All animal experiments were conducted in accordance with NIH guidelines and were approved by the Ethics Committee of Tongji Medical College (Wuhan, China). Male C57BL/6J mice aged 8 weeks were purchased from Tongji Medical College and were provided controllable circumstances (a temperature of 20°C–22°C, 12-h light/dark cycle, and unlimited access to food and water). For a duration of 24 weeks, the mice were provided either a chow diet or a high-fat diet (HFD). Ginkgetin was purchased from Climax Biotech Co., Ltd. (Chengdu, China). Vehicle (DMSO) or ginkgetin (10 mg/kg/day) was administrated intragastrically in the last 8 weeks. The mice were separated into three groups: The chow diet group, the HFD + vehicle group, and the HFD + ginkgetin group.

Fasting body weight (FBW) of all the mice was measured every 4 weeks. Liver weight (LW) was measured after the mice were killed. The indicators related to liver function (the levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) in the serum) and lipid metabolism (the levels of triglyceride and cholesterol in the serum and liver tissues) were measured by commercial kits (Servicebio Co., Ltd., Wuhan, China). The levels of TREM2 in the serum were measured by ELISA kits (Servicebio).

The liver samples were preserved in 4% phosphate-buffered paraformaldehyde once the mice were sacrificed. After being dehydrated, embedded in paraffin, the samples were sliced into sections of 5 μm thick. To assess liver fibrosis, the sections underwent staining with hematoxylin and eosin (HE) as well as Sirius red. Lipid accumulation in the liver was assessed by staining frozen sections with Oil red O. The images were acquired through light microscopy (Nikon, Japan). For immunofluorescence, the sections were incubated consecutively with primary and secondary antibodies and 4’,6-diamidino-2-phenylindole (DAPI). After quenching tissue autofluorescence, the images were acquired using fluorescent microscopy (Nikon). Supplementary Table S1 displayed the antibodies for immunofluorescence.

Equal proteins obtained from liver tissues were subjected to gel electrophoresis and subsequently transferred onto polyvinylidene fluoride (PVDF) membranes. Afterward, the membranes were exposed to primary and secondary antibodies. The protein bands were detected by a chemiluminescence system (Clinix Co., Ltd., China). Supplementary Table S1 displayed the antibodies for Western blot.

The liver tissues were put into the electron microscope fixation solution (Servicebio) immediately after isolation and cut into small blocks about 2 mm × 2 mm × 2 mm. The tissues were stored at room temperature for 4 h away from light and transferred to 4°C for storage. After dehydration, drying, and gold-coated treatment, the prepared specimens were detected by an SEM (SU8010, Hitachi, Japan).

We performed bulk RNA-Seq analysis on the liver tissues of mice with NASH treated with vehicle or ginkgetin (n = 3 per group). Trizol reagent (Servicebio) was utilized to extract total RNA from the liver tissues. mRNA with poly-A was isolated from total RNA and converted to complementary DNA (cDNA). RNA-Seq array was conducted on the Illumina Novaseq6000 platform. After getting the expression matrix, DESeq2 was used to calculate the difference between the two groups. Differently expressed genes (DEGs) were considered as fold change (FC) > 2 and adjust p < 0.05. Further analyses were performed using the corresponding R packages. The raw data of RNA-Seq were available in the GEO database with number GSE235797.

Liver non-parenchymal cells (LNPCs) were isolated from liver tissues of mice with NASH treated with vehicle or ginkgetin as previously described (Mederacke et al., 2015). Then LNPCs were used for scRNA-Seq by 10X Genomics Chromium system (n = 3 per group). Quality control was performed by Seurat (version 4.3.0), and only the single cells with the unique molecular identifier (UMI) between 250 and 5,000 and mitochondrial genes <5% were used for further analysis. The clusters were identified by marker genes according to the CellMarker database (Hu et al., 2023). Visual analysis was performed using R packages. The raw data of scRNA-Seq were available in the GEO database with number GSE235939.

MPI describes all the polarized states of macrophages. Based on scRNA-Seq, the MPI value of each macrophage was calculated according to a well-established model (M0: unstimulated; MI: stimulated with LPS and IFNγ; M2: stimulated with IL4 and IL13) (Li et al., 2019). A higher MPI value indicates that the macrophage is closer to a pro-inflammatory state of M1, and vice versa, a lower MPI value indicates an anti-inflammatory state of M2.

The data were shown as mean ± SD and the difference was compared by t-test. Statistical analyses were carried out using SPSS 23.0 (IBM, United States). p < 0.05 was considered an indicator of statistical significance.

The chemical formula of ginkgetin was shown in Figure 1A. We first evaluated the impacts of ginkgetin on hepatic steatosis in a NASH mouse model induced by HFD. Administration of HFD resulted in a clear increase in FBW, LW, and LW/FBW ratio, which was significantly improved by ginkgetin (Figures 1B–D). And ginkgetin-treated mice also showed reduced levels of hepatic and serum triglyceride and cholesterol (Figures 1E,F). Additionally, HE staining revealed that the hepatocyte steatosis and ballooning were markedly alleviated with reduced NAFLD activity score (NAS) after ginkgetin treatment. And Oil red O staining confirmed the reduced lipid accumulation in ginkgetin-treated mice (Figure 1G). Consistently, the protein expression of FASN and PPARγ, which are associated with lipid metabolism, were significantly downregulated by ginkgetin (Figure 1H).

Immunofluorescence of F4/80 showed a marked increase in macrophage infiltration in mice with NASH, indicating enhanced hepatic inflammation, which was reduced in ginkgetin-treated mice (Figure 2A). And HFD-induced fibrosis, shown by Sirius red, was also improved after ginkgetin treatment (Figure 2A). Moreover, these histological alterations were supported by the level of proteins associated with inflammation (TNFα and p65) and fibrosis (COL1A1), which were overexpressed in NASH mice and reduced after ginkgetin treatment (Figure 2B). In addition, ginkgetin-treated mice showed improved liver function, shown by decreased ALT and AST (Figure 2C).

To further investigate the process of ginkgetin-induced regression of NASH, we performed RNA-Seq analysis of liver tissues from NASH mice treated with vehicle or ginkgetin (n = 3 per group). The two groups could be clearly distinguished by principal component analysis (PCA) (Figure 3A). And 1,199 DEGs were observed, among which 406 were upregulated and 793 were downregulated by ginkgetin (Figures 3B, C). Gene set enrichment analysis (GSEA) showed that the signals associated with lipid metabolism, inflammation, and fibrosis were enriched and downregulated (Figure 3D; Supplementary Table S2). Consistently, the expression of related genes was downregulated by ginkgetin (Figure 3E).

scRNA-seq analysis was conducted to explore the cellular heterogeneity in LNPCs isolated from NASH mice treated with vehicle or ginkgetin (n = 3), which are essential in regulating the process of NASH. After quality control filtering, transcriptomes of 26,160 cells were obtained, including 11,754 cells from vehicle-treated mice and 14,406 from ginkgetin-treated mice. And UMAP dimensionality reduction analysis revealed that the cells were divided into ten clusters according to the marker genes, including endothelial cells, hepatocytes, macrophages, B cells, hepatic stellate cells (HSCs), T-cells, dendritic cells, cholangiocytes, NK cells, and Mast cells (Figures 4A–D). And the cell counts and percent were presented (Figures 4E, F). Remarkably, we observed that endothelial cells and macrophages were the most abundant clusters of LNPCs, accounting for 57.2% of the LNPCs. And 75.3% of endothelial cells were from ginkgetin-treated mice, whereas 76.0% of macrophages were from vehicle-treated mice. Additionally, ginkgetin-induced alteration in gene expression was mainly reflected in macrophages and HSCs with a marked reduction (Figure 4G). These data suggested that the impacts of ginkgetin on NASH may be associated with the reprogramming of macrophages, HSCs, and endothelial cells.

4,419 macrophages were retained in scRNA-Seq analysis, which were further divided into two subclusters, including Kupffer cells (KCs, high expression of Adgre1) and monocyte-derived macrophages (MDMs, high expression of Itgam) (Figures 5A, B). And ginkgetin induced a significant reduction in macrophages, especially in KCs (Figure 5C). In addition, we also assessed macrophage polarization using MPI based on scRNA-Seq. As expected, we observed a marked alteration toward an anti-inflammatory phenotype in both KCs and MDMs after ginkgetin treatment (Figure 5D).

Next, we analyzed the alteration of NASH-associated macrophages (NAMs), which are subcluster of KCs with high expression of Trem2 (25). NAMs emerge during NASH and play a protective role against NASH. And the level of NAMs is positively correlated with the severity of NASH. Consistent with the improvement of lipid accumulation in the liver, NAMs markedly decreased in ginkgetin-treated mice (Figure 5E). Additionally, both Western blot and Immunofluorescence revealed that the level of TREM2 was significantly downregulated by ginkgetin (Figures 5F, G). Meanwhile, we also measured the level of TREM2 in the serum, which is a circulated marker of NASH and has a positive correction with NAMs (Hendrikx et al., 2022). Accordingly, the elevated TREM2 in serum during NASH also decreased after ginkgetin treatment (Figure 5H).

HSC activation, marked by high expression of Acta2, is a core step for liver fibrosis (Higashi et al., 2017). scRNA-Seq analysis revealed a significant decrease in the expression of Acta2 in ginkgetin-treated mice (Figure 6A), which was supported by the Immunofluorescence of αSMA (encoded by Acta2) (Figure 6B). Then, we performed KEGG enrichment analysis of the DEGs specially derived from HSCs. In addition to the pathways associated with lipid metabolism, we also discovered that JAK/STAT pathway was enriched (Figure 6C). Increasing evidence has demonstrated that JAK/STAT pathway was essential for HSC activation, especially IL6/STAT3 pathway and IFNγ/STAT1 pathways (Gao et al., 2012; Deng et al., 2013; Su et al., 2015; Xiang et al., 2018; Marti-Rodrigo et al., 2020). Triggering IL6/STAT3 signaling enhances HSC activation and liver fibrosis, while IFNγ/STAT1 signaling exhibits opposite effects. GSEA revealed that ginkgetin induced a marked downregulation in IL6/STAT3 signaling, but no alteration in IFNγ/STAT1 signaling (Figure 6D). Furthermore, immunofluorescence showed an increase in pSTAT3 and DCN (mainly expressed by HSCs) with a close colocation in NASH mice, which was abolished after ginkgetin treatment (Figure 6E). These results suggested that IL6/STAT3 signaling may be associated with ginkgetin-induced inhibition of HSC activation and liver fibrosis.

Endothelial cells make up the largest cluster of LNPCs, which can be further divided into three subclusters, including liver sinusoidal endothelial cells (LSECs, high expression of Fcgr2b), per-portal endothelial cells (PPECs, high expression of Efnb1), and peri-central endothelial cells (PCECs, high expression of Wnt2) (Figure 7A). Cluster analysis revealed individual transcriptomic features in the three subclusters (Figure 7B). And in vehicle-treated mice, the main type of endothelial cells was PPECs, while LSECs became the dominant type after ginkgetin treatment (Figure 7C). Consistently, immunofluorescence of vWF showed that NASH resulted in a marked decrease of endothelial cells, which was restored to the normal level after ginkgetin treatment (Figure 7D). Furthermore, SEM analysis revealed that the number of LSEC fenestrae markedly reduced due to NASH, but increased in ginkgetin-treated mice (Figure 7E).