Dioscin (>98%) was prepared from Dioscorea nipponica Makino in our laboratory (Yin et al., 2010), which was dissolved with 0.1% dimethylsulfoxide (DMSO) for in vitro experiments, or with 0.5% carboxymethylcellulose sodium (CMC-Na) solution for in vivo tests. CMC-Na, Tris, Sodium dodecyl sulfate (SDS) and 4′,6′-diamidino-2-phenylindole (DAPI) were purchased from Sigma (St. Louis, MO, USA). ALT, AST kits, ALP, TBIL, GSH, GSH-Px, MDA, and SOD were obtained from the Nanjing Jiancheng Institute of Biotechnology (Nanjing, China). A tissue protein extraction kit was obtained from Keygen Biotech. Co., Ltd. (Nanjing, China). In Situ Cell Death Detection Kit (TMR Red; Roche, NJ, USA). RNAiso Plus, a PrimeScript^® RT Reagent Kit with gDNA Eraser (Perfect Real Time) and SYBR^® Premix Ex Taq^TM II (Tli RNase H Plus) were purchased from TaKaRa Biotechnology Co., Ltd. (Dalian, China). A bicinchoninic acid (BCA) protein assay kit was purchased from the Beyotime Institute of Biotechnology (Jiangsu, China).

Sandwich-cultured hepatocytes (SCHs) were isolated from male Wistar rats (200 ± 20 g) by two-step perfusion and purified by 45% isotonic Percoll (Kotani et al., 2011). As shown in Supplementary Figure 2, the hepatocytes were seeded onto collagen -coated plates. Twenty-four hours after seeding, the hepatocytes were overlaid with 0.25 mg/mL Matrigel Basement Membrane Matrix (BD Biosciences, San Jose, CA, USA) to form sandwich configuration and cultured as described previously (Chandra et al., 2005; Yan et al., 2011). After 24 h, the SCHs were exposed to 100 μM of ANIT (dissolved in DMSO, 0.1% final concentration).

The primary cultured hepatocytes were plated into 96-well plates at a density of 5 × 10⁴ cells/mL (100 μL) per well overnight in an incubator and incubated until they reached to approximately 70% confluence. Next, the cells were pretreated with various concentrations of dioscin (50, 100, 200, 400, 800, 1600, and 3200 ng/mL) for 24 h at 37°C, and then the toxicity of the compound was assayed using the MTT method.

The primary cultured hepatocytes were seeded in 96-well plates at a density of 5 × 10⁴ cells/mL (100 μL) for 24 h before treatment. According to the cell proliferation assay kit to add Brdu. After 6 h, the cells were then treated with dioscin at the concentrations of 200, 400, and 800 ng/mL for 6, 12, and 24 h at 37°C, and the cells proliferation were measured using the Brdu method.

The SCHs were plated in 6-well plates at a density of 5 × 10⁴ cells/mL (1 mL) and treated with dioscin at the concentrations of 200, 400, and 800 ng/mL for 24 h, then exposed to 100 μM of ANIT for 24 h. The cells were harvested and re-suspended in 1 mL dichlorodihydrofluorescein diacetate (DCFH-DA) (10 μM) for the detection of ROS level, which was imaged by fluorescence microscope (Olympus, Tokyo, Japan).

Terminal deoxynucleotidyl transferase mediated dUTP nick end labeling (TUNEL) staining and in situ cell apoptosis detection kit (TMR red; Roche, NJ, USA) according to manufacturer’s instructions. Images are taken with a fluorescence microscope (OLYMPUS, Tokyo, Japan).

Male Wistar rats (200 ± 20 g) and male C57BL/6J mice (20 ± 2 g) were purchased from the Experimental Animal Centre of Dalian Medical University, Dalian, China (quality certificate number: SCXK (Liao) 2013-0003). After 1 week of acclimatization, the animals were randomly divided into five groups (n = 10): control group, model (ANIT) group, and high, middle and lose-dose of dioscin-treated groups. The animals in dioscin-treated groups were administered with dioscin at the doses of 60, 30, and 15 mg/kg for rats, and 80, 40, and 20 mg/kg for mice for 7 consecutive days. Liver injury was induced by intraperitoneal (i.p.) ANIT (60 mg/kg for rats and 80 mg/kg for mice) 2 h before the last administration. After 7 days, the animals were sacrificed after an overnight fast. Then, the blood and liver tissue were collected and stored for further analysis. In tests, all animals were housed in a controlled environment at 23 ± 2°C for a period of a 12-h dark/light free access to food and water. All the experimental procedures were approved by the animal protection and use Committee of Dalian Medical University and strictly in accordance with the laws of People’s Republic of China Legislation Regarding the Use and Care of Laboratory Animals.

The serum levels of ALT, AST ALP and TBIL were determined using the commercial kits according to the manufacturer’s instructions.

The levels of GSH, GSH-Px, MDA and SOD in liver tissues were measured according to the manufacturer’s instructions.

The liver tissues were fixed in 10% formalin and embedded in paraffin. Five-micron -thick sections were stained with haematoxylin-eosin (H&E). Images were acquired by light microscopy (Nikon Eclipse TE2000-U, Nikon, Japan), and the degree of liver injury was quantified using Image-Pro Plus 6.0 software. Immunofluorescence staining of tissue slices or formal in-fixed cells for Ntcp and p-PI3K were preformed using primary antibodies, respectively (Santa Cruz, CA, USA) in a humidified chamber at 4°C overnight. After washing twice in PBS, the cells and liver tissue sections were incubated with a fluorescein-labeled secondary antibody for 1 h. Eventually, the cell nuclei were stained with DAPI (5 μg/mL). All samples were imaged using a fluorescence microscope (Olympus, Tokyo, Japan).

Total RNA samples from the cells and livers were extracted using RNAiso Plus reagent following the manufacturer’s protocol. Reverse transcription for cDNA synthesis and quantitative real-time PCR were performed as previously described. The forward (F) and reverse (R) primers for the tested genes are listed in Supplementary Table 1. For each sample, the Ct values for the target gene and GAPDH (as a calibrator) were determined based on standard curves. The calculated relative Ct value of each gene was divided by the relative value of GAPDH. Then, the expression level of each gene in the control group was set to one-fold and used to determine the relative levels in the other samples (n-fold).

The protein samples from the cells and liver tissues were extracted following standard protocols (Beyotime Biotechnology, Haimen, China), and the protein content was determined using a BCA Protein Assay Kit. Proteins were subjected to SDS-PAGE and then transferred to a nitrocellulose membrane. After blocking, the membranes were incubated for 1 h at room temperature or overnight at 4°C. The primary antibodies are listed in Supplementary Table 2. The blots were then incubated with horseradish peroxidise-conjugated antibodies for 2 h at room temperature at a 1: 2000 dilution (Beyotime Institute of Biotechnology, China). Protein expression was detected by the enhanced chemiluminescence (ECL) method and imaged with a Bio-Spectrum Gel Imaging System (UVP, USA). To eliminate variations due to protein quantity and quality, the data were adjusted to GAPDH expression (IOD of objective protein versus IOD of GAPDH protein).

Sandwich-cultured hepatocytes were plated into 6-well plates (5 × 10⁴ cells/mL), and then exposed to PI3K inhibitor wortmannin and Akt inhibitor perifosine (Beyotime, Jiangsu, China) for 2 h. After incubation, the cells were pre-treated with dioscin (800 ng/mL) and then treated with ANIT (100 μM) for 24 h. At last, the protein levels of p-PI3K, p-Akt, Nrf2, GCLm, GCLc, NQO1, HO-1, and the mRNA levels of Nrf2, GCLm, GCLc, NQO1, HO-1 were determined.

Data were evaluated as the mean ± standard deviation (mean ± SD). Statistical analysis of the quantitative data for multiple group comparisons was performed using one-way analysis of variance (ANOVA) followed by Duncan’s test, whereas paired comparisons were performed using the t-test with SPSS software (ver. 20.0; SPSS, Chicago, IL, USA). The results were considered to be significant at ^∗p < 0.05 or ^∗∗p < 0.01.

As shown in Supplementary Figures 3A,B, dioscin (50, 100, 200, 400, and 800 ng/mL) for the primary cultured hepatocytes under 24 h treatment had no effect on cell viability. Dioscin (200, 400, and 800 ng/mL) for 6, 12, and 24 h significantly changed cell proliferation compared with control group. Therefore, dioscin at the concentrations of 200, 400 and 800 ng/mL under 24 h was selected to protect ANIT-induced cholestasis.

As shown in Figure 1A, the livers in control animals showed normal architecture, and the damages including cell necrosis and inflammatory cell infiltration were observed in ANIT-treated animals, which were significantly rehabilitated by the compound. Furthermore, dioscin markedly restored the relative liver weight, serum AST, ALT, ALP and TBIL levels (Figure 1B).

As shown in Figure 2A, compared with control group, dioscin markedly decreased ROS level in SCHs.

As shown in Figure 2B, dioscin significantly reversed the levels of GSH, GSH-Px, SOD and MDA compared with ANIT model groups.

As shown in Figures 3A–C, doscin clearly increased the expression levels of Ntcp in vivo and in vitro based on immunofluorescence assay. The protein levels of Ntcp, OCT1, Bsep and Mrp2 were markedly down-regulated, and the levels of OAT1 were obviously increased in model groups compared with normal groups, which were significantly restored by dioscin (Figures 3D–F and Supplementary Figures 4A–C). Together, these data indicated that dioscin improved cholestasis via regulating the transporters.

As shown in Figures 4A–C, more positive cells with green fluorescence were found in model groups than those of in dioscin-treated groups in SCHs, rats and mice.

As shown in Figures 4D–F and Supplementary Figures 5A–C, dioscin significantly regulated the protein levels of Bak, Bcl-xl, Bax, Bcl-2, Casepase 3 and Caspase 9 compared with ANIT model groups.

Doscin obviously down-regulated p-PI3K levels based on immunofluorescence assay in vitro and in vivo (Figures 5A–C). Compared with ANIT groups, the protein levels of p-PI3K and p-Akt were significantly down- regulated by dioscin, while the levels of Nrf2, GCLm, GCLc, NQO1 and HO-1 were markedly up- regulated by dioscin (Figures 5D–F and Supplementary Figures 6A–C).

As shown in Figure 6A, treatment of SCHs with wortmannin for 2 h significantly inhibited p-PI3K expression, whereas treatment with perifosine for 2 h showed no significant effect on p-PI3K level. In addition, the protein levels of pPI3K, p-Akt, Nrf2, GCLm, GCLc, NQO1 and HO-1 (Figure 6B and Supplementary Figure 7), and the mRNA levels of Nrf2, GCLm, GCLc, NQO1 and HO-1 (Figure 6C) were partially regulated by wortmannin. No obvious changes in p-PI3K levels were found in SCHs by perifosine. These data indicated that dioscin inhibited OS via PI3K/Akt pathway.