Zexie pieces (Alismatis Rhizoma, the rhizome of Alisma plantago-aquatica subsp. orientale (Sam.) Sam., origin from Fujian, China) was obtained from Kangmei Pharmaceutical Co., Ltd., while Baizhu pieces (Atractylodis macrocephalae Rhizoma, the rhizome of Atractylodes macrocephala Koidz., origin from Anhui, China) was purchased from Shanghai Leiyunshang Co., Ltd. Dr. JJ Hou had authorized the two herbs of ZXBZ decoction based on Chinese Pharmacopoeia (2020 edition, Volume I). According to the “Synopsis of Prescriptions of the Golden Chamber,” the ratio of Zexie to Baizhu is 5:2 (w/w). Thus, zexie pieces (600.0 g) and Baizhu pieces (240.0 g) in ZXBZ decoction were socked with water (16.0 L) for half an hour. Then, they were slightly boiled for 2 h. The filtrates were concentrated by reducing pressure at 45°C, and freeze-dried into the water extract, in which the yield was 30.0%. The powder of water extract was stored at −20°C. Before gavage administration in mice, the water extract (75 and 150 mg/ml) was dissolved in 0.5% sodium carboxymethyl cellulose (CMC-Na).

Six weeks old male C57/BL6 mice were purchased from Shanghai Laboratory Animal Co. (Shanghai, China) and fed in specific pathogen-free (SPF)-grade according to requirements of the Institutional Ethics Committee of Shanghai Institute of Materia Medica. The relative humidity was 30%–70%, the light/dark cycle was 12/12 h, and diet and drinking water were provided ad libitum. After 1-week adaptation and a 5-week gubra-amylin NASH (GAN) diet induction (rodent diets with 40 kcal% fat) (Primex or Palm Oil, Research Diets, United States) (Boland et al., 2019), the mice were distributed evenly into four groups according to body weights and ALT, besides normal control diet (NOD) mice were fed with regular diet and water (n = 10). Then the four GAN diet–induced groups were given vehicle (i.g.), obeticholic acid (30 mg/kg, i.g.), and ZXBZ decoction (750 and 1,500 mg/kg, i.g.) every day, respectively. After a feeding period of 17 weeks along with drug treatment of 12 weeks, the mice were sacrificed, and the liver and serum were collected for subsequent analysis.

Blood was collected every 4 weeks via the tail veins of mice since drug or vehicle treatment began when mice had been fed on GAN diet for 5 weeks. The levels of serum ALT, AST, TC, TG, and LDL-C were measured using an automatic Roche biochemical analyzer with Roche kits (ALT, AST, TC, TG, and LDL-C).

Fasting blood glucoses were determined after 6 hours of fasting using the OneTouch Select Simple^® glucose meter (Johnson & Johnson, United States). As for the oral glucose tolerance test (OGTT) and insulin tolerance test (ITT), the mice were fasted for 16 and 6 h, respectively, at the 11th and 13th week of GAN diet induction with the treatment of 1 g/kg glucose (i.g.) or 0.75 U/kg insulin (i.p.), . The glucose baseline levels were measured at 0, 15, 30, 45, 60, 90, and 120 min and calculated by the area under the curve (AUC). The homeostasis model assessment of the insulin resistance (HOMA-IR) level was performed at the 12th week of GAN diet induction and evaluated according to the formula HOMA-IR = [fasting plasma glucose (mmol/L) fasting plasma insulin (ng/ml)]/22.5. The insulin levels were examined using the ELISA kit (Beijing Solarbio Science & Technology Co., Ltd., Beijing, China).

The livers were removed after a 12-week drug or vehicle treatment when HFD induction lasted for 17 weeks and fixed in 10% neutral buffered formalin to dehydrate for 3 days and embedded in paraffin. Serial transverse sections of 3–4 μm were stained with hematoxylin and eosin. To evaluate the degree of NAFLD quantificationally, the liver lesions were observed under low-power microscopy and examined following the NAFLD activity score (Kleiner et al., 2005).

The left lobes of livers were embedded by optimal cutting temperature compound (OCT) and frozen in liquid nitrogen immediately, then, the frozen tissues were cut into lesions and laid flat on glass slides. After staying at room temperature for 5 min, the samples were washed with distilled water and 60% isopropanol twice. Then, the samples were dyed with oil red O working solution for 5 min and terminated with distilled water and redyed with hematoxylin. The samples were observed and screened under a low-power microscope.

At 1, 5, and 9 weeks, urine was collected over 24 h through metabolic cages. The acquired urine samples were stored at −80°C pending sample preparation (Khamis et al., 2017; Feng et al., 2019b). The urine samples were thawed at room temperature before the measurement. The supernatant was diluted with water. The dilution factor was determined by the creatinine level measured in the urine sample by the HPLC procedure according to the Ministry of Health of the People’s Republic of China (People’s Republic of China, 1996). The diluted samples were centrifuged at 14,000 rpm for 10 min at 4°C for the UPLC–MS analysis. The exact UPLC–MS method has been reported before (Feng et al., 2019b). In brief, an LTQ-Orbitrap Velos Pro hybrid mass spectrometer (Thermo Fisher Scientific Corp.) linked to an Ultimate 3000 UHPLC system was used to produce high-resolution mass spectra for metabolomic investigation. A Waters ACQUITY UPLC HSS T3 column (1.8 μm, 2.1 mm × 100 mm) with an online filter was used to separate the samples. The mobile phase consisted of solvent A [100% H₂O (0.1% formic acid)] and solvent B [100% acetonitrile (0.1% formic acid). The elution procedure is as follows: 0–1 min, 99% A; 1–3 min, 99%–85% A; 3–6 min, 85%–50% A; 6–9 min, 50%–5% A; and 9–10 min, 5% A. The flow rate remained constant at 0.5 ml/min. The autosampler and column were held at 4 and 40°C, respectively. The injection volume was fixed at 5 µl for all samples. The quality control (QC) sample was made up of an equal volume (10 μl) of each urine sample, and used to evaluate the peak intensity stability of metabolomics analysis. The QC sample was injected each after five samples.

HepG2 cells were plated in 6-well plates at a density of 1 × 10⁵ cells/ml and incubated in a humidified incubator at 37°C with 5% CO₂. After the confluence reached 80%–90%, the culture medium was changed to the serum-free medium and the cells were starved for 12 h before treatment. Then, the medium was replaced by 1 mM oleic acid (OA)-palmitic acid (PA) = 2:1 or 500 μM PA medium. The ZXBZ decoction was filtered through a 0.22 μm membrane, and diluted into different concentrations. The negative control was 0.2% BSA and (0.1% DMSO or 0.1% ddH₂O) containing medium, and the 10 μM Compound C and EX-527 were used to pre-treat the cells 1 h before ZXBZ intervention. After being co-incubated under different conditions for 24 h, the cells were lysed using 1% TritonX-100 or RIPA with 1% PMSF for the TG content assay or WB experiment, respectively.

Total RNA was isolated using the TRIzol reagent (Yeasen Biotechnology (Shanghai) Co., Ltd.) and a UNlQ10 RNA extraction column kit (Sangon Biotech, Shanghai, China). cDNA was reverse transcribed using the PrimeScript™ RT Master Mix (Takara, Shiga, Japan), and qRT-PCR was analyzed on an ABI 7500 Fast system (ABI, CA, United States) using the Hieff^® qPCR SYBR Green Master Mix (Yeasen, Shanghai, China). All results were normalized to the RPS18 expression and calculated using the 2−(ΔΔCt) method and the primers of qRT-PCT are provided in Table 1.

Mouse livers (20 mg) were lysed using RIPA lysis buffer (Beyotime, China) containing 1% cock-tail (Sigma-Aldrich, St Louis, MO, United States). Total protein lysates were separated on 10% SDS-PAGE gels and then transferred to PVDF membranes (Millipore, United States), and then were incubated overnight at 4°C with antibodies against sirtuin 1 (Sirt1), histone 3, phospho-AMPKα (Thr172), and AMPKα (Cell Signaling Technology, United States), PGC-1α, phospho-ACC1-S79, ACC, phospho-mTOR-S2448, SREBP-1c, PPARα (ABclonal Technology Co., Ltd., Wuhan, China), β-actin, and β-tubulin (Abcam, MA, United States). The membranes were washed and incubated with HRP-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories, Inc., United States) and then detected using an ECL Plus immunoblot detection system (Clinx, Shanghai, China).

SIMCA-P software (version 14.1, Umetrics AB) was employed for multivariate analysis. Principal component analysis (PCA) analysis was carried out to demonstrate the aggregation of QC samples. Bidirectional orthogonal projection to latent structures discriminant analysis (O2PLS-DA) analysis was performed to describe the different metabolic profiles of the NC, vehicle, ZXBZ-L, ZXBZ-H, and positive groups. Orthogonal partial least squares discriminant analysis (OPLS-DA) analysis was used to better investigate the metabolic difference between the NC group and the vehicle group. R2 and Q2 were calculated to evaluate the quality of the model, and 200 permutation tests were performed to test the model. Potential biomarkers were confirmed based on the p < 0.05 and VIP > 1 between the NC and vehicle groups.

The potential urine biomarkers were identified using the metabolite database HMDB (http://www.hmdb.ca/) based on the MS¹ and MS² information. The pathway analysis was performed using MetaboAnalyst (http://www.metaboanalyst.ca/).

GSEA analysis was performed on GSEA 4.2.0 (36, 37) using the Reactome database (https://reactome.org/). The results were visualized using the R code (version 4.0.2) developed by us in RStudio. Paths were sorted by NES (NES < −1 or NES > 1 was considered significantly enriched).

The data from animal experiments are presented as the mean ± SEM, and in vitro experimental data are presented as the mean ± SD. Student’s t-tests were used to compare two groups. Comparisons among multiple groups were made with a one-way analysis of variance (ANOVA). A value of p < 0.05 was considered statistically significant.

The 94% chemical composition of ZXBZ decoction (Batch 20191129) was identified with four assays (Figure 1A). First, the contents of monosaccharides and oligosaccharides were determined using the high-performance liquid chromatography-diode array detection (HPLC-DAD) method (Supplementary Material 1), in which fructose accounted for 21.26%, glucose for 9.97%, and sucrose for 32.44%. The total content of the three saccharides had reached 63.7% of the aqueous extracts. Second, the contents of polysaccharides were analyzed using the phenol–sulfuric acid method (Chen et al., 2019), in which their percentage in the aqueous extract composition was 19.2%. Third, the varieties and contents of amino acids were assayed using the AccQ•Tag pre-column derivation method (Boogers et al., 2008). Fourteen amino acids were detected at a total content of 3.36%, among which the highest amino acid was arginine (1.37%). Fourth, a total of five nucleotides were quantified (Supplementary Material 2) using the HPLC-UV method and their total content was 0.24%. At last, the water content of ZXBZ aqueous extracts was determined using USP 40 <921> method III (rtf), in which the result was 6.93% (Figure 1B). The representative total ion chromatogram of ZXBZ was also characterized (Supplementary Figure S1) and 33 compounds were identified (Supplementary Table S1).

Given the regulatory role of ZXBZ decoction as a classical formula in the hepatic fluid metabolism disorder, we wondered whether it can modulate other metabolic pathways of the liver, especially glucose and lipid metabolism. Thus, the GAN diet–induced NAFLD mouse model was used, where C57 BL/6 mice were pre-fed with the GAN diet for 5 weeks and administered ZXBZ decoction by oral gavage daily for 12 weeks (Figure 2A). The GAN diet is a commonly used diet to induce NAFLD/NASH models, which shares similar characteristics with NAFLD/NASH patients in the respects of histopathology, transcription, and metabolism (Hansen et al., 2020; Radhakrishnan et al., 2020). In this experiment, we selected obeticholic acid (OCA), a candidate for NAFLD treatment in phase III clinical trials (Ratziu et al., 2016), as the positive control for the therapeutic effects. As shown in Figure 2B, the body weights of the vehicle group increased gradually with GAN diet feeding when compared with that of the NC group fed by the normal diet, while the body weights were obviously lowered no matter in the positive group or the low and high dosages of ZXBZ treated (ZXBZ-L and ZXBZ-H) groups than those in the vehicle group. Moreover, the hepatosomatic indexes of OCA, ZXBZ-L, and ZXBZ-H groups and liver weights of OCA and ZXBZ high groups were also markedly lower than vehicles after 12 weeks of continuous administration (Figure 2C).

To assess the effects of ZXBZ decoction on the liver function, the serum biomarkers, aspartate aminotransferase (AST), and alanine aminotransferase (ALT), were detected. Compared to the vehicle group, the ZXBZ-L and ZXBZ-H groups both exhibited lower ALT and AST levels since the fourth week of drug treatment, which was continued until the end of the experiment (Figures 2D,E). Afterward, the pathological changes in liver tissues from different groups were observed by H&E staining and quantified by NAFLD activity score (NAS). The vehicle group with a NAS of 5 had significantly hepatic steatosis, ballooning, and lobular inflammation compared to NC and reached the threshold for the diagnosis of NASH (Brunt et al., 2011). The high dosage ZXBZ treatment could alleviate hepatic steatosis and ballooning of hepatocytes obviously and relieved the lobular inflammation slightly without significance (Figures 2F,G).

The aforementioned results showed that ZXBZ decoction could reduce the levels of ALT and AST, relieve hepatic steatosis, and ballooning degeneration significantly in the GAN diet–induced NAFLD model.

To monitor the metabolic changes, the urine metabolites were collected monthly and studied. The PCA score plot showed good stability and feasibility of the detection method due to the aggregation of points in the QC group (Supplementary Figure S2A). As illustrated by the O2PLS-DA score plot (Supplementary Figure S2B), there was an excellent separation among the NC and vehicle groups, indicating that the urine metabolites in NAFLD mice were significantly changed. The ZXBZ-H and ZXBZ-L groups were close to each other and had similar metabolic phenotypes. In addition, the ZXBZ-H group could be separated from the vehicle group, showing the benefits of ZXBZ at the metabolic level.

An OPLS-DA model was established to search for potential biomarkers of NAFLD between the NC and vehicle groups (Figure 3A and Supplementary Figures S2C,D). In addition, the data were categorized according to different time points. The parameters of the model (the R2Y and Q2 values should be near to 1, indicating a good ability to forecast) were acceptable, indicating good prediction and reliability. Furthermore, the models were validated using permutation tests (n = 200). Variables with VIP values > 1 and p values < 0.05 were identified as significant endogenous biomarkers. The candidate ions were tentatively identified by searching the MS and MS/MS fragments using the online library (http://www.hmdb.ca/). Finally, 60 metabolites were identified as potential biomarkers in the NAFLD model. The results of the detailed identification information and shifting trends of biomarkers are shown in Supplementary Table S1. Overall, 30 metabolites were stably present in all three sampling occasions, and the reversal effect was reflected in the ZXBZ administration groups (Figure 3B and Supplementary Figure S2E).

To further investigate the metabolic pathways disturbed in NAFLD, an online tool, MetaboAnalyst 5.0 software (www.metaboanalyst.ca/), was used and 16 related pathways were found to be affected (Figure 3C). Among them, beta-oxidation of very long-chain fatty acids and oxidation of branched-chain fatty acids were the two most important pathways in the urine metabolite pathways, including L-carnitine and L-acetylcarnitine. As shown in Figure 3D, L-carnitine and L-acetylcarnitine decreased in the vehicle group and increased after ZXBZ treatment. Moreover, the AUC of the two metabolites were 0.9733 and 0.8978, both sensitive to being a biomarker, especially L-carnitine. Moreover, L-carnitine and L-acetylcarnitine were identified using available standards (Supplementary Figure S3).

To study the underlying mechanisms of ZXBZ’s hepatoprotective functions thoroughly, high-throughput RNA-seq technology was used to profile liver transcriptomes of vehicle, NC, and ZXBZ-H groups, respectively (n = 4/group), and the data were analyzed on the free online platform of the Majorbio Cloud Platform (http://www.majorbio.com/) and GSEA (https://www.gsea-msigdb.org/gsea/). The overall significantly differentially expressed genes (SDEGs) in NC and ZXBZ-H groups compared to the vehicle group were shown in the volcano chart (Figure 4A). The details of SDEGs related to NAFLD were shown in the cluster heatmap (Figure 4B), including the genes of lipid metabolism and transport such as Lpl, Accs, and Cd36, nuclear receptors such as Ppara, Pparg, and Rxra, electron transport chain, for instance, Atp4, Cox4, and Nmant, and bile acid metabolism-related genes, such as Cyp7a1. The SDEGs between ZXBZ-H and vehicle groups were further analyzed KEGG pathway and GO annotation, respectively, and the results demonstrated that ZXBZ decoction mainly regulated lipid metabolism, glycan biosynthesis and metabolism, and carbohydrate metabolism (Figures 4C,D). The gene set enrichment analysis (GSEA) was also performed to reveal the potential pathways and biological processes influenced by ZXBZ decoction (Figure 4E), and the results indicated that ZXBZ decoction might balance cellular energy states. The lipogenesis (FAs, cholesterol, and sterols) and hypoxia were enhanced in the vehicle group, whereas the genes involved in the mitochondrial function including ATP synthesis and mitochondrial biogenesis, nitric oxide metabolism, and nuclear receptor (PPAR and LXR) pathways were enriched in the ZXBZ-H group. Interestingly, these pathways are regulated by Sirt1. In addition, autophagy and mTOR pathways indicated the activation of the AMPK/mTOR pathway.

The transcriptomic data suggested that ZXBZ decoction improved energy metabolism, especially lipid metabolism, by affecting energy sensing and regulatory pathways.

Since the multi-omics results clued ZXBZ decoction might influence energy metabolism, particularly lipid metabolism, we assessed the lipid metabolism in NAFLD mice among different groups. The total cholesterol (TC) and low-density lipoprotein cholesterol (LDLC) in serum were elevated in response to the nutritious diet stimulation, while the high dose of ZXBZ could decrease the TC and LDLC (Figure 5A). The parallel results also appeared in the ratio of white adipose tissue (WAT)/bodyweight and WAT weights (Figure 5B). The oil red O staining (Figures 5C,D) and lower hepatic steatosis score (Figure 2E) demonstrated the reduced hepatic lipid accumulation and steatosis intuitively from histology. In addition, the cholesterol and triglycerides (TGs) in liver tissue were reduced in OCA and ZXBZ-L and H groups compared to vehicle group (Figure 5E).

Taken together, we can conclude that ZXBZ decoction could ameliorate lipid disorders in the aspects of reducing circulating lipid levels and hepatic steatosis.

We further confirmed the possible molecular mechanisms based on the remarkable phenotypic improvements in the lipid disorder of NAFLD models and multi-omics analysis results. As shown in Figure 6A, both dosages of ZXBZ treatment increased Sirt1 and activated AMPK, two nutrient sensors responding to NAD⁺/NADH and AMP/ATP, respectively in mammalian cells (Fulco and Sartorelli, 2008; Wang et al., 2011). As reported, Sirt1 activation can suppress de novo lipogenesis by inhibiting SREBP-1c and Chrebp, and promote fatty acid oxidation (FAO) by increasing PPARα and PGC-1α expressions (Cantó and Auwerx, 2009). The corresponding changes in ZXBZ treatment on these genes were consistent with the previous research works. Similarly, the downstream genes of PPARα, such as Cyp7a1, Cyp27a1, and Lxrα (Defour et al., 2012) were also increased under the condition of ZXBZ treatment, which indicated that ZXBZ might regulate bile acid metabolism, stimulate cholesterol clearance, and reverse cholesterol transport from peripheral tissues (Liang et al., 2021). Moreover, the activation of AMPK by ZXBZ consequently inactivated ACC and further reduced FASN (Figure 6B) and SREBP-1c, while increasing the Cpt-1α expression to reduce the fatty acid synthesis and promote the fully oxidation of long-chain and very long-chain FAs (Huang et al., 2017). We also found that AMPK activation downregulated Hmgcr transcription, meanwhile increasing Lpl mRNA levels (Figure 6C), which might reduce cholesterol synthesis and accelerate TG decomposition (Figure 6D) (Day et al., 2021; Trefts and Shaw, 2021). Furthermore, the phosphorylation of mTOR was inhibited in ZXBZ groups, suggesting autophagy was prompted after ZXBZ treatment, which could protect hepatocytes (Kim and Guan, 2021).

Altogether, ZXBZ decoction mainly regulated the energy-sensing network by influencing the Sirt1 expression and AMPK activation to govern lipid metabolism–related protein expressions and activations, consequently inhibiting lipogenesis and boosting lipids utilization in GAN diet–induced NAFLD mice.

Since glucose metabolism is another important assessment indicator in NAFLD, the oral glucose tolerance test (OGTT), HOMA-IR, and the insulin tolerance test (ITT) were measured on the sixth, seventh, and eighth week. As shown in Supplementary Figure S4, insulin resistance was not evident in this model, because there was no change in HOMA-IR and only an 18.65% increase in the AUC of the ITT between NC and vehicle groups. For OGTT with the obvious change between the NC and vehicle groups, we found ZXBZ decoction showed obvious improvement (Figure 7A). In addition, the serum fructosamine levels, which reflected blood sugar levels in the last one to 3 weeks, were also measured at the end of the experiment. Although serum fructosamine level had a slight change under diet induction, the administrated ZXBZ groups showed a weak but statistically significant reduced fructosamine level (Figure 7B). Meanwhile, the expression of gluconeogenesis genes G6pc and Creb which can be upregulated by AMPK activation were significantly hindered by ZXBZ treatment, which indicated that ZXBZ decoction may have regulatory effects in glucose metabolism (Figure 7C).

To sum up, the 12 weeks of ZXBZ administration improved the glucose tolerance, decreased serum fructosamine, and reduced gluconeogenesis gene transcriptions in the GAN diet–induced NAFLD mice model.