The single-cell RNA sequencing data generated in this study are available at the GeneExpression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE218561), RNAsequencing data are available from the corresponding author on reasonable requestOther relevant data are within the manuscript and its Additional files. The “Seurat” R package was used for systematic analysis of this dataset. The analysis workflow primarily included the following steps: First, the CreateSeuratObject function was used to construct the Seurat object, with the initial filtering criteria set as follows: each gene must be expressed in at least three cells, and at least 200 genes must be detected in each cell subsequently, strict quality control was performed to retain high-quality cells with feature gene counts between 200 and 6,000 and mitochondrial gene proportions below 25%. Principal component analysis (PCA) was performed on the top 2,000 highly variable genes, and the top 10 principal components were selected for subsequent analysis using the ElbowPlot function. The FindClusters algorithm (with a resolution parameter of 0.5) was used for cell subpopulation identification, combined with t-distributed stochastic neighbor embedding (t-SNE) for dimensionality reduction and visualization. Finally, biological annotation of cell clusters was performed using known marker genes that successfully identified multiple cell types, including T cells, B cells, epithelial cells, mononuclear phagocytes, fibroblasts, endothelial cells, and NK cells.

InferCNV analysis was performed to detect large-scale somatic chromosomal copy number alterations (CNAs) using reference-normalized gene expression patterns (InferCNV; https://github.com/broadinstitute/InferCNV). We selected 1,000 endothelial cells, 1,000 fibroblasts, 2,000 mononuclear phagocytes, and 2,000 T/NK cells as reference controls to establish baseline expression. The inferCNV object was created using CreateInfercnvObject, and the analysis was conducted with the following default parameters: an expression cutoff of 0.1 (excluding lowly expressed genes), clustering by biological groups (cluster_by_groups = TRUE), denoising (denoise = TRUE), and without Hidden Markov Model smoothing (HMM = FALSE).

We assessed the cellular differentiation state and stemness potential of malignant cells using the CytoTRACE algorithm, which predicts developmental potential based on gene expression signatures. We applied ScanoramaCT before analysis to mitigate batch effects across datasets. CytoTRACE scores were computed for all tumor cells using R, with values ranging from 0 (fully differentiated, low stemness) to 1 (least differentiated, high stemness), enabling the systematic characterization of the stemness hierarchy within the TME.

We performed Gene Set Variation Analysis using the GSVA package (v1.30.0) to evaluate differences in pathway activity among distinct cell types. Subsequently, the differential pathway activity across cell types was statistically assessed using the LIMMA package (v3.38.3), which employs an empirical Bayes approach to enhance the differential expression analysis.

We used the Cellchat software package to infer interactions between different epithelial cell subpopulations and other cells in rat liver cancer tissues and predicted the strength of interactions between cells based on the ligands and receptors expressed by the cells.

Male SD rats (6–8 weeks old, weighing 180–200 g) were purchased from Jihui Laboratory Animal Breeding Company, Ltd. (Shanghai, China). The AIN93M standard feed used in this study was purchased from Trophic Animal Feed High-tech Co., Ltd. (Nantong, China). The rats were kept in an animal facility free of any pathogens and fed DEN carcinogenic solution (0.95 g/mL diluted 100 ppm in triple-distilled water, MeilunBio^®, China) daily for constructing a primary rat HCC model. 8 μmol/100 g GCDCA (Sigma-Aldrich, USA) was intraperitoneally injected twice a week starting from 12 weeks of DEN treatment, and 1.2% cholestyramine (MedChemExpress, USA) was fed to remove excess GCDCA in vivo. JTE-013 was injected intraperitoneally 2 h before GCDCA administration at a dose of 2 mg/kg body weight in the JTE-013 (TargetMol, China) interference model, starting at week 12 of DEN treatment. Each group consisted of 10 rats. All animal experiments were approved by the Animal Care Committee of the Naval Medical University. The ethics approval number was EDWLL-018.

Serum activities of alanine aminotransferase (ALT), aspartate aminotransferase (AST), and total bile acid (TBA) levels were analyzed using ALT, AST, and TBA assay kits (Rayto, China), respectively, according to the manufacturer’s instructions of the reagent vendors.

Sections were placed on a 60 °C roaster for 30 min and deparaffinized in water. Hematoxylin staining solution (Servicebio, China) was used for 8 min, and the sections were rinsed with water until they turned blue. After hydrochloric acid–alcohol differentiation, the slides were rinsed using deionized water, restained with eosin (Servicebio, China) for 2 s, rinsed with tap water, and dried. Finally, the slides were sealed with neutral gum and imaged under a microscope. Each experiment was conducted in triplicate.

The slides were placed on a 60 °C roaster for 30 min, and sequentially dewaxed to water by xylene and gradient ethanol. An appropriate repair method was selected based on the type of antibody detected. The slides were washed thrice with PBS for 5 min each, and the tissues were labeled by drawing a circle with an immunohistochemical pen and closed for 30 min at room temperature using 3% H₂O₂ to block endogenous peroxidase. The slides were blocked with 3% BSA (Servicebio, China) for 30 min at 37 °C. After completion of primary antibody dilution with a universal antibody diluent (Epizyme, China), the primary antibody was added dropwise to the tissue block and incubated overnight at 4 °C in a wet box. The next day, the slides were removed from the refrigerator at 4 °C, rewarmed at room temperature for 10 min, the primary antibody was removed, and the slides were washed thrice with PBS for 5 min. The general secondary antibody was added dropwise to the tissue block, and the slides were incubated at 37 °C for 30 min. Slides were washed thrice with PBS after the incubation was completed, each time for 5 min. After completion of washing, the DAB color development solution (Genetech, China) was prepared for DAB development. We prepared trays containing tap water for the slides to terminate staining. The slides were re-stained with hematoxylin (Servicebio, China) for 6 min, rinsed under running water for several seconds, differentiated with hydrochloric acid–alcohol differentiation solution twice, each time for 1–2 s, and rinsed with running water until the blue color returned. The neutral resin was dropped onto the slide tissue block and sealed, allowing the slides to dry. Images were acquired using a microscope the next day for observation. The first day of immunofluorescence staining was the same as that of immunohistochemistry. A drop of the staining working solution was added to the tissue block on the second day, incubated at room temperature, protected from light for 10 min, washed with PBS, and the nuclei were stained blue with DAPI (Servicebio, China), incubated at room temperature, and protected from light for 10 min. Next, the tissue blocks were incubated with a tissue autofluorescence quencher (Servicebio) for 5 min at room temperature, protected from light, rinsed under running water, and sealed with drops of an anti-fluorescence quenching sealer (Servicebio). Images were captured using a single-photon confocal microscope (Stellaris 5; Leica, Germany). At least three random areas were selected for each slide to capture images. The following antibodies were used for immunohistochemistry: EPCAM (1:100, ab71916; Abcam), KRT19 (1:4000, 10712-1-AP; Proteintech), SOX9 (1:2000, ab185230; Abcam), CD163 (1:4000, ab213612; Abcam), S1PR2 (1:100, sc-365963; Santa Cruz Biotechnology), and CD68 (1:3000, ab31630; Abcam). The following antibodies were used for immunofluorescence: EPCAM (1:100, ab71916; Abcam), KRT19 (1:4000, 10712-1-AP; Proteintech), SOX9 (1:2000, ab185230; Abcam), CD163 (1:4000, ab213612; Abcam), CD206 (1:500, 18704-1-AP; Proteintech), and S1PR2(1:100, sc-365963; Santa Cruz Biotechnology). The quantitative analysis of immunohistochemical staining sections was performed using ImageJ 1.43. Each experimental group consisted of 3 slides, and 3 fields were randomly selected from each slide. Ultimately, the proportion of positive area for each experimental group was the average value of all the calculation results from the fields in that group.

The cells used in the experiments were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). The THP-1 cell line was prepared as RPMI 1640 complete medium using RPMI 1640 medium (BasalMedia, China) with 0.05 mM β-mercaptoethanol (Gibco, USA), 10% fetal bovine serum (FBS; Gibco, Invitrogen), and 1% penicillin and streptomycin (PS; YEASEN, China). F12K medium (Gibco, USA) was used for the NR8383 cell line, and DMEM (BasalMedia, China) was used for Huh7, RH35, and HTLA cell lines. Complete medium was prepared by adding 10% FBS and 1% PS to all cell lines, except the NR8383 cell culture medium, which was prepared by adding 15% FBS and 1% PS. All cells were incubated at 37 °C in an incubator containing 5% CO₂.

The THP-1 cells were treated with 100 ng/mL PMA (MedChemExpress, USA) for 48 h. Fresh complete medium was replaced, and 20 ng/mL IL-4 (PeproTech, China) and IL-13 (TargetMol, China) were added to induce the polarization of THP-1 cells toward M2-type macrophages, and the cells were treated with GCDCA at a concentration of 25 μM for 24–48 h. The JTE-013 inhibitor (10 μM) was added 30 min before the addition of GCDCA. The NR8383 cells were treated with 25 μM GCDCA for 24–48 h and used for subsequent experiments. The JTE-013 inhibitor (10 μM) was added 30 min before GCDCA administration. The concentration of the inhibitor ketoconazole was 20 μM and was added 1 day before the addition of GCDCA. The inhibitor MeTC7 (5 μM) was added 1 day before the addition of GCDCA. Next, we co-cultured Huh7 cells with THP-1 cells, for which THP-1 cells were spread in the upper chamber using a 24 mm co-culture plate (Corning, USA) and treated with 100 ng/mL PMA for 48 h. Next, IL-4 and IL-13 (20 ng/mL) were added to induce macrophage polarization to M2.

macrophages, and 10 μM JTE-013 inhibitor and 25 μM GCDCA were administered according to the subgroups selected. After the polarization was completed, the cells were co-cultured with the human HCC cell line Huh7 in the lower chamber for 72 h. Next, we co-cultured RH-35 cells with NR8383 cells, for which NR8383 cells were spread in the upper chamber using a 24 mm co-culture plate (Corning, USA), and 10 μM JTE-013 inhibitor and 25 μM GCDCA were administered according to the grouping selection, and co-cultured with rat liver cancer cell line RH35 in the lower chamber for 72 h.

Each well was inoculated with 10⁴ cells in a 96-well plate (Corning, Corning, NY, USA), mixed using the cross method, and placed in an incubator at 37 °C with 5% CO₂. After the cells attached to the wall, the old medium was aspirated. The cells were washed with PBS and treated with GCDCA for 24–48 h according to the concentration gradient set in the experiment. The supernatant was removed and replaced with 10% CCK8 solution (10 µL CCK8 + 90 µL medium; Servicebio, China), and incubated for 2 h at 37 °C in an incubator. The absorbance was measured at 450 nm using an enzyme marker (Zenyth1100). Cell viability was calculated using the following formula: cell viability (%) = (absorbance value of experimental group–absorbance value of blank group)/(absorbance value of control group–absorbance value of blank group) × 100%. Each experiment was conducted in triplicate.

Each well was inoculated with 500 cells in a six-well plate (Corning, USA) and placed in an incubator at 37 °C with 5% CO₂. After the cells were attached to the wall on the second day, the cells were administered GCDCA and incubated at 37°C in an incubator containing 5% CO₂ for 7 days, and the cell status was observed every day. After the cell colony grew to the appropriate size, the six-well plate was removed, the supernatant was discarded, the cells were washed twice with PBS, 1 mL of methanol (Macklin, China) was added to each well, and the plate was fixed for 15 min at room temperature. The methanol was removed, and the cells were stained with 0.5% crystal violet staining solution (Servicebio, China) for 30 min at room temperature. The six-well plate was gently rinsed with tap water to remove the excess crystal violet staining solution. The six-well plate was dried at room temperature, and images were acquired. The number of clones in each group was quantified using ImageJ 1.43. Each experiment was conducted in triplicate.

The cell culture medium was prepared by adding 50 mL DMEM/F12 (Gibco, USA), 1 mL 1×B27 (Gibco, USA), 20 ng/mL EGF (MedChemExpress, USA), 10 ng/mL bFGF (TargetMol, China), and 0.4% BSA (Servicebio, China) to the medium. The medium was filtered with a 0.22 μm microporous filter membrane (Merck, Germany) and set aside. Each well was inoculated with 10³ cells in a low adsorption six-well plate (Corning). Next, 25 μM GCDCA was administered after the cells were stabilized on the next day. The cells were incubated in an incubator at 37 °C, 5% CO₂ for 7 days, and observed daily, and imaged under a microscope after the cells became spherical. Each experiment was conducted in triplicate.

The lysate was prepared according to the ratio of PMSF protease inhibitor (100×, Servicebio, China): RIPA tissue cell rapid lysate (Servicebio) = 1:100; 200 μL of the protein lysate was added to each well of the six-well plate (Corning), and the plate was lysed on ice for 40 min. The cell lysate was collected and centrifuged for 15 min at 12,000 rpm at 4 °C. The supernatants were collected. The BCA working solution (Beyotime, China) was prepared, following which 25 µL of the supernatant was taken and mixed with 200 µL of the working solution (supernatant:working solution = 1:8), incubated at 37 °C for 30 min in the dark. Afterward, an enzyme meter (Zenyth1100) was used to measure the absorbance value of the samples at 595 nm, and subsequently, the concentration of proteins according to the standard curve. The supernatant 1/4 volume of 5× SDS-PAGE protein sampling buffer (Beyotime, China) was added, mixed well, and the protein was heated at 100 °C for 10 min to completely denature the protein. The precast gel (GenScript, China) was removed, the protein concentration was calculated, and electrophoresis was performed at 170 V for 45 min. A PVDF membrane (Merck, Germany) was activated in methanol and subsequently placed in a membrane equilibrium solution (GenScript) for 2 min. The gel was washed with deionized water, and the sponge, PVDF membrane, gel, and sponge were sequentially placed in a transfer folder to transfer the membrane. After transfer, the membrane was placed in a protein-free rapid sealing solution (Servicebio) for 15 min at room temperature, and the primary antibody was incubated overnight. The next day, the membrane was removed, the strips were washed with 1×TBST (Servicebio), and the corresponding secondary antibody (Servicebio) was incubated with the primary antibody source. After incubation for 90 min at room temperature, the strips were washed again, and the chemiluminescent horseradish peroxidase substrate (Merck, USA) was mixed in a 1:1 ratio, immersed in the mixture for a few seconds, and developed using a Tanon automated chemiluminescent image analyzer system (Tanon-4600). Each experiment was repeated thrice. The following antibodies were used in the experiments: GAPDH (BioWord, AP0063, 1:5000), EPCAM (1:1000, ab71916, Abcam), KRT19 (1:3000, 10712-1-AP, Proteintech), SOX9 (1:1000, ab185230, Abcam), CD163 (1:1000, ab213612, Abcam), ARG1 (1:5000, 16001-1-AP, Proteintech), CD206 (1:500, 18704-1-AP, Proteintech), AKT(1:1000, 4685S, CST), p-AKT(1:2000, 66444-1-Ig, Proteintech), PI3K(1:5000, 60225-1-Ig, Proteintech), and p-PI3K (1:5000, AF3242, Affinity).

Total RNA was extracted from the cells using a HiPure Total RNA Plus Micro Kit (MGBio, China) according to the manufacturer’s instructions of the reagent company. The cDNA was synthesized using Hifair^® V One-Step RT-gDNA Digestion SuperMix for qPCR Kit (YEASEN, China). The total reaction system was 20 µL, including 15 µL of RNase-free H₂O and total RNA, 4 µL 5× Hifair^® One-Step RT SuperMix (YEASEN, China), and 1 µL gDNA Remover Mix (YEASEN, China). The cDNAs obtained from this synthesis were used for subsequent PCR experiments. Real-time fluorescence quantitative PCR was performed using the Hieff UNICON^® Universal Blue qPCR SYBR Green Master Mix (YEASEN, China) as described by the manufacturer’s instructions. The total reaction system was 20 µL, including 8 µL of RNase-free H₂O, 1 µL of cDNA, 0.5 µL of forward primer, 0.5 µL of reverse primer, and 10 µL of the SYBR Green Master Mix (YEASEN, China). The reaction procedure was standardized: 2 min of pre-denaturation at 95 °C (1 cycle), followed by denaturation at 95 °C for 10 s (40 cycles), annealing and extension at 60 °C for 30 s (40 cycles), and finally melting curve analysis (1 cycle). Information on the primers used in the experiments is provided in the Supplementary Material. Gene expression was quantified using the 2^-ΔΔCt method. The gene expression values were normalized to those of GAPDH. Each experiment was repeated thrice.

THP-1 and NR8383 cells were seeded in six-well plates at a density of 1 × 10⁶ cells/well. THP-1 cells were treated with PMA and pre-treated with GCDCA for 24–48 h after adhering to the plate. Subsequently, the cell culture supernatants were collected, and the concentrations of IL-6 (Thermo Fisher Scientific, USA), TNF-α (Thermo Fisher Scientific), IL-10 (Thermo Fisher Scientific), TGF-β (Cloud-Clone Corp, China), and VEGF (Multi Science, China) in the supernatants were determined according to the manufacturer’s instructions. Each sample was subjected to three repetitions of the experiment.

THP-1 cells were collected during the logarithmic growth phase by centrifugation at 1000 rpm for 5 min. The cell precipitate was collected and washed twice with precooled PBS. The cells were dissolved in PBS containing an Fc receptor blocker (BD, USA), incubated at 4 °C for 15 min to block non-specific binding sites. The cells were dissolved in 100 µL of pre-cooled PBS, and fluorescently labeled CD163 (Biolegend, USA) and CD206 antibodies (Biolegend, USA) were added, incubated at 4 °C in the dark for 30 min, and subsequently washed thrice with pre-cooled PBS. After the incubation, the cells were suspended in 100 µL of Cyto-Fast™ Fix/Perm cell fixation and membrane disruption solution (Biolegend) at room temperature for 20 min. Next, 1 mL of 1× Cyto-Fast™ Perm washing solution (Biolegend) was added and the above antibody incubation steps were repeated to detect anti-CD163 and CD206 antibodies in the cells. The cells were resuspended in PBS in a flow cytometry tube and detected using a flow cytometer (LSRFortessa, BD, USA). Data analysis was conducted using the FlowJo 10.8.1 software. After data collection, median fluorescence intensity (MFI) of CD206 and CD163 was calculated for quantitative analysis. All experiments were performed in three independent replicate samples.

An appropriate number of HTLA cells was inoculated into 96-well cell culture plates. The next day, when the cell density reached about 80%, 200 ng of the GPCR-Tango plasmid solution (dissolved in 200 μL of DMEM) was added, mixed with 400 ng of polyethyleneimine in an equal volume of DMEM, and incubated at room temperature for 20 min. After incubation, the transfection mixture was added to HTLA cells for transfection. After 24 h of transfection, the culture medium was replaced. The agonist ADRB1 (positive control) for GCDCA and S1PR2 receptor was added according to the experimental group settings. Next, 50 μL of the Bright-Glo solution (Promega, USA) was added to each well. Next, the wells were incubated at room temperature for 20 min, and the luminescence intensity was measured using a microplate luminometer (LB963, Centro). The multiplicity of activation for each sample was calculated as the relative luminescence units (RLU) for each condition divided by the control RLU for the medium only.

GCDCA-treated and untreated NR8383 cells were collected for transcriptome sequencing, which was performed by LC-Bio Technologies Co., Ltd. (Hangzhou, China). Each group consisted of three replicates. For data analysis, genes that were not detected in more than 25% of the samples were first removed, and the raw count values were normalized. Differentially expressed genes were calculated using edgeR 3.38.4 with the screening criteria of FDR < 0.05 and |log2FC| > log2(1.5). Finally, pathway enrichment analysis was performed using clusterProfiler 4.4.4.

All data are expressed as mean ± standard deviation (SD). Data were analyzed using R version 4.2.1 and GraphPad Prism 8.0. Differences between two and multiple groups of data were compared by Student’s t-test and one-way analysis of variance (ANOVA), respectively. Statistical significance was set at P < 0.05, implying that the results were statistically significant. All experiments were repeated thrice.

We analyzed the composition of the TME and the degree of malignancy of tumor cells in a DEN-induced rat model of primary HCC using single-cell RNA sequencing to assess the degree of malignancy of tumor tissues and explore the potential mechanisms promoting the development of HCC. We performed single-cell RNA sequencing of rat liver tissues at different time points (0, 4, 8, 12, and 16 weeks) to assess DEN-induced carcinoma. Single-cell clustering caused the clustering of all liver cells into 35 cell clusters (Figure 1A). Several major cell types in the liver were identified after clustering analysis based on classical markers: epithelial cells, endothelial cells, B cells, mononuclear phagocytes, NK cells, T cells, neutrophils, plasma cells, and fibroblasts, as well as the distribution of subpopulations of different cell types at different time points (Figures 1B, C, D). We conducted a proportion analysis of different cells to reveal the expression characteristics and distribution of different cell types in the liver tissues of DEN-induced carcinogenic rats. The relative abundance of each cell type in the liver cancer microenvironment was observed, which also reflected the dynamic changes in the composition of cells during the occurrence of liver cancer (Figure 1E). This suggests that the dynamic remodeling of the immune microenvironment is accompanied by tumorigenesis and development. Next, we used inferCNV to conduct further clustering analysis of epithelial cells. InferCNV analysis is a common method used to assess cellular malignancies. The degree of variation in the chromosomal copy number of target cells was determined by comparing the chromosomes of target cells with those of control cells. A darker color resulted in more CNV variation and higher cell malignancy. Epithelial cells were clustered and analyzed to obtain six cell subpopulations, and inferCNV analysis was performed using endothelial cells, fibroblasts, and T cells as reference cells. The C1 and C2 subpopulations had more significant variations. We calculated the CNV score and compared it with the reference cells, and found that the epithelial C1 and C2 subpopulations were significantly different from the reference cells; therefore, these cell subpopulations were defined as malignant cells (Figures 1F, G). HCC CSCs were identified using the CytoTRACE score to predict the differentiation status of the epithelial cells. CytoTRACE can count genes at the single-cell level to predict the differentiation status and direction of cells. A higher CytoTRACE score leads to higher stemness of the cells and a lower degree of differentiation (32). The CytoTRACE scores were highly heterogeneous in the tumor tissues, with the highest scores in the C1 and C2 subgroups (Figure 1H). The expression of Epcam, Sox9, and Krt19, stemness markers, was analyzed in the subpopulations of epithelial cells; the C1 subpopulation was found to highly express Epcam, Sox9, and Krt19 (Figure 1I). The above experimental results demonstrated that a subpopulation of CSC had high malignancy in rat HCC tissues and that the CSC subpopulation highly expressed Epcam, Sox9, and Krt19.

Several patients with HCC develop jaundice in the late stages, indicating that cholestasis is a very common pathological symptom in these patients. Patients with HCC complicated by jaundice have a significantly shorter survival period and poor prognosis (33, 34). GCDCA is a toxic bile acid synthesized by the human liver. When the liver is damaged, disturbances in bile acid metabolism cause an abnormal accumulation of bile acids, further aggravating liver injury. We used a rat primary HCC GCDCA intervention model to verify whether the toxic bile acid GCDCA has a tumor growth-promoting effect in vivo. The rat model was intraperitoneally injected with 8 μmol/100 g of GCDCA at week 12 of DEN-induced carcinogenesis to humanize the rat bile acid pool. Injections were administered for 4 consecutive weeks to mimic the effect on liver cancer in the presence of cholestasis. The other group was simultaneously fed diets enriched with 1.2% cholestyramine (an anion-exchange resin that binds to intestinal BA and hinders their reabsorption) at week 12, and the effects of GCDCA on DEN-induced primary HCC in rats were observed after 16 weeks (Figure 2A). Rats administered intraperitoneal injections of GCDCA had significantly shorter survival compared to those treated with cholestyramine (Figure 2B). This indicated that GCDCA promotes the progression of liver cancer. Oncological index tests were also conducted on the liver tissues of each rat group. The maximum diameter and number of liver tumors in the GCDCA-treated rats were significantly higher than those in the control group (Figures 2C, D, E). HE staining revealed severe liver damage in this group of rats (Figure 2C). These rats also exhibited higher activities of Aspartate Aminotransferase (AST) and Alanine Aminotransferase (ALT), and higher levels of Total Bile Acid (TBA), resulting in more severe liver injury (Figures 2F, G, H). To further assess the malignancy of the tumor tissue, immunohistochemistry and immunofluorescence staining were performed for the stemness markers EPCAM, SOX9, and KRT19. The positive signal was stronger in the liver after GCDCA treatment (Figures 2I, J; Supplementary Figure 1A). Thus, toxic GCDCA promotes HCC progression and enhances liver cancer tissue stemness.

We directly treated the human HCC cell line Huh7 and the rat HCC cell line RH35 with GCDCA to investigate the mechanism by which GCDCA promotes the development of HCC. The results of the cell proliferation assay demonstrated that GCDCA promoted HCC cell proliferation (Figure 3A). The cloning and sphere-forming assays revealed that GCDCA had no direct effect on HCC cell stemness (Figures 3B, C). The effect of GCDCA in promoting stemness in HCC cells may be realized indirectly by acting on other components of the TME, rather than through direct action on HCC cells.

Recently, an increasing number of studies have demonstrated that macrophage diversity is implicated in the occurrence and development of tumors. Macrophages exhibit distinct functions at different tumor stages. During the initial stages, they promote epithelial cell carcinogenic mutations by generating an inflammatory microenvironment and activating immune responses. They undergo phenotypic polarization in advanced stages, transforming from pro-inflammatory to tissue repair-like phenotypes, and regulate malignant processes such as tumor invasion, migration, angiogenesis, and immunosuppression (35). To identify macrophage subpopulations that interact with the C1 subpopulation in the CSC described above, all mononuclear phagocytes in the liver were further classified into 16 cell clusters, which yielded eight major macrophage subpopulations, and all types of macrophages were present in the livers at different stages of DEN-induced cancer (Figure 4A). We conducted gene set variation analysis (GSVA) to further analyze the functional characteristics of each macrophage subpopulation. This method converts the expression of a single gene into an enrichment score for a gene set, thereby reflecting the differential activity of specific functional pathways or biological processes. GSVA revealed that the macrophage subpopulation associated with bile acid biosynthesis and immunosuppressive signaling pathways was the M1 population (Figure 4B). The tSNE plot demonstrating the expression of Cd163, Spn, Apoc1, Mmp12, Trem1, Cd7, Clec9a, and Mki67 in the macrophage subpopulations revealed that the M1 subpopulation of macrophages highly expressed Cd163 in M2 macrophages (Figure 4C). Strong interactions between C1 subpopulations and M2-type macrophages in CSC were revealed using heat maps (Figure 4D) and by conducting a cellular interaction analysis (Figure 4E). These results indicated a strong interplay between CSC and M2 macrophages in the TME of HCC. M2 macrophages in the TME can interact with CSCs to promote stemness, proliferation, and migration of tumor cells by secreting cytokines, microRNAs, and chemokines (36). This was demonstrated by our findings. Therefore, GCDCA may promote M2 polarization of macrophages by acting on macrophages, which consequently enhances the stemness of tumor cells.

We first treated rat macrophage NR8383 with GCDCA to investigate whether GCDCA can induce M2 polarization in macrophages, and found that the mRNA expression of the M2 macrophage marker genes Mrc1, Cd163, and Arg1 was upregulated (Figure 5A). GCDCA induced the polarization of rat macrophages, as confirmed by the increased expression of CD206, CD163, and ARG1 proteins (Figure 5B). This conclusion was validated in the human acute monocytic leukemia cell line THP-1. THP-1 was first treated with PMA for 48 h to induce differentiation of THP-1 into macrophages, which were subsequently induced to form M2-type macrophages with IL-4 and IL-13, and continued to induce M2-type macrophage polarization in the presence or absence of GCDCA treatment, and we determined the macrophage phenotype using qRT-PCR. The mRNA expression of M2 macrophage marker genes MRC1, ARG1, and TGF-β was significantly up-regulated in GCDCA-treated macrophages compared to controls (Figure 5C). Western blotting revealed the expression of CD206, CD163, and ARG1 proteins consistent (Figure 5D). Flow cytometry revealed an elevated proportion of M2-type macrophages in the treatment group (Supplementary Figures 2A, B). These results indicated that GCDCA promotes macrophage polarization to the M2 phenotype. To further confirm this conclusion, we conducted an ELISA experiment to detect the expression of cytokines IL-6, TNF-α, IL-10, TGF-β, and VEGF in the supernatant of cells after GCDCA treatment. We found that IL-6 and TNF-α, pro-inflammatory factors, were downregulated (Supplementary Figures 2C, D), whereas IL-10, TGF-β, and VEGF, immunosuppressive cytokines, were upregulated (Supplementary Figures 2C, D). These results suggest that GCDCA promotes macrophage polarization to the M2 type, consequently generating a tumor-immunosuppressive microenvironment and facilitating tumor immune escape.

Several molecules, including EPCAM, KRT19, SOX9, and CD133, have been used as CSC biomarkers (37–40). Genetic features associated with EPCAM can be used as key predictors of survival in HCC (41). An in vitro co-culture system of tumor cells and macrophages was constructed to investigate whether the GCDCA-induced polarization of M2 macrophages could promote tumor cell stemness. NR8383 cells were cultured in the upper chamber of six-well Transwell plates, and after completion of cell-induced differentiation, the cells were co-cultured with rat tumor cells RH35 in the lower chamber for 72 h (Figure 5E). Macrophages in the GCDCA-treated group significantly enhanced the expression of HCC cell stemness genes Epcam, Sox9, and Krt19, and upregulated the corresponding protein levels compared with the control group (Figures 5F, G). We investigated whether GCDCA exerted the same effect on human cells. We cultured THP-1 cells in the upper chamber of transwell six-well plates, induced THP-1 differentiation using PMA for 48 h, and added IL-4 and IL-13 to induce M2-type macrophage polarization, and finally provided GCDCA intervention for 24–48 h. M2-type macrophages that had completed the induced differentiation were co-cultured for 72 h with the human HCC cell line Huh7 (Figure 5H). Next, the expression of the stemness marker genes EPCAM, KRT19, and SOX9 was detected by qRT-PCR in HCC cells, and the expression of stemness marker genes was significantly upregulated, similar to that in NR8383 cells. The same was performed to validate the protein levels, and the western blot results were consistent with the qRT-PCR results (Figures 5I, J). The results of the above experiments demonstrated that GCDCA-treated macrophages strongly induced stemness in HCC cells.

Bile acid is an important signaling molecule that maintains homeostasis and energy metabolism by activating bile acid receptors (42). Two membrane receptors exist for common bile acids, the G protein-coupled receptor (TGR5) and sphingosine-1-phosphate receptor 2 (S1PR2), and four nuclear receptors, the farnesol X receptor (FXR), pregnane X receptor (PXR), constitutive androstane receptor (CAR), and vitamin D3 receptor (VDR) (42, 43). Previous studies have demonstrated that GCDCA activates macrophage polarization, enhancing tumor cell stemness. However, exactly which bile acid receptor in macrophages is activated is not yet known. The expression of bile acid receptors in macrophage subpopulations was analyzed using single-cell sequencing to further investigate which receptor GCDCA activates. Only three bile acid receptors, S1PR2, VDR, and NR1I2, were expressed in rat macrophages, with the S1PR2 receptor being widely expressed in different macrophage subpopulations, whereas the other two bile acid receptors were less expressed (Figure 6A). To verify our hypothesis, we used the Tango GPCR assay to test whether GCDCA activates S1PR2, and found that GCDCA activated the S1PR2 receptor (Figure 6B). An analysis of the TCGA database revealed that S1PR2 expression was positively correlated with the expression of both M2 macrophage markers and stemness genes in human HCC tissues (Figures 6C, D). CD68, a TAM marker, is strongly associated with HCC staging (17). Immunohistochemical staining revealed that the expression of CD68, CD163, and S1PR2 was higher in GCDCA-treated liver tissues, whereas the expression of CD68, CD163, and S1PR2 was down-regulated in cholestyramine-treated liver tissues (Figure 6E; Supplementary Figure 3A). These results further confirmed that GCDCA accelerates HCC progression. Immunofluorescence co-localization of the S1PR2 receptor with the M2-type macrophage markers CD206 and CD163 demonstrated a significant increase in the number of S1PR2-positive M2-type macrophages after GCDCA treatment compared to controls, a phenomenon that was reversed by the addition of cholestyramine (Figure 6F). Thus, S1PR2 mediates M2-type macrophage polarization and enhances tumor cell stemness.

To further confirm the conclusion that GCDCA promotes M2 polarization of macrophages through activation of S1PR2 receptors, and thus enhances the stemness of HCC cells, we further verified this in vivo by establishing a rat primary HCC GCDCA intervention model, whereby GCDCA was injected intraperitoneally in the 12th week of carcinogenesis. The other group was pretreated with JTE-013 for 2 h and then administered GCDCA intervention to block the S1PR2 receptor in vivo (Figure 7A). Samples were collected after 16 weeks of DEN-induced cancer, and the results showed that the survival of rats pretreated with JTE-013 was prolonged compared to that of the group treated with GCDCA alone (Figure 7B). The maximum diameter and number of rat liver tumors were significantly reduced after pretreatment with JTE-013 compared to those in the group treated with GCDCA alone, whereas no effect was observed in the group treated with JTE-013 alone (Figures 7C, D, E). HE staining results demonstrated that liver injury was alleviated in rats after JTE-013 pretreatment (Figure 7C). In addition, rats in the JTE-013 pretreated group exhibited lower activities of ALT and AST, and TBA levels than those in the GCDCA-treated group (Figures 7F, G, H). Immunohistochemistry and immunofluorescence staining for EPCAM, SOX9, and KRT19, which are stemness markers, revealed that the positive signals in the liver were attenuated after JTE-013 pretreatment (Figures 7I, J; Supplementary Figure 4A). These results confirmed that pharmacological inhibition of S1PR2 by JTE-013 reversed GCDCA-induced stemness enhancement in tumor cells. To confirm that the action specificity of S1PR2 antagonist JTE-013 stems from its regulation of macrophages, we treated hepatocellular carcinoma cells with the same concentration of JTE-013 alone. Using CCK-8 and clonogenic assays, we investigated the effects of JTE-013 on the proliferation and survival capacity of hepatocellular carcinoma cells. Results indicate that JTE-013 itself does not directly inhibit the proliferation and survival of hepatocellular carcinoma cells (Supplementary Figures 4B, C).

Previous studies have demonstrated that GCDCA induces the polarization of M2 macrophages by activating S1PR2 receptors; therefore, in vitro inhibition of S1PR2 receptors has the opposite effect. To further verify our hypothesis, the NR8383 cell line was pretreated with the inhibitor JTE-013 for 30 min to block S1PR2 receptors, and subsequently, cells were treated with GCDCA for 24–48 h. The expression of the M2-type macrophage marker genes Mrc1, Cd163, and Arg1 was downregulated in the JTE-013-pretreated group compared to the GCDCA-treated group alone, whereas no significant difference was noted in the JTE-013-treated group compared to the control group (Figure 8A). This was also confirmed by the reduced expression of the corresponding proteins (Figure 8B). Similarly, this finding was confirmed on human macrophages, where the expression of M2 macrophage marker genes MRC1, ARG1, and TGF-β was significantly down-regulated in the JTE-013-pretreated group compared to the GCDCA alone group (Figure 8C). The expression of CD206, CD163, and ARG1 proteins was consistent with their (Figure 8D). Previous studies have demonstrated that S1PR2 is highly expressed in macrophages, whereas that of the other two receptors, PXR and VDR, is relatively low. We treated NR8383 and THP-1 cells with their corresponding receptor inhibitors, namely, ketoconazole and MeTC7, to eliminate the influence of these two receptors (PXR and VDR) on the polarization of M2-type macrophages. These inhibitors did not have a significant inhibitory effect on macrophage polarization (Supplementary Figures 5A–H). These results suggest that GCDCA promotes the polarization of M2-type macrophages by activating the S1PR2 receptor.

NR8383 cells pretreated with JTE-013 with or without GCDCA treatment were co-cultured in vitro with the rat HCC cell line RH35 to investigate the effect of S1PR2 receptor inhibition on the stemness of hepatocellular carcinoma cells. Compared with the GCDCA-treated group alone, the expression of the HCC cell stemness genes Epcam, Sox9, Krt19, and their corresponding proteins was significantly downregulated in the JTE-013 pretreated group (Figures 8E, F). The same was again confirmed in human macrophages, where THP-1 cells were induced to differentiate, and PMA, IL-4, and IL-13 inducers were added to induce M2-type macrophage polarization. JTE-013 was administered 30 min before GCDCA treatment, and the treated macrophages were co-cultured with the human HCC cell line Huh7. These results were similar to those described above, and inhibition of the S1PR2 receptor suppressed the polarization of M2-type macrophages and further inhibited the expression of stemness marker genes (Figures 8G, H). Immunofluorescence staining of the M2-type macrophage markers CD206 and CD163 and the stemness markers EPCAM and KRT19 demonstrated a significant increase in the number of CSCs and peripheral aggregation of M2-type macrophages in GCDCA-treated liver tissue. After treatment with JTE-013, the number of CSCs and M2-type macrophages decreased (Figures 8I, J). These results further confirm that GCDCA promotes the polarization of M2-type macrophages by activating the S1PR2 receptor on macrophages and that M2-type macrophages interact with tumor cells, consequently enhancing the stemness of tumor cells.

Previous studies have demonstrated that GCDCA induces M2-type macrophage polarization via S1PR2, which, consequently, enhances tumor cell stemness; however, the specific mechanism remains unclear. Although the number of studies on the mechanism of BA in the tumor immune microenvironment has increased in recent years, the specific mechanism by which GCDCA regulates M2-type macrophage polarization via S1PR2 receptors remains unclear. To explore the changes in gene expression in GCDCA-induced macrophages, we performed a transcriptome sequencing analysis of GCDCA-treated NR8383 cells and screened for genes with significant differences in expression. Volcano plots showed up- regulated and down-regulated differential genes after GCDCA treatment (Figure 9A). KEGG and GO enrichment analysis of all up-regulated differential genes revealed that the pathways enriched for up-regulated differential genes after GCDCA treatment were the PI3K/AKT signaling pathway and the TGF-β signaling pathway, among others (Figures 9B, C). Activation of the PI3K/AKT signaling pathway induces M2-type macrophage polarization (44). NR8383 cells were induced using GCDCA to investigate whether GCDCA functions through this pathway. The levels of AKT, p-AKT, PI3K, and p-PI3K proteins in the cells were detected by Western blotting, which showed significantly upregulated levels of p-AKT and p-PI3K proteins after GCDCA treatment (Figure 9D). Pretreatment of cells with the PI3K inhibitor LY294002 significantly downregulated p-AKT and p-PI3K protein levels as well as protein levels of the M2 macrophage markers CD206 and ARG1 (Figure 9E). These results indicate that GCDCA induces macrophage polarization toward the M2 type through the S1PR2/PI3K/AKT signaling pathway.