GP was extracted three times with 80°C boiled water for 2 h each time. The concentrated filtrate was added into the AB-8 macroporous resins and eluted with water. GPP was obtained by lyophilizing for 48 h with a freeze dryer.

The molecular weight of GPP fractions was determined by high-performance gel permeation chromatography (HPGPC) equipped with a Shimadzu HPLC system (Shimadzu, Kyoto, Japan) fitted with one BRT105-104-102 (8 × 300 mm) (BoRui Saccharide, Yangzhou, China) and a RID-10A detector (Shimadzu, Kyoto, Japan). Briefly, 5 mg/ml of GPP sample was injected and analyzed on the HPGPC system at a flow rate of 0.6 ml/min with 0.05 M NACl at 40°C and monitored with the differential refractive index (dRI). The sample was filtered using a 0.22-μm filter before injection.

The monosaccharide composition was analyzed by ion-exchange chromatography (IC) using a Dionex™ Carbopac™ PA20 column (ICS5000, Thermo Fisher, Waltham, USA). Briefly, 10 mg of GPP sample was dissolved in 10 ml of 3 M trifluoroacetic acid (TFA) at 120°C for 3 h. After evaporation, the solution was dissolved in 10 ml of H₂O, 100 μl was taken, and 900 μl of ddH₂O was added. After 12,000 rpm centrifugation for 5 min, the supernatant was applied to HPIC using H₂O, 15 mM NaOH, and 100 mM NaOAc as the mobile phase with a flow rate of 0.3 ml/min.

Male C57BL/6 9-week-old mice (n = 35) were purchased from the Shanghai Model Organisms Center (Shanghai, China). The mice were maintained in a specific pathogen-free (SPF) grade and under standard conditions of illumination (12-h light/dark cycle). After 1 week of adaptive feeding, the mice were randomly divided into five groups. The control group was fed with a methionine and choline-sufficient (MCS) diet (#519581, Dyets) and received intragastric administration of 0.5% sodium carboxymethylcellulose (CMC-Na) at the same time for 4 weeks, and the model group was fed with an MCD diet (#519580, Dyets) and received intragastric administration of 0.5% CMC-Na at the same time for 4 weeks. In the GPP treatment groups, the mice were fed with an MCD diet and received intragastric administration of either at 150 mg/kg of low dose (LGPP) or 300 mg/kg of high dose (HGPP). In the positive treatment group, the mice were fed with an MCD diet and received intragastric administration of 120 mg/kg of polyene phosphatidylcholine capsules (PPCs). The GPP and PPC powders were dissolved in 0.5% CMC-Na for intragastric administration, and an equal volume of 0.5% CMC-Na vehicle was given to the MCS and MCD mice. GPP (purity: 98.93%) was purchased from Shanghai Winherb Medical Technology (#200607, Shanghai, China). PPCs were purchased from Sanofi (#H20059010, Beijing, China).

After 4 weeks, all mice were sacrificed by cranio-cervical dislocation. The blood sample was collected by extracting the eyeball, and serum was separated after centrifugation (2,500 rpm, 15 min) of the blood for biochemical analyses. The livers were immediately removed, weighed, and washed with PBS. Following evisceration, the liver samples were dissected from the right lobe of the liver. All samples were collected and stored at -80°C for long-term storage.

All procedures performed in the study involving animals were in accordance with the ethical standards of and protocol approved by the Shanghai Model Organisms Center (Shanghai, China).

Hepatic total cholesterol (TC, #A111-1-1), triglycerides (TG, #A110-1-1), superoxide dismutase (SOD, #A001-3-2), malondialdehyde (MDA, #A003-1-2), and hydroxyproline (HYP, #A030-2-1) were measured according to the instruction, and all kits were purchased from Nanjing Jiancheng Bioengineering Institute (Nanjing, China). Serum TC, TG, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), alanine transaminase (ALT), and aspartate aminotransferase (AST) were measured using automated clinical chemistry analyzer (7020, Hitachi, Japan) in the Center for Drug Safety Evaluation and Research, Shanghai University of Traditional Chinese Medicine (Shanghai, China).

Liver tissues from the mice were fixed in 4% polyformaldehyde and embedded in paraffin wax. Sections were cut and stained with hematoxylin and eosin (H&E) and Sirius Red. Oil Red O staining was performed on frozen liver specimens. For the histological diagnosis of NASH, the NAFLD Activity Score (NAS) was evaluated by a pathologist using standards as previously described and NAS ≥5 served as a substitute indicator for the histological diagnosis of NASH (31). Images were quantified using Image-Pro Plus software 6.0 (Rockville, USA), three randomly chosen fields per sample. The percentages of Oil Red O staining and Sirius Red staining were calculated as the average e area of Oil Red O staining and Sirius Red staining relative to the total analyzed area.

To determine the structure and function profile of the gut microbial community, high-throughput sequencing was used to sequence the 16S ribosomal RNA gene v3–v4 regions of the gut microbiota in mouse colon contents. The genomic DNA of the gut microbiota was extracted using the E.Z.N.A.^® Stool DNA Kit (#D4015-04, Omega Bio-tek, Norcross, USA) from colon contents according to the manufacturer’s instructions. The concentration and purity of extracted DNA were determined with NanoDrop 2000 (Thermo Scientific, Waltham, USA). The DNA extract quality was checked on 1% agarose gel. The v3–v4 regions of the 16S rRNA gene were amplified using 338F (5′ ACTCCTACGGGAGGCAGCAG 3′) and 806R (5′ GGACTACHVGGGTWTCTAAT 3′) primers. The multiplexed amplicons were purified using the AxyPrep DNA Gel Extraction Kit (#AP-GX-250G, Axygen Biosciences, Union City, USA). Amplicon sequencing was performed with Illumina MiSeq PE3000 (Illumina, San Diego, USA) at HonsunBio Biotechnology Co., Ltd. (Shanghai, China). The QIIME platform provided a 16S rRNA gene−amplicon−based analysis (https://www.qiime.org). The sequences allowed sorting of the reads into operational taxonomic units (OTUs) by clustering 97% sequence similarity and aligned using the SILVA database. Taxonomic classification was conducted according to various taxonomic ranks (phylum, order, class, family, genus, and species). The percentage of each bacterial species was analyzed and plotted by the R programming language. Sequence data associated with this project have been submitted to NCBI Sequence Read Archive (Accession Number: SRP361247).

Total RNA was extracted from livers of MCS, MCD, and HGPP groups (n = 3). Liver samples were lysed using TRIzol reagent (#15596018, Ambion, Austin, USA), and RNA was isolated with chloroform, followed by isopropanol precipitation and 75% ethanol washing. The RNA was redissolved in RNase-free water, and integrity and purity were assessed using a NanoDrop spectrophotometer (Thermo Scientific, Waltham, USA). RNA-sequencing library preparation was generated using 3 μg of RNA as input material. A sequencing library was generated on the NovaSeq 6000 platform (Illumina, San Diego, USA) by Personalbio Technology (Shanghai, China). After sequencing, the reads were preprocessed with Cutadapt and removed with an alignment quality (<Q20). Gene read counts were estimated with HTSeq (0.9.1) statistics, and then differential expression analysis was performed using DESeq (1.30.0) with a threshold of |log2FoldChange| >1 and P value < 0.05. Heatmaps of gene expression were generated using the R language pheatmap (1.0.8) software package and were generated based on Fragments per Kilobase of exon model per Million mapped fragments (FPKM). Gene Ontology (GO) enrichment was performed using topGO based on the hypergeometric distribution method with P value < 0.05. GO biological processes (BPs) were chosen for displaying the clustering results. Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment was carried out by clusterProfiler (3.4.4) software, focusing on the significant enrichment pathway with P value < 0.05. Gene Set Enrichment Analysis (GSEA) for KEGG pathways was conducted using the GSEA website, and the gene sets were obtained from the MSigDB database (http://software.broadinstitute.org/gsea/msigdb). The GSEA enrichment plotted by R language gseaplot2 (1.2.0) offered the visualization by results with |NES|>1 and P adjust<0.05. Data were uploaded to the NCBI GEO database (GSE205974) and analyzed on the online platform Personalbio GenesCloud (https://www.genescloud.cn).

Total RNA was extracted from homogenizing liver tissue, and cDNA was synthesized using HiScript II Q RT SuperMix for qPCR Kit (Vazyme, Nanjing, China). qPCR was performed by using a Luna Universal qPCR Master Mix (#M3003L, New England Biolabs, Ipswich, USA), and all the primers are presented in Table S1 . GAPDH was used as the endogenous reference control. All primers were purchased from Sangon Biotech (Shanghai, China). The data were analyzed by 2^-ΔΔCt.

This section was performed in accordance with the former studies. Briefly, proteins from liver tissues were extracted by sonication in RIPA lysis buffer with phenylmethanesulfonylfluoride. After centrifugation at 12,000 rpm for 10 min at 4°C, the supernatant of the lysate was collected and the protein concentration was detected by the BCA method. The proteins of samples were fractionated through 8% SDS-PAGE and transferred onto NC membranes. After being blocked with 5% non-fat milk for 1 h at room temperature, the membranes were incubated with different primary antibodies overnight at 4°C. After washing with TBST three times (10 min each), the blots were incubated with secondary antibodies for 1 h. Finally, the signal was visualized using ECL reagent. Pro-Caspase-1 (#YT0652, 1:1,000), Caspase-1 (#YC0002, 1:1,000), β-actin (#YT0099, 1:5,000), and secondary antibody HRP* Goat Anti Rabbit IgG(H+L) (#RS0002, 1:10,000) antibody were purchased from ImmunoWay Biotechnology (Plano, USA). Western blot quantification was analyzed densitometrically using ImageJ software and normalized to β-actin.

Liver tissues were embedded in paraffin, and slides were cut from the paraffin tissue. Slides were heated with sodium citrate for antigen retrieval and then incubated with primary antibodies including TLR2 (#GB11554, 1:300, Servicebio, Wuhan, China), NLRP3 (#YT5382, 1:100, ImmunoWay Biotechnology), ASC (#YT0365, 1:200, ImmunoWay Biotechnology), Caspase-1 (#GB11383, 1:500, Servicebio), and IL-1β (#GB11113, 1:250, Servicebio) overnight at 4°C. Secondary antibody HRP-conjugated Goat Anti-Rabbit IgG(H+L) (#GB23303, 1:200, Servicebio) was incubated for 50 min at room temperature. After PBS washing, peroxidase substrate DAB (#K5007, DaKo, Copenhagen, Denmark) was used as chromogen. The slices were stained, differentiated, and sealed before visualizing and taking photos under the microscope. Images of IHC were quantified using Image-Pro Plus 6.0 and assessed based on the sum value of integrated optical density (IOD, Area × Intensity) (32, 33). The average numbers of IOD were counted from three randomly chosen fields.

Mouse lipopolysaccharide binding protein (LBP) ELISA kit (#PCDBN0177) and mouse IL-1β ELISA kit (#PCDBM0158) were provided by PC Biotech (Shanghai, China). The serum levels of IL-1β and LBP were determined using commercial kits according to the instruction manual based on sandwich ELISA.

Data are presented as median and interquartile range (IQR) using violin plots. Comparisons among multiple groups were performed using one-way analysis of variance (ANOVA) for normal distribution and Kruskal–Wallis test for non-normal distribution. One-way ANOVA was followed by a least significant difference (LSD) test. Comparisons between two groups were performed using unpaired t-test for normal distribution and Mann–Whitney test for non-normal distribution. P < 0.05 was considered statistically significant. Statistical tests used to compare conditions are indicated in figure legends. The violin plots indicated lower quartiles, median, and higher quartiles, from bottom to top. All statistical analyses were performed using GraphPad Prism 8.3.0 software (San Diego, USA) or SPSS 25.0 software (Chicago, USA).

The HPGPC profile showed that the molecular weight of GPP was 1.08 kDa ( Figure 1A ). IC analysis showed the presence of rhamnose (Rha), arabinose (Ara), glucosamine hydrochloride (GlcN), galactose (Gal), glucose (Glc), xylose (Xyl), and mannose (Man) in the ratio of 0.208:0.069:0.015:0.186:0.467:0.018:0.037 for GPP ( Figures 1B, C ).

To explore the potential effects of GPP on NASH, an animal experiment was conducted as shown in Figure 2A . Pathological results showed the phenotype of NASH including ballooning degeneration, steatosis, lobular inflammation, and fibrosis induced by MCD diet. GPP and PPC improved the hepatic steatosis, while HGPP showed a similar efficacy as the PPC group based on the measurement of positive area staining and NAS score ( Figures 2B–E ). Regarding hepatic lipid level, the MCD group showed significantly increased levels of hepatic TC and TG, while the HGPP group showed a significant decreased level of hepatic TC ( Figures 2F, G ). MCD also induced oxidative damage in the liver as reflected by the SOD and MDA indexes. GPP and PPC had a significant ameliorative effect on the hepatic MDA, while HGPP and PPC presented a better effect on the hepatic SOD ( Figures 2H, I ). Meanwhile, HYP assay revealed that both LGPP and HGPP could significantly decrease the liver collagen content, while PPC showed no statistical difference ( Figure 2J ). Furthermore, MCD diet significantly decreased the levels of serum TC, TG, LDL-C and HDL-C, which could be partially recovered by different doses of GPP and PPC, respectively ( Figures 2K–N ). In contrast to the significant downregulation in the PPC group, the level of ALT showed a downward tendency in the LGPP group (P = 0.13) but an upward tendency in the HGPP group (P = 0.07). Moreover, the differences were not statistically significant in the level of AST among five groups ( Figures 2O, P ). These results indicated that GPP could ameliorate the MCD-induced NASH phenotype including hepatic steatosis, oxidative stress, and fibrosis in part. In addition, HGPP showed superior efficacy when compared with LGPP, so the further mechanistical exploration was mainly focused on the HGPP group.

The gut microbiota composition was profiled by sequencing the v3–v4 areas of the 16S rRNA gene. The alpha-diversity reflected by Chao1 and Shannon indexes showed no statistical difference between MCD and HGPP groups ( Figures 3A, B ). Principal coordinate analysis (PCoA) based on Bray–Curtis distance showed a clear separation between three groups, and the HGPP group clustered closer to the MCS group. Differences in analysis of similarity (ANOSIM) were significant for the three groups (P = 0.002, ANOSIM R = 0.5341) ( Figure 3C ). In addition, we observed significant changes in the relative abundance at the phylum and top 20 genus levels ( Figures 3D, E ). LEfSe analysis between MCD and HGPP groups showed that Akkermansia and A2 were enriched in the HGPP group, while Clostridia_uncultured was enriched in the MCD group at the genus level ( Figure 3F ). Random forest ranked the top 15 genera which contribute to the discrimination of MCD and HGPP groups, the top 5 including Lactobacillus, A2, Clostridia_UCG-014_norank, GCA-900066575, and Akkermansia. ( Figure 3G ). The relative abundances of Lactobacillus, Akkermansia, A2, and Clostridia_uncultured were further compared between the MCD and HGPP groups. There was a trend (P = 0.09) for higher relative abundance of Lactobacillus in the HGPP group. Moreover, HGPP was found to have a significantly increased relative abundance of Akkermansia and A2 and a significantly decreased relative abundance of Clostridia_uncultured compared to MCD ( Figure 3H ).

To further understand the potential mechanism underlying the anti-NASH effect of HGPP, the hepatic transcriptional profiles were characterized among different groups of mice. First of all, principal component analysis (PCA) showed a distinct separation among MCS, MCD, and HGPP groups. HGPP treatment resulted in a clear separation in gene expression in the PC2 context of MCD ( Figure 4A ). Volcano plots showed 2,282 genes between the MCS and MCD groups (1,622 upregulated genes, 660 downregulated genes), 1,232 genes between the MCS and HGPP groups (805 upregulated genes, 427 downregulated genes), and 755 genes between the MCD and HGPP groups (132 upregulated genes, 623 downregulated genes) ( Figures S3A, B , Figure 4B ). Venn diagrams depicted the number of overlapping and different DEGs for three group comparisons ( Figures S3C–E ). GO enrichment analysis revealed the top 20 significantly enriched GO terms in the biological process, such as regulation of immune system, leukocyte activation, cell activation, inflammatory response, T cell activation, and cytokine production, among others, between the MCD and HGPP groups ( Figure 4C ). Several KEGG signaling pathways, including chemokine signaling pathway, osteoclast differentiation, platelet activation, leukocyte transendothelial migration, toll-like receptor (TLR) signaling pathway, and nod-like receptor (NLR) signaling pathway, were widely enriched between the MCD and HGPP groups ( Figure 4D ). To be able to systematize the gene expression results, we performed GSEA based on all annotated genes to determine the important signaling pathways. In total, there were 40 pathways with |NES|>1 and P adjust < 0.05 between the MCS and MCD groups, and there were 42 pathways with |NES|>1 and P adjust < 0.05 between the MCD and HGPP groups ( Tables S9, 10 ). The overlapping KEGG pathways between MCS vs. MCD and MCD vs. HGPP were screened, including natural killer cell-mediated cytotoxicity, hematopoietic cell lineage, TLR signaling pathway, NLR signaling pathway, B-cell receptor signaling pathway, Fc epsilon RI signaling pathway, and leukocyte transendothelial migration. A previous study has shown that transcriptional upregulation of the inflammatory pathway is typically initiated by pattern recognition receptors (PPRs) such as TLRs and NLRs (34). MCD consistently led to the activation of the TLR and NLR signaling pathways, showing the importance of these two pathways in the development of a murine MCD model of NASH ( Figures 4E, F ). This result was also found between the MCD and HGPP groups ( Figures 4G, H ). As in the analysis above, the liver gene expression profiles reflected altered gene expressions and signaling pathways when HGPP compared to MCD. The possible mechanism was associated with regulating TLR and NLR signaling pathways.

To investigate whether the anti-inflammation effect of HGPP treatment impacts TLRs and NLRs on NASH, we examined the potential molecular mechanism. Heatmaps of gene expression (FPKM) showed that compared with that in the MCS group, the expressions of TLR genes (Tlr1, Tlr2, Tlr4, Tlr6, Tlr7, Tlr8) (35), inflammasome activation genes (Nlrp1b, Nlrp3, Nlrc4, Naip1, Naip5, Naip6, Nod2, Card9, Aim2, Caspase1, Caspase4, Caspase12, etc.) (36–39), pro-inflammatory genes (Il-1β, Il-18rap, Tnf-α, etc.) (35), and immune-related genes (Cd14, Cd40, Cd80, Cd86, etc.) (40, 41) were upregulated in the MCD group. Conversely, the expressions of genes including Tlr1, Tlr2, Tlr4, Nlrp1b, Nlrp3, Naip1, Naip5, Naip6, Aim2, Caspase1, Nod2, Card9, Il-1β, Il-18rap, Tnf-α, Cd40, and Cd80 were downregulated from mice treated with HGPP ( Figures 5A, B ). Further qPCR revealed that the expressions of TLR2, NLRP3, ASC, IL-1β, and TNF-α genes enriched in TLR and NLR signaling pathways were significantly downregulated in HGPP compared to the MCD group ( Figures 5C–H ). Immunohistochemical staining and quantification results for TLR2, NLRP3, ASC, Caspase-1, and IL-1β showed that the number of IOD was decreased in HGPP-treated mice ( Figures 5I, M–Q ). The expressions of Pro-caspase-1 and Caspase-1 genes in the livers of the HGPP group were verified by Western blotting ( Figure 5J , Figure S4 ). We further investigated the serum level of LBP, and HGPP significantly reduced the LBP compared to the MCD group ( Figure 5K ). The serum level of IL-1β was significantly decreased in HGPP treatment ( Figure 5L ). The IHC quantification revealed the decreased IOD of TLR2, NLRP2, ASC, Caspase-1, and IL-1b in HGPP when compared with MCD ( Figures 5M–Q ). These results suggested that HGPP treatment exerts an anti-inflammation effect involved in the TLR2/NLRP3 signaling pathway.