2,3,5,4′-tetrahydroxy-trans-stilbene-2-O-β-glucoside (trans-SG) and its cis-isomer (cis-SG), as well as Schisandrin B (Sch B), were provided by the Chengdu Chroma Biotechnology Co., Ltd. The reference solutions (0.2 mg/ml) of cis-SG and trans-SG were prepared separately by methanol and stored in brown bottles. Lipopolysaccharide (LPS) was purchased from Sigma-Aldrich (Lot#113M4068V).

The P. multiflorum samples collected from patients were prepared through extraction for eight times using 50% ethanol-water solution via the ultrasound extraction method. Trans-SG was weighed precisely, dissolved in methanol to a concentration of 5 mg/ml, placed in a plain dish, and irradiated with UV light at 365 nm for 60 min. After volatilizing methanol in the fume cupboard, the samples were sealed and preserved in light-proof bags. The concentrations of cis-SG and trans-SG were determined by high-performance liquid chromatography.

Male Sprague Dawley (SD) rats weighing 200 ± 10 g were obtained from the Laboratory Animal Center of the Academy of Military Medical Sciences. They were housed in the Laboratory Animal Center of Military Medical Sciences and maintained under a 12-h light/dark cycle in a controlled temperature (25 ± 2°C) and humidity (50–60%) environment for a period of 1 week before evaluation. The animals were allowed access to food and water ad libitum.

The differential hepatotoxicity potential of these two stereoisomers was evaluated by a previously established rat model (Li et al., 2017), which used a nontoxic dosage of LPS pretreatment to induce minimal inflammatory stress. In a typical procedure, the animals were intragastrically administered with cis-SG or trans-SG or an equivalent volume of normal saline, separately, followed by a tail vein injection of LPS (2.8 mg/kg, Sigma-Aldrich) or normal saline 3 h later. At 7 h after the injection of LPS, the blood and liver samples were collected. The rat liver was fixed in 10% neutral buffered formalin for at least 72 h before being processed for histologic analysis. Paraffin-embedded sections were cut to 4-mm-thick pieces and stained with hematoxylin and eosin for microscopic examination. AST and ALT were also detected.

Human hepatoma cell lines and HepaRG cells were obtained from the Chinese Academy of Food and Drug Administration and cultured in Williams’ Media E (Gibco, United States) supplemented with 10% fetal bovine serum (FBS, Gibco, Canada), 2 mM GlutaMAX (Gibco, United States), 100 units/ml penicillin, and 100 μg/ml streptomycin. The medium was renewed every 2 days. The cells were maintained at 37°C with 5% CO₂ in a cell incubator with a water reservoir.

To prepare the hepatic organoids, the HepaRG cells (400 cells per well) with liver extracellular matrix were seeded into round-bottomed 96-well plates with an ultra-low attachment surface (Costa, Corning) and were incubated at 37°C with 5% CO₂. The culture medium was partially replaced by a fresh medium every 3 days.

Cell viability was detected by trans-SG and cis-SG. For the 2D cultured cell model, 100 μL of culture medium containing cells at a density of 8,000 cells/well was plated in 96-well plates, and after incubation of 24 h, the cells were treated with the drugs for 24 h. As for the 3D cultured HepaRG organoids model, the HepaRG cells were cultured under 3D culturing conditions for 14 days to form the organoids, and the organoids were then treated with testing drugs for 24 h once (as a single treatment) or for three repeated times of exposure at day 0, day 3, and day 5 and tested at day 7 (as repeated treatment). Cell viability was assessed using Alamar blue (Invitrogen) according to the manufacturer’s instructions. A microscope was used to observe the change of organoids.

The levels of lactic dehydrogenase (LDH), aspartate transaminase (AST), and alanine aminotransferase (ALT) in the supernatant gathered after treatment with drugs at different concentrations were tested using a reagent kit (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) according to the manufacturer’s instructions.

High-content imaging and analysis of organoids were performed using the Operetta High-Content Imaging System (PerkinElmer). The organoids were treated with trans-SG and cis-SG for 24 h and washed three times with PBS, and the corresponding probes were added to find the potential lesion mechanisms. All the fluorescence probes used in this study were purchased from Thermofisher. Live and dead cells were detected by calcein AM and EthD-1 probe (LIVE/DEAD™, L3224, 8 μM). Oxidative stress was detected by CellROX and mBcl probe (CellROX™ Deep Red Reagent, C10422, 5 Μm; Monochlorobimane, M1381MP, 100 μM). Mitochondria damage was detected by the MitoTracker probe (MitoTracker^® Deep Red FM, M22426, 200 nM). Steatosis was detected by a Nile Red probe (Nile Red, N1142, 1 μM). Cholestasis was detected by the CMFDA probe (5-chloromethylfluorescein diacetate, C7025, 25 μM). All the probes were incubated for 30 min, the cells were washed three times with PBS, and pictures were taken and analyzed in 3D space.

The chemical compounds in traditional Chinese medicine and their targets were acquired from the TCM-SP and the TCM Database@Taiwan. We then used these keywords, namely, mitochondrial depolarization, depolarized mitochondria, damaged mitochondria, mitochondrial disorders, mitochondrial dysfunction, mitochondrial impairment, and mitochondrial damage, to retrieve the corresponding target information and establish the mitochondrial lesion target database for the subsequent analysis. Based on the Database of Interacting Proteins (DIP), the interacting network between drug compounds and mitochondrial lesion–related targets was established. The network characteristic parameters of the key node proteins, such as degree (D), betweenness centrality (BC), and closeness centrality (CC), were calculated. For the key node proteins with a D value greater than the median of the D value of all nodes, the key drug targets were screened out from the node proteins whose BC and CC values were greater than the median of the two parameters of all nodes. The screen parameters were set as CC ≥ 0.183, D ≥ 2, and BC ≥ 0.000906 and a total of 19 targets were selected according to the network analysis results. Based on the established compound–target database, the candidate mitochondria protective agents were predicted and selected by using a hypergeometric distribution model to enrich the compound–target interaction network involving the 19 targets (p < 0.01).

The DILI cases caused by PM which were then treated with preparations containing Sch B were collected from the Fifth Medical Centre, Chinese PLA General Hospital (inpatient records between 2007 and 2016).

Total RNA was isolated using the RNeasy Mini Kit (Qiagen), after which cDNA was synthesized with reverse transcriptase (ReverTra Ace^® qPCR RT Master Mix, Toyobo) according to the manufacturer’s instructions. The qRT-PCR was performed with SYBR Green Master Mix (TOYOBO) on a Bio-Rad iQ5 real-time PCR detection system. Data were collected using the Bio-Rad CFX Manager software, and the expression of genes within a sample is normalized to the GAPDH expression by the 2^−ΔΔCt method. The primers used in this study are listed in Table 1.

The 3D cultured organoids were isolated as a single cell separately by trypsinization, washed three times with PBS, and made into a cell suspension. The MitoProbe™ JC-1 Assay Kit (Invitrogen, M34152) was used to detect mitochondrial membrane potential. The cell suspension was incubated for 20 min with 10 ug/ml JC-1 at 37°C in the dark. The cells were incubated for 15 min with the FITC-Annexin V/Dead Cell Apoptosis Kit (ThermoFisher company, V13242) in the dark at room temperature for early apoptosis and immediately analyzed by flow cytometry according to the instructions. The TUNEL assay was used to detect late apoptosis, and the cells were incubated only with PI for the cell cycle analysis according to the instructions.

All statistical data are expressed as means and SDs. Unpaired two-tailed Student’s t test was used to compare differences between groups. Differences were considered statistically significant if the p-value was smaller than 0.05. Significance was indicated as *p < 0.05, **p < 0.01, and ***p < 0.001. Experiments were repeated at least three times.

We collected the rest P. multiflorum from the patients, who had been diagnosed as PM—induced liver injury, and tested the content of cis-SG in these samples and normal samples using high-performance liquid chromatography (HPLC). As shown in Figure 1A, there was a significant difference in the cis-SG content in the DILI patients’ samples which was much higher than in the normal samples. To our knowledge, the trans-SG to cis-SG conversion occurred under UV light conditions (Figure 1B). In HPLC fingerprinting, we observed that the intensity of cis-SG and trans-SG was changed after UV light radiation (Figure 1C). Most of the trans-SG converted to cis-SG with UV light treatment (Figure 1D), indicating that the illumination is important for the storage of P. multiflorum.

To determine which of the two isomers was associated with hepatotoxicity, we administered isolated trans-SG and cis-SG to LPS-treated rats to compare the hepatotoxic intensity of the two compounds. Results in Figure 2A show that the administration of cis-SG and LPS result in liver histologic damage and significantly increased plasma alanine aminotransferase (ALT) and aspartate aminotransferase (AST) (Figure 2B). By contrast, the administration of trans-SG and LPS was not associated with liver injury, suggesting that cis-SG is the major hepatotoxic component.

To further reveal the possible hepatotoxic mechanism induced by cis-SG, the monolayer cultured HepaRG cells were used to detect the hepatotoxicity of cis-SG and trans-SG. However, there was no difference between cis-SG and trans-SG–induced hepatotoxicities. The cell viability and related indicators, such as lactate dehydrogenase (LDH), ALT, and AST, did not show any statistical difference with the treatment of cis-SG and trans-SG (Figures 3A,B).

Therefore, the model which could reflect the hepatotoxic difference between cis-SG and trans-SG was required. Next, the 3D cultured HepaRG organoids were further prepared to assess cis-SG and trans-SG–induced hepatotoxicities. Fortunately, due to the enhanced hepatic functions, the results obtained from the 3D cultured HepaRG organoids were consistent with earlier overall animal outcomes. By using the established 3D cultured organoids model, the significantly different hepatotoxic potential between the cis-SG and trans-SG stereoisomers was observed via either single or repeated dose. The cis-SG showed higher toxicity with the lower half-maximal inhibitory concentration (IC₅₀) than did trans-SG in single- or repeated-dose treatment on cells (Figures 3C,D). In addition, the cis-SG and trans-SG–induced cell death was in a concentration-dependent manner. The activity of LDH, ALT, and AST was significantly increased in cis-SG–treated cells (Figures 3E,F). Similarly, the recorded morphological images also confirmed that cis-SG caused more serious cell damage with visible loss of cells and impairment of spheroid structure, especially with the repeated-dose treatment (Figure 3G). Live/dead fluorescence staining was also used to qualitatively indicate the viability of the organoids with drug treatment. Figure 3H showed that more cells were alive (green) in trans-SG–treated organoids than in cis-SG–treated organoids. These aforementioned results suggest that long-term medication may increase the risk of liver injury. More importantly, the 3D HepaRG organoid culture system can be used to reveal the possible mechanism of hepatotoxicity induced by cis-SG.

To demonstrate the potential mechanisms of drug related to hepatotoxicity, multiparametric analysis of assessments in 3D cultured HepaRG organoids treated with trans-SG or cis-SG have been carried out. The cells were stained with selected fluorescence probes and monitored by high content imaging and analysis (HCA) after being treated with trans-SG or cis-SG for 24 h. HCA allows observation of drug-induced hepatotoxic alterations using hyper-multicolor cellular imaging. Different molecular probes were used to indicate the majority of reported changes in oxidative stress, mitochondria damage, steatosis, and cholestasis. For oxidative stress, the generation of reactive oxygen species (ROS) and depletion of intracellular glutathione (GSH) were detected by CellROX™ Deep Red Reagent and monochlorobimane (mBCl). As shown in Figure 4A, both trans-SG and cis-SG did not cause oxidative stress in cells. For mitochondrial damage, the change in mitochondrial membrane potential is the main reason for mitochondrial dysfunctions. MitoTracker^® Deep Red FM was used to analyze the alteration of mitochondrial membrane potential. cis-SG could induce a significant decrease in fluorescence intensity, indicating the decrease of mitochondrial membrane potential (Figure 4B). For drug-induced steatosis, it indicates to the overaccumulation of lipids, mainly triglycerides, inside hepatocytes. Here, Nile red was used to stain and quantify the accumulation of neutral (e.g., triglycerides) and polar (e.g., phospholipids) lipids by its uptake and retention in cells. The images and quantified results showed that trans-SG and cis-SG did not cause steatosis. For cholestasis, the hepatic bile acid transporter substrate 5-chloromethylfluorescein diacetate (CMFDA) was used to stain the canalicular structures. Similar to the steatosis result, there was almost no alteration to the bile canaliculus structures of the 3D cultured HepaRG cells after being treated with trans-SG or cis-SG. These results demonstrated that the cis-SG–induced hepatotoxicity was mainly because of mitochondrial damages in the cells, and these results assist in gaining an understanding of the action mechanism of P. multiflorum–induced hepatotoxicity.

Mitochondrial protective compounds were screened by a computer-aided network pharmacology approach from the natural products database which are listed in Table 2. Among these, Schisandrin B (Sch B) was ranked first and was available to be procured as a pure compound (Figure 5A). The herb containing Sch B, named Schisandra chinensis, is a well-used traditional Chinese medicine to treat acute liver injury in clinical applications. Hence, Sch B was selected as the mitochondrial protective agent to further verify the mechanism of cis-SG–induced liver injury. Firstly, the cytotoxicity of Sch B was tested. There was no significant hepatotoxicity to HepaRG organoids after being treated with Sch B in the concentration range of 0.625–10 μM (Figure 5B). When co-exposed with Sch B, the cis-SG–induced decrease of cell viability was significantly and dose-dependently ameliorated (Figure 5C). And by mitochondrial imaging, it was shown that the mitochondrial protective agent, Sch B, reversed the mitochondrial membrane potential, suggesting that Sch B could decrease mitochondrial damage induced by cis-SG (Figure 5D). The mitochondrial membrane potential was further detected by the MitoProbe™ JC-1 Assay Kit, which in the healthy mitochondria accumulates as a red fluorescent dimer and following mitochondrial injury is released into the cytoplasm as a green fluorescent monomer. As shown in Figure 5E, the flow cytometric results showed that the co-treatment of cis-SG and Sch B effectively reduced the cis-SG–induced mitochondrial damage. In addition, the hepatoprotective effect of the preparations containing Sch B was demonstrated by clinical data with the decrease of ALT and AST levels in patients’ serum (Figure 5F and Table 3).

Mitochondrial damage always induces cell apoptosis. cis-SG–induced apoptosis, indicated by either an increase of BAX and Caspase3 or a decrease of PCNA and Bcl2 gene expressions, was significantly reversed to control group levels (Figure 6A). Further analysis by flow cytometry showed that cis-SG induced 33.58% early apoptosis and 4.41% late apoptosis, while the combination of Sch B and cis-SG treatment showed a decreased proportion of either early or late apoptosis to 15.06% and 0.91%, respectively (Figure 6B). The inhibitory effect of Sch B on cis-SG–induced late apoptosis was also demonstrated by TUNEL analysis (Figure 6C). In addition, cell apoptosis was accompanied by DNA fragmentation. As shown in Figure 6D, the proportion of sub-G1 was increased after being treated with cis-SG, and Sch B could inhibit this effect.