Enterococcus faecalis impairs IGSF9-dependent C1q degradation to accelerate MAFLD-HCC progression

ABSTRACT

The global increase in metabolic associated fatty liver disease (MAFLD)-related hepatocellular carcinoma (HCC) necessitates urgent investigation of its underlying mechanisms. Our study reveals that high fat diet (HFD)-induced gut dysbiosis enriches Enterococcus faecalis, a potential carcinogenic bacterium. Mechanistically, which suppresses hepatic IGSF9 expression, exacerbating liver injury (elevated ALT/AST/ALP), steatosis, fibrosis, and HCC susceptibility. Clinically, reduced hepatic IGSF9 correlates with poor prognosis, particularly in MAFLD-HCC patients. Mechanistically, IGSF9 binds C1QC and E3 ubiquitin ligases (NEDD4/UBE3A) to promote C1QC ubiquitination and degradation; IGSF9 deficiency leads to pathogenic C1q accumulation, accelerating MAFLD progression. Importantly, early interventions targeting this axis - including Bifidobacterium pseudolongum supplementation, overexpressing IGSF9, and C1q neutralization - effectively prevented MAFLD-HCC progression in preclinical models. These findings provide preliminary evidence that the E. faecalis/IGSF9/C1q axis could be involved in MAFLD-HCC, with IGSF9 may serve as both a prognostic biomarker and therapeutic target, which worth further exploration in future studies.

GRAPHICAL ABSTRACT

IGSF9 serves as a biomarker for predicting the clinical prognosis of patients with MAFLD-HCC, exhibiting a gradual decline as MAFLD progresses to HCC. Mechanistically, a high-fat diet leads to an increased prevalence of Enterococcus faecalis in the intestines of MAFLD patients, which subsequently downregulates IGSF9 expression in hepatocytes. The diminished interaction between IGSF9, C1QC and NEDD4/UBE3A results in substantial inhibition of ubiquitination degradation and accumulation of C1q in the liver, thereby exacerbating lipogenesis, inflammation, fibrosis and subsequent transition from MAFLD to HCC. Administration of Bifidobacterium pseudolongidum, IGSF9 activating antibodies/AAV (predicted) and C1q neutralizing antibodies as primary or secondary prevention can offer a therapeutic effect for MAFLD patients.

Introduction

Metabolic associated fatty liver disease (MAFLD), affecting approximately 25% of global population, is a metabolic disorder characterized by liver fat accumulation resulting from impaired lipid metabolism.^1-3 About 13%–38.2% of patients with MAFLD may progress to hepatocellular carcinoma (HCC).^4,5 Progression to HCC in MAFLD happens particularly in those with nonalcoholic steatohepatitis (NASH) and exacerbated by metabolic syndrome.⁶ The prevalence of MAFLD-related HCC (MAFLD-HCC) is on an upward trajectory and will cover the majority of patients with HCC worldwide; while MAFLD-HCC has poor prognosis since it is typically accompanied by older age, multiple metabolic syndromes, difficulties in early-stage diagnosis, and limited drugs for treatment.⁵ Current situation highlights the urgent need to explore new intervention targets and discover feasible and effective therapeutic strategies.

The gut microbiota, composed of trillions of microorganisms, serves as a central regulator of host metabolism. Through the gut-liver axis, gut-dysbiosis has been mechanistically linked to pathogenesis of chronic liver diseases such as MAFLD, NASH, liver cirrhosis, and HCC.⁷ About 7%–30% of patients with MAFLD develop NASH,⁸ where obesity-driven lipid accumulation⁹ coincides with characteristic microbial shifts: reduced α-diversity and an elevated Firmicutes/Bacteroidetes ratio.¹⁰ The microbiome emerges as a promising target for HCC prevention, supported by murine studies demonstrating that antibiotic intervention in advanced disease stages can ameliorate chronic liver injury and suppress hepatocarcinogenesis.¹¹ Importantly, gut-derived microbial metabolites exert sustained regulatory effects on gene expression via enterohepatic circulation.¹²

Members of the immunoglobulin superfamily (IGSF) serve as cell adhesion molecules that mediate cell–cell and cell-extracellular matrix interactions, while also transducing cellular signals in metabolic diseases such as NASH.¹³ IGSF9, a type I transmembrane protein, features an N-terminal domain containing a signal peptide, five immunoglobulin-like domains, and two fibronectin type III domains, along with a C-terminal region bearing an immunoreceptor tyrosine-based inhibitory motif (ITIM) and PDZ-binding motif.^14-17 Our analysis of public databases and clinical samples identified IGSF9 as a novel candidate gene in HCC, particularly MAFLD-associated cases. However, its precise role in metabolic dysregulation-driven hepatocarcinogenesis remains poorly understood. IGSF9 exhibits tissue-specific oncogenic functions: while upregulated in lung cancer and other malignancies,^18,19 it shows no elevation in advanced melanoma. Conversely, breast cancer demonstrates IGSF9 downregulation, correlating with reduced relapse-free survival (RFS) and overall survival (OS).²⁰ IGSF9 loss in breast cancer activates FAK/Akt signaling to induce epithelial‒mesenchymal transition (EMT). Liu et al. reported that IGSF9 might function as an immune-suppressing receptor; anti-IGSF9 treatment significantly enhances T-cell proliferation, increases the secretion of cytokines (perforin, TNF-α, and IFN-γ), and boosts anti-tumor effect of anti-PD-1.¹⁵

Here we uncovered a gut‒liver axis where Enterococcus faecalis (promoted by a high-fat diet) reduces IGSF9, leading to abnormal C1q buildup caused by defective C1QC ubiquitination‒degradation, thereby drives MAFLD-to-HCC progression. Targeting this pathway – such as Bifidobacterium pseudolongum implement – could prevent HCC. These findings nominate microbiome modulation as a translatable strategy for HCC prevention, establishing the gut microbiota as a promising therapeutic target in metabolic liver carcinogenesis.

Materials and methods

Patients and samples

A total of 48 pairs of freshly collected tumor tissues and adjacent normal tissues were randomly selected from hepatocellular carcinoma (HCC) patients who underwent surgical resection and were pathologically diagnosed with HCC at Zhongshan Hospital of Fudan University (Shanghai, China). These samples were utilized for quantitative reverse transcription polymerase chain reaction (qRT‒PCR) and western blot analysis. Additionally, a tissue microarray comprising 131 pairs of HCC patient tissues, along with clinicopathological characteristics such as age, tumor size, and TNM staging etc., was obtained from Zhongshan Hospital. Among these, 106 pairs included complete prognostic information. Recurrence-free survival (RFS) was delineated as the duration from the date of surgery to the occurrence of disease recurrence, while overall survival (OS) was defined as the interval from surgery to death.

The study was approved by the Institutional Ethics Committee of Fudan University (2021-008), and written informed consent was obtained from each participant prior to the commencement of clinical data analysis. All research was conducted in accordance with both the Declarations of Helsinki and Istanbul.

Cell culture and transfection

HEK-293T cells, human HCC cell lines PLC/PRF/5, Huh-7, Hep 3B2.1-7, Hep G2, HCCLM3, MHCC97-H, and the human hepatic stellate cell line LX2 were obtained from the Chinese Academy of Sciences Cell Bank (Shanghai, China). THLE-2 were purchased from Procell (Wuhan, China). All cell lines except THLE-2 were cultured in Dulbecco's Modified Eagle Medium (DMEM; Gibco, 11965092) supplemented with 10% fetal bovine serum (FBS; Gibco, 16140071) and 1% penicillin/streptomycin (Sigma, 381074) at 37°C in a 5% CO₂ atmosphere. THLE-2 cells were cultured using THLE-2 culture medium (Procell, CL-0833). The Huh-7, HCCLM3, and MHCC97-H cell lines were infected with lentiviruses expressing either shRNA targeting IGSF9 or an IGSF9 overexpression construct. Stable clones were subsequently selected using puromycin (Beyotime, ST551). HEK-293T cells were transfected with the over-expression plasmids using Lipo293™ Plus Transfection Reagent (Beyotime, C0522).

Other materials used in cell culture: oleic acid (Selleck, S4707), TGF-β1 (Procell, PCK354), C1q (Hycult, HC2143), Faecalibaculum rodentium (F. rodentium, Biobw, bio-134751), Enterococcus faecalis (E. faecalis, Biobw, bio-74191), Akkermansia muciniphila (A. muciniphila, Biobw, bio-115093), Bacteroides acidifaciens (B. acidifaciens, Biobw, bio-107393), Bacteroides caecimuris (B. caecimuris, Biobw, bio-109058), Bifidobacterium pseudolongum (B. pseudolongum, Biobw, bio-04954), and Clostridium leptum (C. leptum, Biobw, bio-109057).

Animals and models

Igsf9 knockout mice were generated as described previously.¹⁶ All animals were on a C57BL/6J genetic background. Mice were provided with food and water ad libitum and maintained under standard conditions (12 h light–dark cycle) in animal core facility at Fudan University.

For DEN-induced HCC mouse model, male mice aged two weeks received a single intraperitoneal (i.p.) injection of DEN (25 mg/kg, Sigma, N0756). For MAFLD-HCC mouse model, male mice were subsequently fed a high-fat diet (Research Diets, Inc., D12492) starting at six weeks of age (four weeks post-DEN injection).^21-23 Mice were euthanized at indicated time points following the DEN injection. For MAFLD mouse model, male mice aged six weeks were fed HFD for a duration of 28 weeks.²⁴ Blood tests were conducted using Automatic Biochemistry Analyzer (SIEMENS, ADVIA XPT).

For MAFLD C1q monoclonal antibody (mAb) intervention mouse model, male MAFLD mice aged six weeks were intraperitoneally injected with 10 μg C1q mAb (Hycult, HM1096) twice a week while others used mouse IgG2b isotype (BioXCell, BE0086) as control. All mice were sacrificed after 28 weeks.

For MAFLD Enterococcus faecalis intervention co-housing mouse model, male MAFLD mice aged four weeks were administered ABX for two weeks. 5 × 10⁸ colony forming units (CFUs) of Enterococcus faecalis in PBS were fed to mice by gavage three times every week.²⁵ Mice were co-housed together soon after gavage and were sacrificed 28 weeks after co-housing.

For MAFLD-HCC Enterococcus faecalis intervention mouse model, male MAFLD-HCC mice aged two weeks were fed on gavage with 5 × 10⁸ CFUs of Enterococcus faecalis or PBS three times every week for one month after DEN injection.¹¹ Mice were sacrificed at one, four, and eight months after DEN injection.

For MAFLD Bifidobacterium pseudolongum intervention mouse model, male MAFLD mice aged six weeks were fed on gavage with 1 × 10⁸ CFUs of Bifidobacterium pseudolongum or PBS daily. Mice were sacrificed 28 weeks after intervention.

All experimental procedures were approved by the Research Ethics Committee of Zhongshan Hospital and conducted in accordance with institutional guidelines.

Mice 3D NAC-liver spheroid construction

The construction of 3D liver spheroids from mice was performed as previously described.²⁶ In brief, hepatocytes and non-parenchymal cells were isolated from wild-type (WT) or Igsf9^-/- C57BL/6J mice using a two-step perfusion method. The 3D liver spheroids were then generated using the NAC strategy as follows. Hepatocytes and non-parenchymal cells were mixed at a 2:1 ratio and incubated with the NAC-Linker (Puheng Biomedicine). The cell mixture was subsequently seeded into ultra-low adhesion plates at a density of 3000 cells per well and incubated in a CO₂ incubator for 48 h to facilitate spheroid formation for subsequent experiments. For the MAFLD model, 3D liver spheroids derived from mice were cultured using NASH induction medium (Puheng Biomedicine, MI01) for a duration of 10 d following their initial construction.

Other materials and methods

Details of histological staining, immunohistochemical staining, and scoring, immunofluorescence, western blotting, co-immunoprecipitation, GST-pull down, enzyme-linked immunosorbent assay, RT‒qPCR and the molecular docking analysis are described in the Supplementary materials and methods.

Statistical analysis

Categorical variables were presented as frequencies and percentages. Continuous variables were summarized with means and SDs or medians and IQRs. Independent group t-tests compared means of normally distributed data, while the Mann–Whitney U test was used for non-normal data. One-way ANOVA compared multiple groups, using Dunnett's or Tukey's test for independent data. Two-way ANOVA compared more than two groups and variables, followed by Sidak's test. Chi-square or Fisher's exact tests were used for categorical variables, and Spearman correlation analysis for correlation analyzes. The Kaplan‒Meier method was used to create survival curves and the log-rank test was used to compare survival distributions. Statistical significance was determined by a p-value < 0.05. Data analysis and visualization were done using R version 4.1.0.

Result

HFD-induced Enterococcus faecalis colonization promoted MAFLD-HCC progression

A MAFLD-HCC mouse model was constructed. Male mice aged two weeks were injected with DEN intraperitoneally and followed by a high-fat diet starting at six weeks. DEN injection alone was served as control (Figure 1A).^21-23 During MAFLD-HCC, the model showed progressing macrovesicular steatosis detected by Oil red o staining, liver inflammation indicated by macrophage infiltration and fibrosis level detected by Masson staining (Figure 1B). At eight months after DEN injection, mice developed visually visible HCC in the liver (Figure 1C). Subsequently, 16S sequencing on the mice feces induced by DEN alone and DEN combined with HFD (DEN + HFD) for eight months were performed to determine the gut microbiota along MAFLD-HCC progression induced by HFD. We found that HFD changed the composition of gut microbiota and reduced the α-diversity (Figure 1D,E). From the top-ranked species, we found that HFD increased the abundance of bacteria such as Faecalibaculum rodentium and Enterococcus faecalis (E. faecalis) and decreased the abundance of bacteria such as Akkermansia muciniphila, Bacteroides acidifaciens, Bacteroides caecimuris, Bifidobacterium pseudolongum (B. pseudolongum), and Clostridium leptum (Figure 1F). E. faecalis is currently a well-established risk factor for MAFLD or MAFLD-HCC but lacks downstream molecular mechanisms.^27,28 To validate its effect on MAFLD-HCC, we intervene MAFLD-HCC model with E. faecalis (Figure 1G). Intervention of E. faecalis led to more severe hepatic macrovesicular steatosis, inflammation and fibrosis level (Figure 1H). More and larger tumors (10.33 vs 18, p < 0.05; 5.5 vs 10.33 mm, p < 0.05) in the liver of the E. faecalis intervention group mice were also observed (Figure 1I), as well as stronger tumorous Ki67 and AFP staining (Figure S1A). Blood tests also suggested elevated CHOL (9.342 vs 14.75 mmol/L, p < 0.05), DLDL (1.595 vs 3.125 mmol/L, p < 0.05), TG (0.81 vs 1.555 mmol/L, p < 0.05), ALT (277 vs 518.2 U/L, p < 0.05), and AST (285.6 vs 511.7 U/L, p < 0.05) (Figure S1B, C). Furthermore, the application of E. faecalis, after clearing the local gut bacteria of mice by broad-spectrum antibiotics, and exacerbated hepatic macrovesicular steatosis, inflammation and fibrosis. In addition, a similar phenotype was observed in the Ctrl group co-housed with the E. faecalis group (Figure 2A–D). We also constructed 3D liver spheroid models with parenchymal hepatocytes and nonparenchymal cells including endothelial cells, hepatic stellate cells and Kupffer cells from WT mice for validation.²⁶ During MAFLD stimulation, E. faecalis group had more severe levels of fatty infiltration and fibrosis, revealed by stronger Oil red o staining and Sirius red staining (Figure 2E), increased levels of TG (9.24 vs 11.36 nmol, P = 0.04) in the culture medium (Figure 2F), and increased markers of fatty liver (FABP4) and liver cirrhosis (α-SMA) at protein level (Figure S2A).²⁹ The gene expression of markers of lipogenesis (Fasn, Acc1, Acly), inflammation (Il-1β, Il-6, Ifn-γ) and fibrosis (Col1a1, Acta2, Timp1) were increased as well (Figure S2B). These findings indicate that E. faecalis facilitate the progression of MAFLD to HCC.

Figure 1.

Enterococcus faecalis promote MAFLD-HCC. A-C. MAFLD-HCC and DEN-induced HCC mice models were constructed using two-week male WT C57BL/6J mice injected with DEN and subsequently fed with high-fat diet (HFD) or not. Mice were sacrificed one, four and eight months after DEN injection (A). Oil red o, F4/80 and Masson staining (B) during MAFLD-HCC process and terminal gross appearance and HE staining (C) were shown. D-F. Fecal 16 s rRNA sequencing in DEN and DEN + HFD mice, n = 6. (D) PCoA plot, (E) α-diversity indices and top differential gut microbiota (F) between DEN and DEN + HFD group. Statistical analyzes: Unpaired two-tailed Student's t-test. Enterococcus faecalis (E. faecalis) is highlighted in green. G-I. E. faecalis facilitated MAFLD-HCC process. MAFLD-HCC E. faecalis intervention models (G) were constructed based on MAFLD-HCC mice as previously described. 5 × 10⁸ CFUs of E. faecalis or PBS (Ctrl) three times every week were fed to mice by gavage one month after DEN injection. Mice were sacrificed one, four and eight months after DEN injection. Representative Oil red o, F4/80 and Masson staining in mice livers and quantification, n = 6 (H). Scale bar: 200 or 50 μm. Statistical analyzes: Two-way ANOVA and Šídák's multiple comparison test. Representative gross appearance, HE staining and quantification of tumor number and maximum tumor size (I) of the livers eight months after DEN injection, n = 6. Arrowheads indicate tumors in liver. Scale bar: 2.5 mm or 200 μm. Statistical analyzes: Unpaired two-tailed Student's t-test. **: p < 0.01, ***: p < 0.001.

Figure 2.

Enterococcus faecalis promote MAFLD via downregulating IGSF9. A-D. E. faecalis enhanced MAFLD. MAFLD E. faecalis intervention co-housing model (A) was constructed using four-week male WT C57BL/6J mice. ABX were administered via sterile drinking water ad libitum and changed every day for two weeks. 5 × 10⁸ colony forming units (CFUs) of E. faecalis or PBS were fed to mice by gavage three times every week. Mice were co-housed together soon after gavage and were sacrificed 28 weeks after co-housing. Representative Oil red o (B), F4/80 (C) and Masson (D) staining in the mice livers and quantification. Scale bar: 200 or 50 μm. Statistical analyzes: One-way ANOVA and Dunnett's multiple comparison test. E, F. E. faecalis promote MAFLD progression in vitro. Representative HE, Oil red o and Sirius red staining (E) and supernatant TG quantification (F) of 3D liver spheroids under E. faecalis stimulating condition, n = 3. Scale bars: 20 μm. Statistical analyzes: Unpaired two-tailed Student's t-test. G. E. faecalis down-regulated IGSF9 among all top differential gut microbiota. Western blot analysis of IGSF9 protein in MAFLD 3D liver spheroids and primary hepatocytes stimulated with different bacteria culture medium. Experiments were repeated three times. H-K. IGSF9 was downregulated during MAFLD-HCC caused by E. faecalis colonization in vivo. Western blot analysis of IGSF9 protein in mice livers from MAFLD-HCC models, n = 6 (H, I), MAFLD-HCC E. faecalis intervention models, n = 6 (J) and MAFLD E. faecalis intervention co-housing models, n = 6 (K). Statistical analyzes: Two-way ANOVA and Šídák's multiple comparison test. One-way ANOVA and Dunnett's multiple comparison test. *: p < 0.05, **: p < 0.01, ***: p < 0.001.

Enterococcus faecalis-induced hepatic IGSF9 deficiency promoted MAFLD-HCC progression

IGSF9 has been reported negatively correlated with E. faecalis in colon cancer (R = −0.997, P = 0.003).³⁰ Given that our previous study also found a facilitating effect of IGSF9 deletion on breast cancer invasion and migration, we tempted to further explore whether IGSF9 also acted as a key molecule in MAFLD-HCC affected by E. faecalis. IGSF9 was predominantly expressed in tumor cells in several public HCC single-cell datasets (GSE125449, GSE140228, GSE146115, GSE146409, GSE166635, GSE98638) (Figure S3A, B). E. faecalis, among the top 7 bacteria significantly up/down-regulated in MAFLD-HCC, down-regulated IGSF9 after its co-culturing with MAFLD 3D model or primary hepatocytes (Figure 2G). Furthermore, western blot analyzes of liver samples from the above MAFLD-HCC and MAFLD models indicated that E. faecalis and HFD could both down-regulated IGSF9 expression (Figure 2H–K). Meanwhile, only live E. faecalis and its supernatant but not heat-killed bacteria downregulated IGSF9 expression, suggesting that bacterial metabolites mediate this effect (Figure S4).

Next, we studied the role of IGSF9 in WT and Igsf9-knockout (Igsf9^-/-) MAFLD-HCC mice models (Figure 3A). We collected the specimen from Igsf9^-/- mice and its controls for further analysis. Transcriptome sequencing was conducted on the livers of eight-month-old mice, followed by differential expression gene (DEG) analysis for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment. A significant enrichment of pathways associated with lipid metabolism, inflammation, and immune, and fibrosis was revealed in the Igsf9^-/- group (Figure S5A). The degree of macrovesicular steatosis detected by Oil red o staining was more severe in four-and eight-month Igsf9^-/- mice. Severe liver inflammation indicated by macrophage infiltration and heavier fibrosis level detected by Masson staining were also observed (Figure 3B). At eight months after DEN injection, Igsf9^-/- mice developed nearly two-fold higher number (17.83 vs 31.83, p = 0.0072) and the maximal diameters of tumor nodules (7.667 vs 12.83 mm, p = 0.0019) in liver than that in WT mice (Figure 3C), accompanied by the upregulation of Ki67 and AFP staining (Figure S5B). Meanwhile, eight-month Igsf9^-/- mice showed more severe dyslipidemia (higher CHOL, 7.505 vs 13.57 mmol/L, p = 0.0001; DLDL, 1.108 vs 2.657 mmol/L, p = 0.0007; TG, 0.665 vs 1.313 mmol/L, p = 0.0004) (Figure 3D) and liver function defects (higher ALP, 45.9 vs 71.6 U/L, p = 0.0301; ALT, 306.8 vs 627.8 U/L, p = 0.0062; and AST, 319.4 vs 545 U/L, p = 0.0028) (Figure 3E).

Figure 3.

IGSF9 deficiency promotes the MAFLD-HCC process. A–E. IGSF9 knockout promoted MAFLD-HCC. MAFLD-HCC and DEN-induced HCC mice models (A) were constructed using two-week male WT and Igsf9^-/- C57BL/6J mice injected with DEN and subsequently fed with high-fat diet (HFD) or not. Mice were sacrificed one, four, and eight months after DEN injection, n = 6. Oil red o staining, F4/80 staining and Masson staining (B) were used to assess MAFLD severity. Statistical analyzes: Two-way ANOVA and Šídák's multiple comparison test. Scale bar: 200 or 50 μm. Representative gross appearance and HE staining (C) of the livers eight months after DEN injection. Arrowheads indicate tumors in liver. Scale bar: 2.5 mm or 200 μm. Quantification of tumor number (left panel) and maximum tumor size (right panel) between WT and Igsf9^-/- mice were exhibited. Statistical analyzes: Unpaired two-tailed Student's t-test. D, E. Igsf9^-/- mice showed more severe blood lipid level and liver function. Serum CHOL, DHDL, TG (D), ALP, AST, and ALT (E) were up-regulated. Statistical analyzes: Two-way ANOVA and Šídák's multiple comparison test. F. Construction of 3D liver spheroid: Mice 3D NAC-liver spheroids were constructed using hepatocytes from WT or Igsf9^-/- mice and non-parenchymal cells from WT mice. G, H. Hepatocyte IGSF9 deficiency (Igsf9^-/-) group had more severe MAFLD. Enhanced HE, Oil red o, Sirius red staining (G) and up-regulated TG levels in the supernatant (H), n = 3. *: p < 0.05, **: p < 0.01, ***: p < 0.001.

We constructed 3D liver spheroid models with parenchymal hepatocytes from WT or Igsf9^-/- mice and nonparenchymal cells from WT mice (Figure 3F). With or without MAFLD induction, the Igsf9^-/- group had more severe levels of fatty infiltration and fibrosis, revealed by stronger Oil red o staining and Sirius red staining (Figure 3G), increased levels of TG (6.397 vs 7.443 nmol, p = 0.03; 8.533 vs 11.22 nmol, p = 0.0048) in the culture medium (Figure 3H), and increased gene markers at protein level (Figure S6A) and mRNA level (Figure S6B-D). Together, it suggested that E. faecalis-induced hepatic IGSF9 deficiency may further deteriorate MAFLD-HCC. These results indicated that HFD-induced E. faecalis colonization down-regulated hepatic IGSF9.

Loss of IGSF9 associates with poor prognosis in MAFLD-associated HCC

As HCC was the terminal of MAFLD, we validated the downregulation of IGSF9 in HCC in several datasets (GSE64041, GSE76311, GSE45436, GSE121248, GSE25097, and GSE76427) (Figure S7A). In addition, IGSF9 mRNA expression was markedly decreased in 72.9% (35/48) of HCCs, comparing with the adjacent normal tissues (Figure 4A) indicating by qRT–PCR in 48 pairs of frozen HCCs and adjacent normal tissues. Consistently, a significant downregulation of IGSF9 in HCC was observed by western blot analysis (Figure 4B). Then, 131 pairs of HCC specimens with matched adjacent normal tissues from patients underwent curative resection were collected for IGSF9 immunohistochemistry staining. The typical images showed a significant downregulation of IGSF9 in HCC compared with adjacent normal liver cells (Figure 4C–E). Patients were then divided into IGSF9 high and low expression group according to the IHC score of IGSF9 (high: 7–12; low: 0–6). Notably, low IGSF9 expression was significantly correlated with more advanced T staging [especially T3, 37 (28.2%) vs 10 (7.6%), p = 0.007], advanced tumor-node-metastasis (TNM) staging [especially stage III, 33 (25.2%) vs 8 (6.1%), p = 0.011], vascular invasion [39 (29.8%) vs 15 (11.5%), p = 0.013], higher GGT [74 (44, 132.5) U/L vs 48 (37, 69) U/L, p = 0.043] and liver cirrhosis history [75 (57.3%) vs 43 (32.8%), p < 0.001] (Figure 4F, Table 1). Also, HCC patients with low IGSF9 expression appeared worse overall survival (OS) [HR = 0.29 (0.17–0.51), p < 0.001] and relapse free survival (RFS) [HR = 0.49 (0.32–0.76), p = 0.0016] (Figure 4G). Furthermore, low IGSF9 expression was significantly correlated with MAFLD (p = 0.020) but not HBV history (Figure 4F, Table 1), which are common risk factors in liver cirrhosis.³¹ IGSF9 down-regulation in MAFLD-HCC were then validated both in GSE164760 dataset (Figure S7B) and clinical samples (Figure 4H). Together, these data indicated that decreased IGSF9 expression correlated with HCC malignancy and poor prognosis. It might be a valuable indicator for HCC prognosis, especially those with MAFLD.

Figure 4.

Decreased IGSF9 expression correlates with poor prognosis of HCC. A. The ratio (T/N) of IGSF9 mRNA was exhibited as log₂(T/N). Forty-eight paired HCC (T) and adjacent non-tumor tissues (N) from Zhongshan Hospital were used, n = 3. B. IGSF9 protein level was downregulated in HCC tissues. Paired HCC samples (n = 48) were analyzed by western blot. Experiments were repeated three times. Statistical analyzes: Paired two-tailed Student's t-test. C. Scores of IGSF9 immunohistochemistry staining in HCC specimens. Representative immunohistochemistry (IHC) staining intensity of IGSF9 in tumors and paired normal tissues. Scale bar = 200 or 50 μm. D, E. Quantification of IGSF9 IHC staining in specimens, n = 131. Statistical analyzes: Wilcoxon matched-pairs signed rank test. F. Pie chart showing clinical characteristics differences between IGSF9 high (n = 78) and low (n = 53) expression patients. G. Low IGSF9 expression was associated with significantly worse prognosis of HCC patients. Overall survival (OS) and Relapse-free survival (RFS) comparing low-expression group (n = 62) with high-expression (n = 40) of IGSF9 group were estimated using the Kaplan‒Meier method and compared by the log-rank test. H. Representative IHC staining intensity of IGSF9 in MAFLD-HCC and adjacent nontumorous tissue. Scale bar: 300 or 50 μm. ***: p < 0.001.

Table 1.

Correlation between IGSF9 expression and clinicopathologic characteristics of patients with HCC.

Characteristics	IGSF9 low expression	IGSF9 high expression	p value
n	78	53
Age, n (%)			0.476
≤60	56 (42.7%)	41 (31.3%)
>60	22 (16.8%)	12 (9.2%)
Sex, n (%)			0.345
Male	68 (51.9%)	43 (32.8%)
Female	10 (7.6%)	10 (7.6%)
T stage, n (%)			0.007
T1	8 (6.1%)	9 (6.9%)
T2	24 (18.3%)	28 (21.4%)
T3	37 (28.2%)	10 (7.6%)
T4	9 (6.9%)	6 (4.6%)
N stage, n (%)			1.000
N0	76 (58%)	52 (39.7%)
N1	2 (1.5%)	1 (0.8%)
M stage, n (%)			1.000
M0	74 (56.5%)	51 (38.9%)
M1	4 (3.1%)	2 (1.5%)
TNM stage, n (%)			0.011
I	8 (6.1%)	9 (6.9%)
II	27 (20.6%)	28 (21.4%)
III	33 (25.2%)	8 (6.1%)
IV	10 (7.6%)	8 (6.1%)
BCLC, n (%)			0.956
A	40 (30.5%)	26 (19.8%)
B	26 (19.8%)	19 (14.5%)
C	12 (9.2%)	8 (6.1%)
Vascular invasion, n (%)			0.013
No	39 (29.8%)	38 (29%)
Yes	39 (29.8%)	15 (11.5%)
Tumor differentiation, n (%)			0.266
Low/medium	49 (38%)	28 (21.7%)
High	28 (21.7%)	24 (18.6%)
Cirrhosis, n (%)			0.005
No	3 (2.3%)	10 (7.6%)
Yes	75 (57.3%)	43 (32.8%)
MAFLD, n (%)			0.020
No	60 (45.8%)	49 (37.4%)
Yes	18 (13.7%)	4 (3.1%)
HBsAg (< + .0 COI), n (%)			0.887
−	15 (11.6%)	11 (8.5%)
+	61 (47.3%)	42 (32.6%)
AFP (<20 ng/mL), median (IQR)	91.8 (8.725, 2871.8)	39.05 (5.4, 1122.8)	0.299
TB (3.4–20.4 μmol/L), median (IQR)	11.6 (9.275, 14.975)	13 (9.9, 16.95)	0.102
ALB (35–55 g/L), mean ± sd	40 ± 3.0811	41.02 ± 3.1719	0.075
ALT (7–40 U/L), median (IQR)	36.5 (25.75, 52.25)	34 (23, 52)	0.613
AST (13–35 U/L), median (IQR)	32 (24.75, 45.25)	29 (23, 39)	0.089
GGT (7–45 U/L), median (IQR)	74 (44, 132.5)	48 (37, 69)	0.043
PT (10–13 s), median (IQR)	11.7 (10.875, 12.3)	11.55 (10.975, 12.225)	0.852

Chisq test was applied for Age, Sex, T stage, TNM stage, Vascular invasion, Cirrhosis, MAFLD and HBsAg analyzes. Wilcoxon test was applied for AFP, TB, ALT, AST, GGT and Pt analyzes. Yates' correction was applied for N stage and M stage analyzes. T test was applied for ALB analysis. Bold items were considered statistically significant.

Loss of IGSF9 promotes C1q accumulation via inhibiting C1QC ubiquitination degradation

To further explore the molecular mechanisms underlying IGSF9 involved in MAFLD-HCC, we stably knocked down IGSF9 in Huh7 and HCCLM3 cells with relative higher IGSF9 protein levels. (Figure S8A, B). Transcriptome sequencing and KEGG enrichment pathway analysis was conducted using DEGs between WT and IGSF9 knock-down Huh7 cell line. “Complement and coagulation cascades”, the top 1 enriched pathway in shIGSF9 group, was also enriched in Igsf9^-/- mice liver (Figure 5A, Figure S5A). None of genes (C1QA, C1QB, C1QC, C1R, C1S, C2, C3, C4A and C4B) which represent in the classical pathway of complement showed significant correlation with IGSF9 in the TCGA-LIHC dataset (Figure S9A).

Figure 5.

IGSF9 promotes C1QC ubiquitination by enhancing its interaction with NEDD4/UBE3A. A. Pathway “Complement and coagulation cascades” (red) was the top 1 enriched pathway between Huh7 IGSF9-knockdown and ctrl group. KEGG pathways are enriched based on differentially expressed genes (DEGs). Top 15 pathways are listed. B-F. IGSF9 interacted with C1QC. LC-MS/MS analysis of IGSF9 binding proteins (B, C), co-localization using immunofluorescence (D) and CO-IP assays (E) of endogenous interaction between IGSF9 and C1QC in Huh7 cells and exogenous interaction in HEK293T cells transfected with Flag-IGSF9 and HA-C1QC-expressing plasmids (F) were performed. Fluorescence intensity was analyzed by ImageJ. Scale bar: 10 μm. G, H. IGSF9 intracellular domain interacted with C1QC. Myc-IGSF9 extracellular domain (ECD)/Myc-IGSF9 intracellular domain (ICD) and HA-C1QC-expressing plasmids were transfected into HEK293T cells (G). Co-IP was conducted by anti-HA (H). I-J. IGSF9 promoting C1QC degradation by enhancing ubiquitination. C1QC protein levels in primary hepatocytes from WT and Igsf9^-/- mice after treated with 20  μg/mL CHX for 2, 4, 6 and 8 h (I); in Huh7 and HEK293T cells with/without proteasome inhibitor MG132 (5 μM, 6 h) or lysosome inhibitor NH₄Cl (30 mM, 6 h) treatment (J) were detected using western blot. Statistical analyzes: Two-way ANOVA and Šídák's multiple comparison test. K. IGSF9 intracellular domain enhanced ubiquitination of C1QC. Myc-IGSF9 intracellular domain (ICD), Flag-Ub and HA-C1QC-expressing plasmids were transfected into HEK293T cells. Co-IP was conducted by anti-HA. L-P. IGSF9 enhanced interaction between C1QC and NEDD4/UBE3A. LC-MS/MS analysis (L) showing IGSF9 binding proteins NEDD4 and UBE3A. CO-IP assays of endogenous interaction between IGSF9/C1QC with NEDD4/UBE3A in Huh7 cells (M) and exogenous interaction in HEK293T cells transfected with Myc-IGSF9 (ICD)/HA-C1QC and Flag-NEDD4/Flag-UBE3A-expressing plasmids (N) were performed. Their interaction was also detected with/without IGSF9-overexpression (O) or Myc-IGSF9 (ICD)-expressing plasmids transfection (P). ***: p < 0.001.

However, we identified C1QC as a potential IGSF9 interacting proteins by Co-immunoprecipitation coupled with mass spectrometry (Co-IP/MS) (Figure 5B,C). Immunofluorescence exhibited co-localization of IGSF9 and C1QC (Figure 5D). Endogenous and exogenous interaction between IGSF9 and C1QC were detected by co-immunoprecipitation assay in Huh7 cells and HEK-293T cells, respectively (Figure 5E,F). In addition, Myc-tag IGSF9 intracellular (756 aa.–1179 aa.) but not extracellular (21 aa.–755 aa.) fragment bound with C1QC (Figure 5G,H). Specifically, IGSF9 intracellular domains (Q896, E901, G921, N936, V937, D938, E946) were predicted to interact with C1QC domains (Y33, Y114, K115, Q116, Q119, G232) (Figure S10). C1QC is a component of the C1q complex. The stability of C1QC protein in primary hepatocytes from WT and Igsf9^-/- mice or with/without E. faecalis stimulation was detected. CHX chase experiments showed that C1QC protein was more stable in hepatocytes isolated from Igsf9^-/- mice and in hepatocytes stimulated with E. faecalis (Figure 5I, Figure S11A). While the expression of C1q genes (C1qa, C1qb, C1qc) was not altered (Figure S9B). To further investigate the mechanism by which IGSF9 influences the stability of the C1QC protein, we inhibited the ubiquitin-proteasomal degradation pathway and the lysosomal degradation pathway using MG132 and NH₄Cl, respectively. Our findings indicate that IGSF9, particularly its intracellular domains, facilitates the degradation of C1QC via a ubiquitin-proteasomal pathway-dependent mechanism (Figure 5J). CO-IP experiments corroborated that the introduction of IGSF9 intracellular domains resulted in increased levels of C1QC ubiquitination (Figure 5K). Additionally, mass spectrometry analysis identified two E3 ubiquitin ligases, NEDD4 and UBE3A, that interact with IGSF9 (Figure 5L). We validated the interactions between IGSF9, including its intracellular domains, and both ligases through endogenous/exogenous CO-IP and GST-pull down experiments, and confirmed the interaction of C1QC with both ligases (Figure 5M,N; Figure S11B, C). These findings suggest the existence of IGSF9-NEDD4-C1QC and IGSF9-UBE3A-C1QC complexes that may mediate the ubiquitination and subsequent degradation of C1QC. Furthermore, we observed that the presence of IGSF9 or its intracellular domains enhances the interaction of C1QC with NEDD4 or UBE3A (Figure 5O,P). The molecular docking structures showed more interaction sites between C1QC with NEDD4 or UBE3A when IGSF9 appeared (Figure S11D), which validated previous results.

Would IGSF9, through its association with C1QC, affect intracellular C1q produced by hepatocytes? We found C1q accumulation in MAFLD-HCC mice liver from E. faecalis and Igsf9^-/- group (Figure 6A) and the 3D liver spheroid culture medium from E. faecalis and Igsf9^-/- group (Figure 6B). C1q has been reported to exacerbate fatty liver disease and liver cirrhosis.^32,33 In the oleic acid (OA)-induced HepG2 fatty liver model, the number of lipid droplets increased in the cells upon C1q treatment (Figure 6C,D). In a model of TGF-β-induced LX2 fibrosis, the addition of C1q enhanced the level of fibrosis indicated by elevated fibrosis marker genes expression and α-SMA protein (Figure 6E,F). We found the negative correlation between IGSF9 and C1q and the positive correlation between C1q and markers of fatty liver disease and liver cirrhosis by immunohistochemistry in 22 MAFLD-HCC samples (Figure 6G). These results suggest that C1q accumulation derived from IGSF9 deficiency can exacerbate the development of MAFLD.

Figure 6.

C1q exacerbates hepatic steatosis, fibrosis and inflammatory processes. A-B. E. faecalis and loss of IGSF9 led to liver C1q accumulation. C1q content in mice liver (n = 6) and in 3D liver spheroid supernatant (n = 3) from previous experiments were detected using IHC (A) and ELISA (B). Statistical analyzes: Two-way ANOVA and Šídák's multiple comparison test; unpaired two-tailed Student's t-test. C-F. C1q enhanced steatosis and fibrosis in vitro. Oil red o staining of HepG2 cells after stimulated with oleic acid (OA, 20 μg/mL) with/without C1q (1 μg/mL) for 24 h, n = 3 (C, D). Expression of α-SMA protein level (E) and fibrosis marker genes (COL1A1, ACTA2, TIMP1) (F) and in LX2 cells after stimulated with TGF-β1 (5 ng/mL) with/without C1q (1 μg/mL) for 24 h, n = 3. Statistical analyzes: One-way ANOVA and Dunnett's multiple comparison test; two-way ANOVA and Šídák's multiple comparison test. G. C1q was negatively correlated with IGSF9 and was positively correlated with FABP4 and α-SMA in HCC samples. Representative IGSF9, C1q, FABP4 and α-SMA staining and correlation analysis in 22 MAFLD-HCC specimens. *: p < 0.05, **: p < 0.01, ***: p < 0.001.

Targeting E. faecalis/IGSF9/C1q axis inhibited MAFLD-HCC progression

Based on our discovery of the E. faecalis/IGSF9/C1q axis, we would like to find early intervention targets for MAFLD-HCC. In view of the mutual inhibitory effect among the microbiota, we attempted to add the above 5 bacteria decreased in DEN/HFD mice gut along with E. faecalis to co-culture with MAFLD 3D model or primary hepatocytes. Only B. pseudolongum was able to reverse the E. faecalis-induced IGSF9 down-regulation (Figure 7A). We then tested the effects of B. pseudolongum on MAFLD mouse model (Figure 7B). Early intervention with B. pseudolongum in MAFLD mice could relieve hepatic macrovesicular steatosis, inflammation and fibrosis level early in the disease (Figure 7C). Considering that the IGSF9/C1q complex functions as an effector molecular complex influenced by E. faecalis, we have conducted preliminary investigations into the potential mitigating effects of overexpressing IGSF9 and the C1q neutralizing antibody on MAFLD in vitro or in vivo. Lipid droplets in HepG2 fatty liver model decreased after elevated IGSF9 expression (Figure 7D,E), showing the therapeutic potential of manipulating IGSF9 in MAFLD. In MAFLD mice model, early intervention of C1q neutralizing antibodies could also relieve disease progression (Figure 7F,G). These results highlighted the importance of targeting E. faecalis/IGSF9/C1q axis in MAFLD-HCC.

Figure 7.

Targeting Enterococcus faecalis/IGSF9/C1q axis has promising therapeutic effect on MAFLD. A. Bifidobacterium pseudolongum (B. pseudolongum) rescued IGSF9 downregulation caused by E. faecalis in vitro. Western blot analysis of IGSF9 protein in MAFLD 3D liver spheroids and primary hepatocytes stimulated with culture medium from bacteria lacking in DEN + HFD group on the basis of E. faecalis. Experiments were repeated three times. B, C. B. pseudolongum intervention alleviates MAFLD. MAFLD B. pseudolongum intervention model (B) was constructed based on the MAFLD model as previously described. 1 × 10⁸ CFUs of B. pseudolongum or PBS (Ctrl) were fed to mice by daily gavage. Mice were sacrificed 28 weeks after intervention. Representative Oil red o, F4/80 and Masson staining in mice livers and quantification (C) showed alleviation after B. pseudolongum intervention, n = 6. Scale bar: 200 or 50 μm. Statistical analyzes: Unpaired two-tailed Student's t-test. D, E. Overexpressing IGSF9 alleviated steatosis in vitro. Oil red o staining (D) and quantification (E) of HepG2 cells overexpressing IGSF9 or not after stimulated with OA for 24 h, n = 3. Statistical analyzes: One-way ANOVA and Dunnett's multiple comparison test. F, G. C1q neutralizing antibody alleviated hepatic steatosis, fibrosis and inflammatory processes in vivo. MAFLD C1q neutralizing antibody intervention model was constructed based on the MAFLD model subsequently injected with 10 μg C1q antibody two times a week (anti-C1q) or mouse IgG2b isotype (Ctrl) (F). Representative Oil red o, F4/80 and Masson staining in the livers and quantification, n = 6 (G). Scale bar: 200 or 50 μm. Statistical analyzes: One-way ANOVA and Dunnett's multiple comparison test. *: p < 0.05, **: p < 0.01, ***: p < 0.001.

Discussion

Our study identifies a novel molecular mediator, IGSF9, between the gut microbiota and members of complement system in the progression of MAFLD to HCC. As upstream stimulator, E. faecalis decreases IGSF9 levels, leads to less interaction between IGSF9 and C1QC, thereby decreasing C1QC ubiquitination degradation and accumulating C1q. C1q subsequently promoting liver fatty infiltration, inflammation, fibrosis, and the progression to HCC as well. B. pseudolongum, when used for 28 weeks, was beneficial for the recovery of abnormal liver pathologies and functions in mouse models, which offers potential microbiota-targeted therapies for future MAFLD management.

Dysbiosis of gut microbiota increases the risk of HCC in many patients with chronic liver diseases.^7,34 Previous studies have shown that E. faecalis promotes liver tumorigenesis and accelerates colorectal cancer progression.^28,35,36 Microbiota analysis suggests that a negative correlation between the colonization of E. faecalis and IGSF9 expression in colorectal cancer (N = 4, R = −0.997, p = 0.003).³⁰ Here we performed a serial of experiments to demonstrate that E. faecalis downregulates IGSF9 expression in liver. Tumorigenesis is often accompanied by gut barrier damage, leading to increased translocation of microbial metabolites into the bloodstream, the impact of gut microbiota on disease may be closely related to both the intestinal barrier and metabolites.³⁷ In line with that, several studies have shown that E. faecalis and its metabolite bilirubin promote colorectal and liver cancer progression.³⁶ Whether bilirubin mediates E. faecalis reduced IGSF9? It needs further investigation. A healthy gut microbiota (such as B. pseudolongum) has shown to improve conditions such as colitis, colorectal cancer, and liver cancer.^11,38-40 Acetate produced by B. pseudolongum reaches the liver via the portal vein, activates G protein-coupled receptor 43 (GPR43), and inhibits the IL-6/JAK1/STAT3 signaling pathway.^11,41 We speculate that gut microbiota likely regulates IGSF9 expression via their metabolites which enters liver through portal venous system. Supplementation with beneficial gut microbiota is emerging as a promising approach for treating microbiota-related disorders. In mouse models, B. pseudolongum has shown potential therapeutic effects on MAFLD-HCC progression, intestinal inflammation, and antiviral responses.^11,38,41 Furthermore, formulations of B. pseudolongum have been developed to alleviate ulcerative colitis and acute liver injury.⁴² Together with our findings, it is potentially a therapeutic approach for MAFLD management.

The role of IGSF9 in cancer development appears to be tissue context specific. For instance, in nasopharyngeal carcinoma, IGSF9 may be induced by hypoxia and promotes cancer malignancy through Akt signaling pathway.⁴³ In contrast, a study on breast cancer revealed that IGSF9 is transcriptionally activated by TP53, and inhibits epithelial-mesenchymal transition (EMT) as a suppressor of Akt pathway.²⁰ Additionally, due to the potential ITIM motif in the cytoplasmic tail, IGSF9 has been considered as immune-suppressive receptor in lung cancer, monocytic leukemia and colorectal cancer etc.¹⁵ Most studies on IGSF9 have focused its role in cancer progression, with few research on its involvement in the diseases early onset or cancer initiation. Other than increased macrophage infiltration in the Igsf9^-/- MAFLD-HCC mouse model, classical immune suppression was not observed, indicting a more comprehensive role of IGSF9 in diseases. Here, we point out its role in progression of MAFLD to HCC. IGSF9 is seemly far beyond a cell adhesion molecule and immune suppressor receptor. It needs further investigation.

The complement system plays a critical role in immune regulation within the body. C1q is the initiating component of the classical pathway, capable of recognizing and binding antibodies, which in turn activates other complement components such as C1r and C1s, leading to the formation of the membrane attack complex (MAC) and mediating complement-dependent cytotoxicity.⁴⁴ Abnormal complement levels often reflect disease states. In autoimmune diseases, C1q is commonly depleted due to activation of the complement pathway. Conversely, several infectious and inflammatory diseases associate with abnormal accumulation of C1q.⁴⁵ For instance, C1q accumulation has been observed in liver cirrhosis, chronic hepatitis, and alcohol-related liver disease.^32,45 Additionally, C1q-ApoE complexes have been identified as novel biomarkers in viral hepatitis and MAFLD.³³ We revealed novel IGSF9-C1QC complexes associated with the progression of MAFLD. The cytoplasmic tail of IGSF9 is essential for C1QC binding. This binding enhances E3 ubiquitin ligases NEDD4 and UBE3A interaction with C1QC, facilitates degradation of C1QC. More C1QC in C1q, likely more stabilized C1q. Accumulated C1q in liver promotes progression of liver diseases. Therefore, overexpressing IGSF9 by adeno-associated virus (AAV), activating its downstream signaling pathways, or C1q-neutralizing antibodies may rebalance IGSF9-C1q, subsequently intervening in MAFLD progression. Although preliminary in vivo and in vitro results turned out to be good, the effectiveness and detailed protocols need to be further optimized and elucidated.

Collectively, IGSF9 could be a promising target as primary prevention in patients with MAFLD or as secondary prevention in patients with MAFLD-HCC. The downregulation of IGSF9 by E. faecalis suggests a feasible early intervention for high-risk patients with MAFLD.

Funding Statement

This work was supported by the National Natural Science Foundation of China (NSFC) (82173308, 82472669) and the Natural Science Foundation of Shanghai Municipality (#22ZR1411800, #23ZR1412600).

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Author contributions

SC designed and conceived this project. QJ and WT developed methodology. QJ, JL, SW, and LZ performed experiments and generated data. QJ and SC analyzed and interpreted data. LD, WT, and SW provided clinical samples and the related patients' information. LD made suggestions for this project. QJ and JL wrote the draft. SC revised and finalized the manuscript. All authors read and approved the manuscript.

Data availability statement

Gut microbiota 16S rRNA sequencing data is available in SRA at https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1273690.

Ethics approval statement

This study was approved by the Institutional Ethics Committee of Fudan University (2021-008). All procedures involving human data and biological samples were conducted in accordance with the Declaration of Helsinki. Written informed consent was waived by the ethics committee due to the retrospective design and full anonymization of data, which posed no additional risk to participants.

Animal experiments were approved by the Research Ethics Committee of Zhongshan Hospital and performed in compliance with the NIH Guide for the Care and Use of Laboratory Animals and the ARRIVE guidelines.

Abbreviations

MAFLD

metabolic associated fatty liver disease

NASH

nonalcoholic steatohepatitis

HCC

hepatocellular carcinoma

IGSF9

immunoglobulin superfamily member 9

C1q

complement C1q

NEDD4

E3 ubiquitin-protein ligase NEDD4 or Neural precursor cell expressed developmentally down-regulated protein 4

UBE3A

Ubiquitin-protein ligase E3A

FABP4

fatty acid-binding protein 4

α-SMA

alpha smooth muscle actin

Ki67

proliferation marker protein Ki67

CHOL

cholesterol

TG

triglycerides

HFD

high fat diet

DEN

diethylnitrosamine

RFS

relapse-free survival

OS

overall survival

CHX

cycloheximide

TCGA

the cancer genome atlas

AAV

adeno-associated virus