The active compounds of CSP were collected from the Traditional Chinese Medicine Systems Pharmacology database (TCMSP, https://old.tcmsp-e.com/tcmsp.php, updated in 2020). The TCMSP is a specialized systems pharmacology database platform of Chinese traditional herbal medications, which contains the relationships among medications, targets, and diseases. Oral bioavailability (OB) is an important pharmacokinetic parameter for oral TCM, as it indicates the quantity of orally administered unchanged drugs that reach the systemic circulation. The Drug-likeness (DL) index refers to the structural similarity of compounds with known drugs, and it is widely utilized to assess the suitability of a component to be used as medication. Herein, the screening criteria in this research were an OB value of ≥30% and a DL value of ≥0.18.

The potential targets of active compounds in CSP were retrieved from the TCMSP database and Swiss Target Prediction database (STP, http://swisstargetprediction.ch/, updated in 2022). The Uniport database (https://www.uniprot.org, updated in 2022) was used to convert protein names to their corresponding gene names.

NAFLD-related human genes were identified by searching the Gene Cards database (https://www. genecards.org/, version 5.9), Online Mendelian Inheritance in Man database (OMIM, http://omim.org/, updated in 2022), and Therapeutic Target Database (TTD, https://db.idrblab.net/ttd/, updated in 2021). “Non-alcoholic fatty liver disease” was used as the keyword.

The obtained drug targets were mapped to NAFLD disease targets to obtain a Venn diagram of the intersected gene symbols. After collecting the interactions of those targets from above, a complex information network was established. Finally, a visual analysis of the “Components-Genes Network” was conducted using Cytoscape 3.9.1 software. Cytoscape 3.9.1 is a robust bioinformatics software for visualizing molecular interaction networks.

The protein-protein interaction (PPI) network was obtained from the STRING database (https://string-db.org, version 11.5). The PPI network focused on finding hubs in which highly connected proteins were considered to play a crucial role between related ingredients and their probable targets. Afterward, all data were imported into Cytoscape 3.9.1 software to obtain PPI complex diagrams in a representable form to characterize the association between CSP and NAFLD intersection genes. The color and size of nodes represent the degree of the targets.

To further characterize the molecular mechanism of CSP in NAFLD, Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment were performed by the Metascape database (https://metascape.org/, updated in 2022), with p < 0.01, min overlap value 3 and min enrichment value 1.5.

The CSP crude herbs were obtained pharmacy from Guang Dong Provincial Hospital of Chinese Medicine. All of these herbal medicines conformed to the requirements of the Chinese Pharmacopoeia 2020 edition and are in strict accordance with national execution standards. The dosage proportions of Radix Bupleuri (No.2106003, Shanxi, China; 5 g), Paeoniae Radix Alba (No.YPB1E0001, Anhui, China; 10 g), Chuanxiong Rhizoma (No.YPB0L0001, Sichuan, China; 5 g), Aurantii Fructus (No.210502, Jiangxi, China; 15 g), Citrus Reticulata (No. 210600521, Guangdong, China; 10 g), Cyperi Rhizoma (No. YPB1D0002, Anhui, China; 10 g) and licorice (No. YPB1D0002, Gansu, China; 10 g) was 6:6:4.5:4.5:4.5:4.5:1.5. To prepare the aqueous extract of CSP, all crude CSP herbs were decocted with an eightfold volume of distilled water twice, the first time for 1.5 h and the second time for 2 h. Then, the decoction was collected after filtration. Finally, the decoction was concentrated at 1.892 g/ml under reduced pressure using a rotary evaporator (IKA, Germany), and stored at 4°C.

A Q-Exactive mass spectrometer (Thermo Fisher Scientific, Inc., San Jose, CA, United States) and an A Waters, ACQUITY UPLC HSS T3 (2.1 mm × 100 mm, 1.8 μm, Waters, Milford, MA, United States) column were used for LC separation. Gradient elution: acetonitrile/water (0.1% formic acid); flow rate of 0.2 ml/min. The column oven temperature was 45°C. The volume of the injection was 1 μL. Mass spectrometry analysis was performed using an electrospray ionization (ESI) source in positive and negative ion scanning mode. The parameters of the ESI source were as follows: sheath gas flow rate (arbitrary unit, Arb) was 35; aux gas (Arb) was 10; pray voltage was 3500 V; heated capillary temperature, 320°C; the analysis was carried out using a scan from 70 to 1,050 m/z. Scanning mode: full-scan mass spectra were set as follows: resolution, 70000; isolation window 70–1050 m/z, and data-dependent MS2 (dd-MS2) was set as follows: the resolution was 17500; normalized collision energy (NCE) was set at 20, 40, 60.

Six-week-old male C57BL/6 mice were purchased from the Animal Research Laboratory of Guangdong Province (Guangzhou, China). The present study was conducted after approval by the Research Institute of the Animal Protection and Use Committee of Guangdong Provincial Hospital of Chinese Medicine [SCXK(Yue) 2021047]. After adaptive feeding, mice were randomly assigned to five groups (n = 6-7 per group): the control group (Control), the model group (Model), the low-dose of CSP intervention group (CSPL), the medium-dose of CSP intervention group (CSPM) and the high-dose of CSP intervention group (CSPH). The experimental protocols are as follows. Mice were fed a high-fat high-fructose diet (HFHFD, 40% kcal fat content, 20 kcal% fructose, and 2% cholesterol) for 16 weeks plus chronic immobilization stress (CIS, placed mice in a body-fit sized cylinder for 2 h per day for 2 weeks) to construct a NAFLD model. The mice in the control group were fed a normal diet only (13% kcal fat content). The daily dose of CSP decoction for human adults is 1.05 g/kg/d, and the equivalent dose in mice is 9.46 g/kg/d. The equivalent dose was selected as the medium dose; the high dose was doubling of the medium dose, while the low dose was half the medium dose.

Mice were fasted for 6 h, and fasting blood-glucose (FBG) levels were examined by using a glucometer (Countour TS, Bayer, Parsipanny, NJ, United States). Blood lipid profiles, including total cholesterol (TC), triglycerides (TG), high-density lipoproteins (HDL-C) and low-density lipoproteins (LDL-C), and TC and TG contents in the liver were all measured using commercial kits (A111-2-1 for TC, A110-1-1 for TG, A112-2-1 for HDL-C and A113-2-1 for LDL-C, JianCheng Bioengineering Institute, Nanjing, China). The serum levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) were measured by AST and ALT kits (C009-2-1 and C010-2-1, respectively, JianCheng Bioengineering Institute). The concentration of TNFα in the liver was examined by ELISA (EK1352, Signalway Antibody, Nanjing, China). All of the operations were performed strictly in accordance with the manufacturer’s protocols.

Fresh liver samples were collected from the mice and fixed with 4% paraformaldehyde overnight and embedded in paraffin. Small sections from the liver (4 μm thick) were dewaxed with xylene, rehydrated in a graded ethanol series and stained with hematoxylin and eosin (H&E). The frozen liver samples were sliced (8 μm thick) and stained with Oil Red O (ORO) staining solution (G1620, Solarbio, Beijing, China). Nuclear factor κB (NF-κB) (1:300, 8242, CST, United States), sterol regulatory element-binding protein-1(SREBP-1) (1:300, sc-365513, Santa Cruz Biotechnology, United States) and peroxisome proliferator activated receptor (PPARγ) (1:200, sc-7273, Santa Cruz Biotechnology, United States) expression and localization in the liver sections of mice were investigated using immunohistochemical staining. After dewaxing with xylene and rehydration, the liver sections were stained using the EliVision™ plus immunohistochemical staining kits (MXB Biotechnologies, Fujian, China) following the kit’s instruction. The staining was visualized by using a microscope (Olympus, Tokyo, Japan).

Total mRNA was isolated from fresh liver samples using RNAiso Plus (AKF0722A, Takara, Otsu, Shiga, Japan) and reverse-transcribed into cDNA using an Evo M-MLV RT kit (AG11728, Accurate Biology, Hunan, China) according to the manufacturer’s protocol. SYBR Green Master Mix (Rox Plus) (AG11718, Accurate Biology, Hunan, China) was applied to quantify PCR amplification. Relative mRNA expression levels were normalized to the β-actin expression level. Real-time PCR results were calculated using the 2^−ΔΔCt method. The primer pairs used in this study were synthesized by Shanghai General Biotech Co., Ltd., The primer information is listed in Supplemental Table S1.

Total protein samples were isolated from tissue samples using RIPA lysis buffer (P0013B, Beyotime, Shanghai, China) containing both PMSF and phosphatase inhibitors for western blot analyses. A BCA Protein Assay Kit (23225, Thermo Fisher Scientific, Rockford, IL, United States) was used to determine the protein concentration. The protein samples were separated on the 10% or 12% SDS-PAGE gels, and then transferred to a polyvinylidene fluoride (PVDF, Millipore, MA, United States) membranes. After blocking with 5% skim milk, the PVDF membranes were incubated overnight at 4°C with the following primary antibodies: anti-TNFR1 (1:1000, 21574-1-AP, Proteintech, China), anti-phospho-NF-κB p65(Ser536) (1:1000, 8324, CST), anti-IL1β (1:1000, ab6722, Abcam, United States), IL6 (1:1000, bs-6309R, Bioss), iNOS (1:1000, bs-0162R, Bioss), anti-TNFα (1:1000, bsm-33207M, Bioss), anti-SREBP-1 (1:1000, sc-365513, Santa Cruz Biotechnology), anti-PPARγ (1:1000, bsm-4590R, Bioss) and β-actin (1:5000, AC026, ABclonal, China). After that, the PVDF membranes were incubated with the corresponding secondary antibodies. An ECL kit (Tanon, Shanghai, China) was used to visualize the chemiluminescence signals. Protein expression levels were quantified with ImageJ software. The housekeeping protein β-actin was utilized as loading control.

All data in this study are expressed as the mean ± s.d. One-way analysis of variance (ANOVA) followed by LSD or Dunnett’s T3 post-hoc test was applied for the results using the SPSS 25.0 statistical software package. Data for graphing were processed with GraphPad Prism 8.4.3 software. A p value < 0.05 was considered significant.

The phytochemical composition of CSP was evaluated using LC-MS/MS. The total positive and negative ion chromatograms of CSP aqueous extracts, such as quercetin, kaempferol, naringenin, isorhamnetin, and nobiletin, are shown in Supplementary Figure S1 and Supplementary Table S2. This mass spectrometry closely replicated previous findings (Han et al., 2021).

A total of 139 active compounds in CSP were selected from the TCMSP database according to our search criteria (Supplementary Table S3). After taking an intersection of the results of LC-MS/MS analysis, a total of 21 main compounds were collected (Table 1). Meanwhile, the TCMSP database and SwissTarget Prediction database were used to collect the prediction targets of the main compounds of CSP. A total of 244 predicted target genes were acquired after removing duplicated targets (Supplementary Table S4). We also identified a total of 2762 therapeutic targets for NAFLD in 3 different databases (1512 from GeneCard, 1243 from OMIM, and 7 from TTD) (Supplementary Table S5). Once the superfluous targets were eliminated, a total of 2666 known therapeutic targets for NAFLD remained. Ultimately, we obtained 130 overlapping targets that could be instrumental in treating NAFLD using the main compounds of CSP (Figure 1A and Supplementary Table S6). The “Compounds-Genes Network” was constructed using Cytoscape software to elucidate the likely mechanism of CSP acting on NAFLD (Figure 1B). According to network topological parameters, quercetin (MOL000098, degree = 97) exhibited the highest correlation with the target genes of NAFLD, and the rest were kaempferol (MOL000422, degree = 42), naringenin (MOL004328, degree = 27), nobiletin (MOL005828, degree = 25), isorhamnetin (MOL000354, degree = 23), and so on.

The 130 overlapping targets were uploaded to the STRING database to obtain information on predicted interactions. Then, the STRING data were imported into Cytoscape to devise a PPI network (Figure 2). There were 130 protein nodes and 2365 edges remaining in the PPI network after removing targets that lack protein structure and have no interaction with other proteins. Proteins are connected with other proteins; as proteins exhibit a higher degree, they play a more critical role in the central correlation. The top 10 proteins ranked by degree value were TNF, IL6, IL1B, AKT1, PPARG, JUN, TP53, VEGFA, MMP9, and CASP3.

GO and pathway enrichment analyses of 130 overlapping targets were conducted using the Metascape Database (Supplementary Table S7). The GO results showed 1633 biological process (BP), 92 cellular component (CC), and 146 molecular function (MF) terms in total. The top 10 BP, CC, and MF terms are shown in Figure 3A. The terms in BP include response to lipid, inflammation, hormone, etc. The terms in CC include membrane raft, transcription regulator complex, side of the membrane, etc. The MF terms included DNA-binding transcription factor binding, protease binding, cytokine activity, etc. Meanwhile, a total of 193 enriched pathways were selected by KEGG analysis. The top 20 significantly enriched pathways are shown in Figure 3B. The results showed that overlapping targets were mainly enriched in lipid and atherosclerosis pathways, pathways in cancer and chemical carcinogenesis-receptor activation, and the TNF signaling pathway. The results of GO and pathway enrichment analyses indicated that the mechanism of CSP in NAFLD may be highly related to the TNF signaling pathway.

Although, HFHFD treatment could induce hepatic steatosis in mice. Recent studies discovered that combination treatment of HFHFD plus CIS could promote chronic inflammation in the liver and faster development of NAFLD (Corona-Pérez et al., 2017). Therefore, we used this combination treatment to construct the NAFLD model in this study. As shown in Figure 4A, in response to the combination treatment of HFHFD plus CIS, the body weight of mice in the model group increased markedly during the period between 14 and 16 weeks compared with the control group. Meanwhile, CSP intervention inhibited the body weight increase caused by the combination treatment of HFHFD plus CIS in a concentration-dependent manner, as well as the Lee index ( Figure 4B ). As illustrated in ( Figures 4C,D,G,H ), the mice of the model group also exhibited a prominent fatty liver phenotype featuring a pale and enlarged liver, higher liver-to-body weight ratio, increased fat vacuoles shown in H&E staining, and increased red lipid droplets shown in Oil red O staining. The lipid profiles in the liver and in the serum and liver enzyme levels were all elevated in the model group (Figures 4E,F,I). In contrast, after 16 weeks of CSP treatment, the liver size, liver-to-body weight ratio, fat vacuoles, and red lipid droplets were distinctly decreased (Figures 4C,D,G,H). The lipid profiles in liver and blood and liver enzymes were also considerably decreased (Figures 4E,F,I). However, higher HDL-C levels were observed in both the model group and the CSPH group than in the control group. All of these results indicated that CSP had a positive effect on the development of NAFLD in mice fed a HFHFD plus CIS. CSP also decreased FBG levels in mice fed a HFHFD plus CIS (Supplementary Figure S2).

Inflammation is an important factor affecting the progression of NAFLD. The level of inflammation in NAFLD mouse liver tissue in response to CSP was examined using the RNA and protein levels of inflammatory cytokines, including TNFα, interleukin-6 (IL6), interleukin-1β (IL-1β), and inducible nitric oxide synthase (iNOS). Compared with the control group in liver tissue, TNFα, IL-1β, IL-6, and iNOS in the liver of the model group were all significantly increased at the mRNA level (Figure 5A). In contrast, the mRNA levels of hepatic TNFα, IL-1β, IL-6, and iNOS were decreased after CSP intervention (Figure 5A). Moreover, the protein levels of TNFα, IL-1β, IL-6, and iNOS exhibited a similar trend to the mRNA levels (Figures 5B,C).

SREBP-1 and PPARγ, both key transcriptional regulators in hepatic lipogenesis, were subjected to immunochemical staining of the liver in this study. Compared with the control group, a marked increase in the nuclear expression of SREBP-1 and PPARγ in the liver of the model mice and CSP treatment could reverse this tendency (Figure 6A). The protein expression exhibited similar trends to the immunochemical staining (Figures 6B,C). This was consistent with the prediction that PPARG and SREBF-1 have important roles in the PPI network of CSP in NAFLD. We also tested the mRNA expression levels of fatty acid synthases, including stearoyl-CoA desaturase 1 (SCD-1), fatty acid synthase (FASN), acetyl-CoA carboxylase (ACC), and ATP-citrate lyase (Acly). The mRNA expression levels of these fatty acid synthases in the CSP group were all tremendously decreased compared to those in untreated mice in the model group (Figure 6D). Collectively, CSP treatment decreased liver inflammation and suppressed hepatic fatty acid synthesis in mice fed a HFHFD plus CIS.

The TNF signaling pathway is one of the most well-studied inflammatory signaling pathways. In addition, the results of the KEGG pathway enrichment analysis indicated that the overlapping targets of CSP in NAFLD were significantly enriched in the TNF signaling pathway. To determine whether CSP treatment ameliorated liver inflammation by regulating the TNF signaling pathway. The liver TNFα contents were examined using an ELSA kit, and the results showed that TNFα secretion in the liver of the model group was markedly increased compared with that in the control group and decreased after CSP treatment (Figure 7A). TNF signaling-related factors, including TNFR1, its adaptor protein receptor-interacting serine-threonine kinase 1 (RIPK1), TNFR-associated death domain (TRADD), and its downstream transcription factor (NF-κB), were also assessed. Mice in the model group had higher hepatic mRNA expression of TNFR1 and TRADD than mice in the control group (Figure 7B). In contrast, the hepatic mRNA levels of TNFR1 and TRADD in the livers of the CSP groups were significantly reduced (Figure 7B). The protein expression level of TNFR1 in these groups was consistent with its mRNA expression (Figures 7C,D). In addition, compared with the model group, the protein levels of p-NF-κB exhibited a significant reduction in the three-dose CSP group (Figures 7C,D). Immunohistochemical staining in the liver of the CSP intervention groups showed that nuclear NF-κB was significantly reduced compared with that in the model group (Figure 7E). However, the mRNA levels of hepatic RIPK1 were unaffected by both the model group and the CSP groups (Figure 7B). These results indicated that lowering TNFα/TNFR1 expression and ligand availability was required for the therapeutic effect of CSP on NAFLD.