Therapeutic targeting of AREL1 in hepatic stellate cells attenuates MASH-related liver fibrosis Authors: Yang-Wen-Qing Zhang, Xiaoyu Pan, Minghe Zhang, Junyu Zhang, Hui Ren, Xinyu Zong, Xiaomian Li, Kequan Xu, Peng Xia, Tiangen Wu, Weijie Ma, Xi Chen, Chengming Qu, Wenzhi He, Yufeng Yuan Published: 2026 ABSTRACT Fibrosis resulting from metabolic-associated steatohepatitis (MASH) is increasingly recognized as the predominant form of liver fibrosis. Although the activation of hepatic stellate cells (HSCs) is essential for liver fibrosis, the mechanisms underlying HSC activation in MASH remain inadequately understood. Integrated analysis of large-scale single-cell and single-nucleus RNA sequencing data from human healthy and fibrosis samples reveals a distinct subpopulation of HSCs in MASH. AREL1 is a characteristic gene of this subpopulation and is uniquely upregulated in MASH-related fibrosis. HSC-specific knockout of Arel1 markedly attenuates liver fibrosis in MASH model male mice. Mechanistically, AREL1 is regulated by cholesterol and facilitates HSC activation through the AREL1-ILK axis, subsequently activating the PI3K-AKT signaling pathway. Moreover, therapeutic knockdown of Arel1 using vitamin A-modified lipid nanoparticles markedly ameliorates MASH-related liver fibrosis. Here, we show a unique mechanism underlying HSC activation in MASH-driven fibrosis and present the targeted knockdown of AREL1 in HSCs as a therapeutic avenue. Introduction Metabolic-associated fatty liver disease (MASLD) has become a leading cause of chronic liver disease globally, affecting up to 30% of the world population1. Among those affected, approximately one-quarter experience progression to metabolic-associated steatohepatitis (MASH), which is characterized by steatosis, hepatocellular ballooning, lobular inflammation, and fibrosis2,3. The degree of fibrosis is the most robust predictor of patient mortality and is thus a focal point for drug development efforts4. MASH can further progress to cirrhosis and even hepatocellular carcinoma (HCC) and is projected to surpass chronic hepatitis as the primary indication for liver transplantation. In 2024, resmetirom became the first drug to receive accelerated approval from the U.S. Food and Drug Administration (FDA) for the treatment of MASH, marking a significant milestone in the field5. However, clinical trials have shown that only approximately 25–30% of patients achieve a therapeutic response with this agent6. Therefore, identifying effective and actionable therapeutic targets for MASH-related fibrosis remains a critical unmet need. Liver fibrosis is essentially characterized by the excessive production of extracellular matrix (ECM)7, leading to fibrotic scarring of the liver. Various etiologies, including hepatitis B virus (HBV) infection, hepatitis C virus (HCV) infection, alcohol consumption, cholestatic injury, and MASH, can result in liver fibrosis. The activation of hepatic stellate cells (HSCs) into proliferative, migratory, and contractile myofibroblasts is crucial in this process, as these cells are a major source of ECM8. Canonical pathways such as the TGF-β and Notch signaling pathways have been reported to activate HSCs9,10. However, given the unique differences in the pathological microenvironment between MASH and other conditions leading to liver fibrosis, there may be undiscovered MASH-specific pathways that regulate HSC activation. Therefore, precise evaluation of the unique expression profiles and activation pathways of HSCs in the setting of MASH could reveal new therapeutic strategies for MASH-related liver fibrosis. In this work, we characterized the distinct transcriptional profile of HSCs in MASH through integrated analysis of single-cell RNA sequencing (scRNA-seq) and single-nucleus RNA sequencing (snRNA-seq) data derived from normal liver tissues and liver tissues with multiple types of fibrosis. Our analysis revealed a MASH fibrosis-specific subpopulation of HSCs that exhibited a highly activated phenotype. Apoptosis-resistant E3 ubiquitin protein ligase 1 (AREL1) was identified as a characteristic gene of this subpopulation and was specifically upregulated in MASH-related liver fibrosis. HSC-specific knockout of Arel1 markedly attenuated the liver fibrosis induced by both a western diet and a Gubra–Amylin NASH (GAN) diet in male mice. Notably, cholesterol specifically enhanced SREBP2-mediated transcription of AREL1. Mechanistically, AREL1 mediated the ubiquitination of ILK at the K27 residue, increasing its activity, which in turn activated PI3K–AKT signaling and promoted HSC activation. Furthermore, we developed therapeutic vitamin A-modified lipid nanoparticles encapsulating a siRNA targeting Arel1 (VA-siAREL1-LNP) which significantly ameliorated MASH-related liver fibrosis in male mice. In summary, our study reveals a unique pathway driving HSC activation in MASH and offers valuable insights into potential therapeutic strategies for MASH-related liver fibrosis. Results HSCs show distinct transcription signature in MASH fibrosis To investigate the unique expression profile of HSCs in MASH-related liver fibrosis, we collected 52 liver sc/snRNA-seq datasets from four independent sources (Fig. 1A and Supplementary Data 1), namely, liver tissues with no fibrosis (n = 21), with MASH-related fibrosis (n = 21), with alcohol-related liver fibrosis (n = 4), or with primary biliary cholangitis (PBC)-related fibrosis (n = 6). The datasets were integrated and clustered (Fig. 1B and Supplementary Fig. 1A), and HSCs were annotated using established markers, including ACTA2, TAGLN, PRKG1, LAMA2, and C7 (Fig. 1C). A total of 372739 cells were integrated, among which 9967 were identified as HSCs.Fig. 1HSCs show distinct transcription signature in MASH fibrosis.A Schematic overview of the analytical workflow. Created in BioRender. Yuan, Y. (2026) https://BioRender.com/0byydhm. B UMAP plot showing the integration of scRNA-seq and snRNA-seq datasets obtained from healthy human liver tissues and tissues of MASH-related, alcohol-related, and PBC-related liver fibrosis. C Marker genes used for the identification of major cell types. D Comparison of cell type proportions between the samples of healthy and fibrotic tissue, as well as across different etiologies of liver fibrosis. E UMAP plot showing the subclustering of HSCs across all samples. F Comparison of the abundance of HSC cluster 0 among the four experimental groups (n = 52 biologically independent human samples). G Heatmap showing the highly expressed genes in each HSC subcluster. H, I KEGG enrichment analysis of the highly expressed genes in HSC cluster 0 and 2. In (F), data are shown as the mean ± SD. Statistical significance was calculated using a one-way ANOVA with Dunnett’s post hoc test. In (H, I), data were analyzed using a one-sided hypergeometric test. Source data are provided as a Source Data File. HSCs show distinct transcription signature in MASH fibrosis. A Schematic overview of the analytical workflow. Created in BioRender. Yuan, Y. (2026) https://BioRender.com/0byydhm. B UMAP plot showing the integration of scRNA-seq and snRNA-seq datasets obtained from healthy human liver tissues and tissues of MASH-related, alcohol-related, and PBC-related liver fibrosis. C Marker genes used for the identification of major cell types. D Comparison of cell type proportions between the samples of healthy and fibrotic tissue, as well as across different etiologies of liver fibrosis. E UMAP plot showing the subclustering of HSCs across all samples. F Comparison of the abundance of HSC cluster 0 among the four experimental groups (n = 52 biologically independent human samples). G Heatmap showing the highly expressed genes in each HSC subcluster. H, I KEGG enrichment analysis of the highly expressed genes in HSC cluster 0 and 2. In (F), data are shown as the mean ± SD. Statistical significance was calculated using a one-way ANOVA with Dunnett’s post hoc test. In (H, I), data were analyzed using a one-sided hypergeometric test. Source data are provided as a Source Data File. HSCs were subsequently extracted for further dimensionality reduction and clustering. These cells were reclassified into eight distinct subclusters (Fig. 1D). Among the major subpopulations, cluster 2 was mainly present in healthy samples, but cluster 0 was predominantly found in fibrotic tissues (Fig. 1D, E). Notably, cluster 0 was markedly enriched in MASH samples. In addition, cluster 5 was mainly present in PBC fibrosis samples, cluster 6 was primarily associated with alcohol-related fibrosis samples, and cluster 3 was found in both alcohol- and PBC-related fibrosis groups. We further quantified the proportions of each HSC subcluster in different sample types. Cluster 0 was significantly more abundant in MASH samples than in samples from healthy controls and patients with other types of liver fibrosis (Fig. 1F), suggesting that this cluster may represent HSCs with a MASH-specific activation phenotype. The heatmap revealed that the cells in cluster 0 exhibited high expression of fibrosis-related genes such as COL3A1 and PDGFRA (Fig. 1G). Enrichment analysis of the highly expressed genes in each subcluster revealed that cluster 0 was enriched in genes associated with pathways significantly related to HSC activation, including the TGF-β signaling, PI3K–AKT signaling, focal adhesion, and EGF–EGFR signaling pathways (Fig. 1I), which have previously been linked to highly proliferative and activated states of HSCs11,12. In addition, cholesterol metabolism was identified as an enriched pathway. Cluster 2 was significantly enriched in pathways related to quiescent HSCs, such as retinol metabolism and the PPAR signaling pathway13(Fig. 1H), suggesting that cluster 2 represents the quiescent HSC subset. Interestingly, cluster 5 was associated with enrichment in the chemokine signaling and T-cell receptor signaling pathways (Supplementary Fig. 1B), suggesting a potential immunomodulatory role of this subpopulation. Immune activation has been reported to play a critical role in PBC-related fibrosis14. Moreover, cluster 6, which was strongly enriched in alcohol-related liver fibrosis samples, was enriched in genes in the alcoholic liver disease pathway (Supplementary Fig. 1B). Overall, we identified a highly activated HSC subpopulation with a cholesterol metabolism signature that was specifically enriched in MASH-related fibrosis. AREL1 is specifically upregulated in MASH liver fibrosis Given the strong association of cluster 0 with MASH, we sought to identify its characteristic genes through a sequential filtering strategy. First, we defined an initial gene set by intersecting genes highly expressed in cluster 0 with those specifically upregulated in MASH compared to healthy control, alcohol-related fibrosis, and PBC-related fibrosis groups, yielding 2727 candidates. Next, to prioritize genes consistently linked to MASH fibrosis, we cross-referenced this set with two independent transcriptomic datasets: our internal cohort (12 healthy vs. 16 MASH fibrotic livers; Supplementary Fig. 2A and Supplementary Data 2) and a public dataset (GSE: 135251). This refinement identified 20 genes that were consistently upregulated in MASH fibrosis across both datasets (Fig. 2A). Finally, we investigated the association of these 20 genes with fibrosis progression, revealing that only six (AREL1, SPSB1, METTL24, EVC, THBS2, and DLG5) were significantly correlated with the advancement of MASH-related fibrosis (Fig. 2B). UMAP visualization showed that these six genes were primarily expressed in HSC cluster 0 (Fig. 2C and Supplementary Fig. 2B). We evaluated the correlations between these genes and the enriched pathways associated with cluster 0 (Fig. 2H and Supplementary Fig. 2D). Among these, AREL1 exhibited the most significant associations with these pathways (Fig. 2D) and was among the top 100 highly expressed genes in HSC cluster 0 (Supplementary Data 3), suggesting its potential key role in the functions of this cluster. AREL1 is a HECT-type E3 ubiquitin ligase that has not previously been identified to have a role in liver fibrosis.Fig. 2AREL1 is specifically upregulated in MASH liver fibrosis.A, B Venn diagrams intersecting highly expressed genes in cluster 0 HSCs, MASH samples, and MASH samples from our center’s liver sequencing or public dataset (GSE135251). C UMAP plot showing AREL1 and DLG5 expression across HSC subclusters. D Heatmap of correlations between indicated genes and pathways enriched in cluster 0. E, F Raincloud plot showing AREL1 expression across seven HSC subclusters and among the four groups. G Linear regression analysis of AREL1 expression across different fibrosis stages. H Correlation analysis between AREL1 expression and the enriched pathways in cluster 0 (GSE135251, n = 216 biologically independent samples). AREL1 expression between healthy and MASH samples from our center (I) and public dataset (J) (n = 29(I) or n = 160 (J) biologically independent samples). K Correlation between AREL1 and fibrosis-related genes (GSE135251, n = 216 biologically independent samples). AREL1 expression in human and mouse samples (n = 3 (L, M) or n = 8 (N, O) biologically independent samples per group). P, Q Representative histological images of liver samples. AREL1 (red), α-SMA (green), DAPI (blue). Scale bar = 100 μm. R Representative histological images of healthy and non-MASH fibrotic liver samples. AREL1 (red), α-SMA (green), DAPI (blue). Scale bar = 100 μm. S AREL1 expression in normal and various fibrotic livers (n = 4 biologically independent samples per group). AREL1 expression in HSCs isolated from normal and different fibrotic liver (n = 3(T) or n = 6(U) biologically independent samples per group). CCl4: Carbon tetrachloride; DDC: 3,5-diethoxycarbonyl-1,4-dihydrocollidine. In (E, F), Data were analyzed using the Kruskal–Wallis test followed by Dunn’s post hoc test for multiple comparisons. In (G, H, K), Pearson correlation coefficients (r) are shown. In (I, J), Box plots: median (center), 25th/75th percentiles (box), whiskers extend to most extreme points within 1.5×IQR, points beyond are outliers. Statistical significance was calculated by DESeq2 (Wald test; FDR-adjusted P via Benjamini-Hochberg). In (N, O, U), data are presented as the mean ± SD. Statistical significance was calculated using one-way ANOVA with Dunnett’s post hoc test. Source data are provided as a Source Data File. AREL1 is specifically upregulated in MASH liver fibrosis. A, B Venn diagrams intersecting highly expressed genes in cluster 0 HSCs, MASH samples, and MASH samples from our center’s liver sequencing or public dataset (GSE135251). C UMAP plot showing AREL1 and DLG5 expression across HSC subclusters. D Heatmap of correlations between indicated genes and pathways enriched in cluster 0. E, F Raincloud plot showing AREL1 expression across seven HSC subclusters and among the four groups. G Linear regression analysis of AREL1 expression across different fibrosis stages. H Correlation analysis between AREL1 expression and the enriched pathways in cluster 0 (GSE135251, n = 216 biologically independent samples). AREL1 expression between healthy and MASH samples from our center (I) and public dataset (J) (n = 29(I) or n = 160 (J) biologically independent samples). K Correlation between AREL1 and fibrosis-related genes (GSE135251, n = 216 biologically independent samples). AREL1 expression in human and mouse samples (n = 3 (L, M) or n = 8 (N, O) biologically independent samples per group). P, Q Representative histological images of liver samples. AREL1 (red), α-SMA (green), DAPI (blue). Scale bar = 100 μm. R Representative histological images of healthy and non-MASH fibrotic liver samples. AREL1 (red), α-SMA (green), DAPI (blue). Scale bar = 100 μm. S AREL1 expression in normal and various fibrotic livers (n = 4 biologically independent samples per group). AREL1 expression in HSCs isolated from normal and different fibrotic liver (n = 3(T) or n = 6(U) biologically independent samples per group). CCl4: Carbon tetrachloride; DDC: 3,5-diethoxycarbonyl-1,4-dihydrocollidine. In (E, F), Data were analyzed using the Kruskal–Wallis test followed by Dunn’s post hoc test for multiple comparisons. In (G, H, K), Pearson correlation coefficients (r) are shown. In (I, J), Box plots: median (center), 25th/75th percentiles (box), whiskers extend to most extreme points within 1.5×IQR, points beyond are outliers. Statistical significance was calculated by DESeq2 (Wald test; FDR-adjusted P via Benjamini-Hochberg). In (N, O, U), data are presented as the mean ± SD. Statistical significance was calculated using one-way ANOVA with Dunnett’s post hoc test. Source data are provided as a Source Data File. Further characterization revealed that AREL1 expression was higher in MASH samples than in other sample types (Fig. 2F) and was predominantly expressed in cluster 0 (Fig. 2E). Expression patterns of the other five genes across HSC clusters and sample types are shown in Supplementary Fig. 2C. Moreover, AREL1 expression was positively correlated with fibrosis stage in the MASH-related fibrosis samples (Fig. 2G and Supplementary Fig. 3A). Analysis of sequencing data from our center confirmed that AREL1 expression was significantly elevated in MASH samples compared with healthy control samples (Fig. 2I), a finding validated in a public dataset (GSE: 135251) (Fig. 2J). Importantly, AREL1 expression was also significantly correlated with the expression of fibrosis marker genes (Fig. 2K). Western blot and qPCR analyses of human and mouse samples revealed progressively increased AREL1 expression during MASH progression (Fig. 2L–O and Supplementary Fig. 3B, C). Immunohistochemical and immunofluorescence staining confirmed this trend (Fig. 2P, Q and Supplementary Fig. 3D, E), with AREL1 highly colocalized with α-SMA. In contrast, AREL1 expression did not differ significantly between other types of fibrosis and normal tissues (Fig. 2R), nor in mouse fibrosis models of other etiologies (Fig. 2S and Supplementary Fig. 3F–H). Furthermore, analysis of primary HSCs isolated from various mouse fibrosis models confirmed that Arel1 was upregulated only in HSCs from MASH-related fibrosis models (Fig. 2T, U and Supplementary Fig. 3I). In summary, our results demonstrate that AREL1 is a characteristic gene of cluster 0 and is specifically upregulated in MASH-related liver fibrosis. AREL1 promotes HSC activation We next examined the role of AREL1 in HSC activation. We generated an AREL1 knockdown LX2 cell line (Fig. 3A) and found that AREL1 knockdown significantly reduced the expression of fibrosis-related genes (Fig. 3B, C and Supplementary Fig. 4A). Additionally, similar results were obtained in primary mouse HSCs following adenovirus-mediated Arel1 knockdown (Fig. 3D, E and Supplementary Fig. 4B). High migratory and proliferative activity is a hallmark of activated HSCs15. CCK8 assays revealed that AREL1 knockdown reduced the proliferative capacity of LX2 cells, whereas AREL1 overexpression enhanced proliferation (Fig. 3F). These changes were mirrored by alterations in the expression of the proliferation-related gene KI67 (Fig. 3G). Transwell and wound healing assays showed that AREL1 knockdown significantly impaired the migratory ability of LX2 cells, whereas AREL1 overexpression produced the opposite effect (Fig. 3H, I). Activated HSCs also exhibit pronounced mechanical contractility, a critical aspect of maintaining mechanical tension and the cytoskeletal structure within activated HSCs16. Collagen gel contraction assays demonstrated that AREL1 knockdown significantly attenuated the contractile capacity of LX2 cells, whereas AREL1 overexpression enhanced it (Fig. 3J). Additionally, we assessed the functional association between AREL1 and other fibrosis-related genes (DLG5, EVC, METTL24, SPSB1, THBS2). The results showed that compared with AREL1 knockdown alone, combined knockdown of AREL1 with any of these genes did not result in additional suppression of fibrosis markers or further reduction in cell proliferation and migration capacity (Supplementary Fig. 4C–E). In summary, we demonstrated that AREL1 markedly influences HSC migration, proliferation, and contractility.Fig. 3AREL1 promotes HSC activation.A Western blot analysis of LX2 cells with AREL1 knockdown. B, C Analysis of fibrosis-related genes expression in the AREL1 knockdown and control groups of LX2 cells (n = 3 biologically independent experiments). Analysis of fibrosis-related genes expression following adenovirus-mediated Arel1 knockdown in primary HSCs (n = 6(D) or n = 3(E) biologically independent experiments). F CCK-8 assay showing the effects of AREL1 overexpression and knockdown on LX2 cell proliferation (n = 6 biologically independent experiments). G qPCR analysis of KI67 expression following AREL1 overexpression and knockdown (n = 3 or 4 biologically independent experiments). H Wound healing assay showing the effects of AREL1 overexpression and knockdown on LX2 cell migration (n = 3 biologically independent experiments). Scale bar = 200 μm. I Transwell assay showing the effects of AREL1 overexpression and knockdown on LX2 cell migration (n = 3 biologically independent experiments). Scale bar = 200 μm. J Cell contraction assay showing the effects of AREL1 overexpression and knockdown on LX2 cells (n = 3 biologically independent experiments). Western blot results shown in (A) are representative of three biologically independent experiments. In (B, D, F–J), data are presented as the mean ± SD. Statistical significance was calculated using two-tailed unpaired Student’s t test. Source data are provided as a Source Data File. AREL1 promotes HSC activation. A Western blot analysis of LX2 cells with AREL1 knockdown. B, C Analysis of fibrosis-related genes expression in the AREL1 knockdown and control groups of LX2 cells (n = 3 biologically independent experiments). Analysis of fibrosis-related genes expression following adenovirus-mediated Arel1 knockdown in primary HSCs (n = 6(D) or n = 3(E) biologically independent experiments). F CCK-8 assay showing the effects of AREL1 overexpression and knockdown on LX2 cell proliferation (n = 6 biologically independent experiments). G qPCR analysis of KI67 expression following AREL1 overexpression and knockdown (n = 3 or 4 biologically independent experiments). H Wound healing assay showing the effects of AREL1 overexpression and knockdown on LX2 cell migration (n = 3 biologically independent experiments). Scale bar = 200 μm. I Transwell assay showing the effects of AREL1 overexpression and knockdown on LX2 cell migration (n = 3 biologically independent experiments). Scale bar = 200 μm. J Cell contraction assay showing the effects of AREL1 overexpression and knockdown on LX2 cells (n = 3 biologically independent experiments). Western blot results shown in (A) are representative of three biologically independent experiments. In (B, D, F–J), data are presented as the mean ± SD. Statistical significance was calculated using two-tailed unpaired Student’s t test. Source data are provided as a Source Data File. HSC-specific AREL1 knockout inhibits MASH fibrosis progression We next generated an AREL1HSC-KO mouse model in which Arel1 was specifically deleted in HSCs by crossbreeding Arel1-floxed mice with Lrat-cre mice (Supplementary Fig. 5A). AREL1HSC-KO and AREL1fl/fl male mice were fed a Western diet starting at 4 weeks of age and continuing until 24 weeks of age (Fig. 4A). No statistically significant difference in body weight was found between the two groups (Fig. 4B), but AREL1HSC-KO male mice exhibited reduced liver weight (Fig. 4C). Moreover, the serum ALT level and liver hydroxyproline content were lower in the AREL1HSC-KO group (Fig. 4D, E), while hepatic total cholesterol content remained unchanged (Fig. 4G). Histological staining revealed a slight reduction in hepatic steatosis in AREL1HSC-KO male mice (Fig. 4F, H). More importantly, collagen deposition and the presence of activated HSCs were significantly reduced (Fig. 4F, I–L), Fibrosis-related genes were also decreased in AREL1HSC-KO mouse liver tissues (Fig. 4M, N and Supplementary Fig. 5P), indicating a notable improvement in liver fibrosis. We further conducted similar studies in another MASH model (GAN diet) (Supplementary Fig. B). Consistent with the previous findings, the liver weight, serum ALT level, and liver hydroxyproline content were lower in AREL1HSC-KO male mice than in AREL1fl/fl male mice (Supplementary Fig. 5D–F). However, no significant differences in body weight or hepatic total cholesterol content were observed between the two groups (Supplementary Fig. 5C, H). Moreover, hepatic steatosis and fibrosis were significantly reduced in these male mice (Supplementary Fig. 5G, I–M). Fibrosis-related genes were also decreased in AREL1 knockout liver tissues (Supplementary Fig. 5N, O, Q). Collectively, these findings indicate that knocking down Arel1 in HSCs significantly ameliorates MASH-related liver fibrosis in male mice.Fig. 4HSC-specific AREL1 knockout inhibits MASH fibrosis progression.A Schematic diagram showing the procedure for establishing the western diet-induced MASH model in male mice. Created in BioRender. Yuan, Y. (2026) https://BioRender.com/ihkqhoi. Statistical analysis of body weight (B), liver weight (C), serum ALT level (D), and liver hydroxyproline content (E) in AREL1HSC-KO and AREL1fl/fl male mice (n = 8 biologically independent mice per group). F Representative images of HE staining, Oil Red O staining, PSR staining, Masson staining, and immunohistochemical staining for α-SMA and Col1a1. Scale bars = 200 μm. G Statistical analysis of hepatic cholesterol content (n = 8 biologically independent mice per group). Quantification of the Oil Red O staining area (H), PSR staining-positive area (I), Masson staining-positive area (J), α-SMA positive area (K) and Col1a1 positive area (L) (n = 8 biologically independent mice per group). Analyses of fibrosis related markers (n = 8(M) or n = 5(N) biologically independent mice per group). In (B–E, G–M), data are presented as the mean ± SD. Statistical significance was calculated using two-tailed unpaired Student’s t test. Source data are provided as a Source Data File. HSC-specific AREL1 knockout inhibits MASH fibrosis progression. A Schematic diagram showing the procedure for establishing the western diet-induced MASH model in male mice. Created in BioRender. Yuan, Y. (2026) https://BioRender.com/ihkqhoi. Statistical analysis of body weight (B), liver weight (C), serum ALT level (D), and liver hydroxyproline content (E) in AREL1HSC-KO and AREL1fl/fl male mice (n = 8 biologically independent mice per group). F Representative images of HE staining, Oil Red O staining, PSR staining, Masson staining, and immunohistochemical staining for α-SMA and Col1a1. Scale bars = 200 μm. G Statistical analysis of hepatic cholesterol content (n = 8 biologically independent mice per group). Quantification of the Oil Red O staining area (H), PSR staining-positive area (I), Masson staining-positive area (J), α-SMA positive area (K) and Col1a1 positive area (L) (n = 8 biologically independent mice per group). Analyses of fibrosis related markers (n = 8(M) or n = 5(N) biologically independent mice per group). In (B–E, G–M), data are presented as the mean ± SD. Statistical significance was calculated using two-tailed unpaired Student’s t test. Source data are provided as a Source Data File. Cholesterol upregulates AREL1 via SREBP2 Our previous research revealed that AREL1 is specifically upregulated in MASH-related liver fibrosis but not in other types of liver fibrosis. A distinctive feature of MASH-affected livers is the pathological environment enriched with fatty acids17. To investigate whether fatty acids regulate AREL1 expression, we stimulated primary mouse HSCs and LX2 with palmitic acid (PA); however, no significant increase in AREL1 expression was observed at either the protein or mRNA level (Supplementary Fig. 6A). Given the observed enrichment of cholesterol metabolism pathways in cluster 0 HSCs (Fig. 1H) and the strong correlation between AREL1 and cholesterol metabolic pathways (Fig. 2H), together with the fact that cholesterol crystal release from lipotoxic hepatocytes creates a cholesterol-rich microenvironment—a key metabolic hallmark of MASH18,19, we hypothesized that cholesterol may influence AREL1 expression. Notably, cholesterol significantly increased AREL1 expression in mouse primary HSCs and LX2 cells (Fig. 5A and Supplementary Fig. 6B). Given that cholesterol accumulation has been reported to activate HSCs20, we further found that AREL1 knockdown attenuated cholesterol-induced HSC activation (Fig. 5B).Fig. 5Cholesterol upregulates AREL1 via SREBP2.A Analysis of AREL1 expression in primary mouse HSCs or LX2 cells following cholesterol treatment (n = 3 biologically independent experiments). B Analysis of fibrosis-related genes expression (n = 3 biologically independent experiments). C Analysis of AREL1 promoter activity by dual-luciferase assay (n = 4 biologically independent experiments). D Silver staining of DNA pulldown from the AREL1 promoter. E Venn diagram of mass spectrometry‑identified proteins vs predicted AREL1 transcription factors. F qPCR analysis of AREL1 expression (n = 4 biologically independent experiments). G Correlation analysis between AREL1 and SREBP2 expression (n = 216 biologically independent samples). H Analysis of AREL1 expression following SREBP2 overexpression and knockdown (n = 3 or 4 biologically independent experiments). I Analysis of AREL1 promoter activity by dual-luciferase assay (n = 4 biologically independent experiments). J Analysis of AREL1 expression after SREBP2 knockdown with cholesterol treatment (n = 4 biologically independent experiments). K Analysis of AREL1 promoter activity by dual-luciferase assay (n = 5 biologically independent experiments). L ChIP-qPCR analysis of SREBP2 binding to the AREL1 promoter region (n = 4 biologically independent experiments). M Predicted SREBP2 binding sites in the AREL1 promoter (JASPAR database). N Analysis of wild-type and mutant AREL1 promoters activity by dual-luciferase assay (n = 5 biologically independent experiments). O ChIP-qPCR analysis of H3K27ac enrichment at the indicated genes (n = 3 biologically independent experiments). P ChIP-qPCR analysis of SREBP2 binding to the AREL1 promoter region (n = 3 biologically independent experiments). Q–S Raincloud plots showing SREBP2 and cholesterol metabolism pathway expression in HSCs isolated from four fibrosis models and SREBP2 expression from different HSC subpopulations. Chol: cholesterol. In (A–C, F, H–L, N–P), data are presented as the mean ± SD. Statistical significance was calculated using two-tailed unpaired Student’s t test (A, C, H–L, N) or one-way ANOVA with Dunnett’s post hoc test (B, F), or two-way ANOVA with Tukey’s post hoc test (O, P). In (Q–S), data were analyzed using the Kruskal-Wallis test followed by Dunn’s post hoc test for multiple comparisons. Source data are provided as a Source Data File. Cholesterol upregulates AREL1 via SREBP2. A Analysis of AREL1 expression in primary mouse HSCs or LX2 cells following cholesterol treatment (n = 3 biologically independent experiments). B Analysis of fibrosis-related genes expression (n = 3 biologically independent experiments). C Analysis of AREL1 promoter activity by dual-luciferase assay (n = 4 biologically independent experiments). D Silver staining of DNA pulldown from the AREL1 promoter. E Venn diagram of mass spectrometry‑identified proteins vs predicted AREL1 transcription factors. F qPCR analysis of AREL1 expression (n = 4 biologically independent experiments). G Correlation analysis between AREL1 and SREBP2 expression (n = 216 biologically independent samples). H Analysis of AREL1 expression following SREBP2 overexpression and knockdown (n = 3 or 4 biologically independent experiments). I Analysis of AREL1 promoter activity by dual-luciferase assay (n = 4 biologically independent experiments). J Analysis of AREL1 expression after SREBP2 knockdown with cholesterol treatment (n = 4 biologically independent experiments). K Analysis of AREL1 promoter activity by dual-luciferase assay (n = 5 biologically independent experiments). L ChIP-qPCR analysis of SREBP2 binding to the AREL1 promoter region (n = 4 biologically independent experiments). M Predicted SREBP2 binding sites in the AREL1 promoter (JASPAR database). N Analysis of wild-type and mutant AREL1 promoters activity by dual-luciferase assay (n = 5 biologically independent experiments). O ChIP-qPCR analysis of H3K27ac enrichment at the indicated genes (n = 3 biologically independent experiments). P ChIP-qPCR analysis of SREBP2 binding to the AREL1 promoter region (n = 3 biologically independent experiments). Q–S Raincloud plots showing SREBP2 and cholesterol metabolism pathway expression in HSCs isolated from four fibrosis models and SREBP2 expression from different HSC subpopulations. Chol: cholesterol. In (A–C, F, H–L, N–P), data are presented as the mean ± SD. Statistical significance was calculated using two-tailed unpaired Student’s t test (A, C, H–L, N) or one-way ANOVA with Dunnett’s post hoc test (B, F), or two-way ANOVA with Tukey’s post hoc test (O, P). In (Q–S), data were analyzed using the Kruskal-Wallis test followed by Dunn’s post hoc test for multiple comparisons. Source data are provided as a Source Data File. We further found cholesterol significantly enhanced AREL1 transcriptional activity (Fig. 5C). We subsequently identified 364 proteins interacting with the AREL1 promoter region and intersected them with the predicted transcription factors of AREL1, resulting in five candidate genes (Fig. 5D, E). Among these, SREBP2 knockdown resulted in the most pronounced reduction in AREL1 expression (Fig. 5F). Correlation analysis identified a specific positive association between AREL1 and sterol regulatory element-binding protein 2 (SREBP2) (Fig. 5G and Supplementary Fig. 6D), a key transcription factor that regulates cholesterol homeostasis21. We further found that SREBP2 significantly promotes both the expression and transcription of AREL1 (Fig. 5H, I and Supplementary Fig. 6C). We then assessed whether SREBP2 is involved in the regulation of AREL1 by cholesterol and found that SREBP2 knockdown blocked this regulatory effect (Fig. 5J, K). The results of ChIP assays confirmed that SREBP2 binds to the AREL1 promoter (Fig. 5L). Additionally, we used the JASPAR database to predict SREBP2 binding sites in the AREL1 promoter and created point mutations at these sites, finding that these mutations resulted in a significant reduction in SREBP2-mediated regulation of the AREL1 promoter (Fig. 5M, N). We next investigated the mechanism by which cholesterol promotes AREL1 transcription via SREBP2. Interestingly, we found that cholesterol stimulation did not increase SREBP2 protein levels or promote its nuclear translocation in activated HSCs (Supplementary Fig. 6E, F). This finding is consistent with previous reports indicating that the negative feedback regulation of cholesterol is lost in activated HSCs20. Moreover, cholesterol did not enhance the transcription of classical SREBP2 target genes, but specifically increased AREL1 transcription (Supplementary Fig. 6G). This prompted us to investigate whether cholesterol exerts a locus-specific regulatory effect at the AREL1 promoter. Given that cholesterol has been reported to alter chromatin state22–24, we examined classic histone modifications associated with transcriptional activation and repression. We found that cholesterol significantly increased H3K27ac modification in the AREL1 promoter region - a modification closely linked to chromatin accessibility (Fig. 5O and Supplementary Fig. 6H). Additionally, cholesterol enhanced the binding of SREBP2 to the AREL1 promoter (Fig. 5P). Collectively, these results suggest that cholesterol facilitates SREBP2-mediated transcription of AREL1 by increasing H3K27ac modification at the AREL1 promoter. We also found that SREBP2 expression was elevated in HSCs from MASH-related fibrosis patients (Fig. 5Q) and was higher in cluster 0 than in clusters 2–7 (Fig. 5R). Moreover, the cholesterol metabolism pathway was significantly upregulated in MASH-related fibrosis samples (Fig. 5S). Further analysis showed Srebp2 expression in HSCs isolated from male mice with MASH-related fibrosis was significantly higher than that in HSCs isolated from male mice with CCl4- or DDC-induced fibrosis (Supplementary Fig. 6I-J). In summary, our data show that cholesterol regulates AREL1 expression through SREBP2. AREL1 activates HSCs via PI3K-AKT through ILK ubiquitination To explore the mechanism by which AREL1 activates HSCs, we isolated HSCs from AREL1HSC-KO and AREL1fl/fl male mice fed a western diet and conducted transcriptomic sequencing (Supplementary Fig. 7A). Compared with controls, 272 genes were downregulated and 205 genes were upregulated in HSCs from AREL1HSC-KO male mice (Supplementary Fig. 7B). KEGG enrichment analysis revealed significant enrichment of pathways such as Focal adhesion and ECM–receptor interaction (Fig. 6A), indicating that the activation of HSCs was inhibited in AREL1HSC-KO male mice. Importantly, the PI3K–AKT signaling pathway, which plays a crucial role in HSC activation25, was also significantly enriched (Fig. 6A). Correlation analysis further showed a significant association between AREL1 expression and PI3K-AKT pathway activity (Fig. 6D). We hypothesized that AREL1 likely promotes HSC activation through the PI3K–AKT signaling pathway. We further detected significant reductions in the phosphorylation of PI3K and AKT in AREL1HSC-KO male mice (Fig. 6C), and similar changes were observed in LX2 cells (Supplementary Fig. 7C). Notably, inhibition of the PI3K–AKT signaling pathway blocked AREL1-mediated activation of LX2 cells (Fig. 6D and Supplementary Fig. 7D).Fig. 6AREL1 activates HSCs via PI3K-AKT through ILK ubiquitination.A KEGG enrichment of transcriptome from primary HSCs of AREL1HSC-KO and AREL1fl/fl male mice. B Correlation analysis between AREL1 expression and the PI3K–AKT–MTOR signaling score (GSE135251, n = 216 biologically independent samples). C PI3K and AKT phosphorylation in primary HSCs from AREL1HSC-KO and AREL1fl/fl male mice (n = 3 biologically independent mice per group). D Western blot analysis of fibrosis-related genes following AREL1 knockdown and treatment with the LY294002 (n = 3 biologically independent experiments). E Silver staining of the immunoprecipitated products. F Overlap of IP‑MS identified proteins with genes highly correlated with PI3K-AKT-MTOR signaling. G Western blot analysis of fibrosis-related proteins (n = 3 biologically independent experiments). H Correlation between ILK expression and the PI3K-AKT-MTOR signaling score (GSE135251, n = 216 biologically independent samples). I Co-IP assay assessing AREL1-ILK interaction in LX2 cells. J GST pulldown assay assessing AREL1-ILK interaction in LX2 cells. K Immunofluorescence colocalization of AREL1 (green) and ILK (red) in LX2 cells; scale bar = 20 μm. L Schematic of ILK protein domains. Created in BioRender. Yuan, Y. (2026) https://BioRender.com/ppyeav0. M Co-IP assay of AREL1 with full length ILK or truncation mutants. N In vitro ubiquitination assay showing AREL1-mediated ILK ubiquitination. O Analysis of the half-life of the ILK protein in LX2 cells overexpressing AREL1 and treated with CHX (n = 3 biologically independent experiments). P ILK protein expression in LX2 cells treated with MG132 or overexpressing AREL1 (n = 3 biologically independent experiments). Q, R In vitro ubiquitination identifying ubiquitin linkage type (O: only lysine present, R: missing lysine sites). S In vitro ubiquitination showing ILK ubiquitination sites. CHX: Cycloheximide. For (E, I, J, M, N, Q–S), data are representative of three independent experiments with similar results. In (A), data were analyzed using one-sided hypergeometric test. In (B, H); Pearson correlation coefficients (r) are shown. In (C), data are presented as the mean ± SD, Statistical significance was calculated using two-tailed unpaired Student’s t test. Source data are provided as a Source Data file. AREL1 activates HSCs via PI3K-AKT through ILK ubiquitination. A KEGG enrichment of transcriptome from primary HSCs of AREL1HSC-KO and AREL1fl/fl male mice. B Correlation analysis between AREL1 expression and the PI3K–AKT–MTOR signaling score (GSE135251, n = 216 biologically independent samples). C PI3K and AKT phosphorylation in primary HSCs from AREL1HSC-KO and AREL1fl/fl male mice (n = 3 biologically independent mice per group). D Western blot analysis of fibrosis-related genes following AREL1 knockdown and treatment with the LY294002 (n = 3 biologically independent experiments). E Silver staining of the immunoprecipitated products. F Overlap of IP‑MS identified proteins with genes highly correlated with PI3K-AKT-MTOR signaling. G Western blot analysis of fibrosis-related proteins (n = 3 biologically independent experiments). H Correlation between ILK expression and the PI3K-AKT-MTOR signaling score (GSE135251, n = 216 biologically independent samples). I Co-IP assay assessing AREL1-ILK interaction in LX2 cells. J GST pulldown assay assessing AREL1-ILK interaction in LX2 cells. K Immunofluorescence colocalization of AREL1 (green) and ILK (red) in LX2 cells; scale bar = 20 μm. L Schematic of ILK protein domains. Created in BioRender. Yuan, Y. (2026) https://BioRender.com/ppyeav0. M Co-IP assay of AREL1 with full length ILK or truncation mutants. N In vitro ubiquitination assay showing AREL1-mediated ILK ubiquitination. O Analysis of the half-life of the ILK protein in LX2 cells overexpressing AREL1 and treated with CHX (n = 3 biologically independent experiments). P ILK protein expression in LX2 cells treated with MG132 or overexpressing AREL1 (n = 3 biologically independent experiments). Q, R In vitro ubiquitination identifying ubiquitin linkage type (O: only lysine present, R: missing lysine sites). S In vitro ubiquitination showing ILK ubiquitination sites. CHX: Cycloheximide. For (E, I, J, M, N, Q–S), data are representative of three independent experiments with similar results. In (A), data were analyzed using one-sided hypergeometric test. In (B, H); Pearson correlation coefficients (r) are shown. In (C), data are presented as the mean ± SD, Statistical significance was calculated using two-tailed unpaired Student’s t test. Source data are provided as a Source Data file. We then performed immunoprecipitation of AREL1 and mass spectrometry analysis on the enriched proteins (Fig. 6E) and identified 149 enriched proteins. Intersecting these with genes highly correlated with the PI3K-AKT pathway yielded 16 candidates, three of which (ILK, NUDT5, and TRIP4) have been previously linked to PI3K signaling26–28. We further confirmed that ILK knockdown abolished the activation of the PI3K-AKT signaling pathway by AREL1 (Supplementary Fig. 7E). Consistent with this finding, correlation analysis revealed significant associations between the expression of ILK and the activity of the PI3K-AKT signaling pathway (Fig. 6H). ILK is a multifunctional protein that is involved in cell adhesion, cell–matrix interactions, and anchorage-dependent cell growth29 and has been widely associated with fibroblast activity30. Further studies revealed that ILK knockdown also blocked AREL1-induced HSC activation (Fig. 6G and Supplementary Fig. 7F). The interaction between AREL1 and ILK was confirmed by both endogenous and exogenous coimmunoprecipitation assays (Fig. 6I and Supplementary Fig. 7G), and GST pulldown assays further established that this interaction is direct (Fig. 6J). Immunostaining analysis revealed that AREL1 and ILK were predominantly colocalized in the cytoplasm (Fig. 6K). The architecture of ILK contains three distinct domains—an N-terminal ankyrin repeat region, a pleckstrin homology (PH)-like domain, and a C-terminal kinase-like domain29 (Fig. 6L). Deletion of the kinase domain abrogated its interaction with AREL1 (Fig. 6M). We further found that AREL1 significantly promotes the ubiquitination of ILK (Fig. 6N and Supplementary Fig. 7H). However, time-course analysis indicated that AREL1 does not alter the half-life of endogenous ILK (Fig. 6O and Supplementary Fig. 7I). Furthermore, ILK degradation was not influenced by treatment with the 26S proteasome inhibitor MG132 (Fig. 6P and Supplementary Fig. 7J). Thus, we hypothesized that AREL1 does not degrade ILK via the canonical ubiquitin–proteasome system. The results of subsequent ubiquitination assays demonstrated that AREL1 promotes K27-linked ubiquitination of ILK (Fig. 6Q). Mutation of K27 abolished ILK ubiquitination (Fig. 6R). K27-linked ubiquitination is a noncanonical type of ubiquitination known to regulate various properties of proteins, including activity and degradation. It is a key participant in innate immunity31 and has been implicated in the TGF-β signaling pathway32. We further used Phosphosite.org to predict potential ubiquitination sites on ILK (Supplementary Fig. 7K) and selected seven high-confidence sites for validation. Notably, AREL1-induced ubiquitination was reduced in the K154A and K191A ILK mutants (Fig. 6S). However, we found that ubiquitination does not affect the interaction between AREL1 and ILK (Supplementary Fig. 7L). Collectively, these results indicate that AREL1 binds to ILK to promote its K27-linked ubiquitination, thereby activating the PI3K–AKT pathway. AREL1 promotes HSC activation via ILK To further elucidate the role of AREL1-mediated ILK ubiquitination in HSC activation, we first assessed whether ILK is required for the pro-activating effects of AREL1. ILK knockdown abolished the effects of AREL1 on LX2 cell proliferation, migration, and collagen gel contraction capacity (Supplementary Fig. 8A–D). ILK forms a heterotrimeric complex known as IPP with PINCH and parvin. By coordinating bidirectional signaling between the ECM and intracellular compartments, the IPP complex governs critical cellular attributes—such as shape, migration, survival, differentiation, and gene expression—with significant implications for fibrotic pathogenesis29. We found that AREL1 increases the ubiquitination of ILK, thereby promoting its binding to PINCH and parvin (Supplementary Fig. 8E). Focal adhesion kinase (FAK) is a downstream protein of ILK and is a central regulator of focal adhesion turnover, playing a crucial role in force generation and cellular mechanotransduction33. We also observed that AREL1 promoted FAK phosphorylation, an effect that was abolished upon ILK knockdown (Supplementary Fig. 8F). Furthermore, ILK-dependent phosphorylation of PI3K also required FAK (Supplementary Fig. 8G), consistent with previous reports that FAK promotes PI3K phosphorylation34,35. Importantly, AREL1-mediated ILK ubiquitination enhanced ILK-driven FAK phosphorylation and subsequent PI3K activation, whereas mutation of the ILK ubiquitination site abrogated these effects (Supplementary Fig. 8H). In summary, AREL1 enhances ILK function through ubiquitination, thereby promoting PI3K phosphorylation and HSC activation. Therapeutic targeting of AREL1 attenuates MASH fibrosis As HSCs store the majority of the body’s vitamin A through the retinol binding protein receptor (RBPR) that takes up the RBP/VA complex from the bloodstream36, vitamin A-modified lipid nanoparticles have been proposed as a tool for targeted drug delivery to HSCs37. We developed vitamin A-modified lipid nanoparticles for targeted AREL1 knockdown (VA-siAREL1-LNP) (Fig. 7A). Transmission electron microscopy (TEM) revealed the morphology of these LNPs (Fig. 7B), and dynamic light scattering (DLS) was used to determine their particle size and zeta potential (Fig. 7C, D).Fig. 7Therapeutic targeting of AREL1 attenuates MASH fibrosis.A Schematic diagram of the VA-siAREL1-LNP synthesis pathway. Created in BioRender. Yuan, Y. (2026) https://BioRender.com/f3muqzy. B TEM images of VA-siAREL1-LNP; scale bar = 200 nm. C Particle size of VA-siAREL1-LNP. D Zeta potential of VA-siAREL1-LNP. E Western blot analysis of Arel1 expression in primary mouse HSCs following LNPs treatment. F Schematic representation of the procedure for LNPs treatment in a model of western diet-induced MASH. Created in BioRender. Yuan, Y. (2026) https://BioRender.com/5l07wih. G In vivo fluorescence imaging of LNPs in male mice. H Ex vivo fluorescence imaging of the liver and major organs 5 days after intravenous injection of LNPs. I VA-siAREL1-LNP distribution in liver by immunofluorescence: nuclei (DAPI, blue), HSCs (α‑SMA, green); Scale bar = 200 μm. J qPCR of Arel1 expression in primary cells from mice after LNPs treatment (n = 6 biologically independent mice per group). K Representative images of HE, Oil Red O, PSR, Masson, and immunohistochemical staining for α-SMA and Col1a1; scale bars = 200 μm. Statistical analysis of the body weight (L), liver weight (M), serum ALT level (N), liver hydroxyproline content (O), and liver hepatic cholesterol content (P) (n = 8 biologically independent mice per group). Quantification of the Oil Red O staining area (Q), PSR staining-positive area (R), Masson staining-positive area (S), α-SMA positive area (T) and Col1a1 positive area (U) (n = 8 biologically independent mice per group). In (J, L–U), data are presented as the mean ± SD. Statistical significance was calculated using one-way ANOVA with Dunnett’s post hoc test. Source data are provided as a Source Data File. Therapeutic targeting of AREL1 attenuates MASH fibrosis. A Schematic diagram of the VA-siAREL1-LNP synthesis pathway. Created in BioRender. Yuan, Y. (2026) https://BioRender.com/f3muqzy. B TEM images of VA-siAREL1-LNP; scale bar = 200 nm. C Particle size of VA-siAREL1-LNP. D Zeta potential of VA-siAREL1-LNP. E Western blot analysis of Arel1 expression in primary mouse HSCs following LNPs treatment. F Schematic representation of the procedure for LNPs treatment in a model of western diet-induced MASH. Created in BioRender. Yuan, Y. (2026) https://BioRender.com/5l07wih. G In vivo fluorescence imaging of LNPs in male mice. H Ex vivo fluorescence imaging of the liver and major organs 5 days after intravenous injection of LNPs. I VA-siAREL1-LNP distribution in liver by immunofluorescence: nuclei (DAPI, blue), HSCs (α‑SMA, green); Scale bar = 200 μm. J qPCR of Arel1 expression in primary cells from mice after LNPs treatment (n = 6 biologically independent mice per group). K Representative images of HE, Oil Red O, PSR, Masson, and immunohistochemical staining for α-SMA and Col1a1; scale bars = 200 μm. Statistical analysis of the body weight (L), liver weight (M), serum ALT level (N), liver hydroxyproline content (O), and liver hepatic cholesterol content (P) (n = 8 biologically independent mice per group). Quantification of the Oil Red O staining area (Q), PSR staining-positive area (R), Masson staining-positive area (S), α-SMA positive area (T) and Col1a1 positive area (U) (n = 8 biologically independent mice per group). In (J, L–U), data are presented as the mean ± SD. Statistical significance was calculated using one-way ANOVA with Dunnett’s post hoc test. Source data are provided as a Source Data File. First, we confirmed that VA-siAREL1-LNP significantly reduced Arel1 expression in primary mouse HSCs (Fig. 7E). We subsequently assessed the therapeutic effects of these LNPs in a mouse model of western diet-induced MASH (Fig. 7F). The distribution of VA-siAREL1-LNP in vivo was tracked by bioimaging, which showed predominant accumulation in the liver and a gradual dissipation over a 7-day period (Fig. 7G). Organ imaging confirmed that VA-siAREL1-LNP localized primarily to the liver, with some degree of retention in the spleen (Fig. 7H). Additionally, immunofluorescence staining on day five demonstrated strong colocalization of VA-siAREL1-LNP with HSCs (Fig. 7I). Using LNPs encapsulating siAREL1 as a control (siAREL1-LNP), we further demonstrated that VA-siAREL1-LNP significantly enhanced its uptake by HSCs (Supplementary Fig. 9A, B). qPCR analysis further confirmed the effective targeting and knockdown of Arel1 in HSCs (Fig. 7J). Safety assessments revealed no significant differences between the VA-LNP and VA-siAREL1-LNP groups and the PBS group in terms of complete blood counts and major organ histopathology (Supplementary Fig. 9C, D). From a therapeutic perspective, the VA-siAREL1-LNP group exhibited a reduced liver weight (Fig. 7M) and decreases in the serum ALT level and liver hydroxyproline content (Fig. 7N, O). However, no significant changes were observed in body weight or hepatic cholesterol levels (Fig. 7L, P). Pathological staining revealed a slight reduction in hepatic steatosis after VA-siAREL1-LNP treatment (Fig. 7K, Q) and, more importantly, significant reductions in liver fibrosis (Fig. 7K, R–U). The expression of fibrosis-related genes was also significantly reduced after VA-siAREL1-LNP treatment (Supplementary Fig. 10A–C). Additionally, we conducted studies in another mouse model of MASH (GAN diet) (Supplementary Fig. 10D). Consistent with the previous findings, VA-siAREL1-LNP treatment resulted in slight reductions in liver weight, serum ALT level and liver hydroxyproline content (Supplementary Fig. 10F–H), while body weight and hepatic cholesterol levels remained unchanged (Supplementary Fig. 10E, I). Treatment also alleviated hepatic steatosis and fibrosis (Supplementary Fig. 10J–O), accompanied by decreased expression of fibrosis-related genes (Supplementary Fig. 10P–R). Overall, our findings indicate that VA-siAREL1-LNP demonstrates effective therapeutic potential in mouse models of MASH-related liver fibrosis, suggesting substantial clinical application value. SnRNA-seq analysis of mouse livers after treatment We performed snRNA-seq on livers from male mice with western diet-induced MASH that were treated with VA-siAREL1-LNP. Our analysis revealed a reduced proportion of HSCs and a concurrent increase in the proportion of hepatocytes in the treatment group (Fig. 8B, C). Subsequent isolation and examination of HSCs revealed that VA-siAREL1-LNP significantly suppressed PI3K–AKT signaling in this cell population (Fig. 8D), consistent with our previous observations. KEGG enrichment analysis revealed that the downregulated genes in HSCs from the treatment group were associated with HSC activation-related processes, such as focal adhesion and ECM–receptor interactions (Fig. 8E). Conversely, the upregulated genes were enriched in pathways typically associated with quiescent HSCs, including the retinol metabolism and PPAR signaling pathways15 (Supplementary Fig. 11A), suggesting a reduction in HSC activation following treatment.Fig. 8SnRNA-seq analysis of mouse livers after treatment.A, B UMAP of integrated snRNA-seq datasets from livers of male mice in the VA-LNP group and the VA-siAREL1-LNP group. C Cell type proportion comparison between the two groups. D PI3K–AKT–mTOR signaling pathway activity in HSCs between the two groups. E KEGG enrichment analysis based on genes highly expressed in HSCs in the VA-siAREL1-LNP group. F, G UMAP plot showing the subclustering of HSCs across all samples. H Comparison of the abundance of HSC subclusters between the two groups. I Heatmap displaying the highly expressed genes in each HSC subcluster. J, K KEGG enrichment analysis based on the highly expressed genes in HSC cluster 0 and cluster 1. L–N Pseudotime trajectory analysis of HSCs. Each point represents a single cell, colored by pseudotime (L), subcluster identity (M), or treatment group (N). O Expression dynamics of fibrosis-related genes along the HSC pseudotime trajectory. P Expression levels of fibrosis-related genes at different pseudotime points. Q KEGG enrichment analysis based on the genes highly expressed in hepatocytes from the VA-siAREL1-LNP group. R Chord diagram showing ligand–receptor interactions between HSCs and hepatocytes. S Cell communication patterns that were increased (red) or decreased (blue) in the VA-siAREL1-LNP group compared with the control group. T Specific ligand–receptor interactions between cell types that were increased (red) or decreased (blue) in the VA-siAREL1-LNP group. In (D), data were analyzed using the two-sided Wilcoxon rank-sum test. In (E, J, K, Q), data were analyzed using one-sided hypergeometric test. In (T), data were analyzed using a permutation-based test implemented in CellChat. SnRNA-seq analysis of mouse livers after treatment. A, B UMAP of integrated snRNA-seq datasets from livers of male mice in the VA-LNP group and the VA-siAREL1-LNP group. C Cell type proportion comparison between the two groups. D PI3K–AKT–mTOR signaling pathway activity in HSCs between the two groups. E KEGG enrichment analysis based on genes highly expressed in HSCs in the VA-siAREL1-LNP group. F, G UMAP plot showing the subclustering of HSCs across all samples. H Comparison of the abundance of HSC subclusters between the two groups. I Heatmap displaying the highly expressed genes in each HSC subcluster. J, K KEGG enrichment analysis based on the highly expressed genes in HSC cluster 0 and cluster 1. L–N Pseudotime trajectory analysis of HSCs. Each point represents a single cell, colored by pseudotime (L), subcluster identity (M), or treatment group (N). O Expression dynamics of fibrosis-related genes along the HSC pseudotime trajectory. P Expression levels of fibrosis-related genes at different pseudotime points. Q KEGG enrichment analysis based on the genes highly expressed in hepatocytes from the VA-siAREL1-LNP group. R Chord diagram showing ligand–receptor interactions between HSCs and hepatocytes. S Cell communication patterns that were increased (red) or decreased (blue) in the VA-siAREL1-LNP group compared with the control group. T Specific ligand–receptor interactions between cell types that were increased (red) or decreased (blue) in the VA-siAREL1-LNP group. In (D), data were analyzed using the two-sided Wilcoxon rank-sum test. In (E, J, K, Q), data were analyzed using one-sided hypergeometric test. In (T), data were analyzed using a permutation-based test implemented in CellChat. We further partitioned the HSCs into four distinct subclusters (Fig. 8F). Among these subclusters, clusters 0 and 1 were the most abundant; cluster 0 was found predominantly in the VA-LNP control group, whereas cluster 1 was enriched in the VA-siAREL1-LNP group (Fig. 8G, H). Heatmap analysis revealed elevated expression of fibrogenic genes (e.g., Col1a1 and Col6a3) in cluster 0 (Fig. 8I), along with enrichment of the PI3K–AKT signaling, focal adhesion, and TGF-β pathways (Fig. 8J), indicating a highly activated HSC state. In contrast, cluster 1 was associated with enrichment in the retinol metabolism, PPAR signaling, and fatty acid degradation pathways (Fig. 8K), consistent with a quiescent HSC phenotype. Using Monocle2 for pseudotime trajectory analysis (Fig. 8L), we defined a unidirectional trajectory beginning with cluster 1 (quiescence-associated HSCs) and terminating with cluster 0 (activity-associated HSCs) (Fig. 8M, N). The VA-siAREL1-LNP samples were located primarily near the trajectory origin, whereas the VA-LNP control samples were positioned toward the terminus. Along this trajectory, we observed progressively increasing expression of fibrosis-related genes (Acta2, Col1a1, Col1a2, Col3a1, Tgfbr1, Spp1, Loxl1, and Pdgfrb) (Fig. 8O, Q and Supplementary Fig. 11B). Together, these findings demonstrate that VA-siAREL1-LNP treatment significantly suppresses HSC activation in vivo. Given the notable increase in the hepatocyte population after treatment, we next focused on hepatocyte-specific transcriptomic changes. Differential expression analysis revealed that the upregulated genes in hepatocytes from the VA-siAREL1-LNP group were enriched in the PPAR signaling and fatty acid degradation pathways (Fig. 8Q), suggesting an increase in lipid metabolism. This finding supports our earlier observation that VA-siAREL1-LNP reduced hepatic lipid accumulation in vivo. We then used CellChat2 to infer intercellular communication interactions between hepatocytes and HSCs. VA-siAREL1-LNP treatment substantially reduced the overall interaction strength between these cell types (Fig. 8R and Supplementary Fig. 11C). Pathway-level analysis revealed marked attenuation of key profibrogenic signaling pathways, including the COLLAGEN, PDGF, and SPP1 signaling pathways, with the most pronounced reduction observed in COLLAGEN signaling (Fig. 8S and Supplementary Fig. 11D). Specifically, inflammatory SPP1 signaling derived from hepatocytes was decreased after treatment. Conversely, signaling from HSCs to hepatocyte integrin receptors (e.g., Itga9+Itgb1 and Itga1+Itgb1) via Col6a3, Col4a2, Col1a2, and Col1a1 was significantly reduced (Fig. 8T). These pathways promote hepatocyte inflammation, disrupt hepatic structure and function, and perpetuate HSC activation38,39. In summary, these findings indicate that VA-siAREL1-LNP treatment attenuates pathological crosstalk between HSCs and hepatocytes and reduces proinflammatory signaling in hepatocytes. We next investigated the role of Arel1 in hepatocytes. During MASH progression, Arel1 expression increased in hepatocytes (Supplementary Fig. 12A, B); however, its knockdown failed to downregulate fatty acid metabolism-related genes in primary mouse hepatocytes, affecting only a subset of inflammatory genes (Supplementary Fig. 12C). Moreover, Arel1 knockdown did not alter lipid accumulation (Supplementary Fig. 12D), indicating that Arel1 suppresses MASH progression primarily through its function in HSCs. Discussion HSC activation is a well-established cornerstone of hepatic fibrogenesis. However, in conventional paradigms, HSC activation is often treated as a uniform endpoint across various etiologies. While transcriptional shifts in HSCs during the progression from healthy liver to MASLD and MASH have been mapped in prior studies40, a critical question remains unexplored: do HSCs acquire disease-specific phenotypes in response to fundamentally distinct microenvironments? The metabolic injury milieu of MASH differs profoundly from that of other fibrotic diseases, suggesting that HSC activation pathways may also be etiology dependent. In this study, our cross-disease comparative analysis demonstrates that the HSCs are not homogeneously activated; rather, it comprises distinct subpopulations that exhibit differential distribution across various types of fibrosis. Most notably, we identified a subset of HSCs in MASH-related fibrosis characterized by a highly activated phenotype and a gene expression signature enriched in cholesterol metabolism pathways. AREL1 was defined as a characteristic gene of this subpopulation and was found to be specifically upregulated in MASH-related fibrosis. Although prior findings have implicated AREL1 in inflammatory processes—such as exacerbating pulmonary inflammation through SOCS2 degradation and promoting colonic inflammation41—its function in the liver remains unexplored. In this study, knockdown of AREL1 was demonstrated to significantly suppress the activation of HSCs. In addition, HSC-specific ablation of Arel1 attenuated liver fibrosis in mouse models established by western diet or GAN diet feeding. Mechanistically, AREL1 mediated the ubiquitination of ILK at lysine 27, increasing its activity and leading to activation of the PI3K–AKT pathway, thereby promoting HSC activation. The PI3K-AKT signaling pathway plays a critical role in regulating cell proliferation, survival, adhesion, and migration42. In HSCs, this pathway is central to activation through multiple interconnected mechanisms. First, PI3K-AKT signaling markedly promotes HSC proliferation43,44; its inhibition impairs HSC migration and contractility45. Second, this pathway exerts anti-apoptotic effects that sustain HSC survival during chronic liver injury44. Third, it directly stimulates type I collagen transcription, driving excessive extracellular matrix deposition45. The activation of HSCs is efficiently triggered by platelet-derived growth factor (PDGF), which binds to PDGFR-β, activates Ras, and transduces stimulatory signals through the PI3K-AKT pathway46,47. In addition, the PI3K-AKT pathway interacts and synergizes with TGF-β signaling, another critical pathway driving HSC activation, to promote liver fibrosis progression48–50. Collectively, PI3K-AKT signaling is a key driver of HSC activation and fibrogenesis, thereby promoting liver fibrosis progression. Previous studies have indicated that a high-cholesterol diet exacerbates liver fibrosis in bile duct ligation- and CCl4-induced models. The mechanism involves the accumulation of free cholesterol in HSCs, leading to upregulated TLR4 protein expression and thereby increasing the sensitivity of HSCs to TGF-β signaling51. Another study in a MASH model further confirmed the crucial role of TLR4 and revealed that impairment of SREBP2-mediated feedback regulation of cholesterol homeostasis leads to aberrant cholesterol sensitivity in HSCs52. Under physiological conditions, when cells detect low cholesterol levels, SREBP2 is activated and translocated into the nucleus, where it activates the expression of downstream target genes to promote cholesterol synthesis and uptake; conversely, high-cholesterol environments suppress SREBP2 activity, resulting in the formation of a classical negative feedback loop that maintains cellular cholesterol homeostasis53. However, under MASH conditions, this feedback system in HSCs is disrupted. An increased Scap/Insig ratio causes the SREBP2–Scap complex to be preferentially transported to the Golgi for activation—even under high-cholesterol conditions—rather than being retained in the endoplasmic reticulum, resulting in sustained aberrant activation of SREBP254. Our study shows that cholesterol promotes AREL1 transcription through SREBP2, but this regulation does not involve enhanced SREBP2 activity or nuclear translocation. Instead, cholesterol specifically enriches H3K27ac modifications at the AREL1 promoter region, facilitating stronger binding of SREBP2 to this locus and thereby driving AREL1 expression. Cholesterol and its metabolites have been demonstrated to regulate histone modifications and DNA methylation22,23,55, and even to directly modulate chromatin architecture24—findings that support our results. Although we observed upregulation of SREBP2 in HSCs under MASH conditions, our experimental evidence indicates that this increase does not directly reflect a response to cholesterol levels. Other pathways may contribute to SREBP2 upregulation in this context, a possibility that warrants further investigation. More than ameliorated hepatic fibrosis, we observed that HSC-specific knockdown of Arel1 reduced hepatic steatosis in the MASH model, but direct knockdown of AREL1 in hepatocytes failed to ameliorate lipid accumulation. This discrepancy suggests that the metabolic benefits observed upon AREL1 depletion in HSCs may arise from non-cell-autonomous mechanisms. Indeed, emerging evidence points to a complex bidirectional crosstalk between HSCs and hepatocytes during MASLD progression56,57. Activated HSCs are known to secrete a variety of cytokines, chemokines, and signaling lipids that can directly modulate lipid metabolism in neighboring hepatocytes58,59. Taken together, these findings raise the possibility that AREL1 in HSCs influences hepatocyte lipid handling through HSC-derived paracrine signals. Further investigations are warranted to dissect the molecular mediators underlying this intercellular communication. Nevertheless, targeting AREL1 to inhibit HSC activation may represent a promising therapeutic strategy to simultaneously alleviate both steatosis and fibrosis in MASH. In conclusion, our study identifies AREL1 as a promising therapeutic target for MASH-related liver fibrosis with substantial translational potential. Methods Ethical statement Human studies: The study involving human participants was approved by the Ethics Committee of Zhongnan Hospital of Wuhan University (approval No. 2025095 K) and adhered to the principles outlined in the Declaration of Helsinki. Informed consent was obtained from each patient prior to the study, and consent for reporting and sharing de‑identified clinical data was also obtained. No participant compensation was provided. All human tissue samples used in this study were fully processed and consumed during the analyses. Animal studies: All animal experimental procedures were approved by the Welfare and Ethics Committee for Experimental Animals at Wuhan University Zhongnan Hospital (approval No. ZN2025192) and were conducted in strict accordance with relevant ethical guidelines. All animal care and use adhered strictly to the ethical guidelines outlined in the Guide for the Care and Use of Laboratory Animals by the National Academy of Sciences and the National Institutes of Health. Clinical samples Normal and MASH liver samples were obtained from Zhongnan Hospital of Wuhan University. The cohort included both male and female patients; however, sex was not considered as a variable in the study design a priori. No sex-based analyses were performed because the sample size was insufficient to enable meaningful sex-disaggregated conclusions. The sex of participants was determined based on medical records and self-report. Individual-level sex-disaggregated data are provided in Supplementary Data 2. Animal strains Wild-type male C57BL/6 J male mice were procured from the Wuhan University Animal Experiment Center in Hubei, China. HSC-specific Arel1 knockout male mice (AREL1HSC-KO) were generated by crossing lrat-Cre mice with Arel1 flox mice (AREL1fl/fl). Mice with the genotype Lrat-Cretg/0 x AREL1fl/fl were designated as AREL1HSC-KO, while AREL1fl/fl male mice were used as controls. lrat-Cre mice and AREL1fl/fl mice were purchased from Shanghai Model Organisms Center, Inc. The male mice were housed in a specific pathogen-free (SPF) facility under controlled environmental conditions: temperature 22–24 °C, relative humidity 50–60%, and a 12-h light/dark cycle. At the end of the experiments, mice were humanely sacrificed by carbon dioxide inhalation followed by cervical dislocation. MASH model Four weeks old male mice were administered either the RD western Diet (20% protein, 21% fat, 50% carbohydrates; TP 26304; D12079B; Research Diets) or the GAN diet (22.5% protein, 19.9% fat, 44.9% carbohydrates, 0.2% cholesterol; D09100310; Research Diets) for a duration of 20 weeks. The male mice were then sacrificed and analyzed at the 24th week. For each model, eight male mice were used per experimental group. Mice were randomly assigned to treatment groups, and investigators were blinded to genotype and treatment allocation during data collection and outcome assessment. Only male mice were used to avoid confounding by female hormonal cycles; thus, the results may not be generalizable to females. In vivo ubiquitination assay FLAG-AREL1, HA-ILK, and MYC-ubiquitin expression plasmids were co-transfected into LX2 cells. Before harvesting, following a 6-hour treatment with 10 μM MG132, the cells were lysed and subsequently subjected to sonication. The lysates were boiled at 95 °C for 10 minutes. After centrifugation at 13,000 × g for 15 min at 4 °C, 10% of the supernatant was collected as input, while the remaining supernatant was incubated with HA affinity beads at 4 °C for 4 h. Subsequently, beads were washed three times with high-salt wash buffer (20 mM Tris-HCl, pH 8.0, 300 mM NaCl, 1 mM EDTA, 1% NP-40) and denatured in loading buffer at 95 °C for 10 min. Western blotting was ultimately performed to analyze the samples. DNA pull-down assay The biotin-labeled AREL1 promoter was synthesized using a DNA Biotinylation Kit (BTN90604B, Biolab) through PCR amplification. The biotin-labeled AREL1 promoter was first incubated with streptavidin-coated magnetic beads at 4 °C overnight. Following this binding step, the beads were co-incubated with cell lysates for an additional 3 h, followed by three washes. The protein complexes pulled down were identified using protein mass spectrometry. Construction of VA-siAREL1-LNP In summary, SPC, DOTAP, DSPE-PEG2k-vitamin A, and cholesterol were dissolved together in 5 mL of anhydrous ethanol and transferred to a round-bottom flask. siAREL1 was dissolved in a citrate buffer containing 25% ethanol (50 mM citrate, pH 4) and then slowly added to the lipid mixture, followed by incubation for 20 min. The mixture was subjected to ultrasonication and processed using a liposome extruder with a 100 nm filter membrane. Subsequently, the product was dialyzed using a nanodialysis device with a polycarbonate membrane of 10 nm pore size and then lyophilized to obtain the liposome lyophilized powder. Frozen tissue nuclear dissociation Frozen tissue samples were sectioned into fragments smaller than 0.5 cm and homogenized using a glass Dounce tissue grinder (Sigma, cat. no. D8938). Homogenization was carried out with 25 strokes of pestle A followed by 25 strokes of pestle B in 2 mL of ice-cold EZ nuclei lysis buffer. The resulting homogenate was incubated on ice for 5 min after adding 3 mL of cold lysis buffer. Nuclei were pelleted by centrifugation at 500 × g for 5 min at 4 °C, washed once with 5 mL of ice-cold lysis buffer, and incubated again on ice for 5 min. After a final centrifugation step, the pellet was resuspended in 5 mL of nuclei suspension buffer (NSB; 1× PBS, 0.01% BSA, 0.1% RNase inhibitor [Clontech, cat. no. 2313 A]). For flow cytometric purification, all steps were performed under RNase-free conditions with samples maintained on ice to prevent RNA degradation. Nuclei suspensions were filtered through a 70 μm RNase-free strainer, stained with 0.5 μg/mL DAPI for 5 min on ice, and sorted on a BD FACS Aria III sorter (405 nm laser, 70 μm nozzle, 20 psi sheath pressure). The instrument was pre-calibrated with BD CS&T beads, with sample and collection chambers maintained at 4 °C. A sequential gating strategy was used to eliminate background and enrich intact single nuclei: (1) FSC-A/SSC-A gating to exclude low-scatter debris and non-nuclear background; (2) doublet discrimination via FSC-H/FSC-W and SSC-H/SSC-W plots; (3) DAPI-positive gating to isolate intact nuclei, pooling both 2n and 4n populations to preserve native hepatocyte polyploidy. Nuclei were sorted in purity mode into pre-chilled RNase-free wash buffer supplemented with RNase inhibitor, with ≥ 50,000 single nuclei collected per sample. Post-sort nuclei integrity and purity were validated via microscopy prior to library construction. The purified nuclei were resuspended in 2 mL of NSB, filtered through a 35 μm strainer (Corning-Falcon, cat. no. 352235), and counted. A final concentration of 1000 nuclei/μL was used for loading onto the DNBelab C4 channel. DNBelab C4 library preparation and sequencing Single-cell libraries were constructed using the DNBelab C4 scRNA Preparation Kit (MGI) following the manufacturer’s protocol. Briefly, cells were diluted to a concentration of 1000 cells/mL and loaded into a microfluidic chip. Barcoded beads and droplet-generation oil were sequentially introduced, and droplets were generated using the DNBelab C4 system. Beads capturing mRNA were recovered for reverse transcription, followed by PCR amplification. The resulting cDNA was purified and quantified using a Qubit dsDNA kit (Invitrogen, #Q32854). Libraries of 3′-end transcripts were generated through cDNA fragmentation, size selection, end repair and A-tailing, adapter ligation, PCR indexing, and circularization. Sequencing libraries were purified, quantified with the Qubit ssDNA kit (Thermo Fisher, #Q10212), and sequenced on the DNBSEQ-T7 platform (PE150) using the DNBSEQ-T7RS Reagent Kit v3.0. Single-cell RNA-seq and single-nucleus data processing Public human fibrotic scRNA-seq and snRNA-seq datasets (PRJNA877625, PRJNA731458, PRJNA561303, PRJCA027647) were obtained from the SRA or CNCB. Data were processed using Cell Ranger v8.0.1 with the GRCh38-2024-A reference genome to generate gene–barcode matrices. Mouse snRNA-seq reads were processed using DNBc4tools v2.1.2, aligned to the mouse genome, and converted into gene–barcode matrices via UMI counting and barcode filtering. All resulting data were imported into Seurat v4.4.0 for subsequent analyses. Single-cell RNA-seq and single-nucleus data analysis Seurat objects were generated with thresholds set to min.cells = 3 and min.features = 200, then merged for downstream analysis. Quality control was conducted per sample after splitting by orig.ident. Standard preprocessing steps included normalization, selection of variable features, scaling, PCA, UMAP embedding, neighbor graph construction (using the first 50 principal components), and clustering (resolution = 0.3). Doublets were removed using DoubletFinder (pN = 0.25, data-driven pK, expected rate = min (nCells × 8e − 6, 0.08), adjusted for homotypic doublets). Ambient RNA contamination was corrected using scCDC, and the corrected counts were stored in a “Corrected” assay. Cells were retained if they met the following criteria: 500 ≤ nFeature_RNA ≤ 6000, 1000 ≤ nCount_RNA ≤ 30,000, percent.mt < 20%, and percent.HB < 20%. Post-QC objects were exported in HDF5 format for Python-based analyses. For large public datasets, analyses were performed in Scanpy. After importing h5ad files, library-size normalization (target_sum = 1e4) and log1p transformation were applied, and 3000 highly variable genes (HVGs) were selected using the “seurat” method. Mitochondrial content (pct_counts_mt) was regressed out, followed by scaling. PCA was performed (30 components), batch effects were corrected using Harmony (using orig.ident as the key), and neighborhood graphs were constructed based on Harmony embeddings (n_neighbors = 30). Leiden clustering (resolution = 1.0) and UMAP visualization (min_dist = 0.2, spread = 1.0) were applied. Processed AnnData objects were saved in compact HDF5 format for re-import into R. For mouse snRNA-seq data, Seurat preprocessing included normalization, HVG selection (n = 2000, vst method), scaling, PCA (15 components), clustering (resolution = 0.8), and UMAP visualization (min.dist = 0.1). Harmony integration was applied to correct for batch effects, followed by reclustering (resolution = 1.0) and UMAP visualization (dimensions 1–15, min.dist = 0.2). Cell-type annotation was conducted in R. Global marker genes were identified using presto::wilcoxauc, and manual annotation was guided by curated marker panels, with validation via DotPlot visualizations. Subclustering and functional analysis of HSCs and hepatocytes In public datasets, HSCs were isolated by subsetting cells annotated as celltype1 == “HSC”. A predefined exclusion list was applied to remove potentially confounding genes1. HSCs were re-integrated using the following pipeline: CreateSeuratObject → NormalizeData → FindVariableFeatures → ScaleData → RunPCA → Harmony integration (by orig.ident) → UMAP on Harmony dimensions 1–15 → FindNeighbors/FindClusters at resolution = 0.1. Differential expression analysis was performed using presto::wilcoxauc across tissue type and Seurat clusters, with significance defined as adjusted P < 0.05 (Benjamini–Hochberg) and logFC > 0. Functional enrichment analyses were conducted using clusterProfiler, and heatmaps were generated with ClusterGVis. Targeted queries for specific genes (e.g., AREL1) were performed in both group-wise and cluster-wise comparisons, with expression patterns visualized using scCustomize. Pathway activity scoring was performed with VISION on the Harmony-integrated HSC object, using curated gene sets from MSigDB. For mouse snRNA-seq data, HSCs were subset after applying a blacklist filter60. A new Seurat object was created from raw counts and metadata and processed through a standard pipeline comprising normalization, variable feature selection, scaling, PCA (8 components), UMAP embedding (dimensions 1–8), neighbor graph construction, and clustering at resolution = 0.1. Cell-type and group composition were summarized using scMC. Differential expression was assessed with presto::wilcoxauc across experimental groups and HSC clusters, with significance defined as adjusted P < 0.05 and logFC > 0. Enrichment analyses were performed using clusterProfiler, and heatmaps were generated with ClusterGVis. Hepatocytes (celltype1 == “Hepatocytes”) were processed similarly, starting with blacklist filtering60, followed by normalization, variable feature selection, scaling, PCA (10 components), UMAP embedding (dimensions 1–10), neighbor graph construction, and clustering at resolution = 0.1. Group-wise differential expression analysis was conducted using presto, with significant genes defined by adjusted P < 0.05. Volcano plots were generated to compare VA-siAREL1-LNP against VA-LNP, highlighting genes with |logFC | ≥ 0.25 and adjusted P < 0.05. Enrichment analyses followed the same workflow as applied to HSCs. Cell–cell communication analysis To investigate intercellular signaling between hepatocytes and HSCs, the integrated object was subset to these two cell types and split by experimental group (VA-LNP vs VA-siAREL1-LNP). CellChat objects were constructed (group.by = “celltype1”) using the mouse ligand–receptor database. The standard workflow included expression-based filtering (subsetData), identification of overexpressed genes and interactions (identifyOverExpressedGenes, identifyOverExpressedInteractions), computation of communication probabilities using the triMean method (computeCommunProb), filtering of rare interactions (min.cells = 10), pathway-level probability estimation (computeCommunProbPathway), aggregation (aggregateNet), centrality analysis (netAnalysis_computeCentrality), and functional similarity assessment (computeNetSimilarity). Condition-specific objects were merged (mergeCellChat) for comparative analyses, including interaction number and strength (compareInteractions), pairwise differential networks (netVisual_diffInteraction), heatmaps, signaling-role scatter plots, and low-dimensional embeddings/clustering based on functional and structural similarity (netEmbedding, netClustering). Differential communication driven by gene expression was assessed by re-running identifyOverExpressedGenes across datasets (features.name = “VA-siAREL1-LNP. merged”; thresholds: pct ≥ 0.1, logFC ≥ 0.05, P < 0.05), projecting differential expression onto ligand–receptor pairs (netMappingDEG), and extracting up- and down-regulated interactions (subsetCommunication with ligand.logFC ≥ 0.05). Visualizations included bubble plots (netVisual_bubble), chord diagrams (netVisual_chord_gene), and word-cloud summaries of enrichment (computeEnrichmentScore). Trajectory inference in HSCs The HSC Seurat object was converted into a Monocle2 CellDataSet. Size factors and dispersions were estimated, and genes expressed in at least 10 cells with a minimum expression threshold of 3 were tested for cluster-associated differential expression (differentialGeneTest with fullModelFormulaStr = “~clusters”). Ordering genes were defined by q < 0.01 and used to set the ordering filter. Dimensionality reduction (max_components = 2) and cell ordering (orderCells) generated a pseudotemporal trajectory, which was re-rooted at State 6 to align with biological expectations. Trajectories were visualized by group, cluster, state, and pseudotime. The CellDataSet was converted back to Seurat for state-wise marker discovery (FindAllMarkers; only.pos = TRUE), and selected genes were visualized as stacked violin plots. Pseudotime-dependent differential expression was assessed using natural splines (differentialGeneTest with ~ sm.ns(Pseudotime)) and by state, with results ranked by q value. Expression dynamics were profiled for quiescent HSC markers (e.g., Gfap, Ngfr, Lrat, Pparg) and activated HSC markers (e.g., Acta2, Col1a1, Col1a2, Tgfbr1, Timp1). Statistical analysis Unless otherwise stated, nonparametric Wilcoxon rank-sum tests were used for two-group comparisons. Multiple testing in differential expression analyses was controlled using the Benjamini–Hochberg method, with adjusted P < 0.05 considered significant. All analyses were conducted in a Linux environment using R packages (Seurat, DoubletFinder, scCDC, Harmony, STACAS, presto, VISION, scMC, scMetabolism, CellChat, monocle, clusterProfiler, ReactomePA, org.Mm.eg.db, ggplot2, patchwork, gghalves, ggpubr, ggsci) and Python packages (scanpy, anndata, numpy, scipy, pandas, matplotlib, harmony via scanpy external). Bulk RNA-seq: RNA extraction, library preparation, and sequencing Total RNA was extracted from tissue samples using TRIzol® Reagent (Invitrogen) according to the manufacturer’s instructions. RNA quality was assessed using a 5300 Bioanalyzer (Agilent) and quantified with a NanoDrop ND-2000 spectrophotometer. Only high-quality RNA (OD260/280 = 1.8–2.2, OD260/230 ≥ 2.0, RQN ≥ 6.5, 28S:18S ≥ 1.0, and yield >1 μg) was used for library construction. RNA purification, reverse transcription, library preparation, and sequencing were performed by Shanghai Majorbio Bio-Pharm Biotechnology Co., Ltd. Libraries were prepared using the Illumina® Stranded mRNA Prep, Ligation kit with 1 μg of total RNA. Poly(A)+ mRNA was isolated using oligo(dT) beads, fragmented, and reverse transcribed into double-stranded cDNA with random hexamer primers (SuperScript kit, Invitrogen). The cDNA underwent end repair, phosphorylation, adapter ligation, and size selection (~300 bp), followed by PCR amplification (15 cycles, Phusion DNA polymerase, NEB). Libraries were quantified using Qubit 4.0 and sequenced on an Illumina NovaSeq X Plus platform (2 × 150 bp) for human liver tissue, and on the DNBSEQ-T7 platform (PE150) using the DNBSEQ-T7RS Reagent Kit v3.0 for mouse tissue. Bulk RNA-seq: read mapping and downstream analysis Raw paired-end reads were trimmed and subjected to quality control using fastp with default parameters. Clean reads were aligned to the reference genome using STAR, and gene-level counts were generated. Differential expression analysis was performed using the DESeq2 R package, with significance defined by a Benjamini–Hochberg adjusted P < 0.05. Functional enrichment analyses were conducted using clusterProfiler. Isolation of primary mouse HSCs The mouse liver was perfused through the superior vena cava in a sequence beginning with HEPES phosphate-buffered saline A, followed by 30 mL of 0.5 mg/mL pronase, and ending with 30 mL of 0.5 mg/mL collagenase D. This was followed by a 20-min enzymatic digestion using a cocktail of DNase, collagenase D, and pronase to create a cell suspension. The suspension was initially centrifuged to separate and discard residual hepatic parenchymal cells. The resulting supernatant was then centrifuged at 900 × g for 8 min at 4 °C to isolate non-parenchymal cells. These cells were resuspended in a 15% OptiPrep solution. Subsequently, 5 mL of 11.5% OptiPrep and 2 mL of GBSS solution (C1076-B, Livning) were gently layered onto the cell suspension. The tubes underwent centrifugation at 2000 × g for 20 minat 4 °C with low acceleration, and cells located between the 11.5% OptiPrep and GBSS layers were collected, resulting in purified HSCs. Cell culture HEK293T cells were obtained from the American Type Culture Collection (ATCC, USA, catalog number: CRL-3216). LX2 cells were purchased from Procell Life Science & Technology Co., Ltd. (Wuhan, China, catalog number: CL-0560). All cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (FBS; Gibco) and 1% penicillin–streptomycin mixture (Gibco), at 37 °C in a humidified atmosphere containing 5% CO₂. All cell lines were authenticated by short tandem repeat (STR) profiling and confirmed to be free of mycoplasma contamination using a PCR‑based detection kit. Plasmid construction shRNA sequences were inserted into pLKO.1-puro using AgeI and EcoRI. Overexpression constructs were generated in pcDNA3.1 vectors. Epitope tags (e.g., FLAG, HA, MYC) were added as needed. All constructs were verified by Sanger sequencing. Adenovirus generation and infection The shuttle plasmid pENTR-U6 and the ViraPower Adenoviral Expression System were utilized to generate adenoviral vectors for mouse AREL1 knockdown. Following linearization of the plasmid, the adenoviral vector was transfected into 293 cells using Neofect™ DNA transfection reagent (TF201201, Neofect (Beijing) Biotech Co., Ltd.). Cells were harvested 6–7 days later to obtain the initial adenovirus. After three rounds of amplification, the shAREL1 adenovirus was purified through cesium chloride density gradient centrifugation, and the titer was determined using the 50% tissue culture infectious dose (TCID50) method. Primary mouse HSCs were infected with the adenovirus at a multiplicity of infection (MOI) of 50. Western blotting Cells were lysed using RIPA buffer and subjected to sonication before proteins were loaded onto a 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gel. After separation, the proteins were transferred to a polyvinylidene difluoride (PVDF) membrane and blocked with 5% non-fat milk. The membrane was incubated overnight with primary antibodies at 4 °C. Detection was performed using an ECL chemiluminescence imaging system (Image Lab). All antibodies were used at dilutions recommended by the manufacturer. The antibodies used are listed in the Supplementary file Table 1. RNA isolation and qPCR analysis RNA was extracted using TRIzol reagent (15596026, Invitrogen), and cDNA was synthesized with a reverse transcription kit (R223, Novozymes). qPCR was performed using the QuantiTect SYBR Green PCR Kit (Q311-02/03, Novozymes) to quantify mRNA expression levels, which were normalized to the expression levels of reference genes. The primer sequences used in this study are listed in the Supplementary file Table 2. shRNA knockdown Lentiviral plasmids were co-transfected with packaging plasmids into HEK-293T cells. The resulting viral supernatant was used to infect target cells, followed by puromycin selection to ensure stable integration. The shRNA sequences used in this study are listed in the Supplementary file Table 3. Immunoprecipitation FLAG-AREL1 and HA-ILK expression plasmids were co-transfected into LX2 cells. The cells were then lysed, and the protein supernatant was collected. Affinity beads were incubated with the supernatant, washed, and denatured. Samples were analyzed by Western blotting. Palmitic acid (PA) and cholesterol treatment For the treatment of PA and cholesterol, the cells were incubated with PA (100 μM) or cholesterol (60 μM) in DMEM61. BSA in DMEM was used as a control. GST pull-down assay LX2 cells were transfected with plasmids encoding GST-AREL1 and FLAG-ILK. After 24 h, cells were collected and lysed, and the protein supernatant was obtained, with 10% reserved as the input sample. The GST-AREL1 protein supernatant was incubated with GST resin for 3 h. The resin was then washed three times with GST pull-down buffer (20 mM Tris-HCl, 150 mM NaCl, 0.2% Triton X-100). Subsequently, the FLAG-ILK protein supernatant was incubated with the resin for another 3 h, followed by three additional washes. The interactions were further analyzed by Western blotting. The same method was used to verify interactions between GST-ILK and FLAG-AREL1. Mass spectrometry DNA pull-down using the AREL1 promoter and its IgG control, and immunoprecipitation using AREL1 and its IgG control. No replicates were performed (n = 1 per sample). Proteins were reduced, alkylated, and digested with trypsin. Peptides were separated on an UltiMate 3000 RSLCnano system (Thermo Scientific) using a C18 column, and analyzed on a Q Exactive HF mass spectrometer (Thermo Scientific) in data‑dependent acquisition mode (top20, HCD 28%). Raw data were searched with MaxQuant (v1.6.6.0) against the UniProt human database. Identifications were filtered at 1% FDR. Proteins labeled as reverse, contaminant, or only identified by site were removed. Cell proliferation, migration, and invasion assays Cell proliferation was assessed using the Cell Counting Kit-8 (CCK-8) assay (BS350A, Biosharp). Migration and invasion capabilities were evaluated through scratch assays and BioCoat Matrigel invasion chambers (14342, LABSELECT). Cell contraction assays Cell contraction was assessed using a Cell Contraction Assay kit (CBA-201, Cell Biolabs). Briefly, cells were mixed with Type I collagen, and the gel was loaded into a 24-well plate (500 µL per well), where it was allowed to solidify at 37 °C for 1 h. Subsequently, DMEM containing 10% fetal bovine serum was added, and the cells were allowed to contract for 3 days. Gel contraction was determined by measuring the gel diameter. Histology and immunohistochemistry Tissue samples were fixed in 4% paraformaldehyde, embedded, and sectioned. Antigen retrieval was performed, followed by blocking with a primary antibody, incubation with a secondary antibody, and staining with 3,3’-diaminobenzidine (DAB). Slides were counterstained with hematoxylin, dehydrated, and mounted. For quantitative analysis, five non-overlapping high-power fields were randomly selected from each section and photographed under identical microscope settings. Image analysis was performed using ImageJ software (NIH, Bethesda, MD, USA). Briefly, images were subjected to color deconvolution to separate DAB staining from hematoxylin counterstaining. A uniform threshold was applied across all images to identify positive staining, and the percentage of DAB-positive area relative to the total tissue area was calculated. The average value from five fields per sample was used for statistical analysis. All quantifications were performed by two independent investigators blinded to sample identities. Chromatin immunoprecipitation (ChIP) assay The ChIP assay was conducted using a ChIP Assay Kit (P2078, Beyotime). Briefly, cells were cross-linked with 1% formaldehyde. After sonication and preclearance with Protein A Agarose/Salmon Sperm DNA, the sheared chromatin was immunoprecipitated with either control IgG or antibodies against SREBP2, H3K27ac, and H3K27me3. Following elution, the DNA/antibody complexes were digested with Proteinase K, and the DNA was extracted and subjected to qPCR analysis. The supernatant prior to immunoprecipitation for each sample served as an input control. Flow cytometry assay Male mice were intravenously injected with Cy5.5-labeled LNPs. Livers were harvested after euthanasia, enzymatically digested, and primary hepatocytes, macrophages, HSCs, and LSECs were isolated by density gradient centrifugation. Isolated cells were washed with PBS containing 2% FBS and resuspended in staining buffer for flow cytometric analysis. All procedures were performed on ice. Flow cytometric analysis was performed using a Beckman CytoFLEX flow cytometer, and data were analyzed with FlowJo software (version 10.9). The gating strategy included forward/side scatter to exclude debris, FSC-A vs. FSC-H to select single cells, and Cy5.5-positive gates were set based on unlabeled control samples with a threshold of <1% positivity. No cell-type-specific marker staining was performed for purity assessment, as the analysis focused on Cy5.5 signal uptake across isolated cell populations. A representative figure exemplifying the gating strategy is shown in Supplementary Fig. 13. Dual-luciferase assay Cells were seeded at a density of 1 × 105 cells per well in a 24-well plate and transfected with a luciferase reporter plasmid along with a Renilla control plasmid. After 24 h, the cells were harvested and lysed according to the instructions provided with the Dual-Luciferase Assay Kit (DL101-01, Vazyme). Luciferase activity was then measured using a luminometer. Statistics and reproducibility All data are presented as the mean ± standard deviation. Statistical analyses were conducted using GraphPad Prism software (version 10.5.0). Two-tailed Student’s t tests determined differences between two groups, while one-way ANOVA assessed differences across multiple groups. Each experiment included at least three biological replicates with data collected from independent experiments. A P < 0.05 was considered statistically significant. Normality of data distribution was assumed but not formally tested. No statistical methods were employed to predetermine sample sizes. Our sample sizes are similar to those reported in previous publications62,63. To minimize bias, cells and mice were randomly assigned to experimental groups and received uniform treatment. Investigators were blinded to genotype during data analysis. No data were excluded from the analyses. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Reporting Summary Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Reporting Summary Transparent Peer Review file Source data Source data Source data