The following instruments were used to obtain physical data: a SEPA-300 digital polarimeter (Horiba Ltd., Kyoto, Japan, l = 5 cm) for specific rotations; an UV-1600 spectrometer (Shimadzu Co., Kyoto, Japan) to record UV spectra; a FTIR-8100 spectrometer (Shimadzu Co.) to measure IR spectra; a JNM-ECA800 (800 MHz), JNM-ECA700 (700 MHz), JNM-ECA500 (500 MHz), and JNM-ECS400 and JNM-AL400 (400 MHz) spectrometers (JEOL Ltd., Tokyo, Japan) to determine ¹H NMR spectra; JNM-ECA800 (200 MHz), JNM-ECA700 (175 MHz), JNM-ECA500 (125 MHz), and JNM-ECS400 and JNM-AL-400 (100 MHz) spectrometers (JEOL Ltd.) to record ¹³C NMR spectra in CDCl₃ at room temperature (25°C) with tetramethylsilane as an internal standard; an Exactive Plus Orbitrap mass spectrometer (Thermo Fisher Scientific Inc., Waltham, MA, USA) to measure ESIMS and HRESIMS; an HPLC detector, SPD-10Avp UV-Vis (Shimadzu Co.); and Cosmosil 5C₁₈-MS-II (Nacalai Tesque, Inc., Kyoto, Japan) HPLC columns (4.6 mm i.d. × 250 mm and 20 mm i.d. × 250 mm) for analytical and preparative purposes, respectively.

The following materials and experimental conditions were used for the column chromatography (CC): normal-phase silica gel CC, silica gel 60N (Kanto Chemical Co., Ltd., Tokyo, Japan; 63–210 mesh, spherical, neutral); reversed-phase ODS CC, Chromatorex ODS DM1020T (Fuji Silysia Chemical, Ltd., Aichi, Japan; 100–200 mesh); TLC, pre-coated TLC plates with silica gel 60F₂₅₄ (Merck, Darmstadt, Germany, 0.25 mm, normal-phase) and silica gel RP-18 WF_254S (Merck, Darmstadt, Germany, 0.25 mm, reversed-phase); reversed-phase HPTLC, pre-coated TLC plates with silica gel RP-18 WF_254S (Merck, 0.25 mm); detection was performed by spraying 1% Ce(SO₄)₂-10% aqueous H₂SO₄, followed by heating.

The flowers of Mammea siamensis were collected from the Nakhonsithammarat Province, Thailand, in September 2006, as described previously (Morikawa et al., 2012; Ninomiya et al., 2016). The plant material was identified by one of the authors (Y. P.). A voucher specimen (2006.09. Raj-04) for this plant has been deposited in our laboratory.

Dried flowers of M. siamensis (1.8 kg) were extracted three times with MeOH under reflux for 3 h. Evaporation of the combined extracts under reduced pressure afforded the MeOH extract (463.7 g, 25.66%). An aliquot (413.7 g) of the extract was partitioned into an EtOAc–H₂O (1:1, v/v) mixture to furnish an EtOAc-soluble fraction (110.34 g, 6.84%) and an aqueous phase. An aliquot (89.45 g) of the EtOAc-soluble fraction was subjected to normal-phase silica gel CC [3.0 kg, n-hexane–EtOAc (10:1 → 7:1 → 5:1, v/v) → EtOAc → MeOH] to give 11 fractions [Fr. 1 (3.05 g), Fr. 2 (2.86 g), Fr. 3 (11.71 g), Fr. 4 (1.62 g), Fr. 5 (4.15 g), Fr. 6 (6.29 g), Fr. 7 (2.21 g), Fr. 8 (2.94 g), Fr. 9 (10.23 g), Fr. 10 (11.17 g), and Fr. 11 (21.35 g)]. Fraction 5 (4.15 g) was subjected to reversed-phase silica gel CC [120 g, MeOH–H₂O (80:20 → 85:15, v/v) → MeOH → acetone] to afford six fractions [Fr. 5-1 (115.7 mg), Fr. 5-2 (2789.8 mg), Fr. 5-3 (515.4 mg), Fr. 5-4 (430.0 mg), Fr. 5-5 (119.2 mg), and Fr. 5-6 (110.0 mg)] as reported previously (Ninomiya et al., 2016). Fraction 5-2 (517.0 mg) was purified by HPLC [Cosmosil 5C₁₈-MS-II, MeOH−1% aqueous AcOH (85:15, v/v)] to give mammea A/AC cyclo F (35, 4.6 mg, 0.0019%) (Morel et al., 1999; Prachyawarakorn et al., 2000; Guilet et al., 2001) together with mammeas A/AA (17, 101.2 mg, 0.0418%), A/AC (19, 112.9 mg, 0.0466%), A/AA cyclo D (24, 2.7 mg, 0.0011%), E/BC cyclo D (29, 14.0 mg, 0.0058%), and E/BD cyclo D (30, 1.8 mg, 0.0015%) (Mahidol et al., 2002). Fraction 5-3 (515.4 mg) was purified by HPLC [Cosmosil 5C₁₈-MS-II, MeOH−1% aqueous AcOH (85:15, v/v)] to give mammea A/AA cyclo F (34, 13.2 mg, 0.0010%) (Prachyawarakorn et al., 2000; Guilet et al., 2001) together with 19 (45.6 mg, 0.0035%), 24 (14.9 mg, 0.0011%), mammeas A/AB cyclo D (25, 46.4 mg, 0.0035%) and A/AC cyclo D (26, 30.1 mg, 0.0023%). Fraction 6 (6.29 g) was subjected to reversed-phase silica gel CC [200 g, MeOH–H₂O (80:20 → 90:10 → 95:5, v/v) → MeOH → acetone] to afford 10 fractions [Fr. 6-1 (44.7 mg), Fr. 6-2 (157.2 mg), Fr. 6-3 (928.8 mg), Fr. 6-4 (3117.0 mg), Fr. 6-5 (128.8 mg), Fr. 6-6 (487.1 mg), Fr. 6-7 (230.8 mg), Fr. 6-8 (280.5 mg), Fr. 6-9 (102.9 mg), and Fr. 6-10 (96.5 mg)] as reported previously (Morikawa et al., 2012; Ninomiya et al., 2016). Fraction 6-4 (536.2 mg) was purified by HPLC [Cosmosil 5C₁₈-MS-II, MeOH−1% aqueous AcOH (90:10, v/v)] to give kayeassamin I (1, 7.2 mg, 0.0032%) (Win et al., 2008b), mammeasin E (2, 16.5 mg, 0.0073%), and 35 (11.0 mg, 0.0049%) together with mammeasins A (4, 65.8 mg, 0.0293%) and B (5, 21.6 mg, 0.0096%), surangin B (12, 58.2 mg, 0.0259%), 17 (17.0 mg, 0.0076%), mammea A/AB (18, 10.7 mg, 0.0048%), and 19 (112.6 mg, 0.0501%). Fraction 7 (2.21 g) was subjected to reversed-phase ODS CC [47.0 g, MeOH-H₂O (60:40 → 80:20 → 90:10, v/v) → MeOH → acetone] to afford five fractions [Fr. 7-1 (187.3 mg), Fr. 7-2 (912.0 mg), Fr. 7-3 (275.2 mg), Fr. 7-4 (30.0 mg), and Fr. 7-5 (44.0 mg)]. Fraction 7-2 (912.0 mg) was purified by HPLC [column: Cosmosil 5C₁₈-MS-II, detection: UV (230 nm), mobile phase: MeOH-1% aqueous H₂O (85:15, v/v)] to give mammeasin E/BC (23, 99.0 mg, 0.0076%) (Yang et al., 2005). Fraction 7-3 (275.2 mg) was purified by HPLC [Cosmosil 5C₁₈-MS-II, UV (230 nm), MeOH-1% aqueous AcOH 85:15, v/v] to give 1 (52.1 mg, 0.0040%), 2 (34.1 mg, 0.0026%), and mammeasin F (3, 19.5 mg, 0.0015%). Fraction 9 (10.23 g) was subjected to reversed-phase silica gel CC [300 g, MeOH–H₂O (80:20 → 90:10, v/v) → MeOH → acetone] to afford five fractions [Fr. 9-1 (2809.0 mg), Fr. 9-2 (5678.0 mg), Fr. 9-3 (385.9 mg), Fr. 9-4 (422.0 mg), and Fr. 9-5 (51.9 mg)] as reported previously (Morikawa et al., 2012; Ninomiya et al., 2016). Fraction 9-1 (544.5 mg) was purified by HPLC [Cosmosil 5C₁₈-MS-II, MeOH−1% aqueous AcOH (85:15, v/v)] to give deacetylmammeas E/AA cyclo D (31, 1.3 mg, 0.0005%) (Mahidol et al., 2007) and E/BB cyclo D (32, 6.1 mg, 0.0023%) (Mahidol et al., 2007) together with kayeassamins E (9, 28.6 mg, 0.0113%), F (10, 98.7 mg, 0.0390%), and G (11, 43.4 mg, 0.0171%), deacetylmammea E/BC cyclo D (33, 18.6 mg, 0.0073%), and benzoic acid (10.9 mg, 0.0043%).

A solution of kayeassamin A (8, 9.0 mg) in dry-toluene (2.0 mL) was treated with 2,3-dichloro-5,6-dicyano-p-benzoquinone (DDQ, 10.0 mg) and the solution stirred at room temperature (25°C) for 4 h. The aqueous solution was saturated with sodium hydrogen carbonate (NaHCO₃) and extracted with EtOAc. The EtOAc extract was washed with brine then dried over anhydrous magnesium sulfate (MgSO₄) and filtered. Removal of the solvent under reduced pressure gave a residue, which was purified by HPLC [Cosmosil 5C₁₈-MS-II, MeOH−1% aqueous AcOH (85:15, v/v)] to give kayeassamin I (1, 3.8 mg, 46%). Through the similar procedure, mammeasin E (2, 3.3 mg, 38%) and mammeasin F (3, 2.0 mg, 17%) were obtained from surangins D (14, 9.6 mg) and C (13, 12.7 mg), respectively.

The experiment was performed in accordance with previously reported methods (Matsuda et al., 2001; Lee et al., 2012; Koseki et al., 2015) with slight modifications. In brief, the assay was performed in 48-well microplates (Sumitomo Bakelite Co., Ltd., Tokyo, Japan). The reaction solution was pre-incubated with or without a test sample (5 μL/well, dissolved in DMSO), in a potassium phosphate buffer (40 mM, pH 6.5, 490 μL/well) containing substrate (0.35 nmol of testosterone, Tokyo Chemical Industry Co., Ltd., Tokyo, Japan) and NADPH (10 nmol, Oriental Yeast Co., Ltd., Tokyo, Japan) at room temperature (25°C) for 20 min. The enzymatic reaction was initiated by the addition of rat liver S9 fractions (10 μL/well, dissolved in the phosphate buffer, 20.6 μg/well, Oriental Yeast Co., Ltd., Tokyo, Japan, lot no. 109031513) at 37°C for 30 min. After incubation, the reaction mixture was immediately heated in boiling water for 2 min to stop the reaction. Then the reaction solution of each well was transferred to a microtube and extracted with 500 μL of EtOAc. After the microtube was centrifuged (10,000 rpm, 5 min), an aliquot of each EtOAc phase (300 μL) was transferred into another tube. The solvent in the tube was evaporated and the residue was dissolved in 30 μL of acetonitrile containing an internal standard (I.S.) fludrocortisone acetate (20 μg/mL, Sigma-Aldrich, Co., LLC, St. Louis, USA). An aliquot of 2 μL was injected into the HPLC under the following conditions [Instrument: a series LC-20A Prominence HPLC system (Shimadzu Co., Kyoto, Japan); Detection: UV (254 nm); Column: Cosmosil 5C₁₈-MS-II (Nakalai Tesque Inc., Kyoto, Japan, 5 μm particle size, 2.0 mm i.d. × 150 mm); Column temperature: 40°C; Mobile phase: MeOH–H₂O (60:40, v/v); Flow rate: 0.2 mL/min; retention time: 13.5 min for testosterone and 8.0 min for I.S. A similar procedure that described above was carried out for the control tubes. The 5α-reductase inhibitory activity was determined from the following equation using the peak area ratios (r = testosterone/I.S.). Experiments were performed in triplicate or quadruple, and IC₅₀ values were determined graphically. The 5α-reductase inhibitor finasteride (Tokyo Chemical Industry Co., Ltd., Tokyo, Japan) was used as a reference compound.

Control (C): enzyme (+), test sample (–); Test (T): enzyme (+), test sample (+); Blank (B): enzyme (–), test sample (+).

Values are expressed as mean ± S.E.M. One-way analysis of variance (ANOVA), followed by Dunnett's test, was used for statistical analysis. Probability (p) values < 0.05 were considered significant.

The male sex hormones, androgens, play a crucial role in the development, growth and function of the prostate, and other androgen-sensitive peripheral tissues. In the prostate gland, androgens are involved in benign prostatic hyperplasia and prostate cancer, as well as in skin disorders, such as acne, seborrhea, androgenic alopecia, and hirsutism. Among the androgens, testosterone is the most abundant in serum and secreted primarily by the testicles and ovaries. The enzyme steroid 5α-reductase catalyzes the conversion of testosterone to the most potent natural androgen, 5α-dihydrotestosterone (Yamana et al., 2010; Yao et al., 2011; Azzouni et al., 2012). Therefore, inhibition of testosterone 5α-reductase could be useful for the treatment of the above diseases. To date, three types of 5α-reductases, chronologically named types 1, 2, and 3 5α-reductases, have been described (Yamana et al., 2010; Azzouni et al., 2012; Titus et al., 2014). A type 2 and 3 5α-reductase inhibitor, finasteride, is currently marketed worldwide as a drug for benign prostatic hyperplasia and is also used in the treatment of hair loss (Heinzl, 1999; Tosti and Piraccini, 2000) and in the prevention of prostate cancer (Coltman et al., 1999). Therefore, 5α-reductase is considered a useful therapeutic target in the treatment and prevention of the above deceases. In particular, many heterocyclic compounds based on oxygen and nitrogen atoms often have good antiproliferative activity against a variety of solid tumor cell lines and are expected to be seeds of new anticancer agents (Sharma et al., 2018; Petel et al., 2019).

During our characterization studies on bioactive constituents from Thai natural medicines (Manse et al., 2017; Morikawa et al., 2018; Tanabe et al., 2018; Kobayashi et al., 2019), a methanol extract of the flowers of M. siamensis was found to inhibit 5α-reductase activity (IC₅₀ = 2.4 μg/mL). In order to investigate new 5α-reductase inhibitors, we conducted a search for the bioactive constituents from the flowers of M. siamensis.

In our previous report we described the isolation of 26 coumarins: mammeasins A (4, 0.0293%), B (5, 0.0115%), C (6, 0.0008%), and D (7, 0.0047%), kayeassamins A (8, 0.0578%), E (9, 0.0113%), F (10, 0.0390%), and G (11, 0.0171%), surangins B (12, 0.0271%), C (13, 0.0571%), and D (14, 0.0632%), 8-hydroxy-5-methyl-7-(3,7-dimethyl-octa-2,6-dienyl)-9-(2-methyl-1-oxobutyl)-4,5-dihydropyrano[4,3,2-de]chromen-2-one (15, 0.0015%), 8-hydroxy-5-methyl-7-(3,7-dimethyl-octa-2,6-dienyl)-9-(3-methyl-1-oxobutyl)-4,5-dihydropyrano[4,3,2-de]chromen-2-one (16, 0.0012%), mammeas A/AA (17, 0.0494%), A/AB (18, 0.0048%), A/AC (19, 0.1056%), A/AD (20, 0.0022%), E/BA (21, 0.0045%), E/BB (22, 0.0194%), A/AA cyclo D (24, 0.0035%), A/AB cyclo D (25, 0.0097%), A/AC cyclo D (26, 0.0109%), B/AB cyclo D (27, 0.0016%), B/AC cyclo D (28, 0.0062%), E/BC cyclo D (29, 0.0058%), and deacetylmammea E/BC cyclo D (33, 0.0073%), as described previously (Morikawa et al., 2012; Ninomiya et al., 2016). In the present study, we additionally isolated kayeassamin I (1, 0.0072%) and mammeasins E (2, 0.0099%) and F (3, 0.0015%), from the methanol extract of M. siamensis flowers as shown in Figure 1, together with six coumarins: mammeas E/BC (23, 0.0076%) and E/BD cyclo D (30, 0.0015%), deacetylmammeas E/AA cyclo D (31, 0.0005%) and E/BB cyclo D (32, 0.0023%), and mammeas A/AA cyclo F (34, 0.0010%) and A/AC cyclo F (35, 0.0068%), using normal-phase silica gel and reversed-phase ODS column chromatographic purification steps, and finally by HPLC (Figure 2).

Compound 1 was obtained as pale yellow oil with a negative optical rotation ([α]D25 −50.4 in CHCl₃), and its molecular formula was deduced to be C₂₆H₃₂O₆ by high-resolution ESIMS (HRESIMS) measurement. As shown in Figure 3, the HPLC analysis suggested that 1 was obtained as an inseparable mixture (ca. 1:1 ratio). The ¹H and ¹³C NMR spectra spectroscopic properties (Table 1, CDCl₃) of 1, which were assigned with the aid of DEPT, DQF-COSY, HSQC, and HMBC experiments, were in accordance with those of kayeassamin I except for the observation of duplicate signals (1a and 1b) measured by high resolution 800 MHz NMR spectrometer: two primary, a tertiary, and two vinyl methyls [1a: δ 1.04 (3H, t, J = 7.4 Hz, H₃-4^‴), 1.11 (3H, t, J = 7.4 Hz, H₃-3′), 1.52 (3H, s, 2″-CH₃), 1.55 (3H, s, H₃-10″), 1.64 (3H, d, J = 0.9 Hz, H₃-9″); 1b: δ 1.03 (3H, t, J = 7.4 Hz, H₃-4^‴), 1.09 (3H, t, J = 7.4 Hz, H₃-3′), 1.48 (3H, s, 2″-CH₃), 1.57 (3H, s, H₃-10″), 1.67 (3H, d, J = 0.9 Hz, H₃-9″)], five methylenes [1a: δ 1.51, 1.95 (1H each, both m, H₂-2′), 1.71, 1.91 (1H each, both m, H₂-5″), 1.78 (2H, qt, J = 7.4, 7.1 Hz, H₂-3^‴), 2.09 (2H, m, H₂-6″), 3.26 (2H, t, J = 7.1 Hz, H₂-2^‴); 1b: δ 1.53, 1.96 (1H each, both m, H₂-2′), 1.71, 1.91 (1H each, both m, H₂-5″), 1.78 (2H, qt, J = 7.4, 7.1 Hz, H₂-3^‴), 2.09 (2H, m, H₂-6″), 3.26 (2H, t, J = 7.1 Hz, H₂-2^‴)], a methine bearing an oxygen function [1a: δ 5.43 (1H, br t, J = ca. 8 Hz, H-1′); 1b: δ 5.43 (1H, br t, J = ca. 8 Hz, H-1′)], four olefinic protons [1a: δ 5.06 (1H, qt, J = 0.9, 7.1 Hz, H-7″), 5.53 (1H, d, J = 10.2 Hz, H-3″), 6.60 (1H, br s, H-3), 6.78 (1H, d, J = 10.2 Hz, H-4″); 1b: δ 5.06 (1H, qt, J = 0.9, 7.1 Hz, H-7″), 5.55 (1H, d, J = 10.2 Hz, H-3″), 6.61 (1H, d, J = 0.9 Hz, H-3), 6.79 (1H, d, J = 10.2 Hz, H-4″)], and a hydrogen-bonded hydroxy proton [1a: δ 14.47 (1H, s, 7-OH); 1b: δ 14.47 (1H, s, 7-OH)]. This evidence allowed us to revise the structure of kayeassamin I as a mixture (1a and 1b) of ca. 1:1 inseparable stereoisomers in the 2″ position. The absolute configuration of the 1′-position in 1 has been assumed to be S by comparison of the optical rotation with that of similar compounds (Win et al., 2008b). To confirm the stereochemistry, we carried out chemical correlation between 1 and kayeassamin A (8), which has been reported to be in the 1′S form by the modified Mosher's method (Win et al., 2008a). Thus, oxidation of 8 with 2,3-dichloro-5,6-dicyano-p-benzoquinone (DDQ) gave 1. Consequently, the absolute configuration in the 1′ position of 1 was confirmed to be S.

Mammeasin E (2) was also obtained as an inseparable mixture (ca. 1:1 ratio, Figure S1) with a negative optical rotation ([α]D25 −58.9 in CHCl₃). In the negative-ion ESIMS of 2, a quasimolecular ion peak was observed at m/z 453 [M – H]⁻, and HRESIMS analysis indicated the molecular formula was C₂₇H₃₄O₆. The ¹H and ¹³C NMR spectra (Table 2, CDCl₃) of 2 were similar to those of 1, except for the signals due to the 3-methyl-1-oxobutyl moiety in the 8-position [2a: δ 1.03 (6H, d, J = 6.6 Hz, H₃-4^‴ and H₃-5^‴), 2.27 (1H, m, H-3^‴), 3.14 (2H, d, J = 6.7 Hz, H₂-2^‴); 2b: δ 1.03 (6H, d, J = 6.6 Hz, H₃-4^‴ and H₃-5^‴), 2.27 (1H, m, H-3^‴), 3.14 (2H, d, J = 6.7 Hz, H₂-2^‴)] instead of the 1-oxobutyl moiety of 1. As shown in Figure S2, the connectivity of the quaternary carbons in 2 were elucidated on the basis of DQF-COSY and HMBC experiments. Thus, the DQF-COSY experiment on 2 indicated the presence of the following partial structures: C-1′-C-3′; C-3″-C-4″; C-5″-C-7″; and C-2^‴–C-5^‴ shown in bold lines. In the HMBC experiment, long-range correlations were observed between the following proton and carbon pairs: H-3 [2a: δ 6.61 (1H, d, J = 0.9 Hz); 2b: δ 6.59 (1H, d, J = 1.0 Hz)] and C-2 (2a: δ_C 159.6; 2b: δ_C 159.6), C-4a (2a: δ_C 101.0; 2b: δ_C 101.1); the hydrogen-bonded hydroxy proton [2a: δ 14.51 (1H, s); 2b: δ 14.51 (1H, s)] and C-6 (2a: δ_C 105.8; 2b: δ_C 106.0), C-7 (2a: δ_C 163.0; 2b: δ_C 163.0), C-8 (2a: δ_C 104.5; 2b: δ_C 104.6); H-1′ [2a: δ 5.40 (1H, br t, J = ca. 8 Hz); 2b: δ 5.40 (1H, br t, J = ca. 8 Hz)] and C-3 (2a: δ_C 107.0; 2b: δ_C 107.1), C-4a; H-3^‴ [2a: δ 5.53 (1H, d, J = 10.2 Hz); 2b: δ 5.54 (1H, d, J = 10.2 Hz)] and C-6, C-2″ (2a: δ_C 83.0; 2b: δ_C 83.1), 2″-CH₃ (2a: δ_C 27.3; 2b: δ_C 27.5); H-4″ [2a: δ 6.79 (1H, d, J = 10.2 Hz); 2b: δ 6.78 (1H, d, J = 10.2 Hz)] and C-5 (2a: δ_C 156.0; 2b: δ_C 156.0), C-6; H₂-5″ [2a: δ 1.71, 1.90 (1H each, both m); 2b: δ 1.71, 1.90 (1H each, both m)] and C-2″, 2″-CH₃; H-7″ [2a: δ 5.06 (1H, qt, J = 1.0, 7.1 Hz); 2b: δ 5.06 (1H, qt, J = 1.0, 7.1 Hz)] and C-9″ (2a: δ_C 25.5; 2b: δ_C 25.6), C-10″ (2a: δ_C 17.6; 2b: δ_C 17.7); H-9″ [2a: δ 1.64 (3H, d, J = 1.0 Hz); 2b: δ 1.67 (3H, d, J = 1.0 Hz)] and C-7″ (2a: δ_C 123.1; 2b: δ_C 123.0), C-8″ (2a: δ_C 132.6; 2b: δ_C 132.5), C-10″; H-10″ [2a: δ 1.52 (3H, s); 2b: δ 1.54 (3H, s)] and C-7″-9″; and H₂-2^‴ and C-1^‴ (2a: δ_C 206.2; 2b: δ_C 206.2). On the other hand, the molecular formula of mammeasin F (3) was determined to be the same as that of 2, C₂₇H₃₄O₆, by HRESIMS measurement. The ¹H and ¹³C NMR spectroscopic properties (Table 2, CDCl₃) of 3, which were observed to be duplicate signals caused by its inseparable mixture (ca. 1: 1 ratio, Figure S1), were quite similar to those of 2 except for the signals due to the 2-methyl-1-oxobutyl moiety in the 8-position [3a: δ 0.98 (3H, t, J = 7.5 Hz, H₃-5^‴), 1.25 (3H, d, J = 6.7 Hz, H₃-3^‴), 1.46, 1.89 (each 1H, both m, H₂-4^‴), 3.89 (1H, m, H₂-2^‴); 3b: δ 0.98 (3H, t, J = 7.5 Hz, H₃-5^‴), 1.26 (3H, d, J = 6.7 Hz, H₃-3^‴), 1.46, 1.89 (each 1H, both m, H₂-4^‴), 3.89 (1H, m, H₂-2^‴)]. Finally, 2 and 3 were derived by DDQ oxidation of surangins D (14) (Ngo et al., 2010) and C (13) (Verotta et al., 2004; Yagi et al., 2006), respectively. Based on this evidence, the stereostructures of 2 and 3 were determined to be as shown.

To characterize the active constituents of this plant material, the inhibitory effects of 30 isolates (1–13, 17–20, 22–29, 31–35) against 5α-reductase were examined. As shown in Table 3, mammeasins E (2, 22.6 μM), A (4, 19.0 μM), and B (5, 24.0 μM), kayeassamins E (9, 33.8 μM), F (10, 15.9 μM), and G (11, 17.7 μM), surangin C (13, 5.9 μM), and mammeas A/AA (17, 19.5 μM), E/BB (22, 16.8 μM), and A/AA cyclo F (34, 23.6 μM), were found to inhibit testosterone 5α-reductase (Table S1).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.