Methanol (HPLC grade) was purchased from Wokai Chemical Technology Co., Ltd. (Shanghai, China). Acetonitrile (HPLC grade) was obtained from Merck Chemicals (Shanghai, China). Formic acid (HPLC grade) was purchased from TCI Chemical Industry Development Co., Ltd. (Shanghai, China). Other reagents and chemicals were of analytical grade and purchased from Kelong Chemical Reagent Factory (Chengdu, China).

C-8-O-β-D-glu (purity above 98%) was purchased from Chengdu Chroma-Biotechnology Co., Ltd. (Chengdu, China), and was dissolved in DMSO (MP Biomedicals LLC, France) at the concentration of 96 mM as stock solution.

The entire research was carried out on human liver cell line L-02 cells purchased from KeyGEN BioTECH (Nanjing, China). The cells were cultivated in DMEM medium (GIBCO, United States) supplemented with 10% FBS (GIBCO, United States) and antibiotics (100 U/ml of penicillin and 0.1 mg/ml of streptomycin). They were incubated at 37°C, 5% CO₂ in LabLine CO₂ Incubator (Thermo, United States) and subcultured every 2 days.

C-8-O-β-D-glu stock solution was diluted to 96 μM with cell culture medium. After being incubated for 0, 6, 12, 24, and 48 h, 10 μL sample were analyzed (n = 3) by HPLC. Agilent 1260 Infinity HPLC system (Agilent, United States) was applied to conduct the analysis on a Zorbax Eclipse Plus C₁₈ column (4.6 × 250 mm, 5 μm, Agilent, United States) at 30°C. The analyte was eluted by 0.1% phosphoric acid water: methanol (20:80) at 1 ml/min for 10 min.

Exponentially growing cells were plated in 96-well plate (Costar, United States) at the density of 6 × 10³ per well and grew in incubator for 24 h. At the same time, the culture medium with 0.1% DMSO were added into wells without cells to zero the OD value. The adhered cells were treated with different concentrations of C-8-O-β-D-glu (0, 12, 24, 48, and 96 μM) prepared in DMEM medium supplemented with 0.1% DMSO and cultured for 24 h. Then the supernatants were carefully removed, and 20% 3-(4, 5-dimethylthiazol-2-yl) 2, 5- diphenyltetrazolium bromide (MTT) were added. After 4 h, MTT-formazan crystals were dissolved by 150 μL DMSO. The absorbance of the solution was measured at 570 nm (n = 6). The influence of different concentrations on cells viability was calculated by the percentage of viable cells between drug experimental groups and the CG.

Exponentially growing cells were plated in 96-well plate at the density of 6 × 10³ per well and grew in incubator for 24 h. Then the cells were treated with different concentrations of C-8-O-β-D-glu (0, 24, 48, and 96 μM) prepared in DMEM medium supplemented with 0.1% DMSO for another 24 h. After that, the medium was removed and the cells were washed with PBS. Then cells were stained by 50 μL freshly prepared Rho123, 10 μM (Beyotime, China), per well. After 30 min incubation without light, the dye was removed. Cells were washed with PBS and then exposed to Bisbenzimide H 33342(10 μM, Sigma, United States) for 15 min in incubator for imagination.

Cells were imaged under High Content Screening ImageXpress^® Micro (Molecular Devices, United States). The detection conditions were set as follows: the first channel wavelength was 350 nm/460 nm irradiation for Bisbenzimide H 33342 labeled nuclei. The second channel wavelength was 507 nm/530 nm irradiation for Rho123 labeled mitochondria. Five images were captured per well for image analysis performed with MetaMorph image processing. Cells number was directly counted by the software. Average nucleus area, DNA content and MMP were calculated based on the data recorded.

All metabolite biomarkers were identified based on exact molecular weight, retention time and MS/MS information at first. Then, several databases including Human Metabolome Database (HMDB¹), Metlin², massbank³, LipidMaps⁴, and mzclound⁵ were used for further confirmation. To further determine the metabolic patterns of differential metabolites in each group, the dataset was scaled by heat-map package in R(v3.3.2). The samples and metabolites were bidirectionally clustered. To determine the relevant metabolic pathways, the metabolic pathway analysis of potential biomarkers was performed using Kyoto Encyclopedia of Genes and Genomes (KEGG⁶) database. And possible metabolic pathways were identified by metabolic pathways enrichment and topology analysis through MetPA⁷ database.

Data were analyzed by SPSS 21.0 for Windows (SPSS Inc.). Results were represented as mean ± SD and evaluated using the two-tailed unpaired student’s t-test or one-way analysis of variance. The p < 0.05 was considered to be significant and p < 0.01 to be very significant.

The stability of C-8-O-β-D-glu (96 μM) in cell culture medium was analyzed within 48 h. The concentration of C-8-O-β-D-glu incubated for 0, 6, 12, 24, and 48 h were shown in Table 1. The RSD (<2%) of drug concentration at each time point indicated that C-8-O-β-D-glu was stable in cell culture medium which guaranteed the study was focused on C-8-O-β-D-glu instead of other chemicals.

To investigate the effects of C-8-O-β-D-glu on L-02 cells viability and select the concentrations for the follow-up studies, MTT assay was performed at first. In CG, the same FBS and antibiotics were added to eliminate the blank interference. As shown in Figure 2, 12 and 24 μM C-8-O-β-D-glu significantly increased the cells viability (p < 0.05), while 48 μM exert no influence. However, when concentration increased to 96 μM, significant inhibition of the viability of L-02 cells after 24 h was observed (p < 0.05). It indicated that lower concentrations of C-8-O-β-D-glu increased cells viability, but high concentration led to liver injury. Consequently, 24, 48, and 96 μM were selected as the low, medium and high concentrations of C-8-O-β-D-glu, respectively in the follow-up studies.

To further explore the effects of C-8-O-β-D-glu on L-02 cells, HCA was applied to examine the changes of nuclei and mitochondria after the cells being exposed to the drug for 24 h. The representative cell staining images of the four groups were shown in Figure 3, which showed an increase of cells number in LCG and a decrease in HCG when compared with CG. Based on staining on nuclei by Bisbenzimide H 33342, the number of L-02 cells was counted. The results were in accord with the MTT assay (Figure 4A) that HCG showed significant cytotoxicity compared with CG (p < 0.01). It was also observed that the average nucleus area and average DNA content of L-02 cells in HCG were significantly increased (Figures 4B,C) (p < 0.01). As for mitochondria, the MMP was calculated from the intensity of Rho123 staining divided by cells number. As shown in Figure 4D, the MMP of HCG was significantly higher than that of CG (p < 0.01). However, no statistical difference of cells number, average nucleus area, average DNA content and MMP can be found among LCG, MCG and CG. These results above suggested that the hepatotoxic mechanisms of C-8-O-β-D-glu at high concentration might be related to the changes in nuclei and mitochondria.

All cell samples were analyzed by UPLC-MS/MS in both positive and negative ionization modes, and the representative BPC (Figure 5), of four groups were obtained under the optimal conditions. The chromatographic peaks considered to be the representative chemical fingerprints of endogenous metabolites were detected within 18 min with some remarkable differences observed among the four groups.

The QC samples were used to correct the deviations of analysis results from mixed samples and the errors caused by analytical instrument. PCA score plots (Figures 6A,B) showed that the QC samples gathered together in both positive and negative mode, which indicated that the test samples and instrument were stable in long-term run.

To obtain the difference of metabolic components among four group samples, multivariate statistical analysis method was used to screen the metabolites of each sample. PCA score plot (Figures 6C,D) showed a great separation trend of CG and drug experimental groups (R2X = 0.215, Q2 = 0.048 in negative mode; R2X = 0.223, Q2 = 0.025 in positive mode). Supervised analysis including PLS-DA and OPLS-DA was applied to identify the differences and outliers between CG and drug experimental groups. The cross test parameters R2X, R2Y, and Q2 values of PLS-DA model were 0.312, 0.989, and 0.805 in negative mode and 0.371, 0.994, and 0.841 in positive mode, which suggested that fitness and prediction of the model were good. Moreover, the permutation test was applied to evaluate whether the model was overfitting. The results shown in Figures 7C,D indicated that the model was valid. As shown in PLS-DA score plot (Figures 7A,B), every group was clearly separated from each other in both positive and negative modes, which indicated that there were significant metabolite differences among the groups. It also suggested that the effects of C-8-O-β-D-glu on L-02 cells were related to the concentration. OPLS-DA was carried out to further confirm the potential differential metabolites of drug experimental groups compared with CG, respectively. The OPLS-DA score plot was shown in Figures 7E–G. The VIP values of the metabolites were calculated based on this to screen the potential biomarkers.

VIP values reflected the influence of each variable, and a larger distance indicated a more important projection. Therefore, all the metabolites were selected according to the VIP values from OPLS-DA firstly. The corrected p-values (q-values) between CG and drug experimental groups were also applied to screen the differential metabolites. Ultimately, 42 endogenous metabolites contributing most to the separation of drug experimental groups from CG were selected, which might account for the effects of C-8-O-β-D-glu on L-02 cells. These metabolites were identified according to MS/MS information combined with online database information. There were 18 and 5 identified differential metabolites between CG and LCG or MCG, respectively. As for biomarkers of hepatotoxicity of C-8-O-β-D-glu, 26 metabolites were identified as differential metabolites between HCG and CG. Changes of all the differential metabolites between CG and drug experimental groups were shown in Figure 8, a heat map.

The pathways influenced by the changes of above metabolites were identified by metabolic pathway enrichment and topology analysis through MetPA database (Figure 9). The -log(p) value from the pathway enrichment analysis and the pathway impact value from the pathway topology analysis were calculated by MetaboAnalyst 4.0. The higher the -log(p) value and the pathway impact value, the more important the pathway. Four crucial pathways of C-8-O-β-D-glu on L-02 cells were finally identified based on the pathway impact and -log(p) value, which were summarized in Table 2 suggesting that different concentrations of C-8-O-β-D-glu would affect L-02 cells viability through the impacts on the pathways of alanine, aspartate and glutamate metabolism, arginine and proline metabolism, phenylalanine metabolism, beta-alanine metabolism in different ways.

Remarked dysregulation of alanine, aspartate and glutamate metabolism were observed in both LCG and HCG. In addition, low concentration of C-8-O-β-D-glu would impact arginine and proline metabolism. High concentration of C-8-O-β-D-glu would impact phenylalanine metabolism and beta-alanine metabolism. There was no important pathway involved in MCG vs. CG. Based on these metabolic pathways, relevant important differential metabolites were selected and summarized in Table 3. As can be seen in Figure 10, relative intensities of most differential metabolites varied in a dose-dependent manner. For example, contents of L-aspartic acid, (N(omega)-L-arginino)succinic acid, L-proline, N-phosphocreatine, trans-2-coumaric acid and L-tyrosine decreased with increasing dose, while the content of N(6)-(1,2-dicarboxyethyl)-AMP increased with increasing dose.

Because of the poor solubility of C-8-O-β-D-glu, the concentration wasn’t high enough to detect the IC₅₀ of C-8-O-β-D-glu. So the concentration with maximum solubility (96 μM) was used as the high dosage in MTT assay.

The results of MTT assay suggested that C-8-O-β-D-glu in low concentrations (12 and 24 μM) increased L-02 cells viability. However, HCA results didn’t show statistical difference in LCG and MCG compared to CG. It suggested that C-8-O-β-D-glu at lower concentrations did not affect the cells number or cells status of the nuclei and mitochondria of L-02 cells, which was inconsistent with MTT assay. Although the results of MTT assay implied a rise of mitochondrial activity in LCG, MMP analyzed by HCA remained unchanged. It indicated the mitochondria in LCG were not significantly affected. Since there was an upward trend in cells number of LCG (1925 ± 283 in LCG, 1594 ± 163 in CG) measured by HCA, the difference between the cells viability and cells number might be caused by the difference between the two experimental methods. The results of MTT assay and HCA experiment both indicated that high concentration (96 μM) of C-8-O-β-D-glu inhibited cells growth. Moreover, significant changes in cell nuclei and mitochondria were found in HCG. Normally, the cells injured by toxic substances showed nucleus shrinkage caused by cell apoptosis and necrosis. However, both the average nucleus area and DNA content were increased in HCG which needed further study. Normal MMP is mainly dependent on the proton pump located in the gap between mitochondrial matrix and mitochondrial membrane so that different protons between the matrix and the membrane maintain a certain electrochemical gradient. MMP reflects the properties of electron transport chain and changes in pathological conditions. Respiratory chain of mitochondria is the main site of ROS production (Brand and Nicholls, 2011). Under normal condition, mitochondria produce a small amount of ROS to maintain the normal physiological functions of cells. However, the reduced proton reflux increases MMP in pathological condition. Then electrons are leaked which resulted in producing a large number of ROS (Su et al., 2016). Finally, excessive ROS causes oxidative damage to cells (Ray et al., 2012). Thus, we conjectured that the increase of MMP level in HCG resulted in L-02 cells damage by affecting the production of ROS.

Changes of intracellular metabolites reflect abnormalities of cells’ physiological state. The extent of important metabolites change has a relationship with C-8-O-β-D-glu concentration. More importantly, the contents of several metabolites in LCG and HCG changed in different trends compared to CG, including L-aspartic acid, (N(omega)-L-arginino)succinic acid and L-proline. These metabolites might be more relevant to the different effects of C-8-O-β-D-glu on L-02 cells.

In present research, three concentrations (24, 48, and 96 μM) of C-8-O-β-D-glu affected different metabolic pathways in varying degrees to induce different effects on L-02 cells viability. The schematic diagram of these pathways was shown in Figure 11. The metabolism of amino acids occurs mainly in liver. Therefore, physiological changes of liver are usually accompanied with the abnormality of amino acid metabolism. Previous metabonomics studies have shown that anthraquinones would induce liver injury by disordering amino acid metabolism, including tryptophan metabolism (Zhang et al., 2016), phenylalanine metabolism, alanine, aspartate and glutamate metabolism (Li et al., 2017), L-threonine and serine metabolism (Zhang et al., 2015). In this research, we found amino acid metabolism was involved in most identified pathways.

Low concentration of C-8-O-β-D-glu might promote L-02 cells viability by regulating the expression of alanine, aspartate and glutamate metabolism, arginine and proline metabolism. High concentration of C-8-O-β-D-glu also affected alanine, aspartate and glutamate metabolism, but to the contrary trend. Thus, differential metabolites including L-aspartic acid, (N(omega)-L-arginino)succinic acid, N(6)-(1,2-dicarboxyethyl)-AMP and N-acetyl-L-aspartic acid could be considered as potential biomarkers of C-8-O-β-D-glu affecting liver cells viability. In addition, arginine and proline metabolism was significantly changed in LCG. High concentrations of C-8-O-β-D-glu disrupted phenylalanine metabolism and beta-alanine metabolism. The results indicated that the mentioned metabolites were closely related to promoting or toxic effects of C-8-O-β-D-glu. In MCG, there was no significant change in cellular metabolic pathway. This was consistent with MTT results that medium concentration had little effect on cell viability.

L-aspartic acid is an important non-essential amino acid which participates in many biochemical processes including urea cycle, gluconeogenesis, and malate-aspartate shuttle etc. Rising content of L-aspartic acid in LCG indicated a gain in energy generation via these processes. N-phosphocreatine and L-proline were also up-regulated in LCG. N-phosphocreatine is an important energy source for cells. Promotion of these processes related to energy generation will result in an increase in L-02 cells viability.

The content of L-aspartic acid, as well as its by-products (N(omega)-L-arginino)succinic acid and N-acetyl-L-aspartic acid were decreased in HCG. One important physiological function of L-aspartic acid is providing an ammonia molecule for the urea cycle to remove excess ammonia (Mew et al., 2011). Thus, the decrease in L-aspartic acid content will cause ammonia accumulation and damage to the liver. (N(omega)-L-arginino)succinic acid is synthesized from L-aspartic acid by the loss of ammonia. The decline of (N(omega)-L-arginino)succinic acid in HCG conformed to the decrease of L-aspartic acid which can be produced from oxaloacetate via transamination by the function of aspartate aminotransferase (AST). AST is also one of the most commonly used indicators for diagnosis of liver disease. When the hepatocyte is damaged, the permeability of the cell membrane increases, and the AST outflows from the cell (Otto-Ślusarczyk et al., 2016). The decrease in L-aspartic might be caused by the outflow of AST. Based on these facts, we speculated that high concentration of C-8-O-β-D-glu induced L-02 cells damage via reducing AST in cells, inhibiting the production of L-aspartic acid, increasing the amount of ammonia and finally aggravating the liver injury. In addition, L-proline in HCG was significantly reduced. It has been reported that L-proline can protect cells by directly cleaning up ROS, protecting and up-regulating the antioxidant enzymes, and maintaining the key redox molecules (Krishnan et al., 2008; Szabados and Savouré, 2010; Natarajan et al., 2012). According to our previous speculation about the result of HCA, high concentration of C-8-O-β-D-glu induced L-02 cells damage might be related to oxidative stress. Therefore, the decreasing of L-proline probably aggravated the damage caused by oxidative stress.

This study revealed that the different effects of C-8-O-β-D-glu on liver L-02 cell with varying concentrations, which prompts us to pay more attention to dosage applied in clinic. But the dosage-relationship of effect or toxicity remained obsure which will be investigated in the further experiment.