Adlay seeds (Coix lacryma-jobi var. ma-yuen (Rom. Caill.) Stapf) were purchased in 2016 from the Daya District Farmers’ Association, which is directed by Taichung District Agricultural Research and Extension Station, Council of Agriculture, Taiwan. The raw plant material (batch number: 3A0015) was identified by Dr. Ming-Hong Yen. A voucher specimen (code no. KMU-Coix 001) was stored in the Graduate Institute of Natural Products, College of Pharmacy, Kaohsiung Medical University and Joben Bio-Medical Co., Ltd., Taiwan. The exact name of plant material has been checked on the authoritative website in Taxonomy: http://www.worldfloraonline.org. Coix lacryma-jobi var. ma-yuen (Rom. Caill.) Stapf was indicated as an accepted name in genus Coix (family Poaceae). The seeds were dried at room temperature by an air circulator and dehulled with a grinding mill. The grinded particles of adlay were then separated with an industrial-designed AB collector to obtain the AB. It was further grinded into powder and sieved through 20-mesh (aperture = 0.84 mm) for supercritical fluid extraction. The recovery of AB from whole adlay seeds (including hull, testa, bran and endosperm) was 9.8–10.8% (Figure 1).

Hepatic TG and TC were measured by ELISA kits, No. 10010303 and No. 10007640 purchased from Cayman Chemical Company (Ann Arbor, MI, United States), respectively. Fecal bile acid was analyzed with the Bile Acids kit (Product No. 450) obtained from Trinity Biotech Plc. (Wicklow, Leinster, Ireland). The anti-AMPK, anti-p-AMPK, anti-rabbit IgG, and mouse IgG bodies were purchased from Cell Signaling Technology Inc. (Danvers, MA, United States). Anti-FAS, anti-LPL, and anti-β-actin bodies were obtained from Abcam Inc. (Cambridge, MA, United States). Standard laboratory chow diet (No. 5001) was purchased from PMI^® Nutrition International (Brentwood, MO, United States). Standard mixtures for qualitative and quantitative analysis, F.A.M.E (Fatty acid methyl esters) Mix RM-4 (methyl linoleate, methyl oleate, methyl palmitate, methyl stearate), were purchased from Supelco^® (Sigma-Aldrich, St. Louis, MO, United States). Experimental animals in this work were fed either the standard chow diet or the HFD adapted from previous study (Yu et al., 2011). The nutrition facts of the standard chow diet were 3.36 kcal/g, containing 58.0% carbohydrates, 28.5% proteins and 13.5% fats. The HFD was 3.93 kcal/g, containing 44.53% carbohydrates, 21.88% proteins and 33.59% fats in one portion of 89.8% (wt/wt). 10% (wt/wt) of lard and 0.2% (wt/wt) of cholesterol were the rest of ingredients added, respectively (Sigma-Aldrich, St. Louis, MO, United States).

AB-SCF was extracted by an industrial-scaled and regularly-validated supercritical fluid system (NATEX Process Technology GmbH, Ternitz, Austria) settled in the Joben Bio-Medical Co., Ltd. (Pingtung, Taiwan). 2–3 kg AB for each batch was weighed accurately and supplied into the instrumental vessel to process the extraction. Detail parameters for the SCF solvent, SC-CO₂, were optimized as 30–35 MPa and 40–60°C at a flow rate of 30–35 kg CO₂ per hour, along with 60–75 min period for the balance between the recovery and production capacity. Every extraction was terminated depending on whether the yield was less than 0.1%. The yields of AB-SCF from AB were averagely 14–18%. The recovery of AB-SCF from whole adlay seeds was 1.37–3.24% (Figure 1).

Lipid metabolism of hamsters have been reported that closely resemble to human beings. Thus, hamsters were usually considered as the first and appropriate animal model for estimating hypolipidemic effects (Suica et al., 2016). Male Golden Syrian hamsters (6 weeks old) were purchased from the National Laboratory Animal Center (NLAC), Taipei City, Taiwan. All of them were housed under standard temperature (25 ± 1°C) and 50–60% relative humidity of conditions with a 12 h/12 h light-dark cycle. Standard chow diet and distilled water were provided ad libitum. Before the initiation of experiments, the hamsters were accommodated for 1 week to be stabilized and familiarize to the environment. All animal experimental protocols were supervised by the institutional animal care and use committee (IACUC) of Chia-Nan university (Tainan, Taiwan). The study conformed to the guidelines of the protocol CN-IACUC-105008R approved by the IACUC ethics committee.

The human equivalent dose (HED) of AB-SCF for hamsters were converted from the recommended daily dose for an adult human (assumed as 60 kg) which is 500 mg per day (one capsule/serving/day). HED between human and hamsters were calculated with a conversion coefficient, i.e., 7.4, based on body surface area (issued by the US Food and Drug Administration: http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ucm078932.pdf). The formula would be as follows:

HED of hamsters = (Recommended daily dose of human/kg) × 7.4 (coefficient) = [500 (mg)/60 (kg)] × 7.4 = 61.67 mg/kg (designed as 1×)

After 1-week acclimatization, 70 hamsters were randomly divided into seven groups (n = 10/each group): C, a blank control group fed with standard chow diet and water; HFD, an HFD induced hyperlipidemic group fed with water; EM, a reference group for the vehicle (solubilizer) of AB-SCF fed with HFD and emulsifier (prepared with Tween 80/Span 80 in the ratio of 5:1 v/v, then diluted with sterilized RO water to 5%). Experimental groups were induced with HFD and administrated AB-SCF/emulsifier with a dose-ascending manner, i.e., AB-SCF×0.5, 30.84 mg/kg/BW (body weight); AB-SCF×1, 61.67 mg/kg/BW; AB-SCF×5, 308.35 mg/kg/BW; AB-SCF×10, 616.70 mg/kg/BW. The food intakes and water consumptions were daily monitored. The body weights were recorded weekly. All hamsters were fast for 16 h and sacrificed with 95% CO₂ asphyxiation after the complete experimental period of 8 weeks. Serum, hepatic and fecal biochemical data, along with the energy metabolic mechanism(s) from proteins of liver tissues were further investigated.

Biochemical data related to serum lipid and lipoprotein levels, such as triglyceride (TG), total cholesterol (TC), low density lipoprotein cholesterol (LDL-C), high density lipoprotein cholesterol (HDL-C), LDL-C/HDL-C ratio (a predictor of cardiovascular risk, Jukema et al., 2005). Blood samples were collected with cardiac puncture and immediately centrifugated at 1500 × g (4°C) for 15 min in anticoagulant-treated tubes (Greiner Bio-One GmbH, Frickenhausen, Germany) to obtain the serum samples. Fecal samples were gathered within 2 days (48 h) before the termination. The feces were dried, powdered, weighted and extracted with Folch Solution (chloroform/methanol 2:1, vol/vol). An aliquot of the organic phase was then dried and resuspended in isopropyl alcohol (Mera et al., 2015). Lipid and lipoprotein profiles including TG, TC, HDL-C, LDL-C were measured by the use of an automated clinical chemistry analyzer, Fuji Dry-Chem 4000i, and its dedicated biochemical slides (Fujifilm, Tokyo, Japan).

Fecal lipid and bile acid were assessed. The feces were dried, weight and then grounded as fine powder in a mechanical blender. Aliquots of ground feces were well-mixed with sodium borohydride and then subjected to strong alkaline hydrolysis at 120–130°C for 12 h. The extracted fecal bile acid was measured enzymatically measured with a commercial bile acids kit (Product No. 450-A from Trinity Biotech Plc., Wicklow, Leinster, Ireland).

Hepatic lipid profiles (TC and TG) and oxidative stress markers i.e., MDA (malondialdehyde) and GSH (glutathione) were evaluated. All liver tissues were carefully collected, washed three times in ice cold saline, blotted individually on ash-free filter paper, weighted and aliquoted into few parts and frozen stored at −80°C. Before further analysis of hepatic TG and TC, each piece was homogenized with Tris-buffer (Sigma-Aldrich, St. Louis, MO, United States). The centrifuged supernatants were extracted by chloroform-isopropanol-NP40 (7:11:0.1, v/v) with a bullet blender (Lee et al., 2015). After the centrifugation again at 12,000 × g (4°C) for 10 min, hepatic TG and TC levels of the supernatants were measured in triplicate by using commercial enzymatic kits for TG (No. 10010303) and for TC (No. 10007640) from Cayman Chemical Company (Ann Arbor, MI, United States).

For the estimations of GSH and MDA, liver tissues were homogenized in phosphate buffer saline (PBS) 50 mM pH (7.4) and potassium phosphate buffer 10 mM pH (7.4), respectively. GSH levels were carried out with a commercial glutathione assay kit (product CS0260 from Sigma-Aldrich, St. Louis, MO, United States). This principle of MDA assessment depends on its formation as an end product of lipid peroxidation which reacts with thiobarbituric acid to produce thiobarbituric acid reactive substance (TBARS). TBARS, a pink chromogen, would be detected at 532 nm in a Spectrophotometry. A TEP (1,1,3,3-tetraethoxypropane) standard (Sigma-Aldrich, St. Louis, MO, United States) was used to build a standard curve against which readings of the samples were plotted (Noeman et al., 2011).

Bio-markers related to energy-balance, lipoprotein metabolism and oxidative-stress in liver, e.g., AMPK (adenosine-monophosphate-activated protein kinase), p-AMPK (phosphorylated-AMPK), FAS (fatty acid synthase) and LPL (hepatic lipoprotein lipase) were investigated to elaborate the energy metabolic mechanism. Protein extraction was conducted by homogenizing each liver sample in 1 ml lysis buffer (containing 10 mM-HEPES, pH 7.8), 10 mM KCl, 2 mM MgCl₂, 1 mM dithiothreitol (DTT), 0.1 mM EDTA, and 0.1 mM phenylmethylsulfonyl fluoride) at 4°C. Meanwhile, 80 μl of 10% NP-40 solution used for breaking the nuclear membrane within a cell was added as well. After the homogenization, the lysates were centrifuged for 2 min at 14,000 × g. Equal amounts of lysed protein (30 µg/lane) were loaded onto SDS-polyacrylamide gels, and electrophoretically transferred to a PVDF membrane (Bio-Rad Laboratories, Hercules, CA, United States). After blocking with 5% (w/v) skim milk in 0.1% (v/v) Tween 20-containing PBS (PBST) for 1 h at room temperature, the membrane was incubated with the following specific primary antibodies for 1 h at room temperature: anti-AMPK (1:1000), anti-phospho-AMPK (1:1000), anti-FAS (1:1000), anti-LPL (1:1000), and anti-β-actin (1:25000) antibodies (in 5% w/v skim milk with PBST). Antibody recognition was detected with the respective secondary antibody, either anti-mouse IgG or anti-rabbit IgG antibodies linked to horseradish peroxidase. Antibody-bound proteins were detected with the ECL western blotting analysis system (Amersham, Aylesbury, UK). The expression of β-actin was used as loading control. Relative protein expressions were quantified densitometrically with an AlphaImager 2200 (Alpha Innotech Corp., San Leandro, CA, United States), and processed using AlphaEaseFC software in referring to the β-actin reference bands in triplicate (Wei et al., 2017).

An aqueous concentrated HCl (conc. HCl; 35%, w/w) catalyzation was conducted to prepare the fatty acid methyl esters (FAMEs) of AB-SCF for GC analysis (Ichihara and Fukubayashi, 2010). Briefly, an 8% (w/v) HCl prepared in methanol/water (85:15, v/v) was diluted in 9.7 ml of concentrated HCl with 41.5 ml of methanol. Toluene (0.2 ml), methanol (1.5 ml), and the 8% HCl solution (0.3 ml) were added sequentially to the AB-SCF. The final HCl concentration was 1.2% (w/v). This solution (2 ml) was incubated at 45°C overnight or heated at 100°C for 1.5 h. The catalyzed AB-SCF was dried, then dissolved in EtOAc, filtered through 0.22 µm filter and subjected to GC analysis.

The analysis of major compositions and the establishment of AB-SCF fingerprint were carried out with a gas chromatography system (Trace GC Ulture/ITQ 900, Thermo fisher Scientific, United States) with a flame ionization detector (FID). The capillary column was RT^®-2560 (100 m × 250 μm  ×  0.2 μm) coated with biscyanopropyl polisiloxane as stationary phase (Restek Corporation, Bellefonte, PA, United States). The column oven temperature was programmed 150°C (held for 2 min), increased to 220°C at a rate of 35°C/min (held for 1 min), then raised to 225°C a rate of 0.5°C/min (maintained for 1 min). The other parameters were as follows: injection temperature, 225°C; detector temperature, 250°C; carrier gas, Helium at 1 ml/min; injection volume, 1 μl. The relative percentage of each major component in AB-SCF was quantified based on the peak area integrated by Thermo Xcalibur™ data analysis program (Thermo fisher Scientific, United States). Qualitative and quantitative analysis of AB-SCF (C16:0 Palmitate, C18:0 Stearate, C18:1 Oleate, C18:2 Linoleate) was carried out in comparing with the F.A.M.E Mix RM-4 standards.

10.0614g (density = 0.922 g/ml) of AB-SCF was firstly subjected into a Sephadex^® LH-20 column in an environment of Dichloromethane (DCM):Methanol = 1:1 for gel filtration (Fine Chemicals AB, Uppsala, Pharmacia). Sephadex^® LH-20 is composed by cross-linked dextran for molecular sizing natural products in accordance to molecular weight. The purpose would be to divide the nonlipid substances from apolar/lipid mixtures (Wells and Dittmer, 1963). Seven fractions, AB-SCF-S1 to S8, were yielded after passing AB-SCF through the Sephadex^® LH-20 column. Thin-layer chromatography was monitored with silica gel 60 F₂₅₄ and RP-18 F_254S TLC plates (Merck, Darmstadt, Germany) with substance visualized by 10% (v/v) H₂SO₄/ethanol spray (Supplementary Figure S1). After brief screen of ¹H NMR (Nuclear Magnetic Resonance) spectroscopy on every AB-SCF-Sn fraction, AB-SCF-S1 was the main ingredient composed of 78.92% triglyceride, along with AB-SCF-S4 and S5 were detected 8.39% for mostly fatty acids. AB-SCF-S2 (192.50 mg, 1.91%) and S3 (282.27 mg, 2.86%) were observed with interesting minor signals different but still mixed with triglyceride and fatty acids. Meanwhile, the last three fractions, AB-SCF-S6 to S8, occupied only 0.58% (58.36 mg) in AB-SCF with signals of oil and complicate mixtures (Supplementary Figure S2).

AB-SCF-S3 was combined into AB-SCF-S2 as S2’ (totally 479.77 mg, 4.77% weight/total weight of eluents) and fractioned with silica gel (230–400 mesh, Merck, Darmstadt, Germany) open column and stepwise eluted with n-hexane:EtOAc from 80:1–100% into 17 subfractions (S2′-1–17). With the continuous monitor of ¹H NMR, S2′-7 which was different from oil constituents were separated with another silica gel open column from the solvent system of n-hexane:EtOAc from 80:1–100%. S2′-7-1 to 5 were obtained. S2′-7-5-1 to 9 were further got carried out by silica preparative TLC n-Hexane:DCM (1:1). However, the amounts of these fractions were too low (all less than 2 mg) to be re-measured by ¹H NMR. The other fraction considered not being fatty acids from S2′, the S2′-11, was directly processed by silica preparative TLC n-hexane: DCM (1:2) to yield two mixed geometric isomers, (1) (Rf = 0.58) and (2) (Rf = 0.58). (3) was filtered out from S2′-11-6 by using DCM and Methanol (Rf = 0.38) (Supplementary Figures S3, S4).

Data were presented as mean ± standard deviation (SD) from different and independent experiments. Values were evaluated by one-way ANOVA, followed by Duncan’s multiple range test using the 9.0 Statistical Analysis System (SAS Institute, Cary, NC, United States). Difference was considered significant when p-value was <0.05.

The effects of AB-SCF×0.5, AB-SCF×1, AB-SCF×5, AB-SCF×10 on the changes of body weight (BW) and daily food intake (DFI) were recorded in the model of high-fat diet induced hyperlipidemic male hamsters (Table 1). After an 8-week administration, the external appearances and health conditions of all hamsters were remained ordinarily with no adverse effects observed. BWs were stable and steadily increased in each group. High fat diet induced groups were generally heavier than C with significance (p < 0.05). Meanwhile, AB-SCF×10 displayed significantly lower BW than EM, AB-SCF×5, AB-SCF×1 and AB-SCF×0.5 (p < 0.05), and close to C at the 6th to 8th week. The daily food intake (DFI) data showed no difference among all groups.

After the 8 weeks administrations, effects of AB-SCF×0.5, AB-SCF×1, AB-SCF×5, and AB-SCF×10 groups on the changes of serum and fecal lipid and lipoprotein profiles were examined after the termination (Table 2). TG and TC levels of the C group were 47.30 ± 9.59 and 53.60 ± 3.44 mg/dl, respectively. However, HFD and EM had been increased significantly to 328.50 ± 30.15 and 360.90 ± 30.22 mg/dl (p < 0.05). All AB-SCF administrated groups, compared to the HFD and EM, exhibited markedly lower data in not only TG and TC, but also in LDL-C (p < 0.05). In addition, HDL-C levels were significantly higher (p < 0.05) in all AB-SCF groups than in HFD and EM. The ratios of LDL-C/HDL-C, a predictor of cardiovascular risk, were dose-dependently and meaningfully decreased (p < 0.05) in AB-SCF administrated hyperlipidemic hamsters as 2.60 ± 0.66, 1.69 ± 0.50, 0.49 ± 0.10, 0.48 ± 0.26, respectively (with 0.11 ± 0.04 in C, 4.73 ± 1.04 in HFD, and 4.30 ± 0.90 in EM).

In feces, the lipid and lipoprotein profiles were generally higher in high fat diet induced hyperlipidemic hamsters. Only TC were expressively alleviated (p < 0.05) by AB-SCF×1, AB-SCF×5, AB-SCF×10 with 27.00 ± 1.41 mg, 26.00 ± 2.83 mg, 26.5 ± 0.71 mg in every 100 mg feces (11.00 ± 0.82 from C, 35.00 ± 4.08 from HFD, 34.00 ± 0.83 from EM and 39.00 ± 1.41 from AB-SCF×0.5). Bile acids (μmol/100 mg feces) in HFD (116.08 ± 9.73) and EM (113.28 ± 3.10) was obviously elevated (p < 0.05) than in C (17.37 ± 1.46). AB-SCF groups would improve the excretions of bile acids with significances in dose-dependent manner.

Hepatic TG and TC were both aggravatedly higher in HFD and EM groups than the other ones (Table 3 and Figure 2). In contrast, AB-SCF had significantly ameliorated the levels of hepatic TG and TC (p < 0.05). Especially in the AB-SCF×10 group, hepatic TG and TC of hyperlipidemic hamsters were almost retrieved back to levels of the standard chow diet ones.

As compared to the C in the measurement of MDA and GSH (179.10 ± 11.30; 15.90 ± 0.60), HFD (362.10 ± 16.4; 13.70 ± 0.40) and EM (359.50 ± 8.30; 13.60 ± 0.30) showed significant increases and depletions, respectively. In lipid peroxidation of hyperlipidemic hamster livers, AB-SCF groups exhibited promising protective affections with dose-dependent manner (MDA: AB-SCF×0.5, 215.70 ± 7.80; AB-SCF×1, 203.20 ± 9.40; AB-SCF×5, 190.50 ± 9.40; AB-SCF×10, 177.80 ± 5.80. p < 0.05). Meanwhile, the GSH results were positively regulated in the relatively higher doses of AB-SCF×5 and AB-SCF×10 which were indicated as 14.70 ± 0.30 and 14.90 ± 0.50, respectively. Thus, AB-SCF might be able to be considered as an alleviating agent against the accumulation and lipid peroxidation in liver.

To reveal the molecular-biochemical mechanism of energy metabolism exerted by AB-SCF, actions targeted on hepatic proteins from hyperlipidemic and normal hamsters, such as fatty acid synthase (FAS), lipoprotein lipase (LPL), AMPK and p-AMPK, were investigated. As results presented in the Figure 3 and Supplementary Figure S5, the expression of FAS was elevated in HFD, along with AMPK, p-AMPK were down-regulated. It is noteworthy that, just in opposite, the expression of hepatic FAS was decreased in the AB-SCF administrated hamsters with a dose-depend manner. Meanwhile, hepatic LPL, AMPK and p-AMPK had improved significantly in the both of AB-SCF×5 and AB-SCF×10 groups. These evidences suggested that the consumption of AB-SCF may exhibit the capacities in mitigating hyperlipidemia through the modulation of hepatic fat metabolism.

In Table 4, a nutrition fact list was offered. The gross energy of every 100 g AB-SCF is 793.57 ± 1.88 Kcal. Lipids (81.13 ± 0.25%) was the most abundant ingredient with no trans fats was found comprising. Proteins constituted 15.85 ± 0.15%, meanwhile, no carbohydrates, including sugar, were found in AB-SCF.

Supercritical fluid extraction with the solvent, SC-CO₂, was characterized by its apolar property, and known as an effective approach to obtain the oil/apolar substances. The analysis focused on the lipid compositions of AB-SCF with anti-dyslipidemic/hypercholesterolemic capacities is a considerable issue to be investigated as well. As the results presented in the Table 5, four major fatty acids, palmitic acid (C16:0; 13.07%), linoleic acid (C18:2; 28.59%), oleic acid (C18:1; 56.95%) and stearic acid (C18:0; 1.39%) were identified. Unsaturated fats majorly occupied up to 85.18% (linoleic acid + oleic acid) along with 14.16% saturated fats (palmitic acid + stearic acid) were detected. Fingerprint of AB-SCF lipids was established as shown in the Figure 4. Optimized and detailed analytic parameters for quality control using gas chromatography system were established in this work (section 2.8 and 2.9).

The complete flow chart for separation and isolated pure compounds from AB-SCF were presented in the Figure 5. Detailed chromatographic works were recorded in Supplementary Figures S1–S4 as described in the section 2.9. Three compounds, 3-O-(trans-4-feruloyl)-β-sitostanol (1) (Condo et al., 2001), 3-O-(cis-4-feruloyl)-β-sitostanol (2) (Akihisa et al., 2000) and β-sitosterol (3) (Chaturvedula and Prakash., 2012), were the obtained with the yields of 14.724 mg (0.1463%, 1463.41 ppm), 1.636 mg (0.0016%, 162.60 ppm) and 25.07 mg (0.2491%, 2491.70 ppm), respectively. Each compound was identified by MS, ¹H NMR, ¹³C NMR and comparison with authentic samples or with published data.

3-O-(trans-4 feruloyl)-β-sitostanol (1): EISMS m/z 591.71 [M-H]⁻. ¹H NMR (CDCl₃): δ _H 0.6–2.0 (β-sitostanol moiety, ∼50H), 3.91 (s, 3H), 4.82 (tt, J = 11.0, 4.8 Hz, 1H), 5.88 (s, 1H), 6.27 (d, J = 15.9 Hz, 1H), 6.91 (d, J = 8.1 Hz, 1H), 7.03 (d, J = 1.7, 1H), 7.06 (dd, 8.2, 1.8, 1H), 7.59 (d, J = 15.9 Hz, 1H). ¹³C NMR: 11.96, 12.05, 12.24, 18.71, 19.01, 19.80, 21.29, 23.04, 24.20, 26.05, 27.62, 28.25, 28.62, 29.12, 31.99, 33.90, 34.16, 35.48, 35.48, 36.16, 36.78, 39.96, 42.57, 44.67, 45.81, 54.22, 55.89, 56.15, 56.40, 73.71, 109.22, 114.66, 116.18, 122.98, 127.11, 144.36, 146.71, 147.79, 166.81 (Supplementary Figure S6).

The ¹H NMR (CDCl₃) data of 3-O-(cis-4-feruloyl)-β-sitostanol (2), a rotated geometric isomer minorly mixed in (1) were listed as follows (Supplementary Figure S7): δ 7.11 (d, J = 8.2 Hz, 3H), 6.87 (d, J = 8.2 Hz, 2H), 6.76 (d, J = 13.0 Hz, 2H), 6.47 (d, J = 15.9 Hz, 1H), 6.06 (d, J = 15.9 Hz, 0H), 5.80 (d, J = 4.0 Hz, 1H), 5.77 (d, J = 4.2 Hz, 1H), 5.39 (s, 1H), 4.37 (q, J = 7.2 Hz, 1H), 3.65 (s, 3H). (1) and (2) would interconvert to be each other, and steadily exist with a ratio of 9:1 observed according to the integral quantities of ¹H signals.

β-sitosterol (3): EIMS m/z 414.72 [M⁺]. ¹H NMR (CDCl₃): H-3 (δ _H 3.55, dtt, J = 27.0, 10.9, 4.6 Hz, 1H), H-5 (δ _H 5.34, m, 1H), H-19 (δ _H 0.91, d, J = 6.6 Hz, 3H), H-24 (δ _H 0.85, d, J = 7.6 Hz, 3H), H-26 (δ _H 0.82, s, 3H), H-27 (δ _H 0.80, s, 3H), H-28 (δ _H 0.67, s, 3H), H-29 (δ _H 1.00, s, 3H). ¹³C NMR: 11.99, 12.21, 18.91, 19.16, 19.53, 19.95, 21.21, 23.19, 24.43, 28.38, 29.28, 31.78, 32.04, 36.64, 37.39, 39.91, 40.63, 42.42, 42.45, 45.96, 50.26, 56.19, 56.90, 71.92, 77.00, 121.84, 129.41, 138.46, 140.91 (Supplementary Figure S8).

¹H NMR and ¹³C NMR spectra were performed on JEOL JNM-ECS 400 MHz NMR Spectrometer (¹H, F400 MHz; ¹³C, 100 MHz) and Varian Mercury Plus 400 MHz FT-NMR (¹H, 400 MHz; ¹³C, 100 MHz) in CDCl₃. Mass spectra were obtained from Waters 2695 Separations Module (ESI-MS).