Scleromitron diffusum (Willd.) R. J. Wang (10 batches) were purchased from Jiangxi and identified by Prof. Tianai Gao from Shanxi Provincial Food and Drug Inspection Institute. Cellulase was obtained from Zhejiang Yinuo Biotechnology Co., Ltd (Zhejiang, China). The 36 flavonoid standards (seen in Table 3) were purchased from Sigma-Aldrich and Steraloids. Fenofibrate, formic acid, methanol, isopropanol, and acetonitrile were HPLC grade from Sigma. Macroporous adsorbent resin AB-8 was obtained from Chengdu Grecia Chemical Technology Co., Ltd. 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2-H-tetrazolium bromide (MTT), RMPI 1640 medium, fetal bovine serum (FBS), and trypsin were purchased from Wuhan Servicebio Biological Technology Co., Ltd (Wuhan, China). Triglyceride (TG), total cholesterol (TC), aspartate transaminase (AST), alanine transaminase (ALT), γ-Glutamyltranspeptidase (γ-GT), superoxide dismutase (SOD), glutathione (GSH), lactate dehydrogenase (LDH), alkaline phosphatase (ALP), and malondialdehyde (MDA) assay kits were supplied by Nanjing Jiancheng Bio-engineering Institute (Nanjing, China). Ethanol, methanol, formic acid, and acetonitrile were of chromatographic grade.

Dried whole Scleromitron diffusum (Willd.) R. J. Wang powders were sieved through a 60 mesh sieve (0.25 mm), degreased by ultrasonic for 30 min, and dried to obtain defatted powders. The conditions of the four extraction methods are shown in Table 1.

TFH and Ey-TFH were purified by macroporous adsorption resin AB-8 with ethanol (95%, v/v). Next, eluates from AB-8 resin were concentrated at 45°C under reduced pressure and dried. The resulting dry powders were dissolved in a methanol-water solution (7:3, v/v). After vortexing and centrifugation, the supernatants were collected and analyzed by UPLC-MS/MS.

The method described by Li YL. et al. (2020) was used to determine the yields of the total flavonoids. First, 0.6 ml of the total flavonoids purified solution was transferred to a 10 ml volumetric flask and mixed with 0.3 ml sodium nitrite (5%, w/v), 0.3 ml aluminum nitrate (10%, w/v), and 4 ml sodium hydroxide (4%, w/v). The final volume of the mixture was adjusted to 10 ml with methanol. Then, the absorbance of the mixture at 500 nm was determined by a UV-vis spectrophotometer (MAPADA instruments Co., Ltd., Shanghai, China). The contents of the total flavonoids were expressed as rutin equivalents through the standard calibration curve (y = 14.07x+0.0116, r = 0.9990). The total flavonoids yields were calculated as follows: Yield  ( mg / g ) = the mass of extracted flavonoids  ( mg ) the mass of dried sample  ( g )

AB SCIEX 5500 QTRAP was used. Selected/multiple reaction monitoring with positive and negative ion switching mode was performed. The ion source parameters of positive ion mode were as follows: Source temperature, 550°C; Gas 1, 55; Gas 2, 50; CRU, 30; ISVF, 5500 V. The ion source parameters of negative mode were as follows: Source temperature, 550°C; Gas 1, 55; Gas 2, 50; CRU, 30; ISVF, -4500V. The data were analyzed using MultQuant and Analyst software.

Equal amounts of all standards were mixed to prepare quality control (QC) samples which were used to ensure the precision of the analytical method.

A serial dilution was used to prepare calibration solutions from 125 ng/ml stock solutions of 36 flavonoid standards. Regression lines were calculated by the Least Squares Method. The X-axis showed concentrations (pg/ml) of different solution standards, and Y-axis denoted corresponding peak areas. Correlation coefficient values (R) and linear range were obtained.

The TFH and Ey-TFH samples were used for UPLC-MS/MS analysis and repeated three times.

The data were imported into SIMCA 14.1 software to analyze differential components of TFH and Ey-TFH. Hierarchical Clustering Analysis (HCA) was used to analyze the difference between each method. Principal component analysis (PCA) was a multivariate statistic method to investigate the correlation of multiple variables. Orthogonal partial least squares discrimination analysis (OPLS-DA) was a supervised multivariate pattern recognition method that was used to screen out potential differential components.

HL-02 cells (obtained from BOSTER Biological Technology Co., Ltd.) were cultured in RMPI 1640 with 10% FBS, penicillin (100 U/mL), and streptomycin (100 μg/ml) under 5% CO2 at 37°C. When cells were in a logarithmic growth phase, they were treated separately with different concentrations of TFH and Ey-TFH (0, 31.3, 62.5, 125 μg/ml) for 12 h and then exposed to H₂O₂ (200 μmol/L). The six flavonoid compounds, including apigenin, chrysin, genistein, isovitexin, naringin, and vitexin (1, 2, 4, 8, 16, 32, 64 μmol/L), were added in the same way as Ey-TFH. The non-treated HL-02 cells were used as the control group.

The protective rates for HL-02 cells were determined by the MTT method. The OD was measured, and the protection rates were calculated according to the following equation: Protective rate ( % ) = OD administration − OD model OD control − OD model × 100

For the experiment, 60 rats (provided by the Experimental Animal Center of Shanxi Medical University, No. 22829) were fed adaptively for one week and randomly divided into six groups. Animals were given either a standard chow diet (control group, n = 10) or a high-fat diet (HFD, 45% calories from fat, Jiangsu Synergy Medical Bioengineering Co., Ltd., n = 50) for three weeks. After successfully modeling, positive control fenofibrate (83 μmol/kg) or Ey-TFH (low-dose, 60 mg/kg; medium-dose, 90 mg/kg; high-dose, 120 mg/kg) solved in distilled water were administrated intragastrically to the assigned HFD groups once every day for two weeks. Meanwhile, the control and model groups were administered the same distilled water (vehicle). Orbital blood and liver tissue of rats were collected and frozen at −80°C for further analysis. According to the manufacturer’s protocols, TG, TC, γ-GT, GSH, SOD, MDA, LDH, AST, ALT, and ALP were investigated in rats’ serum. Rats were killed by decapitation. The livers were quickly separated, washed with saline, and weighed for liver index evaluation. The liver index was calculated according to the following equation: Liver index = liver weight ÷ body weight 

Finally, liver tissues were fixed in 4% neutral formalin solution, embedded in paraffin, and were stained with hematoxylin and eosin (H&E). Subsequently, the liver sections were observed under an Olympus iX-51light microscopy (×100).

The data were expressed as means ± standard deviation (SD). Statistical analysis was performed using GraphPad Prism 8.4.0 software (San Diego, CA, United States). A T-test was used to compare the differences between groups. p < 0.05 was considered significant.

The yields of the total flavonoids of Scleromitron diffusum (Willd.) R. J. Wang separately extracted by bio-enzymatic, reflux, ultrasonic, and ultrasound-assisted enzymatic methods were shown in Table 2. The yield of total flavonoids from bio-enzymatic was higher than yields obtained by reflux and ultrasonic. The bio-enzymatic method extracted the yield of total flavonoids comparable to the ultrasound-assisted enzymatic method. The results indicated that enzymes played a major role in harvesting more flavonoids.

The contemporaneous quantification of 36 flavonoid compounds was validated. The extracted ion chromatograms (XIC) of 24 representative flavonoid standards are shown in Figure 1A. The retention time, mass information, standard curves, and concentration range of 36 flavonoids standards are shown in Table 3. As seen in Figure 1A and Table 3, each component could be well separated, and the peak shape was sharp and symmetrical. High correlation coefficient values (R² > 0.99) were obtained, indicating strong linearity at a relatively wide range of concentrations. The relative standard deviation (RSD) of 24 representative QC samples is shown in Figure 1B. The precisions were < 30%. The obtained results demonstrated that this UPLC-MS/MS method met the methodological requirements.

The quantitative determination of 24 representative flavonoids components in TFH and Ey-TFH are shown in Table 4 (Supplementary Sheet S1). The total levels of flavonoids in the studied extracts ranged from 201 to 207 ng/g. There was no difference between TFH and Ey-TFH in contents. In sharp contrast, compared with TFH, the contents of six flavonoids components (genistein, isovitexin, luteolin, luteolin-7-O-glucoside, naringin, and vitexin) were higher in Ey-TFH (p < 0.05 or p < 0.01). Furthermore, Ey-TFH contained these three components (apigenin, chrysin, and dihydrokaempferol), which were not detected in TFH.

In order to investigate the differences among the four preparation methods, a widely used tool, PCA, was applied. The results were shown in Figures 2C,E,G. According to PCA relevant results (Figure 2E), three principal components were included as the optimal number of latent variables. R ² _X and Q ² were important parameters of the PCA model. The quality of the model was described by the Q ² and R ² (R ² _X or R ² _y) parameters. Q ² denoted the predictability, and R ² indicated the goodness of fit. In this study, R ² _X = 0.86, Q ² = 0.543. Both parameters were greater than 0.5, and all the samples fell within the 95% confidence interval, suggesting that the model was reliable. The samples could be classified into two separate categories based on the PCA scores plot (Figure 2E). Samples of bio-enzymatic method (A1, A2, A3) belonged to category 1, while samples of reflux method (B1, B2, B3), ultrasonic method (C1, C2, C3), and ultrasound-assisted enzymatic method (D1, D2, D3) belonged to category 2. That is to say, Ey-TFH belonged to category 1, while TFH belonged to category 2. That implied dissimilarity between Ey-TFH and TFH. According to Figure 2C, the overlaps among the samples of reflux method (B1, B2, B3), ultrasonic method (C1, C2, C3), and ultrasound-assisted enzymatic method (D1, D2, D3) were apparent. In other words, there was no significant difference between these three methods. Distance to model (DModX) was applied to measure mismatch with normal. As shown in Figure 2G, the DmodX values of all samples were within the control limits, indicating that all samples had reliable quality.

OPLS-DA was also conducted to explore differential components contributing to the group separation. OPLS-DA was a kind of supervised pattern recognition method. It was designed to remove the variation not associated with the response being studied. An improved level of group classification and a better understanding of the variations responsible for classification was constructed. A permutation test was adopted to reveal overfitting (Figure 2B). All the left points of R² and Q² were lower than the rightmost original ones. The vertical axis intercept by the regression curve of Q2 points was less than zero. These indicated that this OPLS-DA model was reliable and free from overfitting. There were two principal components, as shown in the OPLS-DA fitting results (Figure 2D). The R ² and Q ² values were nearly close to 1 (R ² _X = 0.88, R ² _Y = 0.98, Q ² = 0.87), suggesting the model was validated and considered highly relevant. Greater R ² and Q ² values indicated that OPLS-DA analysis could better reflect between-group differences. Figure 2F shows that the samples of the bio-enzymatic method (A1, A2, A3) were distributed into a cluster distinct from those of the reflux method (B1, B2, B3), ultrasonic method (C1, C2, C3), and ultrasound-assisted enzymatic method (D1, D2, D3). Namely, Ey-TFH was distributed into a cluster distinct from TFH. Similar to the PCA loading plot, the separated distributions implied discrimination between TFH and Ey-TFH. Also, OPLS-DA was able to separate the samples of reflux, ultrasonic, and ultrasound-assisted enzymatic methods, which were overlapped completely in the PCA loading plot. The DmodX value of all samples (Figure 2H) showed that there were no abnormal points, demonstrating that the data were reliable.

Table 5 illustrates the variable importance in projection (VIP) values of the OPLS-DA model. The greater VIP value, the stronger contribution to the model explaining. VIP score >1 was usually used as a threshold. VIP value confirmed the assessment results of differential components, as shown in section 3.3. For three compounds only found in Ey-TFH, contents shown in Table 4 and VIP values of apigenin and chrysin were higher than those of ihydrokaempferol. Taking both aspects into consideration, apigenin and chrysin, but not ihydrokaempferol, were selected for the subsequent in vitro experiments.

A volcanic map was drawn to present the individual components that contributed to group classification in the next step. The results are depicted in Figure 3. The x-axis was the logarithm (base 2) of fold changes (FC). The y axis was the negative logarithm (base 10) of p-values (as it has been calculated in Table 4). Two vertical dashed lines represented FC < 2 or FC > 2, respectively. The horizontal dashed lines indicated p = 0.05. The upper layer of Figure 4 shows a comparison between TFH and Ey-TFH. Green or red represented significantly decreased or increased components in terms of FC and p-value, respectively. Red triangles represented components (genistein, isovitexin, naringin, and vitexin) that met the filtering criteria (FC > 3.5 and p < 0.01). The lower layer of Figure 3 compared extracts prepared by reflux, ultrasonic, or ultrasound-assisted enzymatic method. There was no significant difference in components among them.

After the comprehensive screening process, six representative differential components (vitexin, isovitexin, naringin, genistein, apigenin, and chrysin) were filtered for the subsequent cellular tests.

This part explored the hepatoprotective activities of TFH, Ey-TFH, and six further screened difference components. The hepatoprotective properties against H₂O₂ (200 μmol/L, 4 h) induced damage on HL-02 hepatocytes were studied. As shown in Figure 4A, the protective rates of TFH and Ey-TFH increased dose-dependently. Ey-TFH possessed the highest protection rate of all the tested concentrations compared with TFH (p < 0.05). According to Figure 4B, six differential components (vitexin, isovitexin, naringin, genistein, apigenin, and chrysin) all exerted significant protective effects on HL-02 hepatocytes at 8 and 16 μmol/L. In this light, it can be speculated that the hepatoprotective effect of six differential components might contribute to the superior protection of hepatocytes of Ey-TFH.

Excess hepatic fat accumulates continuously at the early stage of NAFLD. This condition makes the liver extremely vulnerable to other damaging factors such as a plethora of oxidative stress, dysregulated hepatocyte apoptosis, and inflammation. These factors act together or separately, which would drive the progress from simple steatosis to nonalcoholic steatohepatitis. The therapeutic effect of Ey-TFH on NAFLD was investigated with a rat model of high-fat diet-induced obesity. We firstly investigated the weight gain and liver indexes. As shown in Figures 5C,D, Ey-TFH could reduce the risk of excessive weight gain and liver indexes. Next, we measured liver fat accumulation markers, serum TG and TC, after being treated with Ey-TFH. TC and TG contents ameliorated significantly after medium-dose and high-dose Ey-TFH treatment (Figures 5E,F). Taken together, Ey-TFH suppressed hepatic lipid accumulation.

We tried to learn more about the improvement effect of Ey-TFH for NAFLD. Therefore, HE staining was used to assess the injury in liver sections. The result showed that Ey-TFH significantly reduced hepatic injury (Figure 6A). The liver injury markers such as ALT, AST, and LDH were measured. Large amounts of ALT, AST, and LDH were released from the damaged hepatocyte. The ALT, AST, and LDH sharply decreased after administration of Ey-TFH (Figures 6B–D). This showed that Ey-TFH significantly inhibited the elevation of liver injury markers. In addition, ALP and γ-GT, which were recognized markers of chronic inflammation, dramatically decreased in Ey-TFH groups (Figures 6E,F). These dates suggested that Ey-TFH protected the liver and blocked inflammation progress at the onset of NAFLD.

After high-dose Ey-TFH administration, as shown in Figure 7, the activities of antioxidant enzymes, SOD, and GSH increased significantly. MDA contents, which represented the peroxidation levels of the membrane lipid, were reduced significantly by Ey-TFH intervention. The results clarified that Ey-TFH treatment could attenuate oxidative stress by reducing oxidative levels and promoting antioxidative processes.