjunction with the similar negative CE’s observed near 207 nm in the experimental spectra of 1 and 2 (Figure 3), these CE’s served to define the (7R,8S:7′R,8′S) absolute configuration for the furofuran-type lignan core of compounds 1 and 2.

Figure 1.

Structures of the furofuran-type lignans (−)-medioresinol and (+)-medioresinol.

Figure 2.

Calculated electronic circular dichroism (ECD) spectrum of the Boltzmann-averaged spectra of (−)-medioresinol in MeOH. The σ-value (artificial line broadening) was set to 0.29 eV. The UV wavelength correction of 3 nm was applied to generate this graph.

Figure 3.

Experimental (black) and calculated (red) ECD spectra in MeOH for compound 1 diastereomers: (A) 8″R and (B) 8″S. Calculated spectra are Boltzmann-averaged with an artificial line broadening (σ = 0.27 eV) and a UV wavelength correction of 7 nm.

Having established the (7R,8S:7′R,8′S) absolute configuration of the common furofuran-type lignan core of compound 1, the C-8″ absolute configuration was investigated by computing the ECD spectra of the (8″R) - and (8″S)-diastereomers of compound 1 using time-dependent density functional theory (TDDFT) at the mPW1PW91/6–311+G(2d,p) level in MeOH. Comparative analysis of the experimental and calculated ECD spectra of the (8″R) and 8″S-diastereomers (Figure 3) indicated strong agreement with the (8″R)-diastereomer, thus establishing the absolute configuration of this new sesquilignan 1 as (7R,8S:7′R,8′S:8″R), with a systematic name of (7R,7′R,8S,8′S,8″R)-4′,4″,9″-trihydroxy-3,5,3′,3″-tetramethoxy-4,8″-oxy-7,9′:7′,9-diepoxy-8,8′-sesquilignan-7″-one, and who has been named beilschmiedianin A.

Considering a putative biosynthetic relationship between the C₆–C₃ unit in compound 1 and the lignan moiety in compound 2, it is reasonable to propose that the benzylic hydroxy group at C-7″ in 2 derives via reduction of the corresponding carbonyl function of compound 1, resulting in either a (7″R) or a (7″S) absolute configuration. The experimental ECD spectrum of compound 2 (Figure S17, Supporting Information) retained a strong negative CE at ~210 nm, attributable to the medioresinol core, along with weak Cotton effects in the 220–300 nm range. However, these low-intensity CE features did not allow for any meaningful and conclusive stereochemical assignment.

Based upon the limited configurational information inferred from ECD spectrum, it was necessary to take recourse to computational analysis of both the ¹H NMR coupling constants (Tables S2 and S3, Supporting Information) and specific rotations for the (7″S,8″R-erythro configuration) and (7″R,8″R-threo configuration) diastereomers of compound 2. Conformational analysis of the (7″R,8″R-threo-diastereomer) of compound 2 revealed that four out of 14 conformers exhibited an intramolecular hydrogen bond between the C-7″ hydroxy group and the methoxy group on ring B. This interaction significantly skewed the 7″,8″-dihedral angle from approximately −175° (as observed in the other conformers) to around −78°. Owing to this pronounced conformational deviation, these four conformers were excluded from our subsequent analysis. The coupling constant analysis convincingly confirmed the threo configuration via ³J_7″,8″ values of 8.49 and 3.60 Hz for the threo and erythro diastereomers, respectively, compared to the 6.6 Hz experimental value of compound 2. Furthermore, the experimental specific rotation of −11.32 closely matched the calculated value of −12.14 for the threo diastereomer of 2. Notably, an enantiomer of compound 2 has been previously reported by Nguyen-Ngoc and co-workers,²⁴ who observed a specific rotation of +2.5 in CHCl₃. Together, these findings corroborated the (7R,8S: 7′R,8′S:7″R,8″R) absolute configuration for the newly identified dilignan 2.

Molecular orbital (MO) analysis (Figure 4A) of the lowest-energy conformer 6 of (−)-medioresinol revealed that the diagnostic negative exciton couplet around 240 nm in the calculated ECD spectrum originated from the excitations with negative rotatory strengths at 250.25 and 233.21 nm. These electronic transitions were associated primarily with MO100 (HOMO-3) → MO106 (LUMO+2) and MO103 (HOMO) → MO105 (LUMO+1), involving π → π* transitions between the aromatic rings at C-7 and C-7′. The positive CE at ~230 nm corresponded to a transition with positive rotatory strength at 229.29 nm, assigned to MO102 → MO105, indicating electronic excitation (π → π* transitions) from ring A to rings A and B. A pronounced negative CE in the 205–210 nm region was attributed to a transition at 205.99 nm, which involved the electron delocalization from ring B (MO100/MO101) to ring A (MO104).

Figure 4.

Molecular orbitals involved in key transitions in the calculated ECD spectrum of the lowest-energy conformer 6 of (−)-medioresinol (A) and conformer 7 of 1 (B) at the mPW1PW91/6-311+G(d,p) level in MeOH. Orbitals are plotted with 0.03 e⁻/au3 isovalues.

For compound 1, MO analysis of the lowest-energy conformer 7 (Figure 4B) suggested that the positive CE at ~330 nm in the experimental spectrum resulted from the excitation with positive rotatory strength at 313.02 nm, due to the electronic transition from MO153 to MO155 (LUMO), an electronic transition involving a π → π* transition from the B-ring to the benzoyl moiety (ring C). The broad negative CE at 265–280 nm could be assigned to an excitation having a negative rotatory strength at 293.98 nm, which matched the electronic transition from MO152 to MO155 (LUMO), involving the lone pair electrons of the benzoyl carbonyl group and the π system of the B-ring (n → π* transition). On the other hand, the positive CE at 240 nm in the experimental spectrum could be ascribed to an excitation with positive rotatory strength at 230.75 nm, caused by the electronic transition from MO153 (HOMO-1) to MO156 (LUMO+1) or MO158 (LUMO+3). This transition corresponds to a π → π* transition involving ring B and rings A and B. Finally, the negative CE around 210 nm was linked to a transition at 208.32 nm (MO154 (HOMO)→ MO162), involving π-electron delocalization across the A- and B-rings.

Compound 3 was obtained as a brown resin. Its molecular formula of C₂₆H₃₀O₁₄ was determined from the observation of a deprotonated molecular ion at m/z 565.15653 [M − H]⁻ (calcd for C₂₆H₂₉O₁₄⁻, 565.15628) present in the HRESIMS data. Its IR spectrum showed a broad absorption at 3340 cm⁻¹ and sharp signals at 1584 and 1610 cm⁻¹ typical for hydroxy and aromatic groups, respectively. The ¹H, ¹³C, and HSQC NMR spectra of 3 (Table 2) indicated the presence of six aromatic methines, 11 oxymethines, and one oxymethylene. The presence of a para-substituted aromatic ring was substantiated by signals at δ_H 7.35 (2H, d, H-2′/H-6′) and δ_H 6.83 (2H, d, H-3′/H-5′) with a coupling constant of J = 8.6 Hz, while the doublets at δ_H 6.16 (H-6) and δ_H 5.97 (H-8) having a coupling constant of J = 2.2 Hz indicated a 1,2,3,5-tetrasubstituted phenyl ring in 3. A correlation between H-2 (1H, d, J = 11.8 Hz, δ_H 4.92) and H-3 (1H, d, J = 11.8 Hz, δ_H 4.45), with a large coupling constant of J = 11.8 Hz, showed these protons to be pseudodiaxially coupled and oriented (Table 2). In the HMBC spectrum, H-2 correlated with C-1′ and C-2′/C-6′ of the para-substituted phenyl ring. Furthermore, H-2 and H-3 correlated to the carbonyl carbon resonating at δ_C 192.9 (C-4). Additionally, H-2 and the H-8 (1H, d, J = 2.2 Hz, δ_H 5.97) aromatic methine proton correlated with the oxygenated tertiary carbon C-9 (δ_C 165.9) in the HMBC spectrum. These observations, along with the correlation of H-8 and H-6 (1H, d, J = 2.2 Hz, δ_H 6.16) of the tetrasubstituted phenyl ring to the quaternary carbon at δ_C 102.8 (C-10), indicated the presence of a dihydroflavonol core (Figure S23, Supporting Information), similar to (+)-taxifolin (12). This was confirmed by a literature search showing this core to be dihydrokaempferol^26,27 with the main difference being at C-4 and C-5 with a 5 to 6 ppm shift, and also all the carbons of the A ring. Further analysis of the HMBC and NOESY data indicated that these differences arose from the attachment of a sugar unit at C-5 (Figures S23 and S25, Supporting Information). In addition, a comparison of the chemical shifts with literature data confirmed this position of attachment for the monosaccharide unit.²⁸ The first sugar moiety, attached to the dihydroflavonol aglycone, was assigned as a pyranose due to a HMBC correlation between the H-1″ anomeric proton (1H, br. d, J = 2.0 Hz, δ_H 5.69) and the carbon resonating at δ_C 60.9, corresponding to a CH₂ unit (Figure S23, Supporting Information). The proton at C-4″ (1H, ddd, J = 10.4, 5.1, 3.6 Hz, δ_H 3.93) in this sugar unit was axially oriented, consistent with the large coupling constant observed for H_ax-5″ (1H, t, J = 10.9 Hz, δ_H 3.74). Proton H-3″ (1H, br. t, J = 3.5 Hz, δ_H 3.98), was thus assigned in an equatorial orientation as a result of the small coupling constants with its adjacent protons. H-2″ (1H, dd, J = 3.6, 2.1 Hz, δ_H 4.06) was similarly equatorially oriented. Finally, H-1″ was positioned axially since an NOE correlation was observed with H_ax-5″ (Figures S25 and S26, Supporting Information). A second sugar moiety was attached to C-2″ based on the HMBC correlations with its anomeric proton H-1‴ (1H, d, J = 1.5 Hz, δ_H 4.97) (Figure S23, Supporting Information). Following the same reasoning, H-4‴ (1H, t, J = 9.5 Hz, δ_H 3.40) and H-5‴ (1H, dq, J = 9.3, 6.3 Hz, δ_H 3.74) (the minor differences in the J-values of H-4‴ and H-5‴ here and above may result from signal overlap of H-5″) were both assigned in axial orientations due to their observed large coupling constant (J ≈ 9.5 Hz). The methyl group resonating at δ_H 1.28 and appearing as a doublet (3H, d, J = 6.2 Hz) was attached at C-5‴ (δ_C 70.6) based on the COSY correlations of these protons with H-5‴ in the COSY spectrum (Figure S24, Supporting Information). Based on this analysis, the sugar moieties were assigned as an arabinopyranose and a rhamnopyranose. To confirm the absolute configurations of the sugar units in 3, an acid hydrolysis followed by derivatization and a ¹H NMR data analysis were performed, following the method of Inagaki et al.²⁹ The analysis of the proton spectra revealed four signals (d, J = 5.9 Hz, δ_H 5.40; d, J = 8.5 Hz, δ_H 5.65; d, J = 1.4 Hz, δ_H 5.71; dd, J = 3.3, 1.0 Hz, δ_H 5.85) corresponding to l-arabinose and l-rhamnose when compared with literature information (Figure S28, Supporting Information).²⁹ In addition, irradiations at δ_H 5.65 and δ_H 5.85 in a 1D-selective TOCSY experiment displayed resonances at δ_H 4.65 and δ_H 4.76, corresponding to the characteristic proton of the derivatized l-arabinose and l-rhamnose (S-H-1′), respectively (Figures S29 and S30, Supporting Information).²⁹ Finally, the (2R and 3R) absolute configurations of compound 3 were defined by direct comparison with the ECD data of (+)-taxifolin (12), isolated from the ethyl acetate partition, and also compared with literature.^30,31 Therefore, the structure of compound 3 was proposed as (2R,3R)-dihydrokaempferol-5-O-β-l-arabinosyl-(2→1)-α-l-rhamnopyranoside.

Table 2.

¹H and ¹³C NMR (700 and 175 MHz) Data of Compounds 3 and 4 (Methanol-d₄).

	3		4
position	δH, m (J in Hz)	δC, type	δH, m (J in Hz)	δC, type
1
2	4.92, d (11.8)	84.6, CH	4.93, d (11.7)	84.5, CH
3	4.45, d (11.8)	74.3, CH	4.40, d (11.7)	74.5, CH
4		192.9, C		193.0, C
5		159.2a, C		160.2, C
6	6.16, d (2.2)	97.9, CH	6.26, d (2.1)	99.2, CH
7		170.8, C		170.7, C
8	5.97, d (2.2)	99.4, CH	5.99, d (2.1)	99.5, CH
9		165.9, C		166.1, C
10		102.8, C		103.2, C
1′		129.6, C		129.6, C
2′	7.35, d (8.6)	130.3, CH	7.35, d (8.5)	130.3, CH
3′	6.83, d (8.6)	116.1, CH	6.83, d (8.6)	116.1, CH
4′		159.1a, C		159.2, C
5′	6.83, d (8.6)	116.1, CH	6.83, d (8.6)	116.1, CH
6′	7.35, d (8.6)	130.3, CH	7.35, d (8.5)	130.3, CH
arabinose
1″	5.69, br. d (2.0)	96.3, CH	5.24, d (3.8)	100.1, CH
2″	4.06, dd (3.6, 2.1)	77.3, CH	3.99, dd (5.5, 3.8)	71.2, CH
3″	3.98, br. t (3.5)	70.6, CH	3.81, dd (5.8, 3.7)	72.8, CH
4″	3.93, ddd (10.4, 5.1, 3.6)	65.5, CH	3.97, dt (7.7, 3.8)	66.6, CH
5″ax	3.74, t (10.9)	60.9, CH2	3.80, dd (11.7, 7.8)	63.3, CH2
5″eq	3.56, dd (11.1, 5.0)		3.61, dd (11.7, 4.0)
rhamnose
1‴	4.97, d (1.5)	102.5, CH
2‴	3.86, dd (3.3, 1.7)	72.4, CH
3‴	3.67, dd (9.2, 3.2)	72.2, CH
4‴	3.40, t (9.5)	73.8, CH
5‴	3.74, dq (9.3, 6.3)	70.6, CH
6‴	1.28, d (6.2)	18.1, CH3

^aExchangeable signals.

Compound 4 was obtained as a yellow resin. Its molecular formula of C₂₀H₂₀O₁₀, was supported by the observation of a deprotonated molecular ion at m/z 419.09879 [M − H]⁻ (calcd for C₂₀H₁₉O₁₀, 419.09837) present in the HRESIMS data. A broad absorption band observed in the IR spectrum at 3342 cm⁻¹ and a sharp band around 1611 and 1585 cm⁻¹ indicated the presence of hydroxy and aromatic groups, respectively. The ¹H and ¹³C NMR spectra showed comparable signals to those of compound 3, suggesting the presence of a dihydrokaempferol unit (Table 2). However, when compared with compound 3, only six oxymethines were detected along with one anomeric proton, suggesting that one sugar unit was present in this moiety. A HMBC correlation between H-1″ (1H, d, J = 3.8 Hz, δ_H 5.24) and C-5 (δ_C 160.2) confirmed the attachment of this saccharose at C-5. The sugar was found to be a pyranose, based on the HMBC correlation between one of the protons of the CH₂ group and the anomeric carbon C-1″ (δ_C 100.1) (Figure S37, Supporting Information). The proton at C-4″ (1H, dt, J = 7.7, 3.8 Hz, δ_H 3.97) was determined to be in an axial orientation to correspond to the 7.7 Hz coupling constant observed. H-3″ (1H, dd, J = 5.8, 3.7 Hz, δ_H 3.81) was placed in an equatorial orientation, consistent with its 3.7 Hz coupling constant. In turn, H-2″ (1H, dd, J = 5.5, 3.8 Hz, δ_H 3.99) was positioned equatorially due to its small coupling constants of 5.5 and 3.8 Hz. Based on this analysis, the sugar was assigned as an arabinopyranose. Acid hydrolysis, similar to the procedure described for compound 3, was performed to confirm the identity and the absolute configuration of the saccharide.²⁹ After comparison and similarities with the spectra of compound 3, the sugar was assigned as l-arabinose. The (2R,3R) absolute configuration of compound 4 was determined as described for compound 3. Accordingly, the structure of compound 4 was proposed as (2R,3R)-dihydrokaempferol-5-O-β-l-arabinopyranoside.

Moreover, NMR and OR comparison with the literature confirmed the isolation from B. yunnanensis of (–)-9,9′-O-diferuloylsecoisolariciresinol (5),³² (+)-9,9′-di-O-feruloyl-5,5′-dimethoxysecoisolariciresinol (6),³³ wikstresinol (7),³⁴ 9′-O-(E)-feruloyl-5,5′-dimethoxylariciresinol (8),³⁵ and N-trans-feruloyltyramine (9),³⁶ 3-hydroxy-1-(4-hydroxy-3,5-dimethoxyphenyl)-propan-1-one (13),^37,38 evofolin B (14),³⁹ and (–)-(7R,8S,7′E)-9-O-(E)-feruloyl-4,9-dihydroxy-3,3′-dimethoxy-4′,7-epoxy-8,5′-neolignan-7′-en-9′-al (15),⁴⁰ in addition to the previously mentioned catechin (10),¹³ epicatechin (11),¹⁴ and (+)-taxifolin (12).¹⁵

Those compounds isolated in sufficient quantities (5 and 7-12) were tested against the OVCAR3 ovarian and MDA-MB-435 breast human cancer cell lines. Of these, compounds 5, 8, and 12 each showed a selective activity against the ovarian cell line with IC₅₀ values of 0.51, 3.4 and 9.3 μM, respectively (Table 3).

Table 3.

Cytotoxic Activities of Compounds 5 and 7-12.

IC50 value (μM)
compound	OVCAR3	MDA-MB-435
5	0.51	>10
7	>10	>10
8	3.4	>10
9	>10	>10
10	>10	>10
11	>10	>10
12	9.3	>10
paclitaxel	0.005	0.0025

To investigate the cytotoxicity of the most potent compound 5 further, a dose-response study against the OVCAR3 cell line with increasing concentrations tested (16 nM to 10 μM) was conducted. The results showed that 5 is cytotoxic to OVCAR3 cells, with an IC₅₀ value of 521.3 ± 84.6 nM (Table 3). Additional dose-response curves at different time points for OVCAR3 and OVCAR8 cells were generated. It was observed that while 0.1 μM of 5 did not induce discernible cytotoxicity, there was no significant difference between the 1 and 10 μM concentration levels for both cell lines. Therefore, 1 μM of 5 was confirmed to be the optimal concentration (Figure 5). In addition, in previous investigations, compound 5 exhibited cytotoxicity against other human cancer cell lines such as lung (A549, EC₅₀ = 2.52 μM) and breast (MCF-7, EC₅₀ = 5.46 μM) carcinomas.⁴¹ Western blot analyses were performed for cleaved PARP (cPARP) as a marker for apoptosis. Treatment with 5 (1 μM) led to a significant increase in cPARP levels at 24 h in OVCAR3 cells. For this experiment, GAPDH served as the loading control (Figure 6). Additionally, compound 5 increased significantly the apoptotic cell population (Figure 7) and decreased the expression of BCL-2, one of the anti-apoptotic proteins (Figure 8).

Figure 5.

Dose-response over time curves against the OVCAR3 and OVCAR8 cancer cell lines for compound 5.

Figure 6.

Western blot assay of compound 5 against the OVCAR3 cancer cell line.

Figure 7.

Annexin-V/Propidium Iodide staining assay of compound 5 against the OVCAR3 cell line (*p<0.05).

Figure 8.

Western blot analysis of the reduction of BCL-2, an anti-apoptosis protein, by compound 5 in the OVCAR3 cell line.

In summary, this study has provided a first phytochemical and cytotoxic investigation of Beilschmiedia yunnanensis. This species was found to produce a variety of compounds, including lignans and flavonoids and four new compounds belonging to these two classes were purified and characterized. Thus, beilschmiedianins A (1) and B (2) are herein reported as new lignans along with the new flavonoids [(2R,3R)-dihydrokaempferol-5-O-β-l-arabinosyl-(2→1)-α-l-rhamnopyranoside (3) and [(2R,3R)-dihydrokaempferol-5-O-β-l-arabinopyranoside (4). Beilschmiedia yunnanensis also showed cytotoxic potency against the ovarian cancer cell line OVCAR3 with (–)-9,9′-O-diferuloylsecoisolariciresinol (5), 9′-O-(E)-feruloyl-5,5′-dimethoxylariciresinol (8) and (+)-taxifolin (12) having an IC₅₀ of 0.51, 3.4 and 9.3 μM, respectively. Mechanistic studies on compound 5 showed that it reduces BCL-2 protein expression in the cells, likely increasing cPARP levels and inducing apoptosis.

EXPERIMENTAL SECTION

General Experimental Procedures.

Optical rotations were measured on an Anton-Paar MCP 150 polarimeter (Anton-Paar, Ashland, VA, USA). A Hitachi U-2910 UV spectrometer (Hitachi, Tokyo, Japan) was used to record the UV data. ECD spectra were measured using a JASCO J-810 spectropolarimeter (JASCO, Easton, MD, USA). IR spectra were obtained on a Thermo-Nicolet 6700 Fourier-transform infrared spectrometer equipped with an ATR platform (Smart iTX Diamond) (Thermo Scientific, Waltham, MA, USA). 1D and 2D NMR spectra were obtained either on a Bruker AVIII400HD, AVIII600HD or a Bruker AVIII700HD NMR instrument using default experiments, and the data processed using TopSpin software (Bruker, Billerica, MA, USA). HRESIMS data were acquired on a Thermo Q-Exactive Orbitrap mass spectrometer equipped with a Vanquish-H UHPLC and an Accucore^™ reversed-phase-MS column (Thermo Scientific, 2.1 × 50 mm, 7.5 μm, 80 Å) (Thermo Fisher Scientific). The samples were diluted to 5 μg/mL with HPLC grade MeOH. HESI source parameters were derived from Thermo’s Source Auto-Defaults function which is based on LC flow rate and were not further optimized. A gradient method was used, with a flow rate of 0.7 mL/min with solvent A (H₂O + 0.1% formic acid) and solvent B (CH₃CN + 0.1% formic acid), starting at 2%, held for 0.5 min followed by a gradient from 2% to 98% B from 0.5 to 1.5 min. The eluent was held at 98% B until 2.5 min followed by a rapid return to 2% B at 2.7 min. The column was then equilibrated at 2% B until 3.2 min. The instrument was operated in the positive- or negative-ion mode as indicated for each compound, with a resolution of 35,000, an AGC target of 5e5, a maximum IT of 100 ms, a number of scan ranges of 1, and a scan range from 200 to 1500 m/z. The divert valve opened at 0.45 min and closed at 3.0 min. The source was positioned on the D ring, fully back and centered. The data were analyzed using FreeStyle^™ 1.8 SP2 QF1 software version 1.8.65.0 (Thermo Fisher Scientific). HPLC separations were performed on either a Waters HPLC system equipped with a 600 controller, a 717 Plus autosampler, and a 2487 dual wavelength absorbance detector or a Hitachi HPLC system equipped with a L-2200 autosampler, a L-2400 UV detector, and two PrepPumps. The columns used for these analyses were Phenomenex reversed-phase (C₁₈) and silica gel (normal phase) semi-preparative columns (10 × 250 mm, 5 μm, 100 Å) and a Phenomenex reversed-phase (C₁₈) preparative column (21.2 × 250 mm, 5 μm, 100 Å). A combiFlash column silica gel 25 g (Biotage SNAP Cartridge, KP-sil 25 g) was utilized. Sephadex LH-20 (GE Healthcare Bio-Sciences AB) was used for open column chromatographic separation as well as silica gel column chromatography (Sorbtech, Sorbent Technologies). HPLC grade solvents and analytical grade reagents were purchased from Fisher Scientific, Alfa Aesar and MilliporeSigma.

Plant Material.

The branches of Beilschmiedia yunnanensis Hu (Lauraceae) were collected in May 2009 (screening sample) and December 2010 (isolation sample), in Lao Cai Province, Sapa District, Sin Chai locality, Vietnam by D.D.S. and T.N.N, who also identified this species. Voucher specimens (original collection Soejarto et al. 14383 for the screening sample; Soejarto et al. 14835 for the isolation sample) of this species have been deposited in the John G. Searle Herbarium of the Field Museum, Chicago, IL, USA, under the accession numbers FM-2290170 (for the screening sample) and FM-2319985 (for the isolation sample).

Extraction and Isolation.

The air-dried branches of B. yunnanensis (2.4 kg) were extracted with 100% MeOH (6 L) by maceration at room temperature with occasional shaking for 24 h, for a total of five days (5 × 6 L). The macerate from each extraction was filtered, dried using a rotavapor (40°C), and combined to give a total of 161.3 g of a dark-brown gummy semi-solid. A portion (150 g) of this was suspended in MeOH-H₂O (9:1, 1.5 L) and partitioned with hexanes (3×1.5 L), which was dried to give a hexanes-soluble partition (D1, 7.9 g). In turn, the MeOH-H₂O (9:1) partition was completely dried and resuspended in distilled H₂O (1 L) and successively partitioned first with CHCl₃ (5×1.5 L) and next with EtOAc (5×1.5 L). This provided 10.0 g of the CHCl₃-, 15.8 g of the EtOAc-, and >110 g of the H₂O (D4)-soluble partitions, respectively. The CHCl₃ and EtOAc partitions were re-suspended with their respective solvent (1 and 2 L, respectively) and were “partially detannified” using 1% NaCl (2× 1 L).⁴² After the “detannification” step, 1.5 g of CHCl₃-insoluble material and 6.3 g of a CHCl₃-soluble partition (D2) were obtained. For the EtOAc partition, 7.2 g of a soluble extract (D3) and 0.14 g of an interfacial material (D3a) were obtained. The remaining 1% NaCl partition was re-extracted with 4× 1.5 L of EtOAc to give 4.0 g of recovered material (D3Na).

Aliquots of the above partitions (D1, D2, D2as, D3, D3a, D3Na and D4) were tested for cytotoxic activity against the MDA-MB-231 breast, OVCAR3 ovarian, MDA-MB-435 melanoma, and HT-29 colon human cancer cell lines. The CHCl₃ partition (D2) was the only one found active, with inhibitory activity against HT-29 cells causing 21% survival at 20 μg/mL. An aliquot (5.3 g) was fractionated by silica gel column chromatography (60 × 5 cm) using gradient mixtures of CH₂Cl₂/MeOH (100:0, 100:1, 98:1: 95:1, 90:1, 80:1, 70:1 60:1, 50:1, 40:1, 30:1, 20:1, 10:1, 5:1, 2:1, 1:1, and 0:1). The fractions were combined based on the similarities of their TLC profiles to afford a total of 24 pooled fractions (D2F1-D2F24), which were tested against the above four cancer cell lines.

Fraction D2F7 (102.3 mg) showed 16, 12, 13, and 9% growth inhibition against the MDA-MB-231, OVCAR3, MDA-MB-435, and HT-29 cancer cell lines, respectively, at 20 μg/mL. A portion of this fraction was purified by HPLC, using the preparative reversed-phase column mentioned above and a gradient of MeOH-H₂O (50% to 90% of MeOH, flow rate 7 mL/min for 40 min), yielding compounds 13 (t_R 11.02 min, 1.6 mg), 14 (t_R 11.55 min, 1.3 mg), 8 (t_R 30.19 min, 7.1 mg), 15 (t_R 34.34 min, 0.9 mg), and 5 (t_R 39.15 min, 27.3 mg). Fraction D2F8 (100.0 mg) showed 57, 20, 26, and 15% growth inhibition against MDA-MB-231, OVCAR3, MDA-MB-435, and HT-29 cells, respectively, at 20 μg/mL. A portion of this fraction was purified by HPLC, using the normal-phase semi-preparative column mentioned above, with isocratic elution by hexanes and denatured EtOH (80:20, flow rate 3 mL/min, for 45 min), yielding compounds 2 (t_R 40.25 min, 1.1 mg) and 7 (t_R 18.33 min, 3.7 mg). Fraction D2F9 (130.2 mg) showed 55, 22, 38, and 11% growth inhibition against MDA-MB-231, OVCAR3, MDA-MB-435 and HT-29 cells, respectively, at 20 μg/mL. The fraction was purified using HPLC having the normal-phase semi-preparative column described above, with the same elution profile used for fraction D2F8, yielding an impure compound 6 (t_R 27.77 min) and compound 1 (t_R 36.74 min, 1.6 mg). A 1.7 mg sample of purified compound 6 was obtained by continued precipitation and centrifugation using MeOH. The same gradient and column were used to purify compound 9 (t_R 16.56 min, 48.2 mg) from fraction D2F10 (201.0 mg), which showed 52, 25, 42, and 11% growth inhibition against MDA-MB-231, OVCAR3, MDA-MB-435 and HT-29 cells, respectively, at 20 μg/mL.

The detannified EtOAc partition (D3) was found to have a slight cytotoxicity against MDA-MB-231 and, thus, was fractionated (3 g) by Sephadex LH-20 column chromatography (65 cm × 3.5 cm), using CH₂Cl₂-MeOH (1:1). Altogether, 360 fractions (4 mL each) were collected and combined according to their TLC similarities to give 10 pooled fractions (D3F1-D3F10), which were tested against OVCAR3 and MDA-MB-435 cells. Fraction D3F4 (446.9 mg) showed 16.9 and 10.4% growth inhibition against OVCAR3 and MDA-MB-435 cells at 20 μg/mL. Subsequently, this was mixed with 1.79 g of silica gel, dried, and loaded onto an empty cartridge, and fractionated in a CombiFlash unit equipped with a column containing 24 g silica gel (Biotage SNAP Cartridge, KP-sil 25 g), using an isocratic program of CHCl₃-MeOH-H₂O (15:6:1, flow rate 10 mL/min, for 100 min). This afforded 206 fractions, which were pooled to provide 14 major fractions (D3F4F1- D3F4F14) based on their TLC profiles. Fraction D3F4F4 (79.1 mg) showed 9.9 and 8.9% growth inhibition against OVCAR3 and MDA-MB-435 cells, respectively, at 20 μg/mL. A 68.4 mg aliquot was purified by HPLC using the reversed-phase semi-preparative column described above, with a gradient solvent system composed of CH₃CN/H₂O (from 10% to 20% of CH₃CN for 50 min then 20% to 50% for 20 min, flow rate of 2 mL/min). From this separation, compounds 4 (t_R 46.99 min, 2.8 mg) and 3 (t_R 50.92 min, 3.4 mg) were obtained. Fraction D3F4F5 (34 mg) showed 10.6 and 8.6% growth inhibition against OVCAR3 and MDA-MB-435 cells, respectively, at 20 μg/mL. A portion was purified by HPLC using the reversed-phase semi-preparative column described previously, with a gradient solvent system composed of MeOH/H₂O (20% to 60% of MeOH, flow rate of 2 mL/min, for 30 min). Compound 3 (t_R 29.62 min, 1.0 mg) was obtained from this fraction.

Fraction D3F4F5 (34 mg) showed 10.6 and 8.6% growth inhibition against OVCAR3 and MDA-MB-435 cells, respectively, at 20 μg/mL. A portion was purified by HPLC using the reversed-phase semi-preparative column described previously, with a gradient solvent system composed of MeOH/H₂O (20% to 60% of MeOH, flow rate of 2 mL/min, for 30 min). Compound 3 (t_R 29.7 min, 1.0 mg) was obtained from this fraction. Fraction D3F6 (448.1 mg) showed 16.0 and 10.0% growth inhibition against OVCAR3 and MDA-MB-435 cells, respectively, at 20 μg/mL. A part of this fraction was purified by HPLC, using the reversed-phase preparative column described above, with a gradient solvent system of MeOH/H₂O (20 to 60% MeOH, flow rate 8.00 mL/min, for 60 min). From this separation, compounds 10 (t_R 21.73 min, 48.8 mg), 11 (t_R 29.90 min, 92.3 mg), and 12 (t_R 42.76 min, 36.0 mg) were obtained.

Acid Hydrolysis of Compounds 3 and 4.

The acid hydrolysis and the sugar identifications of the flavonoid glycosides 3 and 4 were conducted following a published procedure.²⁹ Thus, ca. 1 mg of each compound was dissolved in 500 μL of HCl (1 M) and heated at 100 °C for 1 h. The solution was diluted with deionized water and neutralized by performing ion-exchange chromatography over Amberlite IRA-400(Cl), filtered, and dried. Each dried sample was dissolved in pyridine-d₅ with 3 mg of l-cysteine methyl ester hydrochloride, heated for 1 h at 60 °C, and kept overnight at room temperature. The ¹H NMR spectrum of each sugar was recorded, using TOCSY, COSY, and NOESY experiments, as necessary.

ECD and NMR Calculations.

The 2D structures of compounds 1, 2, and (–)-medioresinol were initially sketched using Maestro, Schrödinger Release 2020-4: Maestro version 12.6.144 (Schrödinger, LLC, New York, NY, 2020), followed by 3D structure generation and energy minimization at physiological pH (7.4) using the LigPrep, Schrödinger Release 2020–4: LigPrep, (Schrödinger, LLC, New York, NY, 2020) module within the Schrödinger suite, employing the OPLS3e force field. Conformational searches were performed in the gas phase using MacroModel, Schrödinger Release 2020-4: MacroModel (Schrödinger, LLC, New York, NY, 2020) with the Mixed Torsional/Low-Mode sampling approach. An energy window of 10 kcal/mol was applied to retain all relevant low-energy conformers, and redundant structures were excluded using an RMSD threshold of 0.5 Å. The remaining conformers were geometry-optimized using the mPW1PW91/6-311+G(2d,p) basis set and vibrational frequency analysis, as described in our previous work.⁴³ All optimized structures were confirmed as true minima by the absence of imaginary frequencies. These optimizations were carried out in MeOH using the polarizable continuum model (PCM)⁴⁴ as implemented in Gaussian 16 Rev. B.01 (Gaussian, Inc., Wallingford, CT, 2016).⁴⁵ Time-dependent density functional theory (TDDFT)⁴⁶ calculations at the mPW1PW91/6-311+G(2d,p) level were then employed to simulate the ECD spectra, considering 100 singlet excited states. ECD curves were generated using SpecDis 1.71⁴⁷ with a Gaussian band shape and half-band widths between 0.22 and 0.29 eV to match experimental spectra. Boltzmann-weighted populations of each conformer were determined at 298 K. Molecular orbital (MO) visualization was carried out using Avogadro 1.2.0⁴⁸ with isosurface values between 0.02 and 0.04 e−/au³. NMR chemical shift calculations were also conducted at the same level of theory in methanol.

Compound 1 (beilschmiedianin A) [(7R,7′R,8S,8′S,8″R)-4′,4″,9″-trihydroxy-3,5,3′,3″-tetramethoxy-4,8″-oxy-7,9′:7′,9-diepoxy-8,8′-sesquilignan-7″-one].

Light brown gum/solid; [α]²⁰_D −7.9 (c 0.13, CH₃OH); UV (MeOH) λ_max (log ε) 209 (4.58), 229 (4.26), 280 (3.94), 309 (3.82) nm; ECD (MeOH) λ_max (Δε) 203 (− 3.13), 211 (− 1.89), 240 (+ 1.68), 266 (− 0.59), 283 (− 0.53), 330 (+ 0.71); IR ν_max 3380, 1591, 1505, 1458, 1271, 1222, 1121, 1029 cm⁻¹; ¹H NMR (400 MHz, methanol-d₄) and ¹³C NMR (100 MHz, methanol-d₄), see Table 1; positive-mode HRESIMS m/z 583.21715, calcd for [C₃₁H₃₄O₁₁+H]⁺, 583.21739.

Compound 2 (beilschmiedianin B) [(7R,7′R,7″R,8S,8′S,8″R)-9″-feruloyl-4′,4″-dihydroxy-3,5,3′,3″-tetramethoxy-4,8″-oxy-7,9′:7′,9-diepoxy-8,8′-dilignan-7″-ol].

Off-white solid; [α]²⁰_D −11.32 (c 0.13); UV (MeOH) λ_max (log ε) 207 (4.56), 228 (4.23), 281 (3.85), 322 (3.78) nm; ECD (MeOH) λ_max (Δε) 200.5 (+ 1.06), 209 (− 3.82), 241 (− 0.58), 279 (− 0.41); IR ν_max 3384, 1592, 1515, 1271, 1124, 1032 cm⁻¹; ¹H NMR (400 MHz, methanol-d₄) and ¹³C NMR (100 MHz, methanol-d₄), see Table 1; negative-mode HRESIMS m/z 759.26626, calcd for [C₄₁H₄₄O₁₄ - H]⁻, 759.26583.

Compound 3 [(2R,3R)-dihydrokaempferol-5-O-β-l-arabinosyl-(2→1)-α-l-rhamnopyranoside].

Brown resin; [α]²⁰_D −7.1 (c 0.23, CH₃OH); UV (MeOH) λ_max (log ε) 203 (4.40), 223 (4.31), 283 (4.03), 318 (3.75) nm; ECD (MeOH) λ_max (Δε) 202 (− 1.41), 217 (+ 4.76), 238 (+ 4.22), 289 (− 3.21), 325 (+ 2.32); IR ν_max 3339, 1610, 1584, 1247, 1125, 1074, 1053 cm⁻¹; ¹H NMR (600 MHz, methanol-d₄), and ¹³C NMR (150 MHz, methanol-d₄), see Table 2; negative-mode HRESIMS m/z 565.15653, calcd for [C₂₆H₃₀O₁₄ - H]⁻ 565.15628.

Compound 4 [(2R,3R)-dihydrokaempferol-5-O-β-l-arabinopyranoside]:

Yellow resin; [α]²⁰_D −6.4 (c 0.13, CH₃OH); UV (MeOH) λ_max (log ε) 203 (4.28), 223 (4.23), 283 (3.96), 318 (3.76) nm; ECD (MeOH) λ_max (Δε) 202 (− 4.55), 219 (+ 4.96), 237 (+ 3.86), 289 (− 4.54), 330 (+ 2.51); IR ν_max 3341, 1611, 1585, 1250, 1124, 1071, 1008 cm⁻¹; ¹H NMR (600 MHz, methanol-d₄) and ¹³C NMR (150 MHz, methanol-d₄), see Table 2; negative-mode HRESIMS m/z 419.09879, calcd for [C₂₀H₂₀O₁₀ - H]⁻ 419.09837.

Bioassays.

Cell Culture.

Two high-grade serous ovarian carcinoma (HGSOC) cancer cell lines (OVCAR3, OVCAR8), and a melanoma cell line (MDA-MB-435) were purchased from the American Type Culture Collection (ATCC) or acquired from the NCI 60-cell line panel. OVCAR3 and MDA-MB-435 cells were grown in RPMI 1640 medium (Thermo Fisher, #11875085) and OVCAR8 in DMEM medium (Thermo Fisher, #11995065), both supplemented with 10% fetal bovine serum (FBS) (GeminiBio, #100-106) and 1% penicillin/streptomycin (P/S) (Thermo Fisher, #15070063). OVCAR8 cells were grown in DMEM medium (Thermo Fisher, #11965092) with 10% FBS and 1% P/S. Cultured cells were maintained at 37 °C with 5% CO₂ in a humidified incubator. Cells were passaged a maximum of 20 times. Cell lines were tested as mycoplasma-free and validated using short tandem repeat analysis.⁴⁹

CellTiter-Blue Cell Viability Assay.

Altogether, 5,000 cells per well were allowed to attach overnight after being seeded in a clear flat-bottomed 96-well plate. The cells were treated with the test extracts, fractions (2 and 20 μg/mL), compounds (25, 5, 1, 0.2, 0.04 μM) and paclitaxel (Taxol^™) (10, 2, 0.4, 0.08, 0.016 μM) suspended in DMSO. To achieve a wide dosage range, the final vehicle concentration was set at 0.25%. Cells were incubated for 8, 24, 48, or 72 h. To determine cell viability, 18 μL of Cell Titer-Blue^® (Promega, #G8082) reagent was incubated in the wells for 3 h to measure cellular protein content. GraphPad Prism Software was used to generate dose-response curves after the treatment measurements were normalized to vehicle.⁴⁹

Immunoblot Analysis.

T75 flasks were seeded with 30,000 OVCAR3 cells per flask, allowed to adhere overnight, and treated with compound 5 (1 μM), paclitaxel (Taxol^™) (50 nM), and vehicle control (DMSO) for 24 hours. After culturing and treatment, cells were lysed in RIPA lysis buffer (50 mmol/L Tris, pH 7.6, 150 mmol/L NaCl, 1% Triton X-100, 0.1% SDS) with protease (Roche Applied Science, #11697498001) and phosphatase (Sigma, #P0044) inhibitors. All steps were done on ice for 45 min. The cells were incubated at –80 °C and centrifuged (4°C, 10 min, 13,000g). The Bradford Assay (Bio-Rad, #5000205) was used to determine the protein concentrations in the lysate, and proteins (30 μg) were resolved on an SDS-PAGE gel. Proteins were transferred to a nitrocellulose membrane. Following treatment for 20 min at room temperature with a 5% milk block, the membranes were exposed to the primary antibodies at a concentration of 1:1,000 overnight at 4° C. SuperSignal West Femto substrate (Thermo Fisher, #34096) was used for signal visualization after incubation of the membrane with 1:1,000 anti-rabbit secondary antibody (Cell Signaling Technology; #7074). FluorChem E system (ProteinSimple) was used for imaging.⁵⁰

Annexin V/Propidium Iodine Staining.

T25 flasks were seeded with 10,000 OVCAR3 cells per flask, allowed to adhere overnight, and treated with compound 5 (1 μM), paclitaxel (Taxol^™) (50 nM), and vehicle control (DMSO) overnight. Cells were subjected to an Annexin V–FITC/Propidium Iodide Apoptosis Assay (Nexcelom Biosciences), according to instructions provided by the manufacturer. Using K2 Cellometer (Nexcelom Biosciences), fluorescence was detected, and the FCS Express program (De Novo Software) allowed analysis of the data.⁵¹ For quantification of fluorescent signal each replicate contained at least 2,000 cells. In GraphPad Prism Software, Tukey’s multiple comparisons or Dunnetťs multiple comparisons to vehicle control within each group were used to generate the statistics using a one-way ANOVA. The p-value was set at 0.05.

Supplementary Material

The Supporting Information is available free of charge on the ACS Publication website, including the mass, NMR, CD, and UV spectra of compounds 1 to 4. Tables supporting the calculated ECD for compound 1 (Table S1) and NMR data for compound 2 (Tables S2 and S3) are also included (PDF).

Data Availability

The raw NMR data have been uploaded to the Natural Products Magnetic Resonance Database (NP-MRD, https://np-mrd.org/) website, with a deposition ID of NP0351363 (compound 1), NP0351364 (compound 2), NP0351365 (compound 3), NP0351366 (compound 4) and NP0351367 (compound 5) being assigned.