ISX–TWIST1 reprograms CD47–inflammasome signaling to reshape the immune microenvironment in liver cancer

Abstract

Immunotherapies have shown promise effectiveness in cancer treatment, yet the complex tumor microenvironment poses challenges to their efficacy. The interaction between immune checkpoints in malignant and immune cells is crucial for cancer cells to evade the host’s immune response, but the impact of tumor cell immune checkpoint molecules on disease progression remains incompletely understood. Our study uncovers a significant link between the CD47 signals, inflammasomes and intestine-specific homeobox (ISX), correlating strongly with lesion count, disease stage, and lymph vascular invasion. Elevated CD47 expression activates CD47–SIRPα signaling to promote M2-like macrophage polarization, accompanied by inflammasome activation and cytokine production within the tumor microenvironment and memory T cell differentiation within the hepatic microenvironment, intensifying disease progression. In xenograft and chronic hepatic tumor model featuring a liver-specific ISX mutant, the elimination of M2-like TAM macrophages and TRM cells halts disease advancement. Through transcriptomic analysis and molecular evidence, we unveil the interaction of ISX with TWIST1, triggering CD47-SIRPα and inflammasome activation by binding to specific degenerate sequences (“-GGDWYR-“) in the promoter regions of CD47-SIRPα signals and inflammasome-related genes. These findings underscore the pivotal role of the ISX-CD47 axis in liver disease and tumor progression. They offer promising insights into potential treatments for liver disease, shedding light on new therapeutic strategies with the potential to improve patient outcomes.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s40164-026-00777-1.

Highlights

• ISX signals show a strong correlation with CD47-SIRPα signals and inflammasome-related genes (NLRP1 and NLRP3), leading to adverse prognosis in HCC patients.

• ISX genetically transactivates CD47 signals and inflammasome-associated genes (NLRP1, NLRP3, NLRC4, AIM2, ASC, and IL-1β) in hepatoma cells.

• ISX-regulated inflammasome and CD47 activities contribute to disease progression in xenograft and chronic primary DEN-induced tumor models.

• ISX promotes hepatic M2-like macrophage polarization, especially TAM population (Tlr7^hi Laur1 + Pltp+Slamf9+), exacerbating hepatitis progression in both xenograft and chronic primary DEN-induced tumor models.

• ISX promotes hepatic memory T cells differentiation, especially tissue-resident memory T cells (TRM, Zfp683 + Itgae+ Cxcr6 + Cd69+), exacerbating hepatitis progression in both xenograft and chronic primary DEN-induced tumor models.

• The ISX-TWIST1 protein complex modulates the signals of CD47 and inflammasome activities, influencing subsequent oncogenic processes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40164-026-00777-1.

Introduction

Cancer affects a significant portion of the global population, with at least 20 million new cases and 9.7 million cancer-related deaths worldwide in 2022, including those with hepatocellular carcinoma (HCC) [1]. Despite extensive research into its prevalence, the quest for advanced treatments remains imperative to enhance the efficacy of curing cancer patients. In recent years, immunotherapy has emerged as a pivotal and promising strategy in cancer treatment [2–4]. This innovative approach has shown notable advantages, including long-lasting responses, reduced side effects, and improved survival rates when compared to conventional treatments like chemotherapy [3]. However, the dynamic nature of the tumor microenvironment poses challenges to the effectiveness of curing cancer. Ongoing advancements in targets, mechanisms, and treatment modalities are actively shaping and enhancing the contributions of immunotherapy to the continually evolving landscape of cancer treatment.

Inflammasomes in the innate immune system play a crucial role in triggering inflammation in response to various harmful stimuli [5, 6]. Mounting evidence indicates that inflammasomes are involved in the onset and progression of several inflammatory disorders, including metabolic conditions, cardiovascular diseases, liver diseases, inflammatory bowel diseases, rheumatoid arthritis, and neurologic disorders, as well as cancer [7]. Inflammasomes activities are important signals in both tumor cells and macrophages. When macrophages experience xeno-stimulation or damage signals, pattern recognition receptors (PRRs) recognise these signals [6]. This leads to the formation of inflammasome platforms that consist of PRRs, associated speck-like proteins (ASC), and pro-caspase-1 [6, 7]. Activated caspase-1 cleaves pro-IL-18 and pro-IL-1β to produce mature IL-18 and IL-1β, and also induces the synthesis of other inflammatory cytokines [7]. Caspase-1 also cleaves GSDMD, which releases inflammatory factors and exacerbates the inflammatory response [8]. In liver, several types of inflammasomes, including NLRPs, are activated by the infections of HBV and HCV, affecting the pathological process of viral hepatitis through its downstream secretion of inflammatory cytokines interleukin-1β (IL-1β) and IL-18 or induction of pyroptosis resulting from cleaved gasdermin D (GSDMD) [9]. Accumulating reports showed that inhibition of the inflammasome or anti-inflammasome effects blocked tumour progression in a mouse model of HCC [10, 11]. While the regulatory mechanisms of inflammasomes in immune cells have been revealed, their functional regulations in immune, tumour and parenchymal cells are still unclear [12] .

Intestine-specific homeobox (ISX) is a homeobox transcription factor induced by HCV infection, environmental pollutants (AhR signals) and an inflammatory cytokine, IL-6 [13, 14], that regulates the expression of various downstream targets, including PD-L1/PD-1 axis [15–17]. ISX impacts the survival of hepatocellular carcinoma (HCC) patients and is associated with tumour characteristics [13]. Additionally, ISX influences vitamin A metabolism and also regulates high-density lipoprotein receptors and the cholesterol transporter scavenger receptor class B type 1 [18, 19].

This research aimed to investigate the intricate interplay between CD47 signals, ISX, and NLRPs in the tumor microenvironment, with a particular focus on its impact on liver disease prognosis. The findings seek to uncover a novel genetic approach for stimulating immune checkpoint and inflammasome signals, which may promote the progression of liver disease and influence its prognosis. Additionally, the study aims to identify potential therapeutic targets for the treatment of this devastating condition.

Materials and methods

Patients

In a retrospective analysis, we included 625 patients (465 men and 160 women; mean age, 61.7 ± 12.12 years; range, 21–85 years) with confirmed hepatocellular carcinoma who underwent curative hepatectomy between August 2018 to July 2023 at three medical centers (Chung Ho Memorial Hospital, 355 patients; National Health Research Institutes (NHRI) BioBank, 200 patients; Changhua Christian Hospital, 34 patients and Chi Mei Medical Center, 36 patients). Normal liver tissues adjacent to tumors from patients with HCC were detected with no pathogenic signs (such as high nuclei/cytoplasm ratio) observed in tumor parts by microscopy analysis. None of the patients underwent any preoperative treatment. The study was conducted with approval (KMUH-IRB-20200048 and 20180037) from the ethics committee of the Kaohsiung Medical University Chung Ho Memorial Hospital. Written informed consent was obtained from each patient. The pathological diagnosis and classification of variables were based on the criteria recommended in the General Rules for Clinical and Pathological Study of Primary Liver Cancer. Clinicopathological characteristics collected for analyses included sex, age, glutamic oxaloacetic transaminase, glutamic-pyruvic transaminase, triglyceride (TG), Albumin, α-Fetoprotein, Bilit. Tissue specimens obtained during the operation were immediately stored in liquid nitrogen until further analysis. All patients underwent routine and regular follow-up care at our outpatient department and were carefully monitored once every 6 months for five years.

Plasmids

Full-length ISX cDNA was PCR-amplified from a human placenta cDNA library (GIBCO/BRL) and different truncated ISX cDNAs were subcloned into the pEGFP/C1 vector (CloneTech) to express the GFP-tagged ISX protein. Full-length TWIST1 (a gift from Dr. Yang MH, NYMU, Taipei, Taiwan) were inserted into the mCherry or Flag-tagged TWIST1. The PLKO.1.puro or .neo vector was used as a backbone for shRNA constructs targeting ISX (5’- CAAACTTGCATCCCTGTGCTA-3’), NLRP1 (5’- CGTTTCCTATTGGGCCTGTTA-3’), NLRP3 (5’- GAGACTCAGGAGTCGCAATTT-3’, CD47 (5’- GCCTTGGTTTAATTGTGACTT-3’) and TWIST1(5’-AGTCCGCAGTCTTACGAG-3’).

Animal and cell culture

ISX knockout mice were obtained from the Jackson Laboratory and housing according to Animal Center’s (Kaohsiung Medical University, Kaohsiung, Taiwan) protocols. The study was conducted with approval (IACUC-107134) from the ethics committee of Kaohsiung Medical University. Human hepatocellular carcinoma (HCC) cell lines, including HepG2 (RRID: CVCL_0027), Hep3B (RRID: CVCL_0326), SNU-423 (RRID: CVCL_0366), Huh7 (RRID: CVCL_0336), PLC/PRF/5 (RRID: CVCL_0485), HA22T (RRID: CVCL_7046), and HCC36 (RRID: CVCL_VI190), were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) or the Bioresource Collection and Research Center (BCRC, Hsinchu, Taiwan) between January and March 2023. All cell lines were authenticated prior to use by morphology, karyotyping, and PCR-based short tandem repeat (STR) profiling. Mycoplasma contamination was routinely monitored using the Universal Mycoplasma Detection Kit (Sigma-Aldrich, MP0050), with the most recent test performed in March 2025 confirming all cell lines were mycoplasma-free. Cells were used within 20 passages after thawing. Cells were cultured in Eagle’s Minimum Essential Medium (MEM; Gibco, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (FBS; Gibco), 2 mM L-glutamine, 1 mM sodium pyruvate, 0.1 mM nonessential amino acids, and 1.5 g/L sodium bicarbonate (Earle’s Balanced Salt Solution). Cultures were maintained at 37 °C in a humidified incubator containing 5% CO₂. Lipopolysaccharide (LPS; Sigma-Aldrich) was used at a final concentration of 100 ng/mL. MAPK inhibitors U0126 (5 or 20 µM; Cell Signaling Technology) and PD98059 (50 µM; Cell Signaling Technology), the NF-κB inhibitor BAY 11-7082 (5 or 20 µM; Sigma-Aldrich), recombinant human IL-1β (10 µg/mL; R&D Systems), recombinant human IL-18 (15 µg/mL; R&D Systems), and recombinant human SIRPα (10 µg/mL; R&D Systems) were applied as indicated.

Next-generation sequencing (RNA-Seq) and data analysis

Next-generation sequencing was performed using the Illumina TrueSeq RNA Library Preparation Kit v2 with polyA selection to obtain 50 cycles of single-end reads32. The reads were subsequently aligned to the March 2022 mouse reference sequence genome (GRCm38.p6) using the software Hisat2 (v2.0.1), after which the sample reads were visualized using the Integrated Genome Browser (Version 9.0.1). Next, differentially expressed genes were identified using the DESeq2 Bioconductor package. The DRDS function was used to calculate the false discovery rate (FDR) statistic for the significance of differentially expressed genes. Log-transformed FPKM of > 0.1 in at least one treatment group was used for the analysis. DEGs with fold changes greater than 1.5 and a log-transformed FDR of 0.05 or less were only used. Means-centered log-transformed FPKM was used to prepare hierarchical clustering heatmaps in Cluster (version 3.0) and Java Tree View (version 3.0). A final significant differential gene list was used for gene enrichment analysis, including Gene Ontology (Biological Process) and the KEGG pathway.

RNA isolation and semi-quantitative real-time PCR

Total RNA from liver tissue RNA were isolated by RNeasy Mini Kit according to instructions from the manufacture (Qiagen, Valencia, CA, USA) and then transcribed into cDNA (Invitrogen) for PCR amplification on a STEPONE Thermocycler (Applied Biosystems Inc.). Semi-quantitative real-time PCR was performed with SYBR Green FastMix (Applied Biosystem) and analyzed using ΔΔCt calculations. All data were normalized to GAPDH expression. All data are expressed as the mean ± SD of at least 3 experiments.

Western blotting and immunohistochemical analysis

Western blotting and immunofluorescence staining were performed using standard protocols as previously described, with antibody sources, catalog numbers, and working dilutions specified below [13]. The primary antibodies used in this study were Actin polyclonal antibodies (1:5000 dilution; A2066; Sigma–Aldrich), GFP polyoclonal antibodies (1:500 dilution; ab13970; Abcam), FITC-conjugated anti-rabbit IgG (111-095-003), rhodamine-conjugated anti-mouse IgG (115-295-146), and alkaline phosphatase-conjugated anti-rabbit IgG antibody (111-055-003) (1:500 dilution; Jackson ImmunoResearch Laboratories), and F4/80 monoclonal antibody (1:200 dilution; sc-377009; Santa Cruz Biotechnology). The Ki-67 goat polyclonal antibody (1:200 dilution; sc-7846; Santa Cruz Biotechnology). The ISX (GTX49096), TWIST1 (GTX127310), N-cadherin (GTX127345), E-cadherin (GTX100443), Snail1 (GTX125918), Slug (GTX128796), mCherry (GTX128508), CD80 (GTX04260), CD47 (GTX53912), SIRP1a(GTX112645) and Cyclin D1 (GTX112874) (1:500 dilution) primary antibody was obtained from GeneTex International Corp. The Flag (#14793), NLRP1 (#56719), NLRP3 (#15101), NLRC4 (#12421), Caspase1 (#3866; #24232), IL-1beta (#12703; #31202), AIM2 (#12948; #63660), CD206 (#24595), F4/80 (#30325) and ASC (#13833; #67824) (1:1000 dilution) primary antibody was obtained from Cell Signaling Technology. All the experiments were repeated at least 3 times.

Luciferase reporter assays

The NLRP1 promoter (between positions − 1237 and + 25), CD47 promoter (between positions − 1250 and + 24), NEK7 promoter (between positions − 1260 and + 28) and NLRP3 (between positions − 1284 and + 35) were cloned from human placenta genomic DNA and was used to construct a pGL3 luciferase reporter plasmid.3 The expression constructs and two reporter constructs, pSV40-Rluc and pGL3-NLRP1(or CD47, NEK7, NLRP3)/Fire luciferase (Promega Co., Madison, WI, USA), were cotransfected with ISX into 2 ⋅ 10⁵ Huh7 cells. The cells were harvested 16 h after the transfection, and the relative luciferase activity was measured according to the manufacturer’s instructions. All data are expressed as the mean ± s. d. of at least three experiments.

Chromatin immunoprecipitation (ChIP) and two-step chromatin immunoprecipitation (ChIP) assays

Chromatin immunoprecipitation (ChIP) assays were performed using SNU-423 cells as previously described, with minor modifications. Briefly, cells were crosslinked with 1.0% formaldehyde for 10 min at room temperature, and the reaction was quenched with 125 mM glycine. Cells were lysed using RIPA buffer supplemented with protease inhibitors. Chromatin was sheared by sonication for 10 min at M2 intensity to generate DNA fragments of approximately 300 bp. Immunoprecipitation was carried out by incubating chromatin with anti-GFP antibody (Abcam, ab13970) or normal rabbit IgG (Cell Signaling Technology) overnight at 4 °C with rotation. Protein A/G agarose beads (30 µL; Thermo Fisher Scientific) were added and incubated for 1 h. For two-step ChIP (ChIP-ChIP) assays, eluted complexes from the first immunoprecipitation were subjected to a second immunoprecipitation using anti-mCherry antibody (GeneTex, GTX128508) or IgG control. DNA was purified and analyzed by quantitative PCR (qPCR). ChIP–qPCR validation was performed for selected high-affinity targets, including CD47, NLRP1, NLRP3, and NEK7.

Cell proliferation assay

A colorimetric immunoassay (Roche) for the quantification of the cell proliferation was performed based on the measurement of BrdU incorporation during DNA synthesis according to the manufacturer’s instructions. The cells (10⁴) cultured in 96-well plates were incubated at 37 °C for 16 h and then growth-arrested prior to indicated treatment. The BrdU labeling reagent was added into the medium for 2 h incubation at 37 °C. Absorbance values were measured at 450 nm using a VersaMax ELISA Microplate Reader.

Wound-healing assay

The human hepatoma cells overexpressing ISX (or TWIST1) or transfected with shRNA that were seeded in 6-cm culture plates were scratched using a pipette tip to create a gap, followed by incubation at 37 °C and imaged every 12 h using a digital camera attached to a microscope. For each gap, the average width was computed from three measurements taken at the top, middle, and bottom end of the microscopic field.

Transwell invasion assay

The Transwell invasion assay was performed using a Transwell chamber (Life Technologies) with a Matrigel-coated filter. Human hepatoma cells (1 × 10⁵) overexpressing ISX or transfected with ISX mutant were added to 250 µL of serum-free media and plated onto the upper chamber of the Transwell. The upper chamber was then transferred to a well containing 700 µL of media supplemented with 10% FBS and incubated for 18 h. Cells may actively migrate from the upper to the lower side of the filter using FBS as an attractant. Cells on the upside were removed using cotton swabs, and the invasive cells on the lower side were fixed, stained with a 0.1% crystal violet solution, and counted under a light microscope. The experiment was repeated three times.

Proximity ligation assay

Proximity ligation assays (PLA) were performed to assess protein–protein interactions between ISX and TWIST1 using the Duolink^® In Situ Red Starter Kit Mouse/Rabbit (Sigma-Aldrich, St. Louis, MO, USA), according to the manufacturer’s instructions. Cells were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100, and blocked prior to antibody incubation.

Mouse monoclonal anti-ISX antibody (Santa Cruz Biotechnology) and rabbit polyclonal anti-TWIST1 antibody (Cell Signaling Technology) were applied, followed by incubation with species-specific PLA probes (anti-mouse MINUS and anti-rabbit PLUS). Ligation and rolling-circle amplification were performed as specified by the kit protocol, and PLA signals were detected using a red fluorescent detection reagent. Nuclei were counterstained with DAPI, and coverslips were mounted for imaging.

PLA signals were visualized using confocal microscopy. Quantification was performed by counting PLA puncta per nucleus in at least 100 cells per condition, averaged from three independent experiments.

ELISA analysis

The levels of mouse GPT, triglyceride, GOT, IL-6, IL-18, IL-1β and TGFβ were determined by following the manufacturer’s instructions on the ELISA kit from R&D Systems and Cusabio. First, blood was collected from the heart of a mouse and centrifuged. Then, serum was collected from the supernatant after centrifugation. Next, the serum was subjected to three repeated ELISA assays. All concentrations were calculated by referring to a standard curve of purified targets provided in the ELISA kit.

Primary hepatocyte isolation

Liver digestion was adapted from star protocols [20]. Livers were harvested and placed in cold DMEM, minced with scissors, and immediately transferred to tubes containing 10 ml of digest buffer. Minced livers were digested for 30 min at 200 rpm, 37 °C, with additional vigorous manual shaking every 5–8 min. Digested livers were filtered through 100 μm strainers and topped up with cold DMEM containing 10% FBS to 50 ml, then spun at 300 g for 10 min. The supernatant was removed, and the cells were washed a second time with 30 ml of PBS buffer. The remaining pellet was treated with red blood cell lysis buffer, washed, and counted for staining.

Single-cell RNA (scRNA)-sequencing

The alive liver cells of three mice in each group (Control and Isx null) were dissociated according to the above procedure and pooled for further analysis. For a target recovery of 20,000 single cells, 18,000 live cells were loaded into the 10× Genomics Chromium Single Cell chips. Next, libraries were prepared using the Chromium Single Cell 3′ GEM Library & Gel Bead Kit v3 (10× Genomics) according to the manufacturer’s instructions and sequenced on an Illumina Novaseq 6000 with a sequencing depth of at least 100,000 reads per cell and 150-bp (PE150) paired-end reads.

Analysis of 10× scRNA-seq data

Raw sequencing data generated by 10x Genomics were processed using the Cell Ranger (v3.0.0) count pipeline, including read alignment, barcode assignment, and UMI quantification. The reference genome index was constructed using Mus musculus GRCm38.p6. Median sequencing saturation was determined based on the Cell Ranger-reported saturation metric. Cells with fewer than 500 UMIs or with mitochondrial gene content exceeding 50% of total UMIs were excluded. Additionally, genes expressed in fewer than three cells per sample were filtered out. After quality control, 10,000 cells from wild-type and 10,000 cells from Isx-null mice were retained for downstream analysis. Clonotype distributions in tSNE space and expression profiles of liver-specific marker genes (listed in Table S2) were visualized using Loupe Browser (10x Genomics).

Identification of cluster-specific genes and marker-based classification

To identify cluster-enriched genes, the average expression of each gene was calculated across all cells within the target cluster and compared to its mean expression across all other clusters. Genes were then ranked by differential expression, and the top-ranking genes (DEGs) from each cluster were selected for further analysis. For hierarchical clustering, pairwise correlations between clusters were computed, and gene expression values were centered prior to visualization in a heatmap.

Pseudo-time analysis

Cell trajectory and pseudo-time analyses were conducted using the Monocle R package (version 2.18.0; Bioconductor) with the reverse graph embedding algorithm. To initiate the analysis, a set of genes representative of distinct differentiation stages was selected for cell ordering. These genes were used in principal component analysis (PCA), followed by the orderCells function to project each cell into a high-dimensional space defined by their expression profiles. Next, Monocle 3 was applied to construct a developmental trajectory by generating a principal tree, where each cell was iteratively mapped to its nearest vertex in a manner that best captured its transcriptional identity. This process refined the trajectory structure while accounting for gene selection and dimensionality reduction, ensuring a coherent lineage model. After establishing the tree, a terminal node was automatically designated as the root, and the geodesic distance of each cell from this root was calculated to define its pseudo-time. Differential gene expression analysis along the trajectory was then performed using predefined cell clusters from the Seurat object, with statistical significance set at P < 0.01.

Mass cytometry

Mass cytometry was performed by the China Medical University CyTOF core using a modified panel of metal-conjugated antibodies. Isolated cells were washed with Maxpar staining buffer (Fluidigm) and blocked with anti-mouse CD16/32 (BioLegend), followed by incubation with the antibody cocktail for 1 h. Cells were then washed with staining buffer and incubated with cell surface solution for 1 h, followed by two additional washes with staining buffer (Table S2). The cells were then incubated with cell intercalation solution for 1 h, followed by two additional washes with staining buffer and fixation. Mass cytometry data were normalized using EQ Four Element Calibration Beads (Fluidigm). tSNE analyses were performed using Cytobank Premium (Cytobank Inc.).

Model TWIST1-ISX-dsDNA complex structure

Prediction of human TWIST1 (UniProt ID Q15672), human ISX (UniProt ID Q2M1V0) and dsDNA complex was accomplished by AlphaFold 3 server (https://www.alphafoldserver.com) [21]. Two copies of TWIST1 as well as ISX were set. We use multiple random seeds (n = 100) to ensure high accuracy. The top ipTM + pTM model was analyzed in this study. All protein-DNA complex structure graphical figures were generated using PyMOL version 2.5.5 (Schrodinger, LLC).

Co-culture stimulation conditions

SNU-423 (Huh 7) cells and macrophage cells derived from THP1 were co-cultured in 96-well plates for 24 h in RPMI medium. After 24 h, macrophage cells were collected for analysis.

Statistical analysis

Quantitative data are presented as mean ± standard deviation (SD). Comparisons between two groups were performed using an unpaired two-tailed Student’s t-test. Correlations between ISX, NLRP1, NLRP3, CD47, and TWIST1 expression levels were assessed using Pearson’s correlation analysis. Categorical variables were analyzed using the χ² test or Fisher’s exact test, as appropriate, and multiple-group comparisons were performed using one-way analysis of variance (ANOVA). A P value < 0.05 was considered statistically significant.

Results

Strong correlation between CD47, inflammasomes, and ISX signals and adverse prognosis in HCC patients

To validate the candidate effectors in HCC patients influenced by heightened inflammatory effects, transcriptomic analyses were conducted on tumor masses from HCC patients with higher levels of inflammation compared to those with lower levels. These analyses revealed a strong correlation between CD47 signaling—an immune checkpoint modulator on macrophages [22]—and the transcription factor ISX, linking them to inflammasome activities, including NLRPs expression, NF-κB signaling, NEK7, and IL-18 production (Fig. 1a). A retrospective analysis of HCC patients was conducted to assess the relationship between inflammasome activities, CD47 signaling, and the transcription factor ISX in the context of chronic liver disease. The mRNA levels of inflammasome-related genes, including NLRP1, NLRP3, CD47 signaling, and ISX, were measured in 625 paired HCC and adjacent healthy tissue samples. Patients with elevated levels of NLRP1 or NLRP3 mRNA (n = 137 and 158, respectively) exhibited significantly shorter overall survival compared with those with lower expression levels (n = 488 and 467, respectively; P < 0.001; Fig. 1b). Similarly, high CD47 mRNA expression (n = 151) was associated with reduced survival relative to low expression (n = 474; P < 0.001; Fig. 1b). Consistently, patients with high ISX expression (n = 163) showed poorer overall survival compared with those with low ISX expression (n = 462; P < 0.001; Fig. S1a). The baseline characteristics of mRNA expression for inflammasome-related genes (NLRP1 and NLRP3), CD47, and ISX in HCC patients were further analyzed (Table 1 and S1). The findings revealed significant differences (p < 0.001) in adverse prognostic indicators, including lymphovascular invasion (LVI), tumor number, size, and TNM stage, between the high- and low-expression groups of CD47 and ISX. Significant differences in tumor numbers (p = 0.0011), size, and LVI (p < 0.001) were also observed between the high- and low-expression groups of NLRP1. Similarly, the high- and low-NLRP3 groups displayed significant differences in TNM stage, LVI (p < 0.001), and tumor numbers (p = 0.0154) (Table 1). Intriguingly, the mRNA levels of ISX, a transcription factor induced by HCV infection [23], showed significant correlations with CD47 and NLRPs (NLRP1 and NLRP3) expression in HCC patients (Pearson’s r = 0.8626, 0.7967, and 0.6949, respectively; Fig. 1c–e). Analysis of inflammasome-relevant protein expression (NLRP1, NLRP3, NLRC4, Caspase-1, IL-1β), CD47 signaling, and ISX in high-inflammatory-level tissue samples revealed a closely aligned expression pattern between inflammasome proteins and ISX in HCC tissue samples (Fig. 1f). Confocal immunofluorescent staining further confirmed elevated protein levels of CD47 and NLRPs (red) in tumor tissue compared to adjacent healthy tissue, along with increased ISX levels (green) (Figs. 1g and S1b). Overlapping expression (yellow) between ISX and CD47 or NLRP3 was highlighted by pink arrows in Fig. 1g. To investigate the potential regulatory influence of ISX on CD47 and inflammasome activation, Huh7 and SNU-423 hepatoma cells were treated with lipopolysaccharide (LPS) at a concentration of 100 ng/ml to induce an inflammasome response [24]. LPS treatment significantly upregulated CD47, ISX, and inflammasome-related genes (NLRP1, NLRP3, and NF-κB p65), which was effectively attenuated by the MAPK (U0126, 5 and 20 µM) and NF-κB signaling inhibitor BAY-11-7082 (5 and 20 µM) (Figs. 1h–j and S1c). The inhibition extended to the suppression of inflammatory cytokine secretion, specifically IL-1β and IL-18 (Fig. 1h–j). Furthermore, RNAi-mediated knockdown of ISX blocked the LPS-induced upregulation of CD47, inflammasome-related genes, and cytokines (Fig. 1h–j). These findings suggest a clinically significant regulatory relationship among ISX, CD47, and inflammasomes in HCC, with potential implications for prognosis and therapeutic targeting.

Fig. 1.

Activation of inflammasomes is significantly associated with poor prognostic factors in hepatocellular carcinoma (HCC). a Heat map illustrating differential gene expression patterns in HCC tumor tissues compared with paired adjacent non-tumor liver tissues. b Kaplan–Meier overall survival analyses of HCC patients stratified by NLRP1, NLRP3, and CD47 expression. Patients were classified into high- and low-expression groups using survival-based ROC-derived cutoffs (3.1-fold for NLRP1, 3.7-fold for NLRP3, and 4.3-fold for CD47 relative to paired non-tumor tissues). c–e Correlation analyses between ISX and CD47 (c), NLRP1 (d), and NLRP3 (e) mRNA expression levels in HCC tumor tissues, measured by quantitative RT–PCR and normalized to GAPDH using the ΔΔCt method. f Co-expression of ISX with inflammasome-related proteins (NLRP1, NLRP3, NLRC4, caspase-1, IL-1β), CD47, SIRPα, and TWIST1 in human HCC specimens. g Representative immunofluorescence images showing co-localization of ISX (green) with CD47 or NLRP3 (red) in HCC tumor (T) and paired adjacent non-tumor (N) tissues. Nuclei were counterstained with DAPI (blue); arrows indicate co-localization. Tumor and non-tumor regions were identified by hepatopathologists using matched serial sections. h Semiquantitative RT–PCR analysis of NLRP1, NLRP3, p65, CD47, and ISX expression in SNU-423 cells treated with lipopolysaccharide (LPS), NF-κB inhibitors, or ISX shRNA. i, j IL-1β and IL-18 levels in culture supernatants from Huh7 and SNU-423 cells treated with LPS, MAPK or NF-κB inhibitors, or ISX shRNA, measured by ELISA. Survival curves were analyzed using the Kaplan–Meier method and compared by the log-rank test. Correlations were assessed using Pearson’s correlation coefficient. Quantitative data are presented as mean ± SD and were analyzed using an unpaired two-tailed Student’s t-test, as appropriate. P < 0.05 was considered statistically significant

Table 1.

Baseline characteristics of 625 hepatocellular carcinoma (HCC) patients

Group	NLRP1		p-value	NLRP3		p-value	CD47		p-value	ISX		p-value
	Low	High		Low	High		Low	High		Low	High
	n = 488(%)	n = 137(%)		n = 467(%)	n = 158 (%)		n = 474(%)	n = 151 (%)		n = 468(%)	n = 157 (%)
Lymphovascular invasion
No	311 (63.73)	55 (40.15)	< 0.001*	292 (62.53)	74 (46.84)	< 0.001*	306 (64.56)	62 (41.06)	< 0.001*	300 (64.10)	67 (42.68)
Yes	177 (36.27)	82 (59.85)		175 (37.47)	84 (53.16)		168 (35.44)	89 (58.94)		168 (35.90)	90 (57.32)	< 0.001*
Size(cm)
< 2.5	68 (13.93)	16 (11.68)		62 (13.28)	22 (13.92)		68(14.35)	16 (10.60)		67 (14.35)	17 (10.83)
2.5 ≦ < 5	260 (53.28)	36 (33.58)		239 (51.18)	67 (42.41)		252 (53.16)	57 (37.75)		254 (54.43)	59 (37.58)
5≦	160 (32.79)	75 (54.74)	< 0.001*	166 (35.52)	69 (43.67)	0.1432	154 (32.49)	78 (51.65)	< 0.001*	147 (32.22)	81 (51.59)	< 0.001*
Number of tumors
1	383 (78.48)	89 (64.96)		364 (77.94)	108 (68.35)		391 (82.49)	80 (52.98)		371 (79.27)	101 (64.33)
1<	105 (21.52)	48 (35.04)	0.0011*	103 (22.06)	50 (31.65)	0.0154*	83 (17.51)	71 (47.02)	< 0.001*	97 (20.73)	56 (35.67)	< 0.001*
Modified TNM
I	57 (11.68)	12 (8.76)		259 (55.46)	63 (39.87)		277 (58.44)	45 (29.80)		267 (56.05)	55 (35.03)
II	332 (68.03)	92 (67.15)		131 (28.05)	38 (24.05)		131 (27.64)	40 (26.49)		130 (27.78)	40 (25.48)
III(IIIA and IIIB)	99 (20.29)	33 (24.09)	0.4478	77(16.49)	57 (36.08)	< 0.001*	66 (13.92)	66 (43.71)	< 0.001*	71 (16.17)	62 (39.49)	< 0.001*

Hepatocellular carcinoma patients were classified into two groups- “low” and “high” according to survival receiver–operator characteristic (ROC) curve analysis. The cutting points of NLRP1, NLRP3, CD47 and ISX separately were 3.1, 3.7, 4.3 and 3.0 times of the mRNA expression in HCC tumors than that of the neighboring healthy tissues.SD, standard deviation. Statistical analysis of categorical variables was carried out by one-way ANOVA; *, p < 0.05

Patients: 625 HCC patients from three medical centers [Chung Ho Memorial Hospital (391HCC), Taiwan Liver Cancer Network (200 HCC) and Changhua Christian Hospital (34 HCC)] were enrolled into theNLRP1 cohort study from May 2017 to May 2022

Genetic transactivation of CD47 and inflammasome-associated genes by CD47-ISX axis drives macrophage polarization

To explore the regulatory role of ISX in CD47 expression and inflammasome activation, we analyzed the impact of ISX-induced transactivation on CD47 and inflammasome-associated genes in doxycycline (Dox)-induced ISX-expressing hepatoma cells. ISX induction in Huh-7 and SNU-423 cells led to a time-dependent increase in mRNA and protein levels of CD47 and inflammasome-related genes, including NLRP1, NLRP3, NLRC4, AIM2, ASC, and IL-1β, along with elevated cytokine secretion (IL-1β and IL-18) (Figs. 2a, b, and S1d). Promoter analysis revealed that GFP-tagged ISX transactivated luciferase activity driven by the promoters of NLRP1 (-590~-548 bp), NLRP3 (-51~-21 bp), NEK7 (-95~-75 bp), and CD47 (-257~-150 bp) in SNU-423 cells (Fig. 2c). Knockdown of CD47 via RNAi significantly attenuated the ISX-induced upregulation of multiple inflammasome-associated proteins, including NLRP1, AIM2, ASC, caspase-1, and IL-1β, with a more modest effect observed for NLRP3 (Fig. 2d). Notably, recombinant SIRPα, CD47 ligand, treatment significantly upregulated ISX, CD47 and inflammasome-related genes (NLRP1 and NLRP3), which was effectively attenuated by the MAPK inhibitors (U0126, 20 µM; PD 98059, 50 µM) (Figs. 2e and S1e). To validate ISX’s role in modulating macrophage polarization in liver disease, ISX- or GFP-expressing hepatoma cells were co-cultured with PMA-primed THP-1 cells (M0) (Fig. 2f). Co-cultured ISX-expressing cells significantly elevated M2-like macrophage markers (Arginase1, CD206) with minimal changes in M1 markers (iNOS, CD80) (Fig. 2g–j). This effect was reversed by neutralizing antibodies against IL-18 or CD47, but not by anti-IL-1β or IgG, indicating that IL-18 and CD47 signaling promote M2-like polarization. Furthermore, unpolarized THP-1 cells treated with recombinant IL-18 or SIRPα (CD47 ligand) showed increased M2 markers without altering M1 markers (Fig. 2k, l). These findings suggest that CD-47-ISX axis induces CD47 signaling and inflammasome activation, driving M2-like macrophage polarization via IL-18 and CD47.

Fig. 2.

The expression levels of inflammasome-related genes and cytokines were upregulated by ISX. a Relative mRNA expression levels of NLRP1, NLRP3, NEK7 CASP1 and CD47 in Huh7 and SNU-423 cells with DOX-inducible ISX expression system at an 8-hour time point after DOX induction as detected by semiquantitative RT-PCR. Data are presented as means ± SD. a, P < 0.001. b The IL-1β and IL-18 levels in culture media were determined by ELISA. Huh7 and SNU-423 cells were treated with doxycycline (1 µg/mL). c ISX transcriptionally activated luciferase activity driven by promoters of NLRP3, NLRP1, NEK7 and CD47 in SNU-423 cells. Relative luciferase activity was calculated as described in the Materials and Methods section. d Western blot analysis was performed to assess the protein levels of ISX, NLRP1, NLRP3, CD47, AIM2, ERK1/2, p-ERK1/2, pre-IL-1β, CD47, and Caspase-1 in Huh7 cells transfected with ISX and NLRP1 (or NLRP3 and CD47) shRNA. e The mRNA expression levels of ISX, CD47, SIRPα, NLRP1 and NLRP3 were assessed in SNU-423 cells using semiquantitative RT-PCR. THP1 cells (M0) were treated with SIRPα (10 µg/ml), U0126 (20 µM) and PD98059 (50 µM). f Scheme for the co-culture system involving THP1 cells and Huh7 or SNU-423 with varying expressions of ISX, along with the presence or absence of IL-1β, IL-18, and a PD-L1 inhibitor, to evaluate the immune effects. g–j. The mRNA expression levels of M1 markers (iNOS and CD80) and M2 markers (Arginase1 and CD206) were assessed in THP1 cells using semiquantitative RT-PCR. These cells were co-cultured with Huh7 or SNU-423 cells that were treated with IL-1β, IL-18, and a CD47 inhibitor. k The mRNA expression levels of M1 markers (iNOS and CD80) were assessed in THP1 cells (M0) using semiquantitative RT-PCR. THP1 cells (M0) were treated with IL-1β (10 µg/ml), IL-18 (15 µg/ml) or SIRPα (10 µg/ml) alone. l The mRNA expression levels of M2 markers (Arginase1 and CD206) were assessed in THP1 cells using semiquantitative RT-PCR. THP1 cells (M0) were treated with IL-1β (10 µg/ml), IL-18 (15 µg/ml) or SIRPα (10 µg/ml) alone. m Histological analysis of liver tissues from both control and Isx-null mice was performed using hematoxylin and eosin staining. CV, central vein. n A representative heatmap were used to illustrate the patterns of gene expression in control liver tissue compared to those in Isx-null liver tissue. Data are presented as mean ± SD from at least three independent experiments. Statistical significance between two groups was determined using an unpaired two-tailed Student’s t-test. P < 0.05 was considered statistically significant

To investigate ISX’s role in liver function, liver-specific Isx-null mice (Alb-Cre/Isx^fx/fx) were generated (Fig. S1f). These mice showed no significant differences in liver function markers (GOT, GPT, and TG), immune modulators (PD-L1 and kynurenine (Kyn)), liver morphology, or histopathology compared to controls (Alb-Cre/Isx+/+) (Figs. 2m and S1g). Transcriptomic analysis revealed 340 downregulated genes in Isx-null mice, many related to inflammatory responses comparing to control littermates (Fig. S1h). Notably, genes and proteins involved in CD47 signaling (Cd47 and Sirpa) and inflammasomes (Nlrp1, Nlrp3, Aim2, and Il18) were significantly reduced in Isx-null mice (Figs. 2n and S1i). Collectively, these results demonstrate that CD47-ISX axis drives the expression of inflammasome-associated and CD47 genes via MAPKs (ERK1/2) activity. Genetic ablation of ISX effectively disrupted CD47–SIRPα signaling and suppressed inflammasome-related pathways in vivo.

Loss of Cd47–ISX signaling reshapes the tumor microenvironment and inhibits hepatic tumorigenesis

To investigate the pathogenic role of inflammasome activation driven by CD47-ISX signals in the progression of liver disease, we used a xenograft tumor model with human hepatocellular carcinoma (HCC) cells in Isx-null and littermate control mice (Fig. 3a, b) [25]. Eight 12-week-old mice per group received intrahepatic injections of 1 × 10⁷ GFP-labelled SNU-449 cells and were monitored weekly using an in vivo imaging system (IVIS). Visible tumor masses emerged in control mice by the third week post-injection and progressively increased in size (Fig. 3b). In contrast, Isx-null mice displayed delayed tumor development, with minimal or undetectable tumor formation even by the fifth week (Fig. 3b, c). Macroscopically, SNU-449 injections induced numerous white nodules on the liver surface of control mice, whereas substantially fewer lesions were observed in Isx-null counterparts (Fig. 3c). Histological analysis via haematoxylin and eosin (H&E) staining revealed well-defined human hepatocellular carcinoma (HCC) tumor masses accompanied by extensive immune infiltration at the tumor–liver interface in control mice. This was markedly reduced in Isx-null mice (Fig. 3d). Consequently, control mice exhibited significant elevations in hepatic inflammasome-associated cytokines, including IL-1β and IL-18, following tumor cell injection. In contrast, only minor increases were detected in Isx-null mice (Fig. 3e, f).

Fig. 3.

Genetic ablation of ISX disrupts CD47–inflammasome signaling, reduces immune infiltration, and inhibits tumor growth in xenograft models of hepatocellular carcinoma. a-c Experimental protocol, representative IVIS images, and photographs of control and Isx-null mice injected with SNU-449 cells, along with their respective liver tissues. d H&E staining of liver tissues from control and Isx-null mice three weeks after intrahepatic injection of GFP-labeled SNU-449 cells. Control mice showed tumor nodules with prominent immune infiltration, whereas Isx-null mice exhibited minimal or no tumor lesions. Scale bar, 100 μm. e, f ELISA analysis of hepatic IL-1β and IL-18 levels revealed elevated cytokine expression in control mice, but significantly reduced levels in Isx-null mice (n = 5). Data are mean ± s.d.; *P < 0.05, **P < 0.01. g Immunoblotting of peritumoral liver tissues showing reduced expression of CD47, SIRPα, and inflammasome components (NLRP1, NLRP3, NLRC4, ASC, AIM2, caspase-1, IL-1β) in Isx-null mice compared to controls. h The Kaplan-Meier survival curves of control and Isx-null mice injected with SNU-449 cells, along with their respective liver tissues. Single-cell RNA sequencing was performed on pooled liver cells from three mice per group. Cell proportions reflect relative representation after tissue dissociation and are not intended to represent absolute in vivo cellular abundance. i t-SNE/UMAP visualization of hepatic cell clusters. j Marker-based annotation of major liver cell populations. k Relative distribution of cell populations between control and Isx-null groups. Tumor growth and cytokine data are presented as mean ± SD and were analyzed using an unpaired two-tailed Student’s t-test. Survival curves were analyzed using the Kaplan–Meier method. P < 0.05 was considered statistically significant

To further examine the impact of ISX on CD47 signalling and inflammasome activation within the hepatic microenvironment, immunoblotting analyses were performed on liver tissues adjacent to human tumor masses. Tumor formation in control mice led to pronounced upregulation of CD47 signaling components (CD47 and SIRPα) and inflammasome-related proteins (NLRP1, NLRP3, NLRC4, ASC, AIM2, caspase-1 and IL-1β) (Fig. 3g). In contrast, Isx-null mice displayed minimal or no induction of these proteins following tumor formation (Fig. 3g). Notably, Isx-null mice also exhibited significantly prolonged survival compared to control littermates after hepatoma cell injection (Fig. 3h).

To confirm the CD47–ISX–dependent microenvironmental changes identified in the xenograft model, we conducted single-cell RNA sequencing (scRNA-seq) on liver tissues from Isx-null and control mice, with 10,000 viable cells analyzed per group (Fig. 3i). Based on established marker gene expression profiles, the cells were classified into 12 distinct clusters, including T cells, macrophages, Kupffer cells, B cells, endothelial cells (LSECs), human tumor cells, erythroid cells, cholangiocytes, hepatic stellate cells and hepatocytes [26, 27](Fig. 3i–j, Table S2). In Isx-null mice, the relative abundance of T cells (4.11%), macrophages/Kupffer cells (5.4%), and tumour cells (7.25%) was significantly reduced. Meanwhile, the B cell (7.22%) and Hepatocyte Cluster 1 (Hep 1, 9.31%) populations increased compared to control mice (Fig. 3k). No substantial differences were observed in other hepatic cell populations between the two groups (Fig. 3k). Single-cell transcriptomic analysis further revealed that Isx expression was largely restricted to hepatocytes and tumor cells, with minimal expression in immune cell populations, indicating that ISX exerts its immunomodulatory effects through tumor- and parenchymal cell–intrinsic mechanisms. In this context, tumor cell–intrinsic ISX–CD47 signaling is sufficient to promote macrophage M2-like polarization, whereas ISX activity within the hepatic microenvironment is required in vivo to sustain macrophage remodeling and tumor progression. Accordingly, loss of CD47–ISX signaling suppresses tumor formation and reshapes the tumor-associated hepatic immune microenvironment by attenuating CD47 signaling and inflammasome activation, thereby limiting immune infiltration and inflammatory cytokine production.

CD47–ISX signaling drives M2-like TAM polarization and TRM cell differentiation to promote hepatic tumor progression

To further examine underlying regulatory effects in tumour microenvironment, hepatic macrophage subsets was clustered into 9 subsets according to specific population markers to assess the macrophage polarization changes in tumour microenvironment of xenograft tumour models [28] (Fig. 4a, Table S3). Notably, in Isx null mice with human tumour growth, there was a significant decrease in the hepatic macrophage subsets, including monocyte, Cd74 + Cd81+Mif+, and M2-like (TAM, 2a, 2c, and 2d) comparing to control littermates, while M1-like, M2b, and Ly6C^hi monocyte were significantly increased compared to control mice (Fig. 4b). Pseudotime trajectory analysis revealed subtle polarization differences in macrophages, with subset 5 (monocyte macrophages, F4/80 + MHCII^lo) and 6 (M2d, IL-10 + VEGF^hi) showing poor polarization and subset 1 (TAM, Tlr7^hi Laur1 + Pltp+Slamf9+) and 2 (M1, F4/80 + Il1b+cd14 + Cd80+Cd86 + Fcgr3+) exhibiting hyperpolarization in Isx null mice under HFD (Fig. 4c). To access the regulatory effects of Isx on macrophage polarization, macrophages (F4/80+) were also monitored by confocal immunofluorecent staining (IF) in xenograft tumour model. The results showed that elevated levels of both M1-type macrophages (F4/80+(red), CD80+(green); yellow) and M2-like macrophages (F4/80+(red), CD206+(green); yellow) were observed around the tumor zone in control mice (Fig. 4d). In contrast, in Isx null mutant mice, nearly all F4/80 + macrophage cells around the centrilobular zone exhibited characteristics of M1-like macrophages, with fewer or no M2-like macrophages observed (Fig. 4d). Together, these data suggest that Isx-associated activation of Cd47–Sirpα and inflammasome-related pathways is linked to hepatic macrophage remodeling and may contribute to M2-like macrophage polarization, particularly TAM-like subsets, in liver malignancy progression. This interpretation is further supported by functional perturbation assays showing that blocking CD47 or IL-18 attenuates ISX-driven M2-like polarization (Fig. 2g–l).

Fig. 4.

ISX–CD47 signaling drives M2-like TAM polarization and TRM cell differentiation in the hepatic tumor microenvironment. a t-SNE plot of liver macrophages from xenograft tumor-bearing mice, clustered into 9 subsets based on marker gene expression, including monocytes, M1-like, and M2-like TAMs. b Pie charts showing relative proportions of macrophage subsets in control and Isx-null mice. c Pseudotime trajectory of macrophage polarization showing altered differentiation dynamics in Isx-null livers, with reduced M2-like trajectories and enhanced M1-like polarization. d Representative immunofluorescence images showing M1 (F4/80⁺, CD80⁺) and M2 (F4/80⁺, CD206⁺) macrophages in tumor regions. Control mice exhibited both M1 and M2 macrophages, while Isx-null mice predominantly showed M1-like polarization. Scale bar, 100 μm. e t-SNE plot of hepatic T cells from xenograft tumor-bearing mice, clustered into 10 subsets by surface marker expression. f Pie charts showing T cell subset distribution. g Pseudotime analysis of T cell differentiation

Although the percentage of total T cell populations (4.4%) did not differ dramatically like macrophage population between Isx null mice and their control littermates with human tumour growth, T cells were further categorized into ten subsets based on surface markers [27] (Fig. 4e, f, Table S4). Subsets 1 (tissue-resident memory (TRM) T cells, 32.65%), 2 (central memory (TCM) T cells, 2%), and 9 (effector memory (TEM) T cells, 0.95%) were showed reduced numbers in Isx null mice compared to controls littermate with human tumor formation (Fig. 4f). Conversely, subsets 3, (CD4 + Helper T cells, 23.41%), 5 (γδ T cells, 2.26%), 7 (iNKT cell, 12.20%), and 8 (Cytotoxic CD8 + T cells, 4.11%) exhibited abundant in Isx null mice compared to controls littermate with human tumor formation (Fig. 4g). Pseudotime trajectory analysis revealed minimal differentiation among total T cells, with subset 5 (γδT cells, Trdc+Tcrg-C2 + Cd3d+ Il17a+) displaying poor differentiation, while subsets 1 (TRM T cells, Zfp683 + Itgae+ Cxcr6 + Cd69+) and 7 (iNKT cells, Zbtb16 + Mr1+Trav1 + Klrb1c+) demonstrated hyper-differentiation in Isx null mice with human tumor growth (Fig. 4g).

Loss of CD47-ISX–mediated inflammasome activation exacerbates tumor progression in a DEN-induced primary liver cancer model

To investigate the oncogenic potential of ISX-driven CD47 signaling and inflammasome activation in liver tumorigenesis, we employed a diethylnitrosamine (DEN)-induced primary liver cancer model using Isx-null mice and their littermate controls (Fig. 5a). Two-week-old mice received intraperitoneal DEN injections and were monitored for tumor development over eight months (Fig. 5a, b). At the study endpoint, control mice exhibited prominent white tumor nodules (5–9 spots) on liver surfaces, as indicated by yellow arrows, while markedly fewer or no lesions were observed in Isx-null mice (Fig. 5b). Histopathological analysis of H&E-stained liver sections revealed severe oxidative stress-induced damage, hepatocyte ballooning, fatty degeneration, and pronounced immune infiltration in the centrilobular zone of control mice, in contrast to the relatively preserved hepatic architecture and attenuated inflammatory response in Isx-null mice (Fig. 5c). Neoplastic lesions (blue arrows, T) were evident in control livers but absent in Isx-deficient mice (Fig. 5c). Further characterization using immunofluorescence staining demonstrated elevated expression of oncogenic markers Ki-67 (red) and ISX (green) within tumor lesions in control mice following DEN treatment (Fig. S1j).

Fig. 5.

ISX-driven CD47–inflammasome activation fosters a tumor-permissive M2 macrophage environment in DEN-induced liver tumorigenesis. a, b Experimental protocol and representative photographs of control and Isx null mice injected with DEN and TAA and their respective liver tissues. Yellow arrows, tumor. c Histological analysis of mouse liver examined via haematoxylin and eosin staining in the control and Isx null mice injected with DEN and TAA. Yellow arrows, neoplasm. d Analysis of GPT, GOT, IL-6, IL-18 and IL-1β in the control (n = 12) and Isx null mice (n = 12) injected with DEN and TAA. e Western blot analysis of the protein levels of inflammasome signals in liver tissue isolated from the control and the Isx null mice treated with DEN and TAA. f Hepatic microphage (mΦ; F4/80+, brown) level shown in the control and Isx null mice treated with CCl4. g Shown is the use of FlowSOM combined with t-SNE mapping to identify the major cell subsets in the control and Isx null mice treated with CCl4 in an unbiased manner for visualization. h The identity of each cluster color coded in the t-SNE plot in g was further visualized using a FlowSOM heatmap, which can precisely exhibit the combinatory expression level of multiple markers. i Representative tSNE plots of CD45 + CD11b+F4/80 + macrophages of control and Isx null mice treated with CCl4. The color code indicates the expression level of a given marker from low (blue) to high (red). j CyTOF analysis showing increased M1-likeand reduced M2-like macrophages in Isx-null mice compared to controls after DEN treatment. k Confocal immunofluorescent imaging of F4/80, CD80, and CD206 was conducted in control and the Isx null mice injected with DEN and TAA. CD80 and CD206 was visualised in green, F4/80 in red, and DAPI was used to indicate nuclei. Data are presented as mean ± SD from at least three independent experiments. Quantitative data were analyzed using one-way ANOVA followed by Tukey’s post hoc test. P < 0.05 was considered statistically significant

To assess systemic and hepatic injury, plasma levels of liver enzymes (GOT, GPT), the systemic inflammatory cytokine IL-6, and inflammasome-related cytokines (IL-1β, IL-18) were measured post-DEN administration [29, 30]. Control mice exhibited significant elevations in all markers, whereas these increases were largely blunted in Isx-null mice (Fig. 5d). Immunoblot analysis further revealed robust upregulation of hepatic CD47 signaling components (CD47 and SIRPα) and inflammasome-related proteins (NLRP1, NLRP3, ASC, AIM2, Caspase-1, IL-1β) in control livers, with minimal expression changes in Isx-deficient mice (Fig. 5e).

To characterize the immunological landscape within xenograft tumor models, total macrophage populations (F4/80⁺) were first assessed by immunohistochemistry (IHC) (Fig. 5f). DEN treatment significantly increased macrophage infiltration in control mice, whereas macrophage numbers were notably reduced in Isx-null livers (Fig. 5f). To further validate immune alterations within the tumor microenvironment, immune cell composition was analyzed by mass cytometry (CyTOF), revealing eight major immune cell populations (Fig. 5g–h). Consistent with the xenograft tumor model, Isx-null mice showed decreased infiltration of CD8⁺ T cells and macrophages relative to controls after DEN treatment (Fig. 5i). Notably, analysis of macrophage polarization revealed a marked shift toward M1-like macrophages (Ly6C^hi or MHCII^hi) in Isx-null mice, whereas M2-like macrophages (CD206 or MHCII^lo or Tlr7^hi) predominated in control littermates (Fig. 5j, k). CyTOF-based quantification showed that in Isx-null mice, M1-like macrophages (F4/80⁺CD80⁺) constituted over 65% of total hepatic macrophages, whereas M2-like macrophages (F4/80⁺CD206⁺) represented less than 25%. By comparison, control mice exhibited an M2/M1 ratio of 0.5–0.6, which dramatically declined to 0.01–0.005 in Isx-deficient mice. Confocal immunofluorescence corroborated these findings, showing abundant macrophage infiltration with strong expression of CD80, CD206, and F4/80 within tumor lesions in control mice. In contrast, although overall macrophage abundance was reduced in Isx-null livers, the remaining macrophages exhibited relatively enhanced CD80 expression on a per-cell basis, consistent with an M1-like phenotype. (Fig. 5k). These data suggest that the ISX–CD47–inflammasome axis promotes a tumor-permissive, M2-polarized microenvironment essential for carcinogen-induced tumorigenesis, as consistently observed in both xenograft and DEN-induced primary liver cancer models.

ISX and TWIST1 displayed an interaction pattern and a strong correlation with the clinical expression of CD47 and NLRPs in HCC patients

To further characterize the transcriptional regulation, the nuclear binding partners of ISX were determined in Huh7 hepatoma cells that overexpressed ISX with the GFP tag sequences using two-dimensional electrophoresis and tandem mass spectrometry [31]. The results showed that two out of the 14 candidate proteins from three independent experiments were consistent with the TWIST1 sequence. We examined seven hepatoma cell lines and found that the ISX expression levels were higher in aggressive hepatoma cell lines such as Hep 3B, Hep G2, SNU-423, and HA 22T than they were in Huh 7, Alexander hepatoma cell line PLC5, and benign hepatocytes (Chang liver CNL) [32] (Fig. 6a).

Fig. 6.

ISX is shown to interact with TWIST1 in vitro and in vivo. a Western blotting analysis of the protein levels of ISX, TWIST1, Snail and Slug in various liver cancer cell lines. CNL, Chang normal hepatocytes. b ISX association proteins determined by Western blot in the immunoprecipitation of Huh7 and SNU-423 cells. c TWIST1 association proteins determined by Western blot in the immunoprecipitation of Huh7 and SNU-423 cells. d Subcellular localisation of wild-type ISX and the vector determined by confocal immunofluorescence imaging in Huh7 cells. Wild-type ISX and vector, green; TWIST1, red; DAPI, blue; Yellow arrows, colocalization. e, f PLA of ISX and TWIST1 interaction in Huh7 cells and HCC samples. Red foci indicate the close proximity of the two proteins. NC: negative control. Yellow arrows indicate interactions of ISX with TWIST1 in the cells. N, normal; T, tumor. g–j The mRNA expression levels of TWIST1 were significantly correlated with the mRNA levels of ISX, NLRP1, NLRP3 and CD47 in HCC samples (n = 625). k, l. The mRNA expression levels of CD47 were significantly correlated with the mRNA levels of NLRP1and NLRP3 in HCC samples (n = 625). m The Kaplan-Meier survival curves of patients with HCC harbouring low and high TWIST1 mRNA expression. Based on the cutoff TWIST1 mRNA value of 2.7-fold change compared with the neighbouring normal tissue, the study population was divided into high ISX with high TWIST1, low ISX with low TWIST1 and low ISX with high TWIST1 expression groups. The results are shown as the mean ± s.d. Each experiment was repeated at least three times

To confirm the interactions between ISX and TWIST1 in vitro, immunoprecipitants of ISX and TWIST1 from hepatoma cells were examined using immunoblot assays. Endogenous TWIST1 was detected in ISX immunoprecipitants, and endogenous ISX was detected in TWIST1 immunoprecipitants (Fig. 6b, c). These results were further confirmed with confocal immune fluorescence staining analysis, in which GFP-tagged ISX (ISX-GFP; green) and mCherry-tagged TWIST1 (TWIST1-mCh; red) were co-localized primarily in the cytoplasm and, to a lesser extent, in the nuclei of Huh 7 (Fig. 6d). The interactions between endogenous ISX and TWIST1 were further evaluated in hepatoma cells and tumours samples of HCC patients with a proximity ligation assay (PLA). A significant amount of ISX-TWISTmacrophage polarization1 protein complex (discrete red spots) was detected in the cytoplasm and, particularly, in the nuclei of SNU-423 hepatoma cells (Fig. 6e). Additionally, the interaction patterns were detected in liver tumour tissue from patients with HCC (Fig. 6f; NC, negative control for PLA). To further explore the potential clinical impact among ISX, TWIST1, CD47, and NLRPs in patients with HCCs, we examined the expression levels of ISX, TWIST1, CD47 and NLRPs (NLRP1 and NLRP3) mRNA in HCC tumors samples from 625 HCC patients. The results showed that ISX mRNA levels correlated strongly with the level of TWIST1 mRNA expression in tumor samples of HCC patients (Pearson correlation coefficients, r = 0.8122, p < 0.001; Fig. 6g). Also, TWIST1 mRNA levels correlated strongly with the level of NLRP1, NLRP3, and CD47 mRNA expression in tumour samples of HCC patients (Pearson correlation coefficients, r = 0.797, 0.6663, and 0.737, respectively, p < 0.001; Fig. 6h–j). Of note, CD47 mRNA levels correlated strongly with the level of NLRP1 and NLRP3 mRNA expression in tumour samples of HCC patients (Pearson correlation coefficients, r = 0.7002 and 0.6428, respectively, p < 0.001; Fig. 6k, l). Significantly, when compared with the patient’s prognosis information, the survival time in HCC patients with a high mRNA expression of both ISX and TWIST1(high group, n = 58) was significantly shorter than those HCC patients with a low mRNA expression of either ISX, TWIST1 or both (low ISX group, n = 62 or both low, n = 346) (p < 0.001) (Fig. 6m). The results suggest ISX-TWIST1 complex displays a strong correlation with the clinical expression of CD47 and NLRPs in HCC patients.

The ISX-TWIST1 protein complex modulates the signals of CD47 and inflammasome activities and subsequent oncogenic processes

To investigate the regulatory impact of the ISX–TWIST1 complex on CD47 signals, inflammasome activation, and oncogenic activity, we conducted experiments involving forced expression or gene-specific interference of this specific complex in hepatoma cells (Huh 7 and SNU-423 cells). The results revealed that simultaneous forced expression of both ISX and TWIST1 significantly enhanced the mRNA expression of CD47 compared to the effects induced by individual ISX or TWIST1 expression alone in hepatoma cells (Fig. 7a). Moreover, co-expression of ISX and TWIST1 led to a notable increase in the protein expression of the CD47-SIRPα axis and inflammasome-related proteins (NLRP1, 3, NLRC4, AIM2, ASC, and pro-IL-1β) to a greater extent than the effects observed with ISX or TWIST1 expression alone in hepatoma cells (Fig. 7b). Additionally, the co-expression of ISX and TWIST1 significantly heightened the secretion of inflammasome cytokines (IL-1β and IL-18) compared to cells expressing ISX or GFP alone in hepatoma cells (Figs. 7c and S2a). Notably, hepatoma cells transfected with TWIST1 alone did not consistently activate inflammasome-related proteins or induce subsequent inflammatory cytokine secretion. The regulatory effects of the ISX–TWIST1 complex on oncogenic activities were examined by assessing proliferation, wound healing and cell migration activity. As anticipated, the co-expression of ISX and TWIST1 significantly promoted these oncogenic activities compared to cells expressing ISX or GFP alone in hepatoma cells (Figs. 7d and S2b).

Fig. 7.

The ISX-TWIST1 protein complex exhibited a regulatory role in both poor prognosis and inflammasome activities. a The mRNA levels of CD47 were analysed using semiquantitative RT-PCR. The Huh-7 and SNU-423 cells were transfected either with ISX, TWIST1, or a combination of both. b The protein levels of ISX, TWIST1, NLRP1, NLRP3, NLRP4, AIM2, ASC, pre-IL-1β, CD47 and SIRPα in Huh7 and SNU-423 cells were analysed using Western blot. The Huh-7 and SNU-423 cells were transfected either with ISX, TWIST1, or a combination of both. c The levels of IL-18 in the culture media were assessed using the ELISA method. Huh7 and SNU-423 cells were transfected either with ISX, TWIST1, or a combination of both. d The cell invasion abilities in Huh7 and SNU-423 cells were evaluated using the Transwell assay and the cells were transfected either with ISX, TWIST1, or a combination of both. e The mRNA levels of CD47 were analysed using semiquantitative RT-PCR. The Huh-7 and SNU-423 were transfected either with ISX, ISX shRNA, TWIST1 shRNA, or a combination of both. f Western blot was performed to analyse the protein levels of proliferation and epithelial-mesenchymal transition (EMT) markers in Huh7 and SNU-423 cells. The cells were transfected either with ISX, ISX shRNA, TWIST1 shRNA, or a combination of both in order to investigate their effects on protein expression levels. g The levels of IL-18 in the culture media were determined using the ELISA technique. Huh7 and SNU-423 cells were transfected either with ISX, ISX shRNA, TWIST1 shRNA, or a combination of both. h The cell proliferation in Huh7 and SNU-423 cells was assessed by measuring the BrdU (bromodeoxyuridine) incorporation activity. The cells were transfected with ISX, ISX shRNA, TWIST1 shRNA, or a combination of these constructs. i The cell invasion abilities in Huh7 and SNU-423 cells were evaluated using the Transwell assay and the cells were transfected with ISX, ISX shRNA, TWIST1 shRNA, or a combination of these constructs. j–l. The mRNA expression levels of M1 markers (iNOS and CD80) and M2 markers (Arginase1 and CD206) were assessed in THP1 cells using semiquantitative RT-PCR. These cells were co-cultured with Huh7 or SNU-423 cells, which were then transfected with either ISX or TWIST1. Data are presented as mean ± SD from at least three independent experiments. Comparisons between two groups were performed using an unpaired two-tailed Student’s t-test. P < 0.05 was considered statistically significant

Additionally, the introduction of gene-specific RNAi targeting ISX or/and TWIST1 into hepatoma cells resulted in a significant reduction in CD47 mRNA levels compared to cells transfected with ISX-GFP (Fig. 7e). Similarly, the use of gene-specific RNAi against ISX or TWIST1 led to a marked suppression of both the protein levels of the CD47-SIRP axis and key inflammasome-related genes (NLRP1, NLRP3, NLRC4, AIM2, ASC, and pro-IL-1β), as illustrated in Fig. 7f. This effect extended to the secretion of inflammasome cytokines, including IL-1β and IL-18, as demonstrated in Figs. 7g and S2c. Importantly, these inhibitory effects on CD47 and the inflammasome were accompanied by a subsequent attenuation of oncogenic activities, such as cell proliferation and Transwell migration, induced by ISX expression in hepatoma cells, as evidenced by the data presented in Fig. 7h, i.

To investigate the functional impact of the ISX–TWIST1 complex on macrophage polarization, we carried out experiments involving the enforced expression of this specific complex in hepatoma cells (Huh 7 and SNU-423 cells) co-cultured with THP1 (M0). We assessed the polarization markers of M1 (iNOS and CD80) or M2-like (Arginase1 and CD206) by evaluating mRNA expression in THP1 cells. In the co-culture setup with Huh 7 hepatoma cells expressing the forced ISX-TWIST1 complex, a significant repression of M1-like polarization markers was observed, coupled with an increase in M2-like markers when compared to co-culture with hepatoma cells transfected with either ISX or TWIST1 alone (Fig. 7j, k). Similar results were observed with SNU-423 hepatoma cells (Figs. 7l and S2d). The results suggest that the ISX-TWIST1 complex promotes the levels of inflammasome-relevant and CD47 signaling genetically, leading to oncogenic activities and influencing the polarization of surrounding macrophages.

ISX interacts with the bHLH domain of TWIST1 through the homeobox domain to formation of a functional complex

To investigate the detail interaction mechanism of ISX-TWIST1, the interacting domains in the complex were analysed using truncated ISX [31] or TWIST1 [32], together with the anti-GFP tag Abs. The results demonstrated that ISX interacted with the wild-type and truncated TWIST1 mutants, but not the TWIST1 mutant with the deletion of its basic DNA-binding domain (ΔbHLH; encoding 109–163 aa fragment (Fig. 8a). Reciprocally, TWIST1 interacted with the wild-type and truncated ISX mutants by deleting its N-terminal or C-terminal end (∆5 ISX mutant, deletion of 1–81 amino acids; Δ3 ISX mutant, deletion of 190–254 amino acids), but not with the truncated ISX mutants, which deleted homeobox domain (ΔHD ISX mutant, deletion of 81–141 amino acids) (Fig. 8b). The same as Fig. 8a. The interaction between ISX and TWIST1 mutants was further verified by a PLA with hepatoma cells. The results showed a significant amount of discrete red spots in the cytoplasm and, particularly, in the nuclei in SNU-423 hepatoma cells transfected with both GFP-tagged ISX and Flag-tagged wild TWIST1, but not with ΔbHLH (ΔB) TWIST1 mutants (Fig. 8c and S2e). The interactions between the TWIST1 domain and ISX were closely observed. Utilizing three-dimensional structural modelling predicted by the AlphaFold 3 server [21], as well as assessments of surface electrostatic force distribution and solvation energy, it was revealed that basic DNA-binding domain (BHLH) regions on TWIST1 exhibited promising binding sites for the homeobox domain of ISX (Figs. 8d, e and S2f-g). The genomic binding affinities of the ISX-TWIST1 complex on target genes were evaluated through chromatin-immunoprecipitation (ChIP-ChIP) assays. Hepatoma cells transfected with GFP (mCherry)-tagged ISX-TWIST1 demonstrated a notable enhancement in genomic binding levels on the promoter regions of CD47, NLRPs (NLRP1 and NLRP3), and IL-18, as determined by the ChIP-ChIP DNA assay. In contrast, hepatoma cells transfected with TWIST1ΔB mutants did not exhibit a similar increase (Figs. 8f and S2h).

Fig. 8.

ISX formed a functional complex by interacting with the bHLH domain of TWIST1 through its homeobox domain. a and b The presence of GFP-tagged truncated ISX mutant proteins or mCherry-tagged truncated TWIST1 mutants was identified in the immunoprecipitates using anti-mCherry or GFP antibodies. c A PLA was performed to investigate the interaction between ISX and truncated TWIST1 in Huh7 cells. In this assay, the close proximity of the two proteins is shown by red foci, indicating their physical interaction. Yellow arrows are used to indicate the specific interactions between ISX and TWIST1 within the cells. d, e Overall structure is colored by chains (green: TWIST1-bHLH; yellow: ISX-HD). The ipTM and pTM of the model are 0.71 and 0.75, respectively. One TWIST1-bHLH and one ISX-HD complex structure are shown. f A ChIP-ChIP assay was conducted using chromatin samples from transformed Huh7 cells that expressed ISX and TWIST1, as well as their respective variants, in Huh7 cells. g, h Huh7 cells were transfected with ISX, TWIST1, and their respective variants, and their invasive potential was assessed by measuring their abilities to proliferate and migrate using BrdU and Transwell assays. i Analysis of targeted genomic fragments using anti-ISX-TWIST1 ChIP-ChIP sequencing. j, k Genomic binding fragments were located on the promoters of genes in inflammasomes and NF-κB signaling pathways; all peak calling sequences for these analytes were found within 1 kbp of the promoter region. Data are presented as mean ± SD from at least three independent experiments. Comparisons between two groups were performed using an unpaired two-tailed Student’s t-test. P < 0.05 was considered statistically significant

As expected, the ISX-TWIST1 complex synergistically promoted the secretion of IL-1β and IL-18, compared to hepatoma cells expressing ISX or TWIST1 alone, as well as ISX ΔHD mutants or TWIST1ΔB mutants (Fig. S2i). To further assess the oncogenic characteristics regulated by the ISX-TWIST1 complex or mutants in hepatoma cells, cell proliferation and Transwell invasion assays were conducted. In line with previous findings, cells expressing the wild-type ISX-TWIST1 complex exhibited a significant increase in cell proliferation and invasion activities compared to cells transfected with mock or ΔHD ISX mutants (or ΔbHLH (ΔB) TWIST1 mutants) (Fig. 8g and h).

Furthermore, in order to gain global insights of the oncogenic activities induced by ISX-TWIST1 complex that induced macrophage polarization and subsequent pathogenic progression, the target genes regulated by ISX-TWIST1 complex in hepatoma cells were evaluated using chromatin immune precipitation (ChIP-ChIP) sequencing with anti-ISX/TWIST1 mAbs (Fig. 8i). The peak calling of 150 bp genomic DNA fragments localised within the 1.5-Kb promoter region of each target gene was considered an ISX downstream target (Fig. 8i). Interestingly, 68 genes showed high binding affinity and were targeted as positive ISX downstream genes, including genes involved in NF-κB signals (TLRs, MYD88, NFKBs, IKBKs, and RELs), inflammasome assembly (NLRPs, NLRCs, NEK7, AIM2, and CASPs), membrane pore formation (GSDMs), inflammatory cytokines (IL1B, and IL18), and immune checkpoint regulators (CD47 and SIRPα) (Fig. 8j, k). MEME analysis for the binding motifs of target genes revealed the conserved degenerate sequence, -GGWTYARG-(Fig. S2j).

In conclusion, this manuscript highlights a novel ISX-TWIST1 signalling pathway that elucidates how tumor cells modify the tumor microenvironment to favor M2-like macrophages. This underscores the significance of the M2-like macrophage population in tumor progression, including hepatocellular carcinoma (HCC). Additionally, the involvement of the immune checkpoint CD47 in ISX-related oncogenic activity provides further evidence supporting various anti-CD47 tumor therapies currently undergoing preclinical trials.

Discussion

Immunotherapy has emerged as a promising strategy for cancer treatment, but its effectiveness is often hindered by the complexity of regulatory networks and the tumor microenvironment (TME), particularly the immunosuppressive microenvironment that shapes HCC progression, metastasis, and therapeutic resistance [33]. Continuing to the regulation of PD-1/L1 and CTLA4 16, in this study, we identify a novel regulatory axis—CD47–ISX–NLRPs—driven by the homeobox transcription factor ISX. Importantly, genetic ablation of CD47–ISX signaling in tumor models significantly diminished M2-lik TAM infiltration and TRM T cell differentiation, highlighting CD47–ISX as a master regulator of the tumor immune microenvironment. Functioning as a multifunctional molecular switch, ISX integrates diverse environmental cues—including inflammatory stimuli, aryl hydrocarbon receptor (AhR) activation, and HCV infection—which may collectively influence tumor progression and responsiveness to immunotherapies such as anti-PD-1/L1, anti-CTLA4, and anti-CD47 [34].

The CD47–SIRPα axis is a key regulator of macrophage activation and immune tolerance, particularly in the context of tissue-resident and tumor-associated macrophages that shape cancer progression and immune escape [35]. CD47, often overexpressed in tumors, binds SIRPα on macrophages to deliver an inhibitory “don’t eat me” signal that prevents phagocytosis and promotes immune evasion [36]. This interaction recruits SHP-1/2 phosphatases, suppressing NF-κB and STAT1 signaling and driving an M2-like phenotype characterized by Arg1, IL-10, and TGF-β expression. In HCC, elevated CD47 reinforces M2 polarization through SIRPα–SHP–PPARγ–mediated metabolic reprogramming [35]. Therapeutically, CD47 blockade restores macrophage phagocytosis and antitumor activity, and several inhibitors—such as HX009 (CD47/PD-1 dual antibody) [37] and evorpacept (ALX148) [38]—are in clinical evaluation. Our findings identify the ISX–TWIST1 complex as an upstream driver of this immune evasion pathway; its inhibition downregulates CD47, promotes M1 polarization, and enhances antitumor immunity, while CD47 antagonists further augment NK and dendritic cell functions [39, 40]. Consistently, Haji and Ogawa reported that TET3-overexpressing macrophages (“Toe-Macs”) represent a pathogenic subset across chronic inflammatory diseases including MASH and NSCLC, where TGF-β1 and CCL2 induce TET3–STAT3–dependent activation of NLRP3, IL1B, and CD274, enhancing inflammasome signaling and PD-L1 expression [41, 42]. These Toe-Macs and ISX-driven TAMs share a convergent TGF-β1/CCL2–STAT3–NLRP3 axis that sustains chronic inflammation and immune tolerance [7]. Together, these insights position ISX and TET3 as complementary transcriptional and epigenetic regulators linking metabolic dysfunction to tumor immune evasion, providing a rationale for therapies targeting macrophage reprogramming in liver cancer and related inflammatory diseases.

Interleukin-18 (IL-18), a pleiotropic IL-1 family cytokine, bridges innate and adaptive immunity via MyD88-dependent NF-κB and MAPK signaling through the IL-18Rα/β complex [43]. Primarily produced by macrophages, Kupffer cells, and epithelial cells after inflammasome activation, IL-18 exerts context-dependent effects on immune regulation [44]. Under chronic or metabolic stress, it promotes M2-like macrophage polarization through STAT3 and PPARγ activation, enhancing Arg1, IL-10, and TGF-β expression to drive fibrosis and immune tolerance [45], whereas during acute inflammation or with IL-12, it amplifies IFN-γ–mediated M1 activation. In steatohepatitis, hepatocyte-derived IL-18 sustains an M2-dominant, immunosuppressive milieu [46]. It also supports TRM survival and cytotoxicity via the IL-18R–MyD88–T-bet axis and indirectly through macrophage-derived IL-15 and TGF-β [47]. Inflammasomes, particularly NLRP family members, regulate IL-1β and IL-18 maturation and are implicated in chronic liver inflammation and HCC [48].

The dynamic interplay between macrophages and tissue-resident memory T (TRM) cells critically shapes the hepatic tumor microenvironment (TME) [48]. The ISX–TWIST1–CD47–inflammasome axis drives immune remodeling by transcriptionally activating CD47 and NLRP1/3, elevating IL-18, IL-1β, and TGF-β to reinforce M2-like TAM polarization and sustain dysfunctional TRM-like clusters. This feed-forward circuit links innate and adaptive immune suppression, promoting tumor persistence and fibrosis. Therapeutically, disrupting CD47–SIRPα signaling may restore macrophage phagocytic activity and attenuate the cytokine loop that maintains pathogenic TRM cells, while modulation of inflammasome or IL-15/TGF-β pathways could rebalance TRM responses toward protective immunity [47]. Such interventions, potentially combined with immune checkpoint blockade (anti–PD-1/PD-L1 or anti–CTLA-4), may overcome resistance in inflammation-driven, ISX-high HCC. Moreover, monitoring TRM/TAM markers such as CXCR6, CD69, PD-1, and CD47 could provide prognostic insight and guide therapeutic stratification.

In summary, this study reveals the pivotal role of the CD47–ISX–NLRPs axis in the progression of hepatocellular carcinoma (HCC), demonstrating how CD47 signaling collaborates with ISX and TWIST1 to activate tumor-promoting genes via specific promoter motifs (“-GGDWYR-”). The strong association between CD47, ISX, and NLRPs expression levels and key clinical parameters—including lesion number, disease stage, and lymphovascular invasion—highlights their promise as prognostic biomarkers.

Impact and implications

This study uncovers the critical role of the CD47–ISX–NLRPs axis in driving hepatocellular carcinoma (HCC) progression, elucidating how CD47 signaling cooperates with ISX and TWIST1 to transcriptionally activate tumor-promoting genes through distinct promoter motifs (“-GGDWYR-”). The strong correlation between the expression levels of CD47, ISX, and NLRPs and key clinical parameters—including tumor burden, disease stage, and lymphovascular invasion—positions these molecules as promising prognostic biomarkers for HCC.

Importantly, genetic ablation of CD47–ISX signaling in tumor models significantly diminished M2-like tumor-associated macrophage (TAM) infiltration and impaired tissue-resident memory (TRM) T cell differentiation, highlighting CD47–ISX as a master regulator of the tumor immune microenvironment. Functioning as a multifunctional molecular switch, ISX integrates diverse environmental cues—including inflammatory stimuli, aryl hydrocarbon receptor (AhR) activation, and HCV infection—which may collectively influence tumor progression and responsiveness to immunotherapies such as anti-PD-1/L1, anti-CTLA4, and anti-CD47.

These findings not only provide a mechanistic rationale for targeting the CD47–ISX–NLRPs axis as a novel therapeutic strategy but also identify actionable biomarkers for patient stratification. Moreover, this work opens new avenues to enhance the efficacy of current immunotherapies, particularly in liver cancer and other macrophage-driven malignancies.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Author contributions

L.-T. Wang performed experiments, analyzed data, and drafted the manuscript. M.-H. Lin and S.-N. Wang contributed to experimental design and clinical sample analysis. Y.-C. Li, C.-Y. Chai, J.-M. Hsu, S.-S. Chiou, and H.-Y. C. Chiou assisted with methodology and data interpretation. S.-K. Huang and M.-C. Hung provided conceptual guidance and critical revision of the manuscript. S.-H. Hsu supervised the study, secured funding, and finalized the manuscript. All authors read and approved the final manuscript.

Funding

Open access funding provided by National Taiwan Normal University. This work was supported by Kaohsiung Medical University Hospital research grants (KMUH 106-6R33, KMUH107-7R46, KMUH107-7R34); the Ministry of Science and Technology/National Science and Technology Council, Taiwan (MOST108-2314-B-037-065-MY3, 109-2326-B-003-001-MY3, 110-2314-B-037-072-MY3, 111-2320-B-037-031, NSTC-111-2320-B-400-026, 112-2314-B-003-001, 112-2320-B-037-004, 113-2320-B-037-028-MY3, 113-2320-B-003-003, 114-2314-B-003-001-MY3, 114-2320-B-003-006-MY3, 114-2314-B-037-023, 114-2320-B-039-063); Kaohsiung Medical University grants (KMU-DK(A)113009, KMU-DK(A)115004); T-Star Cancer Center (NSTC 113-2634-F-039-001 and NSTC 114-2634-F-039-001); and the Research Center for Precision Environmental Medicine, Kaohsiung Medical University within the Higher Education Sprout Project funded by the Ministry of Education (KMU-TC114A01).

Data availability

All sequencing datasets generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under accession number PRJNA1158586. Additional materials are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study involving human participants was approved by the Institutional Review Board of Kaohsiung Medical University Chung-Ho Memorial Hospital (KMUH-IRB-20200048 and KMUH-IRB-20180037). Written informed consent was obtained from all participants prior to sample collection. All animal experiments were conducted in accordance with institutional guidelines and approved by the Institutional Animal Care and Use Committee of Kaohsiung Medical University (IACUC-107134, IACUC-108162).

Consent for publication

Not applicable. No individual personal data are included.

Competing interests

The authors declare no competing interests.